Prepared by BMRB International

Home , Ammerland, Erzgebirgskreis, Oberbergischer Kreis, Odenwaldkreis, Vogelsbergkreis, Wittmund (district)

Technical report on the E Living study

for the Institute for Social and Economic Research (ISER)

Prepared by BMRB International.

Fieldwork between September and December 2001

Page 1 of 74

Produced for ISER by BMRB International – April 2002

Authors: Rick Loyd & Sinéad McDonald BMRB Number: 2713 457 E-Living Wave 1 Technical Report

1 Introduction...... 4 1.1 Background and Aims of the Study...... 4 1.2 Methodological Overview...... 5 1.3 Sampling and response rates...... 6 1.4 Weighting...... 6 1.5 Geographical coverage NUTS regions ...... 7 2 Telephone outcome codes...... 8 2.1 Introduction...... 8 2.2 Outcome Codes ...... 8 2.3 Outcome definitions...... 9 3 Telephone Response Rates - Germany, Israel, Italy, Norway and the UK...... 18 3.1 Outcomes ...... 18 3.2 Response Rates...... 20 4 Bulgaria...... 21 4.1 Introduction...... 21 4.2 Fieldwork and Questionnaire ...... 21 4.3 Sampling and Survey Design...... 22 4.4 Contact records...... 23 4.5 Response rates...... 24 4.6 Weighting...... 24 4.6.1 Respondent and Household Weights ...... 24 5 Germany...... 27 5.1 Fieldwork and Questionnaire ...... 27 5.2 German RDD sample...... 27 5.3 Outcome Codes and response rates ...... 28 5.4 Weighting...... 29 5.4.1 Introduction ...... 29 5.4.2 Respondent Weighting...... 29 5.4.3 Household weighting ...... 33 6 Israel ...... 35 6.1 Fieldwork and Questionnaire ...... 35 6.2 RDD sample in Israel...... 35 6.3 Outcome Codes and Response Rates...... 36 6.4 Weighting...... 37 6.4.1 Introduction ...... 37 6.4.2 Respondent Weighting...... 37 6.4.3 Household weighting ...... 41 7 Italy...... 43 7.1 Fieldwork and Questionnaire ...... 43 7.2 Italian RDD sample ...... 43 7.3 Outcome Codes and response rates ...... 43 7.4 Weighting...... 44 7.4.1 Introduction ...... 44 7.4.2 Respondent Weighting...... 45 7.4.3 Household weighting ...... 48 8 NORWAY...... 49 8.1 Fieldwork and Questionnaire ...... 49 8.2 Norwegian RDD sample...... 49 8.3 Outcome Codes and response rates ...... 49 Page 2 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

8.4 Weighting...... 50 8.4.1 Introduction ...... 50 8.4.2 Respondent Weighting...... 51 8.4.3 Household weighting ...... 54 9 United Kingdom...... 56 9.1 Fieldwork and Questionnaire ...... 56 9.2 EPSeM Sample...... 56 9.3 Outcome Codes and response rates ...... 56 9.4 Weighting...... 57 9.4.1 Introduction ...... 57 9.4.2 Respondent Weighting...... 58 9.4.3 Household weighting ...... 61

Appendices A: Description of EPSeM and other RDD techniques B: EN NUTS regions and other geographies

Page 3 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

1 INTRODUCTION

1.1 Background and Aims of the Study

The research study is part of a larger project of research funded under the EU Information Society Technologies (IST) programme, and entitled e-Living: Life in a Digital Europe. This technical report, on the first of the two waves of fieldwork, describes the issues surrounding data collection for each of the six countries. The survey will provide valuable data on the uptake and usage of information and communications technology (ICT) at a domestic level in the Bulgaria, Israel, Germany, Norway, Italy and the UK.

The data from e-Living will be used by the academic, public and private sectors to understand the usage of ICT within European households over the two waves of fieldwork. This document reports on the fieldwork for wave one between September and December 2001 and the second wave will follow in the corresponding period a year later.

The Institute for Social and Economic Research (ISER) has responsibility for the design of the study. BMRB International conducted the fieldwork from their Ealing telephone centre in four countries and sub-contracted interviewing to partners in Bulgaria and Israel.

The questionnaires were harmonised as far as possible in each of the countries and this document contains a version of both the telephone and face-to-face questionnaires. These detail the full range of data collected about individuals. Over the two waves, the data will be used to:

• Describe, model and predict the impact of digitalisation on individuals and households in Europe.

• Explain differential rates of change

• Model the interaction between impacts and effects at different levels: individual, household, community, social network, region and nation.

• Describe, explain and model temporal changes in the uptake and usage of ICTs by individual levels and within households.

• Undertake specific analysis of the educational, employment and environmental implications of digitalisation.

Page 4 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

1.2 Methodological Overview

In designing the study a major requirement was to make the data collected from each of the six countries as consistent as possible. Telephone interviewing was favoured because it was cost-effective and could provide wider geographical coverage than clustered face-to-face assignments.

Table 1 shows the estimated telephone ownership amongst residential households by country.

Table 1: Percentage of households with a fixed telephone line Country Telephoned percentage Bulgaria 50 Germany 98 Israel 95 Italy 89 Norway 75 United Kingdom 93 Source: various national websites – some figures relate to 2000, and these should all be considered estimates.

In all countries except Bulgaria telephone ownership is large enough for telephone interviewing. Bulgarian fieldwork was therefore conducted face-to-face using paper interviewing (PAPI) with a local agency’s field-force.

BMRB conducted the fieldwork in four countries using the Computer Assisted Telephone Interviewing (CATI) facilities at its telephone unit in Ealing, West London. Native speaking interviewers were used to conduct fieldwork in Germany, Italy, Norway and the UK.

In Israel interviewers were required to speak Hebrew, Russian and/or Arabic and because of this it was decided that Israeli fieldwork would be best conducted locally rather than from the UK. The agency used, Teleseker, has worked with BMRB before on other projects and was also able to provide RDD sample that covered the three ethnic groups.

To ensure consistency between countries a BMRB researcher visited Sofia and an executive from Teleseker visited BMRB. We are confident that the data collected in Israel and Bulgaria are consistent with that of the other participatory countries.

The questionnaires used in each country were almost identical –there were only slight differences in Israel and Bulgaria and these only affected a few questions.

Page 5 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

1.3 Sampling and response rates

In each country a probability design was used. For countries using telephone interviewing, Random Digit Dialling (RDD) samples were used. A summary of the variants of these techniques is included in Appendix A, and issues specific to each country are discussed in the relevant chapters.

In Bulgaria a multi-stage face-to-face PAPI sampling design was used. The sampling frame was the electoral register which is regularly used as it has good coverage of the residential Bulgarian population.

This report details the response rates for each country in some detail. The calculation of response rates for face-to-face jobs is relatively straightforward and, in the UK, well documented. The equivalent calculations for telephone studies are more complex, and to an extent subjective, because not all numbers in the sample are part of the universe of interest. A full list of outcome codes for the telephone samples is provided in chapter 3 on response rates.

1.4 Weighting Weights are provided at both individual and household level.

The respondent weights for data in each country are comprised of different components:

• design weights to reflect differential probabilities of selection; • weights to correct for varying response rates amongst certain groups; • non-response demographic weights to align the sample profiles to population estimates.

To help with assessment of response bias interviewees who refused to participate in the full interview were requested to answer four brief questions – two personal, age and gender, and two household facts: number of people living in their house and whether or not there was a PC at that address.

The possibilities for the third component of weighting vary between countries in terms of the scope and timelines of available information. To be consistent it was decided to use sex interlocked with age and then region as a rim within each country. The data files for each country contain the design weight and the final weight (which is the combination of the different elements of weighting) and these are at a respondent level

Page 6 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

E-living also collects data at a household level. There is limited availability of household population profiles and so household weights could not be constructed in the same way as for respondent weights. Estimated Household weights have, however, been derived from population profiles and are provided with the data.

1.5 Geographical coverage NUTS regions

ISER requested that NUTS region (NUTS3) be attached to the data for participating countries that are part of the European Union – namely, the UK, Italy and Germany. NUTS are hierarchical geographic administrative areas, and NUTS3 are smaller areas than NUTS1 areas.

In the UK NUTS3 areas are amalgams of local and unitary authorities, in Italy provinces, and in Germany, kreise or counties. Norway use counties rather than NUTS regions. The level of NUTS geography in each European country is included in appendix B, with a full list of geographical classifications for the other participatory countries.

1.6 Achieved sample

Table 2 below shows the number of interviews obtained in each country at Wave 1.

Table 2: Number of Wave 1 Interviews by country Country Wave 1 Interviews Bulgaria 1750 Germany 1753 Israel 1753 Italy 1762 Norway 1756 United Kingdom 1760

A database of respondents’ contact details is available for wave 2.

Page 7 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

2 TELEPHONE OUTCOME CODES

2.1 Introduction

All telephone samples were generated using Random Digit Dialling (RDD) techniques. The use of RDD has developed significantly over the last 15 years and has evolved at differing rates in countries according to the popularity of telephone research, the penetration of telephone owning households and the way a country’s telephone numbers are organised. RDD numbers are easier to generate in countries that use a prefix for an area code. Appendix A discusses RDD techniques more generally and the next section will look at each country individually. In the UK EPSeM, the purest form of RDD sampling, was used. In Germany, Italy, Norway and Israel this technique can not be used currently so numbers were generated using a list-assisted RDD method. The next section defines the outcome codes achieved in each country using RDD sample.

2.2 Outcome Codes Table 3 shows the range of outcomes that may result from using RDD sample.

Table 3: Telephone Fieldwork Outcomes Line Outcome Classification Number T1 Total Sample Issued T2+T3+T4+T5

A Business Number (including Faxes) Deadwood B Dialler - incomplete Deadwood C Dialler - site out of service Deadwood D Number unobtainable Deadwood E Claimed that no-one over 16 lives here Deadwood T2 Total Deadwood A+B+C+D+E

F Unavailable during fieldwork Contact G Soft Refusal Contact H Unknown at Number Contact I Hard Refusal Contact J Full Interview Contact K Abandoned Interview Contact L Respondent quit out of interview Contact M Stopped Interview Contact N Selected Respondent has hearing problem Contact O Selected Respondent has language problem Contact P Duplicate number Contact

Page 8 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

T3 Total Contacts F+G+H+I+J+K+L+M+N+ O+P

Q Pure 10+ Non-Contact R Probably Pure 10+ Non-Contact S At least one other outcome with 10+ Non-Contact unsuccessful T4 Total Non-Contact (10+ unsuccessful Q+R+S calls)

T Appointment Unresolved at end of f'work U General Call Back Unresolved at end of f'work V No answer/Answering machine Unresolved at end of f'work W Engaged Unresolved at end of f'work X Dialler - no answer Unresolved at end of f'work Y Dialler - busy Unresolved at end of f'work Z Dialler - nuisance hangup Unresolved at end of f'work AA Dialler - unknown error Unresolved at end of f'work AB Dialler - General failure code Unresolved at end of f'work T5 Total - Unresolved at end of fieldwork T+U+V+W+X+Y+Z+AA+ AB

2.3 Outcome definitions This section defines the outcome codes in more detail, and groups the codes under four classifications: Deadwood, Contact, Non-Contact and Unresolved to help define response rates.

BMRB’s telephone unit uses a device called an autodialler (or dialler) to maximise the efficiency of fieldwork. The autodialler dials numbers automatically so interviewers need not dial telephone numbers themselves. The dialler operates in two modes: predictive and power dialling.

In predictive mode the dialler calls more numbers than there are interviewers waiting for a call. The dialler reviews the following parameters to calculate how many numbers to dial:

• number of connected calls over the recent period (exact length determined by the dialler operator) = N1 • the number of attempted calls in the same period = N2 • a setting on how "aggressive" the dialling should be = A

Page 9 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

• the number of interviewers waiting for a connection = N3

Upon at least one interviewer becoming free, the dialler will then determine how many telephone numbers to dial from the following formula:

N3 + ( (N2/N1 - 1) x A x N3)

For example, where 50% of recently dialled numbers resulted in a connected call (i.e. twice as many calls were attempted as connected so N2/N1 = 2), the "aggressive" mode is 70%, and 10 interviewers are waiting for a connection, then 17 numbers (=10 + (2-1) x 70% x 10)) will be dialled for the 10 interviewers. Should this result in more then ten connections then the dialler will automatically hang up any extra connections, resulting in a ‘nuisance hangup’ (see outcome code Z).

While predictive dialling is the most efficient method of dialling, it will only work effectively when there is a large team of interviewers calling at the same time so sufficient data is available to accurately predict how many numbers should be dialled. When there is a smaller team or the dialler does not have enough data, the dialler will behave as if it is in power dialling mode.

Power dialling is less efficient than predictive dialling, in that the dialler does not over-dial for each requested piece of sample. Instead for each interviewer requesting a call, the dialler will call one number. Should that number fail to connect, it will automatically dial a second number. Power dialling has to be used when nuisance calls are re-issued, as OFTEL regulations prevent making more than one nuisance call to a household in a six month period.

Whether dialling in predictive or power dialling mode, a key strength of the dialler is that unobtainable numbers, no replies, engaged numbers and other unconnected calls are dealt with automatically by the dialler, resulting in interviewers spending time as productively as possible.

The outcome codes for a telephone sample are more diverse than for face-to-face. This is because, for RDD sample, not all numbers generated will be residential and others will be rung repeatedly during fieldwork without a connection being made or contact established with a person. The level of non-residential numbers and numbers where no contact is ever made varies between RDD samples within a country and between countries. This makes sample management difficult, because interviews must be achieved in the fieldwork period at the highest possible response rate. Hence issuing too much sample too early in fieldwork may result in the interviews being achieved at a lower response rate with a large amount of unresolved sample that has not been worked thoroughly during fieldwork. Alternatively not issuing enough sample means that interviewers may be booked

Page 10 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report on shifts where there are insufficient numbers to be dialled. The profile of sample outcomes at the end of fieldwork will depend on the distribution of the RDD numbers generated for fieldwork and the success of predicting and managing the call outcomes that arise during fieldwork.

The following list describes the outcomes that may arise from fieldwork.

T1 Total Sample of numbers issued at the start or during fieldwork

T2 Deadwood

• A Business Number

RDD sample is screened by suppliers against a register of known business numbers. However, the coverage of these registers varies across countries and depends on how each country’s telephone numbers are allotted to service providers. The highest proportion of business numbers in the sample for the five countries was in the UK at 23%, other countries had much smaller proportions - single figures in each country. The business number category includes fax numbers.

• B Dialler – incomplete

The call did not go through at all. This may be an incomplete or incorrect number which does not even connect as an unobtainable.

• C Dialler - site out of service

This stands for “site out of order” and means unobtainable – see D.

• D Number unobtainable

These are numbers that, when dialled, give an unobtainable tone and so are not considered working numbers and are therefore not part of the universe.

• E Claimed that no-one over 16 lives here

The e-living study has a universe of adults aged 16 or over. Some residents of households claimed that no one in this age group lived at the house, and so nobody eligible for interview was available. These households do not therefore form part of the universe of the study. It is unlikely that there are many of these types of households and across all five countries there were only 22 altogether.

Page 11 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

T3 Contacts

• F Unavailable during fieldwork

When a randomly selected individual was identified for interview a request was made to speak to that person who may or may not have been away at the time of the call. On occasions that person may have been away for some time and so was not available during fieldwork and as such could never be part of the study. These outcomes were rare – fewer than 20 in each country apart from Italy, where response rates were higher than anticipated, and Israel, where there were three surveys covering each of the three language areas.

• GSoft Refusal

These are people who, after an initial refusal, were re-contacted in the anticipation that they would agree to be interviewed at a more convenient time. However, they again declined to be part of the study when they were contacted later. When contacted later they did not complete an interview although technically they did not refuse again.

• H Unknown at Number

This outcome is very rare, fewer than 15 across all countries, and is probably best considered a form of refusal. As RDD is not named sample theoretically this outcome should not occur. However, it may have occurred when a respondent was initially selected for interview but was out at the time and so another call was made later to try and interview the selected individual. When this later call was made, it was claimed by someone in the house (possibly the individual with whom the interview was sought, or by another resident) that the chosen individual did not live there. The person in question could of course have moved, but it may just have been a ploy to terminate the attempted interview.

• I Hard Refusal

A hard refusal is where the respondent strongly refused to participate in the survey, and it was felt that a further call would neither be ethical, nor have a more successful outcome. If a respondent was initially classified as a ‘soft refusal’ and, on being called back, refused again they automatically became a hard refusal.

• J Full Interview

Page 13 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

These are full interviews with respondents and number approximately 1,750 in each country.

• K Abandoned interview

This outcome is the result of a respondent, at some point during the interview, deciding they no longer wished to continue and so the data collected are incomplete and do not constitute a full interview. K, L & M are alike and the totals for each country are broadly similar, whilst the breakdown within each group varies.

• L Respondent quit out of interview

This is essentially the same as K above – codes K and L are coded by interviewers.

• M Stopped Interview

This is different from K above in that the respondent stopped the interview temporarily with the intention of continuing later at a more convenient time. However, the interviews were occasionally not completed later.

• N Selected Respondent has hearing problem

Contact was made with a selected respondent but an interview was not possible because the respondent could not hear the questions well enough to answer them.

• O Selected Respondent has language problem

Contact was made with a selected respondent but an interview was not possible because the respondent could not understand or answer the questions in the language(s) spoken. This outcome, as expected, was much more prevalent in Israel than the other countries.

• P Duplicate number

A true duplicate number would only result from the RDD number propagation process generating two different numbers that rang in a household where there was more than one phone line. Multiple line ownership in the UK is relatively minor so the chances of this happening are minuscule. A more likely

Page 14 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

explanation of this outcome is that respondents have taken part in other research which they have confused with the e-living study. This is a form of refusal and it happened infrequently in Italy, Norway and the UK but more commonly in Germany where people are generally less receptive to telephone research.

Page 15 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

T4 Non-Contacts

It is usual in telephone research to set a minimum number of times that a piece of sample will be dialled. This will vary between projects but is often ten attempts. The number of times a piece of sample is actually called depends on the outcome of the initial attempts, but whilst a piece of sample remains unresolved then it will be dialled repeatedly on different days and at different times. When a number had been dialled at least ten times, it will continue to be dialled until actively removed from the sample management system. The numbers that have been called ten times or more are non-contacts and can be categorised in the three groupings (Q,R and S) below. The assumption here is that for the large majority of these numbers, however many times a number is called, no contact with someone in a residential address will be made.

• Q Pure 10+

These are all numbers that have been dialled at least ten times, and very often more, and the number has rung every time without ever having been picked up.

• R Probably Pure 10+

These are numbers where the majority (usually all bar one) of outcomes have resulted in an undisturbed ringing tone. The exceptions may have been an engaged tone resulting from the exchange being busy or a simultaneous call dialled to that number.

• S At least one other outcome with 10+ unsuccessful

To make fieldwork efficient when a number has been called 10 times without reply the number is taken out of the ‘active sample’. These numbers are not strictly 10+ any longer and may not fall under the umbrella of “probable non residential numbers” in the same way that Q and R do. However, even some of these numbers where one contact has resulted, may still be business numbers and hence subsequent callings are unproductive.

T5 Unresolved outcomes

Codes T to AB are what may be described as unresolved outcomes at the end of fieldwork. There will always be a limited amount of these as it is not cost-effective to continue calling until each number has a final ‘resolved’ classification. With

Page 16 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report careful management of sample the number of unresolved sample can be kept to a minimum.

• T Appointment

This is an unresolved outcome in that appointments were made but not kept. These were kept to a minimum but in some cases cannot be avoided.

• U General Call Back

A holding outcome used to re-activate sample later.

• V No answer/Answering machine

These are coded by the interviewer and the large majority will be answering machines. The code used does not distinguish between the two outcomes.

• WEngaged

Interviewers hear the engaged tone. This could be at the house or exchange level.

• X Dialler - no answer

No answer from the dialler

• Y Dialler – busy

Same as engaged.

• Z Dialler - nuisance hangup

These can only happen when the dialler is in “predictive” mode. The sample has been called by the dialler, the call has connected but there is no interviewer available. The sample is not automatically re-issued but can be reactivated..

• AA Dialler - unknown error

There are a number of reasons for these dialler errors. They are mainly produced due to a technical reason. This can be simply that the interviewer logged into a wrong station or the set-up or sample did not work properly. It could have been something very simple such as the sample having an incorrect Page 17 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

variable which is likely to happen from time to time. These will not have actually been called and are re-activated automatically.

• AB Dialler - General failure code

Same as AA. 3 TELEPHONE RESPONSE RATES - GERMANY, ISRAEL, ITALY, NORWAY AND THE UK

3.1 Outcomes

Table 4: Sample Counts and Percentages in each country List No Outcome(s) Germany Israel Italy Norway UK T1 Total Sample Issued 7653 8721 6841 6035 11828 T2 Total Deadwood 696 2663 1671 983 4768 T3 Total Contacts 5142 5094 4475 4238 4959 Total Non-Contact (10+ T4 unsuccessful calls) 875 777 504 560 1893 Total - Unresolved at end T5 of fieldwork 940 187 191 254 208

T2 Total Deadwood 9% 31% 24% 16% 40% T3 Total Contacts 67% 58% 65% 70% 42% Total Non-Contact (10+ T4 unsuccessful calls) 11% 9% 7% 9% 16% Total – Unresolved at end T5 of fieldwork 12% 2% 3% 4% 2%

The distribution of outcome types varied greatly between each country. The percentages in table 4 are based on total sample issued. Deadwood is the highest in the UK reflecting the use of EPSeM, rather than list-assisted RDD sample. However, Italy, and to a lesser extent the other countries, have significant amounts of deadwood too. The classification of telephone sample as deadwood is relatively easy and final. Some of the other sample classifications are more difficult and subjective. A face- to-face probability design would define a response rate as Interviews/(Sample Issued -Deadwood). However, in face-to-face studies ‘10+ unsuccessful contacts’ and ‘Unresolved’ are not outcomes and hence the response rates calculations are not complicated by them.

RDD techniques that were used in this study are described in more detail in the chapter discussing each country’s fieldwork and in Appendix A. RDD number propagation will inevitably generate phone numbers that are classifiable as deadwood. But it will also generate others that, when dialled repeatedly on different days and at different times, never connect and so are thought to be non-usable. In table 4, these are called Non-Contacts (T4) and the numbers of these also vary between each country.

Page 18 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

When deciding how much sample will be needed to obtain 1750 interviews at the best response rate possible, the proportions of deadwood and non-contacts as well as the response rate itself make the decision difficult. And the final amount of sample is a matter of fine judgement with the aim to minimise the amount of unresolved sample when fieldwork is complete.

In Germany the proportion of unresolved sample is highest because both the proportion of deadwood and response rates were lowest (and lower than anticipated).

After fieldwork the varieties of ‘10+ unsuccessful calls’ mean that there are a number of different ways of calculating Response Rate. Tables 3 and 4 have defined and shown the distribution of outcomes for each country. Table 5 below shows a range of response rate calculations for each country ranging from the pessimistic to the optimistic.

Table 5: Response Rates by Country U List No Germany Israel Italy Norway K J Full Interviews 1753 1753 1762 1756 1760 T1 Total Sample Issued 7653 8721 6841 6035 11828 T2 Total Deadwood 696 2663 1671 983 4768 T3 Total Contacts 5142 5094 4475 4238 4959 T4 Total Non-Contact (10+ unsuccessful calls) 875 777 504 560 1893 T5 Total - Unresolved at end of fieldwork 940 187 191 254 208

Q Pure 10+ 436 777 288 388 1619 R Probably Pure 10+ 41 0 47 61 83 At least one other outcome with 10+ S unsuccessful 398 0 169 111 191 Percentage of 10+ that would never have (Q+R)/T4 had a reply 54.5% 100.0% 66.5% 80.2% 89.9%

T5 Total - Unresolved at end of fieldwork 940 187 191 254 208 AC No Contact and so possible 10+ 28 0 45 72 198 No Contact and so possible 10+ and called AD >=7 times 23 0 25 13 30

RR1 Response Rate Definition 1: J/(T1-T2) 25.2% 28.9% 34.1% 34.8% 24.9% RR2 Response Rate Definition 2: J/(T1-T2-T5) 29.1% 29.9% 35.4% 36.6% 25.7% Response Rate Definition 3: J/(T1-T2-T5- RR3 T4) 34.1% 34.4% 39.4% 41.4% 35.5% Response Rate Definition 4: J/(T1-T2-Q-R- RR4 AD) 27.1% 33.2% 36.6% 38.3% 33.0%

Page 19 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

3.2 Response Rates

Response Rate 1 (RR1) is, in a sense, analogous with a face-to face response rate and in each country the minimum of all the rates in table 5. It assumes that all ‘non-contacts’ and ‘unresolveds’ are refusals and hence the response rates stated are very low.

Response rates 2 and 3 take the alternative view to RR1 and separately remove ‘unresolveds’ and then ‘non-contacts’ from the denominator of the response rate calculation. This has the effect of ensuring that RR2 and RR3 are higher than RR1.

Response Rate 2 removes the ‘unresolveds’ from the denominator which are relatively small (between 2% and 4% of sample issued) in each country apart from Germany where it’s 13%. The difference between RR1 and RR2 is about one percentage point apart from Germany where the difference is almost 4%.

Response Rate 3 improves the RR2 figures by removing all outcomes coded as ‘10+’ from the response rate denominator. The number of ‘10+’ outcomes again varies between country and are largest in the UK where EPSeM sampling was used. RR3 is the most favourable of the four rates quoted and assumes that none of the ‘non-contacts’ would have generated an interview because the numbers could not have connected with a residential household.

Response Rate 4 makes a slight modification to RR3 by adjusting the numbers of ‘unresolveds’ and ‘non-contacts’ used in the definition of RR3. For ‘unresolveds’ it predicts how many of the outcomes might be expected to end up as ‘pure non-contacts’. Similarly for all ‘non-contacts’ only some of the outcomes can be classified as ‘pure’ or ‘probably pure’. These are outcomes from dialling where no-contact has been made throughout the call history for individual numbers. When only these types of ‘non-contact’ and ‘unresolved’ outcomes are removed, rather than T4 and T5 in total, the response rates for RR4 are not as high as for RR3.

The definitions of RR1 through RR4 show the subjective approaches to calculating telephone response rates. The full count of outcomes codes is listed for each country in the relevant chapters so further calculations can be made by readers.

Page 20 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

4 BULGARIA

4.1 Introduction

Table 1 has highlighted the low telephone ownership in Bulgaria which made interviewing a representative sample by telephone impossible and so a face-to-face design was used in Bulgaria.

4.2 Fieldwork and Questionnaire

Fieldwork in Bulgaria was conducted face-to-face using PAPI (Pen and Paper) in- home interviews by Vitosha Research who were sub-contracted by BMRB. A BMRB executive visited Sofia, prior to the commencement of fieldwork to facilitate the smooth running of the fieldwork.

The questionnaire for Bulgarian fieldwork was, as far as possible, identical in construction to the questionnaire used for telephone interviewing in the other five countries. Key issues with regard to the face-to-face administration of the questionnaire were related to complex routing. Masking and complicated filters that are possible on computer-administered surveys were not feasible for the Bulgarian element of the study, and therefore more detailed routing instructions were printed on the Bulgarian questionnaire. In certain cases this allowed responses to be given on the Bulgarian survey that would not have been permitted by the masking on the CATI system. However, as far as possible, the questionnaire was designed to minimise situations where this would occur.

A pilot was conducted in September 2001, and the interviewers were briefed both by a researcher from BMRB who travelled to Sofia for the pilot, and by the Vitosha Research project team and their Field Manager. Following the pilot some minor amendments were made to the questionnaire to make it more suitable for the face-to-face interview.

For the main stage a briefing session was held on October 6, 2001 with the heads of regional teams and some of the interviewing team. The Vitosha Research interviewer network consists of 28 regional teams, each with a team leader experienced in fieldwork. The briefing session included an explanation of the purpose of the survey, an overview of survey methodology, specific features of the questionnaire and of specific questions, sampling methodology, and specific fieldwork requirements, and also the reconstruction of a fieldwork situation.

Fieldwork dates for the main stages were October 6, 2001 – November 3, 2001.

Page 21 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

4.3 Sampling and Survey Design

The sample was representative of the universe to be covered - the population of Bulgaria aged 16 and over. Sampling points were drawn with probability proportionate to population size and representative of the whole of Bulgaria. Vitosha Research employed the following survey design for a sample of 1750 interviews at a response rate of 71%.

The design was a two-stage selection. The first stage included random selection of 175 Primary Sampling Units (PSUs) which were electoral sections. This sample was drawn from the list of electoral sections (the primary units in the last Parliamentary elections of June 17, 2001). Electoral sections of different size (depending on the number of people in each section) were proportionally represented in the sample. There are 12,787 electoral sections and the average size of an electoral section is about 500 people. Electoral sections cover the whole territory of the country and provide access to the entire population. VR used one of the most complete and up-to-date lists of electoral sections which includes the number and territorial location of registered voters and a unique ID number to identify electors. The coverage of the sampling frame for electoral sections was therefore both the best option available and as close to exhaustive as possible.

Selection of electoral sections included in the sample used systematic sampling. Within the list of electoral sections a cumulative measure of size based on the number of people in each electoral section is computed. The number of people in all sections was then divided by the number of primary sampling units (175) to be included in the sample. The result of this division is then the ‘selection interval’ (SI).

A random start (RS) within the range between 1 and SI was chosen. The second section is the one which contains RS+SI, the third - RS+2SI, etc. The number of PSUs (electoral sections) for the sample was 175.

The sample was proportionally distributed over the country and included all types of locations (cities, towns, villages) such that it was representative of the whole population of the country.

At the second stage a fixed number of names/addresses in each cluster (electoral section) was selected at random. The respondents within the cluster were chosen at random from the Central register (the computer centre of the ESGRAON system). The ESGRAON system covers the whole territory of Bulgaria and the whole population. It is based on personal ID numbers: ten digit numbers where the first six are composed by the birth date DDMMYY. These numbers are used for administrative purposes (taxation, social insurance, address registration, etc.) The Page 22 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report personal numbers enable samples to include people living in a specific territory within a given age range. The result of the selection at the second stage was a list of respondents including personal ID, name, community, and address. The second stage included random selection of 14 respondents. The issued sample size was 2450. The expected non-response was about 20-25%, and that made the expected size of the real sample about 1750. VR used the method outlined by BMRB for Face to Face probability surveys to calculate the response rate and achieved this using the quality procedures detailed in the following section.

4.4 Contact records

Interviewers kept identical Contact Records to show the outcome of their calls at each address. These were paper documents with space to record the selected address, the date and time of calls, the outcome of each call, and the final outcome at each address (successful interview, refusal, no contact with household etc). VR used a standard template for the Contact Record provided by BMRB and the data collected were sent back to BMRB for analysis.

The full breakdown of outcome codes is listed in table 6 below.

Table 6: Bulgarian Outcomes Number of sampling points (PSUs) 175 Number of addresses issued 2450 Total Refusals & Ineligible Addresses 698 Number of interviews stopped underway: 2 Number of net (completed) interviews: 1750

Reasons for refusals 1. Address/respondent ineligible 237 1.1 Insufficient contact details 19 1.2 Empty dwelling 35 1.3 Not a private dwelling 5 1.4 Respondent has moved 152 1.5 Property sold/rented out 13 1.6 Respondent not known at address 13 2. No contact 165 2.1 No contact made with responsible adult after up to 78 three calls 2.2 No contact made with named respondent 87 3. Refusal 94 3.1 Personal refusal by named respondent 55

Page 23 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

3.2 Proxy refusal, on behalf of the named respondent 39 4. Other unsuccessful 202 4.1 Broke appointment 2 4.2 Incapacitated at home during survey period 11 4.3 Away/ in hospital during survey period 148 4.4 Language difficulty prevented interview 2 4.5 Named person has died 16 4.5 Named person is too young 23 Total refusals and ineligible addresses 698

4.5 Response rates

The response rate in Bulgaria was the highest of the six countries. This was expected as it was the only country to use face-to-face fieldwork. Of the issued sample, 2450 addressees, 1750 (71%) resulted in an interview. The conventional approach to face-to-face response rates is to remove addresses classified as deadwood from the denominator in response rate calculations. After this adjustment the headline response rate for Bulgaria rises to 79%.

4.6 Weighting

4.6.1 Respondent and Household Weights The two stage design in Bulgaria ensured that addresses were selected with equal probability and so the sample was self weighting and no design weights were needed. Furthermore, the additional questions to estimate response bias used elsewhere were not asked in Bulgaria and so the second stage weighting used in the other countries was not applied either. Demographic profile weights at a respondent level were used for sex and age interlocked and then region was applied afterwards. This respondent level weight was then divided by number of adults in the household to provide an estimate of household weights to be used on household level data.

Names of weighting variables used on the SPSS data. Respondent weights Dem1: Final respondent weight using rim weighting by sex crossed with age followed by region. Household weights Dem2: Dem1 weight from above divided by number of adults eligible for interview.

Page 24 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

Table 7: Respondent profiles according to weighting scheme used Unweighted dem1 Population Sex Male 46.6 48.6 48.4 Female 53.4 51.4 51.6 Age 16 to 19 3.9 7.5 7.5 20 to 24 6.3 8.7 8.7 25 to 34 16.3 16.2 16.2 35 to 44 13.8 18 18.0 45 to 54 17.1 15.6 15.6 55 to 64 17 15.7 15.7 65 to 74 15.3 12.1 12.1 75 & over 9.9 5.8 5.8 Missing 0.4 0.4 0.4 Work Status In paid work 34.3 36.3 N/A Unemployed 21.9 25.3 N/A Retired from paid work altogether 37 29.1 N/A On maternity leave 1.3 1.5 N/A Looking after family or home 1.4 1.5 N/A Full-time student/at school 3.1 5.5 N/A Long term sick or disabled 0.9 0.8 N/A Household Size 1 11.8 9.6 N/A 2 28.3 25.2 N/A 3 21.9 23 N/A 4 21.8 25 N/A 59.19.6N/A 64.24.6N/A 71.11.3N/A 80.90.8N/A 90.30.3N/A 10 0.5 0.5 N/A More than 10 0.1 0.1 N/A Number of PCs Don't Know 0.3 0.2 N/A One 4.9 5.8 N/A Two 0.3 0.4 N/A Region Sofia - City (BULG.) 13.1 14.9 14.9 Varna, Dobruch, Shoumen 9.1 9.9 9.9 Veliko Tarnovo, Gabrovo, Lovec 13.3 10.7 10.7 Vidin, Vratca, Montana 7.5 11.9 11.9 Razgrad, Rousse, Silistra, Targovishta 9.6 7.3 7.3 Bourgas, Silven, Iambol 10.4 14.4 14.4 Pazardjik, Plovdiv, Smolian 13.8 8.9 8.9 Blagoevgrad, Kustendil, Pernik, Sofia 12.2 11.4 11.4 Page 25 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

Region Kurdjali, Stara Zagora, Haskovo 10.9 10.6 10.6 No household population were available for comparison.

Table 8: Household profile according to weighting scheme used Household Size Unweighted dem2 1 11.8 21.1 2 28.3 28.8 3 21.9 20.8 4 21.8 19.8 5 9.1 5.9 6 4.2 2.5 7 1.1 0.6 8 0.9 0.3 9 0.3 0.1 10 0.5 0.2 More than 10 0.1 0 Number of PCs Don't Know 0.3 0.3 None 94.5 94.2 One 4.9 4.9 Two 0.3 0.5 Region Sofia - City 13.1 17.4 Varna, Dobruch, Shoumen 9.1 9.9 Veliko Tarnovo, Gabrovo, Lovec 13.3 10.6 Vidin, Vratca, Montana 7.5 12.2 Razgrad, Rousse, Silistra, Targovishta 9.6 7.2 Bourgas, Silven, Iambol 10.4 13.9 Pazardjik, Plovdiv, Smolian 13.8 8.5 Blagoevgrad, Kustendil, Pernik, Sofia 12.2 11 Region Kurdjali, Stara Zagora, Haskovo 10.9 9.3

Page 26 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

5 Germany

5.1 Fieldwork and Questionnaire

Fieldwork for Germany was conducted by native German BMRB interviewers from the BMRB telephone centre in Ealing. All interviewers working for BMRB receive 2½ days training, covering both the theory and practical elements of their job.

The German questionnaire was translated and back-checked by BMRB interviewers, an independent translation agency, and by the Digital-living German partner. The survey was set up initially in English and the German translation was then imported to create a German script identical in format and routing to the final English script. This allowed for absolute consistency between the surveys.

Interviewing in Germany was made more difficult by the fact that no cold-calling is permitted within Germany itself. The Germans are therefore less open to telephone interviewing and therefore a higher refusal rate was anticipated for the German survey. Every effort was made to overcome this by having a small, but highly skilled interviewing panel working on the survey, all of whom have experience of working on key projects for BMRB. By keeping very tight controls over the interviewers it was possible to steadily improve performance as the survey progressed. In this way, as high as possible a response rate for Germany was achieved, although the final rate was, as predicted, still lower than in the other four countries,.

In order to further help the German response rate it was also decided to introduce an incentive to encourage response. For the final stages of interviewing, respondents on the main survey were told that a donation would be made to the International Red Cross to the value of 1500 Euros, on behalf of all German respondents. Those respondents who had previously refused and were being re- contacted were additionally offered a personal incentive of 10 Euros.

Fieldwork dates for the main stage of the German survey were 12th September – 12th December 2001.

5.2 German RDD sample Further to the approach described in Appendix A, the database was balanced to county using county level population data. This meant that the RDD sample generated was representative of all residential households rather than just those listed. Geographical postcodes were applied based on Postal codes using a file obtained from the German Post office. Page 27 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

Respondents were randomly selected for interview within the household using the “last birthday” rule. 5.3 Outcome Codes and response rates

Table 9: German Outcome Codes Line German Number Outcome Classification y T1 Total Sample Issued T2+T3+T4+T5 7653 A Business Number Deadwood 306 B Dialler - incomplete Deadwood 36 C Dialler - site out of service Deadwood 121 D Number unobtainable Deadwood 230 E Claimed that no-one over 16 lives here Deadwood 3 T2 Total Deadwood A+B+C+D+E 696 F Unavailable during fieldwork Contact 20 G Soft Refusal Contact 1745 H Unknown at Number Contact 6 I Hard Refusal Contact 1260 J Full Interviews Contact 1753 K Abandoned interview Contact 6 L Respondent quit out of interview Contact 67 M Stopped Interview Contact 19 N Selected Respondent has hearing problem Contact 55 O Selected Respondent has language problem Contact 52 P Duplicate number Contact 159 T3 Total Contacts F+G+H+I+J+K+L+M+N+O+P 5142 Q Pure 10+ Non-Contact 436 R Probably Pure 10+ Non-Contact 41 S At least one other outcome with 10+ unsuccessful Non-Contact 398 T4 Total Non-Contact (10+ unsuccessful calls) Q+R+S 875 T Appointment Unresolved at end of fwork 11 U General Call Back Unresolved at end of fwork 433 V No answer/Answering machine Unresolved at end of fwork 467 W Engaged Unresolved at end of fwork 1 X Dialler - no answer Unresolved at end of fwork 0 Y Dialler - busy Unresolved at end of fwork 0 Z Dialler - nuisance hangup Unresolved at end of fwork 0 AA Dialler - unknown error Unresolved at end of fwork 28 AB Dialler - General failure code Unresolved at end of fwork 0 T5 Total - Unresolved at end of fieldwork T+U+V+W+X+Y+Z+AA+AB 940

German response rates vary according to definition used from 25.2% to 34.1%. Germany had 9% of numbers that were deadwood and a further 875 (11% of sample issued) 10+ unsuccessful calls. Response rate figures for Germany are shown in Table 5.

Page 28 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

5.4 Weighting

5.4.1 Introduction

RDD sample is constructed so that any number (working phone number or not) has the same, or a very similar, chance of being generated as any other number. However, the sample supplied is of course not named sample and once a number has been attributed to a residential address a random selection of an individual within a household needs to be made.

At households where contact was made with an individual who, as the selected respondent or on behalf of the selected respondent, refused to participate in the full survey that individual was requested to answer three questions. Additionally, Interviewers coded the sex of the individual they spoke to. The three questions individuals were asked were:

What was your age at your last birthday? How many people are there in your household including yourself? How many computers, if any, do you have in your home?

Age and sex details collected are usually from those answering the phone rather than from a randomly selected individual and hence can’t be used reliably to correct for non-response bias. However, the household size and number of computers information can be used to profile the full interviews against those who conducted the “mini-interview” and provide additional household details. This information can be used to apply corrective weighting to adjust for some non- response bias prior to standard demographic profile weighting.

There are two levels of weighting for the data - respondent level and household level. The type of weighting that should be used for analysis depends on whether the specific analysis is respondent or household based.

5.4.2 Respondent Weighting

There are three separate components used to derive respondent weighting: weights to equalise unequal selection probabilities (design weights); weights to compensate for differential non-response amongst survey sub-groups (full interviews compared with mini-interviews) demographic weighting to correct for sex, age, region and other profiles (profile weighting dependant on availability of population figures).

Page 29 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

The design weight, part I, will be a weight to compensate for the fact that a single individual will be selected from each household irrespective of the number of occupants; this weight will be proportional to the number of eligible individuals (adults 16 or over) identified in the household.

Part II weight will use the profiles collected from the mini-interview on household size and number of PCs to refine the profile of the sample who gave a full interview. Before doing so, the mini-interview information will be adjusted (weighted by household size – part I weight) to reflect the number of people in households, rather than just households.

The multiplication of the weighting factors from I and II will be a pre-weight applied to the data before part III rim weighting algorithm is run. Part III will correct the profile of the sample (preweighted by I x II) by sex, age, and region.

The calculation of part II weights can be done in different ways and two approaches are discussed below and have been added as variables on the data.

Method 1 for calculating the part II weights uses the profile of all responders to the two questions regarding household size and number of PCs. In Germany there were 1753 full interviews and 594 mini interviews with answers to both the household size and number of PC questions. The 1753 interviews are then weighted to the profile of the 2347 (1753+594) with respect to the household size and PC ownership. This has the effect of producing weights that are close to 1, as the profile of the 1753 interviews must be close to that of the 2347 as it represents 75% of all response to the household size and number of PCs questions.

The alternative, method 2 for calculating part II, is to consider the 594 as representative of all the 1262 refusals. This means that the 1753 interviews will instead be weighted by a combination of the two groups - with the 594 representing all 1262 refusals. This will mean that the full interview data will be weighted to look more like the refusal survey than with Method 1 as the 1753 will now represent about 58% of all responders.

The data have two final respondent weights (depending on which calculation of weight II the user wishes to employ) which can be turned on or off in SPSS. The component parts of these weights are also supplied individually.

The demographic profiling part III will follow the previous stages (the combined effect of part I and part II) and be worked with both versions of part II weighting The demographic rims used for this are first sex and age interlocked, and then region.

Page 30 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

Table 10 shows the three stages of respondent weighting and its effect on sex, age, household size, number of PCs and region.

Names of weighting variables used on the SPSS data. Respondent weights Wt1: Design weight to correct for probability of respondent selection Wt3per: Combination of part I weight with part II weight (under method 1) Wt3pera: Combination of part I weight with part II weight (under method 2) Dem1: Final respondent weight using wt3per as a pre weight before rim weighting by sex crossed with age followed by region. Dem1a: Final respondent weight using wt3pera as a pre weight before rim weighting by sex crossed with age followed by region. Household weights Wt3hhd: Preweight part II (method 1) only Wt3hhda: Preweight part II (method 2) only Dem2: Dem1 weight from above divided by number of adults eligible for interview. Dem2a: Dem1a weight from above divided by number of adults eligible for interview.

Table 10: Respondent profiles according to weighting scheme used Unweighted wt1 wt3per dem1 wt3pera dem1a universe

Male 45.8 47.5 46.9 48.3 46.5 48.3 48.3 Female 54.2 52.5 53.1 51.7 53.5 51.7 51.7 Age 16 to 19 6.4 10.4 9.7 5.4 9.3 5.4 5.4 20 to 24 4.9 6.4 6.1 6.6 6 6.6 6.6 25 to 34 15.1 14.6 14.2 17.7 14 17.7 17.7 35 to 44 22.3 21.9 21 19.5 20.4 19.5 19.5 45 to 54 15.3 16.5 16.4 15 16.3 15 15.0 55 to 64 16 15.3 15.9 16 16.4 16 16.0 65 to 74 12.9 10.3 11.4 11 12.2 11 11.0 75 & over' 6.6 4.3 4.9 8.4 5.3 8.4 8.4 System 0.4 0.3 0.3 0.3 0.3 0.3 0.3 Work Status In paid work 51.8 52.3 50.8 50.8 49.9 50.6 N/A Unemployed 2.9 2.7 2.7 2.9 2.7 3 N/A Retired from paid work 25.7 20.9 23 26.1 24.4 26.3 N/A altogether On maternity leave 2.2 2.2 2.2 2.4 2.2 2.5 N/A Page 31 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

Looking after family or home 8.1 8.6 8.8 7.8 8.9 7.8 N/A Full-time student/at school 8.3 12.2 11.4 8.8 10.8 8.7 N/A Long term sick or disabled 1.0 1.0 1.1 1.1 1.1 1.2 N/A Number of Adults 1 32.4 16.3 16.9 19.2 17.3 19.2 N/A 2 45.4 45.6 47 48.5 47.7 49 N/A 3 15 22.6 21.5 19.8 20.8 19.5 N/A 4 5.5 11 10.4 8.9 10.1 8.8 N/A 5 1.4 3.6 3.4 2.9 3.3 2.8 N/A 6 0.3 0.9 0.8 0.6 0.8 0.6 N/A Household Size Don't Know 0.1 0.1 0.1 0.1 0.1 0.1 N/A 1 23.4 11.8 12.6 14.9 13.2 15 N/A 2 30.5 28.5 30.3 31.3 31.4 31.7 N/A 3 18.7 22.8 21.9 21.6 21.3 21.3 N/A 4 17.7 22.1 21 20 20.3 19.8 N/A 5 7.2 10.4 10 8.7 9.7 8.7 N/A 6 1.5 2.8 2.7 2 2.6 2 N/A 7 0.7 1.2 1.2 1.1 1.2 1.1 N/A 8 0.1 0.1 0.1 0.1 0.2 0.1 N/A 9 0.1 0.1 0.1 0.1 0.1 0.1 N/A Number of PCs Don't Know 0.1 0.1 0.2 0.2 0.2 0.2 N/A None 38.6 31.9 37.5 40.6 41.1 43.3 N/A One 41.6 44.6 40.8 39.3 38.4 37.3 N/A Two 12.8 14.8 13.6 12.8 12.8 12.2 N/A Three 4.2 5.5 5.1 4.6 4.8 4.4 N/A Four or more 2.7 3.1 2.8 2.7 2.7 2.6 N/A Region Schleswig-Holstein 3.4 3.3 3.3 3.4 3.3 3.4 3.4 Hamburg 2.1 1.8 1.8 2.1 1.8 2.1 2.1 Niedersachsen 9.5 9.5 9.8 9.6 9.9 9.6 9.6 Bremen 0.8 0.7 0.7 0.8 0.7 0.8 0.8 Nordrhein-Westfalen 22.7 23.5 23.1 21.9 22.9 21.9 21.9 Hessen 7.8 7.7 7.6 7.4 7.5 7.4 7.4 Rheinland-Pfalz 5.8 5.9 6.0 4.9 6.0 4.9 4.9 Baden-Wurttemberg 13.1 13.3 13.1 12.8 13 12.8 12.8 Bayern 17.5 17.5 17.6 14.8 17.7 14.8 14.8 Saarland 1.7 1.7 1.8 1.3 1.8 1.3 1.3 Berlin 3.2 2.8 2.8 4.1 2.9 4.1 4.1 Brandenburg 2.8 2.9 3.0 3.2 3.0 3.2 3.2 Mecklenburg-Vorpommern 1.1 1.1 1.1 2.2 1.2 2.2 2.2 Sachsen 3.5 3.2 3.3 5.4 3.3 5.4 5.4 Sachsen-Anhalt 2.8 2.6 2.6 3.2 2.6 3.2 3.2 Thuringen 2.3 2.4 2.4 3.0 2.4 3.0 3.0

Page 32 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

5.4.3 Household weighting

The level of household population figures available is less detailed than at a respondent level and varies between country. The household weights therefore have been derived using the respondent information and the same approach has been adopted in each country for consistency.

The household weighting will be similar to respondent weighting but will not include the part I weighting. For each respondent a household weight will be derived from the final two respondent weights by taking the final individual weights and dividing them by the part 1 design weight of individuals.

Table 11 below show the effect of the household weight. No population figures were available for households.

Table 11: Household profiles according to weighting scheme used Household Size Unweighted wt3hhd dem2 wt3hhda dem2a 1 23.4 24.7 28.1 25.6 28.3 2 30.5 31.8 31.6 32.6 31.9 3 18.7 17.7 17 17.1 16.7 4 17.7 16.6 15.4 15.9 15.3 5 7.2 6.8 5.8 6.5 5.8 6 1.5 1.4 1.1 1.4 1.1 7 0.7 0.7 0.7 0.7 0.7 8 0.1 0.1 0.1 0.1 0.1 9 0.1 0.1 0.1 0 0.1 Don't Know 0.1 0.1 0.1 0.1 0.1 Number of PCs Don't Know 0.1 0.1 0.1 0.1 0.1 None 38.6 44.7 47.9 48.6 50.7 One 41.6 37.4 35.5 34.7 33.4 Two 12.8 11.6 10.9 10.8 10.4 Three 4.2 3.8 3.4 3.5 3.2 Four or more 2.7 2.4 2.2 2.2 2.1 Region Schleswig-Holstein 3.4 3.5 3.5 3.5 3.5 Hamburg 2.1 2.1 2.5 2.1 2.5 Niedersachsen 9.5 9.6 9.5 9.8 9.5 Bremen 0.8 0.8 0.9 0.7 0.9 Nordrhein-Westfalen 22.7 22.4 21 22.2 21 Hessen 7.8 7.6 7.3 7.6 7.3 Rheinland-Pfalz 5.8 5.8 4.7 5.8 4.7 Baden-Wurttemberg 13.1 12.9 12.5 12.8 12.5 Bayern 17.5 17.6 14.7 17.6 14.7 Saarland 1.7 1.8 1.3 1.8 1.3 Berlin 3.2 3.3 4.7 3.3 4.7

Page 33 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

Brandenburg 2.8 2.8 3 2.9 3 Mecklenburg- 1.1 1.2 2.3 1.2 2.3 Vorpommern Sachsen 3.5 3.6 6 3.6 6 Sachsen-Anhalt 2.8 2.8 3.4 2.7 3.4 Thuringen 2.3 2.3 2.8 2.3 2.8

Page 34 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

6 ISRAEL

6.1 Fieldwork and Questionnaire

Fieldwork in Israel was conducted over the telephone, using CATI for Hebrew and Russian, and pen and paper for the Arabic element of the study.

The questionnaire was translated into the three languages by the Israeli agency, Teleseker, and these were then checked by BMRB researchers and by a BMRB- approved translation agency. The translations were also sent to ISER for back- checking by the Israeli partner in the study.

The project research manager and Data Processing executive from Teleseker came to BMRB in Ealing for three days before fieldwork started to check the CATI set- up of the script and to ensure the filtering and ordering were as comparable as possible. Israel used NIPO software (whereas BMRB used Quancept), and this was installed temporarily in BMRB to allow for parallel testing of the surveys. This was a vital stage of the set-up as it allowed for rigorous testing of the script and highlighted areas which may have proved problematic at the analysis stage had they not been addressed at set-up.

There were 3 separate surveys in Israel to cover the 3 language groups – Hebrew, Russian and Arabic. The sampling was structured as far as possible to reflect the correct ethnic (and therefore language) groups. The surveys were all based on the final version that was tested at BMRB, and translations were of this questionnaire to ensure full comparability. For the Arabic, the interviewing was conducted on paper and then transferred into a CATI script before the data was run.

Fieldwork dates for Israel were 23rd September – 13th November 2001.

6.2 RDD sample in Israel The land-line telephone owning household penetration in Israel is over 95. The proportion of unlisted households is relatively low at approximately 10-15%. The directory file that lists all residential numbers (apart from ex-directory) is therefore a good basis from which to select a sample to act as ‘seed’ numbers to generate an RDD sample (see Appendix A).

Residential phone numbers are eight digits long in Israel – the first four digits act as a prefix or area code. The directory file covered the vast majority of area codes, if not all, so each residential number has a very similar, if not an equal, chance of being included. Thus, the RDD sample provided a genuine random sample with little bias.

Page 35 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

BMRB’s partner agency in Israel generated RDD sample themselves and BMRB have discussed with them how they ensured it was representative of Israel’s telephoned population. Any sample must give the correct coverage to three ethnic groups – Hebrews (70%), Russians (15%) and Arabs (15%). Area population statistics were checked against the generated sample before fieldwork began.

Respondents were randomly selected for interview within the household using the “last birthday” rule.

6.3 Outcome Codes and Response Rates

Table 12: Israeli Outcome Codes Line Number Outcome Classification Israel T1 Total Sample Issued T2+T3+T4+T5 8721 A Business Number Deadwood 615 B Dialler - incomplete Deadwood 0 C Dialler - site out of service Deadwood 0 D Number unobtainable (including Fax) Deadwood 2034 E Claimed that no-one over 16 lives here Deadwood 14 T2 Total Deadwood A+B+C+D+E 2663 F Unavailable during fieldwork Contact 169 G Soft Refusal Contact 436 H Unknown at Number Contact 0 I Hard Refusal Contact 2191 J Full Interviews Contact 1753 K Abandoned interview Contact 123 L Respondent quit out of interview Contact 0 M Stopped Interview Contact 0 N Selected Respondent has hearing problem Contact 84 O Selected Respondent has language problem Contact 316 P Duplicate number Contact 22 T3 Total Contacts F+G+H+I+J+K+L+M+N+O+P 5094 Q Pure 10+ Non-Contact 777 R Probably Pure 10+ Non-Contact 0 S At least one other outcome with 10+ unsuccessful Non-Contact 0 T4 Total Non-Contact (10+ unsuccessful calls) Q+R+S 777 T Appointment Unresolved at end of fwork 58 U General Call Back Unresolved at end of fwork 4 V No answer/Answering machine Unresolved at end of fwork 122 W Engaged Unresolved at end of fwork 3 X Dialler - no answer Unresolved at end of fwork 0 Y Dialler - busy Unresolved at end of fwork 0 Z Dialler - nuisance hangup Unresolved at end of fwork 0 AA Dialler - unknown error Unresolved at end of fwork 0 AB Dialler - General failure code Unresolved at end of fwork 0 T5 Total - Unresolved at end of fieldwork T+U+V+W+X+Y+Z+AA+AB 187

Page 36 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

Israeli response rates vary from 28.9% to 34.4% according to the definition used. Israel had 31% of numbers that were deadwood and a further 777 (9% of sample issued) 10+ unsuccessful calls. Response rates figures for Israel are shown in Table 5.

6.4 Weighting

6.4.1 Introduction

What was your age at your last birthday? How many people are there in your household including yourself? How many computers, if any, do you have in your home?

6.4.2 Respondent Weighting

There are three separate components used to derive respondent weighting: weights to equalise unequal selection probabilities (design weights);

Page 37 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

weights to compensate for differential non-response amongst survey sub-groups (full interviews compared with mini-interviews) demographic weighting to correct for sex, age, region and other profiles (profile weighting dependant on availability of population figures).

Part II weight will use the profiles collected from the mini-interview on household size and number of PCs to refine the profile of the sample who gave a full interview. Before doing so, the mini-interview information will be adjusted (weighted by household size – part I) to reflect the number of people in households, rather than just households.

The calculation of stage II weights can be done in different ways and two approaches are discussed below and have been added as variables on the data.

Method 1 for calculating the part II weights uses the profile of all responders to the two questions regarding household size and number of PCs. There were 1756 full interviews and 567 mini interviews with answers to both the household size and number of PC questions. The 1756 interviews are then weighted to the profile of the 2333 (1756+567) with respect to the household size and PC ownership. This has the effect of producing weights that are close to 1, as the profile of the 1756 interviews must be close to that of the 2333 as it represents 75% of all response to the household size and number of PCs questions.

The alternative, method 2 for calculating part II, is to consider the 567 as representative of all the 2679 refusals. This means that the 1756 interviews will instead be weighted by a combination of the two groups - with the 567 representing all 2679 refusals. This will mean that the full interview data will be weighted to look more like the refusal survey than with Method 1 as the 1756 will now represent about 40% of all responders.

Page 38 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

The demographic profiling part III will follow the previous stages (the combined effect of part I and part II) and be worked with both versions of part II weighting. The rims used for this are first sex and age interlocked by followed region.

Table 13 shows the three stages of respondent weighting and its effect on sex, age, household size, number of PCs and region. Names of weighting variables used on the SPSS data. Respondent weights Wt1: Design weight to correct for probability of respondent selection Wt3per: Combination of part I weight with part II weight (under method 1) Wt3pera: Combination of part I weight with part II weight (under method 2) Dem1: Final respondent weight using wt3per as a pre weight before rim weighting by sex crossed with age followed by region. Dem1a: Final respondent weight using wt3pera as a pre weight before rim weighting by sex crossed with age followed by region. Household weights Wt3hhd: Preweight part II (method 1) only Wt3hhda: Preweight part II (method 2) only Dem2: Dem1 weight from above divided by number of adults eligible for interview. Dem2a: Dem1a weight from above divided by number of adults eligible for interview.

Table 13: Respondent profiles according to weighting scheme used Unweighted wt1 wt3pe dem1 wt3pera dem1a Univers r e Sex Don't Know 0.1 0.2 0.2 0.2 0.3 0.2 0.2 Male 37.8 37.9 37.8 48.3 37.5 48.3 48.2 Female 62.1 61.9 61.9 51.5 62.3 51.5 51.2 Age 16 to 19 13.2 18.2 18.1 9.9 17.6 9.9 9.9 20 to 24 11.6 14.2 14.3 12.3 14.4 12.3 12.2 25 to 34 25.8 22.7 22.7 20.8 22.7 20.8 20.8 35 to 44 17.3 16.8 16.6 17.3 16 17.3 17.2 45 to 54 11.6 12.6 12.6 16 12.3 16 16.0 55 to 64 8.5 7.3 7.4 9.5 7.9 9.5 9.5 65 to 74 8.3 5.9 5.9 7.8 6.5 7.8 7.8 75 & over' 3.3 2 2 6.1 2.2 6.1 6.1 System 0.5 0.4 0.4 0.3 0.4 0.3 0.2 Working Status Page 39 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

In paid work 52.1 50.2 50 52.7 49.1 52.2 N/A Unemployed 12.5 13.4 13.5 11.9 13.7 12.2 N/A Retired from paid work 12.4 9.2 9.3 15.3 10.1 15.3 N/A altogether On maternity leave 0.9 0.8 0.8 0.6 0.9 0.7 N/A Looking after family or home 7.5 8 8.1 6.3 8.5 6.5 N/A Full-time student/at school 13.5 17.3 17.3 12 16.7 11.6 N/A Long term sick or disabled 1 1.1 1.1 1.3 1.1 1.4 N/A Number of Adults 1 21.6 8.6 8.6 10.4 8.7 10.6 N/A 2 35.3 26.7 26.7 30.2 26.2 29.3 N/A 3 21 23.9 23.6 23.5 22.8 22.9 N/A 4 9.5 14.3 14.5 12.9 15 13.5 N/A 5 8 15.1 15.2 13.5 15.5 13.8 N/A 6 3.1 7 7.1 5.8 7.4 6.1 N/A 7 1 2.6 2.6 2.3 2.6 2.3 N/A 8 0.6 1.7 1.8 1.4 1.9 1.5 N/A Number of people in Household Don't Know 0.3 0.1 0.1 0.1 0.1 0.1 N/A 1 9.4 3.5 3.6 5.3 4.1 5.8 N/A 2 20.3 14.3 14.2 18.2 14.3 17.6 N/A 3 17 15.5 15.4 15.7 15.1 15.4 N/A 4 17.5 17.8 17.7 17.5 17.6 17.7 N/A 5 18.7 22.8 22.7 21.5 22.3 21.4 N/A 6 8.6 12.1 12.1 10 11.9 9.9 N/A 7 4.2 6.8 6.8 5.9 7.0 6.0 N/A 8 2.2 3.5 3.5 2.8 3.7 2.9 N/A 9 0.6 1 1.1 0.9 1.1 1.0 N/A 10 0.8 1.4 1.5 1.1 1.5 1.2 N/A More than 10 0.6 1.3 1.3 0.9 1.4 1.0 N/A Number of PCs None 35.1 31.1 32.2 35.1 38.8 40.9 N/A One 50.7 53.3 52.5 49.6 47.3 45 N/A Two 11.5 12.6 12.4 12.3 11.2 11.4 N/A Three 2.1 2.2 2.2 2.3 2 2.1 N/A Four or more 0.7 0.8 0.7 0.7 0.7 0.6 N/A Region Jerusalem District 8.7 8.4 8.4 10.8 8.4 10.8 10.8 Northern District 20.4 22.3 22.4 15.9 23 15.9 15.9 Haifa District 13.8 13.4 13.4 13.6 13.2 13.6 13.6 Central District 21.8 23 22.9 23.3 22.4 23.3 23.4 Tel Aviv District 20.1 18.2 18.2 20.2 18.1 20.2 20.2 Southern District 12.9 12.5 12.5 13.6 12.7 13.6 13.6 Judea, Samaria 2.2 2.2 2.2 2.5 2.1 2.5 2.5 Language Hebrew 75.9 73.8 73.6 74.6 72.4 74 N/A Russian 11.6 10 10.1 13.4 10.5 13.6 N/A Page 40 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

Arabic 12.5 16.2 16.4 12 17.1 12.5 N/A

6.4.3 Household weighting

Table 14 below show the effect of the household weights. No population figures for households are available.

Table 14: Household profiles according to weighting scheme used Household Size Unweighted wt3hhd dem2 wt3hhda dem2a Don't Know 0.3 0.3 0.3 0.3 0.3 1 9.4 9.6 13 10.8 14.4 2 20.3 20.3 23.9 20.5 23.4 3 17 16.9 15.9 16.5 15.5 4 17.5 17.4 16.2 17.3 16.3 5 18.7 18.6 16.5 18 16.2 6 8.6 8.6 6.7 8.2 6.5 7 4.2 4.2 3.4 4.2 3.4 8 2.2 2.2 1.7 2.2 1.7 9 0.6 0.6 0.5 0.6 0.5 10 0.8 0.8 0.6 0.8 0.6 More than 10 0.6 0.6 1.3 0.7 1.3 Number of PCs None 35.1 36.4 40.3 44.2 47.1 One 50.7 49.6 46.3 43.6 40.9 Two 11.5 11.3 10.7 9.9 9.6 Three 2.1 2 2 1.7 1.8 Four or more 0.7 0.7 0.7 0.6 0.6 Region Jerusalem District 8.7 8.7 11.4 8.8 11.5 Northern District 20.4 20.5 14.3 21 14.2 Haifa District 13.8 13.8 13.6 13.6 13.7 Central District 21.8 21.8 22.3 21.4 22.4 Tel Aviv District 20.1 20.1 21.8 20 21.9 Southern District 12.9 12.9 14.1 13.1 14.0 Judea, Samaria 2.2 2.2 2.3 2.1 2.3

Page 41 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

7 ITALY

7.1 Fieldwork and Questionnaire

Fieldwork for Italy was conducted by native Italian BMRB interviewers from the BMRB telephone centre in Ealing. All interviewers working for BMRB receive 2½ days training, covering both the theory and practical elements of their job.

The Italian questionnaire was translated and back-checked by BMRB interviewers, an independent translation agency, and by the Digital-living Italian partner. The survey was set up initially in English and the Italian translation was then imported to create an Italian script identical in format and routing to the final English script. This allowed for absolute consistency between the surveys.

Fieldwork dates for the Italian stage of the survey were 12th September – 31st October 2001.

7.2 Italian RDD sample Further to the approach discussed in Appendix A the database was balanced according to Province populations. Province and Region codes were applied based on Post code/exchange combinations.

Respondents were randomly selected for interview within the household using the “last birthday” rule.

7.3 Outcome Codes and response rates

Table 15: Italian Outcome Codes Line Number Outcome Classification Italy T1 Total Sample Issued T2+T3+T4+T5 6841 A Business Number Deadwood 425 B Dialler - incomplete Deadwood 755 C Dialler - site out of service Deadwood 371 D Number unobtainable Deadwood 116 E Claimed that no-one over 16 lives here Deadwood 4 T2 Total Deadwood A+B+C+D+E 1671 F Unavailable during fieldwork Contact 279 G Soft Refusal Contact 1189 H Unknown at Number Contact 2 I Hard Refusal Contact 986 J Full Interviews Contact 1762 K Abandoned interview Contact 12 L Respondent quit out of interview Contact 153 M Stopped Interview Contact 18 N Selected Respondent has hearing problem Contact 10 O Selected Respondent has language problem Contact 8 Page 43 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

P Duplicate number Contact 56 T3 Total Contacts F+G+H+I+J+K+L+M+N+O+P 4475 Q Pure 10+ Non-Contact 288 R Probably Pure 10+ Non-Contact 47 S At least one other outcome with 10+ unsuccessful Non-Contact 169 T4 Total Non-Contact (10+ unsuccessful calls) Q+R+S 504 T Appointment Unresolved at end of fwork 1 U General Call Back Unresolved at end of fwork 142 V No answer/Answering machine Unresolved at end of fwork 1 W Engaged Unresolved at end of fwork 4 X Dialler - no answer Unresolved at end of fwork 0 Y Dialler - busy Unresolved at end of fwork 1 Z Dialler - nuisance hangup Unresolved at end of fwork 4 AA Dialler - unknown error Unresolved at end of fwork 38 AB Dialler - General failure code Unresolved at end of fwork 0 T5 Total - Unresolved at end of fieldwork T+U+V+W+X+Y+Z+AA+AB 191

Italian response rates vary according to definition used from 34.1% to 39.4%. Italy had 25% of numbers that were deadwood and a further 504 (7% of sample issued) 10+ unsuccessful calls. Response rate figures for Italy are shown in Table 5.

7.4 Weighting 7.4.1 Introduction

What was your age at your last birthday? How many people are there in your household including yourself? How many computers, if any, do you have in your home?

Page 44 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report conducted the “mini-interview” and provide additional household details. This information can be used to apply corrective weighting to adjust for some non- response bias prior to standard demographic profile weighting.

7.4.2 Respondent Weighting

The calculation of part II weights can be done in different ways and two approaches are discussed below and have been added as variables on the data.

Method 1 for calculating the part II weights uses the profile of all responders to the two questions regarding household size and number of PCs. In Italy there were 1762 full interviews and 424 mini interviews with answers to both the household size and number of PC questions. The 1762 interviews are then weighted to the profile of the 2186 (1762+424) with respect to the household size and PC ownership. This has the effect of producing weights that are close to 1, as the profile of the 1762 interviews must be close to that of the 2186 as it represents 81% of all responses to the household size and number of PCs questions. Page 45 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

The alternative, method 2 for calculating part II, is to consider the 567 as representative of all the 999 refusals. This means that the 1762 interviews will instead be weighted by a combination of the two groups - with the 424 representing all 999 refusals. This will mean that the full interview data will be weighted to look more like the refusal survey than with Method 1 as the 1762 will now represent about 64% of all responders.

Table 16 shows the three stages of respondent weighting and its effect on sex, age, household size, number of PCs and region.

Table 16: Respondent profiles according to weighting scheme used Unweighted wt1 wt3per dem1 wt3pera Dem1a Univers e Page 46 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

Sex Male 40.3 42.2 41.1 47.9 40.1 47.9 48.1 Female 59.7 57.8 58.9 52.1 59.9 52.1 51.9 Age 16 to 19 10.3 14.5 13.6 5.1 12.7 5.1 5.0 20 to 24 8.4 11.8 11.1 7.6 10.4 7.6 7.4 25 to 34 19.4 20.1 19.7 18.8 19.4 18.8 18.3 35 to 44 22.9 19.4 19.5 17.5 19.6 17.5 17.1 45 to 54 15.1 15.3 14.9 15.6 14.6 15.6 15.2 55 to 64 10.7 9.4 9.9 14 10.4 14 13.6 65 to 74 8.9 6.5 7.7 11.8 8.8 11.8 11.5 75 & over 3.9 2.7 3.3 9.2 3.8 9.2 9.0 System 0.3 0.3 0.3 0.3 0.3 0.3 0.3 Work Status In paid work 50.2 49.1 48.2 48.2 47.4 48.1 N/A Unemployed 5.6 6.6 6.6 5.3 6.5 5.4 N/A Retired from paid work 15.4 11.8 13.3 22.8 14.7 22.8 N/A altogether On maternity leave 0.6 0.4 0.5 0.3 0.5 0.4 N/A Looking after family or home 13.3 11.9 12.7 13.3 13.5 13.4 N/A Full-time student/at school 14.5 19.8 18.4 9.6 17.1 9.4 N/A Long term sick or disabled 0.4 0.3 0.3 0.5 0.4 0.6 N/A Number of Adults 1 21.7 8.8 11 14.3 12.9 15.9 N/A 2 36.8 30 32 35.8 33.9 36.9 N/A 3 21.6 26.4 25.3 23.4 24.3 22.5 N/A 4 15.3 25 22.8 19.4 20.7 18 N/A 5 4.1 8.5 7.7 6 7.1 5.7 N/A 6 0.5 1.1 1 0.9 0.9 0.8 N/A 7 0.1 0.2 0.1 0.2 0.1 0.2 N/A Household Size Don't Know 0.1 0 0 0 0 0.1 N/A 1 11.9 4.9 6.5 9.9 8 11.1 N/A 2 20.4 15.4 17.4 22.7 19.2 23.6 N/A 3 24.4 24.1 23.8 23.5 23.5 23.1 N/A 4 30.1 36.1 34.1 29.3 32.2 28.1 N/A 5 10.1 14.6 13.7 11 12.9 10.6 N/A 6 2.4 3.7 3.5 2.9 3.2 2.7 N/A 7 0.3 0.5 0.5 0.3 0.5 0.3 N/A 8 0.2 0.4 0.4 0.4 0.4 0.3 N/A 9 0.1 0.1 0.1 0.1 0.1 0.1 N/A Number of PCs None 46.4 41.8 47.7 52.3 53.1 56.8 N/A One 43.5 46.9 42.1 38.6 37.8 35 N/A Two 8.2 9.2 8.3 7.3 7.5 6.7 N/A Three 1.5 1.5 1.3 1.2 1.2 1.1 N/A Four or more 0.5 0.6 0.5 0.5 0.4 0.5 N/A Region Page 47 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

Nord Ovest 11.2 10.1 10.5 10.7 10.8 10.7 10.7 Lombardia 17.2 16.2 16.2 16 16.1 16 16.0 Nord Est 13.1 12.5 12.6 11.7 12.8 11.7 11.7 Emilia-Romagna 7.9 7.4 7.5 7.2 7.6 7.2 7.2 Centro 8.6 8.4 8.4 10.4 8.4 10.4 10.4 Lazio 8.1 7.6 7.6 9.2 7.7 9.2 9.2 Abruzzo-Molise 2.8 2.8 2.8 2.8 2.8 2.8 2.8 Campania 9.6 11.6 11.2 9.4 10.9 9.4 9.4 Sud 9.3 10.5 10.3 11.3 10.1 11.3 11.3 Sicilia 8.9 9.3 9.4 8.4 9.5 8.4 8.4 Sardegna 3.2 3.5 3.4 2.9 3.3 2.9 2.9

7.4.3 Household weighting

Table 17 below shows the effect of the household weights. No population figures are available for households.

Table 17: Respondent profiles according to weighting scheme used Household Size Unweighted Wt3hhd Dem2 Wt3hhda Dem2a Don't Know 0.1 0.1 0.1 0.1 0.1 1 11.9 15.2 21.3 17.8 23.2 2 20.4 22 26.3 23.3 26.6 3 24.4 23.1 21 22.1 20.3 4 30.1 27.7 22.2 25.7 21.2 5 10.1 9.2 7.1 8.5 6.8 6 2.4 2.1 1.7 1.9 1.6 7 0.3 0.3 0.2 0.3 0.2 8 0.2 0.2 0.2 0.2 0.1 90.10000 Number of PCs None 46.4 54.5 60.3 61.1 65.4 One 43.5 36.9 32.4 31.5 28.2 Two 8.2 7 5.9 6 5.1 Three 1.5 1.2 1.1 1.1 0.9 Four or more 0.5 0.4 0.4 0.3 0.3 Region Page 48 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

Nord Ovest 11.2 11.7 12.2 12.2 12.3 Lombardia 17.2 17 17.2 16.9 17.1 Nord Est 13.1 13.4 12.2 13.6 12.3 Emilia-Romagna 7.9 8 7.7 8.1 7.7 Centro 8.6 8.6 10.5 8.5 10.4 Lazio 8.1 8.1 9.8 8.1 9.8 Abruzzo-Molise 2.8 2.8 2.6 2.8 2.6 Campania 9.6 9.2 7.8 8.9 7.7 Sud 9.3 9.1 9.7 8.8 9.7 Sicilia 8.9 9.1 7.7 9.2 7.8 Sardegna 3.2 3.1 2.7 3 2.7

8 NORWAY

8.1 Fieldwork and Questionnaire

Fieldwork for Norway was conducted by native Norwegian BMRB interviewers from the BMRB telephone centre in Ealing. All interviewers working for BMRB receive 2½ days’ training, covering both the theory and practical elements of their job.

The Norwegian questionnaire was translated and back-checked by BMRB interviewers, an independent translation agency, and by the Digital-living Norwegian partner. The survey was set up initially in English and the Norwegian translation was then imported to create a Norwegian script identical in format and routing to the final English script. This allowed for absolute consistency between the surveys.

Fieldwork dates for the Norwegian element were 12th September – 18th November 2001.

8.2 Norwegian RDD sample Further to the approach discussed in Appendix A the database was balanced by county so that the sample was as representative as possible of the telephoned household population.

Respondents were randomly selected for interview within the household using the “last birthday” rule.

8.3 Outcome Codes and response rates

Table 18: Norwegian Outcome Codes Line Number Outcome Classification Norway T1 Total Sample Issued T2+T3+T4+T5 6035 Page 49 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

A Business Number Deadwood 524 B Dialler - incomplete Deadwood 12 C Dialler - site out of service Deadwood 125 D Number unobtainable Deadwood 321 E Claimed that no-one over 16 lives here Deadwood 1 T2 Total Deadwood A+B+C+D+E 983 F Unavailable during fieldwork Contact 4 G Soft Refusal Contact 719 H Unknown at Number Contact 1 I Hard Refusal Contact 1577 J Full Interviews Contact 1756 K Abandoned interview Contact 13 L Respondent quit out of interview Contact 82 M Stopped Interview Contact 2 N Selected Respondent has hearing problem Contact 21 O Selected Respondent has language problem Contact 9 P Duplicate number Contact 54 T3 Total Contacts F+G+H+I+J+K+L+M+N+O+P 4238 Q Pure 10+ Non-Contact 388 R Probably Pure 10+ Non-Contact 61 S At least one other outcome with 10+ unsuccessful Non-Contact 111 T4 Total Non-Contact (10+ unsuccessful calls) Q+R+S 560 T Appointment Unresolved at end of fwork 0 U General Call Back Unresolved at end of fwork 138 V No answer/Answering machine Unresolved at end of fwork 48 W Engaged Unresolved at end of fwork 0 X Dialler - no answer Unresolved at end of fwork 21 Y Dialler – busy Unresolved at end of fwork 3 Z Dialler - nuisance hangup Unresolved at end of fwork 0 AA Dialler - unknown error Unresolved at end of fwork 39 AB Dialler - General failure code Unresolved at end of fwork 5 T5 Total - Unresolved at end of fieldwork T+U+V+W+X+Y+Z+AA+AB 254

Norwegian response rates vary according to definition used from 34.8% to 41.4%. Norway had 16% of numbers that were deadwood and a further 560 (9% of sample issued) 10+ unsuccessful calls. Whichever response rate definition is used Norway had the highest response rates of all the telephoned countries. Response rates figures for Norway are shown in Table 5.

8.4 Weighting

8.4.1 Introduction

RDD sample is constructed so that any number (working phone number or not) has the same chance, or very similar chance, of being generated as any other number. However, the sample supplied is of course not named sample and once a

Page 50 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report number has been attributed to a residential address a random selection of an individual within a household needs to be made.

For households where contact was made with an individual, who as the selected respondent or on behalf of the selected respondent, refused to participate in the full survey they were requested to answer three questions. Additionally, Interviewers coded the sex of the individual. The three questions individuals were asked were:

What was your age at your last birthday? How many people are there in your household including yourself? How many computers, if any, do you have in your home?

Age and sex details collected are usually from those answering the phone rather than from a randomly selected individual and hence can’t be used reliably to correct for response bias. However, the household size and number of computers information can be used to profile the full interviews against those who conducted the “mini-interview” and provide additional household details. This information can be used to apply corrective weighting to adjust for some non-response bias prior to standard demographic profile weighting.

8.4.2 Respondent Weighting There are three separate components used to derive respondent weighting: weights to equalise unequal selection probabilities (design weights); weights to compensate for differential non-response amongst survey sub-groups (full interviews compared with mini-interviews) demographic weighting to correct for sex, age, region and other profiles (profile weighting dependant on availability of population figures).

The design weight, part I, will be a weight to compensate for the fact that a single individual will be selected from each household irrespective of how many individuals were identified there; this will be proportional to the number of eligible individuals (adults 16 or over) identified at the household.

Part II weight will use the profiles collected from the mini-interview on household size and number of PCs. The mini-interview information will be adjusted (weighted by household size) to reflect people in households, rather than just households. Page 51 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

The multiplication of the weighting factors from I and II will be a pre-weight applied to the data before part III rim weighting algorithm is run. These will correct the profile of the sample (preweighted by I x II) by sex, age and region.

The calculation of part II weights can be done in different ways and two approaches are discussed below and have been added as variables on the data.

Method 1 for calculating the part II weights uses the profile of all responders to the two questions regarding household size and number of PCs. There were 1756 full interviews and 593 mini interviews with answers to both the household size and number of PC questions. The 1756 interviews are then weighted to the profile of the 2349 (1756+593) with respect to the household size and PC ownership. This has the effect of producing weights that are close to 1, as the profile of the 1756 interviews must be close to that of the 2349 as it represents 75% of all responses to the household size and number of PCs questions.

The alternative, method 2 for calculating part II, is to consider the 593 as representative of all the 1584 refusals. This means that the 1756 interviews will instead be weighted by a combination of the two groups - with the 593 representing all 1584 refusals. This will mean that the full interview data will be weighted to look more like the refusal survey than with Method 1 as the 1756 will now represent about 53% of all responders.

Table 19 shows the three stages of respondent weighting and its effect on sex, age, household size, number of PCs and region.

Page 52 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

Dem1: Final respondent weight using wt3per as a pre weight before rim weighting by sex crossed with age followed by region. Dem1a: Final respondent weight using wt3pera as a pre weight before rim weighting by sex crossed with age followed by region. Household weights Wt3hhd: Preweight part II (method 1) only Wt3hhda: Preweight part II (method 2) only Dem2: Dem1 weight from above divided by number of adults eligible for interview. Dem2a: Dem1a weight from above divided by number of adults eligible for interview.

Table 19: Respondent profiles according to weighting scheme used Unweighted wt1 wt3pe dem1 wt3pera Dem1a Univers r e Sex Male 48.2 49.4 48.9 48.9 48.4 48.9 48.8 Female 51.8 50.6 51.1 51.1 51.6 51.1 51.2 Age 16 to 19 5.9 9.4 8.8 6.0 8.2 6.0 6.0 20 to 24 4.8 6.3 6 7.8 5.7 7.8 7.8 25 to 34 19.1 18 17.6 19.2 17.2 19.2 19.2 35 to 44 22.4 21.9 21.2 18.4 20.6 18.4 18.5 45 to 54 19.4 19.9 19.7 17.3 19.4 17.3 17.4 55 to 64 14.2 13.2 13.6 12.3 13.9 12.3 12.3 65 to 74 8.1 6.8 7.6 8.9 8.4 8.9 9.0 75 & over' 5.8 4.3 5.4 9.9 6.3 9.9 9.9 System 0.2 0.2 0.2 0.2 0.2 0.2 Work Status In paid work 65.6 65.4 63.8 61 62.4 60.6 N/A Unemployed 1.4 1.4 1.4 1.3 1.3 1.3 N/A Retired from paid work altogether 15.4 12.4 14.4 19.5 16.3 19.7 N/A On maternity leave 1.2 1.2 1.2 1.2 1.2 1.2 N/A Looking after family or home 2.1 1.9 1.9 1.9 2 1.9 N/A Full-time student/at school 9.3 12.8 12 10.4 11.4 10.3 N/A Long term sick or disabled 5.1 4.9 5.2 4.7 5.5 4.9 N/A Number of Adults 1 33 17.4 19.1 21.6 20.9 22.8 N/A 2 51 53.9 53.9 54.3 53.7 53.9 N/A 3 11 17.5 16.9 14.9 16.4 14.8 N/A 4 3.8 8.1 7.3 6.6 6.5 6.1 N/A 5 0.9 2.4 2.2 2 1.9 1.9 N/A Page 53 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

6 0.2 0.7 0.7 0.6 0.6 0.5 N/A Household Size 1 21 11.1 12.9 15.4 14.6 16.6 N/A 2 32.9 32.1 32.9 34.6 33.5 34.5 N/A 3 16.2 18.5 18 16.8 17.6 16.7 N/A 4 19 22.9 21.9 20.1 20.9 19.8 N/A 5 8.2 10.8 10.1 8.8 9.5 8.6 N/A 6 2.3 3.8 3.5 3.4 3.2 3.2 N/A 7 0.3 0.5 0.5 0.5 0.5 0.5 N/A 8 0.1 0.3 0.2 0.2 0.2 0.2 N/A More than 10 0.10000 0N/A Number of PCs None 24.8 20.4 25.4 29.9 29.9 33 N/A One 46.7 46.6 43.7 41.7 41.1 39.6 N/A Two 19.9 22.6 21.1 19.8 19.9 19.1 N/A Three 5.9 6.9 6.5 5.8 6.1 5.6 N/A Four or more 2.6 3.5 3.2 2.8 3 2.7 N/A Region

Akershus 13.3 13.6 13.4 10.5 13.2 10.5 10.5 Aust-Agder 2.5 2.5 2.5 2.3 2.6 2.3 2.3 Buskerud 6.2 6.1 6.1 5.3 6.1 5.3 5.3 Finnmark 1.7 1.6 1.7 1.7 1.8 1.7 1.6 Hedmark 3.9 3.5 3.6 4.2 3.7 4.2 4.2 Hordaland 10.1 10.6 10.6 9.7 10.5 9.7 9.7 More og Romsda 4.5 4.8 4.8 5.4 4.8 5.4 5.4 Nord-Trondelag 2.4 2.7 2.7 2.8 2.7 2.8 2.8 Nord-land 4.7 4.8 4.9 5.3 4.9 5.3 5.3 Oppland 3.4 3.6 3.6 4.1 3.6 4.1 4.1 Oslo 10 8.8 8.9 11.3 9 11.3 11.3 Ostfold 6 6.1 6.1 5.6 6.1 5.6 5.6 Rogaland 9.5 9.5 9.4 8.3 9.3 8.3 8.3 Sogn og Fjordane 2 2.2 2.1 2.4 2.1 2.4 2.4 Sor-Trondelag 4.6 4.6 4.5 5.9 4.4 5.9 5.9 Telemark 3.1 2.9 3.0 3.7 3 3.7 3.7 Troms 3.4 3.6 3.7 3.4 3.7 3.4 3.4 Vest-Agder 3.9 3.8 3.8 3.5 3.7 3.5 3.5 Vestfold 4.9 4.7 4.8 4.8 4.8 4.8 4.8

8.4.3 Household weighting

Page 54 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

Table 20: Household profiles according to weighting scheme used Household Size Unweighted Wt3hhd dem2 Wt3hhda Dem2a 1 21 23.7 27.5 26.2 29.2 2 32.9 32.9 33.5 32.9 32.9 3 16.2 15.4 13.8 14.7 13.6 4 19 17.8 16 16.7 15.6 5 8.2 7.6 6.6 7.1 6.3 6 2.3 2.1 2 1.9 1.8 7 0.3 0.3 0.4 0.3 0.4 8 0.1 0.1 0.1 0.1 0.1 More than 10 0.1 0.1 0 0 0 Number of PCs None 24.8 31.3 36.5 36.9 40.4 One 46.7 42.7 39.9 39.2 37.3 Two 19.9 18.2 16.7 16.8 15.8 Three 5.9 5.4 4.8 5 4.6 Four or more 2.6 2.4 2.1 2.2 2 Region Akershus 13.3 13 10 12.8 9.9 Aust-Agder 2.5 2.6 2.3 2.6 2.3 Buskerud 6.2 6.1 5.2 6 5.1 Finnmark 1.7 1.7 1.6 1.8 1.6 Hedmark 3.9 4.1 4.8 4.2 4.8 Hordaland 10.1 10.1 9.4 10 9.4 More og Romsda 4.5 4.5 5.1 4.5 5.1 Nord-Trondelag 2.4 2.4 2.5 2.4 2.5 Nord-land 4.7 4.8 5.2 4.8 5.2 Oppland 3.4 3.4 3.9 3.5 3.9 Oslo 10 10.1 12.6 10.2 12.6 Ostfold 6 6 5.5 6.1 5.6 Rogaland 9.5 9.3 8.1 9.1 8.1 Sogn og Fjordane 2 2 2.1 1.9 2.1 Sor-Trondelag 4.6 4.5 5.9 4.4 5.9 Telemark 3.1 3.2 4 3.3 4 Troms 3.4 3.5 3.2 3.5 3.2 Vest-Agder 3.9 3.9 3.6 3.9 3.6 Vestfold 4.9 5 5 5 5

Page 55 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

9 UNITED KINGDOM

9.1 Fieldwork and Questionnaire

Fieldwork for the UK was conducted by BMRB interviewers from the BMRB telephone centre in Ealing.

The UK survey was piloted on 3rd July 2001 in Ealing. Representatives from ISER (Randy Banks, Jon Burton) were in attendance as well as researchers from BMRB. The pilot is a vital element of any survey, and particularly for a survey with the complexity of the Digital living project. Following the pilot, refinements were made to the questionnaire which improved the flow and structure of interviewing, and some questions were removed to keep the total length within the maximum planned time of 25 minutes. The pilot was also used to test the survey introduction to ensure that the most effective wording was used to maximise response rate.

The main fieldwork for the UK started on 3rd September 2001. The interviewing team was fully briefed by the BMRB research team and Heather Laurie, Survey Manager at ISER was in attendance.

UK fieldwork was conducted between 3rd September – 18th November 2001.

9.2 EPSeM Sample A full description of this type of RDD number generation is provided in Appendix A.

Respondents were randomly selected for interview within the household using the “last birthday” rule.

9.3 Outcome Codes and response rates

Table 21: UK outcome codes Line Number Outcome Classification UK T1 Total Sample Issued T2+T3+T4+T5 11828 A Business Number Deadwood 2777 B Dialler - incomplete Deadwood 160 C Dialler - site out of service Deadwood 1683 D Number unobtainable Deadwood 148 E Claimed that no-one over 16 lives here Deadwood 0 T2 Total Deadwood A+B+C+D+E 4768 F Unavailable during fieldwork Contact 18 G Soft Refusal Contact 321 H Unknown at Number Contact 5

Page 56 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

I Hard Refusal Contact 2677 J Full Interviews Contact 1760 K Abandoned interview Contact 10 L Respondent quit out of interview Contact 118 M Stopped Interview Contact 1 N Selected Respondent has hearing problem Contact 6 O Selected Respondent has language problem Contact 10 P Duplicate number Contact 33 T3 Total Contacts F+G+H+I+J+K+L+M+N+O+P 4959 Q Pure 10+ Non-Contact 1619 R Probably Pure 10+ Non-Contact 83 S At least one other outcome with 10+ unsuccessful Non-Contact 191 T4 Total Non-Contact (10+ unsuccessful calls) Q+R+S 1893 T Appointment Unresolved at end of fwork 0 U General Call Back Unresolved at end of fwork 1 V No answer/Answering machine Unresolved at end of fwork 9 W Engaged Unresolved at end of fwork 0 X Dialler - no answer Unresolved at end of fwork 5 Y Dialler - busy Unresolved at end of fwork 2 Z Dialler - nuisance hangup Unresolved at end of fwork 4 AA Dialler - unknown error Unresolved at end of fwork 186 AB Dialler - General failure code Unresolved at end of fwork 1 T5 Total - Unresolved at end of fieldwork T+U+V+W+X+Y+Z+AA+AB 208

UK response rates vary according to definition used from 24.9% to 35.5%. The UK was the only country used where EPSeM sample was provided and consequently the UK had 40% deadwood. There were also nearly 1,900 (16% of sample issued) 10+ unsuccessful calls and this is why the range of response rates varies by more than 10 percentage points. Response rate figures for the UK are shown in Table 5.

9.4 Weighting 9.4.1 Introduction

EPSeM sample is constructed so that any number (working phone number or not) has the same chance of being generated as any other number. However, the sample supplied is of course not named sample and once a number has been attributed to a residential address a random selection of an individual within a household needs to be made.

What was your age at your last birthday? How many people are there in your household including yourself? How many computers, if any, do you have in your home?

Page 57 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

9.4.2 Respondent Weighting

Part II weight will use the profiles collected from the mini-interview on household size and number of PCs. The mini-interview information will be adjusted (weighted by household size – part I weight) to reflect people in households, rather than just households.

The calculation of part II weights can be done in different ways and two approaches are discussed below and have been added as variables on the data.

Method 1 for calculating the part II weights uses the profile of all responders to the two questions regarding household size and number of PCs. There were 1760 full interviews and 559 mini interviews with answers to both the household size and number of PC questions. The 1760 interviews are then weighted to the profile of the 2319 (1760+559) with respect to the household size and PC ownership. This Page 58 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report has the effect of producing weights that are close to 1, as the profile of the 1760 interviews must be close to that of the 2319 as it represents 76% of all responses to the household size and number of PCs questions.

The alternative, method 2 for calculating part II, is to consider the 559 as representative of all the 2687 refusals. This means that the 1760 interviews will instead be weighted a combination of the two groups - with the 559 representing all 2687 refusals. This will mean that the full interview data will be weighted to look more like the refusal survey than with Method 1 as the 1760 will now represent about 40% of all responders.

Table 22 shows the three stages of respondent weighting and its effect on sex, age, household size, number of PCs and region.

Page 59 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

Table 22: Respondent profiles according to weighting scheme used Unweighted wt1 wt3per dem1 wt3per dem1a a Sex Male 43.4 43.9 43.7 48.7 43.4 48.7 48.8 Female 56.6 56.1 56.3 51.3 56.6 51.3 51.2 Age 16 to 19 6.3 10.3 9.7 6.1 8.7 6.1 6.1 20 to 24 6.4 8.8 8.5 7.4 8 7.4 7.5 25 to 34 21 19.2 19.3 18.6 19.4 18.6 18.7 35 to 44 22.8 21.9 21.6 18.8 21.2 18.8 18.9 45 to 54 15.6 16.9 16.7 16.3 16.3 16.3 16.4 55 to 64 13.1 11.9 12.3 12.8 13 12.8 12.9 65 to 74 8.9 6.8 7.4 10.3 8.3 10.3 10.3 75 & over' 5.5 3.7 4.1 9.2 4.7 9.2 9.2 System 0.5 0.4 0.4 0.4 0.4 0.4 Work Status In paid work 60.4 62 61.1 58.1 59.8 57.6 N/A Unemployed 5.2 5 5.1 4.5 5.2 4.7 N/A Retired from paid work 18.7 14.6 15.8 23 17.5 23.1 N/A altogether On maternity leave 0.6 0.6 0.6 0.4 0.6 0.4 N/A Looking after family or 5.8 6 6 4.9 6.1 5 N/A home Full-time student/at school 6.2 9.2 8.7 6.3 7.9 6.2 N/A Long term sick or disabled 3.1 2.6 2.8 2.8 3 3 N/A Number of Adults 1 31.8 16 16.9 20.4 18.2 21.1 N/A 2 48.4 48.7 49.8 51.3 51.7 52.4 N/A 3 12.4 18.8 17.8 15.8 16.3 14.8 N/A 4 5.3 10.6 10 8.3 8.9 7.7 N/A 5 1.4 3.6 3.3 2.4 2.9 2.3 N/A 6 0.6 1.9 1.8 1.4 1.7 1.3 N/A 7 0.1 0.4 0.4 0.4 0.3 0.4 N/A Household Size Don't Know 0.1 0 0 0 0 0 N/A 1 21.7 11 11.8 15.6 12.9 16.2 N/A 2 31.5 29.6 30.8 34.4 32.8 35.3 N/A 3 18.5 20.6 20.1 18.5 19.3 18.1 N/A 4 17.8 22.1 21.4 18.5 20.3 18 N/A 5 6.8 10 9.5 7.7 8.8 7.4 N/A 6 2.854.844.53.8N/A 7 0.7 1.5 1.4 1.1 1.2 1 N/A 8 0.1 0.1 0.1 0.1 0.1 0.1 N/A 10 0.1 0.1 0.1 0 0.1 0 N/A More than 10 0.1 0 0 0 0 0 N/A Number of PCs Don't Know 0.2 0.1 0.2 0.1 0.2 0.2 N/A Page 60 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

None 38.1 32.2 36.0 40.4 41.7 44.8 N/A One 42.7 44.9 42.4 39.8 38.7 36.7 N/A Two 12.9 15.7 14.7 13.3 13.3 12.3 N/A Three 44.64.443.93.7N/A Four or more 2.2 2.5 2.4 2.4 2.2 2.3 N/A Region North East 4.3 4.5 4.5 4.3 4.6 4.3 4.3 North West 12.7 12.6 12.6 11.5 12.6 11.5 11.5 Yorkshire /Humberside 8.1 8 8.0 8.5 8.1 8.4 8.4 East Midlands 5.6 5.7 5.7 7.1 5.7 7.1 7.1 West Midlands 8.6 8.9 8.9 8.9 8.8 8.9 8.9 East Anglia 7.6 7.4 7.4 9.2 7.4 9.1 9.1 London 10.9 10.7 10.5 12.3 10.3 12.3 12.3 South East 12.5 12.3 12.1 13.6 11.9 13.6 13.6 South West 8.6 8.3 8.3 8.4 8.3 8.4 8.4 Wales 5.5 5.6 5.6 4.9 5.6 4.9 4.9 Scotland 11.8 11.9 12.0 8.6 12.2 8.6 8.6 Northern Ireland 4 4.1 4.4 2.7 4.5 2.7 2.7

9.4.3 Household weighting The level of household population figures available is less detailed than at a respondent level and varies between country. The household weights therefore have been derived using the respondent information and the same approach has been adopted in each country for consistency.

The household weighting will be similar to respondent weighting but will not include the part I weighting. For each respondent of the final two respondent weights a household weight will be derived by taking the final individual weights and dividing them by the part 1 design weight of individuals.

Table 23 below show the effect of the household weights on household size and number of PCs. No population figures are available for households.

Table 23: Household profiles according to weighting scheme used Household Size Unweighted Wt3hhd Dem2 Wt3hhda Dem2a Don't Know 0.1 0.1 0.1 0.1 0.1 1 21.7 22.9 28.8 24.4 29.5 2 31.5 32.2 33.8 33.3 34.2 3 18.5 17.8 15.5 16.9 15.1 4 17.8 17.1 14.1 16.1 13.7 5 6.8 6.4 5.1 5.9 4.9 6 2.8 2.6 2.1 2.4 2 7 0.7 0.6 0.5 0.5 0.4 8 0.1 0.1 0.1 0.1 0.1 10 0.1 0.1 0 0 0 Page 61 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

More than 10 0.1 0.1 0 0.1 0.1 Number of PCs Don't Know 0.2 0.2 0.2 0.2 0.2 None 38.1 42.4 47.7 48.3 52.1 One 42.7 39.7 36.3 35.7 33.1 Two 12.9 12 10.5 10.7 9.7 Three 4 3.7 3.3 3.3 3 Four or more 2.2 2.1 2.1 1.8 1.9 Region North East 4.3 4.3 4.1 4.4 4.1 North West 12.7 12.7 11.6 12.7 11.6 Yorkshire /Humberside 8.1 8.1 8.5 8.1 8.5 East Midlands 5.6 5.7 7 5.7 7 West Midlands 8.6 8.6 8.6 8.6 8.6 East Anglia 7.6 7.5 9 7.5 9 London 10.9 10.7 12.6 10.5 12.7 South East 12.5 12.4 13.9 12.1 13.9 South West 8.6 8.6 8.9 8.6 8.8 Wales 5.5 5.5 4.8 5.5 4.8 Scotland 11.8 11.9 8.5 12.1 8.5 Northern Ireland 4 4 2.5 4.1 2.5

Page 62 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

Appendix A : EPSeM and Other RDD Techniques

Supplier

SSI is a 22 year old company dedicated to the provision of survey samples. SSI has 1200 clients engaged in survey research in the US, Canada and Europe. As well as offering list assisted telephone samples, they have recently developed EPSeM (Equal Probability of Selection Method) sampling for clients requiring a more rigorous sampling approach. SSI supplied RDD sample for each of the four countries where the research was being conducted by BMRB’s fieldforce. Israeli and Bulgarian local agencies provided RDD samples.

Obtaining random telephone samples – an overview of RDD methods

Introduction

For the e-living study two types of RDD sample propagation techniques are going to be used: list-assisted and EPSeM. This appendix provides further detail about each method. There are a number of issues to consider when using RDD and these are discussed below.

List-assisted RDD

Early random number propagation techniques generated a significant proportion of unusable telephone numbers such that the process was not cost-effective. Further developments introduced the use of ‘seed’ numbers that are known to be working lines and these are used as starting points from which to propagate telephone numbers by subtracting or adding 1 (or more) to the last digit, or randomising the last two digits of the seed number. This is the basis for the list-assisted methodology, where the seed numbers used are a sub-sample (referred to by SSI as an element sample). The selection of the ‘seeds’ is critical and should be done in a random manner as using seed numbers automatically excludes some STD or area codes and this means that phone numbers in those exchanges can not be generated thereby excluding residents in these households.

The process still has a slight weakness in that some numbers can never be generated, if an STD or area code is not represented in the element sample. However, overall the ‘seed’ methods of RDD are very effective and are frequently used for a national samples.

SSI's European RDD databases begin with a file of Directory listed telephone numbers for each country for which such data is available. The records for a given country may or may not have complete postal codes as part of the address. The records for a given country may or may not have any ancillary geographic information such as county or county-equivalent information.

As each country is updated, the following procedures are performed in order to ensure that the database is as representative of the telephone household population as possible:

Confirm that the area codes are current and legal. Where available, compare to an official list of area codes. Confirm that the postal codes are complete and accurate. Where available, compare to an official list of postal codes. Page 63 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

Apply geographic codes. The method and level of precision of this coding depends on the availability of a data set that can be used to apply these codes using codes that already exist on the database. Using the lowest level of coding for which population statistics are available, ensure that the seed database is representative.

Database Balancing

In some countries, the listed household data were not representative geographically This may be due to variations in listed rates or in telephone penetration. In order to compensate for these irregularities, SSI balances its seed databases. The level to which the balancing is done depends on the coding levels available on the database and the availability of statistical information for that level of coding. In most countries balancing has been accomplished at the county level or equivalent. Balancing ensures that the seed database is distributed by geography in the same proportion as population or households. For example, if a county represents 10% of the national population, then the database seeds will be selected such that 10% of the database contains a systematic nth selection of eligible seeds from that county.

It must be noted that the balancing of the seed database does not guarantee that an RDD sample generated from this database will be equally balanced since telephone geography does not necessarily map to postal or census geography.

EPSeM RDD samples

This approach is called EPSeM (Equal Probability of Selection Method) and has been developed very recently by Survey Sampling Incorporated (SSI). It has been used successfully on a regional survey in Wales by another research agency recently and BMRB currently uses it on a national survey for the UK Tourist Boards. EPSeM does not use ‘seed’ numbers; it takes as its starting point a database of all working telephone exchanges in the UK and creates telephone numbers from these. An added refinement is software which screens the generated numbers to check that they are working lines, during the sampling process. EPSeM sampling is described in more detail in the next section.

EPSeM sampling for RDD

Essentially, EPSeM provides samples of the residential population, with both listed and unlisted directory phone numbers represented in all eligible exchange codes. There are C13,000 working consumer exchanges or STD codes in the UK EPSeM database. In the USA the equivalent terminology for exchange is ‘block’; the two terms are used interchangeably throughout this report. The EPSeM database has been created from Oftel data covering only blocks allocated for residential service (some blocks are allocated solely for business use and these are excluded). All residential service providers are included. Excluding the leading zero, most UK telephone numbers are 10 digits long; the first 6 of these are the exchange or block; e.g. in the telephone number 01628 670630, the block is 162867. In the number 020 7230 1212 the block is 207230. The remaining four digits identify the unique number and it is these four digits that are randomly generated.

Page 64 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

Exchange codes are randomly selected with equal probability for all eligible codes. For this survey, eligible exchanges would be those within the United Kingdom. Once a six digit exchange is selected, the last 4 digits of the telephone number are randomly generated (0000- 9999). At this point every potential number has the same chance of being generated.

The generated numbers are then subject to a screening process. A certain proportion of random numbers generated will be non-working phone numbers (non-active). All generated numbers are checked to ensure that they are assigned, working telephone numbers. Similarly the screening checks for known business numbers, although it can not identify all numbers with business usage e.g. small traders or lines with multiple residential/business usage. Non-working lines and business numbers are replaced with alternative random numbers in an iterative process. Overall, the screening ensures that the final selected numbers in each STD are in proportion to the number of working residential numbers in that STD, and the sample ultimately represents exchange codes according to size. Therefore a sample of random numbers is systematically produced with equal probability across all active blocks containing one or more listed numbers, which distributes the sample across counties in proportion to their share of total active blocks.

A hypothetical example may help to illustrate the EPSeM sampling process:

Suppose we required a sample of 13,000 telephone numbers and we assumed that, on average, 50% of all random numbers generated were not active telephone numbers. Across the universe of 13,000 working STDs (working blocks) we would generate two potential phone numbers per STD giving 26,000 in total. EG In block 1628 67 we could generate 01628 675432 01628 672687 In block 20 7230 we could generate 020 7230 1212 020 7230 9999 At this point every potential number has the same probability of being generated. However we suspect that 50% of these are not in use. So when the 26,000 numbers are screened only 13,000 (50%) will be found to be active telephone numbers. This will vary across different exchanges however. e.g. If in block 20 7230, all numbers are in use, then both the generated numbers will be accepted. If in block 1628 67 only 50% of numbers are actually in use, then only 1 of the 2 numbers can be expected to be in use. As the validation is repeated across all blocks, the accepted numbers will populate the blocks in about the same ratios as do all working telephone numbers. The method relies on knowledge of the proportion of numbers in each block that are in use and whether a generated number is in use. Some business numbers will also be generated in the random process and the screening stage also identifies these. To preserve the EPSeM sample these numbers would still be selected but flagged as business numbers. Also a small but significant proportion of the British household population has identified itself as not wishing commercial telephone calls – this database which currently contains around 800,000 records is eliminated from the RDD sample.

Page 65 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

Appendix B:

Bulgaria

BLAGOEVGRAD 101-108 BOURGAS 201-209 VELIKO TARNOVO 401-405 VIDIN 501-503 GABROVO 701-703 DOBRICH 801-802 KUSTENDIL 1001- LOVETCH 1101- 1004 1102 PAZARDJIK 1301 PERNIK 1401- 1403 PLOVDIV 1601- RAZGRAD 1701- 1605 1702 SILISTRA 1901- SLIVEN 2001- 1903 2012 SOFIA 2201- SOFIA-REGION 2301- 2217 2310 TARGOVISHTE 2501- HASKOVO 2601- 2606 2606 IAMBOL 2801- PLEVEN 1501- 2803 1507 VARNA 301-309 ROUSE 1801- 1805 VRATCA 603-606 SMOLIAN 2101 KURDJALI 901 STARA ZAGORA 2401- 2406 MONTANA 1201- SHOUMEN 2702- 1205 2704

Germany

STUTTGART, STADTKR. BOEBLINGEN 8115 8111 GOEPPINGEN 8117 LUDWIGSBURG 8118 HEILBRONN, STADTKR. 8121 HEILBRONN, LANDKR. 8125

SCHWAEBISCH HALL 8127 MAIN-TAUBER-KREIS 8128 OSTALBKREIS 8136 BADEN-BADEN, STADTKR. 8211 KARLSRUHE, LANDKR. 8215 RASTATT 8216

MANNHEIM, STADTKR. 8222 NECKAR-ODENWALD-KREIS 8225

PFORZHEIM, STADTKR. 8231 CALW 8235

FREUDENSTADT 8237 BREISGAU-HOCHSCHWARZWALD 8315

ORTENAUKREIS 8317 ROTTWEIL 8325 TUTTLINGEN 8327 KONSTANZ 8335 WALDSHUT 8337 REUTLINGEN 8415 ULM, STADTKREIS 8421 ALB-DONAU-KREIS 8425 BODENSEEKREIS 8435 RAVENSBURG 8436 INGOLSTADT, KRFR.ST. 9161 MUENCHEN, KRFR.ST. 9162

BAD TOELZ- 9173 DACHAU 9174 WOLFRATSHAUSEN EICHSTAETT 9176 ERDING 9177 FUERSTENFELDBRUCK 9179 GARMISCH-PARTENKIRCHEN 9180 Page 66 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

MIESBACH 9182 MUEHLDORF AM INN 9183 NEUBURG- 9185 PFAFFENHOFEN AN DER ILM 9186 SCHROBENHAUSEN STARNBERG 9188 TRAUNSTEIN 9189 LANDSHUT, KRFR.ST. 9261 PASSAU, KRFR.ST. 9262 DEGGENDORF 9271 KELHEIM 9273 PASSAU, LANDKR. 9275 REGEN 9276 STRAUBING-BOGEN 9278 DINGOLFING-LANDAU 9279 REGENSBURG, KRFR.ST. 9362 WEIDEN I. D. OPF, KRFR.ST. 9363

CHAM 9372 NEUSTADT A. D. WALDNAAB 9374 SCHWANDORF 9376 TIRSCHENREUTH 9377 BAYREUTH, KRFR.ST. 9462 COBURG, KRFR.ST. 9463 BAMBERG, LANDKR. 9471 BAYREUTH, LANDKR. 9472 HOF, LANDKR. 9475 KRONACH 9476 LICHTENFELS 9478 WUNSIEDEL IM FICHTELGEBIRGE 9479

ERLANGEN, KRFR.ST. 9562 FUERTH, KRFR.ST. 9563 SCHWABACH, KRFR.ST. 9565 ANSBACH, LANDKR. 9571

FUERTH, LANDKREIS 9573 NUERNBERGER LAND 9574 ROTH 9576 WEISSENBURG-GUNZENHAUSEN 9577

SCHWEINFURT, KRFR.ST. 9662 WUERZBURG, KRFR.ST. 9663

BAD KISSINGEN 9672 RHOEN-GRABFELD 9673 KITZINGEN 9675 MILTENBERG 9676 SCHWEINFURT, LANDKR. 9678 WUERZBURG, LANDKR. 9679

KAUFBEUREN, KRFR.ST. 9762 KEMPTEN (ALLGAEU), KRFR.ST. 9763

AICHACH-FRIEDBERG 9771 AUGSBURG, LANDKR. 9772 GUENZBURG 9774 NEU-ULM 9775 OSTALLGAEU 9777 UNTERALLGAEU 9778 OBERALLGAEU 9780 BERLIN-WEST, STADT 11100 BRANDENBURG/HAVEL, 12051 COTTBUS, KRFR.ST. 12052 KRFR.ST. POTSDAM, KRFR.ST. 12054 BARNIM 12060 ELBE-ELSTER 12062 HAVELLAND 12063 OBERHAVEL 12065 OBERSPREEWALD-LAUSITZ 12066 OSTPRIGNITZ-RUPPIN 12068 POTSDAM-MITTELMARK 12069 SPREE-NEISSE 12071 TELTOW-FLAEMING 12072 BREMEN, KRFR.ST. 4011 BREMERHAVEN, KRFR.ST. 4012 DARMSTADT, KRFR.ST. 6411 FRANKFURT AM MAIN, KRFR.ST. 6412

WIESBADEN, KRFR.ST. 6414 BERGSTRASSE 6431 GROSS-GERAU 6433 HOCHTAUNUSKREIS 6434 MAIN-TAUNUS-KREIS 6436 ODENWALDKREIS 6437 RHEINGAU-TAUNUS-KREIS 6439 WETTERAUKREIS 6440

LAHN-DILL-KREIS 6532 LIMBURG-WEILBURG 6533 VOGELSBERGKREIS 6535 KASSEL, KRFR.ST. 6611 HERSFELD-ROTENBURG 6632 KASSEL, LANDKR. 6633

WALDECK-FRANKENBERG 6635 WERRA-MEISSNER-KREIS 6636

NEUBRANDENBURG, 13002 ROSTOCK, KRFR.ST. 13003 Page 67 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

KRFR.ST. STRALSUND, KRFR.ST. 13005 WISMAR, KRFR.ST. 13006 DEMMIN 13052 GUESTROW 13053 MECKLENBURG-STRELITZ 13055 MUERITZ 13056

NORDWESTMECKLENBUR 13058 OSTVORPOMMERN 13059 G RUEGEN 13061 UECKER-RANDOW 13062 SALZGITTER, KRFR.ST. 3102 WOLFSBURG, KRFR.ST. 3103

GOETTINGEN 3152 GOSLAR 3153 NORTHEIM 3155 OSTERODE AM HARZ 3156 WOLFENBUETTEL 3158 HANNOVER, KRFR.ST. 3201 HAMELN-PYRMONT 3252 HANNOVER, LANDKR. 3253 HOLZMINDEN 3255 NIENBURG (WESER) 3256 CELLE 3351 CUXHAVEN 3352 LUECHOW-DANNENBERG 3354 LUENEBURG, LANDKR. 3355

ROTENBURG (WUEMME) 3357 SOLTAU-FALLINGBOSTEL 3358

UELZEN 3360 VERDEN 3361 EMDEN, KRFR.ST. 3402 OLDENBURG (OLD.), KRFR.ST. 3403 WILHELMSHAVEN, 3405 AMMERLAND 3451 KRFR.ST. CLOPPENBURG 3453 EMSLAND 3454 GRAFSCHAFT BENTHEIM 3456 LEER 3457

OSNABRUECK, LANDKR. 3459 VECHTA 3460

WITTMUND 3462 DUESSELDORF, KRFR.ST. 5111 ESSEN, KRFR.ST. 5113 KREFELD, KRFR.ST. 5114 MUELHEIM AN DER RUHR, 5117 OBERHAUSEN, KRFR.ST. 5119 KRFR.ST. SOLINGEN, KRFR.ST. 5122 WUPPERTAL, KRFR.ST. 5124 METTMANN 5158 NEUSS 5162 WESEL 5170 AACHEN, KRFR.ST. 5313 KOELN, KRFR.ST. 5315 LEVERKUSEN, KRFR.ST. 5316 DUEREN 5358 ERFTKREIS 5362 HEINSBERG 5370 OBERBERGISCHER KREIS 5374 RHEIN-SIEG-KREIS 5382 BOTTROP, KRFR.ST. 5512 MUENSTER, KRFR.ST. 5515 BORKEN 5554 RECKLINGHAUSEN 5562 STEINFURT 5566 BIELEFELD, KRFR.ST. 5711 GUETERSLOH 5754 HOEXTER 5762 LIPPE 5766 PADERBORN 5774 BOCHUM, KRFR.ST. 5911 HAGEN, KRFR.ST. 5914 HAMM, KRFR.ST. 5915 ENNEPE-RUHR-KREIS 5954 HOCHSAUERLANDKREIS 5958 OLPE 5966 SIEGEN-WITTGENSTEIN 5970 UNNA 5978 KOBLENZ, KRFR.ST. 7111 ALTENKIRCHEN 7132 BAD KREUZNACH 7133 (WESTERWALD) COCHEM-ZELL 7135 MAYEN-KOBLENZ 7137 RHEIN-HUNSRUECK-KREIS 7140 RHEIN-LAHN-KREIS 7141

TRIER, KRFR.ST. 7211 BERNKASTEL-WITTLICH 7231 DAUN 7233 TRIER-SAARBURG 7235 KAISERSLAUTERN, 7312 LANDAU I.D. PFALZ, KRFR.ST. 7313 KRFR.ST. Page 68 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

MAINZ, KRFR.ST. 7315 NEUSTADT A.D.WEINSTRASSE, 7316 KRFR.ST SPEYER, KRFR.ST. 7318 WORMS, KRFR.ST. 7319 ALZEY-WORMS 7331 BAD DUERKHEIM 7332 GERMERSHEIM 7334 KAISERSLAUTERN, LANDKR. 7335 SUEDLICHE WEINSTRASSE 7337 LUDWIGSHAFEN, LANDKR. 7338

SUEDWESTPFALZ SAARBRUECKEN, STADTVERBAND 10041

NEUNKIRCHEN 10043 SAARLOUIS 10044 SANKT WENDEL 10046 CHEMNITZ, KRFR.ST. 14161 ZWICKAU, KRFR.ST. 14167 ANNABERG 14171 FREIBERG 14177 VOGTLANDKREIS 14178 MITTWEIDA 14182 STOLLBERG 14188 ZWICKAUER LAND 14193 DRESDEN, KRFR.ST. 14262 HOYERSWERDA, KRFR.ST. 14264 BAUTZEN 14272

NIEDERSCHLESISCHER 14284 RIESA-GROSSENHAIN 14285 OBERLAUSITZKREIS SAECHSISCHE SCHWEIZ 14287 WEISSERITZKREIS 14290

LEIPZIG, KRFR.ST. 14365 DELITZSCH 14374 LEIPZIGER LAND 14379 MULDENTALKREIS 14383 DESSAU, KRFR.ST. 15101 ANHALT-ZERBST 15151 BITTERFELD 15154 KOETHEN 15159 HALLE/SAALE, STADTKR. 15202 BURGENLANDKREIS 15256

MERSEBURG-QUERFURT 15261 SAALKREIS 15265

WEISSENFELS 15268 ASCHERSLEBEN-STASSFURT 15352 HALBERSTADT 15357 JERICHOWER LAND 15358 STENDAL 15363 QUEDLINBURG 15364 WERNIGERODE 15369 ALTMARKKREIS SALZWEDEL 15370 KIEL, KRFR.ST. 1002 LUEBECK, KRFR.ST. 1003 DITHMARSCHEN 1051 HERZOGTUM LAUENBURG 1053 OSTHOLSTEIN 1055 PINNEBERG 1056 RENDSBURG- 1058 SCHLESWIG-FLENSBURG 1059 ECKERNFOERDE STEINBURG 1061 STORMARN 1062 GERA, KRFR.ST. 16052 JENA, KRFR.ST. 16053 WEIMAR, KRFR.ST. 16055 EICHSFELD 16061 UNSTRUT-HAINICH-KREIS 16064 KYFFHAEUSERKREIS 16065

GOTHA 16067 SOEMMERDA 16068 ILM-KREIS 16070 WEIMARER LAND 16071 SAALFELD-RUDOLSTADT 16073 SAALE-HOLZLAND-KREIS 16074

GREIZ 16076 ALTENBURGER LAND 16077 WARTBURGKREIS 16063 ESSLINGEN 8116 HILDESHEIM 3254 REMS-MURR-KREIS 8119 SCHAUMBURG 3257 HOHENLOHEKREIS 8126 HARBURG 3353 HEIDENHEIM 8135 OSTERHOLZ 3356 KARLSRUHE, STADTKR. 8212 STADE 3359

HEIDELBERG, STADTKR. 8221 DELMENHORST, KRFR.ST. 3401

RHEIN-NECKAR-KREIS 8226 OSNABRUECK, KRFR.ST. 3404 Page 69 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

ENZKREIS 8236 AURICH 3452 EMMENDINGEN 8316 FRIESLAND 3455 SCHWARZWALD-BAAR- 8326 OLDENBURG (OLD.), LANDKR. 3458 KREIS LOERRACH 8336 WESERMARSCH 3461 ZOLLERNALBKREIS 8417 DUISBURG, KRFR.ST. 5112 BIBERACH 8426 MOENCHENGLADBACH, KRFR.ST. 5116

SIGMARINGEN 8437 REMSCHEID, KRFR.ST. 5120 BERCHTESGADENER LAND 9172 KLEVE 5154

EBERSBERG 9175 VIERSEN 5366 FREISING 9178 BONN, KRFR.ST. 5314 LANDSBERG AM LECH 9181 AACHEN, LANDKR. 5354 MUENCHEN, LANDKR. 9184 EUSKIRCHEN 5366 ROSENHEIM, LANDKR. 9187 RHEINISCH-BERGISCHER-KREIS 5378

WEILHEIM-SCHONGAU 9190 GELSENKIRCHEN, KRFR.ST. 5513

STRAUBING, KRFR.ST. 9263 COESFELD 5558 LANDSHUT, LANDKR. 9274 WARENDORF 5570 ROTTAL-INN 9277 HERFORD 5758 AMBERG, KRFR.ST. 9361 MINDEN-LUEBBECKE 5770 AMBERG-SULZBACH 9371 DORTMUND, KRFR.ST. 5913 REGENSBURG, LANDKR. 9375 HERNE, KRFR.ST. 5916

BAMBERG, KRFR.ST. 9461 MAERKISCHER KREIS 5962 HOF, KRFR.ST. 9464 SOEST 5974 FORCHHEIM 9474 AHRWEILER 7131 KULMBACH 9477 BIRKENFELD 7134 ANSBACH, KRFR.ST. 9561 NEUWIED 7138 NUERNBERG, KRFR.ST. 9564 WESTERWALDKREIS 7143

ERLANGEN-HOECHSTADT 9572 BITBURG-PRUEM 7232

NEUSTADT A.D. AISCH- 9575 FRANKENTHAL(PFALZ), KRFR.ST 7311 BAD WINDSHEIM ASCHAFFENBURG, 9661 LUDWIGSHAFEN/RHEIN, KRFR.ST 7314 KRFR.ST. ASCHAFFENBURG, 9671 PIRMASENS, KRFR.ST. 7317 LANDKR. HASSBERGE 9674 ZWEIBRUECKEN, KRFR.ST. 7320 MAIN-SPESSART 9677 DONNERSBERGKREIS 7333 AUGSBURG, KRFR.ST. 9761 KUSEL 7336 MEMMINGEN, KRFR.ST. 9764 MAINZ-BINGEN 7339

DILLINGEN AN DER 9773 MERZIG-WADERN 10042 DONAU LINDAU-BODENSEE 9776 SAARPFALZ-KREIS 10045 DONAU-RIES 9779 PLAUEN, KRFR.ST. 14166 BERLIN-OST, STADT 11200 CHEMNITZER LAND 14173 FRANKFURT/ODER, 12053 MITTLERER ERZGEBIRGSKREIS 14181 KRFR.ST. DAHME-SPREEWALD 12061 AUE-SCHWARZENBERG 14191 MAERKISCH-ODERLAND 12064 GOERLITZ, KRFR.ST. 14263

ODER-SPREE 12067 MEISSEN 14280 PRIGNITZ 12070 LOEBAU-ZITTAU 14286 Page 70 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

UCKERMARK 12073 KAMENZ 14292 HAMBURG 2000 DOEBELN 14375 OFFENBACH AM MAIN, 6413 TORGAU-OSCHATZ 14389 KRFR.ST. DARMSTADT-DIEBURG 6132 BERNBURG 15153

MAIN-KINZIG-KREIS 6435 WITTENBERG 15171 OFFENBACH, LANDKR. 6438 MANSFELDER LAND 15260

GIESSEN, LANDKR. 6531 SANGERHAUSEN 15266 MARBURG-BIEDENKOPF 6534 BOERDEKREIS 15355

FULDA 6631 OHREKREIS 15362 SCHWALM-EDER-KREIS 6634 SCHOENEBECK 15367

GREIFSWALD, KRFR.ST. 13001 FLENSBURG, KRFR.ST. 1001

SCHWERIN, KRFR.ST. 13004 NEUMUENSTER, KRFR.ST. 1004 BAD DOBERAN 13051 NORDFRIESLAND 1054 LUDWIGSLUST 13054 PLOEN 1057 NORDVORPOMMERN 13057 SEGEBERG 1060 PARCHIM 13060 ERFURT, KRFR.ST. 16051 BRAUNSCHWEIG, KRFR.ST. 3101 SUHL, KRFR.ST. 16054

GIFHORN 3151 NORDHAUSEN 16062 HELMSTEDT 3154 SCHMALKALDEN-MEININGEN 16066 PEINE 3157 HILDBURGHAUSEN 16069 DIEPHOLZ 3251 SONNEBERG 16072 SAALE-ORLA-KREIS 16075 EISENACH, KRFR.ST.

Israel

JERUSALEM 11 ZEFAT 21 YIZRE’EL 23,25 AKKO 24 HAIFA 31 HADERA 32 PETAH TIQWA 42 RAMLA 43 TEL AVIV 51-53 ASHQELON 61 KINNERET 22 REHOVOT 44 GOLAN 29 BE’ER SHEVA 62 SHARON 41 Italy

TORINO 12011 VERCELLI 12015 VERBANO-CUSIO-OSSOLA 12013 NOVARA 12009

ASTI 12003 ALESSANDRIA 12001 IMPERIA 8003 SAVONA 8007 LA SPEZIA 8005 VARESE 9021 LECCO 9009 SONDRIO 9019 BERGAMO 9001 BRESCIA 9003 LODI 9011 CREMONA 9007 VERONA 20011 VICENZA 20013 VENEZIA 20009 PADOVA 20003 PORDENONE 6003 UDINE 6007 PIACENZA 5011 PARMA 5009 Page 71 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

MODENA 5007 RAVENNA 5013 RIMINI 5017 MASSA-CARRARA 16011 PISTOIA 16015 FIRENZE 16003 LIVORNO 16007 PISA 16013 SIENA 16019 GROSSETO 16005 TERNI 18003 PESARO E URBINO 10007 MACERATA 10005 ASCOLI PICENO 10003 RIETI 7005 ROMA 7007 FROSINONE 7001 L'AQUILA 1003 PESCARA 1005 CHIETI 1001 CAMPOBASSO 11001 CASERTA 4005 NAPOLI 4007 AVELLINO 4001 FOGGIA 13005 BARI 13001 BRINDISI 13003 LECCE 13007 MATERA 2001 COSENZA 3003 CATANZARO 3001 VIBO VALENTIA 3009 TRAPANI 15017 PALERMO 15011 AGRIGENTO 15001 CALTANISSETTA 15003 CATANIA 15005 RAGUSA 15013 BIELLA 12005 NUORO 14003 CUNEO 12007 ANCONA 10001 VALLE D'AOSTA 19001 VITERBO 7009 GENOVA 8001 LATINA 7003 COMO 9005 TERAMO 1007 MILANO 9015 ISERNIA 11003 PAVIA 9017 BENEVENTO 4003 MANTOVA 9013 SALERNO 4009 TREVISO 20007 TARANTO 13009 ROVIGO 20005 POTENZA 2003 GORIZIA 6001 CROTONE 3005 REGGIO NELL' EMILIA 5015 REGGIO DI CALABRIA 3007 FORLI-CESENA 5005 MESSINA 15009 LUCCA 16009 ENNA 15007 PRATO 16017 SIRACUSA 15015 AREZZO 16001 ORISTANO 14005 PERUGIA 18001 FERRARA 5003 TRENTO 17003 SASSARI 14007 BOLOGNA 5001 CAGLIARI 14001 BOZZANO_BOZEN 17001

Norway

AKERSHUS 1 AUST-ADGER 2 FINMARK 4 HEDMARK 5 MORE OG ROMSDAL 7 NORD-TRONDELAG 8 OPPLAND 10 OSLO 11 ROGALAND 13 SOGN OG FJORDANE 14 TELEMARK 16 TROMS 17 VESTFOLD 19 NORDLAND 9 BUSKERUD 3 OSTFOLD OS 12 HORDALAND 6 SOR-TRONDELAG 15 VEST-AGDER 18

United Kingdom

Page 72 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

HARTLEPOOL & 010101 SOUTH TEESIDE 010102 STOCKTON DURHAM CC 010104 NORTHUMBERLAND 010205 SUNDERLAND 010207 WEST CUMBRIA 100108 HALTON & WARRINGTON 100201 CHESHIRE CC 100202

GT MANCHESTER NORTH 100304 BLACKBURN WITH DARWEN 100405

LANCASHIRE CC 100407 EAST MERSEYSIDE 100508 SEFTON 100510 WIRRAL 100511 EAST RIDING OF 020102 NORTH & NORTH EAST 020103 YORKSHIRE LINCOLSHIRE NORTH YORKSHIRE CC 020205 BARNSLEY, DONCASTER & 020306 ROTHERHAM BRADFORD 020408 LEEDS 020409 DERBY 030101 EAST DERBYSHIRE 030102 NOTTINGHAM 030104 NORTH NOTTINGHAMSHIRE 030105 LEICESTER CITY 030207 LEICESTER CC & RUTLAND 030208 LINCOLNSHIRE 030310 HEREFORDSHIRE 090101 WARWICKSHIRE 090103 THE WREKIN 090204 STOKE-ON-TRENT 090206 STAFFORDSHIRE CC 090207 SOLIHULL 090309 COVENTRY 090310 WALSALL & 090312 PETERBOROUGH 040101 WOLVERHAMPTON NORFOLK 040103 SUFFOLK 040104 BEDFORDSHIRE CC 040106 HERTFORDSHIRE 040207 THURROCK 040309 ESSEX CC 040310 INNER LONDON - EAST 060102 OUTER LONDON - E & NE 060203

OUTER LONDON - W & NW 060205 BERKSHIRE 050101

BUCKINGHAMSHIRE CC 050103 OXFORDSHIRE 050104

EAST SUSSEX CC 050206 SURREY 050207 PORTSMOUTH 050309 SOUTHAMPTON 050310 ISLE OF WIGHT 050312 MEDWAY TOWNS 050413 CITY OF BRISTOL 070101 N & NE SOMERSET, SOUTH 070102 GLOUCESTERSHIRE SWINDON 070104 WILTSHIRE CC 070105 DORSET CC 070207 SOMERSET 070208 PLYMOUTH 070410 TORBAY 070411 ISLE OF ANGLESEY 080101 GWYNEDD 080102 SOUTH WEST WALES 080104 CENTRAL VALLEYS 080105 BRIDGEND & NEATH PORT 080107 SWANSEA 080108 TALBOT CARDIFF & VALE OF 080210 FLINTSHIRE & WREXHAM 080211 GLAMORGAN ABERDEEN CITY, 120101 ANGUS & DUNDEE CITY 120202 ABERDEENSHIRE & NE MORAY EAST LOTHIAN & 120204 THE SCOTTISH BORDERS 120205 MIDLOTHIAN FALKIRK 120207 PERTH & KINROSS AND STIRLING 120208

E & W DUNBARTONSHIRE, 120310 DUMFRIES & GALLOWAY 120311 HELENSBURGH & LOMOND

GLASGOW CITY 120313 INVERCLYDE, EAST 120314 Page 73 of 74

2713457/RL/SMD 15/05/02 E-Living Wave 1 Technical Report

RENFREWSHIRE & RENFREWSHIRE

SOUTH AYRSHIRE 120316 SOUTH LANARKSHIRE 120317 INVERNESS & NAIRN AND 120419 LOCHABER, SKYE & LOCHALSH 120420 MORAY, BADENOCH & AND ARGYLL & THE ISLANDS STRATHSPEY ORKNEY ISLANDS 120422 SHETLAND ISLANDS 120423 OUTER BELFAST 110102 EAST OF NORTHERN IRELAND 110103 WEST & SOUTH OF 110105 NORTHERN IRELAND DARLINGTON 010103 MILTON KEYNES 050102 TYNESIDE 010206 BRIGHTON & HOVE 050205 EAST CUMBRIA 100109 WEST SUSSEX 050208 GT MANCHESTER SOUTH 100303 HAMPSHIRE CC 050311

BLACKPOOL 100406 KENT CC 050414 LIVERPOOL 100509 GLOUCESTERSHIRE 070103 CITY OF KINGSTON UPON 020101 BOURNEMOUTH & POOLE 070206 HULL YORK 020204 CORNWALL & ISLES OF SCILLY 070309

SHEFFIELD 020307 DEVON CC 070412 CALDERDALE, KIRKLEES & 020410 CONWY & DENBIGHSHIRE 080103 WAKEFIELD SOUTH & WEST 030103 GWENT VALLEYS 080106 DERBYSHIRE SOUTH 030106 MONMOUTHSHIRE & NEWPORT 080209 NOTTINGHAMSHIRE NORTHAMPTONSHIRE 030209 POWYS 080212 WORCESTERSHIRE 090102 CLACKMANNANSHIRE & FIFE 120203 SHROPSHIRE CC 090205 EDINBURGH, CITY OF 120206 BIRMINGHAM 090308 WEST LOTHIAN 120209 DUDLEY & SANDWELL 090311 E AYRSHIRE & N AYRSHIRE 120312 MAINLAND CAMBRIDGESHIRE CC 040102 NORTH LANARKSHIRE 120315 LUTON 040205 CAITHNESS & SUTHERLAND AND 120418 ROSS & CROMARTY SOUTHEND-ON-SEA 040308 COMHAIRLE NAN EILAN (WESTERN 120421 ISLES) INNER LONDON - WEST 060101 BELFAST 110101

OUTER LONDON - SOUTH 060204 NORTH OF NORTHERN IRELAND 110104

Page 74 of 74

2713457/RL/SMD 15/05/02 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

E Living, Wave 1, CATI Questionnaire Administered by BMRB to Respondents in the UK, Germany, Italy and Norway

The English version of the E-Living CATI script was the first to be developed by BMRB, and German, Norwegian and Italian translations were based on it. The script was first developed in Quanquest, from which the information here was produced, and centrally administered from BMRB’s telephone unit in London using the Quancept CATI program, from which more detailed information in English, German, Norwegian and Italian has been produced

e-Living: Life in a Digital Europe, an EU Fifth Framework Project [IST-2000-25409] www.eurescom.de/e-living/index.htm PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

1 Introduction

Data collected by this script is returned to the ISER in 3 SPSS system files containing full interview, partial interview and refusal data, merged with data collected by paper in Bulgaria and CATI in Israel; separate free- text files contain open-ended/free text data. All three files will include the respondent serial number, country of interview and a summary interview outcome.

The refusals file contains only that data applicable to refusals, i.e. that collected in response to questions QREF1 through QWHYNOT, as well as coded answers to the question QWHYQT (see note, below)

The full and partial interviews file contain data from all other questions and are identical in content except for QWHYQT, which is applicable only to the partial interviews.

Quanquest metadata has been supplemented here by additional information designed to aid interpretation of the script and of data output and is formatted like these notes to distinguish it from original metadata. To aid your interpretation of the results, you should also note that:

Quanquest assigns the default label of “None of these” to the ‘X’ special code, while Quancept topline frequencies assign the special ‘X’ code the default value of “No answer”

On export to SPSS, “special” non-multi punch values were recoded as follows:

(-99) Inapplicable (Skipped) – set as MISSING in SPSS

(-98) Refused

(-97) Don't Know

(-96) Other

(-95) None

On export to SPSS, multi-punch variables were “exploded” into a series of binary, single punch variables with values as follows:

(-99) Inapplicable (Skipped) – set as MISSING in SPSS

(0) Not mentioned

(1) Mentioned

CATI interviewers may use the Quancept system command “quit” at any point in an interview to abnormally terminate it, although some interviewers mistakenly used it in the case of refusals (see below); data taken from the responses to QWHYQT can therefore be found on both the refusals and partial interviews files. The E-Living script was programmed to jump to the question QWHYQT, in case of a “quit” command, in order to collect data about the reasons for quitting prior to closing down the interview. Non-null responses to QWHYQT indicate that there will be skipped questions in the interview that cannot be accounted for by the routing logic.

www.eurescom.de/e-living/index.htm Page 2 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

Eliving Good morning/afternoon/evening. My name is _____ calling from BMRB International, an independent research company in London. We are carrying out an important research study which is funded by the European Commission. The survey is about people's quality of life and how communications fit in with their needs. It is part of a major project that is taking place in different European countries.

Would you mind answering some questions for this survey?

In order to be sure that our survey represents the population as a whole, we need to select someone at random, so may I ask who in your household, aged 16 or over, had the most recent birthday?

NB: IN MOST CASES A REFUSAL WILL BE A SOFT REFUSAL!

QUANCEPT ITEM:

DQCOUNT DUMMY TO SET COUNTRY FOR INTERVIEW

UK 1 (478) Italy 2 Germany 3 Norway 4 Bulgaria 5 Israel 6

The following, undocumented variable is found in all SPSS files. The question was asked in Israel only; data for all other countries will be set to inapplicable: Variable Valid Values Name Description ILANG Language of interview (1) Hebrew (2) Russian (3) Arabic The following, undocumented variables are found in the full and partial interview files. The questions were asked in Israel only; data for all other countries will be set to inapplicable: Variable Valid Values Name Description RELIG Religion (1) Jewish – Secular (2) Jewish – Traditional (3) Jewish - Religious – national (4) *Jewish - Religious – orthodoxy (5) **Jewish - Religious – other (6) Moslem (7) Christian (8) Druse SECTOR Sector (1) Israeli (2) Russian Immigrants (3) Arabs www.eurescom.de/e-living/index.htm Page 3 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

Variable Valid Values Name Description ORIGIN Origin (1) Born Asia/Africa (2) Born Europe/America (3) 2nd Generation Israeli (4) 1st Generation Israeli – Asian/African descent (5) *1st Generation Israeli – European/American descent IMYR Immigration year (1) 1988 and before (2) 1989 (3) 1990 (4) 1991 (5) 1992 (6) 1993 (7) 1994 (8) 1995-1997 (9) 1998-1999 (10) 2000-2001

QUANCEPT ITEM:

QINT CODE ONE ONLY

Proceed with interview 1 (508) Hard Refusal 2

IF Qint = Hard Refusal THEN ASK: QREF1

QREF1 Just before I go then, it is very helpful for us to understand a little bit about the people who are not willing to answer our survey. So please could you just tell me the number of people in your household and your age?

Yes 1 (509) No 2

www.eurescom.de/e-living/index.htm Page 4 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

IF NOT QREF1 = No THEN ASK: QREF2, QREF3, QREF4

QREF2 How many people are there in your household, including yourself?

1 1 (510) 2 2 3 3 4 4 5 5 6 6 7 7 8 8 9 9 10 0 (511) More than 10 1 Don't Know Y (510) Refused Z

QREF3 And what was your age at your last birthday?

(512 - 514) Numeric Range Don't Know Y (512) Refused Z Permitted Range 1 TO 110 (Numeric Range)

QREF4 And finally, how many computers, if any, do you have in your home?

(515 - 516) Numeric Range Don't Know Y (515) Refused Z Permitted Range 0 TO 50 (Numeric Range)

End of Filter iREF2

Thank you very much for your co-operation and I can assure you that all your answers will be treated with the strictest of confidence. Thank you, Goodbye.

www.eurescom.de/e-living/index.htm Page 5 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QREF5 CODE SEX OF RESPONDENT

Male 1 (517) Female 2 Don't Know Y

QWHYNOT RECORD REASON FOR WHY RESPONDENT REFUSED INTERVIEW

(518 - 521) Don't Know Y (518)

End of Filter iREF

ZREF5 IF QREF5 = Male OR QREF5 = Female OR QREF5 = Don't Know OR QREF1 = No - Termination with data (Quit)

Some interviewers used the “quit” command, and QWHYQT data may therefore also have been collected from refusers (see above).

tnick1 IF QREF1 = Yes - Termination with data (Quit)

See ZREF5

QOLD So firstly then could I ask you how old you were on that birthday?

(522 - 524) Numeric Range Don't Know Y (522) Permitted Range 16 TO 110 (Numeric Range)

www.eurescom.de/e-living/index.htm Page 6 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QHHOLD And could you tell me the number of people living in your household, including yourself?

1 1 (525) 2 2 3 3 4 4 5 5 6 6 7 7 8 8 9 9 10 0 (526) More than 10 1 Don't Know Y (525)

QSEX2 CODE SEX OF RESPONDENT (ASK IF NECESSARY)

Male 1 (527) Female 2 Don't Know Y

We are interested in what people do with their leisure time.

QLEI How often would you say you &olei&?

Most days 1 (528) 2-3 times a week 2 About once a week 3 About once a fortnight 4 About once a month 5 Several times a year 6 Less often 7 Never 8 Don't Know Y

This question is repeated for the following loop values:

- play sport, keep fit or go walking - go to the cinema, a concert, theatre or watch live sport - have a meal in a restaurant or cafe, or go for a drink to a bar or club - attend activity groups such as evening classes - read newspapers, magazines or books - meet with friends

A total of 6 iterations occupying columns (528) to (533)

www.eurescom.de/e-living/index.htm Page 7 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QACT Are you a member of any of the following types of organisation? CODE ALL THAT APPLY

Any social or sport club (including Gym) 1 (534) A residents, school or other local group 2 A trade union 3 An environmental or animal welfare organisation 4 Any other political or campaigning organisation 5 Don't Know Y None of these X

QUANCEPT ITEM:

QTVHRS On average how much time do you spend a day watching television? ENTER ANSWER IN HOURS AND MINUTES. ENTER WHOLE NUMBER OF HOURS ON THIS SCREEN AND MINUTES ON NEXT SCREEN

(541 - 542) Numeric Range Don't Know Y (541) Permitted Range 0 TO 24 (Numeric Range)

QTVMINS NOW ENTER THE REMAINING NUMBER OF MINUTES

(543 - 544) Numeric Range Don't Know Y (543) Permitted Range 0 TO 59 (Numeric Range)

www.eurescom.de/e-living/index.htm Page 8 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QMOBHRS On average how much time do you spend a day on the telephone at home, not including mobile phones? ENTER ANSWER IN HOURS AND MINUTES. ENTER WHOLE NUMBER OF HOURS ON THIS SCREEN AND MINUTES ON NEXT SCREEN

(545 - 546) Numeric Range Don't Know Y (545) Permitted Range 0 TO 20 (Numeric Range)

QMOBMIN NOW ENTER THE REMAINING NUMBER OF MINUTES

(547 - 548) Numeric Range Don't Know Y (547) Permitted Range 0 TO 59 (Numeric Range)

QPCS How many computers do you have in your home now, including any provided by your work? INCLUDE LAP-TOPS AND GAMES CONSOLES USED FOR INTERNET ACCESS EXCLUDE GAMES CONSOLES USED SOLELY FOR GAMES

One 1 (549) Two 2 Three 3 Four or more 4 None 5 Don't Know Y

IF QPCS = None THEN ASK: QGETPC

QGETPC How likely is it that you or your household will get a computer in your home in the next 12 months? Is it .... READ OUT AND CODE ONE ONLY

Very likely 1 (550) Likely 2 Neither likely nor unlikely 3 Unlikely 4 Very unlikely 5 Don't Know Y

End of Filter igetPC

www.eurescom.de/e-living/index.htm Page 9 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

IF QGETPC = Very likely OR QGETPC = Likely THEN ASK: QWHYPC

QWHYPC For which of the following purposes are you or your household thinking about getting a computer? READ OUT EACH AND CODE ALL THAT APPLY

Paid work 1 (551) Voluntary and unpaid work 2 Educational purposes such as private study or school or college work 3 Playing games 4 Personal correspondence or other word processing 5 Household or financial accounts 6 Hobbies 7 For e-mail 8 Browsing or surfing the Internet 9 Don't Know Y Other 0

Other specify... (552 - 555)

End of Filter iyespc

IF QGETPC = Neither likely nor unlikely OR QGETPC = Unlikely OR QGETPC = Very unlikely OR QGETPC = Don't Know THEN ASK: QNOTPC

QUANCEPT ITEM:

www.eurescom.de/e-living/index.htm Page 10 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QNOTPC What is the main reason you do not think you will get a computer in the next 12 months? DO NOT READ OUT - CODE ONE ONLY

No need or use for it 1 (566) Too expensive 2 Don't know how to use one 3 Don't like computers 4 Don't like computer games 5 The children would never stop playing games on it 6 Had one but no longer has it 7 Can bring one home from work if necessary 8 Already use one enough at work 9 Other 0 (567) Don't Know Y (566)

QNOTPC contains the undocumented value, (11) "I don't have enough finances", which was added to the Bulgarian questionnaire and will be valid only for Bulgarian cases End of Filter inoPC

QOLDPC How many computers are no longer in use in your household or have been thrown away in the last five years?

One 1 (568) Two 2 Three 3 Four 4 Five 5 More than five 6 None 7 Don't Know Y

The full and partial interviews file contains the following undocumented variable: Variable Comment Name Description XTTIME2 Time in seconds elapsed between start Not collected and set to inapplicable for Israel and of interview and current position Bulgaria

QUANCEPT ITEM:

www.eurescom.de/e-living/index.htm Page 11 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QUANCEPT ITEM:

IF QPCS = One THEN ASK: QWHENPC, QINTPC

QUANCEPT ITEM:

QWHENPC In which year did your household acquire your computer? ENTER YEAR (4 digits)

(569 - 572) Numeric Range Don't Know Y (569) Permitted Range 1970 TO 2001 (Numeric Range1)

QINTPC Do you have a connection to the internet with this computer?

Yes 1 (573) No 2 Don't Know Y

End of Filter ionepc

IF QPCS = Two OR QPCS = Three OR QPCS = Four or more THEN ASK: QNEWPC, QNETPC

www.eurescom.de/e-living/index.htm Page 12 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QNEWPC In which year did you acquire the newest of your computers? ENTER YEAR (4 digits)

(574 - 577) Numeric Range Don't Know Y (574) Permitted Range 1970 TO 2001 (Numeric Range)

QNETPC How many of your computers have a connection to the internet?

One 1 (578) Two or more 2 None 3 Don't Know Y

End of Filter imorepc

IF QNETPC = One OR QNETPC = Two or more OR QINTPC = Yes

IF QNETPC = One OR QINTPC = Yes THEN ASK: QISDN ELSE ASK: QISDN2

QISDN Is this connection through a standard phone line or a high-speed line such as ISDN or broadband, like cable?

Normal phone line 1 (579) High speed line/broadband 2 Don't Know Y

QISDN2 Are these connections through a standard phone line or a high-speed line such as ISDN or broadband, like cable?

Normal phone line 1 (580) High speed line/broadband 2 Both normal phone line and high speed line 3 Don't Know Y

www.eurescom.de/e-living/index.htm Page 13 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

End of Filter iisdn

End of Filter inetpc

IF QINTPC = No OR QINTPC = Don't Know OR QNETPC = None OR QNETPC = Don't Know THEN ASK: QLIKNET

QLIKNET How likely is it that you will get an internet connection in the next 12 months? Is it READ OUT AND CODE ONE ONLY

Very likely 1 (608) Likely 2 Neither likely nor unlikely 3 Unlikely 4 Very unlikely 5 Don't Know Y

End of Filter inonet

IF QLIKNET = Very likely OR QLIKNET = Likely THEN ASK: QWHYNET

QWHYNET Which of the following reasons would be important for getting an internet connection? CODE ALL THAT APPLY

To work at home 1 (609) To exchange emails with people 2 For games 3 For home shopping or banking 4 For education or information purposes 5 To surf the web 6 Don't Know Y None of these X

End of Filter iliknet

IF TempVa1 > 1 Note: TEMPVA1=NUMB(QWHYNET), i.e. number of responses @ QWHYNET. THEN ASK: QONENET

www.eurescom.de/e-living/index.htm Page 14 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QUANCEPT ITEM:

QONENET And which would be the single most important reason for getting an internet connection? CODE ONE ONLY

To work at home 1 (617) To exchange emails with people 2 For games 3 For home shopping or banking 4 For education or information purposes 5 To surf the web 6 Don't Know Y

End of Filter iment

QUSEPC Do you personally ever use a computer READ OUT AND CODE ALL THAT APPLY

At home 1 (618) At work or for your job 2 Anywhere else 3 Never uses a computer (DO NOT READ OUT) 4

The full and partial interviews file contains the following undocumented variable: Variable Comment Name Description XTTIME3 Time in seconds elapsed between Not collected and set to inapplicable for Israel and XTTIME2 and current position Bulgaria

QUANCEPT ITEM:

www.eurescom.de/e-living/index.htm Page 15 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

IF QUSEPC = At home THEN ASK: QFREQPC

QUANCEPT ITEM:

QFREQPC Would you say you use a computer at home... READ OUT

Most days 1 (622) 2-3 times a week 2 About once a week 3 About once a fortnight 4 About once a month 5 Several times a year 6 Less often 7 Never 8 Don't Know Y

IF QFREQPC = Most days OR QFREQPC = 2-3 times a week OR QFREQPC = About once a week OR QFREQPC = About once a fortnight OR QFREQPC = About once a month THEN ASK: QPCHRS, QPCMINS

QPCHRS When you use the computer at home, on average how much time do you spend a day on this? ENTER ANSWER IN HOURS AND MINUTES. ENTER WHOLE NUMBER OF HOURS ON THIS SCREEN AND MINUTES ON NEXT SCREEN.

(623 - 624) Numeric Range Don't Know Y (623) Permitted Range 0 TO 24 (Numeric Range)

QPCMINS NOW ENTER THE REMAINING NUMBER OF MINUTES

(625 - 626) Numeric Range Don't Know Y (625) Permitted Range 0 TO 59 (Numeric Range)

End of Filter ifreqpc www.eurescom.de/e-living/index.htm Page 16 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QACTSPC Which of the following types of activities do you use this computer for at home? READ OUT AND CODE ALL THAT APPLY

Paid work 1 (627) Voluntary and unpaid work 2 Educational purposes such as private study or school or college work 3 Playing games 4 Personal correspondence or other word processing 5 Household or financial accounts 6 Hobbies 7 For email 8 Browsing or surfing the Internet 9 Don't Know Y None of these X

End of Filter ihomepc

The full and partial interviews file contains the following undocumented variable: Variable Comment Name Description XTTIME4 Time in seconds elapsed between Not collected and set to inapplicable for Israel and XTTIME3 and current position Bulgaria

QUANCEPT ITEM:

www.eurescom.de/e-living/index.htm Page 17 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QTVS How many TV sets do you have in your household? WRITE IN NUMBER

(638) Numeric Range Don't Know Y (638) Permitted Range 0 TO 9 (Numeric Range)

IF QTVS > 0 THEN ASK: QSATTV, QDIGTV

QSATTV Do you have READ OUT AND CODE ALL THAT APPLY.

Satellite TV service 1 (639) Cable TV service 2 Neither (DO NOT READ OUT)^s 3 Don't Know Y

QUANCEPT ITEM:

QDIGTV Do you have a digital television or a digital television service which allows access to the internet or email even if you don't use this facility? INCLUDE SET-TOP BOXES

Yes 1 (643) No 2 Never heard of digital TV service 3 Don't Know Y

IF QDIGTV = No THEN ASK: QLIKDIG

www.eurescom.de/e-living/index.htm Page 18 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QLIKDIG How likely is it that you will get a digital television service in the next 12 months? Is it READ OUT AND CODE ONE ONLY

Very likely 1 (644) Likely 2 Neither likely nor unlikely 3 Unlikely 4 Very unlikely 5 Don't Know Y

End of Filter idig

End of Filter itvs

QHAPPS Do you have any of the following in your household? READ OUT EACH AND CODE ALL THAT APPLY

Washing machine 1 (645) Dish washer 2 Microwave oven 3 Compact disc player/music system 4 Video camera 5 Video player/recorder 6 Digital camera 7 DVD player or drive 8 Don't Know Y None of these X

QUANCEPT ITEM:

QCARSHow many cars, vans or motorbikes does your household have or use, including company cars?

One 1 (655) Two 2 Three 3 Four or more 4 None 5 Don't Know Y

www.eurescom.de/e-living/index.htm Page 19 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QROOMS How many rooms are there in your accommodation, including bedrooms but excluding kitchens, bathrooms? @@ENTER NUMBER

(656 - 657) Numeric Range Don't Know Y (656) Permitted Range 0 TO 20 (Numeric Range)

I'd now like to ask some questions about the importance of communications in your own everyday life.

QFRMOB How often do you speak to friends or relatives by telephone, including both fixed lines and using a mobile?

Most days 1 (658) 2-3 times a week 2 About once a week 3 About once a fortnight 4 About once a month 5 Several times a year 6 Less often 7 Never 8 Don't Know Y

IF NOT ( QACTSPC = For email AND QACTSPC = Browsing or surfing the Internet ) THEN ASK: QUSENET

QUSENET Whether or not you have a computer at home, do you ever... READ OUT AND CODE ALL THAT APPLY

Use the internet 1 (659) Use any kind of email system 2 Neither used^s 3 Don't Know Y

End of Filter iACTS

The full and partial interviews file contains the following undocumented variable: Variable Comment www.eurescom.de/e-living/index.htm Page 20 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

Name Description XTTIME5 Time in seconds elapsed between Not collected and set to inapplicable for Israel and XTTIME4 and current position Bulgaria

QUANCEPT ITEM:

IF QUSENET = Use the internet OR QACTSPC = Browsing or surfing the Internet OR QACTSPC = For email THEN ASK: DQWERNT, QWERNET, DQWER2

DQWERNT DUMMY TO SET QWERNET RESPONSES

at home 1 (3142) at a friend's or relative's house 2 at work or college 3 at a public library or cyber cafe 4 on mobile phone (WAP) 5 Somewhere else 6 Don't Know Y None of these X DQWERNT will be set to “at home” if the respondent says that s/he DOES USE his/her home computer for browsing or surfing the internet @ QACTSPC OR if s/he says that s/he DOES NOT HAVE a connexion to the internet via any of his/her home computers at QINTPC or QNETPC. If set to “at home”, the “at home” option at QWERNET does not display.

QUANCEPT ITEM:

www.eurescom.de/e-living/index.htm Page 21 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QWERNET Do you ever use the internet in any of the following places? READ OUT AND CODE ALL THAT APPLY

at home 1 (664) at a friend's or relative's house 2 at work or college 3 at a public library or cyber cafe 4 on mobile phone (WAP) 5 Somewhere else 6 Don't Know Y None of these X

The “at home” option does not display home internet usage has been previously implied or if the respondent has no home internet connexions as determined at DQWERNT.

DQWER2 DUMMY TO COMBINE DQ and MAIN Q

at home 1 (665) at a friend's or relative's house 2 at work or college 3 at a public library or cyber cafe 4 on mobile phone (WAP) 5 Somewhere else 6 Don't Know Y None of these X This variable is designed to conditionally combine the values of DQWERNT and QWERNET in order to produce a measure which includes the implied home internet usage which might not be asked at QWERNET given the values of DQWERNT. ISER cannot confirm that the the algorithm used to implement this variable was correct and it should not therefore be used.

QUANCEPT ITEM:

IF TempVa2 > 1 Note: TEMPVA2=NUMB(DQWER2), i.e. number of responses @ DQWER2. THEN ASK: DQMOSNE, QMOSNET

www.eurescom.de/e-living/index.htm Page 22 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

DQMOSNE DUMMY TO SET RESPONSES AT QMOSNET

at home 1 (666) at a friend's or relative's house 2 at work or college 3 at a public library or cyber cafe 4 on mobile phone (WAP) 5 Don't Know Y Other 0

Other specify... (667 - 670)

DQMOSNE is set to “at home” if the respondent has previously said @ QINTPC or QNETPC that s/he has no connexions to the internet via any of his/her home computer and, in which case, the “at home” option is not displayed @ QMOSNET

QUANCEPT ITEM:

QMOSNET And which one place do you use the internet the most? CODE ONE ONLY

at home 1 (677) at a friend's or relative's house 2 at work or college 3 at a public library or cyber cafe 4 on mobile phone (WAP) 5 Don't Know Y

The “at home” option is not displayed if the respondent has no home internet connexions as determined @ DQMOSNE.

End of Filter iwer

www.eurescom.de/e-living/index.htm Page 23 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QNETYRS For how many years have you been using the internet (including email)?

Less than 1 year 1 (678) 1 year 2 2 years 3 3 years 4 4 years 5 5 years 6 6 years 7 7 years 8 8 years 9 9 years 0 (679) 10 years or more 1 Don't Know Y (678)

End of Filter inet

IF QUSENET = Use any kind of email system OR QACTSPC = For email THEN ASK: DQMAIL, QMAIL, QFREEM

DQMAIL DUMMY TO SET RESPONSES FOR QMAIL

On a home PC 1 (680) At home as part of a digital TV service 2 At a friend's or relative's house 3 At work or college 4 Using a cyber cafe or similar facility 5 On mobile phone (WAP) 6 Don't Know Y

DQMAIL is set to “on a home PC” if the respondent has previously said that she uses email from home (@ QACTSPC) or does not have a home internet connexion (@ QINTPC or QNETPC), in which case the the “on a home PC” option @ QMAIL is not displayed.

QUANCEPT ITEM:

www.eurescom.de/e-living/index.htm Page 24 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QMAIL Do you use EMAIL in any of the following places?

READ OUT AND CODE ALL THAT APPLY

On a home PC 1 (708) At home as part of a digital TV service 2 At a friend's or relative's house 3 At work or college 4 Using a cyber cafe or similar facility 5 On mobile phone (WAP) 6 Don't Know Y None of these X Other 0

Other specify... (709 - 712)

The “on a home PC” option is not displayed if the respondent has already said that s/he uses email from home or does not have a home internet connexion as determined at DQMAIL.

QUANCEPT ITEM:

QFREEM How often do you communicate with friends or relatives by email?

Most days 1 (720) 2-3 times a week 2 About once a week 3 About once a fortnight 4 About once a month 5 Several times a year 6 Less often 7 Never 8 Don't Know Y

IF QMAIL = On a home PC OR QMAIL = At home as part of a digital TV service OR QACTSPC = For email THEN ASK: QFREQEM

www.eurescom.de/e-living/index.htm Page 25 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QFREQEM How many email messages do you usually send from home? Is it READ OUT AND CODE ONE ONLY

More than 5 a day 1 (721) Between 1 and 5 a day 2 Between 1 and 5 a week 3 Less than 1 a week 4 Don't Know Y

End of Filter ihome

QLETTER Does using email mean you now... READ OUT AND CODE ALL THAT APPLY

Write fewer letters than before 1 (722) Use the phone less than before 2 Neither of these (DO NOT READ OUT) 3 Don't Know Y

End of Filter iemail

The full and partial interviews file contains the following undocumented variable: Variable Comment Name Description XTTIME6 Time in seconds elapsed between Not collected and set to inapplicable for Israel and XTTIME5 and current position Bulgaria

QUANCEPT ITEM:

IF DQWER2 = at home OR QACTSPC = Browsing or surfing the Internet THEN ASK: QNETHOM, QNETHRS, QNETMIN, DQHOWNT, QHOWNET, QIMPUSE, QONLINE

www.eurescom.de/e-living/index.htm Page 26 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QUANCEPT ITEM:

QNETHOM How often do you use the internet at home?

Most days 1 (726) 2-3 times a week 2 About once a week 3 About once a fortnight 4 About once a month 5 Several times a year 6 Less often 7 Don't Know Y

QNETHRS When you use the internet at home, on average how long per day do you spend using it? ENTER ANSWER IN HOURS AND MINUTES. ENTER WHOLE NUMBER OF HOURS ON HIS SCREEN AND MINUTES ON NEXT SCREEN

(727 - 728) Numeric Range Don't Know Y (727) Permitted Range 0 TO 24 (Numeric Range)

QNETMIN NOW ENTER THE REMAINING NUMBER OF MINUTES

(729 - 730) Numeric Range Don't Know Y (729) Permitted Range 0 TO 59 (Numeric Range)

DQHOWNT DUMMY TO SET QHOWNET RESPONSES

A home computer 1 (731) A digital TV service 2 A mobile phone 3 A dedicated internet terminal 4 Don't Know Y Other 0

Other specify... (732 - 735)

www.eurescom.de/e-living/index.htm Page 27 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

DQHOWNT is set to “a home computer” if the respondent has already said that s/he uses a home computer for “Browsing or surfing the Internet” @ QACTSPC, in which case the “a home computer” option is not displayed @ QHOWNET.

QUANCEPT ITEM:

QHOWNET Do you use the INTERNET at home on any of the following? READ OUT AND CODE ALL THAT APPLY

A home computer 1 (736) A digital TV service 2 A mobile phone 3 A dedicated internet terminal 4 Don't Know Y None of these X Other 0

Other specify... (737 - 740)

The “a home computer” option is not displayed if it home internet usage can be implied as determined @ DQHOWNT.

During coding, “modem” was added to the precode frame from the “other specify …” responses and will show up on the SPSS output files

QUANCEPT ITEM:

QIMPUSE Which one of these is the MOST important use of the internet at home to you?

READ OUT AND CODE ONE ONLY

To work at home 1 (746) To exchange emails with people 2 For games 3 For home shopping or banking 4 For educational or informational purposes 5 To surf the web 6 Don't Know Y

www.eurescom.de/e-living/index.htm Page 28 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QONLINE In the last THREE MONTHS have you done any of the following on-line.

READ OUT AND CODE ALL THAT APPLY

Shopped on-line 1 (747) Used banking services 2 Used library or similar services 3 Used travel or holiday information services 4 Used educational services 5 Obtained medical assistance 6 Obtained information about the environment 7 Downloaded music 8 Applied for a job or got job information 9 Anything else that we have not already talked about (TYPE IN)^o 0 (748) Not used in last three months (DON'T READ OUT) 1 Don't Know Y (747)

IF QONLINE = Downloaded music THEN ASK: QMUSIC

QUANCEPT ITEM:

QMUSIC When you have downloaded music from the Internet in the last three months, how many discs or tapes, if any, would you say this has replaced in that period that you might otherwise have bought?

(759 - 760) Numeric Range Don't Know Y (759) Permitted Range 0 TO 99 (Numeric Range)

End of Filter imusic

www.eurescom.de/e-living/index.htm Page 29 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

IF QONLINE = Shopped on-line THEN ASK: QBUYS

QBUYSIn the last three months, have you shopped on-line and bought any... READ OUT AND CODE ALL THAT APPLY

Books 1 (761) Videos or DVDs 2 CDs or tapes 3 Computer hardware or software 4 Clothing 5 Groceries 6 Travel tickets 7 Tickets for events (e.g. cinema, theatre, concerts, sport) 8 None of these (DON'T READ OUT) 9 Don't Know Y Other 0

Other specify... (762 - 765)

IF NOT ( QBUYS = None of these (DON'T READ OUT) OR QBUYS = Don't Know ) THEN ASK: QSPENT

QUANCEPT ITEM:

QSPENT About how much in total have you spent on goods or services on-line in the last 3 months? @@ENTER ANSWER TO THE NEAREST POUND

(808 - 815) Numeric Range Don't Know Y (808) Permitted Range 0 TO 90000000 (Numeric Range)

End of Filter ibuy

End of Filter ishop

www.eurescom.de/e-living/index.htm Page 30 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

End of Filter inethom

The full and partial interviews file contains the following undocumented variable: Variable Comment Name Description XTTIME7 Time in seconds elapsed between Not collected and set to inapplicable for Israel and XTTIME6 and current position Bulgaria

QUANCEPT ITEM:

IF NOT QUSEPC = Never uses a computer (DO NOT READ OUT) THEN ASK: QKNOWPC

We are trying to find out how comfortable people feel with computers. The following questions are about this, but they are not in any way a test.

QUANCEPT ITEM:

www.eurescom.de/e-living/index.htm Page 31 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QKNOWPC Do you know how to... READ OUT AND CODE ALL THAT APPLY

download a file from the web to a computer 1 (816) construct a web page 2 send a file by email 3 cut and paste 4 reboot a computer 5 copy a file to a floppy disc 6 None of these (DON'T READ OUT) 7 Don't Know Y

End of Filter inouse

DQATT DUMMY TO SET LOOP VARIABLE OATT

The internet is a mystery to me 1 (817) I am interested in new technologies 2 Computers are intimidating to use 3 Computers can be fun 4 Everyone depends on computers too much nowadays 5 Computers will make life much easier if you have one 6 The Internet is fun 7 The Internet is very useful to me 8 Don't Know Y

If the respondent has said that s/he DOES NOT “Use the internet” at QUSENET and DOES NOT use a home computer for “Browsing or surfing the Internet” at QACTSPC, the “The Internet is fun” and “The Internet is very useful to me” options are set at DQATT and are not displayed in the QAT1 or QATT loops.

QUANCEPT ITEM:

IF QUSEPC <> At home THEN ASK: QAT1 ELSE ASK: QATT

www.eurescom.de/e-living/index.htm Page 32 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QAT1 Even though you do not use a computer at home. To what extent do you agree with the following statements about computers and technology. &oatt&

Strongly agree 1 (825) Slightly agree 2 Neither agree nor disagree 3 Slightly disagree 4 Strongly disagree 5 Don't Know Y

This question is repeated for the following loop values:

- The internet is a mystery to me - I am interested in new technologies - Computers are intimidating to use - Computers can be fun - Everyone depends on computers too much nowadays - Computers will make life much easier if you have one - The Internet is fun - The Internet is very useful to me

A total of 8 iterations occupying columns (825) to (832)

QATT To what extent do you agree with the following statements about computers and technology. &oatt&

Strongly agree 1 (833) Slightly agree 2 Neither agree nor disagree 3 Slightly disagree 4 Strongly disagree 5 Don't Know Y

This question is repeated for the following loop values:

A total of 8 iterations occupying columns (833) to (840)

www.eurescom.de/e-living/index.htm Page 33 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

End of Filter iqatt

Now I'd like to ask a few questions about yourself.

QMARRY Firstly, could you please tell me your current marital status?

Married 1 (841) Living as a couple 2 Separated 3 Divorced 4 Widowed 5 Never been married 6 Don't Know Y

QDOBMON Could you please tell me your full date of birth? FIRST ENTER THE MONTH

January 1 (842) February 2 March 3 April 4 May 5 June 6 July 7 August 8 September 9 October 0 (843) November 1 December 2 Refused Z (842)

QDOBDAY NOW ENTER THE DATE IN &QDOBMON&

(844 - 845) Numeric Range Refused Z (844) Permitted Range 0 TO 31 (Numeric Range)

www.eurescom.de/e-living/index.htm Page 34 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QDOBYR NOW ENTER THE YEAR

(846 - 849) Numeric Range Refused Z (846) Permitted Range 1900 TO 1986 (Numeric Range)

The full and partial interviews file contains the following undocumented variable: Variable Comment Name Description XTTIME8 Time in seconds elapsed between Not collected and set to inapplicable for Israel and XTTIME7 and current position Bulgaria

QUANCEPT ITEM:

QWORK Which of the following best describes your current situation? (IF DOING A PAID WORK PLACEMENT AS PART OF STUDY THEN CODE AS IN PAID WORK AND NOT FULL-TIME STUDENT)

In paid work 1 (850) Unemployed 2 Retired from paid work altogether 3 On maternity leave 4 Looking after family or home 5 Full-time student/at school 6 Long term sick or disabled 7

www.eurescom.de/e-living/index.htm Page 35 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

IF QWORK = Unemployed OR QWORK = Looking after family or home OR QWORK = Long term sick or disabled OR THEN ASK: QEVWORK

QEVWORK Have you ever had a paid job for a period of at least six months?

Yes 1 (851) No 2 Don't Know Y

End of Filter iwork

IF QWORK = Retired from paid work altogether OR QEVWORK = Yes THEN ASK: QSTOPWO, QJT1, QFIRM1, QEMP1, QMANAG1

QSTOPWO When did you stop working? ENTER YEAR (four digits)

(852 - 855) Numeric Range Don't Know Y (852) Permitted Range 1900 TO 2001 (Numeric Range)

QJT1 Please tell me the exact job title and describe fully the sort of work you did.

(856 - 859)

Don't Know Y (856)

www.eurescom.de/e-living/index.htm Page 36 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QFIRM1 What did the firm/organisation you worked for actually make or do?

(860 - 863)

Don't Know Y (860)

QEMP1Were you an employee or self-employed?

Employee 1 (864) Self-employed 2 Don't Know Y

QMANAG1 Did you have any managerial duties or did you supervise any other employees?

Yes 1 (865) No 2 Don't Know Y

End of Filter iret

QUANCEPT ITEM:

IF QWORK = In paid work OR QWORK = On maternity leave THEN ASK: QJT2, QFIRM2, QEMP2, QMANAG2, QPERM, QNOWORK, QWHERWO

QJT2 Please tell me the exact job title and describe fully the sort of work you do.

(866 - 869)

Don't Know Y (866)

www.eurescom.de/e-living/index.htm Page 37 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QFIRM2 What does the firm/organisation you work for actually make or do at the place where you work?

(870 - 873)

Don't Know Y (870)

QEMP2Are you an employee or self employed?

Employee 1 (874) Self-employed 2 Don't Know Y

QMANAG2 Do you have any managerial duties or do you supervise any other employees?

Yes 1 (875) No 2 Don't Know Y

QPERM Is your current job:

a permanent job 1 (876) a seasonal, temporary or casual job 2 a job done under contract or for a fixed period of time? 3 Don't Know Y

www.eurescom.de/e-living/index.htm Page 38 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QNOWORK How many people are employed at the place where you work? AT THE SITE WHERE THEY WORK ONLY

1 - 2 1 (877) 3 - 9 2 10 - 24 3 25 - 49 4 50 - 99 5 100 - 199 6 200 - 499 7 500 - 999 8 1000 or more 9 Don't know but fewer than 25 0 (878) Don't know but 25 or more 1 Don't Know Y (877)

QWHERWO Do you work mainly:

At home 1 (879) At work premises 2 Driving or travelling around 3 Or at one or more other places? 4 Other 5 Don't Know Y

IF NOT QWHERWO = At home THEN ASK: QTIMEWO, QTRAVEL

QTIMEWO About how much time does it usually take for you to get to work each day, door to door? ENTER MINUTES ONE WAY JOURNEY ONLY. IF NO USUAL, GIVE AVERAGE.

(908 - 910) Numeric Range Don't Know Y (908) Permitted Range 0 TO 300 (Numeric Range)

www.eurescom.de/e-living/index.htm Page 39 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QTRAVEL How do you usually travel to or from work? (IF MORE THAN ONE MEANS OF TRANSPORT, CODE THE ONE USED OVER LONGEST DISTANCE)

By train, including underground/metro 1 (911) By car 2 By train and car 3 By bus 4 By motorbike / scooter / moped 5 Cycling 6 On foot 7 Other 8 Don't Know Y

IF QTRAVEL = By car OR QTRAVEL = By train and car

IF dqcount = UK THEN ASK: QMILES ELSE ASK: QKM

QMILES How many miles in total do you usually drive a day to get to work? ONE WAY JOURNEY ONLY

(912 - 914) Numeric Range Don't Know Y (912) Permitted Range 0 TO 500 (Numeric Range)

QKM How many kilometres in total do you usually drive a day to get to work?

(915 - 917) Numeric Range Don't Know Y (915) Permitted Range 0 TO 500 (Numeric Range)

End of Filter iuk

End of Filter icar

www.eurescom.de/e-living/index.htm Page 40 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

End of Filter ihowork

QHOURS How many hours do you usually work a week in your current job, including any overtime? ENTER HOURS

(918 - 920) Numeric Range Don't Know Y (918) Permitted Range 1 TO 100 (Numeric Range)

IF QEMP2 = Employee THEN ASK: QSCHED, QOWNPAY, QPAYPER

QSCHED Do you mostly set your own work schedule? By that I mean that you can decide when you work on different aspects of your job?

Mostly set own schedule 1 (921) Do not set own work schedule 2 Don't Know Y

QUANCEPT ITEM:

QOWNPAY What is your usual gross pay - that is including any overtime, bonuses, commission, tips or tax refund, but before any deductions for tax, national insurance or pension contributions, union dues and so on? ENTER AMOUNT TO THE NEAREST POUND

(2608 - 2616) Numeric Range Don't Know Y (2608) Refused Z Permitted Range 0 TO 999999999 (Numeric Range)

QUANCEPT ITEM:

www.eurescom.de/e-living/index.htm Page 41 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QPAYPER What period does that cover?

Week 1 (928) Fortnight 2 Four weeks 3 Calendar month 4 Quarter 5 Six months 6 Year 7 Other (TYPE IN) 8 Don't Know Y

End of Filter iemploy

IF QEMP2 = Self-employed THEN ASK: QPAY, QPERIOD

QUANCEPT ITEM:

QPAY On average, what was your gross WEEKLY or MONTHLY income from this job/business over the last 12 months? ENTER AMOUNT

(2618 - 2626) Numeric Range Don't Know Y (2618) Refused Z Permitted Range 0 TO 999999999 (Numeric Range)

QUANCEPT ITEM:

QPERIOD Was that weekly or monthly income?

Weekly 1 (935) Monthly 2 Don't Know Y

End of Filter iself

www.eurescom.de/e-living/index.htm Page 42 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QUALIF What qualification does someone usually need to be able to do your job? DO NOT READ OUT

No qualifications/GCS E or O-level D- G/CSE 2- 5/Scottish O or Standard Grade D- E or 4-7 1 (940) NVQ1 or SVQ1/City and Guild Craft/Scotvec modules/clerical or commercial qualification/appre nticeship 2 NVQ2/SVQ level 2/City and Guilds Advanced or Part II/Scotvec Higher 3 GCSE A-C /O-level pass /CSE grade 1/O or Standard Grades A-C or 1-3 4 NVQ/SVQ level 3/ ONC/OND/BTEC/ Scotvec National/City and Guilds Full Technological/GN VQ 5 DUMMY CODE 6 AS Levels/Scottish Highers 7 A levels/Higher School Certificate/Certifica te of 6th Form Studies 8 DUMMY CODE 1 9 NVQ/SVQ level 4/ HNC/HCD/BTEC or Scotvec Higher/Teaching or nursing qualification (not degree) 0 (941) DUMMY CODE 2 1 First degree 2 DUMMY CODE 3 3 Masters 4 PhD 5 Other 6 Don't Know Y (940)

www.eurescom.de/e-living/index.htm Page 43 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

IF QUALIF = Other THEN ASK: qualifo

qualifo ENTER OTHER TEXT FOR QUALIF (QUALIFICATION TO DO YOUR JOB)

(942 - 945)

None of these X (942)

End of Filter ioth

IF dqcount = Germany

Note: NULL filter; actions removed

End of Filter iqge

IF QUSEPC = At work or for your job THEN ASK: QCOMHRS

QCOMHRS How many hours a day do you usually use a computer as part of your job? ENTER NUMBER OF HOURS

(946 - 947) Numeric Range Don't Know Y (946) Permitted Range 0 TO 20 (Numeric Range)

IF QWERNET = at work or college THEN ASK: QNETWOR

www.eurescom.de/e-living/index.htm Page 44 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QNETWOR How often would you say you use the internet, for work purposes?

Most days 1 (948) About once a week 2 Less often 3 Never 4 Don't Know Y

End of Filter iwernet

QCOMUSE Of the following computer tasks, which best describes the way you MOSTLY use your computer at work? READ OUT.

Word processing 1 (949) Web design or management 2 Spreadsheets / database 3 E-mail or internet 4 Design, analysis, or desk-top publishing 5

Programming/netw ork systems management, PC support 6 Don't Know Y

QCOMUSE was (mistakenly) implemented as a multiple response question

QUANCEPT ITEM:

QCOMYRS For how many years have you been using a computer at work?

Less than one year 1 (956) 1-3 years 2 4-6 years 3 7-10 years 4 More than 10 years 5 Don't Know Y

End of Filter iwocom

www.eurescom.de/e-living/index.htm Page 45 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QHOMEWO How often do you work at home on weekdays, during normal working hours as part of your job?

Most days 1 (957) 2-3 times a week 2 About once a week 3 About once a fortnight 4 About once a month 5 Several times a year 6 Less often 7 Never 8 Don't Know Y

IF QHOMEWO = Never THEN ASK: QHOMHRS ELSE ASK: QHOMHR2

QHOMHRS Even though you never work at home during normal working hours, how many hours if any, do you work at home for your job in the evenings or weekends? ENTER HOURS. IF LESS THAN ONE HOUR PUT 0.

(958 - 959) Numeric Range Don't Know Y (958) Permitted Range 0 TO 99 (Numeric Range)

QHOMHR2 In total then, how many hours a week do you usually work at home for your job including any work done in the evenings or at the weekends? ENTER HOURS. IF LESS THAN 1 HOUR PUT 0.

(960 - 961) Numeric Range Don't Know Y (960) Permitted Range 0 TO 99 (Numeric Range)

End of Filter ihomewo

IF QACTSPC = Paid work

www.eurescom.de/e-living/index.htm Page 46 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

IF NOT QHOMHRS = 0 THEN ASK: QPROP1

QPROP1 What proportion of time of your work at home, if any, is spent working with a computer?

Most of the time 1 (962) About half the time 2 Not much of the time 3 None at all 4 Don't Know Y

IF NOT ( QPROP1 = None at all OR QPROP1 = Don't Know ) THEN ASK: QPROP2

QPROP2 What proportion of time of your work at home is spent working with an on-line connection to your workplace?

Most of the time 1 (963) About half the time 2 Not much of the time 3 None at all 4 Don't Know Y

End of Filter iprop

End of Filter ihomhrs

End of Filter ihrwork

QIMP How important is it to improve your computer skills to keep up with your job?

Very important 1 (964) Quite important 2 Not important 3 Don't Know Y

End of Filter iemp

IF QWORK = Full-time student/at school THEN ASK: QSTUDY, QSTUHRS

www.eurescom.de/e-living/index.htm Page 47 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QSTUDY Could you tell me what type of educational institution you are attending?

High or secondary school 1 (965) College 2 University 3 Don't Know Y

QSTUHRS On average, how many hours a day do you usually use a computer for your studies? ENTER HOURS

(966 - 967) Numeric Range Don't Know Y (966) Permitted Range 0 TO 12 (Numeric Range)

IF NOT QSTUHRS = 0 THEN ASK: QLONG

QLONG For how many years have you been using a computer in your education?

Less than one year 1 (968) 1-3 years 2 4-6 years 3 7-10 years 4 More than 10 years 5 Don't Know Y

End of Filter istuhrs

QSTUHOM How often do you study at home during weekdays, that is during normal course time?

Most days 1 (969) 2-3 times a week 2 About once a week 3 About once a fortnight 4 About once a month 5 Several times a year 6 Less often 7 Never 8 Don't Know Y

www.eurescom.de/e-living/index.htm Page 48 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

IF NOT ( QACTSPC <> Educational purposes such as private study or school or college work AND QSTUHOM = Never ) THEN ASK: QPROP3

QPROP3 When you do study at home at any time of the week, what proportion of time is spent working with a computer?

Most of the time 1 (970) About half the time 2 Not much of the time 3 None at all 4 Don't Know Y

IF NOT ( QPROP3 = None at all ) THEN ASK: QPROP4

QPROP4 When you study at home, what proportion of time is spent working with an on-line connection to your place of education?

Most of the time 1 (971) About half the time 2 Not much of the time 3 None at all 4 Don't Know Y

End of Filter iprop3

End of Filter iPCstud

QSKILLS How important is it to improve your computer skills to keep up with your education?

Very important 1 (972) Quite important 2 Not important 3 Don't Know Y

End of Filter istu

IF NOT QWORK = Full-time student/at school THEN ASK: QEDINST, QTEA

www.eurescom.de/e-living/index.htm Page 49 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QEDINST Could you tell me what type of educational institution you last attended full-time?

Primary or elementary school 1 (973) High or secondary school 2 College 3 University 4 Don't Know Y

QTEA How old were you when you left full-time education?

14 or under 1 (974) 15 2 16 3 17 4 18 5 19 6 20 7 21 8 22 9 23 0 (975) 24 1 25 2 26 3 27 4 28 5 29 6 30 7 31 and above 8 Still studying 9 Don't Know Y (974)

End of Filter iftstu

QUANCEPT ITEM:

www.eurescom.de/e-living/index.htm Page 50 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QHIQUAL What is the highest qualification that you have? DO NOT READ OUT

No qualifications/GCS E or O-level D- G/CSE 2- 5/Scottish O or Standard Grade D- E or 4-7 1 (1022) NVQ1 or SVQ1/City and Guild Craft/Scotvec modules/clerical or commercial qualification/appre nticeship 2 NVQ2/SVQ level 2/City and Guilds Advanced or Part II/Scotvec Higher 3 GCSE A-C /O-level pass /CSE grade 1/O or Standard Grades A-C or 1-3 4 NVQ/SVQ level 3/ ONC/OND/BTEC/ Scotvec National/City and Guilds Full Technological/GN VQ 5 DUMMY CODE 6 AS Levels/Scottish Highers 7 A levels/Higher School Certificate/Certifica te of 6th Form Studies 8 DUMMY CODE 1 9 NVQ/SVQ level 4/ HNC/HCD/BTEC or Scotvec Higher/Teaching or nursing qualification (not degree) 0 (1023) DUMMY CODE 2 1 First degree 2 DUMMY CODE 3 3 Masters 4 PhD 5 Other 6 Don't Know Y (1022) None of these X

www.eurescom.de/e-living/index.htm Page 51 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

IF DQCOUNT = Germany THEN ASK: QGERQUA

www.eurescom.de/e-living/index.htm Page 52 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QGERQUA What is the next highest qualification that you have?

No qualifications/GCS E or O-level D- G/CSE 2- 5/Scottish O or Standard Grade D- E or 4-7 1 (1024) NVQ1 or SVQ1/City and Guild Craft/Scotvec modules/clerical or commercial qualification/appre nticeship 2 NVQ2/SVQ level 2/City and Guilds Advanced or Part II/Scotvec Higher 3 GCSE A-C /O-level pass /CSE grade 1/O or Standard Grades A-C or 1-3 4 NVQ/SVQ level 3/ ONC/OND/BTEC/ Scotvec National/City and Guilds Full Technological/GN VQ 5 DUMMY CODE 6 AS Levels/Scottish Highers 7 A levels/Higher School Certificate/Certifica te of 6th Form Studies 8 DUMMY CODE 1 9 NVQ/SVQ level 4/ HNC/HCD/BTEC or Scotvec Higher/Teaching or nursing qualification (not degree) 0 (1025) DUMMY CODE 2 1 First degree 2 DUMMY CODE 3 3 Masters 4 PhD 5 Other 6 Don't Know Y (1024)

End of Filter IQGER

www.eurescom.de/e-living/index.htm Page 53 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

The full and partial interviews file contains the following undocumented variable: Variable Comment Name Description XTTIME9 Time in seconds elapsed between Not collected and set to inapplicable for Israel and XTTIME8 and current position Bulgaria

QUANCEPT ITEM:

QPHBILL Approximately how much is your household's usual phone charges, that is for fixed phones only, and including the cost of line rental, calls and any internet charges?

IF NECESSARY: by charges I mean your usual phone bill or bills

(1037 - 1045) Numeric Range Don't Know Y (1037) Permitted Range 0 TO 999999999 (Numeric Range)

QPHPER What period does that cover?

Week 1 (1046) Fortnight 2 Four weeks 3 Calendar month 4 Two months 5 Quarter 6 Six months 7 Don't Know Y Other 0

Other specify... (1047 - 1050)

www.eurescom.de/e-living/index.htm Page 54 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QUANCEPT ITEM:

QMOBILE How many mobile phones, if any, does your household have?

One 1 (1051) Two 2 Three or more 3 None 4 Don't Know Y

IF QMOBILE = One THEN ASK: QWAP

QWAP Is this a WAP phone, that is, linked to the internet?

Yes 1 (1052) No 2 Don't Know Y

End of Filter imob1

IF QMOBILE = Two OR QMOBILE = Three or more THEN ASK: QNOWAPS

QNOWAPS How many of these are WAP phones, that is, linked to the internet? @@WRITE IN NUMBER

(1053 - 1054) Numeric Range Don't Know Y (1053) Permitted Range 0 TO 10 (Numeric Range)

End of Filter imob2

IF NOT (( QMOBILE = One OR QMOBILE = Two OR QMOBILE = Three or more ) AND QHHOLD = 1 )

www.eurescom.de/e-living/index.htm Page 55 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

IF QMOBILE <> None AND QMOBILE <> Don't Know THEN ASK: QOWNMOB

QOWNMOB Do you personally have a mobile phone?

Yes 1 (1055) No 2 Don't Know Y

End of Filter inomob

End of Filter iownmo1

DQMOB DUMMY TO SET MOBILE OWNERSHIP

Yes 1 (1056) No 2 Don't Know Y

DQMOB is set to “yes” if the respondent answered “yes” @ QOWNMOB, or if s/he responded “one”, “two” or “three or more” @ QMOBILE

QUANCEPT ITEM:

IF dqmob = Yes

IF QWORK = In paid work OR QWORK = On maternity leave THEN ASK: QMOBIMP

QMOBIMP How important is your mobile phone for your work ? Is it..

Very important 1 (1057) Important 2 Not very important 3 Not at all important 4 Don't Know Y

End of Filter iworkmo

www.eurescom.de/e-living/index.htm Page 56 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QMOBFRI How important is it to you for social calling, for instance to friends? Is it...

Very important 1 (1058) Important 2 Not very important 3 Not at all important 4 Don't Know Y

QMOBFIX When you use a phone at home, which do you prefer to use?

Your fixed phone / landline 1 (1059) Your mobile phone 2 Don't Know Y

QMOBCAL On average, how many calls would you say you make a day on your mobile?

Less than one 1 (1060) 1-5 2 6-10 3 11-20 4 21-30 5 31-50 6 More than 50 7 Don't Know Y

QMOBTEX On average how many text or SMS messages do you send a day?

Less than one 1 (1061) 1 – 5 2 6 – 10 3 11 – 20 4 21 – 30 5 30 – 50 6 More than 50 7 Never send text messages 8 Don't Know Y

QMOBBIL Overall, can you tell me how much your mobile telephone bill is per month including pre-pay cards? ENTER TO NEAREST POUND

(1062 - 1070) Numeric Range Don't Know Y (1062) Permitted Range 0 TO 900000000 (Numeric Range)

www.eurescom.de/e-living/index.htm Page 57 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

End of Filter ipermob

QPHO To what extent do you agree with the following statements. ...

Strongly agree 1 (1071) Slightly agree 2 Neither agree nor disagree 3 Slightly disagree 4 Strongly disagree 5 Don't Know Y

This question is repeated for the following loop values:

- I enjoy speaking to people on the phone - I could spend hours on the phone given the chance - I only use the phone when I have to - I need the phone to organise my everyday life

A total of 4 iterations occupying columns (1071) to (1074)

QENV To what extent do you agree with the following statements. @@...@@ Do you

Strongly agree 1 (1075) Slightly agree 2 Neither agree nor disagree 3 Slightly disagree 4 Strongly disagree 5 Don't Know Y

This question is repeated for the following loop values:

- We worry too much about the future of the environment, and not enough about prices and jobs today. - I try to buy organic food or food with a low level of pesticides

A total of 2 iterations occupying columns (1075) to (1076)

The full and partial interviews file contains the following undocumented variable: Variable Comment Name Description XTTIM10 Time in seconds elapsed between Not collected and set to inapplicable for Israel and XTTIME9 and current position Bulgaria

QUANCEPT ITEM:

www.eurescom.de/e-living/index.htm Page 58 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QUANCEPT ITEM:

I have just a few more questions about yourself and your home.

The next section of the script collects information about the sex, age, employment status and relationship of other household members (if any) to the household reference person (HRP), as well as the relationship of the respondent to the HRP (if someone other than the respondent is the household reference person). It is implemented as a series of loops, each with a maximum of 11 iterations, though the routing of a respondent through the loops depends on data other than the number of people in his/her household in a way that is not immediately obvious. In particular.

The immediately following filter checks if there is more than one person in the household (QHHOLD > 1) AND, IF NOT, skips this section entirely). OTHERWISE:

IFf the respondent is the household reference person (QHRP = 1) person AND there is only one other person in the household (QHHOLD = 2), THEN data about that other person is captured in the variables corresponding to the first iteration of the loops, i.e QEX[1], QAGE[1], QFULLEM[1] and QRELATE[1]. (This is the only circumstance under which variables corresponding to the first iteration of the loops will contain data.)

IF the respondent is the household reference person (QHRP = 1) AND there is more than one other person in the household (QHHOLD > 2), THEN data about those other people are captured in the variables corresponding to the second and subsequent iterations of the loops to a maximum of QHHOLD , i.e QEX[2] .. QSEX[QHHOLD], QAGE[2] .. QAGE[QHHOLD], QFULLEM[2] .. QFULLEM[QHHOLD] and QRELATE[2] .. QRELATE[QHHOLD].

IF someone other than the respondent is the household reference person (QHRP = 2), THEN: a) name, sex, age and employment status data about the household reference person is captured in the variables corresponding to the 11th iteration of the loop, i.e. QNAME[11], QSEX[11], QAGE[11] and QFULLEM[11]; b) the relationship of the respondent to the household reference person is captured in QRELATE[11]; c) data about other members of the household is captured in the second and subsequent iterations of the loops to a maximum of QHHOLD-1, i.e. QEX[2] .. QSEX[QHHOLD-1], QAGE[2] .. QAGE[QHHOLD-1], QFULLEM[2] .. QFULLEM[QHHOLD-1] and QRELATE[2] .. QRELATE[QHHOLD-1]

(This is the only circumstance under which data will be found the the variables corresponding to the 11th iterations of the loops.)

In all cases, QFULLEM is asked only if the member of the household 16 years of age or older; if not, this question is skipped and will be Inapplicable on output.

www.eurescom.de/e-living/index.htm Page 59 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

IF NOT QHHOLD = 1 THEN ASK: QHRP

When analysing the kinds of people who use home computers and their attitudes towards them, it is extremely helpful to have some kind of background on the their household structure. I would therefore like to ask you just a few questions about your household.

QUANCEPT ITEM:

QHRP You said at the beginning that there are &QHHOLD& people in your household. Firstly, who is mainly responsible for paying the rent or mortgage or has the main responsibility for the accommodation? IF 2 NAMES GIVEN TAKE OLDEST IF NOT RESPONDENT, TYPE IN NAME OF HOUSEHOLD REFERENCE PERSON ON NEXT SCREEN

Me (respondent) 1 (1077) Other person 2 Don't Know Y

IF QHRP = Other person THEN ASK: QHRP1

QHRP1 ENTER NAME OF HOUSEHOLD REFERENCE PERSON

(1108 - 1111)

Don't Know Y (1108)

End of Filter iqhrp

QUANCEPT ITEM:

www.eurescom.de/e-living/index.htm Page 60 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

DQHOLD DUMMY TO SET LOOP FOR COLLECTING NAMES

other person 1 (1112) 2nd person 2 3rd person 3 4th person 4 5th person 5 6th person 6 7th person 7 8th person 8 9th person 9 10th person 0 (1113)

QUANCEPT ITEM:

QNAMES Can you tell me the first name of the &onames& in your household, other than yourself?

TYPE IN NAMES, DO NOT TAKE THE RESPONDENT'S NAME THIS IS COLLECTED LATER IF RESPONDENT REFUSES THEN TYPE PERSON 1, THEN PERSON 2, etc.

(1114 - 1117)

Don't Know Y (1114)

This question is repeated for the following loop values:

- other person - 2nd person - 3rd person - 4th person - 5th person - 6th person - 7th person - 8th person - 9th person - 10th person

A total of 10 iterations occupying columns (1114 - 1117) to (1150 - 1153)

A bug in the Quancept script caused all data in this loop to be lost when the number of household members (QHHOLD) was greater than 10; this happened in two cases, serial numbers 119083 (UK) and 447628 (Norway)

www.eurescom.de/e-living/index.htm Page 61 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QUANCEPT ITEM:

DQNAME DUMMY TO CONTROL THE LOOP FOR DEMOGRAPHICS OF OTHER MEMBERS OF HOUSEHOLD

[+name1+] 1 (1318) [+name2+] 2 [+name3+] 3 [+name4+] 4 [+name5+] 5 [+name6+] 6 [+name7+] 7 [+name8+] 8 [+name9+] 9 [+name10+] 0 (1319) [+name11+] 1 Don't Know Y (1318) A bug in the Quancept script caused all data in this loop to be lost when the number of household members (QHHOLD) was greater than 10; this happened in two cases, serial numbers 119083 (UK) and 447628 (Norway)

QUANCEPT ITEM:

QSEX IF NOT CLEAR ASK - Is &odtails& male or female?

Male 1 (1328) Female 2 Don't Know Y

This question is repeated for the following loop values:

- [+name1+] - [+name2+] - [+name3+] - [+name4+] - [+name5+] - [+name6+] - [+name7+] - [+name8+] - [+name9+] - [+name10+] - [+name11+]

A total of 11 iterations occupying columns (1328) to (1338) www.eurescom.de/e-living/index.htm Page 62 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QAGE What is &odtails& age?

(1339 - 1341) Numeric Range Don't Know Y (1339) Permitted Range 0 TO 120 (Numeric Range)

This question is repeated for the following loop values:

- [+name1+] - [+name2+] - [+name3+] - [+name4+] - [+name5+] - [+name6+] - [+name7+] - [+name8+] - [+name9+] - [+name10+] - [+name11+]

A total of 11 iterations occupying columns (1339 - 1341) to (1369 - 1371)

QUANCEPT ITEM:

IF NOT QAGE < 16 THEN ASK: QFULLEM

www.eurescom.de/e-living/index.htm Page 63 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QFULLEM Is &odtails& in full-time paid employment, that is working 16 hours or more a week?

Yes 1 (1408) No 2 Don't Know Y

This question is repeated for the following loop values:

- [+name1+] - [+name2+] - [+name3+] - [+name4+] - [+name5+] - [+name6+] - [+name7+] - [+name8+] - [+name9+] - [+name10+] - [+name11+]

A total of 11 iterations occupying columns (1408) to (1418)

End of Filter i16

QUANCEPT ITEM:

www.eurescom.de/e-living/index.htm Page 64 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QRELATE Could you tell me the relationship of &odtails& to &temp1&?

Husband 1 (1419) Wife 2 Cohabiting partner 3 Father 4 Mother 5 Son/daughter 6 Brother/sister 7 Grandchild 8 Grandparent 9 Other relative 0 (1420) Friend/tenant/other 1 Don't Know Y (1419)

This question is repeated for the following loop values:

- [+name1+] - [+name2+] - [+name3+] - [+name4+] - [+name5+] - [+name6+] - [+name7+] - [+name8+] - [+name9+] - [+name10+] - [+name11+]

A total of 11 iterations occupying columns (1419 - 1420) to (1439 - 1440)

End of Filter ione

The full and partial interviews file contains the following undocumented variable: Variable Comment Name Description XTTIM11 Time in seconds elapsed between Not collected and set to inapplicable for Israel and XTTIM10 and current position Bulgaria

QUANCEPT ITEM:

QUANCEPT ITEM: www.eurescom.de/e-living/index.htm Page 65 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QUANCEPT ITEM:

QCHORES About how many hours do you personally spend on housework in an average week, such as time spent cooking, cleaning and doing the laundry? ENTER NUMBER OF HOURS

(1441 - 1442) Numeric Range Don't Know Y (1441) Permitted Range 0 TO 99 (Numeric Range)

www.eurescom.de/e-living/index.htm Page 66 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

DQINC Could you please tell me your household's total gross monthly income from work, pensions, benefits and investments.

Less than £500 1 (1443) £500 - £999 2 £1,000 - £1,999 3 £2,000 - £2,999 4 £3,000 - £3,999 5 £4,000 - £4,999 6 £5,000 - £10,000 7 Over £10,000 8 Less than 1,610,000 Lira 9 1,610,000 - 3,199,999 Lira 0 (1444) 3,200,000 - 6,499,999 Lira 1 6,500,000 - 9,499,999 Lira 2 9,500,000 - 12,999,999 Lira 3 13,000,000 - 15,999,999 Lira 4 16,000,000 - 32,199,999 Lira 5 Over 32,200,000 Lira 6 Less than 1,600 DEM 7 1,600 - 3,249 DEM 8 3,250 - 6,499 DEM 9 6,500 - 9,749 DEM 0 (1445) 9,750 - 12,999 DEM 1 13,000 - 16,249 DEM 2 16,250 - 32,499 DEM 3 Over 32,500 DEM 4 Less than 6,500 Kroner 5 6,500 - 12,999 Kroner 6 13,000 - 26,299 Kroner 7 26,300 - 39,399 Kroner 8 39,400 - 52,499 Kroner 9 52,500 - 65,699 Kroner 0 (1446) 65,700 - 131,399 Kroner 1 Over 131,400 Kroner 2 Don't Know Y (1443) Refused Z

www.eurescom.de/e-living/index.htm Page 67 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

Values of DQINC are set according to the country of interview @ DQCOUNT; options not appropriate to the country are masked out @ QTOTINC and QPERSON.

QUANCEPT ITEM:

IF NOT ( QHHOLD = 1 ) THEN ASK: QTOTINC

www.eurescom.de/e-living/index.htm Page 68 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QTOTINC Could you please tell me your household's total gross monthly income from work, pensions, benefits and investments.

Less than £500 1 (1449) £500 - £999 2 £1,000 - £1,999 3 £2,000 - £2,999 4 £3,000 - £3,999 5 £4,000 - £4,999 6 £5,000 - £10,000 7 Over £10,000 8 Less than 1,610,000 Lira 9 1,610,000 - 3,199,999 Lira 0 (1450) 3,200,000 - 6,499,999 Lira 1 6,500,000 - 9,499,999 Lira 2 9,500,000 - 12,999,999 Lira 3 13,000,000 - 15,999,999 Lira 4 16,000,000 - 32,199,999 Lira 5 Over 32,200,000 Lira 6 Less than 1,600 DEM 7 1,600 - 3,249 DEM 8 3,250 - 6,499 DEM 9 6,500 - 9,749 DEM 0 (1451) 9,750 - 12,999 DEM 1 13,000 - 16,249 DEM 2 16,250 - 32,499 DEM 3 Over 32,500 DEM 4 Less than 6,500 Kroner 5 6,500 - 12,999 Kroner 6 13,000 - 26,299 Kroner 7 26,300 - 39,399 Kroner 8 39,400 - 52,499 Kroner 9 52,500 - 65,699 Kroner 0 (1452) 65,700 - 131,399 Kroner 1 Over 131,400 Kroner 2 Don't Know Y (1449) Refused Z

www.eurescom.de/e-living/index.htm Page 69 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

Options not appropriate to the country of interview are masked out according to the values set @ DQINC.

End of Filter iinc

IF QHHOLD = 1 THEN ASK: QPERSON

www.eurescom.de/e-living/index.htm Page 70 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QPERSON Could you please tell me your total gross monthly income from work, pensions, benefits and investments. Is it...

Less than £500 1 (1453) £500 - £999 2 £1,000 - £1,999 3 £2,000 - £2,999 4 £3,000 - £3,999 5 £4,000 - £4,999 6 £5,000 - £10,000 7 Over £10,000 8 Less than 1,610,000 Lira 9 1,610,000 - 3,199,999 Lira 0 (1454) 3,200,000 - 6,499,999 Lira 1 6,500,000 - 9,499,999 Lira 2 9,500,000 - 12,999,999 Lira 3 13,000,000 - 15,999,999 Lira 4 16,000,000 - 32,199,999 Lira 5 Over 32,200,000 Lira 6 Less than 1,600 DEM 7 1,600 - 3,249 DEM 8 3,250 - 6,499 DEM 9 6,500 - 9,749 DEM 0 (1455) 9,750 - 12,999 DEM 1 13,000 - 16,249 DEM 2 16,250 - 32,499 DEM 3 Over 32,500 DEM 4 Less than 6,500 Kroner 5 6,500 - 12,999 Kroner 6 13,000 - 26,299 Kroner 7 26,300 - 39,399 Kroner 8 39,400 - 52,499 Kroner 9 52,500 - 65,699 Kroner 0 (1456) 65,700 - 131,399 Kroner 1 Over 131,400 Kroner 2 Don't Know Y (1453) Refused Z

Options not appropriate to the country of interview are masked out according to the values set @ DQINC. www.eurescom.de/e-living/index.htm Page 71 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

End of Filter ionep

Finally, a few questions about your life overall.

QTALK How many people would you say you know outside your immediate family you could really count on to listen to you when you need to talk?

1-3 1 (1457) 4-10 2 11-20 3 More than 20 4 None 5 Don't Know Y

QLIF To what extent do you agree with the following statements. &olif&

Do you...

Strongly agree 1 (1458) Slightly agree 2 Neither agree nor disagree 3 Slightly disagree 4 Strongly disagree 5 Don't Know Y

This question is repeated for the following loop values:

- Overall the conditions of my life are excellent. - I have enough free time to do what I want. - The environmental conditions in my area are good. - I have good communications with friends.

A total of 4 iterations occupying columns (1458) to (1461)

IF QWORK = In paid work OR QWORK = On maternity leave THEN ASK: QUESTI1

www.eurescom.de/e-living/index.htm Page 72 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QUESTI1 And last of all, to what extent do you agree with the statement: In most ways my working life is close to ideal?

Strongly agree 1 (1462) Slightly agree 2 Neither agree nor disagree 3 Slightly disagree 4 Strongly disagree 5 Don't Know Y

End of Filter ilif

QMOVE Thank you for taking part in this survey. There is a possibility that we may be following up this survey in a year's time to see how communications have developed across Europe. We would really like to be able to re-contact you should we decide to do this. So, could I just ask if you think you will be moving in the next 12 months?

Definitely moving 1 (1463) Possibly moving / Don't know if moving 2 Not moving 3 Refused to say if moving 4 Not keen on next survey 5 Absolute refusal of next survey 6

IF QMOVE = Definitely moving THEN ASK: QNEWNUM

QNEWNUM Do you know already what your new telephone number will be?

Yes 1 (1464) No 2 Don't Know Y Refused Z

End of Filter idefmov

IF Qnewnum = Yes THEN ASK: QKNOW

www.eurescom.de/e-living/index.htm Page 73 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QKNOW Can I just make a note of your new number then please. TYPE IN NEW TELEPHONE NUMBER

(1465 - 1468)

Don't Know Y (1465) Refused Z

End of Filter inewnum

IF Qnewnum = No OR QMOVE = Possibly moving / Don't know if moving THEN ASK: QCONTAC

QCONTAC In that case, could I take the phone number of a friend or neighbour who may be able to give us your new contact details should you have moved in a year's time? TYPE IN

(1469 - 1472)

Don't Know Y (1469) Refused Z

End of Filter icont

QNAME And finally, just to show that I have spoken to someone can I take your full name and address?

Yes 1 (1473) No 2 Don't Know Y

IF Qname = Yes THEN ASK: QTITLE, QFNAME, QSNAME, QADDR1, QADDR2, QADDR3, QADDR4, QADDR5, QPCODE ELSE ASK: qpost

www.eurescom.de/e-living/index.htm Page 74 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QTITLEENTER TITLE (MR / MRS / MISS)

(2408 - 2411)

None of these X (2408)

QFNAME ENTER FIRST NAME

(2412 - 2415)

None of these X (2412)

QSNAME ENTER SURNAME

(2416 - 2419)

None of these X (2416)

QADDR1 ENTER 1ST LINE OF ADDRESS (HOUSE NAME / NUMBER ON STREET / BUILDING)

(2420 - 2423)

None of these X (2420)

www.eurescom.de/e-living/index.htm Page 75 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QADDR2 ENTER 2ND LINE OF ADDRESS (STREET)

(2424 - 2427)

None of these X (2424)

QADDR3 ENTER 3RD LINE OF ADDRESS (TOWN)

(2428 - 2431)

None of these X (2428)

QADDR4 ENTER 4TH LINE OF ADDRESS (COUNTY)

(2432 - 2435)

None of these X (2432)

QADDR5 ENTER 5TH LINE OF ADDRESS (COUNTRY)

(2436 - 2439)

None of these X (2436)

www.eurescom.de/e-living/index.htm Page 76 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QPCODE ENTER POST CODE

(2440 - 2443)

None of these X (2440)

QUANCEPT ITEM:

qpost Would you mind then if I just took your post code so that we can look at our data by region?

(2650 - 2653)

Refused Z (2650) Not exported to SPSS

The following undocumented variables are included in the full and partial interview files:: Variable Comment Name Description XTTIM12 Time in seconds elapsed between time Not collected and set to inapplicable for Israel and marker XTTIM11 and current position Bulgaria XTOTIME Interview duration in seconds Not collected and set to inapplicable for Israel and Bulgaria ISCO International standard classification of See occupations (ISCO-88) code http://www.ilo.org/public/english/bureau/stat/class/ isco.htm NACE Nomenclature of economic activities in See the European Union (NACE 4 1993 http://www.econ.ucl.ac.be/ires/Base_de_donnees/ REV.1) code nomenclatures/nace/NACE4_93_ENGL.html NUTS Nomenclature of Territorial Units for NUTS coding for UK partial interviews may Statistics, level 3 (NUTS 3) code (UK, contain some missing data. Germany, Italy, Norway)/??? NUTS codes in the data do not necessarily match www.eurescom.de/e-living/index.htm Page 77 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

Variable Comment Name Description (Bulgaria/Israel) the official NUTS 3 coding frames for reasons unknown; this is under investigation See http://www.cordis.lu/en/src/d_014_en.htm for NUTS codes. We have no information at the moment regarding frames used to code this variable for Bulgarian and Israel; this is under investigation.

The following undocumented variables are included in all SPSS files: Variable Comment Name Description SDATED Day on which interview started SDATEM Month in which interview started SDATEY Year which interview started STIMEH Hour at which interview started Not collected and set to inapplicable for Israel and Bulgaria STIMEM Minute at which interview started Not collected and set to inapplicable for Israel and Bulgaria STIMES Second at which interview started Not collected and set to inapplicable for Israel and Bulgaria EDATED Day on which interview ended Not collected and set to inapplicable for Israel and Bulgaria EDATEM Month in which interview ended Not collected and set to inapplicable for Israel and Bulgaria EDATEY Year which interview ended Not collected and set to inapplicable for Israel and Bulgaria ETIMEH Hour at which interview ended Not collected and set to inapplicable for Israel and Bulgaria ETIMEM Minute at which interview ended Not collected and set to inapplicable for Israel and Bulgaria ETIMES Second at which interview ended Not collected and set to inapplicable for Israel and Bulgaria

QUANCEPT ITEM:

End of Filter iyes

QUANCEPT ITEM:

www.eurescom.de/e-living/index.htm Page 78 of 79 PUBLIC: E LIVING, WAVE 1, CATI QUESTIONNAIRE Monday, 10 June 2002

QUANCEPT ITEM:

IF vquit1 = 1 THEN ASK: QWHYQT

Thank you very much for your time then and I can assure you that all your answers will be treated with the strictest of confidence. Thank you, Goodbye.

QUANCEPT ITEM:

QWHYQT ENTER COMMENTS ON WHY RESPONDENT QUIT OUT OF INTERVIEW

(2444 - 2447)

None of these X (2444) QWHYQT is asked whenever an interviewer uses the Quancept “quit” command to terminate an interview early. The “quit” command was also sometimes used in error for refusers, so QWHYQT may also contain responses from them.

ZQUIT IF vquit1 = 1 - Termination with data (Quit)

End of Filter iquit

www.eurescom.de/e-living/index.htm Page 79 of 79 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

e-Living D7.1 - ICT Uptake and Usage: A Cross- Sectional Analysis

Yoel Raban (ICTAF) Tal Soffer (ICTAF) Pencho Mihnev (Virtech Ltd.) Kaloyan Ganev (Virtech Ltd.)

e-Living: Life in a Digital Europe, an EU Fifth Framework Project [IST-2000-25409] www.eurescom.de/e-living/index.htm PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

Table of Contents

1 Objectives of this report...... 3 2 E-Living Survey Description ...... 3 3 Sample demographics...... 4 4 ICT Take-up ...... 5 4.1 PC & Internet take up...... 5 4.2 Mobile phones take up...... 8 4.3 Internet diffusion...... 9 5 ICTs use ...... 12 5.1 Mobile use...... 12 5.2 Internet use ...... 15 5.3 Internet & mobile: complementary or substitutes...... 18 6 ICTs impact ...... 19 6.1 TV watching and Internet use ...... 19 6.2 Effects of Email use ...... 20 6.3 Quality of life and Internet use ...... 22 7 Summary and conclusions ...... 22 7.1 ICTs take up...... 23 7.2 ICTs use...... 24 7.3 The Impact of ICTs ...... 25 8 Bibliography...... 25

www.eurescom.de/e-living/index.htm Page 2 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

1 Objectives of this report The objective of the work package to which this report contributes is to describe, explain and model the changing patterns of uptake and usage of ICT products and services across Europe. This objective relates to both wave I and wave II of the e-Living survey (see next paragraph), which will together constitute a panel study. The objective of this report is to describe and analyze ICTs uptake, use and impact in the cross- sectional database of the six countries participating in the wave I survey. Previous research has shown that demographics (education, income, etc.) play a key role in explaining the variety of ICTs take up patterns across and within countries. Several researchers have demonstrated the importance of demographic variables in explaining the penetration rate of ICT products and services, as described in WP3 state of the art review in 2001 (OECD 1998 and 2002, NTIA 1995 and 1998, Clemente 1998, and many others). The state of the art review for WP3 has also shown that explaining and predicting ICTs use, and especially the impact of ICTs use, is still a formidable and challenging task. One of the main tasks of the work package is to use sample demographics, as well as other variables, from the e-Living data in order to explain and predict the patterns of ICTs uptake and usage, and their impact on European households. The analysis is limited at this point, since wave I provides only cross-sectional data, and panel data will be available only after wave II. Panel surveys directly measure behavioural change at the level of the individual sample member and thus supply information that cannot be obtained in a cross- sectional survey. Based on cross-sectional data it could be more difficult to explain the causal links between ownership, use and impact of ICTs. We, therefore, do not attempt to explain the causality between variables at this stage. We do, however, try to estimate the impact of different variables (especially demographics) on ICTs take up and use. The words “impact” and “predictor” in the text do not necessarily infer causality. Another estimation difficulty posed by cross-sections is the inability to include network and habit formation effects (see Liebowitz and Margolis 1998 for explanation of network effects). Present ownership level of ICTs is strongly affected by the size of the network of users in previous periods, whereas cross-sectional data lack information about the past. The use of ICTs could be affected by habit formation (Dynan 2000), so the current amount of use is expected to be correlated with past usage levels, which is missing in cross-sectional data. We uses descriptive analysis to explain the major relationships between variables in the wave I dataset. For each ICT, mainly mobile phone and Internet, we attempt to find predictors of ownership, and use, guided by previous research. After finding such predictors, multivariate analysis is used to arrive at a general linear/logistic regression model enabling us to estimate the relative importance of each predictor. We use the pooled sample to arrive at a general explanatory model for each ICT, and then use the same explanatory variables to analyze each of the six countries separately. Pooling all six countries together add variation of the sample, and increase the sample size. However, the impact of various variables could be different across countries, so separate regressions are needed for country comparisons. The e-Living survey is described in the next paragraph, followed by some details of the main demographic variables used in our analysis. We then analyze the uptake of mobile phones and the diffusion of Internet users in the sample. Paragraph five describes the patterns of mobile phones use and Internet use in the six countries comprising the e-Living sample. In the last paragraph we bring a preliminary analysis of some of the impacts of ICTs use. 2 E-Living Survey Description The e-Living survey is a household panel survey carried out in six countries: UK, Norway, Germany, Italy, Bulgaria and Israel1. The aim of wave 1, conducted in October to December 2001 was to recruit a representative sample of roughly 1750 households within each country by computer assisted telephone interviewing (CATI) in all countries except Bulgaria where telephone penetration was insufficient for this method to be practical. As a result face to face interviewing (CAPI) was used in Bulgaria. In the CATI method random digit dialling was used to select households at random. One adult (aged 16+) in each contacted household was selected using the ‘last birthday method’ and was asked to respond to the

1 For full details see E-Living Deliverable D6: Wave I Documentation and Integrated Dataset. See www.eurescom.de/e-living/index.htm

www.eurescom.de/e-living/index.htm Page 3 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

survey. The questionnaire covered standard socio-demographics, ICT ownership and use as well as details of work, education, skills and social interaction. The survey also contained modules on attitudes to ICTs, the environment, measures of quality of life and some prospective questions on likelihood of purchase of ICTs2. Final response rates are shown in Table 1.

Table 1: e-Living Wave 1 response rates

Final achieved sample (N) Achieved response rates (% of contacts who completed the survey) UK 1760 36% Italy 1762 42% Norway 1753 42% Germany 1756 35% Israel 1750 39% Bulgaria 1753 79% Total 10534 This means that, for example, some 4889 people were contacted in the UK to achieve the sample of 1760 whilst in Bulgaria it was only some 2219. This illustrates not only the higher response rates (and thus less biased sample) that face to face interviewing achieves but also, perhaps, a higher propensity amongst Bulgarians to respond to surveys given that response rates for CAPI in the UK are often as low as 50-60%. Wave 2 of the survey will go into the field in October to December 2002 and will attempt to re-interview all those who responded at wave 1 together with their partner if present. 3 Sample demographics Education level in the e-Living sample is based on the highest qualification attained by respondents. The qualification scale is complicated, and goes from “no qualification” up to a PhD University degree. For statistical analysis purposes we grouped the scale to 5 (Table 2) and 3 levels (primary (or less), secondary and higher education)3. Education is expected to be a strong predictor of ICTs ownership and use, especially with regard to Internet use, which still requires a high level of PC skills.

Table 2: Education level by country (sample)

Country Up to Primary school Secondary school I Secondary Post Secondary University School II

UK 41.5% 19.9% 11.8% 8.5% 18.3%

Italy 16.9% 58.4% 10.5% 1.6% 12.5%

Germany 38.3% 27.1% 10.4% 15.4% 8.9%

Norway 13.3% 33.9% 11.1% 0% 41.7%

Bulgaria 11.6% 37.6% 17% 19.5% 14.3%

Israel 6% 18.2% 29.1% 18.2% 28.5%

2 The UK translation of the questionnaire with routing codes is available at http://www.eurescom.de/e-living/deliverables/e-Living-D4- Wave-1-Questionnaire-FINAL.zip 3 The classification scale may not be the same for every country, and a unified education level measure for all 6 countries has not been developed yet.

www.eurescom.de/e-living/index.htm Page 4 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

Sample monthly family income in Euros is described in

Table 3, including Bulgaria where a different scale was used. We can see that in Norway there is a relatively large group of families with incomes above 4787 Euros. Norway is also the country with the highest per capita GDP. Family income is expected to be an important predictor of ICTs ownership and use.

Table 3: Monthly family income by country sample (Euros)

Country >798 799-1594 1595-3190 3191-4786 4787+

UK 25.1% 17.8% 25.7% 16.7% 14.7%

Italy 28% 36.8% 26.5% 4.3% 4.4%

Germany 9.3% 26.6% 40.4% 15.4% 8.2%

Norway 1.7% 11.5% 23.8% 18% 45.1%

Israel 17.3% 29.1% 27.8% 12.9% 12.8%

<51 Euros 51-101 Euros 102-152 Euros 153-253 Euros 253+ Euros Bulgaria 11.80% 26.40% 21.30% 20.10% 9.50% Other important demographics are: : Gender: Sample data for all countries were weighted according to a method set out in D64, so that it will be closer to population proportions. The proportion of women is slightly higher than men on average – 51.5%. Gender could be a differentiating factor with respect to ICTs ownership and use, in which case it would be important to analyze the formation of gender gaps. : Age: The sample age distribution is similar in all countries with an average age of 45-47 years, except for Israel, where the average is lower – 41 years. Age is an important explanatory variable of diverse consumption patterns of individuals and families in general, and is expected to be a strong predictor of ICTs ownership and use. : Household size: In all countries except Israel the most frequent number of householders is 2. In Israel there is a relatively large share of households with 5 or more dwellers. Family size could affect both ownership and use of ICTs. : Marital status among respondents is similar across countries. 63% of the respondents said that they are married. Ownership and use of ICTs could differ between married and non-married individuals, especially with respect to daily inter-family communications. Work status: The situation here is different across countries, and the proportion of individuals that were working at the time of the survey was between 36% (Bulgaria) to 61% (Norway). Work status is known to be a key explanatory variable of ICTs ownership and use, especially for ICTs that are used both in the homes and in the workplace (mobile phones and PCs). 4 ICT Take-up

4.1 PC & Internet take up Ownership of home PCs and Internet access is all 6 countries are described in Figure 1. Norway leads with 63% PC households, 54% Internet households, and 35% households with broadband5 Internet access. Internet access rates are similar in Israel, UK, Germany, and Italy – 30% to 40%, while broadband households take up rates are still low, with the exception of Germany.

4 E-Living Deliverable D6: Wave I Documentation and Integrated Dataset. See www.eurescom.de/e-living/index.htm 5 Defined as ISDN, cable or ADSL access.

www.eurescom.de/e-living/index.htm Page 5 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

Figure 1: PC ownership and Internet Access in e-Living households

By comparison, the eEurope Benchmarking Report for 2002 states that Internet penetration rate in the EU reached 38% in December 2001. In the US 56% of households owned a PC, and 50% had Internet access in September 2001 (A nation online, 2002). Buying intentions of home PC among non-PC households are within the range of 5% to 10%, although this measure is known to be an over estimation of eventual real actions. Buying intentions of Internet access among non-Internet households are similar in magnitude – 2% to 8%.

Figure 2: PC ownership and Internet Access by income level (excluding Bulgaria)

PC ownership and Internet access, including broadband access, are strongly dependent on family monthly income levels, as shown in Figure 2. PC ownership increases from 17% in households with monthly income of 800 Euros, to 83% among households with monthly income of 4800 Euros and above. The same trend is also apparent in Bulgaria, where PC ownership and Internet access climb from zero for low incomes (less than 51 Euros) to 26% and 13% respectively for higher family incomes (more than 254 Euros). A similar pattern can be found with respect to the education level of respondents (Figure 3). PC ownership and Internet access increases from 29% and 18% (respectively) for householders with primary school education to 72% and 58% for householders with higher education degree. This pattern is also found in the Bulgarian sample, where PC ownership increases from 3% (Internet access – 2%) for householders with primary education to 18% (Internet access – 10%) for householders with higher education. Similar patterns are found in the US by recent studies, “A Nation Online 2002”, and “The UCLA Internet Report 2001”.

www.eurescom.de/e-living/index.htm Page 6 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

Figure 3: PC ownership and Internet Access by education level

Figure 4describes the main location where householders access the Internet. The general pattern is similar with the exception of Bulgaria. In most countries between 52% and 60% of Internet users are home users, except for Bulgaria (15%). The workplace (or college) is used to access the Internet by 29% to 42% of users in all 6 countries. Accessing the Internet in public libraries and cyber cafes is not a common practice in most countries with the exception of Bulgaria. In Bulgaria these public places are used to access the Internet by 43% of all Internet users. This pattern of behavior in Bulgaria could be attributed to several factors, among which are: • The high costs of PC & Internet equipment and access (in relation to the incomes) • The technological infrastructure of the public telephone network is relatively unreliable for Internet use (usually the Cyber Cafes maintain dedicated high-speed leased lines vs. the slow dial-up connections which regular user can afford at home) • The Internet service is rather new, attracting at the moment only a fairly small percentage of the population, mainly early adopters groups of young persons with limited financial means.

Figure 4: Place where Internet is used the most

www.eurescom.de/e-living/index.htm Page 7 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

4.2 Mobile phones take up Mobile ownership rates in the e-Living country samples range from 77% (Norway) to 9% (Bulgaria). Israel, UK and Italy follow Norway with 68% to 71%, and Germany with 61%. Mobile ownership is also strongly related to demographics, such as income and age (Figure 5, Table 4). In Bulgaria mobile ownership in the 55+ age groups is negligible, mainly due to relatively low family incomes and high ownership and use prices. The coefficients in Table 4 measure the effect of a unit increase in the independent variable on the log odds of the dependent variable. The negative impact of age on mobile ownership is very significant in all countries, as also observed in Figure 5. Household’s income6 has a positive impact on mobile ownership in all countries, except in Italy and in Norway. Gender is significant in five countries, where woman ownership rate is smaller than men (the UK is the exception). The amount of Interest in new technology (scale) is a positive predictor of mobile ownership in five countries (except in Israel). Mobile phones are a necessity in the workplace, and in many cases are simply given to employees by their employers (work status is a partial measure of this effect). Work status is a significant predictor of mobile ownership in four countries (Germany is the exception). Education level has a positive significant effect only in Italy, Germany and Israel. Household size has a negative significant effect only in Israel. This may be an indication for the presence of young children, or older persons, who may not own a mobile phone. Individuals in larger families could also share mobile phones for economical reasons. The number of close friends is positively related to mobile ownership only in Israel.

Figure 5: Mobile ownership by age for each country

We can see, that in the pooled sample most of the explanatory variables have significant effects on mobile ownership. Pooling the 6 countries together results in increased range of variation, especially in differentiating variables, such as household’s size and the number of close friends. The relatively low explanatory power could be partially attributed to the lack of “network effect” (large number of users increase probability of purchase), which could only be measured in time series.

Table 4: Mobile ownership (0,1) logistic regression results (Statistically significant beta values, p < 0.05)

Variable / Country UK Italy Germany Norway Bulgaria Israel All

Age -0.037 -0.029 -0.042 -0.033 -0.046 -0.039 -0.025 Household size -0.157 -0.057 Gender (0,1=woman) -0.402 -0.457 -0.871 -0.610 -0.625 -0.208

6 We used continuous family income variable generated from categories mid-points. www.eurescom.de/e-living/index.htm Page 8 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

High school (0,1) 1.104 0.634 Higher education (0,1) 0.998 0.547 1.052 Number of close friends 0.029 Work status (0,1) 0.713 0.835 0.333 1.237 0.799 0.611 Household income 0.083 0.256 5.565 0.107 0.228 Interest in new technology (scale) 0.286 0.396 0.142 0.155 0.220 0.366 Frequency meeting friends 0.151 0.174 0.109 0.043 Constant 1.723 -0.774 Nagelkerke R2 0.310 0.264 0.313 0.203 0.317 0.293 0.337

4.3 Internet diffusion Respondents in all six countries were asked for how many years they have been using the Internet. Transforming years of Internet experience to the first year of Internet use enabled us to build diffusion curves of Internet users. By pooling all six countries together, we are able to analyze the pattern of Internet users (all users, including home users) diffusion, as depicted by Figure 6. The dots are the real net and cumulative additions of Internet users in all six countries (not weighted for different population sizes). The smooth curves approximate the real data by using a logistic function, allowing it to go foreword in time to year 2006. We can see that according to diffusion theory (Rogers 1995), we are already past most of the life cycle phases (innovators, early adopters, early majority and late majority), entering into the “laggards” phase. We can also see that the sample diffusion rate approaches a saturation level of 47%7, where the sample proportion of respondents that use PC is 55%.

Figure 6: Internet users diffusion (pooled sample)

When we compare individuals according to their Internet experience we find that “innovators” and “early adopters” (aggregated together as on group) belong to higher income households, and have much higher personal incomes than individuals from the “early majority” and “late majority” segments (monthly personal incomes of 4,250 Euros, 3,200 Euros, and 2,700 Euros respectively). Another demographic variable that distinguishes between diffusion segments is the level of education/qualification, where 56% of “innovators and early adopters” have a University degree, compared to 21% in the “late majority” segment. Innovators

7 The results reflect average levels for the pooled sample, not weighted for differences in population sizes. www.eurescom.de/e-living/index.htm Page 9 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc and early adopters differ from the other segments with respect to leisure activities and attitudes towards ICTs. They are involved in sproting activities more often, are less intimidated by computers, and find the Internet more usefull than individuals in other segments.

Figure 7: Internet users diffusion by country

The diffusion of Internet users in all six countries is shown in Figure 7. Norway exhibits the fastest diffusion rate approaching a saturation level of about 65% to 70%, followed by the UK,Germany, Israel and Italy with saturation levels of 40% to 55%. The fast diffusion rate in Norway could be partially explained by relatively higher income levels, and higher proportion of individuals with a University degree. It is interesting to note, that in all 6 countries there was a sharp drop in the yearly additional Internet users in years 2000 and 20018. After the fast growth in 1998 and 1999 the momentum weakened in most countries as the Internet “bubble” bursted in years 2000 and 2001.

Figure 8: Internet additional users per year

PC skills are one of the major barriers to Internet penetration. Figure 6 and Figure 9 show clearly, that Internet use is correlated with PC use (see Figuera 1999 for similar findings). In fact, the correlation between Internet use (0,1) and PC use (0,1) is 0.723 (!). The Internet might be considered as an “experience good”, so Individuals who have no PC skills cannot really experience the Internet.

8 As the e-Living survey was conducted during September and December 2001 the figures for 2001 may be under-estimated. www.eurescom.de/e-living/index.htm Page 10 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

Figure 9: Internet users and PC skills

PC use is strongly dependent on demographics, such as age, education, and income. 82% of respondents aged 16-24 are PC users, compared to only 14% in age group 65+. PC use among University graduates is 80%, compare to only 35% among primary school graduates. It is therefore expected that Internet use would also be strongly linked to demographics. To estimate the combined effects of PC skills and demographics on Internet use, we performed a logistic regression, described in Table 5.

Table 5: Internet personal use (0,1) logistic regression results (Statistically significant beta values, p < 0.05)

Variable / Country UK Italy Germany Norway Bulgaria Israel All

PC use (0,1) 3.232 3.609 2.472 5.057 2.506 3.658 Age -0.043 -0.046 -0.062 -0.036 -0.061 -0.036 -0.042 Household size -0.193 -0.137 -0.236 -0.212 Gender (0-man, 1=woman) -0.375 -0.389 -0.224 High school (0,1) 0.690 Higher education (0,1) 1.870 2.204 0.947 1.094 0.823 Marital status (0,1) -0.505 Work status (0,1) 0.545 0.544 Household income 0.120 0.143 0.359 0.147 0.140 Interest in new technology (scale) 0.280 0.384 0.3044 0.245 0.663 0.308 0.341 Constant -2.545 -3.276 -3.288 -2.404 Nagelkerke R2 0.691 0.668 0.607 0.698 0.781 0.532 0.688 PC use (“Do you personally ever use a computer”?) is by far the strongest predictor of Internet use in all countries with the exception of Bulgaria. The beta values for PC use are very large, implying that acquiring PC skills increase the odds of Internet use dramatically. In other words, the main barrier to Internet use is lack of PC skills. Age, education and the degree of personal interest in new technology, also have significant effect on the probability of Internet use. Internet use declines with age in all countries, but increases with the level of education (with stronger impact for higher education), and with the interest in new technologies. Household’s income is a positive predictor of Internet use in all countries excluding Norway and Bulgaria. Household’s size has significant negative impact on Internet use in Italy, Germany and Israel. In larger households there www.eurescom.de/e-living/index.htm Page 11 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc are probably younger children, that don’t use the Internet yet. Gender (0-men, 1-woman) has a negative effect on Internet use only in Germany and in Israel. 66% of non-PC users (30% of total sample) claim that they are not willing to buy a PC because they have no need or use for it. Some of the characteristics of PC “averse” individuals are: : 72% of them are couples or singles. : 61% earn less than 1,600 Euros per household per month. : 57% of them don’t own a mobile phone. : 53% of them retired from work (72% are not working). : 49% of them are above 65 years old. Similar findings are shown in the US (A Nation Online 2002), where Internet non-users are from low-income families and low level of overall education. Possible reasons for choosing to be a non-user in the US include: high costs of ownership and use, lack of confidentiality and content concerns.

The exact reasoning behind the unwillingness to use PCs among 30% of the e-Living sample is not mentioned explicitly. We can, therefore only suggest possible measures that may convert non-Internet users into users. One such measure is the use of “disruptive technologies”9 that will enable a much-improved human interface for the PC (voice activation, for example).

Such “disruptive technologies” could be developed for other digital appliances, such as digital TV and mobile phone. Some researchers are skeptical about the ability of alternative digital appliances to convert Internet non-users into users. The introduction of user-friendly terminals to non-PC users did not reduce the complexity or unfamiliarity with the Internet (Lelong and Beaudouin, 2001).

Digital TV is starting to penetrate Europe (30% of households in the UK, 15% in Norway), and Israel (30% of households). However, digital TV is not perceived, at present, as an alternative Internet access mode, and TV Internet use is negligible (0%-3%). However, once Internet applications could be easily accessed through digital TV set top boxes by using the remote control, Internet access may also find a receptive audience in this “PC averse” segment.

More individuals in the e-Living sample use mobile phones to access the Internet, 10% to 12% in Norway, Germany, and UK, and 9% in Israel. However, mobile Internet use is not the main access mode, which still is the home PC. Mobile phones, especially 3rd generation phones, may also help in activating latent demand for Internet use in the “PC averse” segment, provided that potential users could enjoy the use of real user- friendly interfaces.

Additionally, new contents and applications might also be needed in order to bring “PC averse” individuals into the information society. Most of the contents and applications on the Internet today are developed for the younger generations, since they are the heaviest Internet users today.

5 ICTs use This paragraph describes the patterns of mobile phones and Internet use in the e-Living sample.

5.1 Mobile use Figure 10 describes the daily use of mobile phones (number of calls and messages) across countries. Israel leads in voice calls with an average of 7.7 calls per day, followed by Bulgaria (4.9), and other countries with 3 to 4 calls per day. Norway and the UK lead with respect to the number of SMS messages send per day (2.8). In Israel, 67% of mobile phones owners never use SMS, compared to 25% - 35% in other countries. One possible reason is that Israel has 4 cellular providers, each with a different cellular technology (GSM, TDMA, CDMA, iDEN), which until recently did not allow for an exchange of messages between them.

9 Products based on disruptive technologies are typically cheaper, simpler, smaller, and, frequently, more convenient to use. Harvard’s professor Clayton M. Christensen coined the term. www.eurescom.de/e-living/index.htm Page 12 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

Figure 10: Mobile daily use per country

Explaining the differences in the patterns of mobile use across countries is not an easy task. There are many influencing variables, some of which were not present in the e-Living questionnaire. The following are general relationships that apply for the pooled e-Living sample: : Voice calls and SMS messages seem to be complementary rather than substitutes (correlation of 0.12510). : Age is a powerful predictor of the number of voice calls (-0.138), and number of text messages (- 0.367). In Israel the average age in the e-Living sample was much lower than for any other country (41 years compared to 45 on average). : Household size and number of closed friends have a positive “network” effect on mobile use (correlations of 0.145 and 0.138 for voice, lower for text messages). In Israel household’s sizes are larger than in other countries (4.2 persons compared to 3.2 on average), and the number of close friends is also higher (9 compared to 6.6 on average). : Education/qualification level has a positive effect on the number of voice calls (0.137), and negative effect on the number of text messages (-0.068). : Mobile calls and SMS use is positively affected by household income level (0.167 and 0.091 respectively). : Work status (0,1 variable) positively affects the number of calls (0.166), and negatively affects the number of text messages (-0.073). Linear regression of the number of voice calls and text messages gives the results shown in Table 6. Table 6: Mobile voice use (number of calls per day) regression results (Statistically significant standardized beta values, p < 0.05)

Variable / Country UK Italy Germany Norway Bulgaria Israel All

Age -0.113 -0.116 -0.133 -0.083 Household size 0.089 Gender (0-man, 1=woman) -0.162 -0.121 -0.148 -0.248 -0.142 -0.164 High school (0,1) 0.070 Higher education (0,1) 0.122 Number of close friends 0.096 0.158 0.123

10 All Pearson correlations are significant at the 0.01 level (2-tailed). www.eurescom.de/e-living/index.htm Page 13 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

Work status (0,1) 0.165 0.148 0.293 0.078 0.098 Household income 0.168 0.230 0.301 0.167 0.085 Frequency of meeting friends 0.114 0.094 0.040 Daily fixed phone use (minutes) 0.084 0.143 0.332 0.046 Adjusted R2 0.105 0.126 0.155 0.118 0.192 0.133 0.112 Gender and household’s income have significant impact on mobile voice use in most countries (Bulgaria and Norway are the exception). Women mobile voice use is lower than that of men (5.4 calls per day compared to 3 calls per day on average in all six countries), and gender gap is significant in five countries. A positive income effect is consistent with our expectation, and its magnitude differs across countries (higher in Italy and Germany, lower in the UK and Israel). Work status has a positive impact on voice use in Italy, Norway, Bulgaria and Israel. Employed individuals may use their mobile phones for work purposes too, and therefore make more calls than unemployed individuals. Age impact mobile voice use negatively in the UK, Norway and in Israel. Younger persons spend much more time using their mobile phones, than older people. Mobile and fixed phones are used as complementary rather than substitute services in the UK, Germany and in Bulgaria (more on this in paragraph 5.3). Mobile voice use is positively related to the number of close friends (UK, Israel), and to the frequency of meeting friends (Italy, Norway). The impact of close friends is part of a more general “network effect”, which is positively related to mobile phones ownership and use. Against our expectation, education level and household size do not have a significant impact on voice use in any country, except for the pooled sample. The low explanatory power could be attributed to missing variables (tariffs for example), and to the linearity imposed on the model. In many cases mobile phones are given to employees in the workplace, and their use is free of charge, at least up to a certain limit. In such a case family income effect might be under estimated. Higher mobile voice use in Israel could be partially explained by a younger population (41 years old on average, compared with 45 in the e-Living sample), larger households11 (4.2 compared to 3.2 in the e-Living sample), and larger networks of close friends (9 persons compared to 6.6 in the e-Living sample). Similar findings are described in a detailed study of ICTs use and social networks in Europe (Smoreda and Thomas 2001), and in WP6. Table 7: Mobile SMS use (number of text messages per day) regression results (Statistically significant standardized beta values, p < 0.05)

Variable / Country UK Italy Germany Norway Israel All

Age -0.266 -0.314 -0.277 -0.349 -0.157 -0.277 Household size 0.107 0.151 Higher education (0,1) -0.069 Number of close friends 0.109 Marital status (0,1) -0.180 -0.158 -0.191 -0.185 -0.130 -0.151 Work status (0,1) -0.123 -0.068 -0.157 -0.059 Household income 0.103 0.093 0.083 0.118 Frequency of meeting friends 0.072 0.091 0.060 Daily fixed phone use (minutes) 0.086 Adjusted R2 0.157 0.214 0.181 0.210 0.132 0.155 Table 7 describes the regression explaining mobile messaging use. Age has the strongest negative impact on mobile messages use (Figure 11) in all five countries12, as documented in many studies (see WP6, for example). Another variable with negative impact in all five countries is marital status. Married people use mobile messaging considerably less compared to unmarried people, which is compatible with highest use in younger ages. We have positive income effects on mobile messaging use in the UK, Italy and in Norway. Work status has negative impact on messaging use in Italy, Norway and Israel. In those countries, we saw earlier that voice use is positively related to work status. This means that employed persons may prefer voice conversation to text messaging as a mean of communications. Other variables are significant only in

11 Although in Israel household’s size was not significant, it is significant in the pooled sample. 12 Bulgaria was excluded because regression results were not significant. www.eurescom.de/e-living/index.htm Page 14 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc one or two countries, such as household’s size (UK and Israel), the number of close friends (Germany) and fixed phone use (UK).

Figure 11: Mobile phone use by age group

5.2 Internet use Internet home use measures include characterization of Internet applications, and time spent on the Internet per day. Figure 12 describes the average time spent using the Internet at home in the e-Living sample, excluding Bulgaria. Israel leads with 102 and 93 minutes for men and women respectively, while in other countries use levels are similar for men (48 to 69 minutes), but different for women (32 to 58 minutes). In Bulgaria the number of users is small (37 persons in the sample). The average daily Internet use is fairly high – 77 minutes for men and 66 for women.

Figure 12: Internet home daily use per country (minutes)

The level of Internet home daily use may depend on many variables, and the lack of both theoretical and empirical models on this topic makes it difficult to achieve a reasonable explanatory power.

www.eurescom.de/e-living/index.htm Page 15 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

Table 8 describes the results of a linear regression for daily Internet use (minutes). The only variables with significant impact on Internet minutes of use in all three countries are daily time spent watching TV and downloading music online. Internet use is positively related to the time spent watching TV, which may be explained by personal characteristics of householders (heavy TV users are also heavy Internet users). This does not mean that there is no displacement effect of TV by the Internet, as will be shown later. Downloading music may take a long time without broadband Internet access, so it might increase the overall time spent on the Internet. In Italy and in Norway there is a significant gender gap in Internet minutes of daily use, as also shown in figure 12. Employed persons spend less time on the Internet than unemployed persons in Norway, since they have the opportunity to access the Internet at their workplace. The frequency of playing sports (proxy for outdoor leisure activities) has a negative impact on time spent on the Internet in Norway. Engaging in online education activity also increase the time spent on the Internet in the UK and in Italy. Other variables that are significant only in one of the countries are age (Norway), education (UK), marital status (Norway), and income (Italy).

Table 8: Internet daily use (minutes) regression results13 (Statistically significant standardized beta values, p < 0.05)

Variable / Country UK Italy Norway All

Age -0.117 -0.72 Household size 0.097 Gender (0-man, 1=woman) -0.156 -0.149 -0.080 Higher education -0.184 Marital status (0,1) -0.104 Work status (0,1) -0.086 -0.054 Household income -0.150 -0.078 Frequency of playing sports -0.073 -0.079 Daily time spent watching TV 0.190 0.140 0.143 0.167 Online shopping (0,1) 0.044 Online education (0,1) 0.097 0.145 0.057 Online music download (0,1) 0.121 0.194 0.153 0.128 Adjusted R2 0.152 0.121 0.141 0.136

Internet daily use decreases by age group in most countries (Figure 13), although significant age impact was found only in Norway. In Israel and the UK, Internet use has a “U” shape, where younger and older people are heavier users than middle-aged people. In Israel daily use drops from more than 110 minutes in the 16- 24 age group to less than 80 minutes in the 35-44 age group, and then rises to more than 90 minutes for the 65+ age group. By comparison, In the US, the average time spent on the Internet was 9.8 hours per week, 73 minutes per day (UCLA Internet report 2001).

13 The results for Germany, Bulgaria and Israel are not shown due to unsatisfactory results. www.eurescom.de/e-living/index.htm Page 16 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

Figure 13: Internet home daily use by age group

Table 9 describes the most important home Internet uses by country. Educational and informational use is the most important use for the majority of householders in all 6 countries (29% - 44%). Other important uses are Email (17%-22%), web surfing (10%-33%), and work (9%-26%). Shopping and banking is an important use for smaller populations, especially in Norway (21%), Germany (13%), and the UK (9%).

Figure 14: Online uses by experience

Some Internet uses increase in popularity with experience, as Figure 14 shows (travel, banking and shopping). With time, online users tend to be more confident and trustful, especially in uses that require monetary payments. Average online spending (asked for the last 3 months) tends to increase with experience, from around 200 Euros to more than 700 Euros. These findings are supported by Horrigan and Rainie (2002) that are reporting significant increase in Email and other activities, including online transactions, for Americans that are gaining more Internet experience. The UK leads in online spending with 678 Euros, followed by Israel (587 Euros), Norway (558 Euros), Germany (467 Euros), and Italy (356 Euros). Online shoppers don’t seem to favor a particular product category although there are country differences. Books are most popular in Germany (57% of online shoppers), travel in the UK (37%), and clothing in Germany (40%). Average spending differs per product category with travel tickets in the lead (999 Euros), followed by events tickets (748 Euros), computer hardware/software (729 Euros), and CDs (729 Euros). This is a partial explanation why the UK leads in average online spending. The mean online spending for the whole sample is still a small fraction of the monthly personal income (about 3%).

www.eurescom.de/e-living/index.htm Page 17 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

Table 9: Most important home Internet uses (Percent of Internet users)

UK Italy Germany Norway Bulgaria Israel Work 12% 12% 9% 10% 26% 10% Email 22% 17% 20% 18% 16% 21% Games 2% 2% 1% 1% 3% 3% Shopping 9% 2% 13% 21% 0% 4% Education 31% 36% 44% 32% 42% 29% Surf Web 24% 32% 14% 18% 10% 33%

5.3 Internet & mobile: complementary or substitutes It is of interest to explore weather the 2 major ICTs, namely Internet and mobile phone, is used as complementary or substitute products. In the e-Living sample, not all of interviewed persons use both the Internet and a mobile phone - 59% in Norway, 45% in the UK, 39% in Germany, 36% in Israel, and 33% in Italy. The correlations between the daily uses (minutes) of the Internet, mobile and fixed phones are described in Table 10. All correlations are positive (some are not significantly different from zero), which points to complementary relationships. The positive correlations between fixed and mobile phone use was already demonstrated by a number of studies, see Cadima and Pita Barros 2000, for example. The data also points to strong correlation between mobile phone and Internet uses in all countries with the exception of Italy. Correlations cannot be used to explain causality, and are rough indicators of association between variables. A strong association between Internet (especially Email) and mobile uses (including messaging) could imply that the use of both ICTs increase the ability and the tendency to communicate with friends, family members and work colleagues (see Figure 15 for Email and SMS relationships). Causal links between the two could only be found after wave II results are analyzed.

Table 10: Correlations between Internet, mobile and fixed phones daily use (minutes)

Mobile-Internet Mobile-Fixed Internet-Fixed

UK 0.193 ** 0.148 ** 0.201 ** Italy 0.097 0.079 ** 0.125 * Germany 0.165 ** 0.163 ** 0.074 Norway 0.229 ** 0.006 0.041 Bulgaria - 0.194 - 0.13 0.343+ Israel 0.181 ** 0.049 0.195 ** ** Correlation is significant at the 0.01 level (2-tailed). * Correlation is significant at the 0.05 level (2-tailed). + Correlation is significant at the 0.1 level (2-tailed). The positive relations between the uses of various ICTs by householders within the framework of social networks and social capital are described in Smoreda and Thomas (2001) and e-Living D7.4 (Section 3.3 “ICTs and social capital”).

www.eurescom.de/e-living/index.htm Page 18 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

Figure 15: The relationship between SMS and Email use

6 ICTs impact

6.1 TV watching and Internet use There are a number of studies that claim to have found media displacement effects introduced by Internet use, specifically that Internet use reduces time spent watching TV (UCLA Internet report 2001, Clemente 1998, Nie and Erbring 2000). These studies compare Internet users and non-users in cross-sections, from which it is difficult to draw causality conclusions (Gershuny 2002). Anderson and Tracey (2001) use longitudinal time-diary data from the British Home Online study to show that getting Internet access makes very little difference to patterns of individual time use. Rather, micro social transitions such as changes in employment or household status/type have far more significant effects and are often occurring simultaneously with gaining Internet access. Instead of simple comparison, it is more efficient to study the relationships between TV watching and Internet use in a multivariate framework. The causal links between the two will be further revealed after wave II results are analyzed. The e-Living data shows that the time people spent watching TV is significantly lower for Internet users in all countries (183 minutes compared with 136 minutes per day in the UK, for example). However, there are other variables, which might affect TV watching, such as work status, education, and income. A linear regression of the time spent on TV watching per day is described in table 11.

Table 11: Daily TV watching (minutes) linear regression

(Statistically significant standardized beta values, p < 0.05) Variable / Country UK Italy Germany Norway Bulgaria Israel All

Age 0.104 -0.066 -0.076 0.116 -0.051 Household size -0.084 -0.073 -0.087 -0.037 High school (0,1) -0.150 -0.116 -0.105 -0.064 Higher education (0,1) -0.253 -0.189 -0.157 -0.161 -0.157 -0.125 Work status (0,1) -0.245 -0.166 -0.121 -0.154 -0.173 -0.102 -0.148 Household income -0.073 -0.120 0.071 -0.063 Internet use (0,1) -0.100 -0.164 -0.113 -0.114 -0.152 Mobile ownership (0,1) -0.063 Frequency of playing sports -0.081 -0.094 -0.140 -0.061

www.eurescom.de/e-living/index.htm Page 19 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

Frequency of reading books 0.283 0.088 Interest in new technology (scale) -0.080 0.053 0.097 0.087 0.043 Adjusted R2 0.201 0.117 0.105 0.133 0.110 0.082 0.104

Internet use has a negative impact on TV watching in Italy, Norway, Bulgaria and Israel. Education and work status both have negative impact on TV watching in almost all countries. Household’s size also has a negative but smaller impact on TV watching in three countries. Age’s impact is smaller than expected, positive in Germany and Israel, negative in Norway and Bulgaria. Other relevant variables are leisure activities, such as frequency of playing sports, and interest in new technologies. Figure 16 shows that those who have more Internet experience tend to spend less time watching TV and more time using their home PCs, with a crossover for Internet experience of more than six years. The figure also shows that the time spent on the Internet is relatively constant and does not climb up with Internet experience. The time spent talking over the fixed phone seems to decrease slightly with Internet experience. There seems to be a trend of time displacement in the data, but it should be reviewed again in wave II. As already mentioned, the study of causal relationship of time displacement is most efficient in Panel studies.

Figure 16: Time allocation to TV watching and PC use (Pooled sample)

6.2 Effects of Email use Householders in the e-Living sample report that as a result of using Email, they write fewer letters and use the fixed phone less. Figure 17shows, that as Internet experience increases, a larger share of Email Users in the sample report writing fewer letters and using the fixed phone less.

www.eurescom.de/e-living/index.htm Page 20 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

Figure 17: Effects of Email use on writing letters and using fixed phone

Actual behavior shows that in most countries, the average daily time spent on using a fixed phone is higher for individuals who use Email, compared to non-users (except for Norway). However, Email users who said that they have reduced fixed phone use do spend on average less time on the fixed phone than those who did not mention it (30 minutes per day compared with 35 minutes per day).

Table 12 shows regression results of fixed phone use in all six countries. Fixed phone use is probably more affected by habit formation behaviour, and may be better explained in time series, or panel studies, rather than in cross-sections. We can see that Email use has a negative impact on fixed phone use only in Bulgaria, which is also the country with the highest goodness of fit (0.343). In most countries, gender is the variable with the highest impact on fixed phone use (women fixed phone use is larger than men’s). Age is negatively affecting fixed phone use in the UK, Germany and Israel, and has a positive impact in Bulgaria. The number of close friends has a positive impact on fixed phone use in the UK, Bulgaria, and Israel. Other relevant variables are frequency of reading books, and of going to activity groups.

Table 12: Fixed phone use (minutes) linear regression (Statistically significant standardized beta values, p < 0.05)

Variable / Country UK Italy Germany Norway Bulgaria Israel All Age -0.125 -0.110 0.239 -0.135 Household size 0.045 Gender (0-man, 1-woman) 0.231 0.107 0.191 0.111 0.211 0.144 Marital status (0,1) -0.106 Work status (0,1) 0.153 -0.114 -0.223 -0.080 Household income 0.134 0.173 Email use (0,1) -0.178 Number of close friends 0.101 0.418 0.085 0.059 Frequency of activity groups 0.098 0.095 0.197 0.059

www.eurescom.de/e-living/index.htm Page 21 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

Frequency of reading books -0.073 0.081 Fixed or mobile (0, 1-mobile) -0.031 Adjusted R2 0.081 0.021 0.058 0.029 0.3343 0.110 0.035

6.3 Quality of life and Internet use Quality of life was measured in the e-Living project by the level of agreement to the statement “Overall the conditions of my life are excellent”. The average measure of quality of life is similar for Internet users and non-users in most countries, as described in Figure 18. In the UK, Israel and Bulgaria, perceived quality of life is somewhat larger for Internet users. There is also a very slight increase in perceived quality of life between ICTs usage groups, where for users of both mobile phone and Internet the average agreement scale is 4.0 (1 – strongly disagree, 5 – strongly agree), and for non-users it is 3.8. Similar findings are described in the UCLA Internet report (2001): “Compared to non-users, Internet users report slightly lower levels of interaction anxiety, powerlessness, loneliness, alienation, and lack of guiding social norms. Internet users express slightly higher levels of life satisfaction”.

Figure 18: Perceived quality of life and Internet use

Many variables might be involved in explaining perceived quality of life. A detailed regression analysis of quality of life in all 6 countries is described in WP6. The conclusion in WP6 is that “there are few, if any, reliable associations between ICT access and usage and overall quality of life”. 7 Summary and conclusions One of the main tasks of WP3 is to use sample demographics, as well as other variables, from the e-Living data in order to explain and predict the patterns of ICTs uptake and usage, and their impact on European households. The analysis is limited at this point, since wave I provides only cross-sectional data, and panel data will be available only after wave II. Panel surveys directly measure behavioural change at the level of the individual sample member and thus supply information that cannot be obtained in a cross- sectional survey. Based on cross-sectional data it could be more difficult to explain the causal links between ownership, use and impact of ICTs. We, therefore, do not attempt to explain the causality between variables at this stage. We do, however, try to estimate the impact of different variables (especially demographics) on ICTs take up and use. The words “impact” and “predictor” in the text do not necessarily infer causality. Another estimation difficulty posed by cross-sections is the inability to include network and habit formation effects. Present ownership level of ICTs is strongly affected by the size of the network of users in previous periods, whereas cross-sectional data lack information about the past. The use of ICTs could be affected by habit formation, so the current amount of use is expected to be correlated with past usage levels, which is missing in cross-sectional data. www.eurescom.de/e-living/index.htm Page 22 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

7.1 ICTs take up PC and Internet take up is led by Norway (63% and 54% respectively), followed by Israel (60%, 40%), UK (52%, 39%), Germany (52%, 35%), Italy (40%, 30%), and Bulgaria (5%, 3%). By comparison, in the US 56% of households owned a PC, and 50% had Internet access in September 2001 (A Nation Online, 2002). Both PC ownership, and Internet access, is positively affected by family income level and by personal education level. The home is the most popular location for Internet access in most countries (50% to 60% of users) with the exception of Bulgaria (15%), followed by the workplace/college (30% to 40%). Public libraries and cyber cafes are popular Internet access locations especially in Bulgaria, where access and usage costs are still high, and PSTN infrastructure still inadequate for home use. Mobile ownership rates in the e-Living country samples range from 77% (Norway) to 9% (Bulgaria). Mobile ownership is also strongly related to demographics, such as age, gender and family income level. Ownership rates are over 80% for 16-24 years old, going down to 15%-30% for 75+ years old (excluding Bulgaria). The impact’s magnitude of most variables differs across countries. Income effect is largest in Bulgaria and smallest in Norway. Education level is a positive predictor only in Germany, Bulgaria and Israel. Work status is a positive predictor in the UK, Italy, Bulgaria (highest impact) and Israel, perhaps because in these countries more companies give mobile phones to employees. The patterns of Internet users diffusion in the e-Living pooled sample were analyzed with respect to existing diffusion models. The data show that according to diffusion theory (Rogers 1995), we are already past most of the life cycle phases (innovators, early adopters, early majority and late majority), entering into the “laggards” phase. The data also show that the sample diffusion rate approaches a saturation level of 47%14, where the sample proportion of respondents that use PC is 55%. In fact, the correlation between Internet use (0,1) and PC use (0,1) is 0.723 (!). IT seems, therefore, that PC skills are the major barrier to Internet use today. Norway exhibits the fastest diffusion rate approaching a saturation level of about 65% to 70%, followed by the UK,Germany, Israel and Italy with saturation levels of 40% to 55%. The fast diffusion rate in Norway could be partially explained by relatively higher income levels, and higher proportion of individuals with a University degree. It is interesting to note, that in all 6 countries there was a sharp drop in the yearly additional Internet users in years 2000 and 2001. After the “hype” in 1999, created by over optimistic expectation, the Internet “bubble” bursted in 2000 and 2001 in almost perfect synchronicity in all six countries. PC use (“Do you personally ever use a computer”?) is by far the strongest predictor of Internet use (measured by 0,1 variable) in all countries with the exception of Bulgaria. The logistic regression beta values for PC use are very large, implying that acquiring PC skills will increase the odds of Internet use dramatically. In other words, the main barrier to Internet use is lack of PC skills. Demographics also play key role in explainig Internet personal use. Age (negatively), family income level and personal education level (positively), are also strongly affecting Internet use. Another variable with a positive effect on Internet use in all six countries is the degree of interest in new technology (a scale measure). A gender gap appear to be significant only in Germany and in Israel. 66% of non-PC users (30% of total sample) claim that they are not willing to buy a PC because they have no need or use for it. 72% of them are couples or singles. 61% earn less than 1,600 Euros per household per month. 57% of them don’t own a mobile phone. 53% of them retired from work (72% are not working). 49% of them are above 65 years old. These PC “averse” individuals should be encouraged to join the information society by: • Teaching them PC skills. Success here may take longer than expected due to the demographic characteristics of this segment. • Developing new contents. In order to make the Internet more attractive to this segment, new contents and applications might be needed to persuade individuals to experience the Internet. • Develop new and friendly interfaces for PC, that wouldn’t require downloading and installing files. The PC should be “transparent” to individuals in this segment. • Encourage them to use other digital appliances, which may be user-friendlier, such as digital TV and mobile phones. Digital TV is not perceived, at present, as an alternative Internet access mode, and TV

14 Not weighted for difference in country population sizes. www.eurescom.de/e-living/index.htm Page 23 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

Internet use is negligible (0%-3%). The use of mobile phones to access the Internet is becoming more prevalent in the e-Living sample (around 10% of mobile owners), but not as the main access mode.

7.2 ICTs use Mobile use differ across countries, where Israel leads with an average close to 8 voice calls per person per day, but lags in SMS use with only 1.3 messages per day. Norway, UK, Italy and Germany has more similar mobile use patterns with 2.7 to 3.9 voice calls per day, and 2 to 2.8 text messages per day. Israel suffers from the existence of 4 different cellular technologies, and only lately users were able to send messages across networks. Voice calls and text messages are affected differently by similar variables, except maybe to household’s income level. Age seems to be the strongest predictor of SMS use in all six countries. SMS use drops rapidly with age, so other demographics must be “compatible” with it. Marital status, for example, has a negative effect on SMS use in all six countries, but almost none in voice use. Work status is a positive predictor of voice use, but negative predictor of SMS use in Italy Norway and Israel. Higher mobile use in Israel could be partially explained by a younger population (41 years old on average, compared with 45 in the e-Living sample) and larger networks of close friends (9 persons compared to 6.6). The number of close friends is a significant predictor of mobile voice use only in Israel and in the UK. Internet home use in terms of time spent per day is similar in UK, Italy, Germany and Norway, with 48-69 minutes on average for men and 32 to 58 for Women. In Israel Internet use is significantly higher with 102 minutes for men and 93 for women. The level of Internet home daily use may depend on many variables, and the lack of both theoretical and empirical models on this topic makes it difficult to achieve a reasonable explanatory power. Linear regressions (excluding Bulgaria and Israel for poor results) show that the daily time spent on the Internet is affected negatively by gender (Italy, Norway) and work status (UK, Norway). There is a positive relationship between Internet and TV time use in all three countries. Online uses, such as education and music downloading positively affect the time spent on the Internet. The effect of age differs per country. In Israel and in the UK there is a “U” shape and use drops at the younger age groups and climbs up again in the older age groups. Employed persons spent less time on the Internet than unemployed persons (UK, Norway), probably because they can do so in the workplace. Internet experience (number of years) does not seem to affect the average daily time spent on the Internet in the pooled sample (around 60 minutes per day per user). Educational and informational use is the most important Internet application for the majority of householders in all six countries (29% - 44%). Other important applications are Email (16%-22%), web surfing (10%-33%), and work (9%-26%). Shopping and banking are important uses for smaller populations, especially in Norway (21%), Germany (13%), and the UK (9%). Some Internet applications increase in popularity with experience (travel, banking and shopping). With time, online users tend to be more confident and trustful, especially in applications that require monetary payments. Average online spending (asked for the last 3 months) tends to increase with experience, from around 200 Euros for users with less than a year of experience, to more than 700 Euros for users with seven years of experience. Average spending differs per product category with travel tickets in the lead (999 Euros), followed by events tickets (748 Euros), computer hardware/software (729 Euros), and CDs (729 Euros). The UK leads in average online spending with 678 Euros, followed by Israel (587 Euros), Norway (558 Euros), Germany (467 Euros), and Italy (356 Euros). Buying travel tickets online is most popular in the UK (37% of online user), and that is why the UK leads in average online spending. The mean online spending for the whole sample is still a small fraction of the monthly personal income (about 3%). ICTs are used as complementary services rather than substitutes in all countries. Fixed and mobile phone uses are positively correlated in most countries (Norway and Israel are the exceptions). Mobile and Internet uses are also positively correlated in all countries but Italy. A strong association between Internet (especially Email) and mobile uses (including messaging) could imply that the use of both ICTs increase the ability and the tendency to communicate with friends, family members and work colleagues. The positive relations between the uses of various ICTs by householders within the framework of social networks and social capital are described in Smoreda and Thomas (2001) and WP6 (section 3.3 “ICTs and social capital”). The associations of Internet experience with other key variables evident in wave I dataset may give us preliminary clues to some important trends. These trends will become evident only after analysing wave II data. The following are some examples of the important issues waiting for wave II to be resolved:

www.eurescom.de/e-living/index.htm Page 24 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

• The average daily use of the Internet: will it remain relatively constant at about 60 minutes? • Internet uses mix: are we going to see a rise in online shopping and spending? • Mobile phone uses mix: will the rate of growth of SMS exeed that of voice? • Co-evolution of Email and SMS: will the groing use of SMS displace or enhance Email use?

7.3 The Impact of ICTs The impact of ICTs use is a complex issue, and was studied only with respect to TV watching and fixed phone use. ICTs Impacts research will be more efficient once wave II results are compared to wave I, since it should easier to find causality links between uses and impact. There are a number of studies that claim to have found media displacement effects introduced by Internet use, specifically that Internet use reduces time spent watching TV (UCLA Internet report 2001, Clemente 1998, Nie and Erbring 2000). These studies compare Internet users and non-users in cross-sections, from which it is difficult to draw causality conclusions (Gershuny 2002). Instead of simple comparison, it is more efficient to study the relationships between TV watching and Internet use in a multivariate framework. The causal links between the two will be further revealed after wave II results are analyzed. Internet use (measured by 0,1 variable) has a negative impact on TV watching in Italy, Norway, Bulgaria and Israel. The pooled e-Living data show that individuals, who have more Internet experience, tend to spend less time watching TV and more time using their home PCs, with a crossover for Internet experience of more than six years. The data also shows that the time spent on the Internet is relatively constant and does not climb up with Internet experience. The time spent talking over the fixed phone seems to decrease slightly with Internet experience. There seems to be a trend of time displacement in the data, but it should be reviewed again in wave II, after the panel dataset is established. Householders in the e-Living sample report that as a result of using Email, they write fewer letters and use the fixed phone less. Actual data shows, that Email users who said that they have reduced fixed phone use do spend on average less time on the fixed phone than those who did not mention it (30 minutes per day compared with 35 minutes per day). Fixed phone use is probably more affected by habit formation behaviour, and may be better explained in time series, or panel studies, rather than in cross-sections. Regression of fixed phone use show that Email use has a negative impact on fixed phone use only in Bulgaria. In most countries, gender is the variable with the highest impact on fixed phone use (women fixed phone use is larger than men’s).

8 Bibliography Anderson, B., and Tracey, K. 2001 “Digital living; the impact (or otherwise) of the Internet on everyday life.” American behavioral scientist 45, 456-475. Cadima, N., and P. Pita Barros 2001 “The Impact of mobile phone diffusion on the fixed-link network”, CEPR Discussion Paper 2598. Clemente, P. C., 1998 "State of the Net" The New Frontier", New York, McGraw-Hill. Commission of the European communities 2002, eEurope Benchmarking Report, COM(2202) 62 final, Brussels. Dynan, K. E., 2000 “Habit Formation in Consumer Preferences: Evidence from Panel Data”, The American Economic Review Vol. 90, No. 3. Figuera, G. 1999 "An Analysis of International Internet Diffusion" 1999, MIT http://rpcp.mit.edu/Pubs/Theses/gonzalo.pdf) Gershuny, J., 2002 “Web-Use and Net-Nerds: A Neo-Functionalist Analysis of the Impact of Information Technology in the Home”, ISER working paper 2002-01. www.iser.essex.ac.uk/pubs/workpaps/2002-01.php Horrigan, J.B., and L. Rainie 2002 “Getting Serious Online”, Pew Internet & American Life Project, 2002, www.pewinternet.org/reports/toc.asp?Report=55. Lelong, B and Beaudouin, V.2001 “Usages d’Internet, Nouveaux Terminaux et Hauts Debits: Premier Bilan après Quatre Années d’Expérimentations”. Paper for the conference ‘e-Usages’, Paris, 12-14th June.

www.eurescom.de/e-living/index.htm Page 25 of 26 PUBLIC e-living-D7.1-ICT-Uptake-and-Usage-Issue-1.0.doc

Leibowitz, S. and Margolis, S. 1998 “Network Externality" entry in “The New Palgraves Dictionary of Economics and the Law”, MacMillan, http://wwwpub.utdallas.edu/~liebowit/palgrave/network.html Nie, N. H., and L. Erbring, 2000, “Internet Society: A Preliminary Report”. Stanford Institute for the Quantitative Study of Society (SIQSS), Stanford University. NTIA, 1995, 1998, 1999, 2000, "Falling Through the Net, Americans in the Information Age",(www.ntia.doc.gov/ntiahome/digitaldivide/). OECD, 1998, "Information Technology Outlook 1997", Paris. OECD, 2002, “Information Technology Outlook 2002”, Paris. Smoreda, Z., and F. Thomas 2001 “Social networks and residential ICT adoption and use” EURESCOM P903 study "Cross-cultural attitudes to ICT in everyday life". ] http://www.eurescom.de/~ftproot/web-deliverables/public/P900-series/P903/ICT_use_Smoreda.pdf UCLA Center for Communications Policy 2001, The UCLA Internet Report 2001: Surveying the Digital Future. U.S. Department Of Commerce 2002, A Nation Online: How Americans Are Expanding Their Use of the Internet.

www.eurescom.de/e-living/index.htm Page 26 of 26

PUBLIC e-Living-D7.2-Environmental-Impact-Issue-1.1

e-Living D7.2 Environmental Impact of ISTs: A Cross- Sectional Analysis

Alberto Pasquini (Legambiente)

Marisa Velardo (Univ. of Rome "La Sapienza")

Lorenzo Vicario (Legambiente)

e-Living: Life in a Digital Europe, an EU Fifth Framework Project [IST-2000-25409] www.eurescom.de/e-living/index.htm

e-Living-D7.2-Environmental-Impact-Issue-1.1

Table of Contents

1 Introduction 3 2 Identification of the aspects under investigation 3 3 First Aspect: Information and Sensitivity to Environmental issues 4 3.1 Measurement of Environmental Sensitivity 4 3.2 Internet Usage and Environmental Sensitivity 5 4 Second Aspect: Pressure on the Environment due to the use of ICT Devices 8 4.1 The "Electronic Waste" 8 5 Third Aspect: ICT and family lifestyle 10 5.1 Dematerialisation 11 5.2 Decoupling of Time and Space 17 6 Conclusions 19 7 Bibliography 19

www.eurescom.de/e-living/index.htm Page 2 of 1

e-Living-D7.2-Environmental-Impact-Issue-1.1

1 Introduction

There is clear evidence that Information and Communication Technology (ICT) is transforming our society into an Information Society. That is, a society where the access to information and information services, accumulated knowledge and learning opportunities is fast, cheap and efficient and can be done without any significant spatial and temporal constraints. This modification is having a strong influence on human behaviour and the changes induced can have a significant effect on the environment (Greiner 1996, Jokinen 1998). Since, the environmental issues are one of the central aspects in our future we need a better understanding of this relationship between the development of the ICT and the environment.

Recent studies (ASIS 2000, Jokinen 1998) evidence that this relationship is complex and contradictory and that one should avoid simple generalisations. The same studies emphasise that theoretical approaches cannot be adequate to understand this relationship and that empirical investigations are strongly needed.

Existing investigations can be roughly divided into theoretical estimations and forecasting such as (Mitomo 1999), (Lieshout 2000), (Telecommunication Council 1998), and case studies such as those described in (Forseback 2000). Both consider the phenomena having a major impact on the environment, such as telework, and their main indicators, and most of them measure the impact, of ICT on the environment, in terms of energy consumption and related carbon dioxide emissions. Most of the estimates and forecasting are very optimistic and provide neither the underlying quantification model nor the assumptions adopted. Case studies are usually based on small national samples and sometimes have an optimistic interpretation of the results. In addition, in most of the cases, these studies extend and generalise the results obtained without considering the rebound effect as described in Anderson (2001, Section 3).

This study represents a contribution to the state of the art based on the analysis of the first wave of the e- living survey. The study focused on some specific aspects of the relationship between the development of the ICT and the environment that can be considered as emblematic examples of this relationship. The next section describes how these aspects were derived, while the remaining three sections are dedicated to the analysis of the aspects.

2 Identification of the aspects under investigation

This study adopts the "pressure-state-response" model proposed by the OECD in (OECD 1998). This model does not consider the specific role of new technologies in the relation between humans and environment. However, it provides a framework to understand the complex interrelations between environmental pressures, environmental conditions and societal responses and it has been adopted in some studies proposing indicators explicitly designed to investigate the influence of the Information Society on the Environment (Heinonen 2001). A simplified version of this model is presented in Figure 1.

The model is based on the assumption that the human activity (a in Figure 1) exercises pressures on the environment (b) and that these pressures cause some modification to the state of the environment (d). Society (f) receive information (e) about these modifications or information (c) about the possible effect of the human activity, and reacts with policies, regulations, and behavioural changes (g).

ICT may influence directly the human activity, and then the related pressure on the environment, and the quality and extent of the information. For this reason we identified the following aspects, to be investigated and monitored through the survey:

1. the relation between Information (provided through Internet) and sensitivity to environmental problems (links identified with c and e in the figure);

2. the direct pressure on the environment due to the use of ICT devices (a contributor to b in figure);

3. the modifications in the human activity pressure due to the new lifestyles and opportunities offered by ICT (another contributor to b in figure);

www.eurescom.de/e-living/index.htm Page 3 of 1

e-Living-D7.2-Environmental-Impact-Issue-1.1

(b) Pressure: pollution, use of resources, etc. (e) Information

(a) Human (d) State of (f) Society activity Environment

(g) Response: decisions, actions, etc.

Figure 1: Simplified "Pressure-State-Response" Model

The first point considers the influence that the availability of information provided through Internet can have on the environmental sensitivity of the family members, and is investigated in Section 3.

The second point regards the direct pressure on the environment due to the new Information and Communication Technology (ICT) devices used within the families. This is a direct effect that does not require the household members to have any specific intention to reduce their environmental impact and not even a change in their lifestyle. It is just a consequence of buying, using and disposing of an ICT product. Examples include: the increased electric power consumption due to the use of the new ICT products; the saving in energy due to the adoption of intelligent control systems in plants such as air conditioning systems and the increased contribution of the ICT devices to the waste electrical and electronic equipment. This aspect is investigated in Section 4.

The third point concerns the modification of the human activity pressure due to new lifestyles and behaviours induced by the ICT. In some cases this may require to household members a specific intention to reduce their environmental impact, but in general it is just the results of exploiting the opportunities offered by the ICT. For example, thanks to the ICT and under favourable conditions, employed family members can have the opportunity to Telework, changing their working habit and reducing their daily amount of travel and the related environmental impact. This aspect is investigated in Section 5.

3 First Aspect: Information and Sensitivity to Environmental issues

The first issue under investigation is related to the link between human activity and the impact on society of the «pressure-state-respond» scheme. The human activity considered in our inquiry is the usage of ICTs and, more in detail the internet access, under the hypothesis that access to more information related to the environment could modify the environmental sensitivity of the internet users and thus have an impact on society. Clearly a cross-sectional data set cannot prove that this causality occurs but it can enable us to assess whether or not those with internet access also have higher environmental sensitivity.

3.1 Measurement of Environmental Sensitivity

In order to evaluate environmental sensitivity we have inserted in the survey a set of 3 questions with which we constructed a scale of sensitivity. The 3 questions were chosen taking into consideration different aspects: the necessity to avoid and eliminate the influence of different national conditions (e.g. supply of

www.eurescom.de/e-living/index.htm Page 4 of 1

e-Living-D7.2-Environmental-Impact-Issue-1.1 recycling structures for waste disposal), the need to consider the way in which environment sensitivity is manifested (way of acting, way of thinking, behaviours).

Thus the answers to the questions, consisting in a statement, were arranged in order, the possible answers being "I strongly agree" through to "I strongly disagree" on a standard Likert scale. The statements evaluated (questions) were:

· "I try to buy organic food or food with low level of pesticides."

· "We worry to much about the future of the environment and not enough about prices and jobs today."

We have also asked if the person interviewed is a member of an environmental or animal welfare organisation.

Having taken into consideration all answers we created a 4 step scale of environmental sensitivity. The steps are:

1. no sensitivity (disagreement to the two sentences and no membership);

2. low sensitivity (only a positive answer);

3. medium sensitivity ( two positive answers);

4. strong sensitivity (agreement on the two sentences and member of an environmental organisation);

Figure 2 shows results ordered by country. It can be noticed that the level of environmental sensitivity is similar in the 6 countries, but two peculiarities are evidenced: the absence of a category of strong sensitivity in Bulgaria where only 3 out of 1,750 interviewed were members of environmental organisations. The Italian category of No sensitivity is less ample, only 23,1% of the Italians interviewed belong to this class, while in other countries the class ranges from 31 to 34,7%of the population interviewed. This last difference may be due to a larger attention to the quality and origin of food shown by Italian consumers.

40 No sensitivity Low sensitivity 30 Medium sensitivity High sensitivity 20

0 UK Italy Germany Norway Bulgaria Israel

Figure 2: Environmental sensitivity by country

3.2 Internet Usage and Environmental Sensitivity

The relation between the environmental sensitivity, classified using the four classes described above, the penetration of the ICT and the usage of Internet is shown in figures 3 and 4. The first of these figures

www.eurescom.de/e-living/index.htm Page 5 of 1

e-Living-D7.2-Environmental-Impact-Issue-1.1 describes the environmental sensitivity by the number of working computers of the household, showing a clear positive relation between the two.

40 No sensitivity Low sensitivity 30 Medium sensitivity High sensitivity 20

0 0 1 2 3 ³ 4

Figure 3: Environmental sensitivity by number of working computers owned

Figure 4 shows the relation between the access and usage of Internet and the environmental sensitivity and, also in this case, the positive relation between the two is clear. However, environmental sensitivity, penetration of ICT and access to Internet are also clearly related to the level of education and, in a certain measure, to income. It is difficult to say which is the primary relation between these variables. In other words, the current data do not show clearly if Internet penetration and the related access to information promote a higher environmental sensitivity or if both environmental sensitivity and Internet penetration are just a consequence of higher education and income. This will be the subject of future research.

40 No sensitivity Low sensitivity 30 Medium sensitivity High sensitivity 20

0 No access to Internet Access to Internet

Figure 4: Environmental sensitivity by access to Internet

We have tried to verify the hypothesis that the usage of internet is a way to access to more information and to have a strong environmental awareness. The first inquired aspect, on the relation between the usage of Internet and the environmental sensitivity, is the direct use to Internet in order to obtain environmental www.eurescom.de/e-living/index.htm Page 6 of 1

e-Living-D7.2-Environmental-Impact-Issue-1.1 information. The results of this question are shown in Figure 5 and they show that Italian Internet users are currently the most likely to use the net to access environmental information.

100 %

Yes 10

0 No UK Italy Germany Norway Israel

Figure 5: Percentage of internet users who said they had used the net to obtain information about the environment in last three months. Bulgaria excluded due to small n for internet use.

The relation between the environmental sensitivity and the usage of the Internet to obtain environmental information is clearly shown in Figure 6. As we would expect, those internet users with the strongest environmental sensitivity are those who are most likely to have used the net to access environmental information. However it is also interesting to note that even some of those with no environmental sensitivity at all had accessed this kind of information.

50,0 %

44,5 40,0

30,0

20,0 21,3

14,9

10,0 10,8

0,0 None Low Medium Strong

Environmental sensitivity

Figure 6: Percentage of internet users that researched environmental information on it shared between the class of environmental sensitivity.

www.eurescom.de/e-living/index.htm Page 7 of 1

e-Living-D7.2-Environmental-Impact-Issue-1.1 Ling et al (2002) Have shown that overall quality of life is influenced by the perception of the environmental conditions. In particular, good opinions about the environmental conditions, of the respondents living area, are associated with higher scores of their quality of life. However, the overall quality of life has neither a clear association with the access or usage of ICT, nor with the environmental sensitivity.

4 Second Aspect: Pressure on the Environment due to the use of ICT Devices

This Section investigates the direct pressure on the environment due to the new Information and Communication Technology (ICT) devices used within the families. Most applications of the ICT have an impact on environment, for example the one due to the electric power consumption. The most important of these applications, and their influence on the environment are summarised in Table 1.

Table 1: Pressures on the environment due to ICT applications

ICT APPLICATION DESCRIPTION DIRECT EFFECT OF THE APPLICATION Intelligent Plant Plants optimising the performances of Reduction of energy consumption Control Systems house systems such as cooling, heating, (in terms of electric power, lighting systems gasoline, etc.) Any ICT device Any ICT device used by the family needing Increased electric power electric power consumption due to the ICT devices Any ICT device Any ICT obsolete or not working ICT device Increased Electronic Waste needing disposal Intelligent appliances Microprocessors optimising the process of Reduction of resources control systems house appliances such as washing consumption (e.g. water) machines

These applications and their potential final effect on the environment have been studied by several authors, special attention has been given to the possible increases in power consumption. Data concerning the effect on power consumption due to ICT devices and to the spread of Internet have been reported by a series of authors (Laitner, 1999; Laitner, 2000; Mitchell-Jackson, 2001; Romm, 1999). The referenced literature seems to indicate that the additional power needed by these devices is not significant, and, in any case, much lower than the potential savings introduced by the Information Society, even if more comprehensive analyses are still needed. Less attention has been given to the significant impact due to the disposal of electronic devices.

4.1 The "Electronic Waste"

The disposal of ICT devices is becoming an important problem especially because of the continuous reduction of the life span of computers. Some authors claim that because of the advances in chip technology, this lifespan has been reduced from perhaps 4-5 years to approaching 2 years or less (Kiuchi 2001). Our survey did not investigate directly this issue, however the analysis of the data has provided some interesting hints. The amount of computers that have been thrown away by household members, shown in Figure 7, seems to be very low if compared with the PC and Internet take up and with their diffusion over the years, described in Raban et al (2002). Norway is the country were more computers have been thrown away or are no more in use, as shown in Figure 8. However, even in this country of "early adopters", computer owners have thrown away less then one computer, in average, in the last five years.

There are several possible reasons for this discrepancy between data concerning acquisition, lifecycle duration and disposal of computers. Some authors (Matthews, 1997; Cooper, 2000), have evidenced that the real lifecycle of computers can be much longer than expected because computers are recycled through commercial channels, donated to less demanding users or kept as possible back up of the new systems. According to other authors (Kiuchi 2001) several computers are stored in back rooms and offices because people are unwilling or reluctant to discard goods that are perceived as still valuable. These findings have www.eurescom.de/e-living/index.htm Page 8 of 1

e-Living-D7.2-Environmental-Impact-Issue-1.1 been confirmed by the data collected in the E-Scope project (Cooper 2000), although these data concern the UK only. However, the storing capacity of attics, garages, warehouses and children is not infinite and we can expect that an increasing number of these equipments will enter soon the waste stream (Kiuchi 2001). Concerns are also related to the possible presence of materials such as lead, mercury, flame retardants and plastic softeners (Kiuchi 2001). This problem will soon be regulated by two European Directives which were agreed in conciliation negotiations between the European Parliament and Council of Ministers in October 2002. These are the Waste Electrical and Electronic Equipment (WEEE) and Restriction on the use of certain Hazardous Substances in electrical and electronic equipment (RoHS) Directives. They will have to be implemented by Member States by 2005. The Directives were drafted for the following reasons: § The amount of WEEE in the European Community is expanding rapidly. It has been is acknowledged as the fastest growing waste stream. This is despite the apparent reluctance of households to throw away computer equipment, as indicated in our findings. § Recycling of WEEE is not undertaken to a sufficient extent. § The content of hazardous components in electrical and electronic equipment (EEE) is a major concern particularly during the waste management phase. One of the additional aims of the WEEE Directive is to encourage producers to design and produce electrical/electronic equipment which increasingly takes into account; repair, upgrade, re-use, disassembly and recycling as end of life options.

In accordance with the requirements of these two directives producers will have legal and financial responsibility for taking back, collection, treatment and recycling of electrical and electronic equipment and to substitute various substances such as heavy metals and brominated flame-retardants.

80 %

0 0 1 2 3 4 5 >5

Number of computers

Figure 7: Percentage of computer owners who have thrown away (or are not using anymore) one or more computers in the last five years

www.eurescom.de/e-living/index.htm Page 9 of 1

e-Living-D7.2-Environmental-Impact-Issue-1.1

,5 Average of number PCs ,4

0,0 UK Italy Germany Norway Bulgaria Israel

Figure 8: Average number of computers thrown away (or are not used anymore) by computer owners in the last five years, by country

5 Third Aspect: ICT and family lifestyle

This Section deals with the question: "do the new lifestyle of the family members, induced by the penetration of the ICT, modify the pressure of the human activity on the environment?". Several applications of the ICT can have an influence on the behaviour and attitudes of the family members. The most important of these applications, the basic modifications induced by the applications, the related new lifestyle of the family members, and the consequent impact on the environment are summarised in Table 2.

Table 2

ICT APPLICATION BASIC MODIFICATIONS NEW LIFESTYLE OF THE POTENTIAL BENEFITS FOR THE DUE TO THE FAMILY MEMBERS ENVIRONMENT APPLICATION Internet Dematerialisation Exchanging computer Alternative communication applications readable information channel, reducing paper waste including E-mail through telecommunication circuits Electronic Dematerialisation, Accessing magazines, Reducing the consumption of publishing, Decoupling of time books, and newspapers paper in manufacturing and newspapers, and place through electronic media distribution, reducing paper telephone waste, reducing transportation directories to access libraries, reducing the energy required to arrange and maintain books Electronic music Dematerialisation Recording, transferring Alternative music recording and playing music, through reducing the need for row Internet and electronic material and the energy of the media (e.g. MP3), instead production process of CDs

www.eurescom.de/e-living/index.htm Page 10 of 1

e-Living-D7.2-Environmental-Impact-Issue-1.1

Digital photo Dematerialisation Taking, transferring and Alternative picture recording storing pictures with reducing the need for row electronic media instead of material, the energy of the paper production process and the release of chemicals Telework Decoupling of time Different work style based Alternative access for work, and place on the use of info- reducing the need for communications in order to transportation, reducing the eliminate the need to necessity of increasing the size commute of buildings and promoting the decentralisation of cities On-line shopping Decoupling of time Shopping from home Alternative modality for and place through Internet for various shopping, reducing the need for products at various shops transportation and paper consumption by mail-order catalogues Correspondence Decoupling of time Learning from home Alternative access to schools, courses and place through Internet having the reducing the need for chance to access even transportation and increasing remote institutions the efficiency of school equipment On-line Decoupling of time Making reservations at Alternative access for making reservations and place home on the Internet for reservation, reducing the need tickets for rides, plays, and for transportation concerts

These applications and their potential final effect on the environment have been studied by several authors, in most of the cases with a speculative approach. A detailed analysis of the work done in dematerialisation and decoupling of time and place can be found in Anderson et al, 2001).

5.1 Dematerialisation

Dematerialisation consists in the use of new services and goods performing the same functions as the old ones, but with a reduced use of material and energy. Examples are the new personal computers needing fewer material than the old mainframe or the use of electronic services (such as the Telephone Directories on Internet) instead of goods offering the same services (such as the hard copies of the Telephone Directories). Approximately 470.000 tons of telephone books are discarded each year in the United States, and only a marginal 10 per cent of the paper is recycled (Cohen, 1999). The main functions of the telephone directory (data storage and retrieval) are performed far more efficiently on line than on paper. Telephone books go out of date each year, while electronic versions can be up-dated continuously, offer various search capabilities and have no associated printing and distributing costs.

However, innovations introduced by the technology, are mainly directed towards increasing total production rather than saving material and energy. Thus, the total use of material resources may increase more than the savings because of the rebound effect. The rebound effect is an economic and social phenomenon reducing the potential environmental benefits of innovation. The resource savings from innovations of specific products and services are eaten up by extending the consumption of the new, resource effective, products and services, usually because of their lower price. For example it is true that personal computers cost much less and need fewer material than the old mainframe, but there is a much wider diffusion of personal computers if compared with main frames. Another reason of rebound effect is the diffusion of other patterns of resource consumption which originate mainly in a change in consumer behaviour. For example the increase of paper use due to the ease and low costs of processing, storing and printing texts with the support of personal computers. In addition, quite often the new services do not replace completely the old ones, but are used additionally, helping to extend the limits which the old technology would have eventually hit.

www.eurescom.de/e-living/index.htm Page 11 of 1

e-Living-D7.2-Environmental-Impact-Issue-1.1 The degree of replacement depends on several social and economical factors including the ability of the new service to offer all the features provided by the old one, the usability of the related instrument and of the service itself, and the cost.

We could not investigate all the cases of dematerialisation listed in Table 2 within the timescale of this work. Then, we have chosen two representative examples of ICT and Internet enabled services that could replace tangible good and services: the use of digital music (e.g. MP3) with the related substitution of conventional discs and tapes and the usage of Emails replacing conventional mails. These examples have been chosen because: in both cases the adoption of the new service/technology is significant in all the countries of the survey excluding Bulgaria; they evidence very well how the degree of replacement is influenced by several factors, including the functionality and usability of the new service/technology.

5.1.1 Dematerialisation and digital music

Figure 9 shows the percentage of Internet users that downloaded music at least once in the last three months by country. The adoption of this new service/technology seems to be strongly related to the age of the users, as shown in Figure 10, but not to income or education as shown by the regression analysis in Table 3. This is not surprising because the Internet users are relatively homogenous in terms of income and education as described in Raban et al (2002).

50 %

40 42

30 32 30 30 26

0 UK Italy Germany Norway Israel

Figure 9: Percentage of Internet users that downloaded music at least once in the last three months, by country (excluding Bulgaria)

However, the adoption of digital music brings several new possibilities such as transferring or sharing pieces through Internet, but has also some negative points including lower quality and some usability problems. Then, the different, non overlapping features of the two services (conventional and digital music) does not favour a complete replacement of the first with the second, as shown in Figure 11. This figure shows the amount of discs or tapes that the respondents have not bought in the last three months because replaced by music downloaded in the same period. The amount of discs and tapes replaced in three months seems very low. This is also confirmed by the data of Figure 12, that shows the percentage of Internet users who downloaded music in the last three months and those who bought discs or tapes through Internet in the same period of time: "downloaders" are not buying less CDs or tapes than the others.

www.eurescom.de/e-living/index.htm Page 12 of 1

e-Living-D7.2-Environmental-Impact-Issue-1.1

40 % 30

0 16-24 25-34 35-44 45-54 55-64 65-74

Age

Figure 10: Percentage of Internet users who downloaded music in the last 3 months, by age

Table 3: Logistic regression of the effects of being male, age and income on likelihood of downloading music from the net (educational level was also controlled but not reported. Values = logistic coeficients, base sample = internet users in all countries. Bulgaria is excluded due to small n for internet use)

UK Italy Germany Norway Israel Age -0,0679 -0,0754 -0,081 -0,0507 -0,0266 Household income in euro (contrast is > 4786) 798 to 1594 0,9868 1596 to 3190 0,7917 -1,1565 3192 to 4786 0,7752 -1,2689

Male 1,1926 0,7161 0,6322

Constant 2,3179 1,4857

Nagelkerke R2 0,212 0,210 0,230 0,136 0,099

The logistic coefficient can be interpreted as the change in the log odds associated with a one-unit change in the independent variable. The only indicator which is significant in all countries is age and that education is not statistically significant in any countries. Age is a significant indicator with older people being less likely to be a music "downloader". Other variables have a significant effect on downloading music only in some countries. It's interesting to note that income has a positive impact (the odds are increased ) on downloading music in UK whereas it has a negative impact in Italy. The coefficient for male is the change in log odds when we have male compared to female : the log odds of being a music "downloader" increase by 1,1926 ; 0,7161 and 0,6322 for UK, Norway and Israel. However it should be noted that Nagelkerke R square values is low with the models for Israel so these results should be viewed with some caution.

www.eurescom.de/e-living/index.htm Page 13 of 1

e-Living-D7.2-Environmental-Impact-Issue-1.1

32 %

0 Don't know 0 1 2 3 4 5 >5

number of discs and tapes

Figure11: Percentage of internet users by number of discs or tapes not bought in the last three months, because replaced with music downloaded through Internet

90 %

80 79 70 69 60

30 31 CDs or tapes bought

20 Yes 21

10 No Yes No

downloaded music

Figure 12: Percentage of internet users who bought discs or tapes through Internet in the last three months, by the two groups of those downloading and not downloading music

www.eurescom.de/e-living/index.htm Page 14 of 1

e-Living-D7.2-Environmental-Impact-Issue-1.1 5.1.2 Email substitution

The frequency of Emails usage is shown in Figure 13, expressed as the average number of Emails sent per day divided by country. Again Bulgaria is not included because of the very low number of respondents using Emails. The Email usage is related to the age of the users, as shown in Figure 14, but not to income or education. Unlike the digital music, the service provided by Emails seems to cover most of the functionalities and features offered by conventional letters, with additional benefits such as prompt delivery and low cost. Then, the degree of replacement tends to be higher as shown in Figure 15, where respondents have been asked if they are writing fewer letters than before because of the use of Emails.

50 %

Number of e-mails 20 Negligible frequency

Less than 1 a day 10

From 1 to 5 a day

0 More than 5 a day UK Italy Germany Norway Israel

Figure 13: Percentage of internet users who use Email and frequency of use by country (ex. Bulgaria)

50 %

Number of e-mails 20

Negligible frequency

Less than 1 a day 10

From 1 to 5 a day

0 More than 5 a day 16-24 25-34 35-44 45-54 55-64 65-74

Figure 14: Percentage of internet users who use Email and frequency of use by age.

www.eurescom.de/e-living/index.htm Page 15 of 1

e-Living-D7.2-Environmental-Impact-Issue-1.1

70 % 67 60 63

50 54

34 30

0 Negligible frequency From 1 to 5 a day Less than 1 a day More than 5 a day

Number of e-mails

Figure 15: Percentage of email users who claim to write less letters because replaced with Emails by frequency of Emails use.

Both the use of this new technology/service and the tendency to use it in lieu of writing letters seem to grow with the familiarity with Internet as shown in Figure 16 and Figure 17. The first figure reports the percentage of those replacing letters with Emails on the basis of the number of years they are using Internet, and the second the frequency of Email usage on the basis of the number of years they are using Internet.

80%

70%

60%

50% UK Italy 40% Germany Norway 30% Israel

20%

% who mentioned writing fewer letters 10%

0% Less than 1 1 year 2 years 3 years 4 years 5 years 6 years 7 + years year Years using the internet

Figure 16: Percentage of email users who claim to write less letters because replaced with Emails by Internet experience. Bulgaria excluded as too few Internet users.

www.eurescom.de/e-living/index.htm Page 16 of 1

e-Living-D7.2-Environmental-Impact-Issue-1.1

40 %

20 Number of e-mails

15 More than 5 a day

10 From 1 to 5 a day

5 Less than 1 a day

0 Negligible frequency <1 1 2 3 4 5 6 >=7

years

Figure 17: Percentage of internet users using Email with the specified frequency by their experience with Internet.

5.2 Decoupling of Time and Space

Decoupling of time and space is a consequence of the increased possibilities of communicating, accessing and transferring information. Various activities such as shopping, learning and working can be carried out from remote locations, like home and isolated offices. The potential effect on the environment of decoupling of time and place will be substantial if there is a reduction of travels. This possible reduction is strongly debated in the literature with different authors proposing different, and sometime quite contradictory results. Some authors suggested significant saving, expressed in terms of number of trips per week (Forseback, 2000) or carbodioxide emission (Mitomo, 1999). Other authors evidence how the savings could be less than the additional traffic generated by the availability of extra free time and of the worker car (Lieshout, 2000). Most of the previous works are based on theoretical estimations and forecasting, However, even our survey cannot provide an effective and precise answer to the issue of the reduction in travel. Brynin et al (2002) report analysis concerning teleworking "per se" and we concentrate here on travelling for commuting with respect to teleworking.

Figure 18 shows the relation between teleworking, expressed in terms of average number of days worked at home during normal working hours, and travels, expressed in terms of average kilometres done to get to work. The related number of kilometres saved by the respondent thanks to teleworking is shown in Figure 19. These data do not consider the possible additional traffic generated by the availability of extra free time for the worker or by the availability of the worker car for use by other household members. However, recent studies suggest that heavy Internet users (greater than five hours a week) spend less time driving, for all purposes, than average. In particular, a US study reported extensively in (Romm, 2002), has shown that home based business workers spend 1.23 hours a day travelling in cars for all purposes, while home based telecommuters spend 1.39 and conventional workers spend 1.61.

www.eurescom.de/e-living/index.htm Page 17 of 1

e-Living-D7.2-Environmental-Impact-Issue-1.1

35 35 33 30 31 Average of Kms 25

20 21

0 Most days Less often A few days a week Never

Figure 18: Average distance of respondents from their offices by frequency of their working activity at home

400

380

300 Average of Kms

200

100 109

0 Most days A few days a week Less often

Frequency of teleworking during weekdays

Figure 19: Average number of kilometres saved by each respondent per week, by frequency of the working activity at home

www.eurescom.de/e-living/index.htm Page 18 of 1

e-Living-D7.2-Environmental-Impact-Issue-1.1 6 Conclusions

With this first wave of the survey we investigated three aspects of the relation between ICT and environment:

1. the relation between Information (provided through Internet) and sensitivity to environmental problems;

2. the direct pressure on the environment due to use of ICT devices;

3. the modifications in the human activity pressure due to the new lifestyles and opportunities offered by ICT;

For the first point data show that sensitivity to environmental issues is related to Internet penetration, that is, to a more easy access to information through Internet. But, environmental sensitivity is also clearly related to the level of education and to income (even if this latter link is a bit less evident). Then it is difficult to say which is the causal factor that originates sensitivity to environmental issues. In other words, the current data do not show clearly if Internet penetration and the related access to information promote a higher environmental sensitivity or if both environmental sensitivity and Internet penetration are just a consequence of higher education and income. The analysis of the data coming form the second wave will help to clarify this point.

As for the direct pressure on the environment due to use of ICT devices there are already consolidated answer in the literature concerning power consumption, then our survey focused on the other main problem of the electronic waste. We would expect this to grow in the future as the large numbers of ‘old’ PCs which appear to be currently being stored/used within households start to emerge into the waste stream. We hope that the second wave of our survey will allow us to estimate more clearly the real length of the lifecycle of computers used in household and when these devices enter the waste stream. However, the whole matter will be soon regulated by two European Directives which were agreed in conciliation negotiations between the European Parliament and Council of Ministers in October 2002 (the Waste Electrical and Electronic Equipment and Restriction on the use of certain Hazardous Substances in electrical and electronic equipment).

The third point was investigated by considering two representative examples: dematerialisation and decoupling of time and space. Dematerialisation is not automatic with the grow of the technology because substitution of an old service/good/technology with a dematerialised one depends on several factors, however in some cases this phenomenon could bring a significant reduction of environmental pressure. Travel saving thanks to the decoupling in time and space acquired through ICT seems also very promising. However even the saving in travel is dependent on several factors. More investigations are needed to understand if additional traffic is generated by the availability of extra free time for the worker or by the availability of the worker car. Again the second wave, and especially the time diary, will help us to clarify better this point.

7 Bibliography

Anderson, B. (ed.), (2001): e-Living: D3 - State of the Art Review. Public e-Living Project Deliverable. www.eurescom.de/e-living/index.htm. ASIS - Alliance for a sustainable information society (2000): Final project report, Spain Brynin, M., Anderson, B., Yttri, B., (2002) e-living D7.5: Homeworking and Teleworking - A Cross-Sectional Analysis. E-Living Project Deliverable. www.eurescom.de/e-living/ Cohen, N., (1999), Greening the Internet – Ten Ways E-commerce could affect the Environment and what We can do. IMP Magazine, www.cisp.org/imp Cooper, T., Mayers, K. (2000): Prospects for household appliances, Technical Report of the E-Scope project, UK. Forseback, L. (2000): Casa Studies of the Information Society and Sustainable Development, Telia (Sweden), translated in English and published by the Information Society Directorate General of the European Commission www.eurescom.de/e-living/index.htm Page 19 of 1

e-Living-D7.2-Environmental-Impact-Issue-1.1 Greiner, C, Raidermaker F. J., Rose T. (1996): Contribution of the Information Society to Sustainable Development - Report of the Working Circle: A DGXIII-Initiated Group on Sustainability and the Information Society. Heinonen, S. (et al.) (2001): The ecological transparency of the information society, Futures, Vol. 33, Finland Jokinen, P. (et al.) (1998): The Environment in an Information Society, Futures, Vol. 30, no. 6, Finland Kiuchi, T., Gable, C., Cassel, S., Shirman, B., (2001): Computers, E-Waste, and product Stewardship: Is California Ready for the Challenge ?, Technical Report of the Global Futures Foundation, San Francisco, USA. Laitner, J. A (1999): The Information and Communication Technology Revolution: Can It Be Good for Both the Economy and the Climate, Environmental Protection agency, Office of Atmospheric Programs, US Laitner, J. A (et al.) (2000): Re-estimating the Annual Energy Outlook 2000 Forecast using Updated Assumptions about the Internet Economy, Proceedings of the Eastern Economic Association Conference, Crystal City, Virginia Lieshout, M. V., Slob, A. F. L. (2000): ICT and Climate Change, TNO, Delft Ling, R., Yttri, B., Anderson, B., Diduca, D., (2002) e-living D7.4: Age, Gender and Social Capital - A Cross- Sectional Analysis. E-Living Project Deliverable. www.eurescom.de/e-living/ Matthews, H. S. (et al.) (1997): Disposition and End-of-Life Options for Personal Computers, Green Design Initiative Technical Report #97-10, Carnegie Mellon University Mitchell-Jackson, J. (2001): Information Technology and Resource Use, report for the project Energy End- Use Forecasting, US Mitomo, H., Oniki, H. (1999): Information Technology for Sustainable Societies-Public Policy Perspectives in Japan: The Case of Telework, IPTS Report, Vol. 32 OECD (1998): Towards Sustainable Development - Environmental Indicators, Paris. Raban, Y., Soffer, T., Mihnev, P., Ganev, K. (2002) e-living D7.1 ICT Uptake and Usage: A Cross-Sectional Analysis. Public e-Living Project Deliverable. www.eurescom.de/e-living/ Romm , J. (et al.) (1999): The Internet Economy and Global Warming – a scenario of the Impact of E- commerce on energy and the Environment, The Center for Energy and Climate Solutions, Global Environment and Technology Foundation, US Romm, J., (2002): The Internet and the New Energy Economy, in Pamlin D. (ed.): Sustainability at the Speed of Light, Technical report of WWF, Sweden. Telecommunication Council (1998): Addressing Global Environmental Preservation Through info- Communications System, Report, Japanese Ministry of Post and Telecommunications

www.eurescom.de/e-living/index.htm Page 20 of 1

PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

e-Living D7.3 – ICT and Socio-Economic Exclusion Dr. John P. Haisken-DeNew, DIW Berlin, IZA Bonn*

*I appreciate very much many helpful comments from Dr. Ben Anderson and Dr. Malcolm Brynin at the University of Essex, Dr. Conchita D’Ambrosio, DIW Berlin and Bocconi University, Milan, Dr. Markus Pannenberg and Dr. Joachim Frick at the DIW Berlin and also Patrick Dross, DIW Berlin for his excellent research assistance.

e-Living: Life in a Digital Europe, an EU Fifth Framework Project [IST-2000-25409] www.eurescom.de/e-living/index.htm PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

Table of Contents 1 Introduction ...... 4 2 ICT Use and Economic Exclusion ...... 4 2.1 Previous Studies ...... 5 2.2 E-Living Application: Average Wage Effects of Computer Use ...... 7 2.3 E-Living Application: Distributional Earnings Effects of Computer Use...... 10 2.3.1 Distributional Counterfactuals...... 10 2.3.2 Quantile Regression ...... 14 3 ICT Use and Social Exclusion...... 17 3.1 Previous Studies ...... 17 3.2 Operationalisation in eLiving...... 18 3.3 Empirical Application and Interpretation ...... 20 3.3.1 Interpreting Exclusion Graphically ...... 20 3.3.2 Computers, Internet and Social Exclusion...... 20 3.3.3 Cell / Mobile Phones and Social Exclusion ...... 21 3.3.4 Social Exclusion Patterns ...... 25 4 Conclusions...... 27 5 Bibliography ...... 32

www.eurescom.de/e-living/index.htm Page 2 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

Executive Summary This report presents analysis of financial and social exclusion with respect to ICTs in five of the six countries in the e-Living survey. Using well-known econometric methods and e-Living’s unique data it shows that in terms of a wage premium: • Controlling for other variables, workers who use PCs at work earn more. The effect is particularly large in Israel (37% higher earnings for PC users) and the UK (27%). • The PC use effects are particularly large at the lower end of the income scale although this distributional pattern varies across the countries. • Workers in the UK who do not use PCs at work are 17% more likely to be ‘economically’ excluded (earn less than 80% of the median wage). This is the highest figure of all five countries. • There are no overall effects for use of the Internet at work except in Norway (overall an 11.5% effect) where the effect is highest at the upper end of the income distribution. There are significant (and large) effects in Israel and Italy. In Germany there are positive effects at both the lowest and at the upper end of the income range. • There are no effects at all for a range of types of use except for highly skilled computer network/management applications for those in the 50th and 75th deciles in the UK. Overall, whilst something is going on in terms of a PC-use at work wage premium, precisely what is going on, and therefore what actions need to be taken is not at all clear. The results suggest that the effects for PC use in general have more to do with the kinds of people who tend to have/use PCs in the work place than the fact of PC usage itself. It also suggests that simply training workers in specific PC related skills will not have much effect on their earnings returns in most countries. However, all things being equal, workers in the lowest earnings band who uses PCs at work have the greatest association with increased earnings although we must caution any assumption that this link is causal. In terms of social exclusion: • Those who do not have a range of ICTs are considerably more likely to be considered socially excluded. There are differences between countries and for different ICTs. Not having a home PC increases the ‘risk’ of being defined as socially excluded in Israel by 18.5% (and by 35% if they say they cannot afford a PC) but decreases it by 1.3% in Germany. Not having access to a mobile phone increases this ‘risk’ by between 2.5% (Germany) and 44% (Israel). Access to the Internet at home does not make much difference to the ‘risk’ of being defined as socially excluded. It increases the risk by 5.9% in Israel but reduces it by 5.1% in Norway. • Having Internet access is not a strong indicator raising the possibility that access is rather more evenly distributed than might be supposed. Home Internet access alone, in contrast to low ownership of PCs at home due to reasons of cost, may therefore not be that significant as a policy problem. In some countries the main barrier to access for all in the ‘European Information Society’ may not be the cost of Internet access but the purchase and ongoing costs of the PCs required to use it. The lack of some ICTs (home PCs and mobile phones) is an excellent indicator of other kinds of social exclusion in some countries. However the patterns are sufficiently different that cross-country conclusions cannot be drawn and thus cross-country actions and policies to address these patterns are not plausible. We can see the value of ICT access and usage as an indicator of social exclusion but not necessarily a cause of it. It may simply be that the reasons for exclusion from ICTs are the same as those for more traditional social exclusion – low education, income, high risk of unemployment, poor social communication, poor health and so forth. Thus focusing on ICT exclusion (the ‘Digital Divide’) alone without trying to reduce the more familiar structural barriers will have little effect on social exclusion in Europe. Overall we have been able to identify possible associations between ICTs, mean wages, the wage distribution and with a social exclusion index. With additional waves of data, we will be able to test the robustness of these initial cross-sectional results, by allowing for controls for individual specific unobserved heterogeneity. It is this analysis which may provide indications of causal relationships between ICT use and wage/social outcomes and thus provide evidence to support significant public investment in ‘ICT access for all’ in the name of social objectives.

www.eurescom.de/e-living/index.htm Page 3 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

1 Introduction With an estimated fortune of $50 Billion, the richest private person in the world in the year 2002 is William H. Gates, the chairman and CEO of Microsoft Corporation, a maker of computer software and computer operating systems. Virtually every person in the world has been touched in some way by his products. In fact many of today’s nouveau-riche (and in more recent times, nouveau-pauvre!) are indeed in the computer industry. The increase in computer productivity has been documented exhaustively. However, how does this increase in computer productivity spill over into increased labour productivity? Perhaps only a certain few individuals have access to the same computer skills required to create computer operating systems as William Gates, and therefore only a certain few gain in the productivity gains, leaving many somehow excluded from the benefits. It seems today that computers have penetrated all aspects of society. Booking flight and rail tickets over the internet, purchasing music through large discount volume sellers and even ordering and paying for goods using a mobile phone have become as commonplace as any other standard technological innovation in the past. It seems almost impossible to live without computers today. This report examines the role of computers, the internet and mobile phones in economic and social exclusion. For instance, those who have computer skills may have better prospects in a labour market geared toward a production technology depending on computers, thereby earning more than those who do not have these skills. Furthermore, there may be some differential effects depending on whether one is very poor or very rich, not just simply an “average” person. Perhaps computers have become so important in society that “being without” implies some particularly bad hardship. In today’s information society, not having access to information means not being informed of current events, access to inexpensive goods, better opportunities in the job market, etc. Perhaps today’s “haves” and “have-nots” can be defined by access to computers, the internet and mobile phones. Firstly, we will examine information technology and its relation to “economic exclusion”. We will present results from earnings regressions which identify the mean effects. Then we examine the distributional effects of information and communication technology (ICT) usage. We can create some counterfactual scenarios to answer the questions: (a) What would an average person’s wage look like if he/she were to use (not use) a PC at the workplace? (b) What would a particularly poor (rich) person’s wage look like if he/she were to use (not use) a PC at the workplace? (c) How would the distribution of labour earnings change if one removed the positive productivity effects of ICT from the distribution? Next we examine the role of ICT on non-monetary aspects of exclusion, namely so-called “social exclusion”. We focus on a host of goods and characteristics which would be beneficial to somebody, such as having a large apartment, owning a computer at home, having a mobile phone or perhaps other consumer durables, being well educated, in good contact with friends and family, etc. From these characteristics, one can construct an index of social exclusion and ask very similar questions: (d) What does the distribution of social exclusion index look like? What percentage of persons are well below the median, i.e. the ones we should be potentially concerned about as social scientists? (e) What would this exclusion index look like, if one did not have access to various ICT items (i.e. creating counterfactual scenarios)? (f) Would the probability of falling into the lower end of the distribution increase (increased risk of social exclusion) and by how much? For each section, we briefly review the relevant literature to provide a benchmark for the analysis. Then results derived directly from the e-Living data, (Raban et al, 2002) are presented and conclusions are drawn.

2 ICT Use and Economic Exclusion In the United States, the 1980s witnessed a widening in earnings inequality both across skill groups and within narrowly defined groups of workers. The interpretation of these changes dominating the labour literature is an explanation in terms of skill-biased technological change within industries. This refers to the www.eurescom.de/e-living/index.htm Page 4 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD increase in relative demand by firms for highly educated workers as compared to the lower educated. With the growing importance of computers at the workplace and in society in general, this could have profound economic and sociological implications for low skilled, uneducated employees. Krueger (1993) examined the role of the computer as a determinant of wages and whether this computer premium can account for changes in the wage structure in the 1980s. He found significant work-related computer usage premia using American CPS data for 1984 and 1989. In this new and very exciting strand of literature, DiNardo and Pischke (1997), and Entorf and Kramarz (1997) have shown that for Germany and France respectively, cross-sectional results do indeed demonstrate work-related computer usage to generate wage premia. Entorf and Kramarz (1997) however when using a panel and controlling for unobserved individual heterogeneity, find that these computer premia for France are rendered insignificant. DiNardo and Pischke (1997) stress the need for panel data to control for unmeasured individual effects. They find wage effects for using, among other things, a pencil and hand calculator. In contrast, Bell (1996) using British longitudinal data with additional skills and aptitude test information confirms the findings of Krueger (1993) and finds that up to one half of the increase in the return to education can be attributed to various measures of technical skill (computer usage being one of them). Haisken-DeNew and Schmidt (1999) using cross-section data from the German Socio-Economic Panel for 1997, conclude that there indeed is a computer wage premium of around 7%, however after controlling adequately for unobserved individual heterogeneity using panel estimators for 1984-1997, the wage premium is reduced to a mere 1%. Here we provide an outline of the existing American and European literature regarding the role of computer usage (whether at work or at home) and its effect on the wage structure, earnings, and employment prospects and provide a roadmap for future research in this area with particular focus on the use of e-Living panel data. The previous international literature has focussed predominantly on overall main effects. Even if it can be successfully argued that overall effects are small, as some have found under certain conditions, there may be significant between-group averaging of effects, such that in some groups there may be large positive effects whereas in others, perhaps very small or insignificant effects. How are the gains to ICT usage shared between groups? Is there disparity? Who are the "winners"; who are the "losers"? The e-Living panel data will allow an in-depth analysis, examining the effects by many detailed socio-economic and demographic groups. (See Table 1 for the list of socio-economic and demographic indicators to be used.)

2.1 Previous Studies Krueger (1993) links the observed change in the return to education in the United States in the 1980s to increasing popularity of the computer at the workplace. He found that wage differentials gained by those high skilled workers using a computer at work could account for 42% of the increase of the return to education in the private sector in this time period. In the analysis, Krueger used two waves of the October CPS from 1984 and 1989 and found that women were more likely to be using a computer at work, and that in some particular industries more than others, such as the Banking sector, computers were prevalent. He found that males and females aged 29 to 39 and the highest educated tended to use computers the most. Krueger (1993) found raw wage differentials (without any controls) for PC work use in 1984 to be 28%, rising in 1989 to 33%.

One might expect that there are some unobserved positive employer characteristics that are correlated with the existence of PC's in a particular firm, i.e. firms with generous salary packages (efficiency wages) might also as a matter of course provide PC's as a sort of "perk". If this were the case, then one would pick up spurious PC wage differentials simply reflecting the generosity of the employer in an employer-employee rent sharing model. Krueger controls for industry but cannot control for firm size, as it is not asked in the CPS, however Krueger cites Hirschorn (1988) who does not find a strong link between firm size and computer usage. (This contrasts with Haisken-DeNew and Schmidt (1999) who find a positive and strong relationship between firm size and computer use in Germany.) Controls for demographic background, employer characteristics, human capital and union membership reduce the wage impacts of PC use to 14% and 16% respectively. The 1989 CPS differentiates between various computer specific tasks, such as word processing, bookkeeping, CAD, email, inventory control, programming, DTP, spreadsheets, sales and computer games and finds a range of returns for the items, with a high of 15% for email to -5% for DTP and inventory control compared to any "other use" of a PC. Krueger suggests the high return to email might come from the fact that highly paid managers use email intensively. Further, Krueger (1993) draws the link between additional computer-related qualifications or certificates and pay raises for some occupation groups. Thus, Krueger does find large and stable effects for PC usage. However, workers with unobserved skills could be thought to enjoy wage differentials seemingly due to computer usage at work, whereas the real effect came from their ability. By including computer usage at home and its interaction with computer usage www.eurescom.de/e-living/index.htm Page 5 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD at work, any bias in the PC usage effect at work due to omitted factors that are associated with computer use in general, were thought to be eliminated. Indeed Krueger found little change. The wage differential in general for using a computer at the workplace, depending on the specification of the estimation, was found to be between 10 and 15%. Entorf and Kramarz (1994) and Entorf and Kramarz (1997) examine the role of unmeasured ability in the estimations of the computer usage wage premium. They use French Labor Force Survey panel data from 1985-1987 with additional merged firm-level information. For more than 15,000 persons, they have information concerning technology usage at the workplace such that the individuals can be identified over time for a maximum of 3 years in a rotating panel, making up a total of more than 35,000 person-year observations in the panel. In cross-sections, firm effects do not alter the computer usage wage premia. Entorf and Kramarz (1997) can also distinguish between various types of ICT usage, but more importantly, also the kind of usage, i.e. in their terminology "intelligent use" referring to creative use versus "stupid use" referring to robots and assembly lines. They find positive wage effects consisting of two components: whether or not one uses a computer, and also the number of years of experience using a computer. However, they further conclude that indeed, using fixed-effects panel estimation, all computer usage wage premia except for computer experience effects disappear ("differ radically"). The computer experience factor however still remains significant in the panel results. They state, "In particular, to check the effect of NT [New Technologies] on wages, panel data on individuals is necessary, since, as we saw, cross-section data matching workers and firms do not capture the individual ability component of the wage. See Entorf and Kramarz (1997), p. 1504. DiNardo and Pischke (1997) refute the ability to measure true computer effects on wages. Using cross- sectional data for Germany from the German Qualification and Career Survey 1979, 1985-86, 1991-92 they compare German results to the American CPS, replicating Krueger (1993). They include a list of other "office tools" such as pencils, telephones, hand calculator, sitting while working, where one might not expect any particular wage premium to arise. However, they find significant differentials in Germany. The criticism is then, what is actually being measured by a computer usage indicator in a wage regression? They are sceptical of any causal relationship between computer usage and wage premia, "these findings cast some doubt on the literal interpretation of the computer use wage differential as reflecting true returns to computer use or skill." Further, they stress the need for panel data to control for unobserved individual heterogeneity, "Since Krueger relies on cross-section data, he cannot and does not control for individual fixed effects." See DiNardo and Pischke (1997), p. 291. Bell (1996) uses the unique British National Child Development Study from 1981 and 1991 to examine the role of ability and individual heterogeneity in looking at the computer wage premium issue. Here additional test scores are available for reading comprehension and mathematical aptitude. In cross-section, he finds a large significant computer wage premium of 11%, even after controlling for additional skills such as math, planning ability, organisational capabilities. See Bell (1996) Table 5, column (4), p. 28. Using his 1991-1981 difference model, he finds a significant computer wage premium of 13%. He finds, "we show that wages are positively related to these [technical] skills and that there is little evidence that unobserved characteristics of either the individual or the firm are driving the correlation. Furthermore it is the use of these skills in the workplace that is important for wages not simply ability. This suggests that a productivity enhancing interpretation is most appropriate." See Bell (1996), p. 22. In Haisken-DeNew and Schmidt (1999), the German Socio-Economic Panel data set from 1984 to 1997 was used and indeed cross-sectional results from 1997 indicate a highly significant wage premium of 7% for computer usage at work in Germany. They conclude that although cross-sectional evidence may deliver appealing initial results when analysing the wage differential of computer usage, for instance as in Krueger (1993) for the United States, one must include adequate controls for unobserved individual heterogeneity to avoid over-interpreting the results. Simply adding indicators for PC usage at home and interactions between home and work usage are alone insufficient to account for possible "ability" effects. Using all waves and the pooled OLS estimator, the wage premium of using a PC at work is almost 9% and is highly significant, however when using a panel estimator with individual fixed-effects and controlling for computer related skills, this premium all but disappears to 1% and is barely significant, confirming the results of Entorf and Kramarz (1997) for France. This paper extends the results from DiNardo and Pischke (1997) with respect to the data, as one can directly control for unobserved individual heterogeneity using panel data. The GSOEP offers a unique opportunity to examine this question as it provides 14 years of panel information. In stark contrast to Bell (1996), they find that in this analysis for Germany, unobserved individual heterogeneity or ability plays the key role in effectively explaining away the apparent wage premium for using a computer at work.

www.eurescom.de/e-living/index.htm Page 6 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

2.2 E-Living Application: Average Wage Effects of Computer Use The previous studies have shown that many factors are important in addressing the issue of impacts of computer usage on wages. It is not simply enough to collect data on computer usage at the workplace, but rather one must also be able to control for specific tasks that people do with the computers and which specific skills people have. To be able to compare the relative importance of all of these effects, one needs to collect all of these components together and over several years, i.e. a panel. Over time, the changes in exogenous variables will be able to give some insight into the changes in endogenous variables, net of person-specific unobserved heterogeneity. Table 7 lists key questions found in the eLiving questionnaire which will be of critical importance in analysing the impacts of computers and ICT in general. In the past, most analyses have merely focussed on the wage impacts. How are people socially and economically disadvantaged by not having access to ICT, ICT training, ICT at the workplace ? How do these factors differ by country ? In this part, we will use linear regression estimation techniques to model individual wages (labour earnings) as is standard in the literature. Standard human capital factors such as occupation, education/qualification and experience, demographics, and firm characteristics such as sector, firm size, region will be included as the main explanatory variables (Table 7a). This basic model will be augmented with computer use at work (Table 7b), computer use at home (Table 7c), and (years of) experience using computers, components of which have been found for example in Krueger (1993), Entorf and Kramarz (1997) and Haisken-DeNew and Schmidt (1999). We can not only control for standard factors found in the literature, but also for the factors listed in Table 7 , such as intensity of computer use (Table 7d), specific computer tasks (Table 7g), and specific computer skills (Table 7h) simultaneously. People's own attitudes toward computers (Table 7i) may also give insight to otherwise unobserved individual heterogeneity. This will be especially useful in a cross- sectional setting in the first wave before the panel component has been established. The computer use at work and at home indicators can be interacted with any other main explanatory variables to give say sector/occupation specific returns to computer use, or perhaps varying by educational level (Table 7a). Contrary to Krueger (1993), Haisken-DeNew and Schmidt (1999) do indeed find differing firm size PC take- up rates differences and also effects of computer usage. As the panel becomes established, random and fixed effect panel estimators will allow explicit controls for unobserved individual heterogeneity as in Bell (1996), Entorf and Kramarz (1997) and Haisken-DeNew and Schmidt (1999). Bell (1996) in fact only uses two waves of the British NCDS. We examine log earnings (yearly) as a function of demographics (gender, marital status), human capital (labour market experience, education), job characteristics (setting own schedule, hours/week), ICT characteristics (use PC, use Internet, Basic Office computer activities, Network related computer activities). Bulgaria is excluded from the analysis because, as Figure 1 shows, too few Bulgarian workers (20%, n = 134) used a PC at work to enable meaningful statistical analysis1.

1 This number reduces further when we exclude those who did not respond to the income item (28% of Bulgarian workers) and who are self-employed (13% of Bulgarian workers) who were not asked about their ability to control their work schedule. On the assumption that PC use at work will be more widespread in 2002 than 2001, it may be possible to extend this analysis to Bulgaria using wave 2 data. www.eurescom.de/e-living/index.htm Page 7 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

80%

70%

60%

50%

Men 40% Women

30%

20%

10%

0% GB Italy Germany Norway Bulgaria Israel

Figure 1: Proportion of male and female workers in each country who use a PC in their work

We use this data to find the wage impacts (conditional on all other variables, ceteris paribus) of having a PC and the Internet at the workplace for those who are not self-employed2, as shown in Table 1. Here we find that in all countries, Great Britain, Italy, Germany, Norway and Israel, there are significant wage premia paid for having a computer at the workplace. The premia are indeed rather large, ranging from 15.8% in Germany to 37.5% in Israel with Great Britain at 27%, Italy at 22.2% and Norway at 17.9%. However, as mentioned earlier, with only one wave of the eLiving dataset, one cannot control for unobserved individual heterogeneity directly and so these results do not imply causality, merely an association. We must wait for future waves of the data to be collected to examine causality. Thus to some degree, these are to be interpreted as “top boundaries”, with more waves tending to lower these effects. Nonetheless, we can explain about half of the differences between the wages of those having a PC and those not (raw differentials) using other explanatory variables, thereby calculating “net differentials”. For instance, without additional controls, in Great Britain, those using a PC at work enjoy a 55% wage premium, but this drops to 27% with controls listed in Table 1. (Thus showing that some of the high raw return to computer use was in fact due to overly high education, experience or long hours worked of those who used a computer.)

2 As the previous footnote states, the self-employed were not asked the extent to which they controlled their work schedule. www.eurescom.de/e-living/index.htm Page 8 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

Table 1: Wage Impacts of PC/Internet Use at the Workplace

Great Britain Italy Germany Norway Israel Male 0.211 0.213 0.330 0.193 0.267 (4.71)** (3.39)** (6.60)** (6.52)** (4.68)** Married 0.061 0.032 -0.105 0.002 0.122 (1.40) (0.47) (2.16)* (0.07) (2.18)* Experience 0.034 0.044 0.035 0.028 0.028 (4.98)** (4.30)** (4.74)** (5.86)** (3.81)** Experience² -0.001 -0.001 -0.000 -0.001 -0.001 (4.47)** (2.83)** (3.00)** (4.64)** (2.81)** Years Education 0.035 0.047 0.029 0.013 0.025 (5.67)** (6.46)** (5.31)** (4.27)** (4.46)** PC 0.270 0.222 0.158 0.179 0.375 (4.10)** (2.40)* (2.03)* (3.62)** (5.29)** Internet 0.057 0.098 0.114 0.115 0.010 (1.00) (1.13) (1.74) (3.37)** (0.12) Basic Office 0.035 0.115 0.105 -0.011 -0.017 (0.62) (1.38) (1.51) (0.28) (0.24) Network 0.122 -0.002 0.111 0.002 0.080 (2.25)* (0.02) (1.81) (0.07) (1.02) Own Schedule 0.138 0.071 0.062 0.087 0.147 (3.26)** (1.13) (1.34) (2.95)** (2.81)** Hours/Week 0.027 0.016 0.023 0.015 0.015 (14.81)** (5.53)** (11.18)** (10.87)** (8.47)** Constant 0.871 0.537 1.006 2.127 1.040 (7.06)** (3.01)** (8.08)** (25.84)** (7.83)** Observations 655 337 475 767 421 R-squared 0.52 0.39 0.52 0.38 0.42

Note: * significant at 5%; ** significant at 1%, Absolute value of t statistics in parentheses

Source: Own calculations using Wave 1 of the eLiving data set. Interpretation: Here we present mean effects on earnings of PC usage and Internet for each each of the five countries. For instance, overall there would be a 37.5% wage premium associated with using a PC at the workplace in Israel whereas only 15.8% in Germany. Only in Norway are there significant earnings effects of 11.5% for using the Internet at the workplace.

Most countries had strong and significant raw differentials (no controls) for Internet use at the workplace, ranging from 11% to 18%. However, when using identical controls as above, the use of the Internet at the workplace seems only to be significant in Norway at 11.5%. All other countries had insignificant (net) effects. It is also noticeable that there are no effects for type of use (i.e. applications used) apart from highly skilled computer network applications (“Programming/network systems management, PC support”) which had a 12% effect in Great Britain alone. This may suggest that the effects for PC use in general have more to do with the kinds of people who tend to have/use PCs in the work place than the fact of PC usage itself. It also suggests that simply training workers in specific PC related skills will not have much effect on their earnings returns in most countries. It seems clear then that whilst something is going on in terms of a PC-use at work wage premium, precisely what is going on, and therefore what actions need to be taken is not at all clear.

www.eurescom.de/e-living/index.htm Page 9 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

2.3 E-Living Application: Distributional Earnings Effects of Computer Use 2.3.1 Distributional Counterfactuals Combining the two concepts of average (mean) effects and distributional effects, we can pose the question of PC impacts slightly differently. Here we are interested in the question: What would the distribution of wages look like if we could net out the positive effects of PC usage, that is to say, if everyone were paid at the non-PC use level. It would be entirely surprising if the entire distribution were to shift leftwards by exactly the amount of the average effect from the OLS analysis. That would assume that all persons at all points in the distribution would be affected identically. To implement this type of estimation, we employ the tools developed by DiNardo, Fortin and Lemieux (1996) who examine the distributional effects of union membership on the earnings distribution. In our case we will simply exchange “union membership” for “PC use” in examining the earnings distribution. The DiNardo, Fortin and Lemieux (1996) methodology is as follows: 1. We calculate the distribution (kernel density) of wages for all workers, regardless of PC use. This becomes our baseline earnings distribution. 2. Now we start with the counterfactual. We estimate the unconditional probability of not using a PC i.e. just the simple average of not using a PC, say τ (tau). 3. We estimate then a probit model to generate a predicted (conditional) probability of using a PC at the workplace, say pi. Then (1-pi) is the conditional probability that a particular worker “i” does not use a PC.

4. We calculate the person specific weighting factor θi (theta), which is θi = τ / (1- pi). The factor θi (theta) allows one to weight persons proportionally more who are observed not to have a PC and are less likely to have a PC at the workplace, as predicted by the probit model. Persons that are observed not to have a PC at the workplace in spite of a high predicted probability are weighted relatively lower. 5. We calculate the distribution (kernel density) of wages for those workers who do not use a PC at the workplace, but weighting by the factor θi (theta). This becomes our counterfactual earnings distribution. Step (1) yields the distribution of earnings the way we observe it in reality. Step (5) yields the distribution of workers as if they were all paid the non-PC wage rate, i.e. the counterfactual. The difference in the two distributions is the “distributional effect of PC use at the workplace”. In step (3) we use indicators for gender, age, years of education and managerial duties to determine the individual prediction probability. Once we have a distribution for each country, we calculate a country-specific “exclusion threshold”, which is defined to be 80% of the median value (a commonly used, although arbitrary level). This is listed as the “Baseline Risk” in Table 2 and the vertical line to the left side of the distributions in Figure Set 2. Table 2 summarises the results from Figure Set 2. The blue line on all graphs is the distribution of yearly labour earnings for a particular country, as they are observed. The red line is the distribution of labour earnings, given that all were paid at the non-ICT level (the counterfactual). The ICT item is broken down into four components: PC use, Internet, Basic Office computer activities and Network computer activities. Thus, examining Germany in Figure 2(a), we observe that the economic exclusion threshold is defined to be 80% of the median, meaning that those earning less than 19,620 Euro per year are “economically excluded” as we define it, making up 36.8% of the sample. This becomes our benchmark. We then simulate paying everyone at the non-PC pay level and determine what happens to the distribution, as defined by the red line in Figure 2(a)(1). For all points to the left of the exclusion threshold, we see that the red line is above the blue line, meaning that the total number of persons “at risk” of economic exclusion has increased. Therefore the total risk of economic exclusion would increase by 6.7%, given persons were to be paid at the non-PC wage (representing 37.3% of persons). This effect is largest in Great Britain, see in Figure 2(b)(1), at 17.2% additional risk (over baseline) of economic exclusion. Here the entire earnings distribution has shifted leftwards, with even more skewing to the left. Israel and Italy are similar to that of Britain regarding the effects of PC use on the earnings distribution.

Table 2: ICT and Economic Exclusion (Summary of Figure Set 2)

Baseline Risk No No Work No Text & No of Economic Work PC Internet Spreadsheet Networking www.eurescom.de/e-living/index.htm Page 10 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

Exclusion Germany Risk 36.8% +6.7 +0.4 +4.1 +1.4 Share 100% 37.3% 82.9% 51.8% 78.1% Great Britain Risk 34.6% +17.2 +5.0 +7.4 +8.5 Share 100% 35.9% 80.3% 53.2% 61.8% Israel Risk 29.6% +14.6 +6.3 +6.5 +6.9 Share 100% 48.3% 84.1% 69.7% 82.2% Italy Risk 31.1% +15.8 +5.5 +9.2 +4.4 Share 100% 44.5% 81.7% 62.3% 74.8% Norway Risk 24.9% +8.3 +2.4 +3.6 +1.0% Share 100% 28.1% 68.0% 45.7% 66.1% Source: Own calculations using Wave 1 of the eLiving data set. Interpretation: For each country a Baseline Risk is calculated to be 80% of the median earnings. For instance, in Germany, according to this definition, 36.8% of workers are below this threshold and are therefore economically “excluded”. This proportion increases by an additional 6.7% (Increased Risk) for those who do not work with a PC (around 37.3% of the work force).

www.eurescom.de/e-living/index.htm Page 11 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

Figure 2

density: Wage EURO (Yearly) (37.3%) No PC, IR=6.7% density: Wage EURO (Yearly) (82.9%) No Internet, IR=0.4% .04 .03

.02 De De nsit nsit y .02 y

.01

0 0 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 Wage EURO (Yearly) Wage EURO (Yearly) (1) Lack of Work PC (2) Lack of Work Internet density: Wage EURO (Yearly) (51.8%) No Basic Office, IR=4.1 density: Wage EURO (Yearly) (78.1%) No Network, IR=1.4%

.03 .03

.02 .02 De De nsit nsit y y

.01 .01

0 0 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 Wage EURO (Yearly) Wage EURO (Yearly) (3) Lack of Basic Office (4) Lack of Network Exclusion Threshold (80%*Median): 19.62 - Baseline Risk: 36.8% Figure 2(a): ICT and Earnings Distribution: Germany

density: Wage EURO (Yearly) (35.9%) No PC, IR=17.2% density: Wage EURO (Yearly) (80.3%) No Internet, IR=5.0% .03 .03

.02 .02 De De nsit nsit y y

.01 .01

0 0 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 Wage EURO (Yearly) Wage EURO (Yearly) (1) Lack of Work PC (2) Lack of Work Internet density: Wage EURO (Yearly) (53.2%) No Basic Office, IR=7.4 density: Wage EURO (Yearly) (61.8%) No Network, IR=8.5% .03 .03

.02 .02 De De nsit nsit y y

.01 .01

0 0 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 Wage EURO (Yearly) Wage EURO (Yearly) (3) Lack of Basic Office (4) Lack of Network Exclusion Threshold (80%*Median): 20.43 - Baseline Risk: 34.6% Figure 2(b): ICT and Earnings Distribution: Britain

www.eurescom.de/e-living/index.htm Page 12 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

density: Wage EURO (Yearly) (48.3%) No PC, IR=14.6% density: Wage EURO (Yearly) (84.1%) No Internet, IR=6.3% .06 .06

.04 .04 De De nsit nsit y y

.02 .02

0 0 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 Wage EURO (Yearly) Wage EURO (Yearly) (1) Lack of Work PC (2) Lack of Work Internet density: Wage EURO (Yearly) (69.7%) No Basic Office, IR=6.5 density: Wage EURO (Yearly) (82.2%) No Network, IR=6.9% .06 .06

.04 .04 De De nsit nsit y y

.02 .02

0 0 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 Wage EURO (Yearly) Wage EURO (Yearly) (3) Lack of Basic Office (4) Lack of Network Exclusion Threshold (80%*Median): 12.19 - Baseline Risk: 29.6% Figure 2(c): ICT and Earnings Distribution: Israel

density: Wage EURO (Yearly) (44.5%) No PC, IR=15.8% density: Wage EURO (Yearly) (81.7%) No Internet, IR=5.5% .06 .06

.04 .04 De De nsit nsit y y

.02 .02

0 0 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 Wage EURO (Yearly) Wage EURO (Yearly) (1) Lack of Work PC (2) Lack of Work Internet density: Wage EURO (Yearly) (62.3%) No Basic Office, IR=9.2 density: Wage EURO (Yearly) (74.8%) No Network, IR=4.4% .06 .06

.04 .04 De De nsit nsit y y

.02 .02

0 0 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 Wage EURO (Yearly) Wage EURO (Yearly) (3) Lack of Basic Office (4) Lack of Network Exclusion Threshold (80%*Median): 12.40 - Baseline Risk: 31.1% Figure 2(d): ICT and Earnings Distribution: Italy

www.eurescom.de/e-living/index.htm Page 13 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

density: Wage EURO (Yearly) (28.1%) No PC, IR=8.3% density: Wage EURO (Yearly) (68.0%) No Internet, IR=2.4% .04 .04

De De nsit nsit y .02 y .02

0 0 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 Wage EURO (Yearly) Wage EURO (Yearly) (1) Lack of Work PC (2) Lack of Work Internet density: Wage EURO (Yearly) (45.7%) No Basic Office, IR=3.6 density: Wage EURO (Yearly) (66.1%) No Network, IR=1.0% .04 .04

De De nsit nsit y .02 y .02

0 0 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 Wage EURO (Yearly) Wage EURO (Yearly) (3) Lack of Basic Office (4) Lack of Network Exclusion Threshold (80%*Median): 27.22 - Baseline Risk: 24.9% Figure 2(e): ICT and Earnings Distribution: Norway

Similarly, we can observe the effects on the labour earnings distribution for Internet usage at the workplace. The effects in general are all substantially smaller than for PC usage, which we expected, given the results of Table 1. In Germany, for example in Figure 2(a)(2) we see that the Increased Risk (IR) is near zero. However in Britain, as shown in Figure 2(b)(2), it is 5.0% over baseline risk, similarly 6.3% in Israel and 5.5% in Italy. Norway seems to be affected only marginally at 2.4% increased risk. Next, we examine the monetary reward to standard office type computer activities such as text processing and spreadsheets. Although we found from Table 1no significant effects on average, when we ask the counterfactual question for the earnings distribution, we observe an increase risk of economic exclusion of between 3.6% in Norway as in Figure 2(e)(3) and 9.2% in Italy as in Figure 2(d)(3) in the left tail of the distributions. For computer network related activities, we find the highest increased risk of 8.5% in Great Britain as shown in Figure 2(b)(4) and the lowest in Norway at 1.0% as shown in Figure 2(e)(1). 2.3.2 Quantile Regression In quantifying the impact of computer use on wages, the existing literature seems to focus exclusively on average effects, i.e. overall coefficients, which give the average effect of using a PC at the workplace as opposed to not using a PC. However, the previous section suggests that there might be differential effects over the distribution of wages. For instance, it could be that the effects of PC usage might be much higher for those in the lower end of the earnings distribution, i.e. using a PC might dramatically increase their earnings, as has been demonstrated in the previous section. This would suggest analysing the data not with regular OLS regression but rather with quantile regression, evaluating not at the mean but rather at various cuts in the distribution. We could examine the distribution at the 10%-ile (very poor off) , 25%-ile (poor), 50%-ile (median), 75%-ile (well off) and the 90%-ile (very well off). Table 3 illustrates differential effects by various points in the earnings distribution for Great Britain. The endogenous variable is log yearly earnings, explained by indicators for gender, marital status, labour market experience as a quadratic, years of education, hours per week, and whether one can set one’s own schedule at the job. In addition there are ICT related indicators for the workplace, such as computer use, internet

www.eurescom.de/e-living/index.htm Page 14 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD access, basic “office” computer related programs like word processing and spreadsheets and networking or systems operations activities. On average using OLS regression, there is a 27% wage premium associated with using a PC at the workplace in the UK. For the median worker (based on the earnings distribution) this is almost the same at 26.7%. However, this effect could be as low as 18.6% or as high as 36.3% depending on where one sits in the earnings distribution (see the row labelled “PC”). We also see that the effect for highly skilled computer network applications reported in Table 1 is only found in the 50th and 75th income deciles suggesting a ‘rich get richer’ situation. Thus in general UK workers on the low end of the income distribution would gain much more if they had access to a computer at the workplace. This is of course by itself far too simple: the workers must first have sufficient skills to use the computer, and also work in such an environment that the technology of production can incorporate computer use, i.e. there must be adequate demand for this type of skilled labour. Clearly, a simple newspaper boy who uses a computer, will probably not experience a wage increase due to increased productivity. However, all things being equal, putting computers to use together with workers on the left tail of the earnings distribution has the largest impact of increasing earnings although we must caution any assumption that this link is causal.

www.eurescom.de/e-living/index.htm Page 15 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

Table 3: Great Britain: Quantile Regression at Various Percentiles in the Earnings Distribution

Mean (OLS) 10%-ile 25%-ile 50%-ile 75%-ile 90%-ile (1) (2) (3) (4) (5) (6) Male 0.211 0.171 0.222 0.236 0.251 0.257 (4.71)** (1.38) (4.00)** (4.66)** (4.57)** (5.41)** Married 0.061 0.169 0.091 0.024 0.029 -0.007 (1.40) (1.30) (1.54) (0.65) (0.68) (0.13) Experience 0.034 0.034 0.028 0.037 0.042 0.045 (4.98)** (2.56)* (3.81)** (7.20)** (7.34)** (3.96)** Experience² -0.001 -0.001 -0.001 -0.001 -0.001 -0.001 (4.47)** (2.14)* (2.97)** (6.00)** (7.79)** (3.77)** Years Education 0.035 0.022 0.020 0.036 0.052 0.048 (5.67)** (1.90) (3.60)** (3.61)** (4.70)** (4.01)** PC 0.270 0.363 0.327 0.267 0.186 0.280 (4.10)** (3.22)** (3.06)** (4.64)** (3.04)** (2.57)* Internet 0.057 0.157 0.089 0.059 0.078 0.052 (1.00) (1.03) (1.19) (1.11) (1.58) (0.72) Basic Office 0.035 0.049 0.039 0.082 0.004 0.030 (0.62) (0.33) (0.48) (1.53) (0.08) (0.37) Network 0.122 0.210 0.097 0.116 0.117 -0.010 (2.25)* (1.83) (1.80) (2.07)* (2.38)* (0.13) Own Schedule 0.138 0.099 0.158 0.150 0.106 0.049 (3.26)** (1.23) (3.80)** (4.33)** (2.72)** (0.82) Hours/Week 0.027 0.030 0.030 0.028 0.023 0.014 (14.81)** (8.64)** (10.18)** (10.07)** (7.36)** (3.52)** Constant 0.871 0.231 0.738 0.825 1.127 1.744 (7.06)** (0.95) (5.14)** (6.29)** (7.01)** (6.14)** Observations 655 655 655 655 655 655 R-squared 0.52 0.37 0.38 0.37 0.31 0.24

Note: * significant at 5%; ** significant at 1%, Absolute value of t statistics in parentheses. Remaining countries are reported in the following table. Source: Own calculations using Wave 1 of the eLiving data set. Interpretation: Here mean effects as in column (1) are compared to effects at certain points in the earnings distribution, as in column (2) through (7). For instance, overall there would be a 27% wage premium associated with using a PC at the workplace, however this would jump to 36.3% for those very poor off at the 10%-ile of the earnings distribution.

Table 4 shows the condensed results for the four remaining countries, where the coefficients marked with a star indicate significant effects. Germany for instance has an average PC effect of 15.7% - this means that an average worker using a PC would earn 15.7% more (using standard controls). However a worker whose earnings would put him in the lower tail of the earnings distribution would experience much higher returns, e.g. more than double (32.5%) the return at the 10 %-ile. However at the median the effect is no longer significant. In contrast, in Italy significant effects first start at the 25 %-ile and remain significant moving rightwards over the distribution. Clearly the countries analysed here using quantile regression do indeed behave differently at various parts of the earnings distribution and therefore may require different kinds of actions to tackle economic exclusion.

www.eurescom.de/e-living/index.htm Page 16 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

Table 4: Other Countries: Quantile Regression at Various Percentiles in the Earnings Distribution

Percentile Germany Israel Italy Norway Obs 475 421 337 767

Mean (OLS) PC 0.157* 0.374* 0.222* 0.179* Internet 0.113* 0.010 0.097 0.114*

10 %-ile PC 0.325* 0.268* 0.230 0.259* Internet 0.228* -0.191 0.041 0.099*

25 %-ile PC 0.283* 0.464* 0.251* 0.097 Internet 0.148* -0.078 0.177* 0.080*

50 %-ile PC 0.129 0.376* 0.226* 0.107* Internet 0.055 0.212 0.146 0.100*

75 %-ile PC 0.113* 0.313* 0.285* 0.102* Internet 0.066* 0.335* 0.143 0.132*

90 %-ile PC 0.035 0.372* 0.180* 0.086 Internet 0.138* 0.263* 0.231* 0.176*

Note: * significant at 5%, one sided t-test. Source: Own calculations using Wave 1 of the eLiving data set. Interpretation: Here mean effects as in row (1)are compared to effects at certain points in the earnings distribution, 10%-ile through to the 90%-ile. For instance, overall there would be a 37.4% wage premium associated with using a PC at the workplace in Israel, however this would jump to 46.4% for those poor off at the 25%-ile of the earnings distribution in Israel.

3 ICT Use and Social Exclusion There is a considerable literature on social exclusion, including with respect to ICT access. Some of this is covered in other work packages, in particular WP3 (ICT patterns) and WP7 (the relationship between ICT access and gender). Previously we focused on whether access to computers significantly enhances individuals’ productive potential and therefore wages and therefore whether lack of access limits this. The answer to this questions has clear and substantial welfare implications. However, it is possible that certain categories of people are more constrained than others, and so this issue has to be examined along various social dimensions. Older people are well known to be less familiar and happy with computer technology. In some countries, women still lagged behind men in computer usage, but on the other hand job segregation by gender might also have an impact (which may or may not be protective) . Moreover there might be significant variation in the distribution of access and opportunity both regionally and between countries.

3.1 Previous Studies The term social exclusion is used often in a blanket manner and can mean many things to different researchers. As D’Ambrosio et al. (2002) write, most importantly the concept of social exclusion deals with the “inability of an individual to participate in the basic political, economic and social functionings of the society in which he/she lives.” Of interest here is exactly how this concept can be operationalised into observable indicators available to researchers. An individual is considered to be “excluded” if based on many indicators, he/she cannot participate fully in society. Thus simply to be lacking in one particular area does not constitute “exclusion” and therefore we are interested in a multi-dimensional index which summarises information from many domains. In the strictest sense of the term, exclusion deals with not having access to something not because one chose not to have it but rather because it was simply beyond the reach of a person, whether due to budget restrictions or institutional restrictions etc. If longitudinal information were present, we could also focus on persistence of exclusion. Typically periodic dips into and out of exclusion would be weighted lower than long-term exclusion. However in the eLiving database, we currently only have one wave of data to analyse so the longitudinal aspects cannot yet be addressed here. We are interested in the role of information technology in bringing people closer together, empowering them and improving their lives. We have in the previous section examined the wage impact of ICT at the workplace. Here we will examine how ICT is associated with social exclusion.

www.eurescom.de/e-living/index.htm Page 17 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

Eurostat (2000) outline various indicators as main components of a multi-dimensional social exclusion index: (a) Financial Difficulties, (b) Basic Necessities, (c) Housing Conditions, (d) Consumer Durables, (e) Health, (f) Social Contact, (g) Dissatisfaction. This is not the only definition found in the literature. Dekkers (2002) cites many competing definitions, such as those found in Townsend (1979,1993), Whelan and Whelen (1995), Zajczyk (1995), Percy-Smith (2000) etc. For more information, the reader is directed to D’Ambrosio et al. (2002) and Dekkers (2002), who provide a thorough overview of the existing literature on social exclusion. Dekkers (2002) examines also the role of ICT and social exclusion using Belgian household panel data using very similar ICT consumer durables as in eLiving, stressing the importance of “leading” and “laggard” households in their take-up rates with ICT goods. For instance, he finds that poorer households tend to be “laggard” households, where ICT is a top-down phenomenon with the richer households “leading”. This appears to be true for mobile phone use in Belgium, where disproportionately more poor households do not have access to mobile phones.

3.2 Operationalisation in eLiving Using the eLiving data set we will be able to operationalise the previously mentioned D’Ambrosio et al. (2002) concepts as follows: 1. Financial Difficulties: being in upper income class with respect to household income. 2. Basic Necessities: having high education, PC skills, working, having a permanent job, having a job in which one sets one’s own schedule. 3. Housing Conditions: having housing with more than 2 rooms 4. Consumer Durables: having a car, more than one television, having a clothes washing machine, dish washing machine, microwave oven, CD/stereo, video camera, VCR, digital camera, DVD player 5. (Mental) Health: overall life satisfaction 6. Social Contact: talking often on the telephone to friends, satisfaction with communication with friends 7. Dissatisfaction: satisfaction with free time All indicators are coded as zero (0) or one (1), with 1 meaning having a particular good or characteristic and zero (0) not. Tsakloglou and Papdopoulos (2001) and Papdopoulos and Tsakloglou (2002) suggest a method of combining these items into a single index. For the population as a whole, one examines first the overall average of persons having a particular item/good/characteristic, say γi (gamma). Then one ascertains whether each individual has the particular item and if he/she does have the item, then αi (alpha) is equal to 1 and zero if not. Multiplying by the average is an attempt to weight the particular importance of a particular item. If all others have an item and a small number do not, then this small number is considered to be relatively more excluded. If however, in general very few people do not have a particular item, say an expensive car, then even though many would not have such an item, they would not be considered very excluded. Thus each person either has zero (0) when he/she does not have a particular item, or he/she has

γi. The list of items is averaged for every individual and then an overall index of exclusion based on all items is available for each individual. Assuming there were K items/goods/characteristics, the calculation of the index measure would be as follows for each individual “i”:

SocExi = ( [α1i * γ1] + [α2i * γ2] +… + [αKi * γK] ) / K, where α1i , α2i , .., αKi are either 0 or 1, and γ1 , γ2 , ..., γK , range between 0 and 1. Clearly the index SocExi is bounded by 0 and 1, with 0 being complete exclusion, and 1 being complete inclusion. Typically though, the empirical distribution will lie between some number larger than zero and some other number smaller than one. Example results of these calculations are provided in Table 5 converted into percentages.

Table 5: Indicators of Social Exclusion – example data from 3 countries

Item or Characteristic Germany Britain Israel High Household Income 11.8% 20.5% 17.2%

www.eurescom.de/e-living/index.htm Page 18 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

Part of Clubs and Groups 68.9% 56.9% 54.3% Satisfaction with Free Time 22.1% 32.1% 19.2% Satisfaction with Friends 58.3% 80.4% 51.1% Satisfaction in General 22.7% 41.3% 22.8% Highly Educated 9.6% 18.3% 28.8% Knowledge of PC Use 92.3% 93.9% 96.5% Working 68.7% 77.4% 66.7% Permanent Job 59.6% 71.7% 57.0% Can Schedule one's Own Job 34.9% 33.9% 27.6% Durable: More than one TV 53.0% 79.8% 61.3% Durable: Washing Machine 96.7% 97.1% 97.1% Durable: Dishwasher 71.0% 35.4% 37.3% Durable: Microwave Oven 72.5% 92.1% 84.1% Durable: Stereo / CD Player 95.4% 95.1% 76.5% Durable: Video Camera 32.0% 37.4% 34.4% Durable: VCR 84.7% 91.7% 76.0% Durable: Digital Camera 10.6% 20.7% 21.8% Durable: DVD Player 21.0% 32.5% 19.8% Durable: Automobile 92.3% 87.6% 77.6% Dwelling: More than 2 Rooms 87.1% 93.9% 91.5% Talk with Friends on Telephone 51.2% 66.8% 74.9%

Note: a percentage close to 100 indicates high social inclusion. There is one particular drawback of this methodology which is worthy of noting. Although we have many indicators from various domains, we explicitly weight the importance of each particular indicator equally. Perhaps in reality, not having a DVD player is not all that important, whereas having enough rooms in one’s apartment is much more important. We cannot account for this with this measure. If allowed to be determined endogenously, the rank of importance of the domains will be typically different between countries and over time. This might even be true of individuals in a given country at a given time. (This index is ideally suited for longitudinal analysis, whereas we only have one wave of data currently available in the e-Living data set. “Exclusion”, defined very narrowly, would occur when one did not have access to many different goods, etc not just in one period but persistently over time. Nonetheless, we can use this index as a starting point for the first wave.) Having an exclusion index for each individual, one can examine the distribution of the index. Is there clumping around some median or is it spread out with many extreme observations in the low and high ends ? Here we are interested chiefly in what is going on at the left tail of the distribution, not at some mean value. We can define a social exclusion threshold, below which one is considered to be “socially excluded”. Standard in the literature is to define some percentage of the median index value: some value such as 70% or 80% of the median is often used. For the analysis with e-Living data, we focus on 80% of the median as the threshold as in Section 2 where we considered wage effects, although this is arbitrary. The exact social exclusion threshold used will vary between countries as the country-specific medians will vary so we are implicitly controlling for inter-country differences using this approach. Thus, when we examine the distribution of the exclusion index, we can calculate the area to the left of the threshold under the distribution line. This becomes the “population at risk” or “baseline risk”. This is of interest by itself, however we are interested in determining how this distribution changes and/or shifts when we examine various sub-populations. We can again use the tools we have borrowed from DiNardo, Fortin and Lemieux (1996) to look at how various ICT components relate to social exclusion. We have identified four main areas of ICT: (a) having access to a home computer, (b) using a computer at the workplace (c) having access to home internet and (d) having access to a cell/mobile phone. Again, for the reasons given above, Bulgaria has had to be excluded from this analysis due to the low penetration of PCs, Internet and personally-owned mobiles. For each main ICT area, we calculate the baseline distribution of the exclusion index and then calculate the counterfactual of what the social exclusion index would look like www.eurescom.de/e-living/index.htm Page 19 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD for all those who had NO access. For home computer access, we even have additional information asked of those who do not have a home computer as to why it was not present in the home. We can identify two reasons for not having a home computer which satisfy stringent exclusions definitions: not being able to afford a computer, (or computers cost too much) or not knowing how to use a computer. Typically, this implies a leftward shift of some magnitude of the exclusion distribution and an increase in the population at risk. We can calculate the increase in the population at risk and calculate the measure of “increased risk” called IR, as we have discussed in the previous section. Following DiNardo, Fortin and Lemieux (1996), it is not enough merely to examine those who do not have a particular item, but rather to weight those persons proportionately higher who have the least likelihood of having a particular ICT item in calculating the counterfactual. To predict the individual probability of having a certain item, we use indicators for: gender, age, marital status, years of education, household size and household income. These ICT measures themselves are not used in calculating the SocEx index. We cannot claim any causality but we can examine how likely it is that a person is considered socially excluded (based on a host of non-ICT indicators) given that he/she does not have access to particular ICT items.

3.3 Empirical Application and Interpretation 3.3.1 Interpreting Exclusion Graphically For each of the five countries Germany, Great Britain, Israel, Italy, and Norway we try to identify the association between various ICT items and social exclusion. The results are summarised in Figure Set 3 in a collection of four graphs for each country whereas Table 6 provides the summary of the figures in tabular form. Some explanation of how to interpret the graphs is required here. For example, let us take the top left graph from Israel (Figure 3c), which depicts the distributional effects of the social exclusion index for the Israeli population at large as we observe them (the blue line) and as a counterfactual, the social exclusion index for those not having a home PC for whatever reason (red line) and those not having a home PC due to definitional exclusion reasons (green line). From the main title in the overall graphic, we know that the threshold (80% of median) is the value 0.30, implying a baseline risk of 22.4%. That is to say, given that we use this particular (common) threshold, 22.4% of Israelis are found to be suffering social exclusion (below the threshold). However, when examining those persons not having a home PC (some 29.8% of the Israeli population), there is a increased risk of 18.5% of social exclusion above and beyond the baseline risk. This is further exacerbated by those 7% of Israelis not being able to afford a home PC with an increased risk of 35.9% over baseline.

Table 6: ICT and Social Exclusion (Summary of Figure Set 3)

Baseline Risk No No No No No of Social Home PC Afford PC PC At Work Internet Cell Phone Exclusion Germany Risk 15.7% -1.3 -- -3.6 -4.0 +2.5 Share 100% 28.8% Under 5% 20.4% 71.7% 14.4% Great Britain Risk 15.3% +8.3 +25.2 +15.5 +0.2 +26.6 Share 100% 30.5% 5.6% 25.6% 67.7% 10.3% Israel Risk 22.4% +18.5 +35.9 +13.3 +5.9 +44.1 Share 100% 29.8% 7.0% 35.0% 68.4% 8.6% Italy Risk 19.1% +6.1 -- +3.8 +3.2 +27.5 Share 100% 44.1% Under 5% 36.5% 68.5% 6.9% Norway Risk 13.8% +0.1 -- +1.8 -5.1% -- Share 100% 17.8% Under 5% 9.3% 60.1% Under 5% Source: Own calculations using Wave 1 of the eLiving data set. Interpretation: For each country a Baseline Risk is calculated to be 80% of the median social exclusion index. For instance, in Great Britain, according to this definition, 15.3% of the adults are below this threshold and are therefore socially “excluded”. This proportion increases by an additional 8.3% (Increased Risk) for those who do not have a home PC (around 30.5% of the adults). For robustness, only those effects supported by at least 5% of the sample are reported.

3.3.2 Computers, Internet and Social Exclusion For Germany, we observe no association between broadly defined social exclusion and lack of access to computers at home or the workplace. If anything, those not having access to PC’s might even be slightly less

www.eurescom.de/e-living/index.htm Page 20 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD likely to suffer social exclusion. Fewer than 5% answered that they did not have a PC at home because of financial constraints or difficulties, i.e. strictly defined “exclusion”. (The same was true for Norway and Italy). Thus, this particular aspect was dropped from the analysis as the underlying sub-sample size was insufficiently large for robust analysis. However for Great Britain, we observe quite well that those not having access to a PC are likely to be defined as suffering social exclusion based on other factors. With a baseline risk of 15.3%, not having a home computer is associated with an increased risk of social exclusion by 8.3%. For those who said they could not afford a home PC, this is associated with a 25.2% increase over the baseline risk! Approximately one quarter of the sample has no access to computers at home or at work, resulting in a 15.5% increase over the baseline risk. However lack of access to the Internet does not increase the risk in any significant way. A very similar pattern can be found for Israel. The baseline social exclusion risk is 22.4%. Those not being able to afford a PC have an additional increased risk of 35.9%! Clearly those persons in Israel not being able to afford a home PC are fundamentally different with respect to their social attachment as compared to the rest of society. No PC access whether at home or at work is associated with a 13.% increase over the baseline risk. No home Internet access add 3.2% increased risk. Italy displays very moderate effects. Less than 5% said they could not afford a home PC, so this does not appear to be an issue. PC and Internet access have apparently little affect with respect to social exclusion. Norway is the most extreme case, where the baseline risk is the lowest (i.e. the country with the most inclusion). Here as well, less than 5% said they could not afford a PC. Internet and PC access does not appear to have any association with social exclusion whatsoever. In fact, the measures are slightly negative! 3.3.3 Cell / Mobile Phones and Social Exclusion Compared to the mild affects of PC and Internet access on social exclusion, cell phone or mobile phone access has a relatively strong association with social exclusion in most countries (except Norway where so few people do not have one that the analysis is not valid). In Britain the risk increases by 26.6%, in Israel by 44.1%, in Italy by 27.5%. In Germany the effect is quite small at 2.5%. As the costs of cell phones have dropped dramatically over the past 5 years to a point in 2000 where prepaid cell phones in Germany were selling for as low as EUR 25, cell phone penetration has reached seemingly saturation levels. For instance only 10% of adult Britons did not have access to a cell phone (In Italy only 7% and in Norway 4.9%). As this particular ICT item is so inexpensive, it seems hardly plausible for most people not to be able to afford it. Those who indeed do not have access to such an item, must almost assuredly be very different to the average. Indeed we see this in the increased risk results when set beside the low share values.

www.eurescom.de/e-living/index.htm Page 21 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

Figure 3

density: Exclusion Index [0,1] (28.8%) No PC, IR=-1.3% density: Exclusion Index [0,1] (20.4%) No PC Work, IR=-3.6% 7 7 6 6

5 5

De 4 De nsit nsit4 y y 3 3

2 2 1 1 0 0 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Exclusion Index [0,1] Exclusion Index [0,1] (1) Lack of Home PC (2) Lack of PC Work density: Exclusion Index [0,1] (71.7%) No Internet, IR=-4.0% density: Exclusion Index [0,1] (14.4%) No Cell Phone, IR=2.5% 7

6 7

5 6

De 4 De 5 nsit nsit y y 4 3 3 2 2 1 1 0 0 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Exclusion Index [0,1] Exclusion Index [0,1] (3) Lack of Home Internet (4) Lack of Cell Phone Exclusion Threshold (80% of Median): 0.32 --- Baseline Risk: 15.7% Figure 3(a): ICT and Social Exclusion: Germany

density: Exclusion Index [0,1] (30.5%) No PC, IR=8.3% density: Exclusion Index [0,1] (25.6%) No PC Work, IR=15.5% (5.6%) No PC Afford, IR=25.2% 7 7 6 6 5 5 De 4 De 4 nsit nsit y y 3 3 2 2 1 1 0 0 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Exclusion Index [0,1] Exclusion Index [0,1] (1) Lack of Home PC (2) Lack of PC Work density: Exclusion Index [0,1] (67.7%) No Internet, IR=0.2% density: Exclusion Index [0,1] (10.3%) No Cell Phone, IR=26.6% 7 7

6 6 5 5

De 4 De 4 nsit nsit y y 3 3

2 2 1 1 0 0 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Exclusion Index [0,1] Exclusion Index [0,1] (3) Lack of Home Internet (4) Lack of Cell Phone Exclusion Threshold (80% of Median): 0.38 --- Baseline Risk: 15.3% Figure 3(b): ICT and Social Exclusion: Britain

www.eurescom.de/e-living/index.htm Page 22 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

density: Exclusion Index [0,1] (29.8%) No PC, IR=18.5% density: Exclusion Index [0,1] (35.0%) No PC Work, IR=13.3% (7.0%) No PC Afford, IR=35.9% 7 7 6 6 5 5 De 4 De 4 nsit nsit y y 3 3 2 2 1 1 0 0 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Exclusion Index [0,1] Exclusion Index [0,1] (1) Lack of Home PC (2) Lack of PC Work density: Exclusion Index [0,1] (68.4%) No Internet, IR=5.9% density: Exclusion Index [0,1] (8.6%) No Cell Phone, IR=44.1% 7 7

6 6 5 5

De 4 De 4 nsit nsit y y 3 3

2 2 1 1 0 0 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Exclusion Index [0,1] Exclusion Index [0,1] (3) Lack of Home Internet (4) Lack of Cell Phone Exclusion Threshold (80% of Median): 0.30 --- Baseline Risk: 22.4% Figure 3(c): ICT and Social Exclusion: Israel

density: Exclusion Index [0,1] (44.1%) No PC, IR=6.1% density: Exclusion Index [0,1] (36.5%) No PC Work, IR=3.8% 7 7 6 6

5 5

De 4 De 4 nsit nsit y y 3 3

2 2 1 1 0 0 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Exclusion Index [0,1] Exclusion Index [0,1] (1) Lack of Home PC (2) Lack of PC Work density: Exclusion Index [0,1] (68.5%) No Internet, IR=3.2% density: Exclusion Index [0,1] (6.9%) No Cell Phone, IR=27.5% 7 7

6 6 5 5

De 4 De 4 nsit nsit y y 3 3

2 2 1 1 0 0 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Exclusion Index [0,1] Exclusion Index [0,1] (3) Lack of Home Internet (4) Lack of Cell Phone Exclusion Threshold (80% of Median): 0.29 --- Baseline Risk: 19.1% Figure 3(d): ICT and Social Exclusion: Italy

www.eurescom.de/e-living/index.htm Page 23 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

density: Exclusion Index [0,1] (17.8%) No PC, IR=0.1% density: Exclusion Index [0,1] (9.3%) No PC Work, IR=1.8% 7 7 6 6

5 5

De 4 De 4 nsit nsit y y 3 3

2 2 1 1 0 0 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Exclusion Index [0,1] Exclusion Index [0,1] (1) Lack of Home PC (2) Lack of PC Work density: Exclusion Index [0,1] (60.1%) No Internet, IR=-5.1% density: Exclusion Index [0,1] (4.9%) No Cell Phone, IR=-4.2% 7 7

6 6 5 5

De 4 De 4 nsit nsit y y 3 3

2 2 1 1 0 0 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Exclusion Index [0,1] Exclusion Index [0,1] (3) Lack of Home Internet (4) Lack of Cell Phone Exclusion Threshold (80% of Median): 0.42 --- Baseline Risk: 13.8% Figure 3(e): ICT and Social Exclusion: Norway

www.eurescom.de/e-living/index.htm Page 24 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

3.3.4 Social Exclusion Patterns The next step will be to examine the relationship between ICT use and the various components of social exclusion. Thus we evaluate the social exclusion function with increasing number of exclusions and examine the shares of ICT access/usage and the pattern of the relationships. To date we have conducted this analysis for 3 countries – Germany, Great Britain and Israel. For example, in Germany (shown in Figure 4 and in full in Table 8) taking into account all indicators for social exclusion, those 5.2% who experienced 6 exclusions had on average a score of 0.478 (on a scale of 0 to 1) and 86.2% had a home PC. Compare this to those 5.0% of the sample having 14 exclusions with a social exclusion score of 0.278 and only 43.7% having access to a home PC (almost half)! This trend is confirmed with Internet access, having a PC at the workplace, and mobile/cell phone access. In general increasing social exclusion implies decreased ICT participation (home PC, home internet, work PC, mobile phones). We do not focus in this chart on those cells not having at least support of 5% of the observations (c.f. Table 8). Clearly there are very few persons who have almost no items (15-21) or almost all items (0-5 exclusions). Interestingly enough, those experiencing many exclusions (14 and over) still have access to mobile phones to a large extent (at least half). This is probably because the purchase cost of mobile phones has been traditionally quite low, if not subsidised by providers outright, but by locking consumers into long-run contracts or pre-paid cards.

100% 90% 80% Index percent 70% Inclusion Index mean 60% Home PC access 50% % Internet Acces s 40% PC use at work 30% Mobile phone access 20% 10% 0% 6 7 8 9 10 11 12 13 14 Total Number of Exclusions

Figure 4: Germany. This chart only includes those cells which contain more than 5% of the population (i.e. those who had more than 5 but less than 15 exclusions. See Table 8 at end).

Figure 5 presents the results of similar analysis for Great Britain. Here computer and mobile use is higher for those least socially excluded but Internet access is relatively flat. However even those considered quite socially excluded (12 exclusions and above), still have relatively high levels of mobile phone access (almost 80%) in contrast to the pattern in Germany.

www.eurescom.de/e-living/index.htm Page 25 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

100% 90% 80% Index percent 70% Inclusion Index mean 60% Home PC access

% 50% Internet Acces s 40% PC use at work 30% Mobile phone access 20% 10% 0% 56789101112 Total Number of Exclusions

Figure 5: Great Britain This chart only includes those cells which contain more than 5% of the population (i.e. those who had more than 4 but less than 13 exclusions. See Table 9 at end.)

100% 90% 80% 70% Index percent Inclusion Index mean 60% Home PC access 50% % Internet Acces s 40% PC use at work 30% Mobile phone access 20% 10% 0% 6 7 8 9 10 11 12 13 14 Total Number of Exclusions

Figure 6: Israel. This chart only includes those cells which contain more than 5% of the population (i.e. those who had more than 5 but less than 15 exclusions. See Table 10 at end.)

Figure 6 presents the results for Israel. Computer and Internet use is dramatically higher for those least socially excluded. Those considered very much socially excluded (14 exclusions and above), still have high levels of mobile phone access (more than 80%) as was the case with Great Britain but not in Germany. In contrast to Great Britain, Internet access is a better indicator of exclusion. In general however the patterns in each country are relatively similar to one another. We can suggest that lack of ownership and/or usage of these ICTs is a good indicator of ‘social inclusion’ as might be measured by other non-ICT related variables and this is especially so for home PCs, mobile phones and using PCs at work. Interestingly having Internet access at home is a much less useful indicator. It will be interesting to see in future work if the same patterns are found in the other countries surveyed. What we cannot conclude of course is that these relationships are causal. Instead it points to the fact that those small groups of people www.eurescom.de/e-living/index.htm Page 26 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD who are ‘ICT excluded’ are also very likely to be ‘economically and socially excluded’. It remains for future work using longitudinal data to unravel the relationships between the two.

4 Conclusions There appears to be a clear relationship between ICT use/access and decreasing exclusion at the economic and social levels. The richness of the e-Living data set allows us to calculate detailed indicators of economic and social exclusion for each country in the data set. We find a substantial degree of variation of effects between countries, which clearly indicates the advantage of a multi-country empirical analysis – not all countries behave the same way and we can identify the differences. Moreover, the exclusion analysis is based on distributional considerations. All of the analyses have allowed for country specific baselines (means, medians, etc), taking into consideration the country specific nature of what is “desirable” in a given country. This indirectly takes into account country specific institutions, patterns of labour market participation, education levels, behaviour in purchasing consumer durables etc which might otherwise make comparison impossible. When dealing with economic exclusion and ICT, the strongest effects come from computer usage at the workplace. Those not using a computer at the workplace, whether they do not have the required training to do so or because their job does not call for it, seem to be disadvantaged economically. Using simple regression techniques, one can identify a computer wage premium of anywhere between 16% in Germany and 38% in Israel on average (over all of the earnings distribution), depending on the country. However, examining the quantile regression results, where one can examine the PC wage premium at various slices of the earnings distribution, those employees with earnings in the left tail (worse off) of the distribution would benefit much more than workers with median earnings. When examining counterfactual scenarios, such as what the earnings distribution would look like if all were paid according to the non-PC usage wage, we see that employees would be much more likely to fall into the left-tail (poor off, or “excluded” area) of the earnings distribution. This so-called “Increased Risk” of exclusion is at least 6% as in Germany and as high as 17% in Great Britain. Use of the Internet at the workplace does not generally seem to indicate a significant wage premium. Only in Norway could we identify a positive wage premium (on average) of 11.5%. All other countries had insignificant results where we find only a slight increased risk of economic exclusion, ranging from 0.4% in Germany to 6.3% in Israel. In addition we find virtually no effect for different kinds of PC usage suggesting that the effects for PC use in general have more to do with the kinds of people who tend to have/use PCs in the work place than the fact of PC usage itself. It also suggests that simply training workers in specific PC related skills will not have much effect on their earnings returns in most countries. It seems clear then that whilst something is going on in terms of a PC-use at work wage premium, precisely what is going on, and therefore what actions need to be taken is not at all clear. Based on the small size of these effects, it would seem that effects related to ‘PC usage’ as opposed to ‘Internet usage’ should dominate the debate at present. In our analysis of ICT use and social exclusion, we can identify a strong correlation. Based on many accepted indicators, standard in the existing literature, we calculate an overall index of social exclusion. We then examine the shape of the distribution of this exclusion index, for everyone and then for those who do not use ICT. We find a strong increase in social exclusion (left tail of the distribution), for those not using ICT. For instance, those not having a home PC are as much as 18.5% more likely to be considered socially excluded in Israel. This increases even more, to 35.9% for those who say they do not have a PC because the cannot afford one, or simply do not have the skills to use one (“exclusion” in the strict and narrow definition of the word)! There is a fair amount of variation between countries as to the effects of computers on social exclusion. In Germany, Italy and Norway there are hardly any effects of computer use on social exclusion. However there are very strong effects for Great Britain and Israel. It seems being able to communicate by mobile/cell-phone is associated with high levels of social inclusion, especially for Great Britain, Israel and Italy. Those without access to a cell phone in these countries are very likely to be excluded, based on our multi-dimensional exclusion index but in most cases are also a small group in the population (with the exception of Bulgaria). For instance, in Israel, those not having a cell phone (8.6%) have an increased risk of 44.1% of falling in the left tail (exclusion area) of distribution! In Britain and Italy the increased risk is around 26-27%. We can therefore see the value of ICT access and usage as an indicator of social exclusion but not necessarily a cause of it. It may simply be that the reasons for exclusion from ICTs are the same as those for

www.eurescom.de/e-living/index.htm Page 27 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD more traditional social exclusion – low education, income, high risk of unemployment, poor social communication, poor health and so forth. In addition having Internet access is not a strong indicator raising the possibility that access is rather more evenly distributed than might be supposed. Home Internet access alone, in contrast to low ownership of PCs at home due to reasons of cost, may therefore not be that significant as a policy problem. In some countries the main barrier to access for all in the ‘European Information Society’ may not be the cost of Internet access but the purchase and ongoing costs of the PCs required to use it. Thus focusing on ICT exclusion (the ‘Digital Divide’) and in particular ‘Internet exclusion’ alone without trying to directly tackle the more familiar socio-economic barriers will probably have little effect on social exclusion in Europe nor provide much progress against e-Europe objectives. We have been able to identify possible associations between ICTs, mean wages, the wage distribution and also and an index social exclusion. With additional waves of data, we will be able to test the robustness of these initial cross-sectional results, by allowing for controls for individual specific unobserved heterogeneity, found to be important in previous studies on the impacts of ICT. It is this analysis which may provide indications of causal relationships between ICT use and wage/social outcomes and thus provide evidence to support significant public investment in ‘ICT access for all’ in the name of social objectives.

www.eurescom.de/e-living/index.htm Page 28 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

Table 7: Key Components for Economic and Social Exclusion Analysis in "eLiving"

(a) Standard labor market indicators: Wages, work hours, industry, occupation, education Study: All (b) Use of a computer at the work place and since when Study: All (c) Use of a computer at home and since when Study: Krueger (1993), Haisken-DeNew and Schmidt (1999) (d) Intensity of Computer Use Study: eLiving innovation (e) Use of the Internet and intensity Study: Pischner, Wagner and Haisken-DeNew (2000) (f) Influential Job Characteristics Study: DiNardo and Pischke (1997), Entorf and Kramarz (1997) (g) Specific Computer Tasks Study: Krueger (1993), Entorf and Kramarz (1997) • Word processing • Web design or management • Spreadsheets / database • E-mail or internet • Design, analysis, or desk-top publishing • Programming/network systems management, PC support

(h) Specific Skills and their importance for the job Study: eLiving innovation • able to write computer programmes, • able to download a file from the web • able to construct a web page • able to send a file by email • able to cut and paste between programs • able to reboot a computer • able to copy a file to a floppy disc

(i) General Attitudes toward computers Study: eLiving innovation • Generally interested in new technologies • Computers are intimidating • Computers can be fun • Difficulty in understanding new technologies • Over-dependence on computers • Computers will make life easier • Computers are a necessary evil

(j) Leisure time activities Study: eLiving innovation (in connection with IT) • play sport, keep fit or go walking • go to watch live sport • go to the cinema, a concert, theatre or other live performance • have a meal in a restaurant or cafe, or go for a drink to a bar or club • attend activity groups such as evening classes • read newspapers or magazines • read books, whether fiction or non-fiction • meet with friends

www.eurescom.de/e-living/index.htm Page 29 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

Table 8: Social Exclusion Indices for Germany

Number Index Inclusion Home PC Internet PC Cell Phone Exclusions Percent Index Mean Access Access Work Access

3 0.3 50.6% 100.0% 0.0% 100.0% 100.0% 4 2.0 49.6% 100.0% 47.5% 100.0% 100.0% 5 3.0 50.1% 95.3% 13.5% 100.0% 96.3% 6 5.2 47.8% 86.2% 33.7% 91.5% 97.9% 7 9.2 46.1% 91.4% 29.7% 92.4% 93.5% 8 12.9 44.3% 84.0% 38.1% 87.5% 91.8% 9 13.2 42.1% 85.0% 38.3% 82.7% 88.7% 10 14.2 39.4% 67.2% 24.7% 77.2% 90.6% 11 14.9 37.2% 68.0% 33.3% 75.9% 85.2% 12 10.1 34.6% 56.8% 18.3% 80.0% 80.9% 13 5.3 30.7% 37.8% 15.8% 63.9% 73.7% 14 5.0 27.8% 43.7% 14.4% 67.3% 58.9% 15 2.3 24.5% 24.4% 3.0% 34.7% 58.7% 16 1.2 22.3% 41.3% 36.1% 33.8% 50.7% 17 0.8 18.2% 57.9% 20.2% 45.4% 61.4% 18 0.4 15.1% 21.9% 0.0% 41.4% 80.5% 19 0.1 5.9% 0.0% 0.0% 100.0% 0.0% 21 0.1 4.2% 0.0% 0.0% 0.0% 0.0% Source: Own calculations using Wave 1 of the eLiving data set. Interpretation: Using the definition of Social Exclusion above, one groups all those individuals by the number of exclusions suffered by them. Thus 5.2% of the sample experienced 6 exclusions and had an average of 0.478 on the social exclusion index. However, they also had strikingly high levels of ICT. The horizontal lines indicate those effects supported by at least 5% of the sample (column labelled “Index Percent”) are reported.

Table 9: Social Exclusion Indices for Great Britain

Number Index Inclusion Home PC Internet PC Cell Phone Exclusions Percent Index Mean Access Access Work Access

1 0.2 60.7% 100.0% 83.2% 100.0% 83.2% 2 0.7 59.4% 100.0% 8.7% 100.0% 100.0% 3 2.3 57.1% 87.5% 36.4% 87.3% 100.0% 4 3.8 55.1% 93.4% 40.9% 91.8% 100.0% 5 7.9 54.2% 85.9% 35.1% 91.0% 98.5% 6 11.8 51.9% 85.8% 37.1% 86.1% 96.1% 7 13.7 49.7% 81.5% 40.3% 79.0% 96.3% 8 13.5 48.2% 77.5% 35.6% 80.0% 97.0% 9 12.3 45.8% 72.5% 33.6% 75.3% 97.1% 10 11.5 43.0% 51.5% 29.2% 69.4% 79.7% 11 7.6 39.8% 62.9% 28.5% 65.3% 79.6% 12 5.6 36.3% 41.9% 26.0% 61.5% 79.5% 13 3.7 32.9% 37.7% 15.5% 44.8% 74.3% 14 2.9 30.4% 19.7% 8.8% 37.8% 69.5% 15 1.4 26.9% 17.7% 0.0% 25.6% 49.8% 16 0.7 23.3% 34.6% 34.6% 16.7% 9.9% 17 0.3 20.3% 0.0% 0.0% 30.7% 30.7% 18 0.1 15.4% 100.0% 0.0% 100.0% 100.0% 21 0.1 4.3% 0.0% 0.0% 0.0% 0.0% Source: Own calculations using Wave 1 of the eLiving data set. www.eurescom.de/e-living/index.htm Page 30 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

Interpretation: Using the definition of Social Exclusion above, one groups all those individuals by the number of exclusions suffered by them. Thus 7.9% of the sample experienced 5 exclusions and had an average of 0.542 on the social exclusion index. However, they also had strikingly high levels of ICT. The horizontal lines indicate those effects supported by at least 5% of the sample (column labelled “Index Percent”) are reported.

Table 10: Social Exclusion Indices for Israel

Number Index Inclusion Home PC Internet PC Cell Phone Exclusions Percent Index Mean Access Access Work Access

0 0.1 54.3% 100.0% 100.0% 100.0% 100.0% 1 0.1 53.5% 100.0% 100.0% 100.0% 100.0% 2 0.4 52.0% 100.0% 26.3% 100.0% 100.0% 3 2.5 50.4% 90.2% 22.4% 70.0% 100.0% 4 3.8 48.9% 98.3% 34.3% 95.0% 100.0% 5 4.0 47.0% 94.4% 42.7% 84.5% 100.0% 6 6.1 45.7% 94.9% 49.0% 76.6% 94.0% 7 7.6 44.5% 89.0% 55.1% 81.7% 100.0% 8 10.7 42.4% 82.7% 37.8% 78.8% 100.0% 9 8.2 40.5% 71.7% 37.3% 71.1% 100.0% 10 10.4 37.9% 74.8% 37.4% 73.3% 95.3% 11 10.7 35.1% 71.2% 31.4% 65.0% 97.0% 12 10.9 32.4% 63.3% 23.0% 54.2% 90.0% 13 7.3 29.6% 56.2% 28.2% 63.6% 86.1% 14 6.2 27.3% 42.9% 13.3% 36.1% 83.7% 15 4.1 24.1% 26.6% 6.6% 32.4% 72.1% 16 3.6 21.5% 40.2% 9.6% 24.7% 55.3% 17 1.5 18.4% 22.4% 3.4% 18.4% 28.9% 18 1.3 14.7% 33.7% 9.8% 50.9% 97.9% 19 0.5 11.8% 29.7% 20.4% 20.6% 19.5% 20 0.1 8.5% 100.0% 0.0% 52.6% 47.4% 21 0.1 4.4% 0.0% 0.0% 0.0% 100.0% Source: Own calculations using Wave 1 of the eLiving data set. Interpretation: Using the definition of Social Exclusion above, one groups all those individuals by the number of exclusions suffered by them. Thus 6.1% of the sample experienced 6 exclusions and had an average of 0.457 on the social exclusion index. However, they also had strikingly high levels of ICT. The horizontal lines indicate those effects supported by at least 5% of the sample (column labelled “Index Percent”) are reported.

www.eurescom.de/e-living/index.htm Page 31 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

5 Bibliography Autor, D.H, L.F. Katz and A.B. Krueger (1998): "Computing Inequality: Have Computers Changed the Labor Market," Quarterly Journal of Economics, 4, 1169-1213. Bell, Brian D. (1996): "Skill-Biased Technical Change and Wages: Evidence for a Longitudinal Data Set," University of Oxford, mimeo. D’Ambrosio, C, F. Papadopoulos, and P. Tsakloglou (2002): “Social Exclusion in EU Member-States: A Comparison of Two Alternative Approaches”, Working Paper from Bocconi University, Milan, Italy. Dekkers, Gijs (2002): ”Poverty, Dualisation and the Digital Divide” in Cammaerts, B., L. Van Audenhove, G. Nulens and C. Pauwels, Eds., Beyond the Digital Divide, VUB-Press, Brussels. DeNew, John P. and Christoph M. Schmidt (1994): "The Industrial Structure of German Earnings 1980- 1990," Allgemeines Statistisches Archiv 78, 141-159. DiNardo, John E., N. Fortin, T. Lemieux (1996): “Labor Market Institutions and the Distribution of Wages: 1973-1993,” Econometrica, September 1996. DiNardo, John E. and Joern-Steffen Pischke (1997): "The Returns to Computer Use Revisited: Have Pencils Changed the Wage Structure Too?," Quarterly Journal of Economics, February, 291-303. Entorf, Horst and Francis Kramarz (1997): "Does Unmeasured Ability Explain the Higher Wages of New Technology Workers?," European Economic Review 41, 1489-1509. Entorf, Horst and Francis Kramarz (1994): "The Impact of New Technologies on Wages: Lessons from Matching Panels on Employees and their Firms," CREST-INSEE-Paris Working Paper No. 9407. Haisken-DeNew, John P. (1996), "Migration and the Inter-Industry Wage Structure in Germany," Springer Verlag, Heidelberg. Haisken-DeNew, John P. and Christoph M. Schmidt (1997): "Inter-Industry and Inter-Region Differentials: Mechanics and Interpretation," The Review of Economics and Statistics, August, 79, 3, 516-521. Haisken-DeNew, John P. and Christoph M. Schmidt (1998): "Industry Wage Differentials Revisited: A Longitudinal Comparison of Germany and USA (1984-1996)," IZA-Bonn Discussion Paper No. 98. Haisken-DeNew, John P. and Christoph M. Schmidt (1999): "Money for Nothing and Your Chips for Free: The Anatomy of the PC Wage Differential" IZA-Bonn Discussion Paper No. 86. Haisken-DeNew, J. P., and C. M. Schmidt (2001): "Brothers in RAMS: Diffusion of the PC and the New Economy ", DIW-IZA mimeo. Frijters, Paul, John P. Haisken-DeNew and Michael A. Shields (2001): "The Value of German Reunification: German Reunification and Life Satisfaction", IZA Bonn Working Paper #XXX. Hübler, Olaf (2000): "All Goes Faster but Lasts Longer: Computer Work and Overtime Work", ifo Studien, 46 (2), 249-271. Krueger, Alan B. (1993): "How Computers have Changed the Wage Structure: Evidence from Microdata, 1984-1989," Quarterly Journal of Economics, February, 33-60. Papadopoulos, F. and P. Tsakloglu (2002): “Social Exclusion in the EU: Quantitative Estimates and Determining Factors”, mimeo, Athens University of Economics and Business. Percy-Smith, J. (2000): “Introduction: The Contours of Social Exclusion”, in Percy-Smith, J. (Ed.) Policy Responses to Social Exclusion: Towards Inclusion”, 1st Edition, Open University Press, Buckingham. Pischner, Rainer, Gert G. Wagner and John Haisken-DeNew (2000): "Computer- und Internetnutzung hängen stark von Einkommen und Bildung ab", DIW-Wochenbericht 41/00, DIW Berlin. Raban, Y., Soffer, T., Mihnev, P., Ganev, K. (2002) e-living D7.1 ICT Uptake and Usage: A Cross-Sectional Analysis. E-Living Project Deliverable. www.eurescom.de/e-living. Shleifer, A (1986): "Implementation Cycles", Journal of Political Economy, 94, 1163-1190. Townsend, P., (1979): Poverty in the United Kingdom, Hammondsworth, Penguin. www.eurescom.de/e-living/index.htm Page 32 of 33 PUBLIC e-living-D7.3-Social-and-Market-Exclusion-Issue-1.0-MSWORD

Townsend, P., (1993): The International Analysis of Poverty, London, Havester Wheatsheaf. Tsakloglu, P. and F. Papadopoulos (2001): "Identifying Population Groups at High Risk of Social Exclusion”, IZA Bonn Discussion Paper 392, Bonn, German Whelan, B. and C. Whelen (1995): “In What Sense is Poverty Multidimensional?”, in Room, G. (Ed.) Beyond the Threshold, 1st Edition, The Policy Press, Bristol Zajczyk, F. (1995): “Between Survey and Social Services Analysis: an Inquiry on Two Lines and Three Levels” in Room, G. (Ed.) Beyond the Threshold, 1st Edition, The Policy Press, Bristol

www.eurescom.de/e-living/index.htm Page 33 of 33 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

e-Living D7.4: AGE, GENDER AND SOCIAL CAPITAL- A CROSS SECTIONAL ANALYSIS Rich Ling (Telenor) Birgitte Yttri (Telenor) Ben Anderson (Chimera, University of Essex) Deborah Diduca (Chimera, University of Essex)

e-Living: Life in a Digital Europe, an EU Fifth Framework Project [IST-2000-25409] www.eurescom.de/e-living/index.htm PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

Table of Contents 1 Introduction...... 3 2 Age, Gender, Social Capital and ICTs ...... 3 3 Age, Gender and ICT use...... 4 3.1 Television ...... 4 3.2 Fixed line telephony...... 6 3.3 PC, Internet and E-mail use ...... 7 3.4 Mobile voice telephony...... 11 3.5 Mobile text messages (SMS) ...... 11 4 Social integration and ICT use ...... 12 4.1.1 Social capital and the Internet...... 13 4.1.2 Social capital and the mobile telephone ...... 14 4.1.3 Social Capital and Employment Status...... 15 4.2 Method and data...... 15 4.2.1 Informal social interaction ...... 15 4.2.2 Formal groups ...... 18 4.2.3 Close friendship ...... 21 4.3 Summary ...... 24 5 Analysis of Quality of life...... 24 5.1 Quality of Life – a descriptive analysis ...... 25 5.2 Is having household Internet access or using the Internet associated with higher QoL?...... 27 5.3 What are the best predictors of perceived quality of life and do ICTs and social capital have any place?27 6 Conclusion ...... 29 6.1 Age, gender and ICT use ...... 29 6.2 Social integration and ICT use ...... 30 6.3 Quality of life analysis...... 31 6.4 Recommendations for future analysis...... 32 7 Bibliography...... 32

www.eurescom.de/e-living Page 2 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

1 Introduction This report constitutes the second deliverable from the e-Living workpackage 6 – Family, Gender and Youth. Building on the literature review provided by Chapter 5 of Anderson (2001), this report uses wave 1 of the e- Living survey to analyse the relationships between patterns of ICT use, social capital and quality of life. We tackle these issues from a European social perspective by first reporting patterns of access to and usage of ICTs with respect to age and gender (Section 3), then by analysing how these distributions manifest themselves in patterns of social capital (Section 4). We also take account in this analysis of employment status which is known to be a major contributor to quality of life and, we hypothesise, to social capital. Finally we follow this thread through to the analysis of perceived quality of life and its relationship with social capital and thus patterns of ICT usage (Section 5). We therefore draw attention to the potential affects of increased social capital on quality of life and by examining the role of ICT usage in increased social capital, we start to assess whether participation in the information society might, or might not, be significant in affecting perceived quality of life. We feel that this analysis is timely because recent European Commission policy documents (e.g. COM (2002) 263) appear to focus intensely on public services, e-business and the creation of a knowledge economy to further commercial rather than social objectives. Given that a key recommendation of the High Level Expert Group on the European Information Society (HLEG (IS), 1997) was to focus on support for social participation (via social and community interactions) and quality of life, we hope that this report can help to re-start debate on this issue. Given that the results reported here can inform this debate and, in particular, suggest a new approach to the ‘digital en-skilling’ of the youth segment, we see the potential readers of this report as researchers and policy makers with the IS Directorate and the groups of expert advisors who advise on the various e-Europe action plans. We would also expect this report to be of interest to those who are interested in issues of social and digital exclusion at the national government level who may be unaware of the links between social communication and quality of life. Finally we would also expect the report to be of interest to social policy, marketing and innovation strategy groups within a range of service providers because the results point towards certain kinds of services and their significance for particular market segments. In this report we restrict ourselves to analysis of individuals rather than household or family units due to restrictions in the e-Living wave 1 data. Following wave 2, which will collect data on both respondents and their partners (if present), we will be able to extend this analysis to interactions within the household. 2 Age, Gender, Social Capital and ICTs In recent years there has been much debate and not a little research investigating the relationships between ICT use, social capital and perceived quality of life. This was to some extent triggered by Castells’ view of the networked society (Castells, 1996); by discussions of the social potential of ICTs (Kraut et al, 1998); a series of moral panics concerning the ‘de-socialising’ potential of ICTs, and commentary on the middle ground between the two such as Putnam, 2000. The debates have continued in, for example, a recent issues of American Behavioral Scientist (Haythornwaite, 2001) and a review of evidence to date can be found in Wellman (2001). According to Putnam, social capital describes the social networks of the individual along with the various webs of reciprocity. Whereas physical capital refers to physical objects and human capital refers to properties of individuals, social capital refers to connections among individuals–social networks and the norms of reciprocity and trustworthiness that arise from them. In that sense, social capital is closely related to what some have called “civic virtue.” The difference is that “social capital” calls attention to the fact that civic virtue is most powerful when embedded in a dense network of reciprocal relations. A society of many virtuous but isolated individuals is not necessarily rich in social capital (Putnam 2000, 19).

The argument is that as people interact with others in their local milieu, they develop forms of interaction and a sense of identity based on the type and extent of the interactions. They can, in turn, rely on these relationships in other contexts. In this way social capital facilitates the functioning of society (Kavanaugh and Patterson 2001, 497). Clearly, the concept is not a new one. Rather it has been a point of discussion since the early parts of the 20th century for academics as diverse as Jane Jacobs, Pierre Bourdeau, James Coleman and Claude Fischer (Putnam 2000, 19-20).

www.eurescom.de/e-living Page 3 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

Putnam, in particular has examined social capital and its fate in modern North American society. In his work he looks at a broad variety of indicators in order to determine how social capital is faring. He includes items that range from participation in formal organisations, informal social interactions such as dinner parties and local neighbor coffee klatches etc. His general finding is that there are both cohort and life phase effects that are conspiring to reduce social capital in the US. When looking at the age cohort issue, the replacement of the “depression” children’s generation with the more “me” oriented post war, and X generations is having a deleterious effect. Putnam asserts that the tremendous effects of the depression and the Second World War served to develop a generation of persons who were characterised as being particularly given to social interaction. Their passing has not been replaced by subsequent generations with the same focus. However his book was written before September 11th 2001, and it is possible that the events of that date will have long term effects on social capital within the US, perhaps even similar to those of the Depression and the Second World War. While the cohort effect is one impact on social capital, Putnam also examines the effects of various other developments in society. He places particular significance on the effects of suburbanisation and the television. He suggests that the atomising effects of these two post war developments have had the effect of reducing one’s interest in more general social interaction. The greater distances one must travel, as demanded by suburban life, and the time that is taken from other possible social intercourse, as in the case of the TV, have conspired to reduce the social capital in Putnam’s estimation. Research in Europe however, suggests that this decline in social capital is not universal (Johnston 2001). In particular, Johnston has found that the generational-replacement process of declining trust seen in the US is not found in Britain, and neither is the decline in membership of organisations. It is also worth noting that ‘Much of the social capital research is carried out in America and the concept has tended to be exported wholesale, which ignores the cultural context of its conceptualisation’ (ONS, Oct. 2001)

3 Age, Gender and ICT use As described in Raban et al (2002) wave 1 of the e-Living survey gathered information on the use of various ICTs (Internet, e-mail, PC, mobile telephone, traditional fixed telephone and the television)1. That report provided some basic analysis of ownership and usage by age and income. Here we focus on the relative importance of age and gender on usage. Generally we will move from the most traditional to the most recent of the technologies.2 We focus on usage patterns of those who have access on the basis that these patterns indicate actual usage and may be some sort of predictor of future usage patterns given age cohort effects and increasing penetration. Each section concludes by analysing the relative contributions of socio-economic variables, age and gender to these usage patterns using new regression models or those already reported in (Raban et al, 2002).

3.1 Television The data shows that of the time based analyses, television viewing was one of the heaviest ICT uses in the sample – only some 3.2% of respondents reporting no usage at all and this includes those who live in a household without a TV (7.7% of Bulgarian households, 5.7% of Israeli households but less than 3.5% in other countries have no TV). Access then is not a significant issue in any of the countries surveyed and the TV apparently has a well established position in society. The data shows that the respondents in the sample used the TV for over two hours a day on the average.

1 The e-Living wave 1 survey instrument has been published as D4: Wave 1 Questionnaire. See http://www.eurescom.de/e- living/deliverables/e-Living-D4-Wave-1-Questionnaire-FINAL.zip.

2 A note on the analysis. In some cases we compare groups on the amount of time used on a particular technology. In other cases it is the number of messages sent, as with e-mail and SMS. This is based on the type of material included in the questionnaire. www.eurescom.de/e-living Page 4 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

250

200

150 Women Men 100 Mean minutes perday 50

0 16- 20- 25- 30- 35- 40- 45- 50- 55- 60- 65- 70- 75- 80- 85+ 19 24 29 34 39 44 49 54 59 64 69 74 79 84

Figure 1: Mean minutes watching TV per day (pooled sample, base = whole sample)

In addition, this is the only ICT in our data where use appears to increase with age. While the youngest respondents viewed TV slightly more than the young adult and adult groups, when one moves into the ranks of the older adults and particularly the retirees, one sees that there is a significant increase in the time used for TV viewing. The data shows that 45 – 49 year olds viewed TV an average of 126 minutes a day where the 70 – 74 year olds viewed it 177 minutes a day.

Table 1: Results of linear regressions of socio-economic variables on mean minutes per day watching TV3. Figures given are standardised co-efficients (Beta values)4.

UK Italy Germany Norway Bulgaria Israel Age -0.005 0.087* 0.053 -0.070 -0.077 0.143*** Male 0.013 -0.053 -0.006 -0.027 -0.011 0.004 Having a University degree -0.152*** -0.076* -0.097*** -0.185*** 0.058 -0.181*** Household monthly income in 8 categories -0.072* 0.023 -0.132*** -0.007 -0.019 0.045 Number of consumer electronics items5 0.036 -0.020 0.150*** 0.001 0.148*** 0.000 Number of cars -0.131 -0.044 -0.153*** -0.060* 0.008 -0.009 In work -0.267*** -0.154* -0.111*** -0.127*** -0.098** -0.108*** Retired -0.058*** -0.005 0.025 0.072 0.022 0.013

R sq 0.149 0.071 0.110 0.082 0.032 0.063 Given the high incidence of TV use in all countries, we would expect that a model with just socio-economic factors has little predictive power. This proves to be the case in all countries except Germany and, in particular, the UK. The factors associated with usage tend to be broadly similar across the six countries with those who are more highly educated and those in work watching less TV, perhaps due to having less leisure (i.e. non-work) time. This is confounded in the UK at least by the result for retirees which we might expect to show a positive relationship but which shows a small but significant negative one. When other variables are controlled, age has a positive relationship in Israel and Italy suggesting that the differences depicted in Figure 1 may be country specific. Higher income has a negative effect in the UK and Germany, perhaps

3 This analysis is a variant of that reported in D7.1 which did not include gender as an explanatory variable

4 *** p < 0.001, ** p < 0.01, * p < 0.05. These values apply through the report.

5 Washing machine, Dish washer , Microwave , Compact disc player/music system, Video camera, Video player/recorder, Digital camera, DVD player or drive www.eurescom.de/e-living Page 5 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc linked to the paid work/education result, whilst a different relative measure of wealth (number of consumer electronics items) shows a positive relationship in Germany and Bulgaria. Gender makes no difference at all.

3.2 Fixed line telephony Given that five of the six countries were surveyed using CATI it is not possible to assess “digital divide” issues associated with fixed line telephony access in these five. In planning fieldwork, it was estimated that some 50% of the homes in Bulgaria would have access to traditional PSTN telephony and so CAPI was used. However since the e-living survey in Bulgaria did not collect data on whether or not an individual had access to a fixed line phone, we can only estimate this indirectly via non-usage and non-response to the item asking about usage of the fixed line telephone. Figure 2 shows this estimation and it indicates some gender differences as well as some age trends in terms of non-access to (or non-usage of) fixed line telephony in Bulgaria. The rise towards the upper end of the age range is particularly noticeable and further analysis of these patterns to assess the best indicators of non-access would be beneficial given the association of telephony with increased social participation as will be discussed below.

30%

25%

20%

Men 15% Women

10% % of each age group 5%

0% 75+ 16-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74

Figure 2: Percentage of men and women in each age range in Bulgaria who do not use (or do not have access to) a fixed line telephone (base = Bulgarian sample, weighted)

When looking at use, the two countries that distinguished themselves were Bulgaria and Israel. The former reported low fixed line telephone use (a mean of 8.6 minutes per day) and the latter reported high use (a mean of 42.7 minutes a day). The general mean for the entire sample was just under 27 minutes per day and Figure 3 shows the overall distribution by age and gender. This chart suggests major effects for both age and gender.

www.eurescom.de/e-living Page 6 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

50 45

40 35

30 Women 25 Men 20 15 10

5 0 Mean minutesMean per day telephone using fixedline 85+ 16-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84

Figure 3: Mean minutes per day using fixed line telephone (countries pooled, base = whole sample)

With the exception of Bulgaria, it is likely to be difficult to fit demographically based models due to the high penetration. Since the technology is available and used by virtually all groups in society traditional demographic variables provide only marginal insight. This is confirmed by the low r2 values in Table 2 for each country except Bulgaria although if each variable has a U shaped distribution as age appears to do (Figure 3), the linear model may be a poor fit in any case. However the results do show that controlling for other variables, men use the fixed line phone less in all countries. While this finding may play on the stereotype of women as chatters and gossipers, other analysis indicates that it is through this technology that a type of distributed care taking and familial administration takes place. Women make social calls to relatives (their own and those of their husbands), congratulate others on special occasions and coordinate various mundane tasks such as delivering and picking up children (Moyal 1989; Moyal 1992; Rakow 1988; Rakow 1992).

Table 2: Results of linear regressions of socio-economic variables on mean minutes per day using the fixed line telephone. Figures given are standardised co-efficients (Beta values).

UK Italy Germany Norway Bulgaria Israel Age -0.129*** -0.063 -0.122** -0.042 -0.142*** -0.183*** Male -0.225*** -0.117*** -0.180*** -0.115*** -0.073*** -0.225*** Having a University degree 0.014 -0.026 0.024 0.002 0.018 -0.071* Household monthly income in 8 categories -0.062 0.108** 0.035 0.076** 0.130*** 0.195*** Number of consumer electronics items 0.131*** 0.025 0.039 -0.037 0.183*** 0.023 Number of cars -0.078* -0.033 -0.099** -0.067* 0.100*** -0.119*** In work -0.082 0.156 -0.075 -0.017 -0.011 -0.039 Retired 0.020 0.001*** 0.006 0.064 0.094* 0.082

R sq 0.090 0.056 0.050 0.030 0.132 0.094 There is also a negative effect for age and this is a significant result in all countries except Italy and Norway. Education makes no difference except in Israel but household income has a positive effect in all countries except the UK and Germany suggesting that those who are least able to pay, use the telephone less. There are a range of effects for other measures of wealth (number of cars and consumer items) and small but positive effects for the retired in Italy and Bulgaria. What is perhaps most interesting from the regulatory point of view is the persistence of relative wealth indicators as key predictors of usage in several countries.

3.3 PC, Internet and E-mail use Uptake and usage of PCs and the Internet is described in some detail in Raban et al (2002) with respect to country, income, gender and educational background. That analysis showed how important these factors are for access with women, lower income and less well educated groups all showing lower uptake across all

www.eurescom.de/e-living Page 7 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc countries. Here we focus on how age and gender effect usage of PCs, Internet and email. PC use, use of the Internet and sending e-mail are three nested technologies. Currently a PC is generally needed to access the Internet at home and further, ‘Internet use’ is a term that often includes e-mail use. Thus there is a certain overlap between these categories. None-the-less, usage of these three technologies varies in its gender bias as Figure 4, Figure 5, Figure 6 and Figure 7 show.

160

140

120

100 Women 80 Men 60

40 Mean minutes per day per minutes Mean

0 65+ 16-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64

Figure 4: Mean minutes per day using a PC at home (counties pooled, base = all those who used a PC at home once a month or more)

Generally speaking men in the sample reported using the technologies more than women but these differences are marginal for email (Figure 6 and Figure 7) and only hold true for some age groups with respect to PC use (Figure 4) and some countries with respect to overall Internet use (see Raban et al (2002), Section 4.2 and Figure 5). There is a reduction in use as we move through the age range of the respondents although the effect seems to flatten or even become U shaped after 40-44, an effect also noted in Raban et al, (2002).

Table 3: Results of linear regressions of socio-economic variables on mean minutes per day using a PC at home (base = all those who use a PC at home). Figures given are standardised co-efficients (Beta values). Results for Italy and Israel not reported due to poor model performance.

UK Italy Germany Norway Bulgaria Age -0.195*** -0.109 -0.068 -0.334 Male 0.164*** 0.183*** 0.128*** 0.368* Having a University degree 0.041 0.023 -0.010 0.387* Household monthly income in 8 categories 0.000 0.011 0.016 -0.362* Number of consumer electronics items 0.006 0.131*** 0.049 0.096 Number of cars -0.055 -0.182*** -0.019 -0.014 In work -0.054 -0.077 -0.134*** 0.194 Retired 0.081 -0.049 -0.063 0.370*

R sq 0.059 0.073 0.037 0.386 Whilst the regression results in Table 3 again confirm the weakness of simple socio-economic models in predicting PC usage (with the notable exception of Bulgaria) they do at least confirm the gender differential for general PC use. Age turns out only to be significant in the UK. Having a degree is significant in Bulgaria as is household income and being retired a result which is not immediately explicable. Those in work tend to use the PC at home less although this result is only significant in Norway.

www.eurescom.de/e-living Page 8 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

100 90 80 70 60 Women 50 Men 40

Mean minutes per day per minutes Mean 20 10 0 65+ 16-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64

Figure 5: Mean minutes per day using the Internet at home (counties pooled, base = all those who used the Internet at home, see D7.1 Section 4 for country rather than gender differences)

Perhaps not surprisingly, similar patterns are found for overall Internet use as shown by Figure 5 and reported in Raban et al (2002, Section 4.2). However the fact that older users are still making almost as much use of the technologies as those aged 40-50 suggests that whilst the older sectors of society may have lower access (cf Raban et al 2002, Section 3) they do not necessarily make less use once they have them. It also seems to be the case that gender differences reduce with age for both PC and Internet use suggesting that gender differences in the younger groups may be more to do with lifestyle, lifestage and ‘perceived need’ factors than overall gender-defined attitudes to technology per se. These age and gender effects are less marked with respect to email where the distribution of respondent’s email contact with their social network (Figure 6) suggests that social communication itself is highest in the younger age groups (cf Figure 7) and declines with age. As we shall see, SMS may be acting as a substitute for younger respondents.

Women 3 Men

Mean frequencyMean 2

0 65+ 16-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64

Figure 6: Mean frequency of contacting friends and relatives by email (countries pooled, base = all those who used email at home, 7 = Most days, 4 = About once a fortnight, 0 = Never)

On the other hand this could be an age cohort effect such that as the current 20-30 segment matures, this pattern of email communication persists. What is noticeable is that the magnitude of gender difference appears smaller than for fixed line telephony (Figure 3) suggesting either that email encourages more communication by men or less by women. In either case email appears to be a less genderised communication medium than traditional telephony.

www.eurescom.de/e-living Page 9 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

2.5

1.5 Women Men 1 Mean frequency Mean

0.5

0 65+ 16-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64

Figure 7: Mean frequency of sending emails from home (countries pooled, base = all those who used email at home, 4 = More than 5 per day, 2 = , Between 1 and 5 a week, 1 = Less than 1 per week)

Table 4 shows the results of a similar linear regression model to those reported in earlier sections but for the frequency of communicating with friends and relatives by email. These results confirm that, controlling for other variables, age has a significant negative effect whilst being male does, in general, have a significant positive effect. This shows that there is still an effect for age when we control for the size of social networks which we might suggest decrease with age. This reinforces the view that it is the younger age groups who are the high ‘mediated’ communicators. Having a degree also has a positive (and quite marked) effect in most countries. Given that these data have all email users as their base, this may be suggestive evidence of the more geographically dispersed social networks of university graduates who find that email suits their communication needs.

Table 4: Results of linear regressions of socio-economic variables on the frequency of contacting friends and relatives by email (base = all those who used a email at home). Figures given are standardised co-efficients (Beta values). Results for Bulgaria should be viewed with caution due to there being far fewer email users in the sample (n=77), results for Israel are not included due to poor performance of the model.

UK Italy Germany Norway Bulgaria Age -0.168*** -0.244*** -0.283*** -0.316*** -0.299* Male 0.082 0.131* 0.121** 0.122*** 0.144 Having a University degree 0.180*** 0.242*** 0.074 0.189*** 0.342* Household monthly income in 8 categories 0.020 0.096 -0.051 0.070* 0.122 Number of consumer electronics items 0.133*** 0.165** 0.031 0.133*** 0.259* Number of cars -0.100* -0.054 0.049 -0.054 -0.081 In work -0.029 -0.052 0.011 0.005 -0.094 Retired 0.039 0.096 0.095 0.111*** Large social network (has more than 11 people 0.022 -0.009 0.049 0.061 0.079 outside immediate family can rely on)6 Frequency of engaging in outdoor leisure7 0.178*** 0.117 0.176*** 0.187*** -0.088 R sq 0.123 0.185 0.131 0.202 0.205

6 See survey item QTALK for categories

7 - play sport, keep fit or go walking, go to the cinema, a concert, theatre or watch live sport, have a meal in a restaurant or cafe, or go for a drink to a bar or club, attend activity groups such as evening classes, meet with friends. www.eurescom.de/e-living Page 10 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

Having a large number of consumer electronics is associated with heavy email use in all countries although the effect is not significant in Germany, whilst income makes no difference at all. In Norway, the retired are much heavier users of email, which may account for the slight increase in use in this age group shown in Figure 6. Interestingly whilst having a large ‘close’ social network has a positive association with email usage, this result is not significant in any country and may well be associated with ‘youth’ in any case as we shall see in Section 4.2.3. However the frequency of engaging in outdoor leisure is positively associated with heavy email use in all countries except Bulgaria (for which results are unreliable in any case due to small n). This suggests that those who have the most intense social life also make the most use of email, a finding that has been reported in other research and which is discussed below.

3.4 Mobile voice telephony Raban et al (2002, Section 4.1) analyse the uptake and usage of the mobile phone with respect to both SMS messaging and voice calls. Voice based mobile telephony seems to be the most broadly based of the new ICTs considered here. While the users – in particular the heavy users – are younger and more often male, the diffusion of use in society is more even than is the case of SMS (compare Figure 8 and Figure 9).

5 Women 4 Men

3 Mean calls per day 2

0 80+ 16-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79

Figure 8: Mean number of mobile calls made per day (countries pooled, base = all mobile owners)

The regression analysis reported in Raban et al (2002) confirms that older people make fewer calls and this result is significant in the UK, Norway and Israel whilst being female is associated with making fewer mobile calls in all countries except Bulgaria. Interestingly the number of close friends only makes a difference in the UK and Israel whilst household income has a significant positive association with heavy mobile calling in all countries except Bulgaria. It has been suggested elsewhere that the 20-30 segment is a particularly nomadic period in people’s lives and it could be argued that the technology is well adapted to this life phase. In addition, young adults may have a slightly better economic situation in terms of disposable income as they have begun to work but often have not encountered major expenses such as the purchase of a home. Thus, from both a life phase and also form an economic perspective it is plausible that voice mobile telephony is a popular solution. The gendering of the technology is somewhat more difficult to explain. While men seem more willing to use mobile voice telephony, women seem slightly more resistant. This finding may reflect the issues noted above with regard to fixed line telephony, namely that women commit more time and energy to telephony than men (Ling 1998; Rakow and Navarro 1993). Thus, the relative expense of mobile telephony as opposed to the desire for longer, more extensive telephone use may exclude women from the group of heavy mobile telephone users. There may also be issues of gendered display and an effect for work-use given that Raban et al (2002) report positive relationships between being in work and making more mobile calls in all countries except the UK and Germany.

3.5 Mobile text messages (SMS) SMS on the other hand seems to be the prototypical teen application (see Figure 9). Teens almost completely dominated the use of this service. Almost half of the teen users reported sending more than 5 messages per day. No other group was near this level of activity although 25% of the young adults reported this level of activity. As one moves through the age scale there is nearly a logarithmic decrease when looking at the use of this application even though these age groups do have mobile phones (cf. Figure 8 and www.eurescom.de/e-living Page 11 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

Raban et al, 2002)). Regression analyses reported in Raban et al (2002) confirm that older mobile owners in all countries send fewer SMS messages and that there is a strong relationship between being single and sending many SMS messages across all countries. This also holds true for being in work. The analysis also suggests a relationship between higher household income and SMS use and that in Italy and Norway at least, there is a positive relationship between the frequency of meeting friends and SMS use. All of these results also point toward significant use in the younger age groups. SMS is perceived as a relatively cheap and quick form of communication. While the interface is more difficult than, for example, e-mail, teens have mastered this and thus it is not a particularly difficult barrier. In addition, one can observe a specific type of SMS culture among the teens. It has become de rigueur to send and receive messages. Indeed it is a way that teens identify themselves and it is a defining part of the culture. One can ask if teens will grow out of SMS or if they will continue to use it as they mature. There are two considerations here. The first is that the teen period is a particularly important point in life wherein they emancipate themselves from their parents. In this period, contact with peers is important with regards to the establishment of an independent identity. Thus, it is easy to see that a cheap communication medium wherein teens can freely (read: avoid parental intervention) interact with peers is a useful tool in this process. This line of thought would argue that as the teen matures they will drop the use of SMS since the need for intense peer interaction decreases as one moves into adulthood and eventually parenthood. However the need for close co-ordination of family life by mobile-equipped parents and children may lead to increased use by parents over time. We will focus on this analysis in subsequent reports.

10 9 8 7 6 Women 5 Men 4 3 2 1 Meannumber messages per day 0 80+ 16-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79

Figure 9: Mean number of SMS messages sent per day (countries pooled, base = all mobile owners)

On the other hand, the teens of today are the first generation to have used SMS in their daily activities. Thus, it is a part of their communications repertoire. As they mature, SMS will likely be one of the ways that they feel that they can communicate with others as the need arises. Thus, it may be that in general that SMS will spread to other age groups through a cohort effect as current teens age, but that innovation with new services, and especially mobile services will continue in the teen segment. A final thought here is that the technology supporting SMS will change. As future generations of mobile telephony come to market, the limitations of the current SMS technology will fall. In addition, terminals will change the form of text input and perhaps also the types of things that can be sent. Thus, SMS represents a dynamic, and as of now, immature technology which is emerging from ‘play’ into ‘serious use’. 4 Social integration and ICT use Given these patterns of ICT usage, we now turn to an analysis of how their use in mediating personal social networks and wider social integration might enhance what has come to be termed social capital. In this section we discuss the possible relationships between ICT use and social capital as well as discussing the mediating effects of employment status which have also been shown to be important. We then conduct

www.eurescom.de/e-living Page 12 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc analyses to investigate the relative importance of ICT usage, socio-demographics and employment status on our measures of social capital. It seems that one of the contributions of this analysis is the combination of email and mobile telephony to the examination of social capital. Where previous analyses have generally examined the Internet’s contribution to one’s interaction with various social spheres, there have been few analyses of mobile telephony’s effect on this part of life (Mante-Meijer and al. 2001). 4.1.1 Social capital and the Internet At the most utopian levels, and in the earlier days of its diffusion, some suggested that the Internet will have virtually unlimited effect on our ability to develop not only social capital, but to address a panoply of social ills. The open nature of the system suggested that communities of interest will allow individuals to seek out like minded individuals and via these connections to develop a stronger sense of inclusion than is allowed given traditional physical constraints. Some have gone so far as to suggest that it will give rise to a new form of consciousness, will eliminate the need for war, expand resources, eliminate illiteracy, solve the energy crisis, achieve disarmament, topple dictators save the environment, provide us with an endless life span, eliminate the need for factories, crystallise participatory democracy and result ins a rich symbiosis of god and man, without the compulsion of power or law but by the voluntary co-operation of citizens (Masuda in Kumar 1995, 15). To date these predictions have not proved well founded. Putnam examines the potential impact of the Internet in the light of social capital. In this analysis he is less sure about its general effect (Putnam 2000, 170 - 180). On the one hand it can function to isolate people in the same way that he suggests TV does. On the other hand it can assist persons in their development of communities of interest. Putnam does not address the issue of mobile telephony since it does not have the same diffusion in the US as it has in Europe. Beyond Putnam’s preliminary analysis of the Internet’s impact on social capital there have been several other analyses of this issue. The work that really started the discussion of this issue was contributed by Kraut et al. Based on experiences with a small sample of new Internet users they outlined the possibility that use of the Internet can lead to social isolation and depression (Kraut et al. 1998). By way of contrast Katz, Rice and Aspden examined Internet use and social participation. They found that the social dimension is an important glue that holds together the task oriented aspects of various computer mediated interaction (2001). In addition, they found that longer term Internet use is found to be associated with greater levels of social interaction. An analysis that echoes some of the findings of Katz et al is Kavanaugh and Patterson’s analysis of the Blacksburg Electronic Village. They found that the longer people were involved in the Blacksburg Electronic Village, the more likely they were to use the Internet for various types of social capital building (Kavanaugh and Patterson 2001, 505). However, there is also the strong suggestion that computer networks such as that in Blacksburg thrive in communities were there is already extensive social commitment beforehand. It is the “the rich get richer” argument. The net in itself does not generate social capital; it only facilitates already existing tendencies. These analyses do not necessarily decompose the Internet but rather treat it as a unified whole. The Internet i, of course a general name for a whole set of activities that one can currently carry out via a PC. Within this general category one can include, for example, “surf” the world wide web (WWW), e-mail, download files, engage in chat groups, instant messaging and Usenet or one can interact with others via games. Often the term Internet and surfing the WWW are used somewhat interchangeably. Viewing home pages and retrieving information from the Internet is indeed a major draw for many people as the e-Living data shows (Raban et al, 2002). None-the-less, e-mail is by many accounts the most used function associated with the Internet. In addition, in the US, instant messaging is an integral part of teen culture. Keeping these distinctions in mind, Franzen suggests that surfing on the web results in small but significant reductions in the degree to which an individual engages in social activities with friends. By way of contrast, more use of e-mail increases the number of close friends (Franzen 2000). In a similar vein, Anderson and Tracey followed a group of individuals in a panel design. According to their data, and particularly in the case of new users, the use of e-mail increased from T1 to T2. Thus, the more experienced one was on the net, the more likely they were to use e-mail, the most firmly established social channel on the Internet in Europe (Anderson and Tracey 2001). The assertion that e-mail covaries with greater sociability does not go unchallenged. Like Kavanaugh and Patterson, Nie suggests that the Internet does not cause greater sociability, rather, those who are already

www.eurescom.de/e-living Page 13 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc sociable, i.e. those who are better educated, more financially at ease and those who are in the more active phases of their lives, are the most likely to also adopt the Internet. (Nie 2001). According to Nie: In sum, Internet users do not become more sociable because they have used the Internet, but they display a higher degree of social connectivity and participation because they are better educated, better off and less likely to be among the elderly. The point is that they report a steady decrease in social interaction with family and friends the longer they are using the Internet, in terms of either weekly hours or years online (despite the fact that they also report an increase in e-mail use!) (Nie 2001, 429)

Nie’s cross sectional analysis looks carefully at the total use of the Internet. However, when looking specifically at the most social of all Internet activities, i.e. e-mail, their analysis seems to take on a new dimension. Rather than being founded on material gathered in questionnaires it goes over to less structured types of analyses. Thus one is left with a certain sense that the general use of the Internet is deleterious to social interaction, but a critical reading of the material indicates that there may be different dimensions to the issue. 4.1.2 Social capital and the mobile telephone The mobile telephone shares many characteristics with the Internet. First, it is an artefact of the broader development of ICTs that has arisen in the past decades. Where traditional telephony and TV have been a part of the culture for longer periods of time, mobile telephony and the Internet have only become popularised recently. Perhaps because of this, only certain groups have adopted these technologies while others are more cautious in their use of the device. In addition, they both allow for asynchronous text based interaction. When looking at the traditional telephone, there is little doubt in the literature that the device has assisted in the maintenance of social connections (Putnam 2000, 166 - 169; Wellman 1996; Wellman and Tindall 1993). It is worth underscoring maintenance since the device does not generally allow for one to expand their social sphere in the same way that the relatively open Internet may. When comparing mobile telephony to the Internet there are, however many basic differences. These include the cost of becoming a user of the two systems. Whereas the cost of adopting the Internet almost always includes the purchase of a PC, the subscription to an Internet provider and the instillation of software and hardware components, the adoption of a mobile telephone is a far less expensive undertaking. Basically one can obtain an inexpensive telephone for as little as 50 – 100 Euro.8 Another difference is the degree to which the mobile telephone is more a point of personal display than is the equipment associated with the Internet. In addition, the mobile telephone allows of ubiquitous access where, at least up to now, using the Internet has been more fixed geographically. Finally, as noted above, many portions of the Internet, such as usenet groups, mailing lists and chat rooms are open systems wherein one can come into contact with other who were previously unknown. By contrast, mobile telephony allows one, in many cases to limit access (Fortunati 2002). Indeed, mobile telephone numbers are rarely provided in telephone catalogues and the users of pre-paid subscriptions are, in some cases, completely anonymous. Mobile telephone interaction therefore requires prior knowledge of the other. Much of the comment regarding the social consequences of the mobile telephone has focused on its capacity to facilitate co-ordination (Ling and Haddon 2001; Ling and Yttri 2002), and its role as a disturbing influence in society (Ling 1997; Ling 2002). Beyond this there has been some work done describing its potential as a social instrument among teens. Unlike the Internet, the mobile telephone’s potential to assist in the broader project of developing social capital is relatively untouched. This is likely because whereas the Internet is an open system, i.e. one can interact with other, unknown, persons via functions such as chat and Usenet, the mobile telephone requires that one seek out a specific person as with the traditional telephone. At the same time, engagement in the telephone call closes one off from other, co-present activities (de Gournay 2002). This can have a chilling effect on the ability to create face-to-face social relationships. In addition, the ease with which one calls, via both the fixed and the mobile telephone, might cheapen the total meaning of a phone call when compared with a visit. Thus, while the relatively open and unwieldy nature of the Internet is new and worthy of comment, the traditional nature of the mobile telephone means that this issue is not as relevant. Still, the fact that mobile telephony means that one can interact when ever and wherever one is, then there is the potential to cultivate

8 Indeed during subscription campaigns one could get a hand set and be paid 2-3 Euros. This of course assumes that one sign a contract committing to, for example, a minimum of 12 monthly ‘rental’ payments. www.eurescom.de/e-living Page 14 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc friendships regardless of time and place. This leads one to suggest that the mobile telephone could be seen as a tool to assist in the nurturing of social capital and the prevention of social exclusion. 4.1.3 Social Capital and Employment Status Recent social network theorising has also focused on whether people in employment have social networks of a different nature to those of the unemployed (Burt 2000, Wellman & Berkowitz 1997). Previous research has suggested that although the unemployed have large social networks, it is the types of ties within the network that can reinforce their social exclusion. Similarly, with people in paid employment, the ‘weak ties’ in their networks are extremely useful to them, and other research (Häußermann & Petrowsky, 1989; Gallie & Paugam 2000) has shown that people across Europe are most likely to find employment through informal social networks (which are often maintained by telephony) rather than through formal channels. This research showed that it was important for the newly unemployed to maintain various social networks so that the they could find out about work opportunities, could be recommended by personal acquaintances should a vacancy occur, could make informal enquiries and could be contactable by potential employers. The authors also note that the telephone was as important for the unemployed as anyone else in terms of making arrangements to meet. It was often through the phone that people were invited to a range of social events such as coffee gatherings and sports or family celebrations and so without one it was easy to ‘drop out’ of society. It may well be that these results can be extended to the mobile phone and that, given the mobile’s support for instant contact, those without mobiles may be beginning to suffer a new (and not often acknowledged) form of exclusion. Through careful examination of the social capital data we can examine whether or not social capital varies according to employment status and what part, if any, ICT usage may play in mediating this.

4.2 Method and data Wave 1 of the e-Living survey collected data on several measures of sociability. These range from the least intimate (membership in formal organisations), through leisure interaction, frequency of contacting friends and relatives via telephony (fixed or mobile) or email, the significance pof mobiles for social calling, overall satisfaction with social communication and finally the number of close friends. Thus, we gathered information on formal and informal social interaction as well as interaction with one’s inner circle of friends. Obviously, there are several types of social interaction that are not included in this range of possibilities. We did not gather information on familial interactions,9 neighbours, work colleagues, nor did we ask about so called ‘virtual’ friendships. Up to this point we have discussed the various dimensions of social capital and also examined the types of ICTs that were included in the analysis. The next sections provide a descriptive and then explanatory analysis of these social capital variables. In this analysis we focus on four measures of sociability: • Their participation in socially focused leisure activities • Respondents’ membership in formal clubs and organisations • Their satisfaction with their communications with friends • The number of people outside their family on whom they could rely We take these measures to be useful indicators of social capital. As we shall see not all of these variables proved reliable in that some seem to be interpreted differently in different countries. 4.2.1 Informal social interaction We asked about the frequency that one participated in a range of informal social activities. An index of active leisure use was developed using a selection from the six variables in the questionnaire battery 10 by summing the frequency of engaging in all activities. Thus a lower score represents less frequent activity.

9 Aside from the more demographically focused information of family type and the number of persons living in the home.

10 These were: - play sport, keep fit or go walking; go to the cinema, a concert, theatre or watch live sport; have a meal in a restaurant or cafe, or go for a drink to a bar or club; attend activity groups such as evening classes; meet with friends. www.eurescom.de/e-living Page 15 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

UK 15 Italy Germany Norw ay

10 Bulgaria Is r ael Mean aactive leisure score

0 80+ 16-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79

Figure 10: Mean active leisure score by country and age (all sample, weighted)

Figure 10 shows that the rate of engaging in an active leisure life is roughly the same across all the countries except Bulgaria and decreases with age. Teens reported the highest levels of active leisure; indeed their score on the participation index was approximately 25% higher than that of mature adults. By contrast, the elderly, and in particular the oldest age groups were less inclined to participate in these types of leisure activities which is to be expected given the likelihood of declining health and mobility. This chart in itself is a reminder that on this measure of social capital at least, the elderly are the least ‘included’.

Mean active leisure score 0 UK Italy Germany Norw ay Bulgaria Israel

In paid w ork Unemployed Retired from paid w ork altogether On maternity leave Looking after family or home Full-time student/at school Long term sick or disabled

Figure 11: Mean active leisure score by work status and country (all sample, weighted)

Figure 11 shows the same index but for the various employment status groups recorded by the survey. We can see immediately that those in full time education engage in more active leisure than any other group, matched only by those on maternity leave in Norway although with a cell size of 21 this latter result may be unreliable. The chart also shows that students apart, Bulgarians appear to engage in less active leisure than those in the other countries but that there may be few differences between the unemployed and those in paid work. There are also marked differences between those in work and the long term sick/disabled in each

www.eurescom.de/e-living Page 16 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc country except Israel11. It therefore partially confirms the view that employment status is important to some kinds of social capital and needs to be part of any explanatory model. We now present just such a model that looks for best predictors of active leisure from a range of socio- demographic (including employment status) and ICT usage variables (Table 5). These models perform reasonably well for each country and noticeably well for Bulgaria. In each they confirm that age, TV watching and being long term sick or disabled (except in Israel) have negative associations with active leisure. The finding that being long term sick or disabled is significant even when age is controlled shows the strength of this effect. The finding for TV is also reported in the USA by Putnam (2000). Those with higher incomes (strongest effect in Bulgaria and Israel) and those with a degree (especially Bulgaria) all engage in more active leisure. Retired persons in Germany appear to engage in more active leisure whilst those who look after the home do less except in Bulgaria12. Of the ICTs considered, frequency of making mobile calls shows a positive relationship to active leisure in Italy, Germany and Israel. SMS has less of an association except in Bulgaria but emailing friends and relatives has a positive association in all countries (and is significant in all but the UK). Interestingly fixed line telephony only has an effect in Bulgaria which has the lowest penetration of mobiles.

Table 5: Results of linear regression model predicting frequency of active leisure (informal social interaction). Figures are beta values.

UK Italy Germany Norway Bulgaria Israel Age -0.299*** -0.278*** -0.273*** -0.248*** -0.349*** -0.143** Male 0.054 0.158*** 0.025 -0.079** 0.129*** 0.093** Has a degree 0.050 0.042 0.063* 0.067* 0.107*** 0.002 Household income 0.085** 0.064* 0.114*** 0.039 0.166*** 0.128*** Unemployed 0.023 -0.019 0.053 0.034 0.007 -0.027 Retired 0.094* 0.017 0.129** 0.069 0.045 0.018 On maternity leave -0.017 -0.005 -0.050 0.029 -0.015 -0.030 Looking after family/home -0.049 -0.076* -0.009 -0.021 0.024 -0.074* Full time student 0.086* 0.037 0.084** 0.086** 0.164*** 0.082** Long term sick or disabled -0.097** -0.049 -0.062* -0.071** -0.070** 0.012 Partner/Married -0.037 -0.025 0.011 -0.023 -0.020 0.085** Minutes per day using TV -0.126*** 0.005 -0.145*** -0.158*** 0.101*** -0.014 Minutes per day using telephone 0.050 0.040 0.032 0.027 0.137*** 0.083 Minutes per day using PC -0.019 0.015 -0.002 -0.011 0.049* 0.048 Minutes per day using Internet -0.026 0-.040 -0.041 -0.056* -0.015 -0.023 Frequency of Mobile calls 0.052 0.101** 0.067* 0.038 -0.014 0.149*** Frequency of SMS -0.007 0.072* 0.045 0.047 0.083*** -0.007 Frequency of emailing friends and 0.078 0.114* 0.226*** 0.153*** 0.094*** 0.142** relatives Number of emails sent per day 0.107* 0.007 -0.064 0.041 -0.024 -0.028 R sq 0.257 0.287 0.262 0.223 0.446 0.227 Thus, younger men, who are better educated and who have a higher income, who are active users of mobile telephony and e-mail but not TV seem to be the most likely to report participating in active leisure activities. This is the loosest form of socialising that we will examine. It is not these informal activities that are associated with formal groups, rigid timetables, deep emotional commitment or strong rituals. There is flexibility in the way they are organised and in the commitment of the individuals and there is not the same routinised interaction that one might find when looking at more formalised types of participation. This informal type of socialising describes the situation of young adults, particularly those who have better education, well paying jobs and who we know already are the heaviest users of mobiles, SMS and email. It

11 ANOVA results (Tukey post-hoc tests) show significant differences between these groups (p < 0.01 in all cases) for all countries except Israel.

12 This may in part be due to the Bulgarian responses to the ‘work status’ question which recorded very few who were home-makers. This appears to be because Bulgarian women (in particular) do not categorise their activities in these terms and would have said that they were unemployed i.e. ‘not having a job’. www.eurescom.de/e-living Page 17 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc is a period of life wherein one has yet to establish the major commitments of home and family. Thus, one is available to for the type of informal socialising described by this variable. However, beyond simply having the capacity to participate in this type of lifestyle its cultivation can become a type of culture in itself. Thus, the young adult period goes beyond simply being something to experience. Rather it is a goal in itself (Frønes and Brusdal 2000). The culture of youth and the culture of young adulthood is in itself older than the Internet and the mobile telephone. The Yuppie culture of the mid 80’s, coffee bars and young adult oriented networking pre-date the mass acceptance of both of these technologies. None-the-less, the spontaneity allowed by, in particular, the mobile phone appears a useful tool in these situations. Thus, while at this point we cannot assert causality, the likely direction is that the demographic situation of the young adult men described here supports an active, but informally organised, leisure. In addition, the same demographic variables support the adoption and use of ICTs (excluding extensive TV use). The “social” ICTs, in turn contribute to the ability of these persons to participate in an active but informal leisure. This line of reasoning has been called “The rich get richer” approach to explaining the effect of ICTs on social interaction. That is, some persons who already enjoy an active leisure also are able to shape the use of various tools in a way that further supports a pre-existing inclination. Using the thinking suggested by the ‘ domestication’ theorists, a technology is observed and evaluated in terms of a pre-existing social context (Silverstone 1995; Silverstone and Haddon 1996; Silverstone, Hirsch and Morley 1992). Once the technology is procured, it, as well as the users of the technology, are progressively altered such that the technology and the user become house broken. Put into the context of the analysis here, the pre-existing social situation of young well- educated persons who commanded high incomes drives the model of social capital in this context. It is therefore apparent that the ICTs are probably not enabling more people from across society to engage in informal leisure, but that they are the medium of choice for those groups who would be ‘doing the most active leisure’ in any case. 4.2.2 Formal groups The survey also asked respondents how many formal social groups they were members of and a simple index of formal social activity was developed by summing the number of memberships per respondent13. It is important to note that these organisations represent a wide variety of groupings. In some cases they can demand active participation on the part of members, as in the case of school and local groups. On the other hand, many of these organisations only require that the members pay their dues once a year and, in return, receive a monthly or a quarterly magazine. Thus there is a wide gap in the actual commitment of the individual to the organisation and also a gap in terms of the degree to which the individual comes into meaningful social contact with others.14

13 These were: Any social or sport club (including Gym), a residents, school or other local group, a trade union, an environmental or animal welfare organisation, and any other political or campaigning organisation

14 This said, all of these organizations loaded onto the same factor when using a factor analysis. Thus there is likely at least some underlying similarly between the various groups and also the make-up of the individuals who join them. www.eurescom.de/e-living Page 18 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

1.8

1.6

1.4 UK 1.2 Italy Germany 1 Norw ay 0.8 Bulgaria Israel Mean 'joining' score 0.6

0.4

0.2

0 80+ 16-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79

Figure 12: Mean membership of formal organisations (all sample, weighted)

When considering the age distribution of membership the data indicates that formal organisations are primarily the realm of middle-aged persons and Germans and Norwegians appear to do a lot more of it than respondents in the other countries (Figure 12). The data shows that while there is relatively active participation among younger adults, it is not until one is in the middle years that this form of social interaction reaches its peak. Mature adults reported about 20% more formal member ship activity than teens and young adults. Mature adults were between 25 and 50% more likely to join formal organisations than those who were in their elderly years (notwithstanding the result for Israel) with a noticeable drop at retirement that may be driven by leaving an employment related Trade Union. As with informal social interaction, Bulgarians also report less formal group interaction.

1.5

0.5

Mean number of memberships of number Mean UK Italy Germany Norw ay Bulgaria Israel

In paid w ork Unemployed Retired from paid w ork altogether On maternity leave Looking after family or home Full-time student/at school Long term sick or disabled

Figure 13: Mean number of memberships by country and work status (base = all sample, weighted)

Figure 13 shows that in contrast to informal social interaction, those in work are most likely to be a member of formal social groups with the exception of those on maternity leave in Norway (although n for this group is small (21)). There appear also to be differences between the employed and the long term sick/disabled in most countries which is confirmed by statistical analysis.

www.eurescom.de/e-living Page 19 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

Applying the same linear regression model to participation in formal organisations produces less satisfactory results (lower R sq) than for informal social activities. In general age, education and being male (except Norway and Israel) have positive associations although given the inverse U shape of the age distribution, fitting a linear curve may not be appropriate.

Table 6: Significant predictors of participation in formal organisations (linear regression results, figures are beta values)

UK Italy Germany Norway Bulgaria Israel Age 0.082 0.023 0.118* 0.171*** 0.092* 0.210*** Male 0.034 0.063 0.154*** -0.069* 0.006 -0.017 Has a degree 0.135*** 0.085** 0.038 0.111*** 0.056* -0.059 Household income 0.066 0.081* 0.141*** 0.040 0.107** 0.076* Unemployed -0.056 -0.061* -0.039 -0.032 -0.165*** -0.131*** Retired -0.010 -0.045 -0.124** -0.138** -0.177*** -0.128** On maternity leave 0.008 -0.010 -0.008 0.025 -0.006 -0.026 Looking after family/home -0.046 -0.089* -0.014 -0.057* -0.061* -0.026 Full time student 0.004 0.008 -0.045 -0.055 0.009 -0.041 Long term sick or disabled -0.045 0.014 -0.024 -0.083** -0.029 -0.039 Partner/Married -0.007 -0.011 -0.009 0.019 -0.024 0.008 Minutes per day using TV -0.069* -0.002 -0.087** -0.111*** -0.019 0.015 Minutes per day using telephone -0.048 0.032 0.063** -0.019 -0.011 0.022 Minutes per day using PC at home -0.041 0.053 -0.084** 0.003 0.046 0.083* Minutes per day using Internet at home -0.010 0.015 0.042 -0.033 0.082* -0.016 Frequency of Mobile calls 0.010 -0.025 -0.021 -0.004 -0.018 0.176*** Frequency of SMS 0.080* 0.079* 0.020 0.001 0.029 -0.024 Frequency of emailing friends and 0.037 0.210*** 0.036 0.048 -0.074* 0.069 relatives Number of emails sent per day 0.132** -0.109 0.013 0.123*** -0.042 0.016 R sq 0.109 0.114 0.101 0.102 0.082 0.095 Of the variables we are interested in here, age has a significant positive association with being a member of formal groups in all countries except Italy and the UK although we should be aware of the apparent inverse U shaped distribution mentioned above. Gender only makes a significant difference in Germany and Norway (men are more likely to be members than women) whilst being retired shows a negative effect across all countries although again the result is not statistically significant in the UK and Italy. Household income and having a degree both have, in general, positive associations with being a member of formal organisations whilst being unemployed, being a ‘homemaker’ or being sick or disabled have negative associations although these effects are not statistically significant for all countries and their effects are less marked than for the informal social interactions. When looking at ICT use, as we might expect there were only weak connections. Greater TV use has a negative association in the UK, Germany and Norway whilst using the fixed line phone has a positive association in Germany. PC usage has no relationship except in Germany (negative) and Israel (positive); Internet use showed no real effect except for Bulgaria where penetration is lowest and the result may therefore be spurious. Usage of mobiles for calls had no relationship other than in Israel where, as reported elsewhere, usage is highest (Raban, 2002). Text messaging on the other hand had a significant positive effect in the UK and Italy. Use of email with family and friends showed a positive association in Italy and a negative one in Bulgaria (small sample) whilst the number of emails sent, which might be a more meaningful variable in this context, showed a significant positive association in the UK and Norway. As we have suggested above, participation in formal organisations appears to be a middle-aged phenomena which falls away during retirement. In addition, the analysis shows upper income and educated males are also relevant variables in some countries. The only “ICT” variable that produced a broadly significant positive relationship was e-mail. This may reflects the degree to which e-mail has begun to find its place as a mundane part of organising formal groups in those countries where penetration is highest – such as Norway. On the other hand it may be that in these countries, those with Internet access who use email a lot (their social networks members also use email) tend to be the ‘joiners’. Not surprisingly, as with informal social

www.eurescom.de/e-living Page 20 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc interactions, excessive use of the TV and joining organisations have a negative relationship in some countries and we would expect the results to be significant in the remainder with a larger sample size. However, there seems to be some positive, albeit weak, interaction between SMS and participation in formal organisations in some countries. This might be a weak indication that the electronic age is slowly moving out of the commercial world and has become common enough that voluntary and leisure organisation can begin to use the efficiencies and co-ordination opportunities associated with this technology. On the other hand it may simply be that users of these kinds of technologies also happen to be those that join organisations. 4.2.3 Close friendship The final measures of social integration we consider are the number of persons outside their family that the respondent considered to be close friends and the degree to which respondents felt they had good communication with friends. We should note that these are different phenomena and, as we shall see, may be interpreted differently in different cultures. In order to examine the former, we asked respondents to tell the number of persons outside their family they felt that they could “really count on to listen to you when you need to talk”.15 Obviously there are many different types of friendship, and indeed the nature of friendship changes through the life cycle. One can speak of near friends and also those who are acquaintances. One can speak of work colleagues, neighbours and, for the elderly and disabled, carers who also occupy various other types of dimensions when considering friendship. Thus all or some of these may be included in response to this survey item.

10 UK Italy 8 Germany Norw ay 6 Bulgaria Israel 4

Mean number of 'close friends' index friends' 'close of number Mean 2

0 80+ 16-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79

Figure 14: Mean ‘close friends’ index by age and country (whole sample, weighted)

Figure 14 shows that overall there appear to be no age related trends in this variable for all countries except the UK and Israel. In the UK, respondents in the 16-24 and 50-75 categories reported a higher score whilst in Israel it is the youngest and those in the 30-50 age group who score highest. However the pattern for Israel is so different in magnitude to the other counties in the latter group that this measure of social participation may have a different meaning in Israel. It is possible that Israelis interpreted the item to mean a wider social network, perhaps including ‘weak ties’ and thus we should expect out subsequent regression models to produce different results for Israel.

15 The data collected in the questionnaire is ordinal but non-scalar. In order to create a scale the midpoints of the ranges were used. www.eurescom.de/e-living Page 21 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

2 Mean number of 'close friends' index 0 UK Italy Germany Norw ay Bulgaria Israel

In paid w ork Unemployed Retired from paid w ork altogether On maternity leave Looking after family or home Full-time student/at school Long term sick or disabled

Figure 15: Mean ‘close friends’ index by work status and country (whole sample, weighted)

When considering work status we see a range of patterns within each country (the startling result for Italians on maternity leave could be due to a small cell size as n=6). For example Bulgarians who are in work report far higher scores than the unemployed whilst the pattern is reversed (although the difference is perhaps not statistically significant) in the UK and Israel. Turning to our other measure, we asked respondents to rate their communications with friends on a scale from 1-5. Unlike the previous measure, there was little evidence of an age effect and also little evidence of differences between countries other than a greater variability with age (Figure 16). The same holds true for employment status where no significant differences were found between, for example, those in work and those who were unemployed nor were there any between countries effects (not shown).

4 UK Italy Germany 3 Norw ay Bulgaria 2 Is r ael

1 Mean quality of commssocial score

0 80+ 16-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79

Figure 16: Mean ‘Quality of communications with friends’ by age and country (whole sample, weighted)

www.eurescom.de/e-living Page 22 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

Given that there is greater variability shown by responses to the ‘number of close friends’ index we focus on building a predictive model for this indicator. Table 7 shows the results of running such a model and from it we can conclude that these models are extremely poor predictors of this variable in every country (low R sq)16. This could be due to a number of factors including the inappropriateness of the linear model and the effects of unmeasured and un-modelled variables. Given the poor performance of the model, the results should be viewed with some caution. However it would appear that income has a positive association with the ‘number of close friends’ as does being male. This result for gender is interesting in that it is often suggested that women are more accomplished at cultivating social networks. It may be, however, the women have a higher ‘standard’ and thus to be considered a close friend by a woman one must really demonstrate their willingness to be available and thus their responses to this survey item are different to those of men. Being long-term sick or disabled has a significant and positive effect in the UK and Israel although there is a negative (but not significant) effect in the other countries. The positive association in the UK and Israel may reflect the presence of carers in the social networks of the disabled or ill who may be included in those who can be ‘counted on to listen’. As with the other social activity indicators, watching TV has a significant and negative association. Using the fixed line telephone has a positive (and strong) association in Bulgaria whilst using a mobile to make calls has a strong and positive association in Israel where, as we have seen, mobile usage is highest. Whilst the model for Israel performs rather better than the others (R sq = 0.075), the results are not as different as we might have supposed given our suggestion that the indicator was measuring a different phenomenon in Israel. We must therefore leave the analysis of the Israeli pattern of response to this indicator for future analysis.

Table 7: Results of linear regression model predicting ‘number of close friends’ index based on socio-demographics and ICT usage variables. Figures are beta values.

UK Italy Germany Norway Bulgaria Israel Age 0.086 0.072 0.106* 0.089 0.012 0.010 Male 0.096** -0.013 0.056 0.059* 0.078* 0.073* Has a degree -0.013 0.038 -0.032 -0.011 -0.028 -0.034 Household income 0.048 0.085* 0.054 0.029 0.115** 0.155*** Unemployed 0.014 -0.058 0.046 0.018 0.022 -0.043 Retired 0.055 -0.039 0.028 -0.035 0.042 -0.008 On maternity leave 0.045 0.040 -0.029 0.017 -0.007 -0.004 Looking after family/home -0.007 -0.091* 0.040 -0.005 0.026 0.017 Full time student 0.024 -0.022 0.050 0.032 0.040 0.032 Long term sick or disabled 0.087** -0.004 -0.003 -0.045 -0.038 0.061* Partner/Married -0.003 0.001 -0.035 -0.031 -0.016 0.020 Minutes per day using TV -0.101** -0.054 -0.049 -0.062* 0.024 -0.075* Minutes per day using telephone 0.047 -0.028 0.014 0.071* .0129*** 0.027 Minutes per day using PC at home -0.054 -0.019 0.003 0.023 -0.033 0.006 Minutes per day using Internet at home -0.032 0.055 0.001 0.016 -0.017 -0.046 Frequency of Mobile calls 0.066* -0.037 -0.014 -0.015 0.027 0.128*** Frequency of SMS 0.018 0.006 0.109 0.039 0.014 0.017 Frequency of emailing friends and -0.033 -0.060 0.024 0.080 -0.015 -0.034 relatives Number of emails sent per day 0.066 0.000 -0.019** -0.062 -0.022 0.004 R sq 0.044 0.030 0.027 0.024 0.036 0.075 In some respects this finding reflects the basic nature of intimate relationships. Given the extremely tight definition of friendship, i.e. trust in individuals in moments of need, all individuals reported that they had some of these relationships. In addition, the intense nature of the relationships may to some degree exclude mediated interaction. Thus, the number of persons one considers to be close friends does not co-vary with

16 An identical model was run on the ‘satisfaction with communication with friends’ index with similar poor results. The results are not reported here. www.eurescom.de/e-living Page 23 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc interactive ICT use. Thus social interaction with close contacts at the most elemental level does not appear to have been brought into the world of mediated interaction and the best predictors of such social capital have not been included in this model. At the same time TV use is an indicator of a particular type person who has less external social contact and may have a lower ability to develop and maintain these types of relationships.

4.3 Summary Overall then we have presented some preliminary analysis of the relationship between socio-demographics, work status, ICT usage and three different forms of social capital. We have discovered that in some countries, age and gender make a difference for some kinds of social interaction. We have also shown that in general employment status makes a difference to formal and in formal social interaction but not to links with close friends. More particularly we have highlighted the extent to which being disabled or long term sick is associated with lower social capital according the measures we have developed. It remains for future analysis to unpack what, if any, benefits accrue to these groups for those who use ICTs as compared to those who do not. Such analysis can be carried out using e-Living data provided cell sizes are reasonably large which will not be the case in some countries. We have found no evidence that ICT usage is increasing social capital for any particular groups. Rather we have evidence that for the most part, those with ‘high’ social capital also happen to be high users of certain kinds of ICTs (primarily email and mobile telephony), whilst those with ‘low’ social capital tend to be higher users of the TV. Neither of these results should be unexpected. It remains for future work using wave 2 to analyse any longitudinal change in these social capital variables within particular groups (such as the unemployed or sick and disabled) and any associations they may have with changes in ICT usage behaviour. In general the models we have used have performed well with respect to predicting participation in informal leisure, relatively well to predicting formal social activity and very poorly when applied to close social relationships. A clear recommendation is therefore to develop different models to account for these different indicators of social capital and to conduct an analysis of their relative contribution to overall quality of life. The next and final section of this report provides the latter by developing models of quality of life using socio- demographic, social capital and ICT access and usage variables. 5 Analysis of Quality of life Having analysed overall patterns of ICT usage by age and gender and then analysed the relationship between these patterns and ICT use we finally we turn our attention to the contribution these social capital variables make to perceived quality of life (QoL). The study of quality of life is a well known field of research generally focusing on how handicapped people live their everyday life and how medical treatment and care can improve and influence on the quality of life of the individual. A range of questionnaires have been developed to explore different dimensions of quality of life. In order to develop an assessment of quality of life that can be used cross-culturally, the World Health Organisation (WHO) study quality of life in realisation that the concept are related to culture and value systems. The University of Toronto, Centre for Health Promotion study quality of life of the individual in relation to the belief that the individual have physical, psychological and spiritual dimensions. By separating the individual life into three major life domains, Being, Belonging and Becoming, they use a Quality of Life Profile to measure all components, both health and well-being17. However this scale is extremely large and not suitable for a general socio-economic household survey such as e-Living. In contrast, Diener et al’s Satisfaction with Life Scale18 (Diener et al, 1985) scale consists of five questions to measure quality of life and a short set of questions were developed from this scale for use in the e-living questionnaire. These items asked respondents to agree/disagree via a Likert scale with these statements: • Overall the conditions of my life are excellent. • I have enough free time to do what I want. • The environmental conditions in my area are good.

17 See http://www.utoronto.ca/qol/concepts.htm and http://www.utoronto.ca/qol/profile.htm for further details.

18 See http://s.psych.uiuc.edu/~ediener/hottopic/hottopic.html for more information. www.eurescom.de/e-living Page 24 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

• I have good communications with friends. And for those in paid work only: • In most ways my working life is close to ideal. A detailed analysis of the last item (quality of work life) can be found in Brynin et al (2002). Here we examine the relationship between ICTs and QoL using the first four items. As the previous section has noted there has been extensive debate on the ‘effects’ of ICTs, and Internet based services in particular, on social capital, social participation and social time. We see the focus on ICTs as clouding the issue because the effects found are as likely to be a consequence of socio-demographic differences between (for example) internet users and non-users as they are a consequence of Internet use itself. Our conceptual model therefore runs roughly as follows: A person’s QoL score depends on a range of variables including several types of social capital, socio- economic situation, lifestage/lifestyle, personality and attitudes (see Frijters et al, 2002 for a review from the socio-economic perspective). Some ICTs may mediate some of these variables as we have seen in the previous section. Thus the ICTs themselves are not necessarily significant factors, but they may enhance behaviours that are associated with increased QoL. Our analysis will begin with a brief descriptive analysis of the distributions of QoL scores across the samples. It will then present a conceptual mistake. We will imagine a simplistic model where the link between ICT ownership (and usage) is directly linked to QoL. We will then analyse the relative contributions of ICTs and the kinds of socio-economic and social capital variables mentioned above to illustrate this fallacy and explain why a policy focus purely on access to ICTs as a way to drive up QoL is also fallacious.

5.1 Quality of Life – a descriptive analysis Before assessing the associations between ICTs, social capital and quality of life it is worth giving some descriptive analysis of the patterns of responses to these items across countries and social groups. Figure 17 shows the mean scores for each of these items across the six countries. ANOVAs for each item show significant differences between the countries for every one, even the last item which appears visually similar. In general Norwegians record higher scores for all aspects except free time where Germans came slightly higher. Respondents in all countries scored highest on their communications with friends but there are more interesting variations for some of the other items. For example Norwegians are more satisfied with their environmental conditions than any of the others whilst Bulgarians are the least. Indeed Bulgarians score lowest on all types of ‘life quality’ except for communications with friends.

Mean score 2

0 In most ways my Overall the I have enough free The environmental I have good working life is conditions of my time to do what I conditions in my communications close to ideal life are excellent want area are good with friends

UK Italy Germany Norway Bulgaria Israel

Figure 17: Mean scores for the quality of life items across all countries (High score = strongly www.eurescom.de/e-living Page 25 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

agree)

Of course these patterns also vary across different social groups as Figure 18 shows. People aged 25-54 in all countries are less satisfied with the amount of free time they have compared to those aged 16-24 or 75+. Older people in Israel were less likely to have good communications with friends whilst younger people in Bulgaria were less satisfied with their environmental conditions than older people. This is interesting as it suggests younger people in Bulgaria may be more sensitised to environmental issues than older people since their external environmental conditions are likely to be broadly similar.

UK Italy

5 5 4.5 4.5 4 4 3.5 3.5 3 3 2.5

2.5 Mean score Mean score 2 2 1.5 1.5 16-- 25-- 35-44 45-- 55-64 65-74 75+ 16-- 25-- 35-44 45-- 55-64 65-74 75+ 24 34 54 24 34 54

Ge rm any Nor w ay

5 5 4.5 4.5 4 4 3.5 3.5 3 3 2.5

2.5 Mean score Mean score 2 2 1.5 1.5 16-- 25-- 35-44 45-- 55-64 65-74 75+ 16-- 25-- 35-44 45-- 55-64 65-74 75+ 24 34 54 24 34 54

Bulgar ia Israel

5 5 4.5 4.5 4 4 3.5 3.5 3 3

Mean score 2.5 Mean score 2.5 2 2 1.5 1.5 16--24 25--34 35-44 45--54 55-64 65-74 75+ 16--24 25--34 35-44 45--54 55-64 65-74 75+

Key: www.eurescom.de/e-living Page 26 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

Figure 18: Age and country distributions of mean scores for quality of life items. Those aged 65+ are excluded form the quality of work life analysis due to small cell sizes.

The distributions of quality of work life are discussed in more detail in (WP7 report) in relation to the effects of various kinds of workstyles. Reliability analysis of these distributions for each country confirms that they cannot be meaningfully combined (alpha < 0.5 in all countries and is not improved by omitting one or other of the variables) into a quality of life scale. As a result the following analysis uses only the ‘Overall the conditions of my life are excellent’ scale as an outcome variable leaving analysis of the other scales for subsequent research. This has the advantage of enabling us to assess the contribution of other QoL scales to the variance of this overall scale.

5.2 Is having household Internet access or using the Internet associated with higher QoL? This question is relatively easy to answer. Table 8 shows the mean overall quality of life scores a) for those with and without PC based internet access in their households and b) those who did and did not actually use the Internet. The results have been broken down by country. There are significant differences between the quality of life scores for those who do and do not have household Internet access in the UK, Germany and Bulgaria. There are also significant differences between those that do and do not use the Internet (irrespective of where they use it) for the exact same countries. One or more internet-connected PCs in the household Uses the internet No (a) Yes (b) Difference (b-a) No (c) Yes (d) Difference (d-c) UK 3.859 4.125 0.266*** 3.852 4.103 0.252*** Italy 3.848 3.866 0.017 3.866 3.833 -0.033 Germany 3.923 4.047 0.124** 3.919 4.038 0.118** Norway 4.208 4.206 -0.002 4.218 4.201 -0.017 Bulgaria 2.102 2.950 0.848*** 2.052 2.760 0.708*** Israel 3.583 3.841 0.259*** 3.560 3.896 0.336***

Table 8: Mean overall quality of life scores for those with Internet access and those who use the Internet in each country. * signifies significance level of the difference between the means. N in b) and d) is small in Bulgaria (52 and 18419 respectively).

Thus the answer to our simple question of whether Internet access and use is associated with higher quality of life is therefore ‘yes for some countries’. We might therefore conclude that in some countries having access to and using the Internet could effect a person’s overall quality of life. But this would be an unsafe conclusion to draw based on this analysis alone because there may be other characteristics of internet-users in (for example) the UK which have a stronger effect such as employment status, income or age as previous sections in this report have demonstrated.

5.3 What are the best predictors of perceived quality of life and do ICTs and social capital have any place? To answer this question we construct a regression model for each country that attempts to predict overall quality of life from a series of variables, including having Internet access and using the Internet. The variables used in the model are wide ranging but are based either on the existing literature (cf. Frijters et al, 2002) or on hypothesised associations between the variable and overall quality of life in the context of our interest in ‘social capital’. As a survey concerned with socio-economic research on ICTs, e-Living did not collect data on psychological traits such as personality which are well know to effect life satisfaction (Kahneman et al, 1999). As a result the models developed here are unlikely to be as complete (i.e. predictive) as we would wish. Further, as models based on cross-sectional data, they can only uncover associations rather than causal relationships between the variables modelled. The results are reported in Table 9. Note that the last 3 items are the different measures of quality of life which may be likely to contribute to the overall perceived quality of life scale. They are included here to assess their contribution. Surprisingly they do not exhibit co-linearity with the overall quality of life scale when

19 There are more internet users than those who have an internet connected PC at home in Bulgaria due to work, school and university access. www.eurescom.de/e-living Page 27 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc the model is run for each country (Tolerance approximately 0.9, VIF approximately 1.3 in each case Field (2000), p 153). As we can see we find a range of effects in different countries and a range of model performance from reasonably good in the UK (R sq=0.270) to mediocre in Norway (R sq = 0.185). This may be due to the use of linear regression modelling that assumes linear relationships between dependent and independent variable. The most reliable negative associations across most countries were being male, long term sick/disabled, separated/divorced/widowed/un-married although there is some evidence of collinearity in these variables which requires further analysis. The most reliable positive ones were the frequency of informal social interaction/leisure, household income and the three other quality of life indicators of which one, social communication with friends, is explicitly related to social capital as we have seen above. It is interesting to note that very few of the ICT usage variables have proved to have any associations at all. There are no significant associations in any country for simply having ‘in principle’ access to an Internet enabled PC in the household nor for Internet usage. Heavy TV watchers in the UK have a slightly lower QoL whilst those in Bulgaria who send a lot of SMS messages have a slightly higher one. This may well be a lifestage-related effect, as we have seen before, given that there is a strong association between high QoL and being a student in Bulgaria. There is also a positive relationship between QoL and emailing family and friends in Bulgaria that may be an artefact generated by the small number of Internet users in this country. In general then we find no significant and reliable relationships between ICT access/usage and overall quality of life when a range of other variables are taken into account thus confirming our caution in accepting the results of Table 8 at face value. The contribution of the three other quality of life indicators to overall quality of life is clear with a significant positive association in all countries except for social communication in Bulgaria. Finally we can quite clearly see that the contribution made to quality of life by the frequency of engaging in active leisure far outweighs the effects of being a member of formal organisations, which has no effect at all, and the number of close contacts (significant only in the UK). Rather, informal social interaction is similar to the quality of communications with friends in its statistical relationship with quality of life suggesting that it is satisfaction with social communication rather than the number of contacts per se which is the most meaningful indicator of social capital in this context.

Table 9: Statistically significant (p < 0.05) beta values for regression model predicting overall quality of life. Blank cells indicate no significant effect.

UK Italy Germany Norway Bulgaria Israel Age 0.037 -0.077 0.069 0.148** 0.094 -0.035 Male -0.079** -0.004 -0.095** 0.024 -0.066* -0.012 Number in household -0.050 -0.023 0.082* 0.099** -0.029 -0.042 Has a degree 0.021 -0.030 0.006 -0.052 0.029 -0.009 Household monthly income 0.059 0.061 0.186*** -0.002 0.176*** 0.106** Number of consumer durables out of 8 0.043 0.039 0.017 0.015 0.108** 0.109** Unemployed -0.060* 0.048 -0.051 0.034 -0.165*** -0.065* Retired -0.029 0.087 0.017 -0.054 -0.146 0.015 On maternity leave 0.010 0.028 0.011 0.010 -0.021 -0.002 Looking after family/home 0.024 0.074* 0.041 0.021 -0.020 0.011 Full time student 0.027 0.016 0.041 0.051 0.121*** 0.049 Long term sick or disabled -0.100** -0.032 -0.077** -0.122*** -0.020 0.018 Married or partnered 0.068 0.041 -0.064 0.305 Widowed -0.066* -0.127 -0.002 0.007 -0.073 0.172 Separated or divorced -0.127*** -0.085 -0.046 -0.083** -0.095 0.128 Never been married -0.038 -0.091 0.079 -0.087** -0.065 0.270 One or more internet-connected PCs in the -0.063 -0.002 0.023 -0.068 -0.016 0.051 household Uses the internet 0.020 -0.036 -0.043 0.040 0.016 0.006 Has ISDN or broadband -0.036 0.032 -0.024 -0.011 0.014 0.013 Minutes per day using TV -0.076* 0.007 0.002 0.016 -0.016 -0.035 www.eurescom.de/e-living Page 28 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

Minutes per day using telephone -0.012 -0.001 0.024 -0.026 0.026 0.001 Minutes per day using PC 0.027 -0.008 0.015 0.006 0.023 -0.010 Minutes per day using Internet 0.043 -0.056 0.017 -0.023 0.027 0.019 Frequency of Mobile calls -0.036 -0.029 -0.006 0.052 0.036 0.047 Frequency of SMS 0.005 -0.024 0.003 -0.042 0.080** 0.016 Frequency of emailing friends and relatives -0.015 0.035 0.043 0.022 -0.083* -0.026 Number of emails sent per day 0.051 -0.016 0.002 0.047 -0.012 -0.037 Active leisure/ informal activity score 0.151*** 0.137*** 0.074 0.104*** 0.038 0.113** Number of memberships -0.016 0.013 -0.013 -0.026 0.002 -0.001 Number of close contacts index 0.091** 0.010 0.051 0.007 -0.002 0.022 I have enough free time to do what I want. 0.178*** 0.236*** 0.156*** 0.106*** 0.034 0.126*** The environmental conditions in my area are 0.170*** 0.092** 0.267*** 0.136*** 0.187*** 0.192*** good. I have good communications with friends. 0.218*** 0.169*** 0.122*** 0.235*** 0.207*** 0.220*** R sq 0.270 0.190 0.214 0.185 0.259 0.236 As traditional sociologists and economists might predict, ICT access and usage does not, in general, have any association with higher or lower overall quality of life when other variables are taken into account. Rather it is the familiar combinations of gender, disability or illness, wealth, being in paid work, communication with friends and amount of leisure time which show significant associations. More interestingly this analysis does not suggest that ‘cultural capital’ variables, such as education levels, and ICT mediated ‘social capital’ variables such telephony usage or emailing social networks have strong associations either. To return to our simple conceptual model, it does not appear that ICTs are directly enabling behaviours which can be associated with higher or lower perceived quality of life as specified in this model. It may be the case that ICT mediated behaviours not included in this model may have shown some effects. It may also be that in contributing to informal social capital (c.f. Section 4.2.1) ICTs may be indirectly affecting QoL. However our results show that much more work needs to be done to uncover the presumed links between social capital, ICT access, usage and overall quality of life. Unfortunately the use of cross-sectional data, as here, serves to confuse this issue because it cannot say anything about the direct (or indirect) effect, as opposed to association, of the acquisition and use of ICTs on overall quality of life. It may be that those who now (at time 2) use email to contact friends and family have a higher quality of life than they used to (at time 1) but it is still not significantly different from the rest of the population. In this instance the use of ICTs to media social behaviour would be having a positive effect on that group but this effect cannot be analysed using cross-sectional data because we do not know what has changed and for whom. Therefore the most immediate task is to base such analyses not on cross-sectional data, as here, but on longitudinal data which can associate actual changes in overall quality of life scores with changes in socio-economic conditions and/or patterns in ICT usage. This, of course, is exactly what analysis of e-Living wave 2 data will start to provide. 6 Conclusion

6.1 Age, gender and ICT use In summary, age is the demographic characteristic that best covaries with use of the new ICTs. When looking across a range of the ICTs, (overall internet use, use of e-mail, voice mobile telephony and text mobile telephony), the demographic characteristic that is most often significantly correlated is age. This is particularly true in the case of the various modes of mobile telephony. European youth and young adults have embraced this technology in a way that no other group can match. When looking at the strengths of the models, it is clear that mobile telephony is that technology wherein age (and also income) has the best covariance. In addition, youth also shows a significant affinity for traditional PSTN telephony. The only technology that shows the opposite effect is television. Here the models are weak, but they indicate that the elderly are slightly more inclined to use their time viewing television in some countries. The data also shows that, with the exception of traditional PSTN telephony women are not as heavy users as men with it comes to the use of ICTs. The data also shows that education and, in particular, income are relatively strong covariants of ICT use. Overall however what is notable about e-mail use is that its decline with age is less marked than, for example, SMS or mobile calling. The gender differences are also smaller. Thus whilst from a policy point of view we might be concerned with lower access rates for women and the elderly, their actual usage rates may be less of a cause for concern. www.eurescom.de/e-living Page 29 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

However these results have a further policy implication. We have seen how it is the youth segments of the surveyed countries who are the heaviest users of various ICTs and with a particular focus on their use for inter-personal communication. There is currently much policy consideration of ways to engage young people with ICTs to ensure that they have the skills and competencies required to prevent disadvantage (e-Europe 2002, COM (2002) 263). Currently these efforts focus very heavily on actions taken by and within schools and formal sites of ‘learning’. However it may well be that such actions will simply recapitulate existing patterns of exclusion and disadvantage to the extent that they are caused or mediated by structural and societal problems within education systems. In contrast our results show that a different route to ‘digital literacy’ could be to build on the phenomenal success of communication oriented ICTs in the youth segment and to focus policy and investment on creating a ‘digital communication’ society for young people in Europe. Not only might this evade familiar problems of engaging young people with ‘formal learning’ environments but it will place the usage of ICTs within a social rather than an informational context. This would have the advantage of playing to possible network effects of mass communication ICTs so that the dynamics of technology and ‘practice’ diffusion within social groups can be used to further the social policy aims.

6.2 Social integration and ICT use The data indicates that one can observe the greatest covariance between ICT and informal social interaction where mobile telephony/SMS and email appear to have a role. Obviously, this type of social interaction is also characteristic of a life phase. Thus it is likely that the demographic dimensions of informal social interaction were in place before the advent of Internet and mobile telephone mediated contact. None-the- less, these technologies have further facilitated the intensity of the social interaction. As noted above, this suggests that ICTs facilitate social interaction where there is already a core upon which to build The data only partially supports the suggestion that formal organisations can be more functional via the use of new ICTs whilst ICT use seems to be irrelevant to the extent of close relationships. Thus, the suggestions of some commentators that, for example, the Internet would flower into a self- sustaining social centre in itself seems to be unfounded. Rather, well functioning social groups that also use ICT to facilitate their interactions will potentially enjoy better co-ordinated and flexible informal social interaction. It is easy to interpret this as a positive direction, i.e. people will be able to seek out friends for informal social interaction at times that better suit them and in ways that are easier to accommodate. An intriguing additional assertion here is that ICTs can potentially also strengthen counter social tendencies in communities. On the one hand there is that part of society that is well educated, affluent and thus they supplement their connectivity with new technologies that take it to higher levels of interactivity. On the other there is another sector of society that is also well connected, but illicit. This can include gang of youths, drug pushers, prostitutes or terrorists (Ronfeldt and Arquilla 2001). The important thing here is that in both groups there are well-developed bases upon which the technology can build. Finally, the material here seems to indicate that ICTs are irrelevant for those who do not have a well developed network from before. If this is not in place, that is if one lives in an anomic situation without an established social network, or if one lives in a well routinised locally based social situation, then there is little need to use resources to streamline one’s broader social interactions. Quite simply, they do not exist and thus there is not need to bother with them. This comes through in William Gibson’s comments with regards to the Riots in Los Angeles. In an interview he said: A Radio Shack shop (ed. a chain of shops selling consumer electronics gear) was being looted. Next to that there was an Apple shop, and it was untouched. People wanted to steal portable TVs and CD players, not computers. I think this clearly indicated the gaps of culture, or simply the gaps of chances, in our society.(Salza 2002)

While ICTs represent a powerful opportunity to connect socially, the network and secondarily the understanding of the technology’s potential have to be in place before one can conceive of adopting the technology. Thus, it appears beyond the role of ICTs to create social networks in the general population. While certain types of sociability can be developed and even elaborated via mediated interaction, these are marginal in our society. The “chat” friendship that blooms into romance and marriage is the exception as evidenced by its ‘newsworthiness’. It is far more common that social interaction is founded on various forms of physically co- present contact (Ling 1998; Ling 2000). The Internet, e-mail, mobile telephony and SMS can all enhance these interactions once they are established, but one must establish them in the first place. This is not to

www.eurescom.de/e-living Page 30 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc say that relationships that are established via other channels cannot flourish and be as multifaceted as other types of relationships, only that these types of social interactions are in the minority. Further, in situations where social institutions are only precariously in place, one does not often find the support for the development of firm local sociability. Where middle-class and upper middle class parents can encourage their children to play soccer and visit the library while they network with other firmly implanted parents via various media, lower class, and impoverished parents do not have the same possibilities. If we shift to a Durkheimian turn of mind here, one can suggest that what characterises “nomic” societies is, to one degree or another, the presence of a common sentiment. If the communication established between [individuals] is to become real communion, that is to say a fusion of all particular sentiments into one common sentiment, the signs expressing them must themselves be fused into one single and unique resultant. It is the appearance of this that informs individuals that they are in harmony and makes them conscious of their moral unity. It is by uttering the same cry, pronouncing the same word, or performing the same gesture in regard to some object that they become and feel themselves to be in unison (Durkheim 1954, 230)

According to Durkheim, it is through these common rituals that the community becomes conscious of itself, and the individuals are aware of their membership in the broader community. Collins suggests that there are several component elements in the essential version of the Durkhiemian ritual. Collins, who was writing before the widespread use of Internet, noted the elements of 1) face-to-face presence of the group, 2) a common focus of attention, 3) shared emotions and 4) non-practical actions carried out for symbolic ends (Collins 1994, 206). There are two points to be made here. The first is the assertion that these types of binding rituals are missing in some social situations. Given that these basic building blocks of social capital and not in place, the more social group or groups are neither able to further elaborate the social forms that characterise the more “nomic” societies nor are they able to employ the potential advantages provided by ICTs in the further leveraging of this type of access. The second point is the degree to which these types of social ritual can be shorn from their physical basis and taken into the exclusively virtual world. The findings here indicate that a concrete, here and now orientation is essential in first establishing social contacts and second in maintaining the group. This does not mean that parts of both the establishment and the maintenance can be carried out via mediated channels(Ling 2000). However, it seems that one has to touch base occasionally. Without the shared and focused experience of a common meeting, the impulse to maintain the social relationship runs out.

6.3 Quality of life analysis The analysis presented above reveals a clear result. There are few, if any, reliable associations between ICT access and usage and overall quality of life. Instead it is the usual combinations of gender, disability or illness, wealth, being in paid work, communication with friends and amount of leisure time which show significant associations. These results are different for different countries suggesting different policy implications. If we were interested in increasing quality of life in the UK for example we might focus on leisure and disability related policies as well as policies to improve environmental conditions, life balance and communication with friends. In Bulgaria on the other hand we might focus on wealth creation, employment and the environment. This is, in itself, a useful result because it shows how important it is to build a conceptual model linking increased ICT access and use to socio-economic outcomes as a basis for informed policy-making. It is not sufficient to assume that increased penetration of ICTs will lead to improved quality of life because these results suggest that, in this simple model, it will not. On the other hand the model does show that if increased ICT penetration leads to (for example) increased employment, wealth, leisure, free time, better environmental conditions and better communications with friends then quality of life may be positively affected. On the other hand, given the reliable effect for disability or long term sickness, we should seek to focus our ICT policy making efforts in this area where they could, potentially, have a positive effect. It is this indirect effect of ICTs, of the Information Society, on quality of life which need to be modelled and understood in order to determine the likely returns to the massive public investment in, for example, e- Europe. That said we must return to a theme that runs through out this report and indeed the e-Living project. Conducting analyses such as these using cross-sectional data does not provide the necessary evidence for action. Cross-sectional analysis can show which characteristics are associated with lower quality of life www.eurescom.de/e-living Page 31 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc scores at the ‘population’ and ‘average’ level. But what we really want to know is what difference particular ICT related interventions or transitions will make for particular groups of people. For example we might find that having PC or Internet skills increases the chances of finding work for those on the labour market margins. Finding work may lead to an increase in perceived quality of life for those particular people from ‘very low’ to ‘merely low’. To them this is a positive effect of ICTs which could not be captured using cross- sectional data. This, of course, is exactly what analysis of e-Living wave 2 data will start to provide.

6.4 Recommendations for future analysis This report raises more questions than it answers not least to do with adequate indicators, models and methods of analysis. In this final section we draw together the next steps: 1. To carry out further analysis of the social capital indicators to determine which combination converge and which best represent indicators of the various forms of social capital. 2. To carry out comparative analysis of those within groups who report low quality of life and social capital to determine if access to and usage of ICTs makes any differences within these groups. This analysis will inevitably be limited by small sample sizes. 3. To develop better models to predict the formal social activity and close relationship social capital indicators. 4. To analyse the effects of lifestage, the presence of children and interactions between partners’ social capital and quality of life. 5. Finally to make use of the longitudinal nature of the wave 2 data set to analyse any changes in social capital indicators with respect to changes in ICT usage behaviour. Some of these analyses are ongoing work which will be published under the aegis of the e-Living project as additional publications whilst the last two will require wave 2 data and so will form part of the analysis to be completed for D11: Longitudinal Comparative Analysis due to be completed in October 2003. 7 Bibliography Anderson, B (ed) (2001) e-Living D3: State of the Art Review. E-Living Project Deliverable. www.eurescom.de/e-living. Anderson, B., and Tracey, K. 2001. "Digital living: The impact (or otherwise) of the internet on everyday life." American behavioral scientist 45:456-475. Burt, R.S. (2000) “The Network Structure of Social Capital”. Research in Organizational Behaviour, Vol. 22. Castells, M., (1996), The Rise of the Network Society. Oxford: Blackwell. Collins, R. 1994. Four sociological traditions. New York: Oxford. COM (2002) 263, eEurope 2005: An information society for all, Communication from the European Commission, COM (2002) 263 final. Brussels 28.5.2002 de Gournay, C. (Ed.). 2002. Pretense of intimacy in france. Cambridge: University of Cambridge Press. Durkheim, E. 1954. The elementary forms of religious life. Glencoe, IL: The free press. Diener, E., Emmons, R.A., Larsen, R,J., and Griffin, S. (1985) The Satisfaction With Life Scale, Journal of Personality Assessment Vol.49 No.1 1985 e-Europe 2002 (2000) , eEurope 2002: An information society for all: Action Plan. Brussels 14.6.2000. Field, A. (2000) Discovering Statistics Using SPSS for Windows. London: Sage. Fortunati, L (Ed.). 2002. Italy, stereotypes, true and false. Cambridge: University of Cambridge Press. Franzen, A. 2000. "Does the internet make us lonely?" European Sociological Review 16:427-438. Frønes, I. , and Brusdal, R. 2000. På sporet av den nye tid: Kulturelle varsler for en nær fremtid. Bergen: Fagbokforlaget. Frijters, P., Haisken-DeNew, J.P. and Shields, M.A. (2002) The Value of Reunification in Germany: An Analysis of Changes in Life Satisfaction. 5th International Conference of German Socio Economic Panel Users. Berlin, July 3-4, 2002.

www.eurescom.de/e-living Page 32 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

Häußermann, H. and Petrowsky, W. (1989) ‘Die Bedeutung des Telefons für Arbeitslose’, in Forschungsgruppe Telefonkommunikation (ed.) Telefon und Gesellschaft, Vol.1, Volker Spiess, Berlin. Haythornwaite, C. (ed) (2001) The Internet in Everyday Life. American Behavioral Scientist, November 2001, vol. 45, no. 3. HLEG (IS) (1997) Building the Information Society for us all: Final policy report of the high-level expert group. European Commission, April 1997. Johnston, M. (2001) “Social Capital in Britain, Europe and the United States: Recent Evidence and Continuing Questions”. UK Office for National Statistics & Home Office Social Capital workshop, 16th November, 2001. Kahneman, D., Diener, E., Schwarz, N. (1999, Eds), Foundations of Hedonic Psychology: Scientific Perspectives on Enjoyment and Suffering. New York: Russell Sage Foundation. Katz, J. E., Rice, R.E., and Aspden, P. 2001. "The internet, 1995 - 2000." American behavioral scientist 45:405-419. Kavanaugh, A.L., and Patterson, S.J. 2001. "The imact of community computer networks on social capital and community involvement." American behavioral scientist 45:496-509. Kraut, R., et al. 1998. "Internet paradox: A social technology that reduces social involvement and psychological wellbeing?" American psychologist 53:1017-1031. Kumar, K. 1995. From post-industrial to post modern soceity: New theories of the contemporary world. Oxford: Blackwell. Ling, R. 1997. ""one can talk about common manners!": The use of mobile telephones in inappropriate situations." in Themes in mobile telephony final report of the cost 248 home and work group, edited by Haddon, L. Stockholm: Telia. —. 1998. ""she calls, [but] it's for both of us you know": The use of traditional fixed and mobile telephony for social networking among norwegian parents." Kjeller: Telenor R&D. — (Ed.). 2000. Direct and mediated interaction in the maintenance of social relationships. Kluwer: Boston. —. 2002. "The social juxtaposition of mobile telephone conversations and public spaces." in The social consequences of mobile telephones, edited by Kim, S. D. Chunchon, Korea. Ling, R., and Haddon, L. 2001. "Mobile telephony and the coordination of mobility in everyday life." in Machines that become use. Rutgers University. Ling, R., and Yttri, B. 2002. "Hyper-coordination via mobile phones in Norway." in Perpetual contact: Mobile communication, private talk, public performance, edited by Katz, J. E. and Aakhus, M. Cambridge: Cambridge University Press. Mante-Meijer, E. , and al., et. 2001. "Checking it out with the people - ict markets and users in europe." Helkdelberg: EURESCOM. Moyal, A. 1989. "The feminine culture of the telephone: People patterns and policy." Promethius 7:5 - 31. —. 1992. "The gendered use of the telephone: An australian case study." Media culture and society 14:51 - 72. Nie, N.H. 2001. "Sociability, interpersonal relations, and the internet: Reconciling conflicting findings." American behavioral scientist 45:420-435. ONS, UK (2001) “Social Capital; A Review of the Literature”. Office for National Statistics, Oct. 2001 Putnam, R. 2000. Bowling alone: The collapse and revival of american community. New York: Tuchstone. Raban, Y., Soffer, T., Mihnev, P., Ganev, K. (2002) e-living D7.1 ICT Uptake and Usage: A Cross-Sectional Analysis. E-Living Project Deliverable. www.eurescom.de/e-living. Rakow, L.F. (1988). "Women and the telephone: The gendering of a communications technology." Pp. 207- 229 in Technology and women's voices: Keeping in touch, edited by Kramarae, C. —. 1992. Gender on the line. Urbana: University of Illinois.

www.eurescom.de/e-living Page 33 of 34 PUBLIC e-living-D7.4-Age-Gender-Social-Capital-Issue-1.0.doc

Rakow, L.F. , and Navarro, V. 1993. "Remote mothering and the parallel shift: Women meet the cellular telephone." Critical studies in mass communication 10:144-157. Ronfeldt, D., and Arquilla, J. 2001. "Networks, netwars and the fight for the future." First monday 6. Salza, G. 2002. "Interview with william gibson." Silverstone, R. 1995. "Media, communication, information and the 'revolution' of everyday life." Pp. 61 - 77 in Information superhighways: Mulitmedia users and futures, edited by Emmott, S.J. London: Academic press. Silverstone, R., and Haddon, L. 1996. "Design and domestication of information and communication technologies: Technical change and everyday life." in Communication by design: The politics of information and communication technologies., edited by Silverstone, R. and Mansell, R. Oxford: Oxford University Press. Silverstone, R., Hirsch, E., and Morley, D. 1992. "Information and communication technologies and the moral economy of the household." in Consuming technologies: Media and information in domestic spaces, edited by Silverstone, R. and Hirsch, E.: London. Wellman, B. 1996. "Are personal communities local? A dumparian reconsideration." Social Networks 18:347 - 354. Wellman, B., (2001) Computer Networks as Social Networks. Science Vol 293, pp 2031-2034. Wellman, B. & Berkowitz, S.D. (1997) “Social Structures: A Network Approach (updated edition)”. Greenwich, CT: JAI Press. Wellman, B., and Tindall, D. 1993. "Reach out and touch some bodies: How social networks connect telephone networks." Pp. 63 - 93 in Progess in communication sciences. Vol 12, edited by Richards, W. and Barnett, G. Norwood N.J: Ablex.

www.eurescom.de/e-living Page 34 of 34 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1

e-Living D7.5 – Homeworking and Teleworking: A Cross-Sectional Analysis Malcolm Brynin (ISER, University of Essex) Ben Anderson (Chimera, University of Essex) Birgitte Yttri (Telenor)

The authors would like to thank Dr Leslie Haddon and Peter Stollenmayer of the e-Living steering group for their comments on previous versions of this report.

e-Living: Life in a Digital Europe, an EU Fifth Framework Project [IST-2000-25409] www.eurescom.de/e-living/index.htm PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1

Table of Contents 1 Introduction ...... 4 2 What is teleworking? ...... 6 3 The measurement and extent of home-working and telework ...... 7 4 The personal and family characteristics of home-workers and teleworkers...... 10 5 Education and PC skills ...... 13 6 The jobs and pay of teleworkers ...... 14 7 Home-workers’ perceived quality of life ...... 16 8 A Fuller Picture...... 17 8.1 Teleworking...... 18 8.2 PC skills ...... 21 8.3 Pay ...... 21 8.4 Quality of Life ...... 22 9 Conclusions and Implications ...... 23 10 Bibliography...... 25

www.eurescom.de/e-living Page 2 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1

Summary

• Teleworking is a widespread form of work if this is viewed in terms of technology rather than of time. A large proportion of the labour force uses ICTs to work at home at least part of their working week. • Teleworkers must be distinguished from other homeworkers, but both are different from people who always work in the workplace. • Homeworkers as a whole comprise people who use an online connection to work, people use a PC, those who rely on the mobile phone, and those who use no ICTs to work at home. • These groups are different in their characteristics, and these have potential effects on the demand for more extensive teleworking and for any economic benefits which might follow from this. The report highlights the differences and similarities between the groups. • Homeworkers who use PCs but not mobiles or the Internet are more often female. • Self-employment figures prominently in teleworking and homeworking. • All tele and homeworkers are more educated and have higher computer skills than the rest of the labour force. • Homeworkers who use the Internet have the highest educational levels, occupational status, pay and quality of work life when other factors are controlled. PC homeworkers, mobile users and homeworkers are similar in many respects, but PC homeworkers are paid less than other homeworkers. However cross-sectional data cannot prove that becoming a tele/homeworker leads to these outcomes. Teleworking may suit some but not others… • There is no relationship between quality of work life and reduced commuting time in any country except Britain (marginal effect) and Israel (strong effect). This suggests that promoting telework in order to reduce commuting is, on the whole, unlikely to lead to substantial increases in workers’ quality of work life. • Most of these patterns are consistent across countries, which suggests that there are underlying processes that influence tele and homeworking regardless of country.

www.eurescom.de/e-living Page 3 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1

1 Introduction Unlike new computer technologies and processes which directly increase the demand for products and services, and which therefore have immediately visible economic effects, teleworking is simply a new way of doing something old. It is only a relocation of work, an extension of the “take-away” phenomenon which more commonly applies to food or films. It is, therefore, as much a social as an economic process, introducing family as well as employment issues. It is therefore not clear that our concern with teleworking should be primarily economic. Yet teleworking does have the potential for considerable indirect economic benefits. Whether these materialise depends in turn, however, on whether teleworking is genuinely a new phenomenon with the potential for a significant restructuring of the way we work. If it is old wine in new bottles, the new bottling technology and distribution system might themselves be new and economically important innovations. The extent to which teleworking represents both new and different processes forms the focus of this report. Teleworking has in fact been with us for a long time, perhaps dating back to the concept of ‘telecommuting’ that appeared in the mid-1970s (Nilles 1975) and re-appeared with various agendas in the 1980s and 1990s (Haddon and Lewis, 1994). Little changed then, but the currently new ICTs – computers, the internet, mobile telephony – have made many European policy makers think that teleworking will become a socially and economically important mode of work (e.g. COM (2002) 263) even though Haddon (1999) warns us against such ‘technology or policy push’ approaches . Predictions have been made for a very considerable increase in the incidence of teleworking although there are certainly maximum constraints on teleworking: one study estimates the potential for teleworking as 23% of the UK labour force. In Italy this is over 17%, Germany 20%. It is lowest in Greece at less than 11% (ONS 2002). But even figures well below these ceilings represent very large absolute numbers. Increased teleworking could in principle result in real social and economic benefits. The general expansion of ICTs has been strongly linked to the prospect of growth through increased demand and increased efficiency (OECD 2000). More specifically, teleworking might lower the unit cost of labour by reducing overheads, in particular by reducing the demand for office space. It would also minimise travel to work and therefore travel costs, with benefits either to the worker or, if the saving is passed on through reduced wages, to the employer. This would of course have beneficial effects on the environment, though limited by the so-called “rebound” effect – people using time at home for other purposes, such as shopping, which might affect the environment (see Gillespie et al, 1995 and Anderson, 2001: 23-30). Whatever people do with the time transferred from the workplace to home, there is implicitly also a social benefit derived from the freedom to control one’s time – for instance for home care purposes, although a recent review suggests there is little evidence that teleworkers actually do this (Bailey and Kurland 2002). Teleworking might also be viewed as an aspect of “flexible employment”, which is held by many to be an essential component of new forms of working resulting from both changing technologies and changing management techniques (Gregg and Wadsworth 1999). Finally, this apparent shift of control also identifies the teleworker as a consumer. The need to have computer equipment, internet services or mobile communications raises the demand for these, whether paid for by the employee, the employer, or the self-employed. This too has important economic consequences. In sum, a radically enlarged teleworkforce could be viewed as part of a generalised modernisation characterised by new and cleaner technologies, flexible employment, lifestyle choices and both increased consumption and economic growth. Yet many have been the critics of this vision, and there are good reasons for this. First, both telecommuting and the closely related distance learning have, using earlier technologies, had negligible impact on how we work and how we learn. There are fundamental organisational reasons why people work and learn in groups, and this is strongly reinforced by the personal need to interact with people. New technologies cannot alter these basic requirements. Second, in terms of welfare, if teleworking increases for economic reasons, this might include a resurgence of routine work from home, such as telephone sales or data entry, and could be exploitative. The social effects might therefore be negative, especially for women (Webster 1996). Third, the internet bubble has burst, and even if steadier growth occurs over a longer time than some originally predicted, there is no sign that the take-up of new domestic technologies is grounded in previously unmet social needs through which major new commercial developments could be anticipated. Computers may, therefore, be simply one more consumer luxury, like the CD player, the food mixer, the skater board. Assessment of the potential impact of teleworking involves a number of difficulties – not only measurement and projection but even definition. How can evaluations be made if we are unclear what the process actually entails? Felstead and Jewson point to “the wide variation in definitions and terminology used within and between societies and by different groups of researchers” (2000: 47). Telework always signifies work undertaken from a remote site, making use of ICTs, and we restrict this from the outset by focusing mostly www.eurescom.de/e-living Page 4 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1 on ‘home-based telework’ - any paid work conducted in the home with the aid of ICTs. However, more or less expansive definitions are possible even within this framework, depending on whether telework is restricted to replacement of the workplace by the home through use of ICTs or whether any work undertaken at home using ICTs might count. The “spillover” effect (people taking work home outside of normal work hours), where ICTs are used at home to supplement normal work, is only one end of a spectrum of teleworking possibilities. Some such work might also be undertaken at home in normal hours, so teleworking might be occasional. Bailey and Kurland (2002) suggest that to date most studies of telework have focussed too narrowly on the few who are full time teleworkers and not sufficiently on the many more who work at home on a more infrequent basis. At issue also is how fundamental the usage of ICTs might be. For instance, if someone works at home online are they merely doing work they would otherwise have been doing while happening to be on-line? In this case teleworking is effect rather than cause. The impacts of such distinctions have to be estimated (Haddon and Silverstone 1994). The importance of these distinctions derives partly from the characteristics of those who telework. If the extent to which people telework (for instance, pitching a tight definition against a loose one) is distinguished by different characteristics such as type of job, but perhaps also by the age and gender of the teleworker, then it would be reasonable to assume that the different forms of teleworking are important: the variation in the degree and nature of teleworking is not arbitrary but describes different types of employment and of employee. One of the key distinguishing characteristics might be the education and skills, in particular computer skills, which the teleworker brings to the job. These might tell us something about the nature of the person doing the job. For instance, it is necessary to test whether teleworking is polarised between professional workers with university degrees, for whom teleworking is perhaps a benefit, and poorly educated workers who use ICTs not to network with colleagues but to call customers or sell products. At the same time, the distribution of computer skills affects the extent to which teleworking is feasible in the first place. No amount of economic or private need will raise the proportion of the labour force which teleworks from home if IT skills are inadequate to enable this level of independence. In related fashion, it is necessary to test the extent to which teleworking might raise the rewards to work, but this can only be undertaken by controlling for variation in both education and skills. Again, these might tell us something about the meaning of telework. These rewards are important both in terms of productivity and welfare. If teleworking is organisationally beneficial then we might expect teleworkers to receive higher pay than other workers - that is given the nature of their jobs and their relative education. Data on pay can therefore also say something useful about the nature of the teleworking experience. It might be that teleworkers receive higher pay than others, because they are trusted to telework, because they are successfully self-employed, because teleworking makes them more productive, or because they are professional or managerial workers who are highly paid anyway and simply need to telework at least some of the time. Alternatively, some teleworkers might be poorly paid if they are, for instance, traditional homeworkers who merely use ICTs as a work tool (eg homebased telephone sales). Assessment of the rewards to teleworking might help elicit the extent to which the phenomenon is simply another form of traditional homeworking, whether of high or low status. Teleworking might be influenced not only by organisational expediency but by personal or family demands for greater control over the use of time. Indeed, unless it benefits the individual it is difficult to see why teleworking should ever take off to a significant extent. But there are only a few studies on the potential link between telework and quality of life. Van Sells and Jacobs see quality of life as “… a global evaluative term that summarises a person’s reactions to the experiences in his or her life” (1994: 81) and separate quality of life from quality of work life. The latter is reflected in variables such as productivity, creativity, turnover and absenteeism in the organisation – and if and how individual employees identify with or feel alienated from the organisation. Van Sell and Jacobs conclude their study by suggesting the need for more research on effects of telecommuting on individual quality of life at work and away from work. Analysis of the quality of life of teleworkers might help us to see in what measure teleworking does offer real benefits to the individual. These, therefore, are the themes that inform the following discussion. In summary terms, the report examines teleworking, and homeworking more generally, from several perspectives: • the impact of differing definitions of both homeworking and teleworking on measurement of their incidence • identification of types of home-worker by their characteristics • the potential productivity effects of home-working as determined by individual income from work www.eurescom.de/e-living Page 5 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1

• estimation of the effect of education and PC skills on the likelihood of various kinds of home-working • analysis of differences in quality of work between different kinds of home-worker The aim is to test the extent to which forms of home-working (including telework) have distinctive effects and causes over and above the personal characteristics and the family requirements of those who engage in them. The data are from the EU-funded e-Living survey, which interviewed 1,750 individuals in each of six countries – Britain, Bulgaria, Germany, Israel, Italy and Norway - towards the end of 2001. As these data are the basis of all the reports using e-Living data, these are not described here. Details appear more fully in Raban et al (2002). 2 What is teleworking? As we have noted, definitions of teleworking are many and varied. The European Commission has proposed a definition of telework as: a method of organising and/or performing work in which a considerable proportion of an employee’s working time is: away from the firm’s premises or where the output is delivered; and when work is done using information technology and technology for data transmission, in particular the internet (eWork, 2001, p11). This is a general definition, though it does not explicitly include the self-employed. The ECaTT (2000b) definition of home-based teleworking requires the following: • work from home at least one full day per week • use of a PC for this work • use of the phone, fax or email to communicate with colleagues from home • that teleworkers are either salaried or self-employed People with the same pattern of work at home but who spend less than a full day per week teleworking are called ‘occasional’ teleworkers. There are three elements to this definition: technology, intensity and location. The former has two functions: performance of the work and communications. Intensity here means frequency and duration. The location component distinguishes between work at the workplace, at home, or at varying places. The Office of National Statistics (ONS) definition used in the British Labour Force survey has the same distinctions but treats them differently. Teleworkers are paid or unpaid workers who use a phone and a pc, whether they work at home full-time or occasionally (at least one day in the reference week). A narrower group called TC teleworkers, for whom a telephone and a computer are essential for their work, is also defined. We start the analysis presented in this report by defining home-workers to be those in paid work only, but who do any paid work at home. It is possible then to define teleworkers to be a subset of home-workers who use information and communications technologies (ICTs) in their work. However it is not clear which ‘ICTs’ should be included – PCs? Internet? Mobile phones? Fax? Fixed line phones? There is no clear consensus on this issue and therefore any analysis is governed by the research issues it wishes to address. Brocklehurst (1989) suggests we reserve the label ‘telework’ for those homeworkers who use new technology. Given the focus of e-Living on those technologies whose impact is least understood and on those predicted to grow most in penetration, we have chosen to define teleworkers as those home-workers who use PCs and/or the internet, or a mobile phone, during this work. We also distinguish between these forms of technology, which adds a further level of detail to the final classification. In respect of intensity of teleworking, there are several reasons for being interested in teleworkers whether or not they telework full-time or in normal working hours. First, it might give an indication of the potential employee demand for full-time ICT mediated work done from home. Second, it gives a more realistic indication of the demand for ICT facilities needed to work effectively from home. Third it enables a more complete analysis of the effects of home-working of any kind on the lives of the workers and their households. Our definition locates telework at home, using various technologies or services (the net, a PC or the mobile phone), and where this is undertaken full-time or otherwise. We use various elements of e-Living dataset to build a classification of work, including categories for homework and telework based on the above definitions. The first part of our classification relates to location. The survey question asks those in paid work to specify their main place of work. The options are: 1. Mainly work at home www.eurescom.de/e-living Page 6 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1

2. Work at work premises 3. Driving or travelling around 4. Or at one or more other places For simplicity the last two categories are merged into a single category, though this does not form part of our final home/telework classification. The main part of our construction of the schema derives from a series of items relating to frequency of working at home during the day, in evenings and weekends and of using PCs/Internet to do so, and therefore includes measures both of intensity and technology. We then add information on the importance of a mobile phone for the individual’s work (though use of a computer takes primacy over this in the classification scheme). This results in the following categories of worker: 1. People who do any work at home and use the internet to do so (Net Homeworkers) 2. People who do any work at home and use a PC to do so (PC Homeworkers) 3. People who say that their mobile is important for their work but are not internet or PC homeworkers (Mobile Users) 4. People who do any work at home during normal work hours but who do not use a PC or the internet and do not view the mobile phone as important for work (Day Homeworkers) 5. As in (4) but where undertaken in the evenings or at weekends (Overspill Workers) 6. People who work at one or more workplaces – excluding the home (Workplace Workers) Much of the analysis below merges categories 4 and 5. The first of these are in the EcaTT definition “occasional workers”, or in Kraut’s definition (1989) “supplementers”. The second are often called “overspill workers”. In both cases the above scheme treats these as homeworkers, not as teleworkers. Technology is therefore not an issue in this case while the data show that the numbers in (5) are small. In addition, it is probably sometimes difficult to distinguish the two processes in practice, and work flexibility is implied by both categories. There seems little harm in combining the two categories, though distinctions are made between them at various junctures as appropriate. The overall impact of our classification is that we contrast teleworkers with homeworkers and workplace workers; we make no a priori distinction between occasional or overspill teleworkers and more formally defined teleworkers (so exclude a measure of intensity); and we take account of the different technologies which are used to telework (regardless of its function). 3 The measurement and extent of home-working and telework Teleworking has made significant inroads in some countries. According to EcaTT (2000a), the average in ten EU countries is 6%. Finland has the highest proportion of employment who telework at just less than 17% (followed by Sweden and the Netherlands). The UK has nearly 8%, Germany around 6%, Italy nearly 4% and France and Spain the lowest at 3% or less. The proportion of this which is supplementary or occasional varies between roughly a quarter and one third. In contrast, teleworking is around 21% of the US labour force. The ONS survey in the UK found that there were 2.2 million teleworkers in the UK, an increase of 65-70% since 1997 (ONS 2002). The effect of using the various taxonomies discussed above to subdivide the labour force in the countries surveyed is shown in Figures 1-3. These graph the extent of homeworking in the six countries of e-Living for the different definitions. To repeat the point made at the outset: teleworking depends on location, intensity, and technology. However, one element of this features little in the discussion below, namely the potential for teleworking on the basis of work undertaken in several places (or via several jobs) which has been termed ‘mobile work’. Figure 1 demonstrates this. The baseline for the graph is the labour force but the chart excludes those who work at a single place of work, who are therefore treated as a residual. It shows that roughly 5% or less of the labour force have their home as their main workplace in five countries, with Bulgaria at around half that. A far higher proportion work at several places including those who travel for work, although Israel is notably lower. The graph clearly shows that in terms of location, the home-based element of work away from a main place of employment forms a rather small proportion of all such work. Nevertheless, as it is difficult to test the degree to which the ‘mobile work’ group uses ICTs in the process of their work (and this usage is likely to be highly variable), we do not incorporate this group explicitly into our home/telework classification.

www.eurescom.de/e-living Page 7 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1

at home 15 at several plac es

10 % thoseof in paid work 5

0 UK It al y Germany N orw ay Bul gari a Is rael

Figure 1: Place of work of those in paid work and who do not work mainly at work premises (percentage of labour force of each country)

The intensity aspect is shown in Figure 2, revealing the importance of differing thresholds of intensity to estimation of the importance of the homeworking experience. Around two fifths of the work force usually works at home at least part of the week. This provides some basis for estimating the potential for teleworking.

50 % with home as main 40 workplace % working from home at 30 least weekly 20 % doing any work at home % of labour force 10

0 All GB Italy Israel Norway Bulgaria Germany

Figure 2: The intensity of home-based working in each country (percentage of labour force of each country)

www.eurescom.de/e-living Page 8 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1

80 70 60 Net Homeworkers 50 PC Homeworkers

% 40 Mobile Users 30 Day Homeworkers Other 20 10 0 Italy Israel Britain Norway Bulgaria Germany

Figure 3: Distribution of teleworking, mobile telephony, home-working and other employment (percentage of labour force of each country)

The full picture based on the classification, adding technology to location and intensity, is given in Figure 3. Two things should be noted. First, the proportion working at the workplace varies from somewhat over 30% to over 70%, though if Bulgaria is excluded the upper limit is 50%. This means that half or more of the workforce in nearly all countries works from home at least part of the week, or at least finds a mobile phone useful for work. The variation between Britain, Germany and Italy is also rather limited and there seems therefore to be some sort of mode effect. Norway and Israel have the lowest proportions. This might reflect a higher national income in the case of the former but it seems reasonable that geographical limitations in Norway and security problems in Israel might be important factors. In Bulgaria, the opportunity to telework is presumably limited both by income and infrastructural constraints, though as Figure 2 makes more clear, this seems also to apply to any homeworking. The second point to note is the variation between countries in the impact of the home and teleworking elements of the classification. While Israel and Norway have the largest proportions of their workforce who are PC homeworkers, Norway and Britain show the highest proportion of Net homeworkers. Interestingly, even though relatively few Israelis work at several places, they show the highest proportion who say that their mobile phones are important for work. This should remind us that workers in ‘traditional offices’ will still use mobile phones for work purposes for a variety of reasons. The mobile phone is the most frequent basis – presumably but perhaps not always - for work away from the workplace in all countries. Several things of note are obscured by the above: the role of self-employment, the amount of time spent teleworking, and, as mentioned earlier, the distinction between day and overspill homeworkers. Taking self- employment first, although the ONS report (ONS 2000) claims that the predominance of self-employment in telework is declining (ie it is spreading to employees), self-employment explains a large part of teleworking, as shown in Table 1.

Table 1: The percentage of each work category comprising self-employment

Britain Italy Germany Norway Bulgaria Israel Net Homeworkers 23 44 28 13 - 21 PC Homeworkers 25 32 29 22 - 26 Mobile Users 21 40 16 13 41 22 Day Homeworkers 13 29 11 9 26 17 Workplace Workers 3 14 4 4 5 10 All labour force 12 27 11 10 13 18 In all countries except Bulgaria there is no difference between homeworkers, who do not use ICTs to work at home, and the labour force as a whole in terms of the proportion in self-employment. But the self-employed form a far higher proportion of teleworkers than they do of the labour force as a whole. This is especially the case in Britain, Italy, and Germany. In Italy, 44% of net homeworkers are self-employed. However, this

www.eurescom.de/e-living Page 9 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1 relatively sizeable share applies to PC homeworkers in Norway and Israel. To put this another way, across all the countries together, 28% of the self-employed are either net or PC teleworkers, compared to 15% of the employed; 39% of the former and 22% of the latter say a mobile phone is important for their work. However, this makes equally clear that teleworking is important to employees as well as the self-employed. In addition, most teleworkers spend a lot of time working at home. Table 2 shows the proportion of each category that workers at home at least two days a week.

Table 2: Proportion of Each Work Group Working at Home 2 Days a Week or More

Britain Italy Germany Norway Bulgaria Israel Net Homeworkers 75 69 79 51 - 73 PC Homeworkers 37 36 53 36 - 49 Mobile Users 22 24 25 15 22 24 Day Homeworkers 52 72 55 29 24 71 75% of Net homeworkers in Britain and 79% in Norway work at home at least two days a week. This extensive homeworking applies less where a PC is used for this, but in this case the figures are still substantial. However, in some other countries, the frequency of non-ICT based homeworking is also high. Taking all countries together, 40% of net home workers, 27% of PC homeworkers, 12% of mobile users and 33% of homeworkers do some work at home most days in the week. Taking work at home at least once weekly, the figures are respectively: 67%, 42%, 22% and 55%. Frequency of work at home where this is undertaken at all is therefore far from marginal. This does not tell us for what proportion of work undertaken at home the particular technology is used. 50% of teleworkers (net and PC) in Britain use a PC for their work at home at least half of the time. The figure is 34% for Italy, 37% for Germany, 59% for Norway, and 34% for Israel. Thus at least a third of homework requires a PC for a minimum of half of work at home where telework using a PC is undertaken at all. Taking all countries together, 16% of these teleworkers use a PC for most of their work at home. Even telework does not require use of technology all the time. It is a mistake to assume that telework is fundamentally technology based. It is fundamentally work at home, for which technology is important but not always necessary. Much the same caveat applies to intensity of work at home. The high frequencies just shown do not mean that in the aggregate much time is spent working at home. Many of these workers work at home only part of their working time and some work only part-time anyway. The average weekly number of hours a week worked at home by someone who mostly uses the net to do so is 16 in Britain, 11 in Italy, 19 in Germany, only 8 in Norway, and 15 in Israel. For PC homeworkers the figures are 12, 10, 14, 8 and 11 respectively. For other homeworkers they are 8, 12, 14, 5 and 9 (and the figures for mobile users are very similar to these). Not only are the figures low in some countries but they are low across all categories of work. If we were therefore to measure telework on an hourly basis the contribution to total work hours in the population would be fairly small. It was stated above that for most analyses day homeworkers would be combined with overspill or evening/weekend workers (where neither use ICTs for this purpose). This is in part because overspill workers are a small category in absolute terms. Proportionally, their share of general homeworking (the combined category) is 27% in Britain, 12% in Italy, 25% in Germany, 44% in Norway, zero in Bulgaria, and 13% is Israel. In sum, a strict definition of teleworking radically underestimates the real incidence of homeworking. A definition of teleworking based on location, intensity and the form of technology gives a much fuller and more varied picture. The result shows high frequency of home and teleworking, with a substantial share coming from self-employment, though the total number of hours spent working at home at all is not high, and nor is computer technology used for the larger part of this work. Much of the rest of the report looks at the relationship between these difference categorisations and a range of other factors that describe the different types of teleworker and different aspects of the teleworking experience. 4 The personal and family characteristics of home-workers and teleworkers What are the best predictors of a worker having their home as their main place of work? A recent review of telework research suggests that evidence here is still elusive (Bailey and Kurland, 2002). However the studies the authors review suggest that age, gender, education, income, number of people in the household, industry type, whether an employee or self-employed, whether or not a manager and whether a contract or

www.eurescom.de/e-living Page 10 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1 full time worker may all affect the likelihood that someone will work mainly from home. In this section we focus primarily on age and gender. Other aspects will be treated in subsequent sections. Stanworth (1997) differentiates the following types of home-based teleworker: • Employees who work at home for part of their working week, who tend to be male, are well paid and relatively highly skilled • Employees who work at home for all of their working week: these tend to be female and much lower skilled, often undertaking routine clerical work • Freelance homebased teleworkers who work with many clients on a contract basis and are often owners of small businesses This conceptualises teleworking as a reflection of other employment and demogrqphic characteristics. The outcome is a polarisation between those in high-status jobs – usually men - who can use the new form of working to increase their personal welfare and those – mostly women - for whom teleworking might simply be another form of home-employment, another form of drudgery. The recent UK survey by the Office of National Statistics (2002) found that: • Two thirds are mostly professional or managerial, but the “skilled trades” also contribute (through self- employment) • Male teleworkers predominate in occupations with a high proportion of teleworkers such as the professions; women predominate where teleworkers is not common, such as secretarial work • TC teleworkers are more likely to be in the private sector than other teleworkers but there is no gender difference between the two groups • 45% of TC teleworkers are self-employed, compared to 11% of the entire labour force, but the proportion of employed teleworkers is increasing rapidly (the number of employed teleworkers rose by 82% 1997- 2001 compared to 48% for self-employed teleworkers) Such findings give only partial support to the sort of conceptualisation promoted by Stanworth. Teleworking is largely professional or part of the working system of the self-employed, but might not be a new form of traditional, low-paid home-working. We turn now to a more detailed discussion of the kinds of work types defined in Section 2. The aim is to understand what factors may be associated with these employment patterns and thus which may be important to their growth. Figure 4 shows the distribution of the home/teleworking categories on the basis of three age groupings for all countries combined. This procedure has been chosen to make the chart presentationally manageable, though there are some country differences. For instance, compared to the average age within each country, there are far fewer young net homeworkers in Norway, more young PC homeworkers in Italy, Germany and Norway, fewer young mobile users in Norway but more in Bulgaria, fewer homeworkers in Britain, Italy, Bulgaria and Israel. In all other cases the distributions are similar across countries, and the proportion of young people amongst workplace workers is similar to the proportion in the labour force of all countries.

www.eurescom.de/e-living Page 11 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1

50 45 40 35 30 16-29 25 30-44 20 45+ 15 10 5 % of each telework category 0 Net PC Mobile use Homework Workplace homework homework work

Figure 4: The Age Distribution of Tele, Home and Workplace Categories

Figure 4 shows no very great difference in the age distribution on each of the categories of the variable of interest. Age, therefore, is rather neutral in this respect. It is not the case that either young people - who are more familiar with the technology - predominate in teleworking, or alternatively that older people - who are more likely to fill senior positions - do so either. What about gender? Table 3 shows the ratio of men to women in each category of homeworking. One implication of Stanworth’s classification is that we might expect teleworkers to be male regardless of extent of time spent working at home. The first two rows strongly suggest this is the case. This means that women, who might conceivably have most to gain from teleworking, are very much under-represented in the ranks of teleworkers. Mobile-using workers are also much more likely to be male. But homeworkers are, as they have traditionally been, more likely to be female.

Table 3: Ratio of Men to Women in Each Work Category. (In Bulgaria the figures for Net and PC homework are too small to analyse.)

Britain Italy Germany Norway Bulgaria Israel Net Homeworkers 2.6 3.2 4.5 2.5 - 1.4 PC Homeworkers 1.3 1.9 1.4 1.7 - 1.3 Mobile Users 2.4 3.5 3.7 2.7 2.7 2.3 Homeworkers 0.8 0.5 1.1 0.7 0.6 0.5 Workplace Workers 1.0 1.2 0.9 0.7 0.9 0.7 This reveals clear gender differentials. Teleworking is largely male. Net-based teleworking is overwhelmingly male in all countries except Israel. PC homeworking is generally male but less overwhelmingly. Mobile use for work is about as male-dominated as net-based telework. Homeworkers who do not use ICTs for this are largely female. All these groups can be assessed in comparison with workplace workers, in which the gender divide is about equal. It appears, therefore, although this table contains no information on the nature of the actual work done and does not control for other variables, that women tend to fall in to the traditional homework category while men are strongly over-represented in the non-traditional forms of work outside the workplace where the new ICTs are used. It is the case that men are far more likely to be self-employed than women (men constitute 71% of the pooled self-employed labour force in the e-living survey), and self-employment is strongly linked to teleworking. The ONS study (ONS 2002) makes this point (stating too that the occupational and industrial distribution of men and women do not explain the gender difference in teleworking). However, looking at the employed by themselves in eLiving, the male bias remains strong in the three teleworking categories but disappears amongst homeworkers. For instance, 70% of employed Net homeworkers are male, compared to 41% of homeworkers and 46% for the whole labour force (with the countries pooled). The outcome for homeworking is odd given that self-employment is common amongst homeworkers while homeworking tends to be female. In fact there is an interaction between self-employment and gender in this category. Amongst the self-employed in this group homeworkers have the highest proportion of females (51%, compared to 24% for net homeworkers and 29% www.eurescom.de/e-living Page 12 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1 for the whole labour force). This suggests that gender does play an important role in the distinction between homeworking and teleworking, while for homework itself this depends on the rate of female self-employment. It would seem that any self-employed women work at home without using ICTs. A final demographic factor of particular interest is family, as it is often held that homeworking in general might be sought by employees, especially women, to fit in with family needs. As the general relationship between family and ICTs will be examined in detail in Work Package 6 (after wave two, when data from more than one family member will become available), this will be only briefly discussed here. If homeworking is in part designed to fit in with family, then there should be a probability that single-person households would be under-represented in homeworking houses relative to the labour force as a whole. Of course the proportion of the labour force living alone is always small – ranging from 3.1% in Israel to 10.5% in Norway, and so variations around this proportion are not going to make much difference to the demand for teleworking. Nevertheless, if people living alone are less likely to work at home than to work at a workplace, then this is at least an indicator of family demand and/or the need for single persons to use work as a asocial activity. The over-representation of people living in a family occurs only in PC-homeworking households in all countries. It is consistently the case, except in Bulgaria (where the numbers are to small to assess) and in Israel (where nevertheless the proportion is about the same as in the labour force as a whole). The biggest differential applies to Norway where 10.5% of sample individuals but 6.5% of PC-homeworkers live alone. Mobile users tend not to have a higher or lower probability of living in a family. Both net-homeworking and general homeworking produce mixed results, with some countries being under and some over-represented in family households. Taking this on a country basis, in Israel there is very little variation across work forms in the probability of living in a family. Italy, Germany and Norway give mixed results. Only in Britain is there a clear difference. People in the three forms of teleworking have a lower probability of living alone than in the labour force as a whole, and thus are more likely to have a family of some sort. General homeworkers are more likely to be living alone. In sum, family might make some difference to teleworking, but this is slight, and is only consistently the case for PC-homeworking (in all countries) and in Britain (all forms of teleworking). The distinction between day and evening/weekend (overspill) homeworkers, as was stated above, has been obscured in the analysis by combining these into a general homeworking category. However, there are a number of differences between the two in terms of the type of people within each category. This is especially so with age. In Britain, Italy and Germany the average age of day homeworkers is 43-44, compared to around 36 for overspill workers. In Norway and Israel there is no appreciable difference, but it is clear that in general overspill workers are younger than those who work from home during the day, whether occasionally or otherwise. In Italy and Israel they are also far more likely to be male, but the reverse applies to Britain and Norway, though the gap is not as large in this direction, while in Germany there is no gender difference. However, the largest imbalance is not huge: in Israel 32% of day homeworkers are male compared to 52% of overspill workers. Family size varies little across the two categories. Overall, therefore, the division of non-ICT based homeworking is marked by certain demographic differences. If not consistently across countries, when compared to day homeworkers, overspill homeworkers tend to be younger and are more likely to be male. 5 Education and PC skills Does the gender bias shown above mean that teleworking reflects professional, well-paid work while predominantly female homeworking is the reverse? This can be tested through looking at the proportion of each category with a degree, shown in Table 4. The final line of the table shows the average number of graduates in the labour force of each country (in the sample). It is clear that in comparison with this last line in all countries net-homeworking is very much a graduate activity, as is PC-homeworking, though less overwhelmingly. In the case of mobile use there is roughly parity in the comparison with the labour force as a whole, though in Israel these people are less likely to be graduates. But technology does not explain everything. Homeworkers also tend to be more highly educated than the labour force in general. In other words, homeworking of any sort is quite likely to be undertaken by graduates. Non-ICT based homeworking does not appear to be undertaken by people with low levels of education and who might as a consequence be expected to be doing poorly paid and routine work.

www.eurescom.de/e-living Page 13 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1

Table 4: Proportion of each homeworker type that has a degree1

Britain Italy Germany Norway Bulgaria Israel Net homeworkers 62 46 29 80 - 71 PC homeworkers 35 32 23 65 - 51 Mobile users 24 17 12 48 37 29 Homeworkers 40 23 14 64 43 53 Workplace workers 10 13 4 35 18 30 All labour force 24 19 11 52 25 37 Obviously those who use a PC are likely to know more about computers than those who do not, and this might apply especially to people who use a computer at home. In this case, it might be that teleworking could extend further if people had more knowledge of and presumably confidence in how to operate a computer. Table 5 produces a very clear pattern. In the five countries where figures can be analysed, there are effectively three bands of skills: net and PC homeworkers have the highest PC skills (but with net homeworkers coming above PC homeworkers), mobile users and homeworkers are somewhere in the middle, while workplace workers have the lowest level of skills. This may relate to the type of work that people do and this is examined in the next and in the final section. The results suggest that homework is not ICT-based because of a lack of PC skills, though this might play some role. Given that many workplace workers do not need to use a PC in their work, the scores here are also hardly low. Nevertheless, if the scores achieved by net and PC-homeworkers are a reflection of the level of skill needed, then it is possible that many people in the other categories might find it difficult to telework from home effectively with their particular level of skill.

Table 5: Mean PC Skill Scores (range 0-6)

Britain Italy Germany Norway Bulgaria Israel Net homeworkers 4.8 5.1 5.2 5.3 - 5.4 PC homeworkers 3.9 4.1 4.5 4.8 - 4.1 Mobile users 3.0 2.6 3.2 3.8 1.7 2.3 Homeworkers 3.4 2.4 3.1 4.1 0.8 2.5 Workplace workers 2.5 2.3 2.5 3.2 0.5 2.2 To return to the distinction within homeworkers between daytime and spillover work, in terms of both education and PC skills these are clearly two different categories. In Italy, Norway and Israel the latter is much more likely to be undertaken by graduates than is the former. The most extreme difference applies to Israel: 49% of day homeworkers and 76% of spillover workers are graduates. However, the difference is slight and reversed in Britain and non-existent in Germany. Moreover, the proportion of graduates amongst day homeworkers is higher than in the labour force of all countries, though this differences is not always large. The distinction between the two categories is further reinforced by variation between the two in PC skills, which are much higher in all countries amongst spillover workers (an average of 4.3 compared to 2.5 for day homeworkers). Recalling from the previous section that spillover workers tend to be younger and are more likely than day homeworkers to be male, it seems these are two distinct groups. Spillover workers probably differ little from teleworkers. It is not the technology that makes the difference: spillover workers have good PC skills. This section has revealed substantial differences between each group in the variable under review. Teleworkers who uses computers tend to be more highly educated and to have higher PC skills than other groups, but spillover workers are similar; mobile users are intermediate; day homeworkers have greater education and skills than the non-homeworking part of the labour force. 6 The jobs and pay of teleworkers Much of the above may be reflected in occupational differences. Manual workers have on average lower education, work a set number of hours under controlled conditions, and generally do not need to use a PC. Few would be likely to telework: refuse collectors are unlikely to want to take their work home with them. However, mobile phones might be important for workers who work away from a main site but not from home.

1 Though with a caveat: the meaning of a graduate varies across countries, with some countries having a fairly tight definition of what would be included as a graduate qualification, others a looser definition. This could explain some of the within-country differences. www.eurescom.de/e-living Page 14 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1

The ONS study in Britain found that teleworking is dominated by managerial, professional and associate professional/technical workers. However, there were over 300,000 teleworkers amongst the skilled trades - compared to over 500,000 teleworking professionals (ONS 2002). The situation of professional workers and managers is shown in Table 6.

Table 6: The percentage of each work type comprising professionals or managers

Net PC Mobile Home Workplace All Britain 76 49 38 52 22 37 Italy 74 47 35 42 19 32 Germany 56 51 35 36 22 32 Norway 72 67 44 66 37 50 Bulgaria - - 30 41 16 21 Israel 51 39 39 37 20 33 In most countries roughly three quarters of net homeworkers are managers or professional workers. In each country the proportion of PC-homeworkers is less predominantly managerial and professional than this, but still well above average for the labour force. Homeworkers are broadly similar to this, while mobile users have roughly the same profile as the rest of the labour force. No other occupational group consistently has a higher proportion of either teleworkers or homeworkers than appears in the labour force in general. Technicians, for instance, not only have roughly the same proportion of net or PC homeworkers as general workers in all countries, but the proportion is also roughly the same for mobile users and homeworkers. Clerical workers are nearly always under-represented on all forms of tele or homework, which suggests that telework is not a basis for paid domestic drudgery. The mobile phone is of some significance for craft workers, unlike, as shown above, for managerial or professional workers, which implies a fairly strong occupational basis for this technology. The other occupational groups considered are service workers and less skilled manual work. In general, with the slight exception of the mobile phone, both are under-represented in non-premises based work. In sum, teleworking is overwhelmingly a managerial and professional usage, while the mobile phone, while also important to this group, is more strongly used by craft workers. However, it is also to be noted that manual work does make some contribution to both net and PC homework, which is surprising. Taking craft and less skilled manual workers together, and even lumping in skilled agricultural workers, results are shown in Table 7.

Table 7: The percentage of each work type comprising manual workers

Net PC Mobile Home Workplace All Britain 3 18 32 14 30 25 Italy 7 16 27 21 29 25 Germany 20 18 39 26 29 29 Norway 4 8 29 8 23 18 Bulgaria - - 31 19 53 45 Israel 35 32 42 36 40 39 The figures for Net homeworkers in Germany and Israel stand out and require further investigation. It may be due to the relatively high proportion of manual workers in their labour forces (c.f. last column) but Germany is not that different in this regard from, say, Italy. If employers pay certain categories of worker more than others then this presumably reflects their differential productivity as measured by income, whether caused by teleworking itself or the fact that more able, motivated or experienced workers are likely to be teleworkers. The direct productivity effects are far from certain. What is the mechanism? It might be the ability to work better at home. Looking at this in terms of subjective indicators Akselsen et al (2001) show a positive relationship between perceived concentration at home and workers assessment of job performance, though female employees are perceived to be less effective the more days they work at home. This clearly is only a perception and implies that women are more “invisible” than men when they work at home. Alternatively, a productivity effect might result from selection of the best and most willing workers to telework, but Hartman, Stoner and Arora demonstrate in their study of variables affecting telecommuting productivity a “…lack of significant relationships between www.eurescom.de/e-living Page 15 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1 demographic and occupational variables and telecommuting productivity” (Hartman, Stoner and Arora 1991: 224). Table 8 shows the gross monthly pay of categories of homeworker (excluding Bulgaria, and excluding the self-employed, whose pay figures are harder to capture and therefore less reliable than that of employees).

Table 8: Gross Hourly Pay of Categories of Employees (in euros)

Britain Germany Israel Italy Norway Internet homeworkers 22 21 15 12 25 PC homeworkers 17 14 10 10 20 Mobile users 17 14 10 9 20 Homeworkers 17 13 12 13 21 Workplace workers 12 12 8 10 17 Except in the case of Italy, all homeworkers earn more per month than non-homeworkers, while online teleworkers mostly have the highest earnings. Broadly speaking this confirms Table 4, showing the educational distribution, and Table 6, which indicates the strong association between teleworking and managerial/professional work. However, it is of note that when we come to pay there is little to choose between PC-homeworkers, mobile users and other homeworkers. These groups differ quite significantly on a number of characteristics reviewed thus far but there is no significant pay differential, on average, and therefore implicitly rather limited productivity differentials. If this is the case, other factors than productivity must generate the propensity to adopt one work form (eg PC-based telework or homework) rather than another. The two sub-categories of homeworking vary occupationally less than in other respects. In Britain, Italy and Germany day homeworkers are more rather than less likely to be managerial or professional. This is the reverse of the implication of the previous findings that spillover workers are more highly educated and have higher PC skills. It appears that they include in substantial numbers professions other than managerial or professional jobs but which nevertheless tend to be graduate. On the other hand, this reversal is not very large. In Norway spillover workers are more likely to be graduate than day homeworkers (by ten percentage points). In Israel the number of spillover workers is so small it can be discounted. In Germany and Italy day homeworkers exceed spillover workers only by 5-6 percentage points. Only in Britain is the gap in favour of day homeworkers large – as much as twelve percentage points. In these last three countries the gap is filled by clerical or technical workers. The absolute numbers are small, but it does mean that spillover work is not the prerogative of only managers and professional workers. The distinction in pay is difficult to interpret because income questions always suffer from high item nonresponse2. In Britain it is reasonably clear: day homeworkers’ pay is 22 euros, compared to 16 for spillover workers. In Norway too, though here the figures are equal at 19 euros. In Germany and Italy, both with very limited cell sizes, day homeworkers again earn more, but the reverse applies to Israel. It appears that day homework in general conveys higher status and is more highly paid than spillover work but further analysis controlling for other variables is required. To summarise this section, teleworking is predominantly professional, but day homeworking which does not require use of ICTs has some similarities, while spillover work tends to be of lower status and pay. Net homeworkers earn more on average than other workers. Craft workers might use the mobile phone extensively but other ICTS far less so. However, both homework and telework is undertaken by a more than marginal group of blue-collar workers. 7 Home-workers’ perceived quality of life Do people who telework have higher quality of life than other workers? And if they do, is this because they telework or because those who choose to telework already have a higher quality of life? As discussed above, it is possible that teleworkers are paid more for teleworking than they would if they worked normally. It is more probable though that pay differentials between the categories of the variable are more strongly associated with the characteristics of the people within these categories. As income is generally linked to quality of life, in this case causality works from characteristics to teleworking. Equally, homeworking might conceivably improve family life but, alternatively, teleworkers might just have particular family characteristics

2 Roughly 23% of those in paid work refused to answer the income questions. www.eurescom.de/e-living Page 16 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1 which are themselves associated with higher quality of life and which would be present whether the individual teleworked or not. Causality is perhaps too complex to be assessed adequately, especially when dealing with such subjective measures as quality of life although the issue is addressed below. Here we present the quality of life score from e-Living for each of the categories of our home/telework variable. As just noted, the concept is subjective, and producing a short but appropriate scale is not straightforward. The study by the University of Toronto’s Centre for Health Promotion of quality of life relates this to physical, psychological and spiritual dimensions. It conceptualises these across three major life domains: “being”, “belonging” and “becoming3. However this scale is extremely large and not suitable for a general socio-economic household survey such as e-Living. In contrast, Diener et al’s Satisfaction with Life Scale4 (Diener et al, 1985) consists of five questions to measure quality of life and a set of five questions was developed from this scale for use in the e-living questionnaire, one general and four on specific domains. The only item specifically on work is: “In most ways my working life is close to ideal.” Respondents were asked to (strongly) agree or (strongly) disagree with this on a standard Likert (1-5) scale. Can such measures say something about teleworking? This has been tested by Akselsen et al (2001). Data were collected by a survey distributed to 217 workers, as well as their managers, colleagues and social network, in four countries. The main conclusions show that number of days worked at home is weakly associated with overall quality of life. Individual characteristics play an important part in affecting quality of life, especially self-efficacy and gender.

Table 9: Percentages of Employment Groups with High Quality of Life at Work Scores

Britain Germany Israel Italy Norway Bulgaria Internet homeworkers 69 62 48 60 78 - PC homeworkers 79 63 54 51 87 - Mobile users 81 58 45 57 90 78 Homeworkers 80 58 54 52 86 65 Workplace workers 83 55 46 60 81 46 We examine the link between quality of life and teleworking by comparing scores based on our single quality of working life measure - rather than a composite of all the quality of life measures -across the categories of our tele/homeworking variable. Descriptive analysis of the patterns of responses to all the other items across countries and social groups can be found in Ling et al (2002). As most people by far say they strongly agree that their work life is ideal, only this category is used in the analysis represented in Table 9. This shows generally rather limited variation in the scores across the categories of the variable, though rather more across countries. However, net homeworkers in Britain appear to be less willing than other workers to say their working life is ideal, while workplace workers have the highest quality of life, if by an insignificant margin. In Italy, PC homeworkers are somewhat less inclined to say their worklife is ideal but the reverse if the case in Norway. In Germany both net and PC-homeworkers are very slightly more likely to have a better working life. In Israel the picture is more mixed. Overall, it is difficult to assert that either teleworking or homeworking is associated with higher or lower quality of life from this simple analysis. 8 A Fuller Picture The report has so far tried to examine the prevalence of teleworking in contrast to other forms of work participation and to see to what extent the forms of work discussed above vary on the basis of a number of characteristics. If no distinction is made in the extent of teleworking (frequency and duration) then its prevalence can be seen to be high. While this varies across countries, the relative distribution between the different forms (net or PC homeworking, mobile use, and plain homeworking) are roughly the same in all countries (except Bulgaria). When the frequency and duration of teleworking are put into the picture a complication arises. Frequency of telework is high, especially amongst net homeworkers, but in some countries the total time spent teleworking is rather small. A little and often appears to be the rule. How real are the distinctions between the various teleworking and other forms of work? Are they all potentially the same so that teleworking could eventually be the pool of the entire set of people who do not always work at regular workplace (or several workplaces)? The analysis thus far shows that the distinctions

3 See http://www.utoronto.ca/qol/concepts.htm and http://www.utoronto.ca/qol/profile.htm for further details.

4 See http://s.psych.uiuc.edu/~ediener/hottopic/hottopic.html for more information. www.eurescom.de/e-living Page 17 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1 reflect different types of people in some but not in all respects. Amongst personal factors, age seems to matter little since the age distribution varies little across the varying forms of work across the six countries combined. However, there is a strong gender bias. Men appear to predominate in teleworking, especially net homeworking, women in non-ICT homeworking. This in turn might suggest that teleworking is the preserve of people with high-status careers while homeworking is a basis for “pin money” – a traditional way of exploiting female labour demand. In this case, teleworking and homeworking are two distinct forms of work. Homework could not then be characterised as a potential pool for extending telework because it describes different needs and different people. However, the analysis does not bear this out. While net homeworkers are very likely to be graduates, to have professional or managerial jobs, and to be paid more than others, homeworkers do not differ much from the other teleworkers in any of these respects. Moreover, they have roughly equivalent PC skills. Homeworkers are a pool of mostly female workers who could telework but do not - for whatever reason. However, this needs to be tested further and thus forms the basis of a large part of the research reported in this final section. In many respects there is rather little variation between most countries in the relationship between these factors and the forms of work at the heart of this report. The broad cross-country similarity in the distribution of the above forms of work (which, though, does not apply to time spent teleworking) might imply either some sort of organisational imperative, which determines that a certain proportion of the labour force in each country will need to work in a particular way, or a tendency for personal factors in each country to work similarly on the propensity to telework. This question also forms part of the remaining research to be reported. All the above tables and charts test the relationship between work forms and a single characteristic (age, gender, pay etc). This means it is impossible the test the relative weight of these characteristics in their relationship with the work forms. It might for instance be that gender becomes less important as an explanatory factor once other things such as family circumstances are taken into account. This would mean that it is not so much the different work circumstances of men and women which account for differences in participation in teleworking but simply differing levels of family responsibility. The reverse would imply the continuing importance of gender per se. The analysis reported in this section uses two basic multivariate procedures, logistic regression and ordinary least squares (OLS) to test various outcomes. The models focus on the following questions: • What factors affect the likelihood of being a teleworker? • What is the role of PC skills in determining work forms? • What is the impact of the work forms on people’s pay? • Do the work forms, controlling for other factors, influence quality of life? In the first two of these, the work forms which have been the basis of the research reported above are the dependent variable, ie the outcome of interest. What explains these, and what is the relative weight of the explanatory variables? In the other two, the aim is to test the impact of participation in the varying work forms on productivity (pay) and welfare (pay and quality of life).

8.1 Teleworking As a preliminary exercise, logistic regression models were run for each country to see which factors had some statistically significant association with the probability of working at home. The following had the most widespread effects, showing up as positive in all or most countries: • self-employment, • professional or managerial status, • university education, • Computer skills. Mostly, therefore, it is aspects of occupational status that indicate the probability of working at home, though PC skills extend beyond this. Additionally in some countries clerical workers undertake some work at home, and family factors such as household size have a positive effect. It is to be noted that gender is never significant, which might appear odd given the huge gender differential noted earlier. This shows that it is women’s work rather than their personal circumstances that limits their working at home. Women are www.eurescom.de/e-living Page 18 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1 relatively unlikely to have professional or managerial work, to be self-employed (their rate is half that of men in the data) or to have high PC skills (also demonstrated by the data), and these are the factors most strongly associated with work at home. Indeed, when men and women are analysed separately, self- employment has the same effect on working at home for both. Female professional workers are almost as likely to work at home as their male equivalents. Women with high PC skills are also nearly as likely to work at home as similar men. Moreover, family size, where this does have some effect, does so only on men. Hours worked has the opposite effect of what might be expected for women. The length of time a day worked by men is associated with work at home (though only in some countries). It might be expected that women who work full-time might seek to work at home. This does not appear to be the case. It is occupational factors and education which are most strongly associated with work at home, not family matters. Moving now to the teleworking categories, the six countries have been combined into one analysis in order to make the presentation manageable. Each country appears as a variable to control for country-specific effects and Norway is excluded so that the effect of the others can be compared to it. The results are shown in Table 10. The figures show how the variable in question affects the outcome, which in all four cases represents the likelihood of someone being in one work category rather than another. The contrast in each column is with workplace work, so for instance the first column shows the effects of the variables in the table on the likelihood of being a net homeworker rather than a workplace worker. The figures are derived from the odds (the probability of the event occurring relative to the probability that it will not occur). They show the influence, whether positive or negative, of the variable in question pushing odds above unity (unity means no discernible effect). As one example, the relative odds of people in Britain being a net homeworker compared to a mere homeworker are 2.2, which means that the odds are more than doubled. But for Bulgaria the figure is -4.7, which means that the odds are reduced by a factor of nearly five. Finally, the variables have different scales. For instance, the impact of hours worked shows the effect of one extra hour, and this is clearly a small scale compared to the effect of being self-employed. For such variables it is expected that the figures will be small, though possibly still highly significant statistically. First to be noted are the country differences. Overall, and discounting Bulgaria (relatively very unlikely to engage in any homeworking activity), homeworking is the least differentiated from working at the workplace in most countries (but not Germany). Only in Israel is mobile usage strongly differentiated. PC homeworking is pronounced in Britain and Israel (and by inference, Norway is less likely to PC homework than either of these), while net homeworking is strong in Britain and Germany. Second, many of the previous findings are confirmed. Occupational and educational indicators stand out. Having a degree and high computer skills are both positive in all cases, as is self-employment and professional or managerial status. Two additional variables have now been added to the previous analysis to test this further – managerial tasks (management of people, which need not apply to professional or self- employed workers) and whether respondents sets their own work schedule. Both are positive and significant throughout. It is of some interest that in respect of education and PC skills, net and PC workers are similar and somewhat “above” the other two modes of work, but net homeworking stands out most when the occupational indicators are considered. However, even homeworkers are more likely to set their work schedules than workplace workers, which implies relatively high status and responsibility for this form of work as well as for teleworking. Third, some information on PC usage at work has been added, and the results are unexpected. Except in the case of net homeworkers, if people use a PC at work they are less rather than more likely to PC homework. This is difficult to understand and remains to be checked further. Some light is thrown on the issue by information on what those who do use a PC at work use it for (where the reference group the figures are contrasted to are people who use PCs for programming, design, or desk-top publishing). PC homeworkers are most likely to use this for word-processing, perhaps suggesting high-level clerical or administrative work. But the clearest relationship between PC functions and teleworking is with net homeworking, with web management being the most surprising.

www.eurescom.de/e-living Page 19 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1

Table 10: Factors associated with net homeworking, PC homeworking, mobile use and homeworking, compared to workplace workers in each case (figures represent relative odds)5

Net PC Mobile Home Britain 1.8** 1.6*** 1.1 1.1 Italy -1.1 1.0 1.3* -1.1 Germany 1.7* 1.2 -1.2* 1.4** Israel 1.0 1.7*** 1.8*** 1.0 Bulgaria -3.9** -3.8*** -2.6*** -1.2 Male 1.1 -1.8* 2.4*** 1.2 Family size 1.4*** 1.2*** 1.1*** 1.1*** Male*family size 1.0 1.0 -1.1* -1.1* School exams 1.2 1.2** 1.1 1.1 Degree 1.9*** 1.7*** 1.2*** 1.5*** PC skills 1.9*** 1.6*** 1.1*** 1.1*** PC at work 1.4 -1.2* -1.3*** -1.3*** Word processing 1.1 1.5*** 1.1 1.2*** Web control 1.2 1.0 1.0 1.1 Spreadsheets 1.3 1.0 1.0 1.0 Email 1.3** -1.2 1.1 1.0 Self-employed 3.1*** 2.4*** 1.5*** 1.5*** Professional 1.5*** 1.3** 1.1 1.2** Managerial tasks 1.4*** 1.2*** 1.3*** 1.2** Own schedule 1.7*** 1.2** 1.3*** 1.3*** Hours worked 1.01* 1.00 1.02*** 1.00 Male*hours worked 1.00 1.02*** 1.00 1.00 Commute time 1.00 1.00 1.00 -1.01** (minutes)

N 2903 3069 3681 3197 Pseudo R2 .51 .33 .22 .14 More easy to interpret are the results relating to gender, hours worked and to family. Mobile users are predominantly male but PC homeworkers are likely to be female (over and above whatever else induces them to be in this group). All home and teleworkers, including mobile users, are likely to have relatively large families, but, as the next row in the table indicates (showing the interaction between family size and gender), in the case of mobile users and homeworkers these are more likely to be male than female. Net homeworkers and mobile users are likely to work longer than average hours but only in the case of PC homeworkers is there a gender bias here, towards men. Finally, and perhaps of considerable interest, time taken to travel to work is not related to teleworking, and appears to be negatively related to homeworking. This last result is probably an artefact, however, of the coding of the small number of people who work at home all the time as having zero travel time. When this group is excluded this coefficient becomes non-significant too. Thus, people do not work at home because they live some distance from the workplace. What do these findings mean? It is of note, looking at the final line which indicates the explanatory power of each model (and where a pseudo R2 of one would indicate a perfect explanation and zero no explanation of the outcome at all), that net homeworkers are very clearly distinguished from workplace workers with a high pseudo R2. The ability of the variables to distinguish each category of worker decreases moving to the right across the table. The final figure of .14 for homeworkers means that even this group is different from the average workplace workers but also probably that it is more heteregeneous than the other three groups. To

5 Other variables: constant; age and job permanence (non-significant); industry (varied significance); size of workplace (positive and significant for PC homeworkers and mobile users). Significance levels: * probability > 95%; ** >99%; *** >99.9%. www.eurescom.de/e-living Page 20 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1 return to the issue stated at the outset – how different are these four groups – this result alone suggests that they are, but where homeworkers are closer to workplace workers than the other groups. When we look at the actual things that distinguish the various groups – occupational indicators, education, family, and so on – the only group that really stands out is that of the net homeworkers, thus confirming the high pseudo R2. This is a distinct group although comprising both full-time and part-time teleworkers. Broadly speaking the other three groups are rather similar, although PC homeworkers have some similarities with net homeworkers – they seem a “weaker” version of this group, with the significant exception that they are more likely to be female and to do word processing at work perhaps partially confirming Stanworth’s taxonomy. Homeworkers differ from the others, however, in the likelihood of having a place of work relatively close to home.

8.2 PC skills The association between home and teleworking and education is clearcut as is that of PC skills. The above regression analysis included PC skills as a variable but it does not highlight the impact, or give information on differences between countries. Do country differences in education and training help explain the probability of teleworking. The result of attempting to answer this question is graphed in Figure 5. This shows the effect of PC skills on the odds that each mode of work will be adopted in preference to workplace work. In all six countries it is strikingly clear that the effect of computing skills on teleworking is positive - though it varies enormously by mode of work. As all the outcomes are equal to one or more, this means that the association between computer skills and workplace employment (to which each mode is being compared) is less than one. PC skills reduce the odds of being a workplace worker. The second striking thing is the consistency of the effect across countries, though Bulgaria is excluded from the net and PC homeworking analysis. These latter forms of work are most strongly associated with PC skills in most countries, but the effects for all four modes vary little across these. A third striking outcome, though, is that Britain is in a category of its own. The modes which are mostly linked to high PC skills in the other countries vary little in Britain from the other two modes, and the results in Britain are close to one (ie close to having no effect). PC skills do not differentiate modes of work very well in Britain. If someone teleworks or homeworks it is for other reasons and with other backgrounds or capabilities.

4.5 4 3.5

3 Net homework 2.5 PC homework 2 Mobile use

relative odds 1.5 Homework 1 0.5 0 GB Italy Israel Norway Bulgaria Germany

Figure 5: The Effect of PC Skills on Modes of Work (compared to workplace work)

The fact that workplace workers are likely to have limited PC skills compared to tele and homeworkers strongly implies that such a lack of skills might act as a constraint on working at home (although it is possible that these skills are at least partly acquired through working at home). However, the fact that Britain has relatively high rates of teleworking associated with rather low levels of PC skills suggests that this might not be a critical impediment.

8.3 Pay The rewards to teleworking discussed above reflect both the nature of the worker and the nature of the work. What contribution do the various forms of teleworking and homeworking make to this once other personal

www.eurescom.de/e-living Page 21 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1 and employer characteristics are netted out? This cannot be tested by looking only at the average pay of the different work groups. It has to be examined at an individual level where each individual’s pay is looked at in relation to the expected pay for people with different characteristics. This is undertaken through ordinary least squares regression. The analysis in effect reverses the direction of the previous analyses. There the issue of interest was the factors associated with someone’s participation in home or telework. Here it is the effect of this participation on pay. The figures given in Table 11 are the number of euros per hour earned by people in each category of work over and above what they would earn given their gender, age, education, PC skills, occupation and industry – in fact virtually all the variables used in the previous regression analyses. Only the results for the home and teleworking categories are shown and the self-employed are excluded for the reasons given previously.

Table 11: Contribution of forms of work to gross hourly pay, compared to workplace workers (in euros, employees only, controls for gender, age, education, PC skills, occupation, industry)

Britain Germany Israel Italy Norway Net homeworkers 3.7** 7.0*** 3.0*** -1.3 4.8*** PC homeworkers 1.9* 0.6 -1.5 -2.8 1.8 Mobile users 3.1*** 2.0** 1.3 0.5 1.2 Homeworkers 2.5** 0.2 1.2 1.8 1.8* ***=statistically significant at 99.9% level, **=99% level, *=95% Here we have a very varied picture. The category of work which is excluded - those who work in the workplace - is the group of people to whom the other groups are being compared. In Germany and Norway online teleworkers clearly earn a lot more regardless of their education and type of job. It is possible that these workers are seen as more productive by their employers or conversely that Net homeworking is a ‘perk’ awarded to (or demanded by) the most valued and hence highly paid. In Italy there are no homeworking effects at all. In Britain, PC homeworkers earn less than other categories of home and teleworker. It is not the case, therefore, that teleworking is always associated with higher pay - and therefore presumably higher productivity - though it sometimes is. PC homeworking is not that well paid relative to other categories of work once the background and other characteristics of the various workers are taken into account, and seem to earn less even than plain homeworking. Again, this may confirm Stanworth’s taxonomy.

8.4 Quality of Life Pay may have some effect on quality of life as well as reflecting differential productivity or value. How important to quality of work life, therefore, are the differing modes of work when we include pay as a variable, along with a range of other information? Because the e-Living data is derived from a general household survey with an ICT focus, it does not provide a wide range of work-related variables which are generally considered important to quality of work life such as stress, workload, control, flexibility and concentration (Akselsen et al, 2001). It is unlikely therefore that the models we can produce will be highly predictive of quality of work life. However since our aim to assess the relative significance of forms of home-working rather than develop good predictive models of quality of working life this is not a significant problem. The results, showing the tele/homeworking categories only, are given in Table 12, for all countries pooled but with country as a control variable, and first with pay as an explanatory variable, and then without. Other explanatory variables are the same as in previous models but with some minor changes. The range of quality of life is only 1-5 so the above figures sometimes show variables increasing or reducing quality of life by as much as one step in this short scale. As for the outcomes, the variable pay is itself is not significant in this model but as we can see it alters the effects of some other variables considerably, suggesting that in some cases the impact of a specific variable on quality of life works through level of pay. This is clearest in the case of PC homeworking which has a much more powerful effect when pay is not included. However, because of non-response to the pay question, and because the pay variable in this analysis excludes the self-employed, it is possible that sample differences explain some of the apparently different results. In addition all forms of home and teleworking are associated with a higher score than in the case of the missing category, workplace workers. This is most clearly the case for net homeworkers, and then mobile users, while PC homeworkers rank somewhat lower and homeworkers do not differ significantly at all from workplace workers. This contrasts with the results in Table 9 and shows the importance of controlling for other variables when conducting this kind of analysis.

www.eurescom.de/e-living Page 22 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1

Table 12: The Effect of Home and Teleworking on Quality of Work Life (OLS).

With pay Without pay Britain 0.56*** 0.49*** Italy 0.63*** 0.50*** Germany 0.76*** 0.73*** Norway 1.40*** 1.30*** Israel 0.35*** 0.39***

Net homeworker 0.37*** 0.35*** PC homeworker 0.10 0.19** Mobile user 0.24*** 0.23*** Homeworker 0.11 0.05

Adjusted R2 .17 .14 ***=statistically significant at 99.9% level, **=99% level, *=95% The highly significant results for countries are slightly artificial, as the contrast country is Bulgaria, which has a distinctively low quality of work life score (see Ling et al, 2002). It is possible that some of the differences between the other countries are not statistically significant. However, it can be seen that they are substantively different, with Norway highest and Israel lowest. Inter-country differences seem to affect the relationship between quality of work life and work modes too, but running the analysis separately for each country produces very low explanatory power in each case, with the much reduced numbers. However the results of doing so suggest that net homeworkers have the highest quality of work life in Italy, and then in Israel and Germany, with Norway coming out lowest. The scores for PC homeworking are much lower but again most prominent in Israel, Italy and Norway. Mobile usage is associated with high scores in Italy again, followed by Bulgaria and Germany. Israel has the lowest score. Homeworking has a high score only in Germany, followed by Norway. Overall, it would appear that the teleworking modes are associated with high quality of work life primarily in Italy and Germany, but that they explain relatively little of the high quality of work life scores of Norwegians. As for variables not shown (apart from pay), there is a negative relationship between hours worked, commute time and quality of work life – quite possibly there is a missing (i.e. unmeasured in this survey) mediating variable here which may well be work-related stress. However, only in the UK and Israel were there any effects for commuting time which suggests that the assumption that increasing telework might reduce commute time and thus lead to more satisfactory work conditions might not hold generally. It remains to be seen whether or not commuting time has any effect on overall life quality. The self-employed have higher scores (in the models excluding pay). This is of interest given that we have seen many homeworkers to be self-employed. Some of the differences in quality of work life scores for teleworkers may be more to do with the fact that many are self-employed rather than their ‘telework’ status, although some work modes have significant effects nevertheless. Overall there is some support from this data for the idea that certain kinds of telework might be associated with higher quality of work life in certain countries when other important variables are taken into account. The multivariate analysis shows more clearly than the descriptive account discussed earlier what these associations and inter-country differences are. What is not clear from this analysis is what causes these patterns. 9 Conclusions and Implications Teleworking, when narrowly defined as full-time work at home using ICTs, is a growing but still very limited employment pattern. Yet teleworking defined as any work at home is widespread especially amongst the self-employed. Between these two extremes we probably have some indicator of the significance of telework for the economy and society. Whether the predicted benefits will materialise – reduced costs, growth in demand for ICTs, environmental improvement – is highly uncertain, but to base any assessment of the potential benefits (but also costs) on a restricted view of the process could be misleading. Teleworking might use one or indeed several technologies. We focus on the use of an internet connection as the most ‘extreme’ form but we add to this a distinct form of teleworking, work at home based on use of a PC. We then add a more nebulous group for whose work a mobile phone is considered important, and we www.eurescom.de/e-living Page 23 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1 contrast these three groups with other homeworkers, who might work at home during the day rather than at a work premises or bring work home in the evenings or weekends. This last set of people could be considered a potential pool of teleworkers because in principle at least they may be doing home-based work which could be supported by ICTs in the future. The bulk of the report has been concerned with examining the distinctions between these groups in terms of the personal characteristics of the people found to be in them, their education and computer skills, their occupation and pay, and finally their quality of life. Through both descriptive and more complex, multivariate analysis, the aim has been to see to what extent all four groups can be distinguished from those who always work in the workplace, but also to what extent these groups are meaningfully different from each other. If they do not differ from each other then we can say nothing about their likely impact on the long-term demand for teleworking or of the impact, in turn, of this demand on the economy. Equally, if some sorts of homeworkers do not differ from workplace workers then homeworking might be of limited significance. One example of this is pay. This reflects individual productivity and/or value, if not always precisely. A part of this productivity is derived from the person’s education, PC skills, and other factors. If the different work groups tend to receive differing levels of pay once we net out the effect of education, and so on, then it is possible that these reflect productivity differences associated with teleworking. They might not be the outcome of teleworking, but might be facilitated by it. The long-term economic benefit of teleworking can be at least partially assessed through such means. At the same time, the consumption of ICTs and related services for homeworking purposes will depend on the particular form of teleworking adopted. The age of people varies little across the different work categories, but there are substantial gender differences. Mobile users are predominantly male but PC homeworkers more likely to be female. This is consistent across the different countries almost without exception. Some of the male bias in homeworking may be because the self-employed are likely to work from home and are largely male, but not all can be explained by this. Moreover, women who are self-employed and who work at home are particularly unlikely to use ICTs. Education also makes a difference. All forms of non-premises work have a strong graduate bias. While strongest for net homeworkers this includes homeworkers. All four groups are also likely to have higher PC skills than workplace workers. Net homeworkers stand out most in all respects while the other categories do not vary that much. Gender sets the PC homeworkers apart, but otherwise homeworkers, mobile users and homeworkers seem to share fairly similar characteristics. Very broadly, this applies to the distribution of occupations too, with net homeworkers having the highest proportion of managers and professionals. Pay is also highest for this group, but is more variables across other groups, especially when country is taken into account. Only quality of work life diverges from this consistent pattern, with a generally less predictable distribution. With respect to the ‘telework’ effects on quality of work life, this analysis cannot separate cause from effect. We should not naively imagine that converting more of the workforce to Net homeworkers or mobile-using workers would necessarily increase their quality of work life. There may be particular characteristics (not measured here) of Net homeworkers in the UK, Germany and Norway which leads them to have higher quality of work life scores irrespective of the fact that they use the internet whilst working at home. It remains for future (longitudinal) research to analyse any possible changes in quality of (work) life for particular groups (such as those with limited mobility or otherwise disadvantaged) who adopt teleworking as a means of integration into the labour force (SEC (2001) 1428, p21). It also remains to examine the degree to which teleworkers can better integrate work, home and social/community activities (SEC (2001) 1428, p21; SEC (2002) 372, p 27) given our finding that the majority of those who did any work at home did not do so full- time (see Figure 2) and the aggregate time that they do is rather low. This being the case it would seem that ‘telework’ as presently practised is not likely to lead to major change in the lives of those who do it because it is part of a flexible pattern of work rather than a binary work/home-work division and it is not ‘done’ that much. It also means that ‘effects’ will be rather more disparate and harder to detect. It would therefore seem that although the definition of telework can be extended to minimise the role in this of time (frequency or duration of work at home), technology makes a difference. Net homeworkers may be considered currently the most ‘extreme’ form of teleworker and they differ from other types of worker in several important ways. The other three groups are all clearly different from workplace workers and it can therefore be assumed that their work at home is not a matter of chance: it varies with their education, PC skills, job, pay and other organisational factors not measured here. The former do not on average help us distinguish these three work modes, but gender helps us set apart PC homeworkers from the rest. However, the multivariate analysis reveals a number of factors that distinguish these three groups, with the most prominent being the country of the survey. Even though countries seem not to differ very much in the way www.eurescom.de/e-living Page 24 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1 that these various characteristics are associated with modes of work, the prevalence of the workstyles varies by country over and above the effect of these factors. However, Britain does stand out from the other countries in that PC skills vary across the work categories far less. This suggests that either teleworking and homeworking in Britain do not require high PC skills or that those skills are so widespread in the working population that their effect is not significant. The potential for increased teleworking is reinforced by the fact that the characteristics which best predict whether someone will be a net homeworker or PC homeworker or plain homeworker are not by and large mutually exclusive. Net homeworkers and homeworkers stand out in several ways but people’s backgrounds and circumstances do not overwhelmingly predict into which of the four tele and homeworking groups people might fall. This should come as no surprise given the evidence in the literature (Haddon and Silverstone, 1994; Bailey and Kurland, 2002) and from the FAMILIES project qualitative study (FAMILIES, 2002) that the reasons for starting to telework are extremely wide ranging. This does not mean, of course, that teleworking actually has benefits. Net homeworkers appear to have the highest quality of life, and this might derive in part from the fact that they telework, but this cannot be proved directly. Pay tells us more. Here again, except in Italy, net homeworkers are paid the most. Possibly this is because these people are the most productive or they may be the most valued in certain industries. It is impossible to ascertain whether their teleworking contributes to this. But it is of particular note that homeworkers, far from being an exploited group of female, traditional homeworkers, not only are relatively likely to have degrees but to be well paid too. It is the PC homeworkers who, generally female, get paid the least of the four groups, and in some countries even less than workplace workers. If there is an exploited group, this might be it and policy making in pursuit of ‘telework’ would do well to consider them more explicitly. We should also draw attention to the finding that only in the UK and Israel was there any relationship between commute time and quality of work life. This result confirms the view in the literature that on the whole travel reduction is not a major inducement for telework (Bailey and Kurland, 2002, p387). Whilst the relationship between types of telework and overall quality of life warrants further research, this does at least suggest that reducing commute time by promoting telework may not necessarily increase the quality of work life in every country. Given the uncertain outcomes and variability of our results in terms of associations between pay, quality of work life and forms of telework, we hope that the results reported here can help to generate “a much higher quality of debate on teleworking” (HLEG (IS) 1997, p 42). We also hope it has started to provide ‘a detailed assessment of the numbers of men and women currently involved in telework, the activities they are engaged in, the skills required and the social consequences.’ (HLEG (IS) 1997, p43) As we have seen, the answers are not simple and straight-forward. Thus policies to promote telework will also have to be complex and multifaceted taking account of both individual and country/regional variations. 10 Bibliography Akselsen, S. (2001) (ed) “Telework and Quality of Life: Basic Concepts and Main Results”, Eurescom Project P904: The Impact of Telework on a Sustainable Social Development and Quality of Life Anderson, B (ed) (2001) e-Living D3: State of the Art Review. E-Living Project Deliverable. www.eurescom.de/e-living. Bailey, D. E., and Kurland, N.B. (2002) A Review of Telework Research: Findings, New Directions, and Lessons for the Study of Modern Work., J. Organiz. Behav., 23: 383-400. Brocklehurst, M. 1989. The reality and rhetoric of new technology homeworking: some lessons for personnel managers. Personnel review, Vol. 18, No. 2. COM (2002) 263, eEurope 2005: An information society for all, Communication from the European Commission, COM (2002) 263 final. Brussels 28.5.2002. Cowan, R. (1983) More Work for Mother: The Ironies of Household Technology, London: Basic Books. Diener, E., Emmons, R.A., Larsen, R,J., and Griffin, S. (1985) The Satisfaction With Life Scale, Journal of Personality Assessment, 49 (1).

www.eurescom.de/e-living Page 25 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1

ECaTT (2000a) Benchmarking Progress on New Ways of Working and New Forms of Business across Europe, ECaTT Final Report, Empirica: Bonn ECaTT (2000b) Telework Data Report (Population Survey): Ten Countries in Comparison, Empirica: Bonn eWork (2001) Status Report on New Ways to Work in the Information Society, European Commission FAMILIES (2002) Deliverable No 3: Results of the family survey. http://www.families- project.com/doc/families-d3-public.pdf Felstead, A., and Jewson, N. (2000) In Work, At Home: Towards an Understanding of Homeworking, London: Routledge Gallagher, M. (1987) “Redefining The Communications Revolution”, in H. Baehr and Dyer, G. (eds), Boxed In: Women and Television, London: Pandora Press Gillespie, A., Richardson, R., Cornford, J. 1995. Review of telework in Britain: implications for public policy. A report for the Parliamentary Office of Science and Technology, Sheffield, Employment Department, February. Gregg, P. and Wadsworth, J. (1999) “Job Tenure, 1975-98”, in Gregg, P. and Wadsworth, J. (eds.) The State of Working Britain, Manchester University Press. Haddon, L. (1999) Approaches to Understanding Teleworking, Telekronikk, No 4, Telenor, Oslo. Haddon, L. and Lewis, A., (1994) The Experience of Teleworking: An Annotated Review, International Journal of Human Resource Management, 5 (1) Haddon, L. and Silverstone, R. (1993) Teleworking in the 1990’s – A View from Home, A Report on the ESRC/PICT Study of Teleworking and Information and Communication Technology, CRICT Policy Research Unit Haddon, L. and Silverstone, R. (1994) “Telework and the Changing Relationship of Home and Work”, in Mansell, R. (ed) (1994) Management of Information and Communication Technologies: Emerging Patterns of Control, Aslib, London. HLEG (IS) (1997). Building the Information Society for us all: Final policy report of the high-level expert group. European Commission, April 1997. Hartman, R.I., Stoner C.R. and Ray Arora (1991), An Investigation of Selected Variables Affecting Telecommuting Productivity and Satisfaction, Journal of Business and Psychology, 6 (2) Huws, U. (1993) Teleworking in Britain: a Report to the Employment Department, Research Series No. 18, October Huws, U. (1995) Social Europe: Follow up to the White Paper – Teleworking, Office for Official Publications of the European Commission Kraut, R (1989). "Telecommuting: The Trade-offs of Home Work". Journal of Communication, Vol. 39, No. 3. OECD (2000), OECD Information and Technology Outlook, Paris ONS (2002) Labour Market Trends, Vol 11, No 6, The Stationery Office, Norwich. Ling, R., Yttri, B., Anderson, B., and Diduca, D., (2002) Age, gender and social capital- a cross sectional analysis. D7.4, Public deliverable of the e-Living IST project. August 2002. . www.eurescom.de/e-living/. Nilles, J.M. (1975) Telecommunications and Organizational Decentralization, IEEE Transactions on Communications, Com23:1142-1147 Papadakis, M. and Collins, E. (2001) The Application and Implications of Information Technologies in the Home: Where are the Data and What do they Say?, Division of Science Resources Studies, National Science Foundation, USA Schepp, D. (1995) The Telecommuter’s Handbook, McGraw Hill Raban, Y., Soffer, T., Mihnev, P., Ganev, K. (2002) ICT Uptake and Usage: A Cross-Sectional Analysis. D7.1, Public deliverable of the e-Living IST project. August 2002. www.eurescom.de/e-living/. SEC (2001) 1428. e-Inclusion: The Information Society’s potential for social inclusion in Europe. Commission Staff Working Paper, Commission of the European Communities.

www.eurescom.de/e-living Page 26 of 27 PUBLIC e-living-D7.5-Homeworking-and-Teleworking-Issue-1.1

SEC (2002) 372. Information Society Jobs – Quality for change: Exploiting the Information Society’s contribution to managing change and enhancing quality in employment. Commission Staff Working Paper, Commission of the European Communities. Standen, P., Daniels, K., and Lamond, D. (1999) The Home as a Workplace: Work-family Integration and Psychological Well-being in Telework, Journal of Occupational Health Psychology 4 (4): 368-381 Stanworth, C. (1997) “Telework and the Information Age”, New Technology, Work and Employment, 13 (1): 51-62 Sullivan, C. (May, 2001) Living with a Homeworker: Female and Male Partners’ Experiences of The Management of Work And Family and Family Satisfaction, (mimeo, ESRC) Sullivan, C. and Lewis, S. (2001) Home-based Telework, Gender,and the Synchronisation of Work and Family: Perspectives of Teleworkers and their Co-residents, Gender, Work and Organisation, 8 (2): 123- 145. Van Sell, M. and Jacobs, S (1994), Telecommuting and Quality of Life: a review of the literature and a model for research, Telematics and Informatics, 11(2): 81 – 95. Webster, J. (1996), “Revolution in the Office? Implications for Women’s Paid Work”, in Dutton, W. (ed), Information and Communication Technologies: Visions and Realities. Oxford: OU Press.