A COMPARATIVE STUDY OF COMMUTER PATTERNS AND TRENDS IN GREAT BRITAIN, IRELAND AND THE US

Ian N Williams Ian Williams Services and University

1 INTRODUCTION

There are a number of reasons why we need to understand better the forces that shape commuter travel. It continues to be a major user of road and rail capacity in congested peak periods so that forecasting the future pattern of commuter trips is a critical task for most urban passenger models. Commuter travel by car has also been a substantial contributor to the past growth in greenhouse gas emissions so that its future growth trajectory is important to the environment. The quality of the transport facilities that connect homes to workplaces affect the land use policy options available to planners and influence the longer term success of planning policy decisions on residential and business construction location decisions.

Patterns of commuting have changed through time but these changes do not always follow simple linear trends. In many countries the long-run trend of increases in commuting trip lengths of the 20th century had eased by the last decade. We need to understand the extent to which this is a result of changes in the underlying patterns of the behaviour of workers within the socio-economic environment they inhabit and the extent to which it is simply a reflection of changes in the transport supply characteristics such as cost and time of travel that are standard inputs to our transport models.

The reason for initiating the comparative study presented in this paper was to examine those aspects of commuter patterns that have been shown to be similar across a range of countries and a range of years. Our assumption when forecasting travel demand is that such phenomena are more likely to persist in a stable fashion through into the future than those travel patterns and behaviours that have been observed in only some locations or in some years. This analysis also provides guidance on the most appropriate segmentation of commuter types to adopt within travel demand models so as to enable stable and reliable forecasts of future travel demand to be made.

Section 2 outlines the data sources on which the comparative analysis of computing patterns is based. These data are then used in Section 3 to identify the main regularities in commuter travel behaviour across countries and through time. Section 4 speculates on some emerging trends and influences that may change future commuter patterns and so these should be considered within our designs for forecasting models. The reasons why it is important to provide an appropriate degree of segmentation in travel demand models are explained in Section 5. Finally, conclusions are drawn in Section 6 on the manner in which the design of commuter travel demand models should be structured and segmented so as to

© Association for European Transport and Contributors 2012 1

enable them to generate robust and realistic future forecasts for use in assessing land-use and transport policy decisions.

2 DATA SOURCES

This study focuses on commuter travel patterns in three countries: UK, Ireland and the US, with a particular interest in identifying those behavioural characteristics that appear to have similarities across countries and through time.

The main data sources on commuter travel available for each of the three countries are summarised in Table 1. Typically the travel surveys provide detailed data on trends through time and on the characteristics of the trips, whereas the Census data sources provide good spatial differentiation on travel patterns and on the characteristics of the workers. However, for the Census data in particular, a substantial effort is required to convert the data to a form that is adequately consistent between years. The aim should be to avoid the emergence of apparent behavioural trends that in reality are just a side effect of differences between years in ways in which the data collection has either been carried out or processed.

Table 1 Sources of data on commuting by country Country Source Years Comment UK NTS 1985/86, 1989 onwards LFS 1992 onwards Mode and travel duration (mins) Census 1981, 1991, 2001, 2011 Mode and travel distance Ireland Census 1981, 86, 91, 96, 2002, 06, 11 US NPTS 1977, 83, 90, 95 NHTS 2001, 2009 Replaces the NPTS Census 1980, 1990, 2000 ACS 2005 onwards Replaces the Census

Access to sampled individual population Census records on a comparative basis across countries has improved significantly in recent years as a result of the initiative to assemble census data within the International Integrated Public Use Microdata Series (IPUMS - https://international.ipums.org/international/) by the University of Minnesota. This data is assembled on a reasonably common basis at the anonymised individual level for a large number of countries and for many years for each such country.

As yet no detailed data on commuter travel has been published from the most recent 2011 Census in either the UK or Ireland. When published, these data will provide interesting insights into the impacts of the economic recession on commuter behaviour.

There are other data sources that also provide insights into commuter behaviour that are not discussed here. For example, Dargay and Hanly (2003) use British Household Panel Survey data from 1991 onwards to examine throughout a 10- year period the year-to-year changes for individuals in their commuter travel and residential and workplace locations. Lyons & Chatterjee (2008) provide a wide- ranging review of the behavioural forces that influence the long commuter

distances now travelled, drawing innovatively on research from the fields of psychology, sociology and medicine. Williams (2005) reviewed the main influences on commuter patterns, concentrating on the use of 2001 UK Census data to examine cross-sectional influences in the UK so that it complements the focus here on trends through time and on international comparisons.

3 COMMUTING PATTERNS AND TRENDS 3.1 Average trip lengths by segment In this section we examine regularities across countries in the pattern of average commuting trip length for sub-groups of the working population.

Figure 1 Commuter distance (one-way crow-fly kms) by industry (SIC), sex, part-/full-time – internal trips within , South East and East of Source: UK 2001 Census

Table 2 Key to Standard Industry Codes (SIC) Code Industry Code Industry A Agriculture, hunting, forestry I Transport storage and communication B Fishing J Financial intermediation C Mining and quarrying K Real estate, renting and business activities D Manufacture L Public administration & defence, social security E Electricity, gas and water M Education supply F Construction N Health and social work G Wholesale and retail trade, OPQ Other repair of motor vehicles H Hotels and restaurants

Using 2001 Census journey to work data for the wider South East set of regions, Figure 1 illustrates how journey to work distances differ systematically by type of worker. The main findings are:  Males travel further than females in almost every specific sub category;  Full-time workers travel further than part-time workers in almost every specific sub category;  The various sub-categories of worker in higher income service industries (SICs: J, K, L) as well as in transport (SIC I) have longer than average journeys, while those in SICs AB (agriculture) and H (catering) have the shortest average distances.

The same broad pattern of trip length differentiation by segment is found in the rest of the UK, as illustrated by the histogram in Figure 2 for female workers resident in the East Midlands.

15.00

10.00

5.00 F1-15 F16-30 F30+

0.00

AB

F

G

CDE

H I

J F30+ K

L F16-30

M F1-15

N OPQ

Figure 2 Commuter distance (one-way crow-fly kms) for females by Industry and hours worked - residents of the East Midlands Source: UK 2001 Census

Men commute further than women also in the US and in Ireland. In Ireland the excess distance for males relative to females was 28% in 2006 which is little changed from the 27% rate of 1981, despite the overall growth of around 55% in the average trip length for each sex over this 25 year period to give an overall average of 15.6km in 2006.

In Great Britain in 2009 (DfT, 2011), males travelled 16.4km to work on average, which is 51% further than females (10.8km), with an overall average of 13.2km.

The high proportion of part-timers among female workers (Figure 3) tends to push down the overall average trip length for females relative to males in the UK.

In the US (Santos et al. 2011) the average commuter trip length in 2009 was 19.0km, which is significantly longer than in the UK or Ireland.

Figure 3 Number of full- and part-time workers by sex by year, UK (000s) Source: Table EMP01, ONS, http://www.ons.gov.uk/ons/dcp171766_276987.pdf

Having analysed the differences in commuter distances between groups in the workforce we can see how this in turn will impact on the overall growth in commuter travel through time. Figure 3 illustrates that only 10% of the growth in employment in the UK since 1992 has been in the full-time male group - that which has the longest average commuter trip lengths. The 90% majority of employment growth has been in the other categories all of which tend to have relatively shorter trip lengths. This structural change in the workforce in the economy will have acted through time to depress the rate of growth in the overall UK commuter trip length.

When forecasting overall commuter travel demand into the future it will be important to consider what assumptions should be made:  about whether the recent relative growth in the proportion of part-time working, particularly for men, will continue or reverse; and  about the future balance of the proportion of females to males in employment.

3.2 Mode choice trends In this section we examine regularities across countries and through time in the pattern of use of modes and in how this usage differs between segments of the working population.

Figure 4 Commuter choice of mode by year in Great Britain (%) Source: GB NTS

Figure 4 indicates that in more recent years there has been a minor reduction in the proportion of car/van commuters in Great Britain from 69% in 1996 to 67% in 2010. This represents a change in trend from the continuing increase in car use that had occurred through many decades in the last century. Over the last decade both rail and bus commuting have increased significantly albeit each from a small base percentage.

This reduction in car use is pronounced in the larger cities, particularly London where cycle, bus and rail/LU usage have all eaten into the car mode share. For all persons entering Central London in the AM peak period (during this period entrants are mainly commuters), the car share has reduced by more than 2/3rds from a high of 19% in 1982 gradually down to 6% in 2010. Only 2 points of this 13 percentage point reduction occurred between 2002 and 2003 in response to the congestion charging implementation so that most of this reduction in car use appears to have been caused by a combination of increased cost and difficulty in parking and of increased congestion resulting from reductions in road capacity. Analogous reductions in AM peak car entries have been observed in other major UK cities in tandem with decreases, not increases, in speeds (Devereux and Williams, 2010).

Figure 5 illustrates how the pattern of choice of mode for commuting has evolved since 1981 In Ireland.

Figure 5 Daily commuter trips by mode by sex by year in Ireland (000s) Source: Irish Census

It indicates that the size of the female workforce in Ireland grew rapidly throughout the period from 1981 to 2006. For female commuters the right side of Figure 5 shows that: • most of the absolute increase in female commuters was captured by the car driver mode which increased from a 24% share in 1981 to a 63% mode share by 2006; • rail (inc. DART + Luas) grew five-fold in absolute terms which doubled its mode share from 1.7% to 3.4% over this period; • all other modes exhibited major reductions in mode share, even though some of them had increased significantly in absolute terms due to the growth in the female workforce.

The size of the male workforce in Ireland was fairly stable from 1981 until 1996 then grew by 35% up to 2006. The changes in male mode choice from 1981 are: • a major reduction from 24% to 6% in the proportion with no journey to work, which is largely due to the major reduction in the number working in the agriculture, forestry & fishing sector; • this reduction in those working at or from home has been balanced by growth in the shares of car driver from 43% to 54%, of lorry and van (i.e. “other”) from 3% to 13% and of rail from 1.2% to 2.6%; • coupled with small declines in the shares of the modes car passenger, bus, walk, cycle and motor cycle.

In contrast to Europe, in the US the private vehicle mode share for commuter trips has been high for many decades and has increased slightly over time from 87% back in 1977 to 90% by 2009.

The Irish modal data highlights a number of important topics that are relevant to commuter model design and to forecasting.

Firstly, the pattern of mode choice and its trend through time differs considerably between males and females. Figure 6 for the wider South East of England confirms analogous differences in 2001 between males and females in their choice of mode. It breaks the trips into distance bands to enable the differences in available modal supply by distance band to be taken into account. In general, females at any given distance have a greater propensity to use bus or walk modes, whereas cycling and car are relatively more used by males.

Figure 6 Proportion of commuter trips by mode in each distance band by sex, for London, South East and East of England, 2001 Census Sources: UK Census 2001

Secondly, the proportion of the population that works at home impacts on the overall demand for transport. Traditionally in the UK this proportion has been low due to the absence of a large agricultural workforce and to the decline through time in self-employed shop-keepers and others living on the premises. However, as discussed further below in Section 4, there is evidence of growth in the rate of home-working in recent years in Great Britain and the US as illustrated in Figure 7, while the decline in Ireland from the high level in the 1980s appears to have flattened out by 2006.

Figure 7 Proportion of those who work at home by year by country Sources: Irish Census, GB NTS, US Census & NHTS 2009

3.3 Trends in average distance and travel time Having examined some of the cross-sectional influences on average trip lengths we now examine in greater detail the trend in average trip lengths through time and discuss the factors that have influenced its growth rate.

Figure 8 presents the evolution of the average trip length at the national level for all commuter trips and for those by car. In the US the all mode and the car trends are necessarily very similar due to its continuing high car mode share.

Figure 8 Average one-way commuter trip length (kms) by year by country by car and across all modes Sources: Irish Census (excludes those working at home), GB NTS, US NPTS & NHTS

The Census data for Ireland from 1981 onwards covers both the economically depressed decade up to 1991, which had 8% growth in trip length, and the subsequent very rapid urban population and economic growth in the 16-year „Celtic tiger” era up to 2007. This latter period exhibited a further 40% trip length growth in the 11 years to 2002 but just 4% growth in the subsequent 4 years, by which time road congestion had grown substantially. The trip length for mode car has been higher than the overall average for Ireland throughout the period but with a gradual lessening in the excess distance from 24% in 1981 to 11% in 2006, as the proportion commuting by car has increased and as the ratio of females to males among car drivers has increased.

In Great Britain, the overall average commuter trip length grew by 34% in the decade from 1986 to 1996 but since then growth has flattened to increase only by a further 13% to 2010. In particular, car commuter trip lengths have remained relatively stable at around 15 kms since 1996 so that their excess distance above the average has reduced from 13% to 5% by 2010.

In all three countries the increases in average commuter trip lengths of earlier decades have flattened out or reversed in more recent years. Some of the potential underlying reasons for this reduction in trip length growth are discussed below here as well as in Section 4.

Figure 9 Average commuter travel speed (kph) by year by country by car and across all modes Sources: GB NTS, US NPTS & NHTS

Figure 9 illustrates that average commuter speeds were increasing in the US and Great Britain up to around 1996, after which they started to reduce. Part of this reduction is a result of increased road congestion in the peak periods in the urban areas that contain large numbers of jobs. In the US the significant decline in car speeds between 1982 and 2007, resulting from congestion delays (Figure

10), particularly in the congested larger metropolitan areas, has been associated with a change in trend there to a reduction in the average commuter distance travelled in order to compensate for the extra travel time needed.

Figure 10 Annual person-hours of road traffic delay per car commuter in US urban areas, by population size Source: Texas Transportation Institute (2011)

The impact of structural economic change on the growth in average trip lengths can be clearly seen in Ireland where between 1981 and 2006 the average commuter trip length grew by 56% for males, by 55% for females but only by 52% for workers as a whole. The reason for this lower growth for the whole workforce is that the proportion of the workforce that is female (who have relatively short trip lengths) increased strongly from 29% in 1981 to 43% by 2006.

The evidence for the influence of economic growth on the growth of average trip lengths is ambiguous. The evidence from Ireland between 1981 and 2002 is consistent with a strong correlation between the rate of GDP growth and of increase in average trip length. However, in the period from 2002 to 2006 the annual real growth rate in GDP was at the high rate of 5.2%, though down from 10% per annum in the previous decade but as illustrated in Figure 8, the annual growth rate in trip length dropped to 0.6% for 2002 to 2006, down from 3% per annum in the previous decade. This suggests that any economy led growth in trip length was being off-set by some alternative factor, which most likely was the growth in road congestion in and around urban centres.

The UK relationship of trip length growth to GDP growth again is not strong. Figure 8 indicates that there was major growth in trip length throughout the decade 1985 to 1995, though the economic boom had stopped by 1989 and was followed by a strong recession through to 1994. Overall the average annual growth in GDP per capita from 1986 to 1996 was similar to that from 1996 to 2007, though the growth in trip length is much slower in the latter decade. This suggests that whatever impacts on average trip lengths may arise from economic growth, they are less important than other competing influences.

4 EMERGING TRENDS 4.1 Growth in home working A trend that will tend to reduce the growth in total commuter travel is the increase in the proportion of the population that is working at home, as illustrated in Figure 7 above. This recent growth is in part due to improved telecommunication technology which enables home-working in a variety of office-based service occupations to be efficient, so it is a trend that is likely to grow in importance through the future but only in a subset of industry sectors and occupations within them. Its effects will be to reduce the average daily trip rate per worker both through having a greater number who work normally at home as well as through those who reduce the number of weekdays in which they commute to the office by reserving a few weekdays for home working. The reduction in the demand for overall capacity on the road and rail networks by the latter may in practice not be as great as would be expected because:  home working is not evenly spread across the week so that in mid-week the commuter demand is significantly greater than on Fridays;  those needing to travel less frequently to their workplace may avail of the opportunity to relocate for lifestyle reasons to more distant residences in attractive areas, leading to fewer but longer commuter journeys.

The growth in home working is also a side-effect of the recession that has led to an increase in the proportion of those who are self-employed, often working part- time, but who would prefer to change back to full-time employment outside the home, should a job opportunity actually arise. It is less certain whether and how quickly this influence that reduces the level of commuter travel demand will abate through into the future.

This feature can be modelled through use of changes in work trip rates in future years for those industries and occupations where its impacts are likely to be significant.

4.2 Splitting of work trips One important finding from the comprehensive study by McGuckin & Srinivasan (2005) of US commuting travel trends up to 2001 is: “In the context of daily travel the work trip remains as important as it has ever been. Work trips continue to increase as workers are added to the population, but the proportion and number of trips directly from home-to- work continue to decline. Direct, or non-stop trips to work are the minority of work travel - over half of all workers make non-work trips as part of their travel to work. … Therefore, local analysis based on an estimation of home-based-work trips as the only source of travel data may contain bias, since women stop during the work trip more often than men, and trends show that growth in stops made during the work tour varies by race, sex, and ethnicity.” (Page 42)

It would be of interest to examine NTS data to ascertain if a similar pattern of behaviour has emerged in the UK. In Ireland, to mirror the US “Starbucks”

commuter, there is anecdotal evidence of the growth in “breakfast roll” culture, where the ever earlier journey starts in the morning that are needed by long- distance road commuters to arrive close to Dublin city centre before the roads congeal, has led to an increase in the number of road commuters who breakfast in transit to their workplace.

The growth in the proportion of children who are delivered and collected by car to/from school is another cause of the commuter trip splitting trend. Again previous direct journey to work trips are being split into two components: escort to school; and non-home based to work, neither of which is classified in travel surveys as a home-based work trip. The policy change in England towards increased competition between schools, rather than fixed local catchment areas, will lead to longer school trip distances which in turn will tend to increase the demand for escort related diversions from direct journeys to work. These trends help explain the reduction in trip rates per capita for home-based journey to work that can be observed in the GB NTS in recent years, though the long-term reduction in the proportion of the population that is in full-time employment is another major influence on this reduction.

The best approach to modelling this increased splitting of work trips is to make effective use of a tour based representation of travel, and not just adopt a traditional independent trip based representation.

5 WHY IS IT NECESSARY TO SEGMENT TRAVEL DEMAND MODELS?

When developing forecasting models for use in land-use and transport policy testing, it is crucial that these models are designed in a form that maximises their suitability for the task in hand. This implies that the models need to maintain a clear separation between: a) changes in travel demand that are a direct response to the policy stimulus being tested e.g. those responses resulting directly from lower fares or faster travel speeds; and b) those changes in travel demand that are unconnected to the transport system, which result instead indirectly from exogenous demographic, social or economic changes to the population of travellers.

It is sometimes argued that: yes, it may be theoretically beneficial to segment the travelling population in detail in our travel demand forecasting models; but that unfortunately actually forecasting the relative future growth rate of each such segment accurately is so difficult in practice as to negate its theoretical benefits. This argument is not the full story for two reasons.

Firstly, the use of a model that has an appropriate level of traveller segmentation facilitates the application of realistic tests of the sensitivity of its forecasts to exogenous influences. These could be carried out through varying the respective growth rates of the relevant components of the population to a degree that is similar to the relativities observed in past times. This would clarify whether the ranking of policy measures is reasonably robust to uncertainties in exogenous growth rates of individual demand segments or whether the benefits from certain

policies depend critically on our ability to accurately forecast future demographic and socio-economic trends. This increased understanding of the degree of robustness of individual policy measures to future uncertainty would provide valuable insights for use in policy assessment.

Secondly, it is crucial when calibrating the model that the complete set of important behavioural influences for each model stage are included efficiently within its parameter estimation process. If there are changes either over time or through space in the balance within the population between individual segments that each exhibit distinct travel behaviour; and if this rebalancing is not represented explicitly in the segmentation within the model, then there is a real danger that the estimated elasticity of response to travel supply characteristics will be biased. For instance, in the example discussed in Section 3.3 of the changes over time in Irish trip lengths, a trip distribution model that does not take account of the differences in behaviour between males and females and in their relative balance over time within the labour force, will underestimate the rate of increase in distance travelled and so is likely to underestimate the elasticity of transport supply influences. An alternative distribution model example, in which the variability in trip lengths between industry types (as illustrated in Figure 1) is not made explicit through segmenting by industry group, would lead to a major underestimate in the elasticity to supply changes, for reasons that were highlighted by Williams (2005).

This need for a discriminating parameter estimation procedure that can identify the required segmentation of travellers implies that it should use individual rather than grouped (zonal) data records. Any grouping/zoning of the data records will greatly diminish the ability of the estimation procedure to discriminate between potential influences, due to the increased multi-collinearity between explanatory variables that the grouping will create. Even when using individual records, the high level of spatial correlation between variables, such as car ownership, residential density and socio-economic status, means that traditional simple regression techniques may struggle to identify casual structures unambiguously. More recently developed estimation techniques such as Structural Equation Models (SEM) may prove to be more effective (Jahanshahi et al., 2012) in determining the underlying causal structures.

5.1 Current UK Practice The majority of current travel demand models used in transport planning practice in the UK have relatively limited segmentation (other than car ownership) in their mode and distribution choice modelling stages. Exceptions do exist, such as the PRISM model of the West Midlands and the MEPLAN based models of the East Midlands (PTOLEMY), Cambridgeshire (CSRM) and the wider South East (LASER), which take account of a wider range of characteristics of the components of the labour force when estimating commuting travel patterns.

Within the current UK DfT WebTAG guidance, Section 1.2 of Unit 3.1.1 (DfT, 2009) draws attention to the combinatorial explosion that may arise when the number of distinct segments being represented increases. This issue is more problematic within aggregate, rather than disaggregate, model formulations.

Traveller type segmentation requirements are discussed in more detail in the following TAG Units  3.10.2 - Section 1.7, in which minimum requirements for segmentation are discussed, noting that different segmentations of traveller types may be suited to the different demand model stages and that no more segmentation detail is required than that necessary to meet the policy testing aims for which that model has been designed;  3.11.1 – Section 6, in which the requirements for public transport demand models are covered; and  3.15.2 – Section 3.3, which outlines the segmentation by person type that is maintained in the National Trip End Model (NTEM) used to underpin the TEMPRO data provision, stating that “The traveller type disaggregation used internally within the model includes the following, although the outputs are not available at this level of detail: • age (under 16, 16-64, 65+) • gender • employment status (for 16-64 age group) • household car availability • household size (number of adults)”.

The documentation of the current version 6.2 of NTM (WSP, 2011) explains that it segments workplace jobs into 12 aggregate SIC categories. This aggregation relates to the trip attraction variables required by the trip end model rather than being that most suited to commuter modelling. It is available in a form that is further segmented by gender and full-/part time but only in units of jobs in workplace zones. The segmentation of the working population at the residence zone that is used in NTEM is that in the previous paragraph and so it does not include segmentation by industry type.

6 CONCLUSIONS

There are a wide range of behavioural factors that influence trends in commuting patterns including: 1. Trends in car ownership, car operating costs and public transport fares; 2. The effect of travel speed (either by car or public transport) and of congestion on commuter travel distance; 3. Planning policy to encourage higher density urban residential development; 4. Planning policy on central V out of town office and retail employment; 5. Change in industrial structure – the growth in specialised business and professional service jobs and the decline in manufacturing and other manual jobs; 6. Change in the balance of males to females in the workforce; 7. Change in the balance of full-time to part-time; 8. Local imbalances for individual segments of the labour force between the labour supply and demand; 9. The potential impact of growth in tele-working; 10. Trends in multi-stage commuting trips (e.g. initial leg being escort to school).

However, few trip distribution and mode choice models in current UK practice take explicit account of factors 5 to 10 above, while the representation of factors 2 to 4 is often opaque:  The elasticity of the responsiveness of employment location to congestion and how it varies between industry sectors is not well understood;  The sectors of employment that can be captured in city centres have particular labour requirements that need to be explicitly represented in the segmentation of commuter trip distribution models both at the workplace and residential ends.

Increasing the level of segmentation of workers within the trip distribution and mode choice stages for commuting can provide a more realistic representation of most of these behavioural factors. This would lead to more realistic demand elasticities, better founded model hierarchies and more reliable long term forecasts and sensitivity testing. These should more than compensate for the extra computational burden that is created, though this burden should not be overstated since it is generally the assignment rather than other demand model stages that eats up resources.

The segmentation recommendations for individual stages of a commuter demand model are summarised in Table 3, which distinguishes their relative priorities for inclusion in order to ensure that the overall computational burden does not become excessive.

Table 3 Commuter demand model segmentation - recommendations Dimension Stage No. of segments in category Priority Distribution Gender [2] Male/female H Age [4] <25, <35, <50, <65 L Hours [3] >30, >=20, <20 {OR full- / part-time} H Occupation/Industry/SEG As many as possible relevant combinations H Income Less relevant than O/SIC/SEG L Car availability [3] none, <1 per adult in HH, 1+ per adult M

Mode choice Gender [2] Male/female M Hours [3] >30, >=20, <20 {OR full- / part-time} L Car availability [3] none, <1 per adult in HH, 1+ per adult H Income or O/SIC/SEG [<5] M

The priority of a particular dimension will partly depend on the policy use to which the model will be applied, so that in mode choice the use of segmentation by income will increase in importance for models used to test road pricing schemes.

At the trip distribution stage the adoption of a fairly detailed segmentation by occupation, SEG or industry is likely to be more helpful than segmentation by income, since the former is more discriminating in representing local imbalances between labour supply and demand, which are the key determinants of trip length

differentials. They will also provide a reasonable surrogate for income effects when testing pricing policies.

Figure 1 has demonstrated that average trip lengths are significantly different between segments distinguished by sex and hours worked, while Figure 6 has demonstrated that preferences of choice of mode differ significantly between male and female commuters. Figure 3 has demonstrated that the balance of males and females who work full-and part-time within the UK economy has not been stable through the years. Combining these three findings indicates that models that ignore these individual effects will have serious limitations in the reliability of their forecasts and of their ability to test various types of policy measures.

The potential ability to increase the level of detail in traveller segmentation through NTEM is critical to encouraging increased segmentation by gender and by full- / part-time employment within models of commuter trip distribution and mode choice. However, it does not appear that this increased segmentation detail can be accessed directly through TEMPRO at present. Accordingly, consideration should be given by DfT to the development work required to:  Facilitate access via TEMPRO to further segmentation by gender and employment status for trip-ends at both residence and workplace zones;  Introduce to NTEM and then through to TEMPRO a new further segmentation by one of SIC, occupation or SEG for the labour force at both residence and workplace zones. The underlying data required to introduce this segmentation will become available in the near future through the 2011 Census results.

7 BIBLIOGRAPHY

Dargay J.M. and M. Hanly (2003) A Panel Exploration of Travel to Work: an investigation based on the British Household Panel Survey. Presented at BHPS Conference Colchester, Essex.

Devereux L.S. and I.N. Williams (2010) Evaluation of the Urban Congestion Programme. Final Report and Appendices by WSP for UK Department for Transport. http://www.dft.gov.uk/publications/urban-congestion-programme- evaluation/ Access date: 2/6/2012.

Department for Transport (2009) TAG Unit 3.1.2C: Introduction to Modelling - for consultation. http://www.dft.gov.uk/webtag/documents/expert/unit3.1.1c.php. Access date: 3/10/2012.

Department for Transport (2011) Personal Travel Factsheet: Commuting and Business Travel. http://www.dft.gov.uk/statistics/series/national-travel-survey/. Access date: 3/10/2012.

Jahanshahi K., Y. Jin and I.N. Williams (2012). Analysing sources of variability in average travel distance: an investigation based on extended structural equation models and the UK National Travel Survey data. Proceedings of the European Transport Conference 2012, , UK.

Lyons G. and K. Chatterjee (2008) A human perspective on the daily commute: costs, benefits and trade-offs. Transport Reviews, 28(2), 181-198.

McGuckin, N. and N. Srinivasan (2003) Journey to Work Trends in the United States and its Major Metropolitan Areas 1960–2000. Publication No. FHWA -EP- 03-058. Washington, DC: U.S. Department of Transportation, Federal Highway Administration, Office of Planning. ftp://ftp.abag.ca.gov/pub/mtc/census2000/JTW_Trends/PDF/FullReport.pdf Access date: 2/6/2012.

Santos A., N. McGuckin, H.Y. Nakamoto, D. Gray and S. Liss. (2011) Summary of travel trends: 2009 National Household travel Survey. Publication No. FHWA - Pl-11-022. Washington, DC: U.S. Department of Transportation, Federal Highway Administration, Office of Planning. ftp://ftp.abag.ca.gov/pub/mtc/census2000/JTW_Trends/PDF/FullReport.pdf Access date: 2/6/2012.

Texas Transportation Institute (2011) Congestion Data for Your City. Excel spreadsheet, available at http://mobility.tamu.edu/ums/congestion-data/ Access date: 4/10/2012.

Williams I.N. (2005) Travel Demand - The Influence on Commuter Distances of Labour Supply / Demand Imbalances. Proceedings of the European Transport Conference 2005, Strasburg France. http://etcproceedings.org/

WSP (2011) NTEM Planning Data Version 6.2: Guidance Note. Report for DfT. https://www.dft.gov.uk/tempro/downloads.php#temprosystemfiles Access date: 4/10/2012.