social disadvantage and transport in the uk: a trip …social disadvantage and transport in the uk:...

Social disadvantage and transport in the UK:

a trip-based approach

José Moore

Dept of Civil Engineering, Universidad de Concepción.

Dr Karen Lucas and John Bates

Transport Studies Unit, University of Oxford.

Working Paper N° 1063

April 2013

Transport Studies Unit

School of Geography and the Environment

http://www.tsu.ox.ac.uk/

Moore, J. “Social disadvantage and transport in the UK: a trip-based approach”

Working Papers Series - Transport Studies Unit, University of Oxford 2

Social disadvantage and transport in the UK: a trip-based approach

José Moore, Dept. of Civil Engineering, Universidad de Concepción

with

Dr Karen Lucas and John Bates, Transport Studies Unit, University of Oxford

April 2013



Acknowledgements

The main study for this research is funded by the UK Economic and Social Research Council Mid-

Career Fellowship Programme for Dr Karen Lucas (project reference ES/J00023X/1). The National

Travel Survey data for the project is provided through an agreement with the National Centre for

Research (NatCen) and is for academic use only. José Moore’s contribution to the research was

funded by the EU Marie Curie International Researcher Exchange Scheme (IRSES) as part of the

project Transport and Social Exclusion: New Directions and National Comparisons project

(TranSENDaNC) and forms part of the research for his Master’s dissertation in Transport Engineering

that is being undertaken under the supervision of Dr Juan Antonio Carrasco.



ABSTRACT

This TSU Working Paper is the product of a three-month internship visit to the Transport Studies Unitby José Moore, which is funded under the auspices of the EU Marie Curie International ResearcherScientific Exchange Scheme (IRSES)

1between the University of Oxford, Universidad de Concepción,

Chile and the University in Ghent, Belgium. The research it presents is also contributing to an ESRCfunded mid-career fellowship programme to develop new models of the travel behaviour of low incomeand social disadvantaged population groups in the UK, which is being undertaken by Dr Karen Lucasunder the mentorship of John Bates

2.

The purpose of this paper is to firstly report on a literature review that was undertaken to identify theempirical studies that have been undertaken to model transport and social exclusion in differentnational contexts. Secondly, it reports on research to model the travel behaviours of sociallydisadvantaged members of the British population using the UK’s main travel survey, the NationalTravel Survey (waves 2002-2010). Two approaches have been developed within the paper: (1)introducing extra variables of social disadvantaged into the baseline UK National Trip End Model(NTEM); and (2) more refined purpose-based analysis, which considers four different purposes (i.e.work, social, visiting friends and relatives, and services).

The results from this analysis indicate that there are important differences in travel behavioursaccording to household income, possession of a driver’s licence and the presence of children within ahousehold, as well as by dummy indicators for vulnerable population segments (e.g. single parents,economically inactive and unemployed people, non-whites, etc.). Disaggregation by journey purposeidentifies that income has a particularly significant effect on visiting friends and relatives and social tripfrequency and travelled distance. Both findings could have important implications in terms ofmaintaining family bonds and generating social capital.

Vulnerable segments such as non-whites and people with mobility difficulties also have different travelpatterns than the baseline: although both are involved in less trip-making for all the purposesanalysed, non-whites travel further for VFR while people with disabilities tend to perform morelocalised VFR trips. Single parents do not show significant differences for trip-making (only a slightincrease) but their activity patterns proved to be more localised for both VFR and overall trips. Socio-economic variables also proved to have higher fit for mandatory activities which imply that otherfactors (e.g. contextual, psychological, etc.) are affecting social and visiting trips.

There are also important interactions between the three independent variables for travel behaviour(i.e. number of trips, trip length and trip duration) which will require further analysis as well as furtherindicators of transport and social disadvantage that have not been included within the modelsdeveloped for reasons of lack of data and appropriateness at the national spatial scale, which cannotallow for geographical factors.

Key words: travel behaviour, transport disadvantage, social exclusion, income effects, transport

modelling

1TranSCENDaNC: Transport and Social Exclusion: new Directions and National Comparisons

http://www.tsu.ox.ac.uk/research/TranSENDaNC/2

Modelling the Relationships between Transport Poverty and Social Disadvantagehttp://www.tsu.ox.ac.uk/research/mrtpsd/



1. INTRODUCTION

This TSU Working Paper is the product of a three-month internship visit to the Transport Studies Unit

by José Moore, which is funded under the auspices of the EU Marie Curie International Researcher

Scientific Exchange Scheme (IRSES) between the University of Oxford, Universidad de Concepción,

Chile and the University in Ghent, Belgium. The research it presents is also contributing to an ESRC

funded mid-career fellowship programme to develop new models of the travel behaviour of low income

and social disadvantaged population groups in the UK, which is being undertaken by Dr Karen Lucas

under the mentorship of John Bates.

Both projects have been developed in response to growing academic and policy recognition that

inequities in the provision and use of transport services are a major issue in the context of the

continued growth urban mobility and accompanying upgrades in different transport modes, particularly

private motor vehicles. However, studies of the links between travel behaviour and social

disadvantage are still in their infancy in most countries and the mathematical models of travel

behaviour that dominate all levels of transport decision-making rarely include social evaluation criteria

(Van de Voorde and Vanelslander, 2010). It is reasonable to suggest from an overview of the literature

that interdisciplinary socio-theoretical and mathematical explanations have yet to be developed,

although progress has been made in some respects in more recent years (e.g. socio-technical

transition models).

The objective of this paper is to models different trip-based indices of travel behaviour such as trip

generation, distance and duration considering variables in combination with indices of social

disadvantage. The overarching aim of the study is to develop model that can predict the effects of

different transport policies and scenarios on different sub-groups of the population, with a particular

focus on those groups that are already recognised as economically or socially vulnerable. This paper

is divided into seven sections. The next section provides the theoretical fundamentals and recent

empirical evidence related to the analysis of social disadvantage and travel behaviour focusing on

quantitative approaches. Section 3 explains the data and methods used in this study while sections 4,

5 and 6 presents our analytical results from the different models we applied: section 4 provides an

analysis of the NTEM and the inclusion of new variables related to social disadvantage and

socioeconomic and contextual characteristics and section 5 and 6 analyse how these results vary

when considering different trip purposes. Finally, section 7 provides our conclusions and next steps for

the study.

2. LITERATURE REVIEW

Social scientific studies of the transport and concerns of income deprived and socially disadvantaged

individuals, households and communities have been prolific over the past ten or more years. The

literature confirms that there is considerable interest in this topic from academics and policy makers in

the UK (e.g. Farrington, 2007; Preston 2009), as well as from across Europe (e.g. Ohnmacht et al,

2009) and more widely internationally (e.g. Currie et al, 2007; Páez, et al, 2009). Early UK studies

identified multiple relationships between transport disadvantage and income poverty (e.g. Church and

Frost, 1999; TRaC, 2000; Lucas et al, 2001). This has helped to stimulate the interest of policy makers

within central government, and in 2003 the UK Social Exclusion Unit (SEU) published its seminal

report on this subject. The SEU report led to new transport policy for local authorities in England to

deliver accessibility planning in order to reduce transport-related exclusion as part of the local planning

process (Department for Transport, 2006a). Problems such as the poor conceptualisation of social

issues within transport studies, the lack of adequate quantitative data for robust analysis and

commonly perceived difficulties associated with adopting a social equity approach to transport have

nevertheless persisted within UK transport planning authorities (Grieco, 2006). This suggests that the

need for new tools and approaches are needed which are capable of overcoming the current socio-

technical divides that prevails within the transport studies so that we can develop collective

understandings of the interactions between people’s travel behaviours and their economic and social



wellbeing. Our study aims to identify some of the associated conceptual, methodological and

analytical challenges of adopting this hybrid approach.

Theoretical framings

A number of publications have sought to enhance the conceptual understanding of the links between

transport and social disadvantage through the framings of social exclusion, which has dominated the

European policy discourse over the past ten years (e.g. Church and Frost, 2001; Social Exclusion Unit,

2003; Lucas, 2012). They identify that social exclusion is a complex and multi-dimensional process

and that transport disadvantage is only one part of the overall dynamics that create this phenomenon.

Levitas et al. (2007: 9) provide a useful definition of social exclusion in these terms:

“[social exclusion is] a complex and multidimensional process that involves the lack or denial

of resources, rights, goods and services, and the inability to participate in the normal

relationships and activities, available to the majority of people in a society, whether in

economic, social, cultural or political arenas. It affects both the quality of life of individuals and

the equity and cohesion of society as a whole”.

Their definition suggests that the primary concern is with individuals who do not have the same level

of access to goods and services and participation in activities enjoyed by the majority of people in the

societies in which they live. It is primarily this focus on access and participation which has attracted

transport researchers to this field of study, although clearly there are many other factors that may play

a role in determining people’s ability to fully participate in society, including income poverty, low

educational attainment, ill health, language barriers, racism, sexism and classism (Litman, 2003).

Considering these multiple criteria, different segments of the population can be defined as particularly

vulnerable or excluded in transport-related social exclusion (see Table 1).

Table 1 Vulnerable segments

Socially disadvantaged Transport disadvantaged Other

Low Income No car Women Unemployed Rural population Students Unskilled People w/disabilities Children Single parents Poor access Ethnic minorities Elderly

Other factors such as psychological characteristics, time availability, social capital and transport

provision can be crucial in defining whether a person is socially excluded or not. Building on Levitas et

al’s theoretical definition above, Lucas (2012: 107) has provided a conceptual framework for

understanding the contributing factors and interactions between transport and social disadvantage to

create a framework for transport poverty, inaccessibility and social exclusion (see Figure 1).



Figure 1: Relationship between Transport disadvantage, social disadvantage and social exclusion

(Lucas, 2012)

Recent empirical studies

Different efforts have been made internationally to analyse the travel behaviours of disadvantaged

segments of the population. European experience has been well developed during the latest years. In

the UK, transport related social exclusion has been analysed among the elderly (Titheridge et al, 2009;

Su and Bell, 2012), low-income groups (Lucas et al, 2009), rural population (Huby et al, 2007) and its

role regarding access to higher education (Kenyon, 2011). In the northern Irish context, activity spaces

had been used to identify transport disadvantage in rural population (Ahern and Hine, 2012;

Kamrruzzaman and Hine, 2010) and students (Kamrruzzaman and Hine, 2011). In the case of

Mainland Europe, GIS-based analysis of the labour market in Barcelona (Spain) by Cebollada (2009)

concluded that individuals without a car had less job prospects. In Germany, Scheiner (2010) used

structural equation modelling (SEM) to assess the links between trip distance and lifestyle variables

and the importance of access and quality of the transport supply. His findings indicated that trip

distances are most significantly affected by residential location.

In the Canadian context, access to health, food services and jobs had been analysed in Páez et al

(2009) based on vulnerable segments such as the elderly, low-income and single parents respectively.

Their analyses identified the presence of food desserts in the suburbs (i.e. zones with no food services

in a walkable distance) even with vehicle availability in lower-income households. In the case of health

services and the elderly, access was very low throughout the city with the exception of the City Centre,

with no significantly positive effects for vehicle ownership. In a later study, Páez and Farber (2010)

modelled the travel behaviours of people with disabilities using traditional econometric modelling

techniques such as discrete choice models and regression analysis. In the USA, there has been a

focus on car-pooling and car-sharing behaviours of immigrant populations (e.g. Lovejoy and Handy,

2010; Blumenberg, 2008).



Currie et al (2011) offers a comprehensive overview of transport-related social exclusion research in

the Australian context considering methodological frameworks, empirical analysis and international

comparisons. Some of the research included considers equity in public transport (Delbosc and Currie,

2011a; Currie, 2010a), transport and wellbeing (Currie et al, 2009, 2010b), the psychological impacts

of transport poverty (Currie and Delbosc, 2010) and the relationship between transport and social

capital (Currie and Stanley, 2008). The link between social exclusion and transport had also been

analysed in New Zealand, by Rose (2009) using qualitative approaches and Mavoa et al (2012) using

a GIS-based approach to analyse public transport accessibility.

Less research has been conducted in the context of developing countries. In the Latin-American case,

most of the experience is concentrated in Colombia where transport provision has been the most

analysed variable (Cahill Delmelle and Casas, 2012; Jaramillo et al, 2012). In Chile, research has

been undertaken to assess the role of public transport in social disadvantage of individuals (Ureta,

2008) and for the development of indicators of social exclusion at the transit zone level (Jara, 2009). In

the African context, research has been undertaken in South Africa (Lucas, 2011), Kenya (Salon and

Gulyani, 2010) and Nigeria (Ipingbemi, 2010). While in Asia the main studies have been on university

students in Thailand (Limanond et al, 2011) and women in Japan (Zhang et al, 2012; Chikaraishi et al,

2012).

Methodological approaches and quantitative analytical techniques

Figure 2 provides a basic overview of the different methodological approaches to transport and social

disadvantaged research that have been identified through our literature review. It is noted that much of

the early research tended to be qualitative offering case studies, ethnographies and other descriptions

of the problems experienced by different affected population groups and/or areas, as well as

considering and evaluating potential solutions. As these studies have already been reviewed and

synthesised elsewhere (see Lucas, 2012 for an overview), the focus of this paper is on describing the

development of quantitative approaches that have been more prevalent within the literatures in recent

years.

Our review has identified that a significant proportion of the remaining methodological experience is

GIS-based, with a focus either on analysis of accessibility the transport system and/or key economic

and social activities and/or facilities. The rapid development of computational devices and specialized

software in recent years has popularised the use of GIS-based methodologies across transport

studies. The most common way on which GIS-based analysis is used correspond to accessibility

measures which can be helpful when analysing spatial mismatch or the spatial distribution of access

to certain facilities or services (Casas et al, 2009). GIS tools such as AMELIA (e.g. Mackett et al, 2010)

and LUPTAI (Pitot et al, 2006) have also been developed as micro-level decision aids for

policymakers to evaluate the effect of different transport policy options and the use of cluster analysis

to differentiate the transport disadvantaged within integrated models (Duvarci and Yigitcanlar, 2012).



Figure 2 Methods and techniques for social exclusion analysis

Advanced statistical analyses, econometric and other mathematical modelling approaches have been

less prevalent within the literatures, although they have gathered momentum in recent years and a few

studies have also sought to bring together qualitative, spatial and statistical techniques. Here again

there can be a focus on either modelling supply or demand. Some studies have developed hybrid

methods that use may use qualitative and/or GIS as a first step to derive input variables for further

statistical models. The use of Gini Coefficients or Indexes, to indicate how equitable the provision of

transport services and other infrastructures are among the population is an increasing common

approach in this respect (e.g. Delbosc and Currie, 2011; Neutens et al, 2012; Lopez-Suarez, 2008).

It is also clear from these past studies that many different measures of travel behaviour can be used to

quantify the interactions between transport use and social disadvantage. Past studies can be broadly

divided according their unit of analysis and whether they take a transport demand or supply-side focus

(see Figure3).Demand-based studies have considered measures such as trip generation (Roorda et al,

2010), destinations (Scott and He, 2012), mode choice (Mercado et al, 2012; Schmöcker et al, 2008)

and distance travelled (Morency et al, 2011; Mercado and Páez, 2009). Activity-based studies have

included include investigation of trip purpose (Páez et al, 2009: Johnson et al, 2011), participation and

propensity to perform activities (Páez and Farber, 2012), activity duration or time-use (Limanond et al,

2011; Farber et al, 2011, Spinney, 2009) and activity spaces (Schönfelder and Axhausen, 2003). As

our own study focus takes on a statistical modelling, we now describe some of these methodologies

and findings from their application in more detail focusing in particular on their findings in relation to

trip generation, trip purpose and trip duration among socially disadvantaged populations.



Figure 3 Quantitative approaches to social exclusion

Trip generation

Trip generation is usually considered to be the first step of the traditional transport modelling (i.e. 4-

stage model). By implication it equates accessibility with the number trips an individual or household

makes during a certain period of time. This has been used to measure the transport disadvantage of

certain population groups in some past studies. Different results have been obtained within different

contexts and techniques, which have included ordinal probit models (Schmöcker et al, 2005; Roorda

et al, 2008) and negative binomial regressions (Van den Berg et al, 2011).

Schmöcker et al. (2005) analysed shopping trip generation of elderly and disabled people in London,

UK. Their results demonstrated that increased age and walking difficulties resulted in fewer trips; and

variables such as household structure, ethnicity, income, car access and taxi or public transport cards

also had an effect on trip generation. In Canada, Roorda et al (2008) modelled the trip generation of

certain vulnerable groups (low income, elderly, and single parents). Some of their findings were that

the effect of age is cancelled by auto ownership (contrasting with the elderly having the lowest trip

making frequency), and that only minor differences were found for single parent households. Positive

correlations with trip generation included owning a driving licence, being employed, being a student,

having free parking at work, and living in a single-person household. Van den Berg et al (2011)

examined social trips of the elderly in the Netherlands using 2-days travel diaries. Results indicated

that high education and involvement in clubs resulted in higher social trip rates while full time work had

the opposite effect.

Trip distance

Trip distance can be use as measure of an individual’s or household’s overall mobility as well as an

implicit measure of their accessibility to services and thus activity-spaces. Log-linear regressions have

usually been used to analyse this variable (Schmöcker et al., 2005; Farber and Paez, 2010; Morency

et al., 2011; Van den Berg et al., 2011; Maoh and Tang, 2012).However, alternative techniques such

as multilevel analysis (Mercado and Páez, 2009; Loo and Lam, 2012) and structural equation

modelling (SEM) (Scheiner, 2010; Manaugh et al., 2010) have also been used to model trip distances

at the individual level.

When analysing travel distances of senior citizens in London (UK) in the study mentioned above,

Schmöcker et al. (2005) demonstrated that trip distances decreased with age even though

recreational trip distances increased until the age of 80 years. Their findings also included positive

correlations with income, driver`s license and car ownership, whereas urban density, walking

difficulties and trip frequency had negative influence on trip distances.



In the Canadian context, Farber and Páez (2010) have analysed the commute distances of people

with disabilities. Their results demonstrated that requiring attendants to use transportation tools and

the use of “active transport” (i.e. non-motorized modes) implies shorter commuting distances.

Availability of public transportation seemed to not be significant for commuting distance. Morency et al.

(2011) demonstrated that significant spatial variations between different vulnerable groups (low

income, elderly and single parents). Single parents and the elderly had the smallest activity spaces

(trip distances); vehicle ownership effects were non-significant for seniors and transit effects were

ambiguous, small and localized. Maoh and Tang (2012) explained the travelled distances of normal

and extreme commuters in Windsor, Canada. Their findings indicated that socioeconomic and land

use was significant in both normal and extreme commuter distance. Variables such as gender, age,

occupation type, transportation mode, migration status, employment status, mixed land uses and job

concentration near the individual’s residence explained commute distance. However, socioeconomic

factors were significant in normal commute distance while land use factors were more important for

explaining extreme commuting distance. In the case of multilevel modelling, Mercado and Páez (2009)

studied the effect of ageing for different transport modes. Their results indicated that distance as a car

driver decreased with age and variables such as gender, working status, household size, income, land

use, license and car ownership had significant effects on trip distance. Also in Canada, Manaugh et al

(2010) analysed commuting distances as a joint variable with home-work location with positive

correlations (for commute distance) with variables such as full time work, income, vehicle ownership

and being a male. Age and trip frequency had negative influence on trip distance.

In the European context, Van den Berg (2011) used a random effects model to analyse trip distance in

Eindhoven. Her results demonstrated that younger full time workers undertook longer trip distances in

urban as well as rural areas. Visits and joint activities were also related to longer trip distances and

though age has significant effects on travelled distance when controlling for other characteristics,

senior citizens (over 65 years) were found to be as mobile as the younger (working) people. Scheiner

(2010) analysed working, maintenance and leisure trip distances in Cologne, Germany using SEM and

introducing subjective elements including attitudes, lifestyles and location preferences. His results

indicated that neither lifestyles nor location preferences had a strong impact on trip distances with the

exception of leisure trips where lifestyle had the strongest impact of all variables studied.

Loo and Lam (2012) analysed activity spaces of working couples in Hong Kong. Their findings showed

that gender plays an important role in individual mobility. Women with children, low-income level and

living in the suburbs had lower activity spaces than their male counterparts.

Trip duration / travel time

Usually the study of travel times has been modelled using transport network approaches, as travel

times most often depend on network characteristics, transport mode and levels of network use (e.g.

congestion). In the UK, McQuaid and Chen (2012) used a binary multiple logistic regression model by

gender, age and working status to analyse their influence on commuting times and to assess factors

such as presence of children and working hours. Their obtained results indicated that worker’s age,

having children, youngest child`s age, occupation, weekly pay, and public transport ridership were

significant for trip lengths. However, their results suggested the need to disaggregate the effects of

gender, working hours and childcare responsibility when modelling or developing policy.

However, across these literatures the direction of the relationship between social disadvantage and

travel behaviour is not always clear, i.e. do more or longer trips indicate greater affordability and thus

more social advantage or do they indicate less accessibility and therefore greater social disadvantage.

In order to unpack the complexity of the relationship it is important to recognised the significant

interactions between the three output variables, as well as factors such as mode choice, journey

purpose and destination location.



3. DATA AND METHODS

Having explored the literature pertaining to the focus of our study, we now discuss in more detail the

dataset that we have used to develop our modelled analyses and the methods we employed, before

presenting our initial results and finally describing our plans for further research.

Data

The data for our study is the National Travel Survey (NTS) years 2002-2010. This is a household

survey that has run continuously since 1988 and collects information on how, why, when and where

people travel as well as factors which affect personal travel, such as car availability, driving licence

holding and access to key services. Approximately 20,000 individuals of different age, ethnicity and

economic background living in 8,000 households across the UK participate in the NTS each year (DfT,

2010). However, this is cross-sectional survey, which means that the respondents are unlikely to be

the same across different survey years.

Data is collected via face-to-face interviews and a 7-day travel diary to collect information on

households, individual members within the household and all the vehicles to which they have access.

Each household member is then asked to record details of all their trips over a seven day period in a

travel diary, allowing travel patterns to be linked with individual characteristics. Adult members of the

household are in charge of recording the 7-day travel diary in the case of younger children and others

unable to complete the travel diary on their own. This approach can be represented by a hierarchical

structure considering different levels as the variables get more disaggregate and represent with higher

detail the characteristics of the respondent’s travel behaviour.

Figure 4 Levels within the NTS Database (Source: National Travel Survey)

A number of weights are also considered by the NTS in order to improve the accuracy and

representativeness of the data and to adjust for non-response bias. This weighting strategy also

adjusts for the drop-off in the number of trips recorded by respondents during the course of the travel

week (DfT, 2010). The 7-day diary is a particularly important data source as the whole travel week is

recorded. This allows analysis of differences in travel behaviour during labour days and weekends or

holidays.

However, though the NTS is a very complete and detailed database there are still limitations that

prevent some socially relevant disaggregate analysis. For example, there is no information about skills

or education levels other than for Household Reference Person (HRPs), also, household income and

personal income although collected as a continuous variable is only released in income bands, which

may lead to model biases. Weekly car travel expenditure is also not collected and although it would be

possible to construct this variable from the vehicle and public transport fares data, it is a lengthy and

not wholly reliable process.

More fundamentally, no geographical referencing is released below the general regional level (e.g.

London and other Metropolitan regions), which means that it is not possible to model the effects of

local land use and service supply on travel behaviours, which can be important when considering the



issue of social disadvantage. Nevertheless, the NTS provides the best source of data on travel

behaviour in the UK at the present time outside of much smaller bespoke surveys.

Methodology

We used multiple linear regression (MLR) analysis for our study, which relies on describing linear (or

first order) relationships between the independent and dependent variables. Different assumptions are

considered for the formulation of multiple regression models, such as continuity, linearity,

independently and randomly sampled observations, homoscedastic observations, uncorrelated

independent variables and disturbances, and normally distributed disturbances (Washington et al,

2003). This family of models are usually estimated using either least square estimation (OLS) or

maximum likelihood estimation (MLE). As we are dealing with continuous variables, 90% confidence

interval OLS estimation is used. Independent variables considered in the modelling process

represented the different levels of the NTS and the original variables from the National Trip End Model

(NTEM), as already described above. Variables such as person and area type, urban context,

household and individual characteristics were considered (see Annex Table 1).

Dependent variables include trip generation (trips per week), distance (miles per trip) and duration

(minutes per trip). The first is related to travel demand and activity participation (when analysed by

purpose). Trip distance, on the other side, can be used as a measure of people’s activity spaces

(Schönfelder and Axhausen. 2003), indicating the average distance covered by an individual per trip.

A smaller activity space does not necessarily mean a lower level of activity however, because this will

also depend on the quantity and quality of the land uses within this travel area. Finally, trip duration

can be used as an indicator of quality of the travel experience service, i.e. considering the “burden

nature of travel” and the fact that most trips are performed as a derived demand rather than a valued

activity.

Vulnerable segments

Variables for gender, household income and employment status were included in the model to as for

the NTEM, as important for determining differences in people’s overall travel behaviours. A further

seven vulnerable population segments were also identified for inclusion within the model: i) single

parents (family structure), ii) non-whites (ethnicity), iii) elderly (age), iv) rural population (public

transport access), v) persons in households with unskilled HRPs (skills and education), vi)

economically inactive and vii) unemployed (employment). Most of these categories are not

independent and it is possible for an individual to be represented in more than one segment. The

economically inactive are independent from the unemployed, however, because the vi) non-

economically active category includes home-workers and people unable to work due to health issues

while the vii) unemployed category only includes 18-64 year old that are registered unemployed and

actively seeking work.

As seen in table 2, the largest vulnerable groups are the elderly and rural population. This shows that

both age and location-related disadvantages are the most common among the population of the UK

but not necessarily the most transport disadvantaged. Income distribution shows that both single

parents and the elderly are the population segments with the largest population in the lowest income

band and non-whites and people living in rural areas are the ones with largest representation at

highest income, although this is below the national average. However, the effect of family size is not

considered which means that though household income might be high, income per capita could be

small. If we analyse employment status, single parents, non-whites and unskilled HRPs are the ones

with the most unemployed individuals. While for gender distribution even though most segments have

a fair share between males and females, there is an absolute majority for single parents being women.

In the case of car ownership, it is interesting to see how near half of single parents have no cars while

the economically inactive, elderly and non-whites have low car rates when compared to the whole

sample. However, most single parents do have a driving licence and probably higher access to cars

not belonging to their while most non-whites do not.



Table 2 Socio-economic characteristics of vulnerable segments

Singleparents

Non-White

Elderly RuralUn-

skilledHRP

Un-employed

Econ.inactive

WholeSample

Age (years)

Mean 35.2 28.6 74.3 42.1 43.9 33.1 40.2 39.3

Age distribution (%)

0-16 years 0 31.2 0 19.0 0 4.1 0 20.7

17-64 years 99.6 62.3 0 59.2 95.8 95.3 89.0 60.0

65+ years 0.4 6.5 100 21.8 4.2 0.6 11.0 19.3

Household income level (in £ ,000)

Less than 25K 88.6 52.1 79.5 39.0 60.6 70.7 68.5 46.0

25-50K 10.1 29.5 16.0 35.2 31.1 20.6 22.1 33.3

50K or more 1.3 18.4 4.5 25.8 8.3 8.7 9.4 20.7

Employment Status (%)

Full time 25.5 42.8 2.4 44.4 68.0 0 0 44.0

Part time 25.5 12.7 4.6 16.1 24.4 0 0 14.5

Student 3.4 11.6 0 2.7 0 0 0 4.0

Econ. inactive 41.0 25.2 3.4 8.6 7.6 100.0 100.0 10.4

Retired 4.8 7.7 89.5 28.3 0 0 0 27.1

Car ownership

No car (%) 47.6 28.6 33.4 6.4 28.5 38.9 32.0 18.5

No licence (%) 25.1 52.3 41.2 29.4 20.4 34.7 35.5 38.6

Cars per HH 0.54 0.95 0.77 1.53 1.01 1.05 0.94 1.13

Gender (%)

Female 93.0 51.6 54.9 50.7 39.0 39.9 75.0 51.7

N

% of sample 2.0 9.7 16.6 14.9 4.2 1.9 8.2 100.0

No. of cases 3,992 18,925 32,341 29,095 8,205 3,726 16,083 195,018(Source: National Travel Survey)

The model demonstrates that the average British person undertakes 16 trips per week, travels 8 miles

per trip and takes 25 minutes per trip to arrive at their destination. Most of these trips are made by car

(mainly as a driver) and their trips are usually to work, shopping facilities or for social purposes. The

results from our analysis demonstrate significant variations in travel behaviours of the seven

vulnerable population groups we identified.

As table 3 identifies, single parents make the most trips, an important part of these being for escort

purposes, by car and within a relatively local area, as shown in Table 3. Non-whites make the least

number and shortest trips in terms of journey distance but the longest trips in terms of trip duration.

This is a modal effect as nearly a third of their trips are either by public transport (bus) or non-

motorized modes. Non-motorized modes are mostly used by the unemployed (and economically

inactive) and single parents; walking being the main mode in both cases. In the case of cycling,

unskilled HRPs and the unemployed have the biggest shares, doubling the sample mean. In the case

of weekly trip frequency, single parents and rural population are the ones making the most trips.

However, there is an income bias in the case of rural population caused by the fact that they are

relatively the richest vulnerable segment and only 6.4% have no car. On the other hand, for trip

distance and duration single parents, unskilled HRPs and the unemployed are the ones with the

shorter trip distances and hence have the shortest trip durations.

Table 3 Travel behaviour measures of vulnerable segments

Singleparents

Non-White

Elderly RuralUn-

skilledHRP

Un-employed

Econ.Inactive

WholeSample

Trip Frequency (trips per week)

Mean 19.0 13.7 13.3 17.2 17.9 14.8 15.4 16.5

Trip Distance (miles per trip)

Mean(Median)

5.7(2.5)

6.6(4.3)

7.0(3.0)

10.8(8.6)

7.2(3.0)

6.5(3.0)

6.2(4.5)

8.1(3.0)



Travel time (minutes per trip)

Mean 23.3 28.8 24.8 25.4 23.8 27.4 24.7 25.1

Modal Split (% of trips)

Walk 15.8 13.1 11.0 7.1 11.4 17.9 20.9 11.3

Cycle 1.0 1.0 1.0 1.4 3.0 1.4 3.0 1.7

Car Driver 58.3 33.0 48.4 57.1 60.1 44.1 39.1 47.9

Car Passenger 10.0 26.9 24.0 27.6 12.0 22.3 17.2 26.6

Bus 10.6 17.4 11.2 3.2 8.5 10.2 14.1 7.3

Rail 1.0 3.0 0.8 1.0 1.2 1.0 1.8 1.7

Taxi 2.3 1.2 1.5 0.6 1.4 1.7 1.7 1.2

Other 0.8 4.3 2.0 2.1 2.5 1.3 2.3 2.4

Trip Purpose (% of trips)

HB Work 10.7 18.5 2.4 14.7 32.8 4.1 1.8 16.3

HB Business 1.2 1.8 0.7 2.9 3.3 0.9 0.4 2.2

HB Education 0.7 11.4 0.0 5.0 0.4 2.1 0.8 5.4

HB Shopping 16.5 13.3 31.6 15.3 15.3 20.2 23.6 16.0

HB PB 5.7 9.1 16.2 8.4 5.6 12.3 10.2 7.9

HB Social 9.6 10.2 20.8 16.6 10.6 17.4 15.1 15.3

HB VFR 10.4 9.7 10.5 8.4 8.1 14.6 12.8 9.5

HB Holiday 0.3 0.3 0.5 0.6 0.4 0.4 0.5 0.5

HB Escort 26.0 14.5 6.1 11.8 11.3 16.0 21.8 12.2

NHB EB 1.4 1.1 0.3 2.3 2.2 0.3 0.2 1.7

NHB Other 17.6 10.1 11.0 14.1 9.9 11.7 12.9 13.0

Stages

Stages per trip 1.03 1.08 1.03 1.03 1.03 1.05 1.03 1.04

Sample

Total Trips 64,750 198,435 373,371 438,093 124,338 45,427 205,844 2,737,087Obs: The three most frequent modes and purposes are in bold. (PB: Personal business: VFR: Visiting friends and relatives; EB:

Employer’s Business). (Source: National Travel Survey)

Modal split analysis shows that although most vulnerable segments travel by car, a high percentage

use local buses, being particularly high in the case of the non-white population, which is doubles the

population average. In the case of trip purpose, single parents and the unemployed have a higher

percentage of home based (HB) escort journeys, which is as expected in the case of single parents as

the presence of children in the household generally leads to an increase in escort trips. On the other

hand, HB social trips are very low amongst both single parents and non-whites, which may suggest a

time poverty factor for single parents and language barriers for recent non-white immigrants.

HB social trips are higher than for the average population average for the elderly and rural dwellers. In

the case of working and education trips, it is interesting to see how non-whites and the unskilled HRPs

are above the mean in both categories. Another interesting fact is the case of shopping trips where the

elderly and the unemployed are way above the average being this one of the main trip purposes

considered by single parents. The problem is that it is not defined what kind of shopping this is (i.e.

groceries or non-primary products). Finally it is important to notice that most non-home-based trips

(NHB) are performed by single parents.

4. MODEL RESULTS

In this section we outline the main results from our three modelling approaches, namely i) the NTEM

baseline model, ii) the extended NTEM model and iii) the purpose-based analysis.

i) NTEM baseline model

The UK Department for Transport (DfT) has commissioned two major investigations in order to

forecast trip generation at the national level. The first one made use of the NTS data from 1988 to

1996 (WSP, 2000) while the second used data from 1995 to 2006 (WSP, 2009).Results from the first

study were included in the National Trip End Model. The NTEM models trip rates considering 8 home-

based (HB) and 7 non-home-based (NHB) trip purposes. It considers variables such as gender,

person type, household structure, car ownership and area type. However, not all the categories



generate statistically different trip rates for each purpose, although the general level of explanation is

high and consistent.

We recreated the NTEM model as a first step of our analysis in order to further analyse the effect of

additional variables on travel behaviour. The analysis of categorical data presents certain difficulties in

exposition. For reasons of ‘identifiability’ in estimation, it is not possible to obtain a coefficient for each

category. If there is only one categorical variable, then there are two possibilities for model

specification, either (i) to include a regression constant and, after defining a base for the

categorical variable, include all the other levels as dummies, or (ii) to drop the constant, and include

all the levels as dummies. In the first case, the coefficients are the increments to the base, and in the

second they are the absolute values for each level. When there is more than one category variable

then the choice between methods (i) and (ii) remains for one of the category variables (which can be

arbitrarily selected) but for each of the remaining variables we have to select a base level and drop the

associated dummy from the list of the regression variables. Our approach has been use method (ii)

and to select the person-type variables (represented by αperson-type in the Equation given earlier) as

the category for which the absolute values will be estimated for all levels. For all other categories, the

base level will be explicitly stated. It should be noted that these are arbitrary conventions, which do not

affect the results (model fit) but have some relevance for the interpretation. We next present the output

tables from the models and discuss key findings from our interpretation of them.

Table 4 Baseline NTEM model

Trip generation(trips per week)

Trip distance(miles per trip)

Trip duration(minutes per trip)

Variable Coeff. t-ratio Coeff. t-ratio Coeff. t-ratio

Area type

London -1.25 -16.4 0.63 5.5 7.46 42.2Metropolitan Ref. Ref. Ref.Urban big 0.18 2.3 0.76 6.4 0.88 4.8Urban large 0.17 2.0 0.59 4.8 -0.63 -3.3Urban medium 0.23 3.0 0.99 8.5 -0.74 -4.1Urban small -0.11 -1.5 2.29 20.1 0.98 5.6Rural -0.39 -4.8 4.03 33.6 2.94 15.9Person type

Child 12.07 147.7 5.21 42.4 24.09 127.2Full time 16.59 211.3 9.67 81.8 30.65 168.5Part time 18.07 179.7 6.74 44.5 26.70 114.6Student 13.57 99.5 8.22 40.0 32.96 104.3Non-working 14.10 159.4 6.54 49.1 26.64 130.0Retired 12.14 147.5 6.17 49.8 25.77 135.1Gender (female) 0.20 4.6 -1.43 -21.8 -1.87 -18.5N of adults per HH -1.39 -37.4 -0.90 -16.1 0.49 5.6N of cars per HH 2.47 92.6 1.42 35.4 -1.54 -25.0Model fit and ANOVA

Sum of squares

Residual 9,672,475 9,690,154 42,811,474Total 47,206,873 14,943,903 124,303,082DoF 15 15 15F-Ratio 43.034 6.013 21.109Adjusted R

20,80 0,35 0,66

No. of Obs. 142,807 142,807 142,807Obs: Dark cells indicate non-significance at the 90% confidence level.

The area constants show some variation, with London having the lowest frequency and the

highest average trip duration, and rural areas having the greatest average trip length. Overall, the

differences are perhaps less than one might expect, but of course they hide the modal

variation. More interesting are the person type effects; children, students, non-working and

retired persons make the least trips, while part time and full time workers make the most.



These absolute estimates by person type must be interpreted as applying to male persons in

one-adult households with no car in a Metropolitan area.

There is a small gender effect although this is not highly significant, with females making on average

0.2 trips per week more. Trip frequency per person reduces by 1.4 trips per week for each extra

adult in the household (because of the possibility of “sharing” travel, especially shopping) and

increases by about 2.5 trips for each additional car. The positive influence of certain variables

appear to be consistent with previous research in the case of distance travelled by full time

workers (Manaugh et al., 2010), females (Manaugh et al., 2010; Mercado and Paez, 2009). For

average trip length, the directions of the effects are similar except for part time workers and non-

workers who have trip lengths lower than those of full time workers and students. These

results carry over to the average trip duration model, with three exceptions: students spend

longer on travel despite having shorter trip distances compared to full time workers, and the

increasing number of cars within a household reduces the average duration despite longer

journey distances. These are both probably modal effects showing the higher speed for the car. The

third exception is that the travel time increases slightly with the number of adults despite the reduced

distance effect: the reasons for this are not immediately clear.

ii) Extended NTEM model

Additional variables were included in the model for households (i.e. income, presence of children),

individuals (e.g. licence owning, belonging to vulnerable segments) and spatial characteristics (e.g.

density, index of deprivation) in order to capture social effects not previously considered in the NTEM3.

As seen in Tables 5, the vast majority of the additional coefficients are highly significant. and there is

a slight improvement in the adjusted R2

values. In particular, it is clear that there are important

effects on travel behaviour indicators due to household income, the presence of children, and personal

attributes such as being non-white and having a mobility difficulty. The interactions between variables

are probably significant as well (e.g. licence holding and number of cars or being a single parent,

non-white and in low income groups) but as yet these have not been fully explored. There is also a

raised significance in the gender effect for trip frequency with these additional variables.

Table 5 NTEM extended model

Trip generation(trips per week)

Trip distance(miles per trip)

Trip duration(minutes per trip)

Variable Coeff. t-ratio Coeff. t-ratio Coeff. t-ratio

Are

aty

pe

London -1,41 -17,7 0,05 0,5 6,38 33,9

Metropolitan Ref. Ref. Ref.

Urban big -0,17 -2,1 0,52 4,3 0,94 4,9

Urban large -0,15 -1,9 0,41 3,3 -0,37 -1,9

Urban medium -0,31 -3,7 0,50 4,0 -0,24 -1,2

Urban small -0,93 -10,1 1,67 12,1 1,99 9,1

Rural -1,11 -10,3 2,56 15,9 3,34 13,1

Pe

rso

nty

pe

Child 9,94 27,5 8,06 14,9 28,88 33,9

Full time 11,66 32,3 10,68 19,8 35,16 41,3

Part time 13,26 36,2 8,44 15,4 32,07 37,1

Student 10,61 28,4 10,30 18,5 37,56 42,6

Non-working 10,76 29,8 8,68 16,1 31,96 37,5

3This meant that the overall sample that was available for analysis dropped by 14%, to 142,807

individuals, primarily because of the inclusion of the household income variable, which many peoplerefuse to complete within the NTS. These changes in sample size meant there was a related changethe total sum of squares, which makes direct comparison of the NTEM and extended NTEMmodels problematic. In addition, the interpretation of the area type coefficients changes,because the added variables affect the implicit base. In addition to the previous definition, theynow reflect: household income of £1000 p.a., no children in the household, no driving licence,white, no mobility difficulties, not single parent, density of 1 person/acre, and lowest level of Index ofDeprivation (most deprived).



Retired 9,86 27,3 7,93 14,7 30,46 35,7N

TE

MGender (female) 0,49 10,7 -1,43 -20,7 -2,12 -19,5

N of adults per HH -1,06 -25,0 -1,16 -18,3 -0,13 -1,3

N of cars per HH 1,36 42,4 0,69 14,5 -1,76 -23,3

Ho

use

ho

ldin

co

me

per

ann

um

Less than £1,000 Ref. Ref. Ref.

£1,000- £1,999 1,09 2,4 -0,22 -0,3 -0,49 -0,5

£2,000- £2,999 0,41 1,0 -0,70 -1,1 -1,28 -1,3

£3,000- £3,999 -0,03 -0,1 -0,87 -1,4 -1,48 -1,5

£4,000- £4,999 -0,04 -0,1 -0,98 -1,7 -1,99 -2,2

£5,000- £5,999 0,05 0,1 -1,37 -2,5 -2,85 -3,3

£6,000- £6,999 -0,19 -0,5 -0,99 -1,8 -2,25 -2,7

£7,000- £7,999 0,07 0,2 -1,06 -1,9 -2,76 -3,2

£8,000- £8,999 0,66 1,8 -0,92 -1,7 -2,30 -2,7

£9,000- £9,999 0,36 1,0 -0,50 -0,9 -2,45 -2,9

£10,000- £12,499 0,65 1,9 -0,95 -1,8 -3,25 -4,0

£12,500- £14,999 1,16 3,4 -0,61 -1,2 -2,93 -3,6

£15,000- £17,499 1,23 3,6 -0,64 -1,2 -3,05 -3,7

£17,500- £19,999 1,18 3,4 -0,27 -0,5 -2,81 -3,4

£20,000- £24,999 1,71 5,0 -0,13 -0,2 -2,55 -3,2

£25,000- £29,999 1,71 5,0 0,15 0,3 -2,47 -3,1

£30,000- £34,999 1,68 4,9 0,24 0,5 -2,21 -2,7

£35,000- £39,999 1,96 5,7 0,42 0,8 -2,00 -2,5

£40,000- £49,999 1,89 5,5 1,14 2,2 -0,94 -1,2

£50,000- £59,999 1,93 5,6 1,63 3,2 -0,13 -0,2

£60,000- £69,999 1,70 4,8 2,40 4,6 1,07 1,3

£70,000- £74,999 2,14 5,8 2,28 4,2 1,46 1,7

£75,000 or more 1,62 4,7 3,41 6,6 2,77 3,4

Exte

nde

dN

TE

M

Presence of children 2,08 36,0 -1,19 -13,8 -3,19 -23,4

Driving licence 4,51 65,8 0,85 8,3 -2,06 -12,8

Non-white -1,71 -22,1 0,30 2,6 2,23 12,2

Mobility difficulties -1,95 -24,0 -0,61 -5,0 -2,53 -13,1

Single parent 0,96 5,5 -0,88 -3,3 -0,61 -1,5

Log-density -0,08 -2,6 -0,32 -6,9 0,20 2,7

Index of Deprivation 0,13 14,9 0,11 8,8 -0,02 -1,1

Sum of squares

Residual 7.946.527 8.126.825 36.478.293

Total 41.470.027 12.915.028 109.442.258

DoF 44 44 44

F-Ratio 13.688 1.912 6.490

Adjusted R2 0,81 0,37 0,67

No. of Obs. 142.807 142.807 142.807Obs: Dark cells indicate non-significance at the 90% confidence level.

Area type

Results indicate that even though achieving its lowest level for those living in London, trip generation

decreases along with area-type. Living in a less populated area has a direct positive effect on mean

distance due to the more disperse origins and destinations, as expected. These results show some

variation when trip durations are analysed: although they have the lowest mean distance, those living

in London have the highest mean trip duration spending 6.4 additional minutes per trip than those

living in the baseline area (Metropolitan). We can assume that this is an effect of higher levels of

congestion and higher use of slower modes such as bus or walking. Also, additional time spent

travelling may have an effect on trip frequency as an outcome of constrained time budgets.

Variables such as Log-density and index of deprivation were included because we anticipated there

was likely to be an area-based effect based on density and deprivation. The index of deprivation

variable is based on a 1-10 scale with 1 denoting the most deprived and 10 the least deprived areas,

which would suggest it would represent a good indication of whether people living in these areas are

also transport disadvantaged. The results on area-type differ from those for log-density as denser

areas refer to less, more localised and longer trips (e.g. London).



However, trip frequency increased, and mean distance decreased as area-types grew bigger for other

area-types. This negative effect on mean distance is expected to occur because denser zones imply a

higher supply of goods and services within smaller areas which leads to lower average trip distances.

Index of Deprivation has a positive effect on all indicators. This effect is weak considering that the

difference from being the most deprived or the least deprived turns out to be 1 trip per week and 1

mile per trip. As neither variable appears to have had a significant effect on travel outcomes

we will probably drop them from our future models.

Household and individual characteristics

Income was treated both as a semi-continuous (defined as the base 10 logarithm of the mean of each

income band) and as a set of dummy variables (1 if belonging to a particular income band, 0 if not).

However, although the log-linear form for income has a better fit than a linear form, it does

not in fact capture the income effect on trip frequency very well, as demonstrated in the graph in

Figure 5In this case, some income bands lose their statistical significance due to their very low shares

(especially in the lower income bands).

Figure 5 Effect on income on trip frequency

The solid green line indicates the income effect on trip frequency for the fitted model

assuming a functional form relating to log10(income). In this form, it is implied that there will be

an equivalent rise in trip making for a given proportional increase in household income. It can be

seen that relative to the base (< £1000 p.a.), a person from a household with £25,000 p.a. is implied to

make 1.7 more trips per week, while a person with income > £75,000 makes 2.5 additional trips.

Compared with this, we have plotted the 95% confidence intervals (upper and lower bounds) for the

coefficient for each band (since the sample size for each band is relatively low). It can be seen that

there is a tendency for trip rates to fall as we move from the lowest income up to the £6000-7000 p.a.

band. However, it is reasonable to treat these lowest bands as anomalous, since they are well below

the poverty line, and are probably households who have little or no regular income, but nonetheless

have access to other sources of finance.

If we therefore ignore these lowest bands, then a very consistent pattern emerges. It is evident that

there is a steep and more or less linear rise in the number of trips between £6,000 per

annum (which equates with a single person household on welfare benefits) and £25,000

(which is the average household income level), with the increase over this range being about 1.9 trips,

suggesting approximately an additional 0.1 trips per additional £1000 p.a.. There appears to be



no significant increase in trip making after this. In the diagram we have represented this by a

piece-wise linear dashed function. It is clear that the shape is rather different from the

logarithmic form, which is probably dominated by the pattern in the range £6,000-25,000 and thereby

gives a misleading impression of how the trip rate increases for the higher incomes. This

suggests that another model is needed to more adequately capture the effect of income on trip-

making, which is an important finding for guiding our future research.

Presence of children and licence owning were also defined by dummy variables indicating whether the

household had 1 or more children and if the individual had a valid driving licence. In the case of

presence of children within the household, the variable is highly significant for all variables indicating

that those living in households with children were involved in more (extra 2.1 trips per week) but

shorter (1.2 miles and 3.2 minutes less than living with no children) trips during the sampled week.

This suggests a more localised travel pattern for those having children within the household and an

increase in travel demand due to the fact of children being escorted to most of their out-of-home

activities. There is some interaction with the number of cars variable which leads to an increase of 1.4

trips per week, while owning a driving licence implies additional 4.5 trips per week indicating that a

licence owner with 1 car will be making nearly 6 more trips on average in a week than one who has

neither car nor licence. In terms of journey distance, results are similar to the ones for car ownership

with 0.9 extra miles and duration falling by 2.1 minutes. However, these are values per trip: the weekly

consequence of owning a driving licence will be 3.5 extra miles per week and a saving of nearly 15

minutes (plus the additional effect of owning one or more cars). This difference of magnitude related to

car ownership and licence owning on both trip making and distance is consistent with previous studies

in Canada (Morency et al., 2011; Roorda et al., 2010).

Vulnerable segments

In the case of belonging to certain vulnerable segments (e.g. non-whites, people with mobility

difficulties and single parents), a dummy variable was used indicating whether the individual belonged

or not to the mentioned segments. The criteria on which the individuals were identified is based on the

questions available in the National Travel Survey on both the household (single parents) and

individual levels (non-whites, mobility difficulties).

For those identified as non-whites, no significant effect was found for average distance (a 0.3 mile

reduction per trip), trips per week were reduced by 1.7 and average duration increased by 2.3

minutes. These results indicate more localised activity patterns, lower activity frequency and slower

trips, generated by the fact of non-whites being the segment with the highest bus use. People with

mobility difficulties had significant negative effects for all three dependent variables: making on

average 2 fewer trips per week, the average distance was 0.6 miles shorter, and the average

duration 2.4 minutes less. This decrease in trip frequency and travel distance is consistent with

other similar studies conducted in the UK (Schmöcker et al., 2005) and in Canada (Farber and Páez,

2010).

Finally, coefficients related to being a single parent (one-adult households with one or more children)

indicated that on average such persons make an additional 1.0 trips per week, with a lower

average trip length (by 0.9 miles) and lower trip duration of 0.3 minutes. However, the

significance levels of these differences are quite low in all three models and this suggests that

single parents have very similar trip making patterns to single-person households. These

findings are consistent with previous studies for single parents in Canada for both trip-making (Roorda

et al, 2010) and travelled distance (Morency et al., 2011). As previously noted this does not take into

account the interaction effects, which means that belonging to certain vulnerable segments might have

additional effects on travel behaviour when being considered as an interaction with variables such as

gender, income, licence owning, etc.



5. TRIP PURPOSE ANALYSIS

Multiple linear regressions were estimated for the analysis of trip purposes, considering a 90%

confidence level and only the main effects of the same variables used previously (Annex tables 3, 4

and 5). Four trip purposes were considered as most important for this particular analysis: i) commuting

(Home-Based Work), ii) social (Home-Based Social), iii) VFR (Home-Based Visiting Friends or

Relatives) and iv) services (Home-Based Shopping and Personal Business). Commuting is considered

to be an obligatory purpose and it represents the production of goods (in this case money). The

second and third variables represent social travel (a growing proportion of people’s overall travel in

recent years). Social trips represent leisure trips which probably incur additional costs at the

destination, whereas VFR represents trips probably do not incur an additional cost. Service trips

represent the main consumption activities of a household, defined as shopping and personal business

(e.g. going to the doctor or the dentist). Both essential food and secondary-need shopping trips are

included for the sake of simplicity but it is recognised that some purposes may be essential while

others may not.

i) Trip generation (Annex table 4)

In general terms, all models obtained a model fit (Adjusted R2) between 0.6 and 0.8. Overall trips (0.81)

and commuting trips (0.84) were the ones with the highest R2

value and social (0.63) and VFR (0.66)

the ones with the lowest one. This tendency shows that there is a higher fit for mandatory activities as

socioeconomic variables do not perform as well as expected when analysing non-mandatory trips

(Páez and Farber, 2012; Farber and Páez, 2010). Results per variable type are further analysed as it

follows.

Area type

Mixed results are obtained as the difference between the categories and the reference (Metropolitan

built-up urban areas such as Merseyside or Newcastle) seem to be too small to be relevant as trip

frequency is defined on a weekly basis. Area-type variables are mostly non-significant for commuting

trips making them not directly affected by the urban context. However, living in more populated urban

areas shows higher trip frequencies for all the analysed purposes except for social trips. The case of

social trips is interesting as living in zones with higher population could imply higher levels of

interaction with other individuals which does not seem to be the case. Living in London implies less

frequent VFR and Services trips and is also the area-type with the highest negative effect (-1.4 trips

per week) along with living in rural areas (-1.1 trips per week). Other area-based variables such as

log-density and the index of deprivation did not have much effect on trip making as log-density

seemed to be non-significant and the index of deprivation irrelevant for all the purposes analysed.


It can be noted that coefficients related to NTEM original variables such as the number of adults and

cars keep their sign but reduce their value. The number of adults per household shows a positive

effect on commuting trip rates but negative values for the other purposes while the number of cars per

household has a positive effect on all purposes.

Income effects are significant among households with higher income levels. This relationship is

positive for social trips indicating a higher propensity to perform social activities for those in the highest

income bands. These results are consistent with previous experience regarding social trips (Van den

Berg et al., 2011). Also, the fact of social trips probably incurring in additional costs make them less

preferable for those in lower income bands and might be more satisfactory than activities on the

relative’s/friend’s household.

The presence of children in the household has a negative effect on all the trip purposes analysed with

commuting (-0.59) and social (-0.49) being the most negative. This show how there is a reduction on

out-of-home joint activities and a slight, yet statistically significant, effect on commuting.



As for individual characteristics, full-time workers seem to have lower trip frequencies in the case of

social and VFR trips. However, they have the highest frequencies in terms of commuting trips while

part-time workers make the most overall trips. Non-workers and retirees have the highest social and

services trip rates indicating that non-mandatory trip rates rise when temporal restrictions are relaxed

(Van den Berg, 2011; Carrasco, 2008) a tendency that is also demonstrated by the fact that students

and non-workers are the segments with the highest VFR trip rates. In the case of gender, females

generate slightly less commuting (-0.6) and social (-0.3) trips than their male counterparts but still

prove to have higher overall trip frequency, and the difference in both commuting and social trips is

irrelevant on a weekly time scope. Licence owners also prove to have slightly higher trip rates in the

analysed purposes even though the overall effect of licence owning is much bigger than the effect per

purpose.

Vulnerable segments

Turning to the vulnerable segments, in common with the findings of Páez and Farber (2012), the non-

white sample in our models demonstrated higher commuting frequencies but lower social, VFR and

Services frequency. However, the effect of being non-white on services and VFR and Services effect

is very low. The effect on commuting and joint activities (social and VFR) can be considered either as

a trade-off between production and leisure activities, the existence of a smaller supply of leisure

facilities (and services) in the vicinity, or as an effect of recent immigration implying a smaller social

network (e.g. less contacts) and cultural issues implying a higher propensity of performing in-home

social activities.

In the case of people with mobility difficulties, there was no effect on commute frequency and a

negative effect in social (Schmöcker et al., 2005), VFR and services trips. This negative effect is

expected considering the overall decrease in trip frequency previously noted in this paper. However,

the magnitude of the coefficient for social trips is high when compared to the rest and reflects a

tendency to either being involved in less social activities or having most of the social activities based

on the individual´s household.

Although being non-statistically significant for VFR trips, being a single parent implies negative effects

on trip frequency for all the analysed purposes. In the case of commuting, the effect can be explained

by single parent’s employment status (Table 2), indicating a smaller full-time work and a higher

economically inactive share. As for social and services trips, there may be an effect of both financial

(e.g. due to lower income levels) and time constraints (e.g. due to the presence of children and the

fact of being a single adult household). Also, and as noticed in Table 3, there is a trade-off between

trips involving the analysed purposes and others (e.g. escort trips).

ii) Trip distance and duration (Annex Tables 5 and 6)

In this case, results indicated that though consistent, the model fit was clearly low (between 0.2 and

0.4). This tendency indicates that the relationship among the variables might not be linear and that the

use of another model specification such as a log-linear model will allow us to achieve better fit and

hence, more robust results. In the case of trip duration, results are related to those for trip distance

with slight differences among the magnitude of certain coefficients as an outcome of mode choice (i.e.

faster modes). Also, the overall model fit of these family of models seem to be higher than in the trip

distance case (R-squared between 0.4 and 0.7) but still, the use of non-linear model specifications

might show to be a more robust method. Comparing these results with previous studies shows

similarities to results obtained for social trip distance (Van den Berg et al., 2011), and commuting

distance (Maoh and Tang, 2012; Farber and Páez, 2010) and duration (McQuaid and Chen, 2012).

Area type

As expected, higher values are obtained for the coefficients related to less populated areas. This

becomes much higher in the case of VFR and trips to services with 7.9 and 2.5 extra miles per trip,

respectively, when compared to the Metropolitan reference category. Log-density shows a negative

effect on all purposes while the index of deprivation proves to be positive for all purposes. In the case



of mean trip duration, results indicated that those living in London and Rural areas have the highest

durations. However, there is a major difference in the cause of this phenomenon as Londoners are

affected by congestion and the lower speed of public transport while bigger distance between origins

and destinations is the reason for those living in rural areas. Log-density has a positive effect on

duration for all purposes except “services” while the index of deprivation has a negative effect on all

but VFR.


The number of adults per household implies a decrease on distance travelled especially in VFR trips.

On the other side, the number of cars per household implies higher trip distances for all purposes

except for VFR. Household income shows a positive effect on distance for all purposes. However, this

effect is only statistically significant for VFR and overall trips in case of living in households with higher

income levels. The effect is most noted for VFR with 12.2 extra miles when generating over £75,000

p.a. while those in the £40,000-£49,999 p.a. band have 4.0 extra miles. This phenomenon is

interesting to take into consideration due to its consistency with previous experience in Canada

(Carrasco et al., 2008) as people with lower income had spatially more distributed social contacts but

at the same time more localised face-to-face interactions than their higher income counterparts. The

presence of children implies more localised activity patterns, especially for VFR purposes (-3.1 miles

per trip). Attributes such as number of adults have a positive effect on trip duration (except for VFR)

while the number of cars has a negative effect on all purposes as an effect higher speeds due to car

use/availability. Household income, on the other hand, proves to show mixed results as higher

durations are related to higher income level (as well as trip distance). The presence of children has a

negative effect on all the analysed purposes.

Full-time workers and students have the highest travelled distance per trip. This higher trip distance for

full time workers participating in social activities was also identified in previous experience (Van den

berg et al, 2011). In the case of gender, females also have more localised activity patterns in the case

of commuting trips with a 1.9 miles decrease (Maoh and Tang, 2012) and hence, shorter commute

duration (McQuaid and Chen, 2012). This finding might be relevant for further analysis as “women´s

job choice set” might be clearly constrained by the “distance effect”. Driving licence owners show

higher trip distances, as expected, though driving licence ownership seems not having an effect on

commuting trips (when only full and part-time workers are considered). As for trip duration, individual

characteristics show consistency with the results obtained for trip distance. However, modal or speed

effects can be identified when ranking the segments with highest trip distances or durations. For

example, students show to have slower trips than full-time workers in all categories.

Vulnerable segments

Different tendencies were identified in the case of vulnerable segments. Non-whites show higher trip

distances for all purposes (especially for VFR trips with an increase of 1.6 miles per trip) while people

with mobility difficulties and single parents did not have significant variations except for VFR with a

decrease of 2.1 miles per trip. In the case of trip duration, non-whites show higher trip durations

generated by higher bus use while the mobility difficult prove to have smaller durations as an effect of

smaller activity spaces. Being a single parent shows no effect on any of the trip duration of the

analysed purposes. Vulnerable segments were consistent with prior experience in the case of non-

white commuting distance (Maoh and Tang, 2012).

6. CONCLUSIONS

Our paper has identified a number of studies that have analysed potential exclusion processes from

the travel behaviours of socially disadvantaged and vulnerable population groups. In particular, we

have shown that some traditional statistical modelling techniques can be applied to offer new insights

in this respect but that there are also certain vulnerable segments that need to be studied further due

to their lack of attention within the literature to date, especially low-income, the elderly, ethnic

minorities and single parents.



We developed a baseline model based on the UK’s National Trip End Model (NTEM) to compare the

generation, distance and duration level with the NTEM’s original variables and then proposed an

extended model that included more socio-economic characteristics and belonging to different

vulnerable population segments. Our results indicated that the improvement in the model fit is of

nearly 1 or 2% of the variance (though not considering the constant) which in relative terms is not

much. However, the inclusion of additional variables represented the effects of important variables in

terms of theoretical consistency. Income effects proved to be particularly important as trip-making

seems to maintain a steep and more or less linear rise in the number of trips between £6,000

and £25,000 p.a. but there appears to be no significant increase in trip making after this.

Presence of children within the household and household size (number of adults) are highly significant

for all three dependent variables implying more localised travel patterns in both cases and higher trip

frequencies for those having children while additional adults implied lower trip rates as an outcome of

task assignment. Vulnerable segment analysis indicated that people with mobility difficulties and single

parents have more localised activity patterns (also single parents) and lower trip rates (also non-

whites).

In the case of purpose analysis, this shows that there is a higher fit for mandatory activities while

socioeconomic variables do not perform as well as expected when analysing non-mandatory trips.

Area-type categorization captured most of the spatial effect per purpose. These findings indicated that

those living in London and Rural areas had the lowest trip-making rates while trip distance was

negatively related to area-type size especially for VFR trips. Person type was crucial as social trips

rise in segments with fewer temporal constraints (i.e. children, retired and non-working),

demonstrating that non-obligatory trips (i.e. social and services) rise when temporal restrictions are

relaxed. Students travel the farthest (and longest) for visiting friends and relatives, but in the case of

social trips, full time workers travel the furthest. Income effects are especially significant in VFR trip

distances, indicating that family bonds, in terms of face-to-face interactions, might be restricted by

income. This effect also proved to be negative for VFR trip-making while being positive for social trips,

indicating that a trade-off between face-to-face interactions and expenditures is occurring when higher

budgets are available (VFR trips are supposed not to involve expenses while social trips do). In the

case of vulnerable segments, people with mobility difficulties and single parents have more localised

activity patterns while ethnic minorities tend to travel further specially for VFR purposes.

These preliminary results demonstrate that income effects and other indices of social disadvantage

have a significant influence on travel behaviours (or this may be vice versa but is not demonstrated

here). It is clear from our analysis that there are difficulties of interpretation in this respect: at the

simplest, people may travel more because they want to (as evidenced by the income effect) or

because they are obliged to (as evidenced by the presence of children in a household). Hence, the

number of trips per week is not an unambiguous indicator. Similar remarks relate to average distance

and average duration. The ratio of these two is related to average speed, and may be treated as an

index of transport service quality. However, once again, people may elect to travel further (e.g. to

access better quality destinations) or may be obliged to, because of a lack of nearby services. Spatial

analysis poses some problems for surveys such as NTS, where the information about the locality

tends to be restricted. Nonetheless, there are some proxy variables that can be used, for example

ward density, public transport availability and frequencies, as well as the perceived accessibility of

essential services (doctors, post office etc.).

The cost of travel is another important effect on the supply of transport which is difficult to fully capture

using NTS data. Although there is data collected within the travel diary on the cost of public transport

fares, the cost of car trips is not recorded and must be calculated based on the vehicle mileage

data. This requires considerable additional analytical effort and is also not necessarily reliable. There

are other contextual factors such the timing and availability of transport services and perceptions of

personal safety which have also been identified within the literature as affecting an individual’s

willingness to travel to different destinations. Although it could be possible to extract further proxy



measures from the NTS to represent these additional factors this is unlikely to be very useful for

policymakers without a spatial understanding of where these problems are occurring.

There are other indicators such as mode choice and time use that will help us to further analyse travel-

activity behaviour of the social and transport disadvantaged. Our study is still in its preliminary stages

and there are plans to further refine our national level analysis to consider interactions between

different independent variables and as well as mode-based analysis. We also want to re-analyse the

NTS data at the regional level to understand the differences in travel behaviours of low income and

vulnerable groups living in certain parts of the country that are well-known for their low levels of

employment and high intensities of income poverty.

Finally, we will be undertaking a bespoke survey of 700 households in one of these low-income

regions of the travel behaviours of low-income households in order to collect more detailed data on

their travel expenditures, attitudes and perceptions of the transport system. This data will be used to

develop local area models of transport and social disadvantage, which will be used to predict the

effects of different local policy options on socially disadvantaged households and will be reported in

subsequent project reports and working papers.

References

Carrasco, J.A., E.J. Miller, and B. Wellman (2008). How far and with whom do people socialize?

Empirical evidence about distance between social network members. Transportation Research

Record: Journal of the Transportation Research Board 2076 pp.114-122.

Casas, I., Horner, M., and Weber, J. (2009). A Comparison of Three Methods for Identifying

Transport-Based Exclusion: A Case Study of Children's Access to Urban Opportunities in Erie and

Niagara Counties, New York. International Journal of Sustainable Transportation 3 (4) pp.227-245

Cebollada, A. (2009). Mobility and labour market exclusion in the Barcelona Metropolitan Region.

Journal of Transport Geography 17 (3) pp.226-233

Church, A., Frost, M. and Sullivan, K. (1999) ‘Transport and social exclusion in London’ Proceeding of

the European Transport Conference www.etcproceedings.org/paper/download/2528

Currie, G. (2010). Quantifying spatial gaps in public transport supply based on social needs. Journal of

Transport Geography 18 (1) pp.31–41

Currie, G. (2011), New perspectives and methods in transport and social exclusion research. Emerald.

Currie, G., and Delbosc, A. (2010), Modelling the social and psychological impacts of transport

disadvantage. Transportation.

Currie, G., and Stanley, J. (2008). Investigating Links between Social Capital and Public Transport.

Urban Policy and Research 28 (4) pp.529-547

Currie, G., Richardson, T., Smyth, P., Vella-Brodrick, D., Hine, J., Lucas, K., Stanley, J., Morris, J.,

Kinnear, R., Stanley, J., (2010), Investigating the links between transport disadvantage, social

exclusion and well-being in Melbourne – Updated results. Research in transport economics 29.

Currie, G., Richardson, T., Smyth, P., Vella-Brodrick, D., Hine, J., Lucas, K., Stanley, J., Morris, J.,

Kinnear, R., and Stanley, J.(2009). Investigating links between transport disadvantage, social

exclusion and well-being in Melbourne-Preliminary results. Transport Policy 16 (3) pp.97-105



Dargay, J. M. (2001). The effect of income on car ownership: evidence of asymmetry. Transportation

Research Part A 35pp.807-821

Delbosc, A., Currie, G. (2011).Using Lorenz curves to assess public transport equity. Journal of

Transport Geography 19.

Department for Transport (2006a). Full Guidance on Accessibility Planning

http://www.dft.gov.uk/pgr/regional/ltp/accessibility/guidance/gap/nicalguidanceonaccessibi3641.pdf

Department for Transport (2006b) Social Inclusion: Transport Aspects. Final Report. Department for

Transportation, UK. Centre of Transport Studies at Imperial College, Mott Mac Donald and Institute for

Transport Studies at University of Leeds.

Department for Transport (2011). National Travel Survey: 2010. Statistical release.

Duvarci, Y. and Yigitcanlar, T. (2007). Integrated Modelling Approach for the Transportation

Disadvantaged. Journal of Urban Planning and Development 133 (3) pp.188-200

Farber, S., Paez, A., Mercado, R., Roorda, M., and Morency, C. (2011). A time-use investigation of

shopping participation in three Canadian cities: is there evidence of social exclusion? Transportation

38 (1) pp.17-44

Farrington, J. (2007) ‘The new narrative of accessibility: its potential contribution to discourses in

(transport geography)’ Journal of Transport Geography 15pp.319-330

Grieco, M. (2006) ‘Accessibility, Mobility and Connectivity: The Changing Frontiers of Everyday

Routine’ European Spatial Research and Policy (Special Issue) 78 (6)pp.1360-80

Hurni, A. (2006) Transport and Social Exclusion in Western Sydney University of Western Sydney and

Western Sydney Community Forum, Australia

Jara, M. (2009). Indicadores de exclusión social, accesibilidad y movilidad en el contexto urbano

chileno: un análisis desde la perspectiva del comportamiento de transporte. Civil Engineering

undergraduate thesis. Universidad de Concepción, Chile. [In Spanish]

Johnson, V., Currie, G., and Stanley, J. (2011).Exploring transport to arts and cultural activities as a

facilitator of social inclusion. Transport Policy 18 (1) pp.68-75

Levitas, R., Pantazis, C., Fahmy, E., Gordon, D., Lloyd, E. y Patsios, D. (2007) The Multidimensional

Analysis of Social Exclusion. Cabinet Office. UK.

Limanond, T., Jomnonkwao, S., Watthanaklang, D., Ratanavaraha, V., and Siridhar, S. (2011). How

vehicle ownership affect time utilization on study, leisure, social activities, and academic performance

of university students? A case study of engineering freshmen in a rural university in Thailand.

Transport Policy 18 (5) pp.719–726

Litman, T. (2003). Social inclusion as a transport planning issue in Canada. Paper presented at the

Transport and Social Exclusion G7 Comparison Seminar. London, April 2003.

Loo, B.P.Y., and Lam, W.W.Y.(2012). A multilevel investigation of differential individual mobility of

working couples with children: a case study of Hong Kong. Transportmetrica.



Lopez, A., Gutierrez, J., and Gomez, G. (2008). Measuring Regional Cohesion Effects of Large-scale

Transport Infrastructure Investments: An Accessibility Approach. European Planning Studies 16 (2) pp.

277-301

Lovejoy, K., and Handy, S. (2011). Social networks as a source of private-vehicle transportation: The

practice of getting rides and borrowing vehicles among Mexican immigrants in California.

Transportation Research Part A: Policy and Practice 45 (4) pp.248-257

Lucas, K. (2012), Transport and social exclusion: Where are we now? Transport Policy 20.

Lucas, K., Grosvenor T. and Simpson R. (2001) Transport, the Environment and Social Exclusion

York: Joseph Rowntree Foundation/ York Publishing Ltd.

Lucas, K., Tyler, S. and Christodoulou, G. (2009). Assessing the ‘value’ of new transport initiatives in

deprived neighbourhoods in the UK. Transport Policy 16 (3) pp.115-122

McQuaid, R., and Chen, T. (2012). Commuting times - The role of gender, children and part- time

work. Research in Transportation Economics 34 pp.66-73

Mackett, R., Achuthan, K. and Titheridge, H. (2008) ‘AMELIA: A tool to make transport policies more

socially inclusive’ Transport Policy 15: 372-378

Manaugh, K., Miranda-Moreno, L., and El-Geneidy, A. (2010). The effect of neighbourhood

characteristics, accessibility, home-work location, and demographics on commuting distances.

Transportation 37 pp.627-646.

Maoh, H., and Tang, Z. (2012). Determinants of normal and extreme commute distance in a sprawled

midsize Canadian city: evidence from Windsor, Canada. Journal of Transport Geography 25 pp.50-57

Mercado, R., and Páez, A. (2009). Determinants of distance travelled with a focus on the elderly: a

multilevel analysis in the Hamilton CMA, Canada. Journal of Transport Geography 17 (1) pp.65–76

Mercado, R., Paez, A., Farber, S., Roorda M., and Morency, C. (2012). Explaining transport mode use

of low-income persons for journey to work in urban areas: a case study of Ontario and Quebec.

Transportmetrica 8 (3) pp.157-179

Morency, C., Paez, A., Roorda, M., Mercado, R., and Farber, S. (2011). Distance travelled in three

Canadian cities: Spatial analysis from the perspective of vulnerable population segments Journal of

Transport Geography 19 (1) pp.39-50

Neutens, T., Delafontaine, M., Scott, D., De Maeyer, P., (2012), A GIS-based method to identify

spatiotemporal gaps in public service delivery. Applied Geography 32.

Ohnmacht, T., Maksim, H. and Bergman, M.M. (2009) Mobilities and Inequalities Surrey, United

Kingdom: Ashgate

Páez, A. and Farber, S. (2010). Understanding the transportation situation of Canadian adults with

disabilities. Human Resources and Skills. Canada.

Páez, A., and Farber, S. (2012). Participation and desire: leisure activities among Canadian adults

with disabilities. Transportation



Páez, A., Mercado, R., Farber, S., Morency, C., and Roorda, M. (2009). “Mobility and Social Exclusion

in Canadian Communities: An Empirical Investigation of Opportunity Access and Deprivation from the

Perspective of Vulnerable Groups. Human Resources and Social Development. Canada.

Pitot, M., Yigitcanlar, T., Sipe, N., and Evans, R. (2006). "Land Use and Public Transport Accessibility

Index (LUPTAI) tool: the development and pilot application of LUPTAI for the Gold Coast" (2006).

Social Research in Transport (SORT) Clearinghouse. Paper 123.

Preston, J. (2009) ‘Editorial’ in Transport Policy Special Issue 16: 3: 140-142 (Guest editors Lucas, K.

and Stanley, J.)

Roorda, M., Paez, A., Morency, C., Mercado, R., and Farber, S. (2010). Trip generation of vulnerable

populations in three Canadian cities: a spatial ordered probit approach. Transportation 37 pp.525-548.

Scheiner, J. (2010). Social inequalities in travel behaviour: trip distances in the context of residential

self-selection and lifestyles. Journal of Transport Geography 18 (6) pp.679–690

Scheiner, J., and Holz-Rau, C. (2012). Gendered travel mode choice: a focus on car deficient

households. Journal of Transport Geography 24 pp.250-261

Schmöcker, J., Quddus, M., Noland, R., and Bell, M. (2008). Mode choice of older and disabled

people: a case study of shopping trips in London. Journal of Transport Geography 16 (4) pp.257-267

Schmöcker, J., Quddus, M., Noland, R., and Bell, M. (2005). Estimating trip generation of the elderly

and disabled people: An analysis of London data. Transportation Research Record1924 pp.9–18.

Schönfelder, S., and Axhausen, K. (2003). Activity spaces: measures of social exclusion? Transport

Policy 10 (4) pp.273-286

Scott, D., and He, S. (2012). Modelling constrained destination choice for shopping: a GIS-based,

time-geographic approach. Journal of Transport Geography 23 pp.60-71

Sheller, M. and Urry, J. (2006) ‘The new mobilities paradigm’ Environmental Planning A38: 207-226

Social Exclusion Unit (2003) Making the Connections: Final Report on Transport and Social Exclusion,

Office of The Deputy Prime Minister.Social Exclusion Unit. UK.

Spinney, J., Scott, D., and Newbold, B. (2009).Transport mobility benefits and quality of life: A time-

use perspective of elderly Canadians. Transport Policy 16 (1) pp.1-11

TRaC (2000) Social Exclusion and the Provision of Public Transport London: Department of

Environment Transport and the Regions

Urry, J. (2000). Sociologies beyond Societies: Mobilities for the Twenty-First Century. New York:

Routledge.

Urry, J. (2007) Mobilities Cambridge, United Kingdom: Polity Press

WSP (2000), Land-Use Indicators and Trip end Models, Final Report for HETA Division of the

Department of the Environment, Transport and the Regions, January 2000



WSP (2009), Research into Changing Trip Rates over Time and Implications for the National Trip End

Model: Draft Final Report for Department for Transport. May 2009

Van den Berg, P., Arentze, T., and Timmermans, H. (2011). Estimating travel demand of senior

citizens in the Netherlands. Journal of Transport Geography 19 (2) pp.323-331.

Van de Voorde, E. and Vanelslander, T. (2010) (eds.) Applied Transport Economics: a management

and policy perspective Antwerp, Belgium: Uitgerverji De Boeck

Zhang, J., Xu,L., and Fujiwara, A. (2012). Developing an integrated scobit-based activity participation

and time allocation model to explore influence of childcare on women’s time use behaviour.

Transportation 39 (1) pp.125-149



Annex Table 1 Variable description

Variable name Type Description

Person type

Child Dummy 1 if under 16yrs

Full time Dummy 1 if Full time employed and aged 17-64yrs

Part time Dummy 1 if Part time employed and aged 17-64yrs

Student Dummy 1 if Student and aged 17-64yrs

Non-working Dummy 1 if non-working and aged 17-64yrs

Retired Dummy 1 if over 65yrs

Area type

London Dummy 1 if living in London

Metropolitan Dummy 1 if living in built-up area (e.g. Merseyside)

Urban big Dummy 1 if living in urban areas w/ population over 250K

Urban large Dummy 1 if living in urban areas w/ population 100–250K

Urban medium Dummy 1 if living in urban areas w/ population 25–100K

Urban small Dummy 1 if living in urban areas w/ population 3–25K

Rural Dummy 1 if living in rural areas

Urban context

Log-density Continuous Logarithm (base 10) of PSU density (persons per hectare)

Index of deprivation Continuous Based on accessibility, 1 the most deprived 10 the least

Household characteristics

N of adults per HH Ordinal N of adults per HH (0, 1, 2, 3+)

N of cars per HH Continuous N of cars per HH (0-5 cars)

HH income perannum

Dummy 1 if belonging to income band

Children in HH Dummy 1 if there are children in the household

Individual characteristics

Gender Dummy 1 if female

Driving licence Dummy 1 if individual owns a driving license

Social disadvantages

Non-white Dummy 1 if non-white

Mobility difficulties Dummy 1 if individual has mobility difficulties

Single parent Dummy 1 if Single parent



Annex Table 2 Descriptive statistics of the exogenous variables

Variable Share (%) Mean* Min* Max* P75* P50* P25*No. ofcases

Person type

Child 20.7 40,365

Full time 34.5 67,275

Part time 10.7 20,865

Student 3.2 6,240

Non-working 14.3 27,885

Retired 16.6 32,370

Area type

London 15.0 29,250

Metropolitan 14.4 28,080

Urban big 12.5 24,375

Urban large 11.7 22,815

Urban medium 14.8 28,860

Urban small 16.7 32,565

Rural 14.9 29,055

Gender

Female 51.7 100,872

N of adults in HH

1 adult 17.1 33,345

2 adults 57.7 112,515

2+ adults 25.2 49,140

Index of deprivation

1 (Most) 14.4 28,080

2 14.0 27,300

3 10.5 20,475

4 9.9 19,305

5 10.0 19,500

6 8.8 17,160

7 8.2 15,990

8 7.2 14,040

9 4.9 9,555

10 (Least) 12.3 23,985

Personal attributes (as a % of total sample)

Children in HH 45.1 87,945

Driving license 61.4 119,730

Non-white 9.7 18,915

Mobility diff. 11.3 22,035

Unskilled HRP 4.0 7,800

Single parent 2.0 3,900

Continuous variables

Log-density - 1.16 0.18 1.96 1.59 1.27 0.54 195,000

N of cars in HH - 1.3 0 5 2 1 1 195,000

Income (£ ,000) - 35.4 0.8 125 45.0 27.5 13.8 195,000*Only for continuous variables



Annex Table 3 Trip frequency per purpose

Commute SocialVisiting Friendsand Relatives

Services All

B t B t B t B t B tChild - 3.89 19.2 3.49 24.3 3.51 20.9 9.94 27.5Full-time Worker 8.40 21.2 2.99 14.8 2.96 20.6 3.30 19.8 11.66 32.3Part-time Worker 6.35 16.0 3.62 17.7 3.15 21.6 4.23 25.0 13.26 36.2Student - 3.72 17.9 3.37 22.7 3.29 18.8 10.61 28.4Non-worker - 4.25 21.0 3.66 25.5 5.56 33.4 10.76 29.8Retired - 4.25 21.0 3.15 21.8 6.21 37.3 9.86 27.3London 0.04 0.8 -0.11 -2.4 -0.31 -9.9 -0.38 -10.1 -1.41 -17.7Metropolitan (Ref.) Ref. Ref. Ref. Ref. Ref.Urban (over 250K) 0.00 -0.1 0.11 2.6 -0.07 -2.2 -0.08 -2.1 -0.17 -2.1Urban (100-250K) 0.05 0.9 0.06 1.3 -0.14 -4.3 -0.04 -1.1 -0.15 -1.9Urban (25-100K) 0.12 2.2 0.12 2.7 -0.05 -1.5 -0.13 -3.2 -0.31 -3.7Urban (3-25K) 0.09 1.4 0.17 3.6 -0.08 -2.2 -0.22 -5.2 -0.93 -10.1Rural (Ref.) 0.05 0.7 0.05 0.9 -0.28 -6.9 -0.30 -6.1 -1.11 -10.3Female -0.63 -19.6 -0.33 -13.8 -0.03 -1.7 -0.03 -1.5 0.49 10.7N of adults in HH 0.45 16.1 -0.19 -8.2 -0.10 -5.9 0.03 1.4 -1.06 -25.0N of cars in HH 0.06 2.9 0.21 12.5 0.19 15.0 0.22 14.4 1.36 42.4HH income p.a.Less than £1,000 Ref. Ref. Ref. Ref. Ref.£1,000- £1,999 -0.16 -0.3 0.29 1.1 0.13 0.7 -0.14 -0.7 1.09 2.4£2,000- £2,999 0.23 0.5 0.47 1.9 0.26 1.5 -0.14 -0.7 0.41 1.0£3,000- £3,999 -0.29 -0.6 0.26 1.1 0.22 1.4 -0.32 -1.7 -0.03 -0.1£4,000- £4,999 0.04 0.1 0.44 2.1 0.22 1.5 -0.41 -2.3 -0.04 -0.1£5,000- £5,999 -0.11 -0.2 0.23 1.1 0.21 1.5 -0.27 -1.6 0.05 0.1£6,000- £6,999 -0.20 -0.5 0.20 1.0 0.08 0.6 -0.17 -1.0 -0.19 -0.5£7,000- £7,999 -0.35 -0.8 0.13 0.6 0.03 0.2 -0.21 -1.3 0.07 0.2£8,000- £8,999 -0.14 -0.3 0.07 0.3 0.20 1.4 0.05 0.3 0.66 1.8£9,000- £9,999 -0.30 -0.7 0.25 1.2 0.12 0.9 -0.09 -0.5 0.36 1.0£10,000- £12,499 -0.04 -0.1 0.29 1.5 0.06 0.4 0.08 0.5 0.65 1.9£12,500- £14,999 -0.08 -0.2 0.29 1.5 0.01 0.1 0.17 1.1 1.16 3.4£15,000- £17,499 -0.34 -0.9 0.29 1.5 0.02 0.1 0.07 0.4 1.23 3.6£17,500- £19,999 -0.07 -0.2 0.28 1.4 0.05 0.4 -0.03 -0.2 1.18 3.4£20,000- £24,999 -0.24 -0.6 0.41 2.1 -0.02 -0.2 0.07 0.4 1.71 5.0£25,000- £29,999 -0.32 -0.8 0.41 2.1 -0.06 -0.4 -0.02 -0.2 1.71 5.0£30,000- £34,999 -0.51 -1.3 0.51 2.6 -0.17 -1.2 -0.11 -0.7 1.68 4.9£35,000- £39,999 -0.59 -1.5 0.60 3.1 -0.21 -1.5 -0.13 -0.8 1.96 5.7£40,000- £49,999 -0.74 -1.9 0.55 2.9 -0.22 -1.6 -0.09 -0.6 1.89 5.5£50,000- £59,999 -0.92 -2.4 0.59 3.0 -0.25 -1.8 -0.14 -0.8 1.93 5.6£60,000- £69,999 -1.05 -2.7 0.62 3.2 -0.42 -3.0 -0.25 -1.5 1.70 4.8£70,000- £74,999 -0.94 -2.4 0.70 3.4 -0.31 -2.1 -0.14 -0.8 2.14 5.8£75,000 or more -1.28 -3.3 0.74 3.8 -0.57 -4.1 -0.28 -1.7 1.62 4.7Presence of children -0.51 -15.8 -0.49 -16.2 -0.09 -4.0 -0.12 -4.5 2.08 36.0Car License own. -0.12 -2.3 0.62 16.0 0.21 7.6 0.94 29.9 4.51 65.8Non-white 0.33 6.2 -0.45 -9.7 -0.11 -3.5 -0.11 -3.0 -1.71 -22.1Mobility difficulties -0.02 -0.3 -0.74 -16.0 -0.12 -3.6 -0.20 -5.6 -1.95 -24.0Single Parent -0.38 -3.2 -0.15 -1.5 -0.05 -0.8 -0.21 -2.7 0.96 5.5Log-Density 0.01 0.7 -0.03 -2.0 0.00 0.1 -0.01 -0.5 -0.08 -2.6Index of Deprivation -0.04 -6.3 0.03 7.8 -0.03 -7.8 0.02 5.0 0.13 14.9Sum of squaresResidual 497,802 583,522 239,247 879,605 7,946,527Total 3,027,536 1,557,000 709,867 2,965,228 41,470,027DoF 40 44 44 44 44F-Ratio 6.363 3.191 3.171 5.900 13.688Adjusted R

20.84 0,63 0,66 0,70 0,81

No. of Obs. 50,123 84,192 70,970 109,538 142,807Obs: *statistically non-significant at the 95% level. Dark cells indicate non-significance at the 90% confidence level. Boldcoefficients represent those of major interest.



Annex Table 4 Trip distance per purpose


Services All

B t B t B t B t B tChild - 6.97 6.7 10.81 5.2 6.24 14.9 8.06 14.9Full-time Worker 5.50 7.2 9.19 8.9 11.03 5.3 6.83 16.4 10.68 19.8Part-time Worker 3.49 4.5 7.75 7.4 10.42 4.9 6.41 15.1 8.44 15.4Student - 7.75 7.3 13.57 6.3 6.98 16.0 10.30 18.5Non-worker - 7.48 7.2 10.35 4.9 6.23 14.9 8.68 16.1Retired - 7.44 7.2 10.05 4.8 5.74 13.8 7.93 14.7London 0.65 6.2 -0.73 -3.3 0.43 0.9 0.06 0.7 0.05 0.5Metropolitan 0

a0

b0

b0

bRef.

Urban (over 250K) 0.53 5.0 0.62 2.8 0.90 2.0 0.11 1.1 0.52 4.3Urban (100-250K) 0.63 5.8 0.55 2.4 1.86 4.1 0.06 0.7 0.41 3.3Urban (25-100K) 0.91 8.3 0.57 2.5 0.70 1.5 0.17 1.7 0.50 4.0Urban (3-25K) 1.07 8.8 1.11 4.5 3.01 5.9 1.39 13.0 1.67 12.1Rural (Ref.) 1.60 11.4 1.74 6.1 7.93 13.2 2.48 19.8 2.56 15.9Female -1.89 -30.5 -0.65 -5.3 -1.00 -3.9 -0.14 -2.6 -1.43 -20.7N of adults in HH -0.55 -10.3 -0.43 -3.6 -2.26 -9.3 -0.06 -1.2 -1.16 -18.3N of cars in HH 0.28 7.2 0.67 7.7 -0.69 -3.7 0.42 10.9 0.69 14.5HH income p.a.Less than £1,000 Ref. Ref. Ref. Ref. Ref.£1,000- £1,999 -0.71 -0.7 0.04 0.0 1.13 0.4 -1.05 -2.0 -0.22 -0.3£2,000- £2,999 -0.23 -0.2 1.35 1.1 0.57 0.2 -0.92 -1.8 -0.70 -1.1£3,000- £3,999 0.51 0.6 -0.95 -0.8 -0.58 -0.2 -0.52 -1.1 -0.87 -1.4£4,000- £4,999 -0.31 -0.4 -0.30 -0.3 -0.28 -0.1 -0.43 -1.0 -0.98 -1.7£5,000- £5,999 -1.14 -1.3 0.44 0.4 -3.78 -1.8 -0.80 -1.9 -1.37 -2.5£6,000- £6,999 -0.82 -1.0 0.15 0.1 -2.62 -1.3 -0.58 -1.4 -0.99 -1.8£7,000- £7,999 -0.75 -0.9 0.60 0.6 -1.93 -0.9 -0.55 -1.3 -1.06 -1.9£8,000- £8,999 -0.79 -1.0 1.06 1.0 -2.81 -1.3 -0.77 -1.9 -0.92 -1.7£9,000- £9,999 -1.24 -1.5 0.64 0.6 -1.72 -0.8 -0.46 -1.1 -0.50 -0.9£10,000- £12,499 -1.62 -2.1 0.45 0.5 -2.02 -1.0 -0.66 -1.7 -0.95 -1.8£12,500- £14,999 -1.43 -1.9 0.56 0.6 -0.18 -0.1 -0.49 -1.2 -0.61 -1.2£15,000- £17,499 -1.19 -1.6 0.78 0.8 -0.69 -0.3 -0.59 -1.5 -0.64 -1.2£17,500- £19,999 -1.16 -1.5 0.56 0.6 0.22 0.1 -0.46 -1.1 -0.27 -0.5£20,000- £24,999 -0.95 -1.3 1.01 1.0 1.48 0.7 -0.31 -0.8 -0.13 -0.2£25,000- £29,999 -0.87 -1.2 1.38 1.4 0.15 0.1 -0.43 -1.1 0.15 0.3£30,000- £34,999 -0.75 -1.0 1.55 1.6 2.99 1.5 -0.35 -0.9 0.24 0.5£35,000- £39,999 -0.54 -0.7 0.85 0.9 2.77 1.4 -0.39 -1.0 0.42 0.8£40,000- £49,999 -0.21 -0.3 1.83 1.8 3.95 2.0 -0.24 -0.6 1.14 2.2£50,000- £59,999 0.14 0.2 1.55 1.6 7.20 3.6 -0.20 -0.5 1.63 3.2£60,000- £69,999 0.09 0.1 1.52 1.5 7.32 3.6 -0.20 -0.5 2.40 4.6£70,000- £74,999 0.09 0.1 1.24 1.2 9.45 4.4 0.12 0.3 2.28 4.2£75,000 or more 0.70 0.9 2.46 2.5 12.21 6.1 0.25 0.6 3.41 6.6Presence of children -0.25 -4.0 -0.79 -5.0 -3.05 -9.2 -0.40 -6.1 -1.19 -13.8Car License own. -0.12 -1.2 0.30 1.5 3.77 9.3 0.19 2.4 0.85 8.3Non-white 0.85 8.2 0.62 2.6 1.40 3.1 0.49 5.2 0.30 2.6Mobility difficulties -0.10 -0.6 0.30 1.2 -2.10 -4.4 -0.17 -2.0 -0.61 -5.0Single Parent -0.43 -1.8 -0.34 -0.7 -2.07 -2.3 -0.16 -0.8 -0.88 -3.3Log-Density -0.04 -0.9 -0.34 -4.2 0.37 2.2 -0.60 -16.7 -0.32 -6.9Index of Deprivation 0.02 1.9 0.05 2.2 0.48 10.1 0.00 -0.4 0.11 8.8Sum of squaresResidual 741,443 5,741,234 15,058,438 2,429,481 8,126,825Total 1,113,123 7,493,832 18,753,481 3,523,850 12,915,028DoF 40 44 44 44 44F-Ratio 628 584 396 1.121 1,912Adjusted R

20.33 0.23 0.20 0.31 0.37




Annex Table 5 Trip duration per purpose


Services All

B t B t B t B t B tChild - 27.60 17.0 31.49 11.5 25.82 34.1 28.88 33.9Full-time Worker 19.68 12.6 32.91 20.3 34.70 12.7 27.95 37.2 35.16 41.3Part-time Worker 14.31 9.1 32.09 19.5 34.90 12.6 27.93 36.6 32.07 37.1Student - 33.28 20.0 40.67 14.4 30.70 39.0 37.56 42.6Non-worker - 32.45 20.0 35.41 12.9 28.07 37.3 31.96 37.5Retired - 31.73 19.5 34.77 12.7 26.94 35.9 30.46 35.7London 6.73 31.4 4.64 13.4 8.11 13.7 4.08 24.0 6.38 33.9Metropolitan Ref. Ref. Ref. Ref. Ref.Urban (over 250K) 0.24 1.1 1.37 4.0 1.70 2.9 0.33 1.9 0.94 4.9Urban (100-250K) -0.23 -1.0 0.61 1.7 1.60 2.7 -0.53 -3.1 -0.37 -1.9Urban (25-100K) 0.38 1.7 0.80 2.3 0.33 0.5 -0.60 -3.4 -0.24 -1.2Urban (3-25K) 0.82 3.3 2.24 5.8 3.59 5.4 1.62 8.4 1.99 9.1Rural 1.84 6.4 3.26 7.3 10.38 13.2 3.13 13.9 3.34 13.1Female -2.88 -22.7 -1.36 -7.0 -1.47 -4.3 -0.11 -1.1 -2.12 -19.5N of adults in HH 0.59 5.4 0.20 1.1 -1.35 -4.3 1.27 14.1 -0.13 -1.3N of cars in HH -1.32 -16.4 -1.45 -10.7 -3.92 -16.2 -1.65 -24.0 -1.76 -23.3HH income p.a.Less than £1,000 Ref. Ref. Ref. Ref. Ref.£1,000- £1,999 -1.23 -0.6 -1.70 -0.8 3.13 0.9 -2.80 -3.0 -0.49 -0.5£2,000- £2,999 0.84 0.4 1.75 0.9 1.49 0.5 -2.93 -3.3 -1.28 -1.3£3,000- £3,999 -1.23 -0.7 -0.87 -0.5 1.84 0.6 -1.83 -2.2 -1.48 -1.5£4,000- £4,999 -0.07 0.0 -0.31 -0.2 0.79 0.3 -2.11 -2.7 -1.99 -2.2£5,000- £5,999 -3.34 -1.9 0.38 0.2 -5.03 -1.8 -2.70 -3.6 -2.85 -3.3£6,000- £6,999 -0.76 -0.5 -0.04 0.0 -4.19 -1.5 -2.92 -3.9 -2.25 -2.7£7,000- £7,999 -1.70 -1.0 1.15 0.7 -3.62 -1.3 -3.00 -4.0 -2.76 -3.2£8,000- £8,999 -1.52 -0.9 1.12 0.7 -3.52 -1.3 -3.47 -4.6 -2.30 -2.7£9,000- £9,999 -3.63 -2.2 0.94 0.6 -4.56 -1.7 -3.33 -4.4 -2.45 -2.9£10,000- £12,499 -2.79 -1.8 0.11 0.1 -4.95 -1.9 -3.83 -5.3 -3.25 -4.0£12,500- £14,999 -2.57 -1.7 -0.71 -0.5 -3.00 -1.1 -3.87 -5.4 -2.93 -3.6£15,000- £17,499 -2.36 -1.5 -0.60 -0.4 -3.98 -1.5 -4.01 -5.6 -3.05 -3.7£17,500- £19,999 -2.38 -1.5 -0.87 -0.6 -2.81 -1.1 -4.06 -5.6 -2.81 -3.4£20,000- £24,999 -2.48 -1.6 -0.68 -0.4 -1.47 -0.6 -3.90 -5.5 -2.55 -3.2£25,000- £29,999 -2.27 -1.5 -0.26 -0.2 -3.24 -1.2 -4.38 -6.1 -2.47 -3.1£30,000- £34,999 -2.37 -1.6 -0.41 -0.3 0.04 0.0 -4.27 -6.0 -2.21 -2.7£35,000- £39,999 -1.73 -1.1 -1.28 -0.8 -0.09 0.0 -4.44 -6.2 -2.00 -2.5£40,000- £49,999 -1.29 -0.8 0.03 0.0 1.34 0.5 -3.96 -5.6 -0.94 -1.2£50,000- £59,999 -0.57 -0.4 -0.25 -0.2 4.95 1.9 -3.88 -5.4 -0.13 -0.2£60,000- £69,999 -0.57 -0.4 0.18 0.1 5.70 2.1 -4.17 -5.7 1.07 1.3£70,000- £74,999 -0.08 0.0 -0.14 -0.1 8.20 2.9 -3.58 -4.7 1.46 1.7£75,000 or more 0.70 0.5 0.78 0.5 11.34 4.3 -3.18 -4.4 2.77 3.4Presence of children -1.50 -11.9 -2.03 -8.3 -5.48 -12.6 -1.32 -11.1 -3.19 -23.4Car License own. -4.18 -20.3 -2.33 -7.5 0.45 0.8 -2.76 -19.5 -2.06 -12.8Non-white 4.10 19.4 2.40 6.4 3.34 5.6 2.75 16.3 2.23 12.2Mobility difficulties -0.11 -0.3 -1.54 -4.2 -4.26 -6.8 -1.21 -7.7 -2.53 -13.1Single Parent -0.76 -1.6 -0.92 -1.2 -1.51 -1.3 0.61 1.8 -0.61 -1.5Log-Density 0.50 6.1 0.15 1.1 1.06 4.7 -0.37 -5.7 0.20 2.7Index of Deprivation -0.09 -4.2 -0.16 -4.5 0.37 5.9 -0.18 -10.0 -0.02 -1.1Sum of squaresResidual 5,033,072 31,775,283 49,404,205 16,318,618 36,478,293Total 11,033,698 67,579,438 79,364,956 45,575,927 109,442,258DoF 40 44 44 44 44F-Ratio 1,493 2,155 978 4,462 6,490Adjusted R

20.54 0.53 0.38 0,.4 0.7


social disadvantage and transport in the uk: a trip …social disadvantage and transport in the uk:...

Documents