National Institute of Statistics

NATIONAL INSTITUTE OF STATISTICS

ROMANIA

DEVELOPMENT OF DETAILED STATISTICS ON ENERGY CONSUMPTION IN HOUSEHOLDS (SECH Project)

GRANT AGREEMENT No. 30304.2009.003 - 2009.707

FINAL REPORT

GRANT AGREEMENT No. 30304.2009.003 - 2009.707

FINAL REPORT

ENERGY CONSUMPTIONS IN HOUSEHOLDS

1. General presentation

In order to comply with the requirements of Regulation 1099/2008 on Energy Statistics, Member States have to develop further energy statistics related to energy consumption and renewable energy. These detailed statistics are an important tool in the assessment of energy efficiency/ energy savings of each country.

In the “Energy Statistics Working Party”, held in June 2008, it was recognised that surveys of the industry are common practise in EU, therefore improvement of energy consumption statistics in other sectors, such as households, were considered high priority.

Although the Romanian National Institute of Statistics has a long experience in computing the national energy balance, we also considered that there were domains (such as households consumption) that have to be developed and improved for more accurate and reliable results.

Currently, some information on the households consumption of energy commodities were estimated based on data from Household Budget Surveys and other administrative sources. The obtained data were good, but only for a few main energy commodities (such as electricity, natural gas and fuel wood) and too general, with a low (or zero) degree of detailed information regarding other energy carriers and the purpose of the consumption (space heating, water heating, cooking, etc).

The survey dedicated to energy consumption in households, along with the already existing annual surveys in energy statistics carried out by the Romanian NIS, will be an important source of information for policies and other bodies interested in promoting energy efficiency/ energy savings programs.

This project, implemented within the framework of the European Statistical System (ESS), helped us to develop further detailed statistics on energy consumption in the households sector and the data will substitute the estimated data currently used to compute the national energy balance.

2. Scope of work

The objective of this project was to develop further detailed statistics on energy consumption in the households sector, using the coverage recommended by the June 2009 ESWP as a basis, and taking into account the national priorities- that arise from policies pursued at national level- and the current national practices.

The project was performed using a dedicated survey, with face to face interviews of households on a sampling basis, to determine the energy consumption, related costs and type/age of energy consuming appliances, for the reference year 2009.

The data are significant at national level (NUTS 0) and on urban and rural areas.

Based on modelling, the consumption on major end uses was determined (space heating, water heating, cooking, space cooling, lighting and electrical appliances).

3. Description of the work

The project was based on a data collection exercise, by way of a survey of households, based on a sample of dwellings and a survey questionnaire.

Data were collected with the help of 780 interviewers, specialised in collection of data from households, the interviewers being used constantly by NIS for this type of surveys.

Interviewers were briefed by the NIS experts working in regional statistical offices, in order to have a good knowledge of energy units (kWh, GJ, etc) and local distribution and sale of energy commodities. Interviewers were also required to regularly report all problems experienced with the questionnaire or the interviewing process to the county offices.

Throughout the data collection stage questionnaire responses were checked for completeness and internal consistency.

The data were keyed-in on line in a data base that permits additional checks and electronic treatment of data in preparing the necessary report tables at national level.

4. Main activities

4.1. Documentation and project preparations

The project has started in January 2010 with the preparation of the survey, documentation regarding the objectives of the project and the actions to be performed.

It was necessary to read and get familiar with the methodological issues presented in the Recommended Coverage and the definitions sent by Eurostat.

The questionnaire, methodological instructions and guidelines and the results of a previous survey, performed in 1996, were also studied, as well as the documents presented on CIRCA as results of Task Force “Final Energy Consumption in Households” from October 2008.

An important mission of the project leader was to train the staff (including the experts from the regional statistical offices) involved in this project and to prepare the programme of the territorial network.

Also, in order to inform the population about the purpose and schedule of this survey and to allow householders to prepare past invoices and accept the interviewer more readily in their home, NIS created a Press Release and an official notification (Annex 1) that was printed and sent to the households comprised in the sample.

4.2. Questionnaire and Instructions design

Based on the methodological requirements of Eurostat and on past experience of questionnaire and methodology, as well as the experience of other Member States, NIS has designed the national household consumption statistical questionnaire (CEnG) for 2009 (Annex 2; the English version of the indicators are presented in Annex 2_English).

As a guide for interviewers and staff as well, the fill-in Instructions were designed; they contain the main check keys and informative descriptions and pictures (Annex 3).

The statistical instruments (questionnaire and instructions) were printed and distributed to the county statistical offices.

4.3. Sampling

The Romanian Households Energy Survey is based on a two-stage sampling design, with the stratification in the first stage. The sample size is 10920 dwellings with permanent occupancy. Keeping in mind the proposed survey design and the expected response rate (80%), the sample was being design to produce reliable estimates for each area of residence (urban / rural) as well as national estimates for main variables.

First Stage Sampling (master sample EMZOT)

The master sample EMZOT was designed on the basis of the Census of the Population and Dwellings from 2002. Even if the use of a master sample makes the estimation process more complicated, it is the only way to carry out household surveys at the time being.

EMZOT contains 780 Primary Sampling Units (territorial areas), selected separately on the 88 strata. Criteria by which the stratification has been made are county (NUTS-3 level) and area of residence, resulting a number of 88 strata (in Bucharest the selection was done separately for each of the six sectors).

The 780 PSU are distributed in the master sample as follows: 427 PSU in urban area and 353 PSU in rural area. On the first stage, probabilities of inclusion proportional with the size were calculated in each stratum. The size criterion was the number of permanent dwellings of the PSU.

Second Stage Sampling

Within each PSU, 14 dwellings are sampled using a systematic algorithm of selection with random start. The second-stage inclusion probability is the inverse of the sampling interval inside a PSU. At the end, the probability of selection of a dwelling in the sample is calculated as product of two inclusion probabilities: the first-stage inclusion probability and the second-stage inclusion probability.

4.4. Study visit to Statistics Austria

Romanian NIS asked, and it was granted, to use the Austrian experience in households energy consumption surveys.

The main topics of the agenda were:

-Presentation of the Romanian SECH project: expected results of the action, the questionnaire.

-Presentation of Austrian “Domestic energy consumption survey” and “Survey on electricity and natural gas consumption by purposes”

-Presentation of interviewer training elements, validation of primary data, non-response imputation, validation of final data.

-The butt for data adjustment; types of data adjustment;

-Modelling procedures in order to determine energy consumption by end use.

The questionnaire designed by NIS Romania was first discussed, along with the expected results of SECH project.

The surveys conducted by Statistics Austria were presented and discussed: data processing, modeling procedures, data validation and especially the target determination of a household’s thermal energy consumption were the most important items explained during the discussions; the plausibility checks (both during and after the interview), item non response imputation and the shares’ modelling of energy needs for space heating, water heating and cooking was given.

An important discussion was about how to adjust the Austrian average energy consumption values to the Romanian specific energy situation and where to obtain further relevant information in order to build up the target.

The Austrian questionnaires used were discussed in detail and special attention was paid to potential problems in collecting the data.

The results of the IEE project “REMODECE” were analysed with focus on the Romanian results; their applicability for future calculations of electricity consumption was declared to be acceptable.

4.5. Design of IT software for data entry and control of primary data

The IT software was designed, based on the questionnaire and the specifications necessary for data entry and control, and afterwards for imputations. There were discussions between project team and IT experts of NIS, for a better understanding of the terminology, the sense of the questions and the check keys.

The on-line IT software permitted data entry and check for completeness and internal consistency, eliminating the errors at unit level, simultaneously with data checking at central level.

4.6. Data collection and data entry

The period for data collection was April- May 2010 and by the end of June 2010 the data entry was finished.

At regional level working teams were established for data collection, comprising the coordinating staff and the operating staff.

The statistical instruments were distributed and the local working groups assisted the interviewers in identifying the households and provided technical assistance when needed.

The whole activity of statistical data collection, carried out under the supervision of NIS, complied with the legal provisions.

The results of data collection, in terms of response rate, are as follows:

·  14 dwellings were inhabited by two households and 1 dwelling by three households.

·  The total response rate is 88.5%.The urban response rate is 85.9% and the rural response rate is 91.6%.

/ Total /

Urban

/

Rural

Number of dwellings in the sample

/

10920

/

5978

/

4942

Number of dwellings with valid data

/

9661

/

5135

/

4526

Number of households

with valid data /

9676

/

5141

/

4535

4.7 Validation and data processing at national level

Validation of data at national level was a long and time consuming process. The team verified to the highest possible extent if the reported data respected the prescribed methodology.

In phase one, the errors that occurred at questionnaire level (and failed to be solved by the county statistical offices) have been dealt with; they were mainly partial non-responses and inconsistencies in data reporting (for example, the relation between fuel and the device using that fuel, unrealistic quantities or values reported, considering the size of the household and type of the dwelling, etc); the county offices were contacted and asked to solve these errors.

In phase two, the data have been subjected to the following treatment:

·  the missing heated floor space was imputed by the total living floor;

·  if no light lamps were reported, 2 incandescent light lamps / room were imputed;

·  the missing quantities of fuel consumption were calculated from the monetary values by applying average prices and vice-versa;

·  the missing quantities and values were replaced with the average data from donors belonging to the same area of residence with ±10% deviation of characteristic keys (type of building, heated surface and number of persons in household);

·  the quantities of fuel reported in secondary measure unit have been transformed in primary measure unit;

·  quantitative consumption checks against an upper and lower level of fuel consumption (these levels are based on information from various sources, such as REMODECE studies, Statistics Austria, internet, etc) ;

·  quantitative consumption checks against an interval (-50%, +100%) of default values (based on Statistics Austria information) for annual energy use for heating, in GJ/m2;

·  quantitative consumption checks against an interval (-50%, +50%) of default values (based on Statistics Austria information) for:

- annual energy use for water heating, in GJ/person;

- annual energy use for cooking, in GJ/person;

4.8 Weighting and calibration of data

Weighting is a three step procedures:

a) Basic weight calculation. The corresponding basic weight of a sampled dwelling is the inverse of the probability of selection. Each household belonging to a sampled dwelling takes the basic weight of the dwelling.

b) Unit Non-Response Adjustment. To cover the non-respondent sampled households the respondent units are re-weighted. Response Homogeneity Groups are constructed by crossing the following variables: county and residence area. The basic weight of each respondent household belonging to such a group is adjusted with the inverse of the corresponding response rates of the group.

c) Calibration. The final weights are obtained by calibrating the weights issued from the previous steps on certain variables of interest, for which totals are known for the entire population. Demographic variables (population on gender, age groups), locating variables (population on residence area, region (NUTS2 classification)) and number of households for every region are used for this final adjustment. The structure of the population for the above-mentioned variables is known from the legal population, available at 1st of July 2009. The calibration is carried out within each region, by means of CALMAR SAS macro. At the end of this step, the final weights are obtained.

4.9 Final data checking and analyse

The final data have been computed applying the weighting coefficients to all variables, obtaining two sets of data: one at the level of household (in 2009 were 7395749 households) and one at the level of dwelling (7383643 occupied dwellings in 2009).

Due to the small number of dwellings with two or three households (12106), the comparative analysis of these two sets of data revealed differences of less than 0.3% in terms of fuel consumption.

For this reason, and because of the doubts concerning the correct split of ownership regarding the appliances in dwellings with more than one household, we have decided to analyse all data at the level of dwelling.