Embed Size (px)
Citation preview
Evaluating the influence of social
capital on travel behaviour of urban
users: The case of TransMiCable
Estudiante: Víctor Cantillo-García
Asesor: Luis A. Guzmán
Co-asesora: Diana Higuera-Mendieta
Jurado externo: Julián Arellana
Jurado interno: Hernán Ortiz
Tesis de Grado
Maestría en Ingeniería Civil
Énfasis en Ingeniería de Transporte
1. Introduction
The relation between social interactions, travel and transport has attracted the interest of
many researches who have addressed the subject from diverse perspectives. This situation is
partly motivated by the hypothesis that a general view of social contact, beyond the study of
intra-household interactions, might help to explain the motivation and characteristics of travel
behaviour, improving the analysis of aspects such as modal choice, the planning and execution
of activities, and travel-location related decisions (Dugundji et al., 2011). At this point, social
capital is a relevant topic when reviewing social interactions. However, social capital is
considered a complex, polysemic and multidimensional concept, leading to different definitions
and interpretations according to the context of study and viewpoints by researchers (Li et al.,
2005; Paldam, 2000).
Some of the most popular definitions of social capital are the ones proposed by Bourdieu
(1986): “Social capital is the aggregate of the actual or potential resources which are linked to
possession of a durable network of more or less institutionalized relationships of mutual
acquaintance and recognition”; and Putnam (2000): “Social capital refers to the features of
social organization such as networks, norms and social trust that facilitate coordination and
cooperation for mutual benefit”. Besides the individual advantages generated by the
possession of these features, the strength of this interactions can also enhance the sense of
community facilitating the capabilities to solve collective problems. Nevertheless, even though
social capital might be interpreted as a property of communities, some empirical research
suggests that the collective consequences of social capital must be consistent with individual
attitudes (Brehm & Rahn, 1997). Moreover, in the light of these definitions, one can infer that
the stock of social capital of an individual depends on the social connections he can
successfully summon, as well as on the volume of economic, cultural and symbolic recourses
he can benefit from those connections. Likewise, it can be seen that in a determined market
context the economic decisions of consumers (i.e. the choice of transportation mode), are
based not only on the individual’s self-interest, but are also influenced by his social relations
and interactions, which can be measured in the form of social capital. Hence, recognizing the
relevance of social capital in the resource allocation process might help to understand the
economic behaviour of consumers, providing more robust tools for policy evaluation and
decision making (Robison et al., 2012).
On the other hand, in the field of transportation engineering it is relevant the analysis and
understanding of the factors that affect travel choices, that are crucial for the evaluation of
policies, infrastructure investments, demand forecast, fare estimations, etc. In particular, the
selection of transportation mode is defined by attributes of the available alternatives, the
context of the journey and the trip maker, including social capital (Ortúzar & Willumsen, 2011,
chapter 6). Transport demand modellers are then concerned on the definition of variables
related to these categories and exploring new model specifications that might help to improve
the capabilities to simulate mode choice. Given this, is no surprise that there is a growing
interest on the explicit incorporation of perceptions, attitudes, habits and social interactions
into the analysis of travel behaviour. However, there are still some comprehensive challenges
in the representation of subjective data and latent constructs such as social networks, well-
being or social capital (Ben-Akiva et al., 2012).
Given this background, the research aims to determine the effects that social capital has
on travel behaviour, applied to a case of mode choice of a new cable car line (TransMiCable),
located at the urban south periphery of Bogotá, Colombia. To establish these outcomes,
discrete choice models incorporating the latent variable “social capital” are estimated. The
modelling framework allows to estimate the probability of choosing the new service compared
to the probability to continue using the typical mode before the cable was inaugurated, by a
sample of individuals living in the influence zone of the stations. The study population might
be denominated as vulnerable, characterized by low-income and poor accessibility (Sarmiento
et al., 2020). We hypothesize that higher social capital levels are associated with a higher
willingness to use the cable car due to an appropriation process of the new infrastructure
produced by a higher sense of community development. The investigation seeks to contribute
to the discussion of the links between social capital and travel behaviour in the Latin-American
context.
After this introduction, section 2 presents a brief review of related research addressing the
links between social capital and transport. Section 3 includes a description of the study area
and the TransMiCable project. Section 4 describes the data used while section 5 discusses
the methodology and model formulation. Section 6 includes the main results and exposes the
implications of the findings. Finally, section 7 provides the conclusions as well as some
recommendations for future research given the limitations of the present study.
2. Exploring the links between social capital and transport
The links that connect social capital with the transport characteristics of a community are
reciprocal. Most of the studies exploring these links have addressed the effects that transport
have on social capital. In summary, there is some consensus that transport improvements
facilitate social interactions and promote social inclusion by the provision of better accessibility
to opportunities (Östh et al., 2018; Stanley et al., 2010). Mobility connects people to meet over
time and space, allowing to sustain social networks Urry (2012). Poor urban zonal connection
inabilities people to access opportunities causing social exclusion due to travel difficulties that
could be mitigated with infrastructure and policy (Church et al., 2000; Lucas, 2012). For
example, transit oriented developments complemented with policies that encourage public
transport usage, walking and cycling could enhance the creation of social networks, trust and
reciprocity, reducing social exclusion and improving the well-being of transport disadvantaged
groups (Currie & Stanley, 2008; Currie et al., 2010; Kamruzzaman et al., 2014; Stanley &
Lucas, 2008; Stanley & Vella-Brodrick, 2009; Utsunomiya, 2016).
Increasing trip making could promote the sense of community (Stanley et al., 2012), while
policies restricting demand (i.e. congestion charges) reduce the number of social trips affecting
social capital (Munford, 2017). Having said that,, sociodemographic and mobility related
characteristics such as car availability, income or trip frequency are determinants of the
differences in social capital levels among individuals (Carrasco & Cid-Aguayo, 2012; Lucas et
al., 2016). However, despite this evidence, social exclusion, transport disadvantage, and social
capital conform an entwined process with complex non-linear causal relationships that require
more quantitative research for a better understanding Schwanen et al. (2015).
In contrast, fewer studies have evaluated the relation in the opposite direction, that is, the
effects that social capital has on travel behaviour. To start, the relationships between socio-
demographics, activity participation and travel behaviour can be captured with structural
equation modelling (SEM) and discrete choice models. Furthermore, the inclusion of activity
participation can provide a better understanding of travel behaviour than through socio-
demographics alone (Lu & Pas, 1999). Using SEM, (Carrasco & Miller, 2006) assessed the
influence of social networks on the propensity to perform social activities. Even though travel
dynamics and social networks are interdependent Sharmeen et al. (2014), the incorporation of
this interactions is useful to describe the social context that determines the travel behaviour.
Discrete choice models have also been applied in diverse contexts to model the social
influence and social interaction on individual decision making (Brock & Durlauf, 2001; Dugundji
& Gulyás, 2008; Maness et al., 2015; Páez et al., 2008).
It is especially relevant the research by Di Ciommo et al. (2014), who explored the influence
of social capital in the choice of a new metro line in Madrid. They incorporated two dummy
variables as social capital proxies to their model: participation in voluntary activities and
receiving help for tasks. Both resulted significant, while the shift to metro was higher for people
helped and lower for people participating in voluntary activities, so different types of social
capital might have different effects on travel behaviour. However, this approach has some
methodological limitations since the proxy variables used to measure social capital do not
respond to a rigorous definition of the concept.
3. The TransMiCable project
TransMiCable is a cable car mass transit system implemented to complement Bogotá’s
Public Transport System in zones where hillside conditions require alternative solutions to
mobility. Currently, the system counts with one line operating in the administrative area of
Ciudad Bolívar, in the south urban fringe of the city, characterized by originally informal
settlements with high poverty levels. There are plans to extend the system with new corridors.
In fact, a second line project in San Cristóbal, a zone with similar conditions to those of Ciudad
Bolívar, is in studies and design phase.
The current study is framed in the influence zone of the cable in Ciudad Bolívar. The cable
started operations on December, 2018, with the objective of improving the quality of life of the
community. The line has a length of 3.4 km, and is composed of four stations. The travel time
of the route is 13.5 minutes, the capacity is around 3,600 passengers per hour and it moves
close to 22,000 passengers in a typical day. Before the cable, an average displacement
between the terminal stations could take up to one hour and the typical modes used for it were
formal bus and informal paratransit services provided by private vehicles. At the base station
Portal el Tunal, shifts between the cable and the bus rapid transit (BRT) system of the city,
known as TransMilenio, can be made without additional charges, that is, the two systems are
fare and physically integrated. Additional to the cable, public investments in the zone for the
project contemplate a complementary urban redevelopment plan with facilities for cultural,
recreational and social activities, community centres and a program to support home
improvements to reduce geomorphological hazards in the zone (Sarmiento et al., 2020).
The south-west peripheries of Bogotá, including Ciudad Bolívar, are considered to be
economic and spatial segregated, with very poor accessibility considering that high income
households and most opportunities are found to the central and north-eastern areas (Arellana
et al., 2020; Guzman et al., 2017; Guzman & Oviedo, 2018). The population in the influence
zone of TransMiCable in Ciudad Bolívar was then denominated as socially vulnerable,
characterized by low income, informal settlements, transport disadvantages and unplanned
urbanization. This situation led people to a social exclusion conditions where they had to trade-
off valuable resources (i.e. time) to respond to their mobility needs (Oviedo & Titheridge, 2016).
The history is of TransMiCable is well represented by the context of Ciudad Bolívar. The
locality was initially informally urbanized with self-built settlements, a high proportion of rural
to urban migrants and victims of forced internal displacement product of the armed conflict and
the waves of violence that have affected the country in the last century (Madrigal & Sánchez,
2014). Furthermore, the official annexation of Ciudad Bolívar to Bogotá in 1984 and the
provision of public services was conditioned to a series of initiatives and confrontations lead
by the local community to stablish some occupancy norms that facilitated the recognition from
the authorities (Forero & Molano, 2015). This sense of collective work to improve the
neighbourhood conditions emerged once again in 2007 when then a group of social leaders,
motivated by the success of the cable car implemented in Medellín (Bocarejo et al., 2014),
started a mobilization to demand a similar solution for Ciudad Bolívar. These efforts in between
different local administration changes were crucial for the approval and funding of the project
that was developed a decade later under a participatory process including community
members and local government representatives from different sectors (Sarmiento et al., 2020).
Given this background, TransMiCable is an interesting case study to evaluate the influence
that social capital has on travel behaviour of vulnerable users.
4. Data
The analysis relies on two components of the baseline surveys collected for the TrUST
study (Sarmiento et al., 2020), before the inauguration of the cable between February and
November 2018. The population target where adults living within 800 m buffer from the cable
stations without plans of moving for at least two years. A general survey was applied to a
sample of 1,031 individuals. One respondent was randomly selected per households
nominated considering a probability proportional to the number of blocks in the study area.
The general questionnaire gathered sociodemographic, mobility, accessibility, health and
social related information, including a series of preferences and attitudes toward some social
capital indicators. Then, among these people a subsample of 343 participants was choose to
answer a stated preference (SP) experiment to assess the willingness to use the cable service.
When data was collected and digitalized, a debugging process resulted in eliminating 25
observations in the general survey and 3 in the SP, so the final sample sizes for the analysis
were 1,006 and 340 respectively.
4.1. Sample
Table 1 presents and description of the samples, including the characteristics of the
individuals interviewed in the general survey and the SP component. Data shows low
education, low income and high unemployment. Around 54% of the households at the moment
of the data collection subsisted with less than one monthly minimum wage (approximately 250
USD in 2019), and 98% with less than two minimum wages, while the mean household size
was 3.8 residents, so per capita income is considerable low. All households care catalogued
in the two lowest socioeconomic strata (SES), a housing classification system (six categories
according to physical characteristics) associated with income level in Colombia (Cantillo-
García et al., 2019), but the lower zones of the cable hill present higher housing prices and
better urban conditions. The set of indicators and attitudinal questions used to measure social
capital is discussed in Section 5.1.
Table 1 Sample description
General survey SP
Sample Size 1006 340
Attribute Proportion
Time living in the house
< 8 years 32% 26%
8-25 years 36% 39%
> 25 years 33% 35%
Age
18-28 25% 21%
28-41 23% 24%
41-58 25% 26%
> 58 28% 29%
Sex Female 65% 74%
Male 35% 26%
Marital status
Single 20% 21%
Married or domestic partner 53% 53%
Divorced, separated or widow 26% 27%
Education level
Primary 39% 44%
Secondary 46% 43%
Higher education 16% 13%
Occupation Studies or works 64% 59%
Non occupied 36% 41%
Vehicle ownership
Motorcycle 16% 14%
Car 7% 6%
Household income
< 1 Monthly minimum wage 54% 56%
> 1 Monthly minimum wage 46% 44%
Socioeconomic strata
SES 1 84% 96%
SES 2 15% 3%
Owns the house where living 42% 42%
Household size < 4 46% 39%
4.2. Stated preferences (SP)
Stated preferences refer to a range of techniques used to gather information about
preferences in hypothetical scenarios designed by a modeller to represent a specific market
scenario. In our case, the SP component corresponds to a choice experiment where each
respondent faced nine situations where they had to choose whether use the new cable car or
their current typical mode, so we obtained 3,060 pseudo-observations. The cable alternative
was described according to travel time savings, additional waiting time, additional walking time
and cost savings in comparison to the typical mode of the respondent. The levels of the
attributes considered in the choice experiment are presented in Table 2. Lastly, the alternative
cable was chosen in 79% of the total pseudo-observations.
Table 2 Attribute levels in stated preference experiment
Choice situation Travel time
savings (min)
Additional waiting time
(min)
Additional walking time
(min)
Cost savings (COP)
1 10 5 2 $0
2 30 8 2 $800
3 20 10 2 $1,500
4 20 5 4 $800
5 10 8 4 $1,500
6 30 10 4 $0
7 30 5 6 $1,500
8 20 8 6 $0
9 10 10 6 $800
Expected effect on the willingness to
use the cable + - - +
5. Model formulation
The methodological approach for the investigation is based on the specification of discrete
choice models where the construct social capital is explicitly accounted as a latent variable.
The discrete choice models are estimated using the SP data, while social capital levels are
measured with the general survey described in section 4. The theoretical framework and
modelling approach are described below.
5.1. Measuring social capital
Social capital is a multidimensional construct that embeds a diverse range of
manifestations of civic engagement, trust and fairness, so focusing on a single component can
generate results that do not correspond to the concept as a whole (Owen & Videras, 2009).
Each social capital dimension represents a different aspect of the concept, so a good measure
should account for as many as possible. Typical dimensions used to calculate social capital
include interpersonal trust, civic engagement social networks and groups membership, as can
be seen in Table 3, where a set of selected empirical studies measuring social capital are
summarized.
We then define social capital as a second order latent construct formed by six first order
dimensions (latent variables), each of which measures or reflects a set of indicators
established from attitudinal questions included in the general survey, based on questions with
dichotomous and Likert answers. Table 4 describes the indicators for each domain, the scale
used and the response proportion for each category. The categories of the indicators were
ordered in such a way that the relation between them and domain level was expected to be
positive. The domains defined intend to measure the following aspects:
• Groups: Measures the tendency of the individual to belong to civic groups.
• Networks: Measures the availability of networks and connections that might help the
individual in the case of an unfortunate event
• Interpersonal trust: Measures the extension of social connections from which it is possible
to rely in the case of an emergency.
• Institutional trust: Measures the level of trust in the government and some civic institutions.
• Cooperation: The perception of individuals regarding the level of cooperation in the
neighbourhood to improve its condition and solve problems.
• Empowerment: Related to the degree of cooperation, autonomy and joint mobilization to
deal with problems and special circumstances in the neighbourhood.
Table 3 Dimensions of social capital measured in selected studies
Authors Study region Social capital dimensions Approach
Brehm & Rahn (1997)
United states Confidence in government, civic engagement, interpersonal trust
SEM
Paldam (2000) - Trust, networks, ease of cooperation Theoretical discussion
Narayan & Cassidy (2001)
Ghana and Uganda
Groups, generalized norms, togetherness, everyday sociability, neighbourhood
connections, volunteerism, trust Factor analysis
(Li et al., 2005) Britain Neighbourhood attachment, social
network, civic participation Generalized latent models
Owen & Videras (2009)
United States Trust, fairness, groups membership Latent class
Congdon (2010) England Social support, trust, groups membership MIMIC
Savage et al. (2013)
Britain Contacts and connections Latent class
Kamruzzaman et al. (2014)
Brisbane Trust, reciprocity, connections with
neighbours Factor analysis
Neves & Fonseca (2015)
Lisbon Bridging, social participation Latent Class
Utsunomiya (2016)
Japan Trust, networks, participation Regression
Munford (2017) London Social trips made Regression
(Östh et al., 2018)
Sweden Community connectivity Regression
The latent variables where modelled via Multiple Indicator Multiple Causes (MIMIC) model
(Bollen, 1989, chapter 7), which describes the latent variable 𝜂𝑖𝑝 as a function of a set of
observed variables 𝑆𝑖 affected by the parameters ∝𝑖𝑝 to estimate, plus an error term 𝜗𝑖𝑝, as
seen in Equation 1 (structural equations). The incorporation 𝑆𝑖, also known as causes, allows
to capture the heterogeneity of the latent variables through the characteristics of the individual
i. At the same time, the latent variable explains a set of attitudinal indicators Cip, through
ordered probit regressions that recognize the ordinal nature of the indicators. In this case each
categorical response k of indicator p is a function of the latent variable, a set of parameters 𝛾𝑖𝑝
to estimate plus an error term 𝜁𝑖𝑝, and a set of thresholds to estimate, as shown in Equation 2
(measurement equations) and Equation 3.
𝜂𝑖𝑝 =∑ ∝𝑖𝑝 𝑆𝑖𝑝 +𝜗𝑖𝑝 (1)
𝐶𝑖𝑝∗ =∑𝛾𝑖𝑝𝜂𝑖𝑝 + 𝜁𝑖𝑝 (2)
𝐶𝑖𝑝 =
{
1 𝑖𝑓 𝐶𝑖𝑝
∗ ≤ 𝜏1
2 𝑖𝑓 𝜏2 < 𝐶𝑖𝑝∗ ≤ 𝜏2
…𝑘 𝑖𝑓 𝜏𝑘−1 < 𝐶𝑖𝑝
∗
(3)
Figure 1 displays the structure of the model conceived to measure social capital. This the
final structure where only the significant causes for each domain are included, but other
observed variables were also evaluated to capture the heterogeneity of the domains. The
behaviour of the first order domains is completely explained by the MIMIC approach. The
relation between the domains and social capital is formative, so we postulate that social capital
is a composite caused by the domains, represented by a linear combination of the first order
latent variables. In order for this specification to be identified, the repeated observations
approach is implemented where the manifest indicators of the domains are reused for the
second order construct (Van Riel et al., 2017), this additional relations were omitted in the
figure for visualization purposes.
Figure 1 Social capital measurement model structure
Table 4 Social capital domains and indicators
Domain Indicator (question) Scale Response Proportions
1 2 3 4 5
Groups
Belongs to a social group or organization 1: No -> 2:Yes 69% 31% - - -
Belongs to a group to improve the infrastructure of the neighbourhood 1: No -> 2:Yes 93% 7% - - -
Belongs to a group to improve the security of the neighbourhood 1: No -> 2:Yes 92% 8% - - -
Belongs to a group to improve the appearance of the neighbourhood 1: No -> 2:Yes 92% 8% - - -
Belongs to a group to improve the provision of public services in the neighbourhood 1: No -> 2:Yes 98% 2% - - -
Belongs to a group to improve the tolerance among young people 1: No -> 2:Yes 98% 2% - - -
Belongs to a group to improve transport services in the neighbourhood 1: No -> 2:Yes 94% 6% - - -
Networks
In case of an unfortunate event, do you think your personal circle would help you? 1: No -> 2:Yes 2% 98% - - -
In case of an unfortunate event, do you think social organizations would help you? 1: No -> 2:Yes 22% 78% - - -
In case of an unfortunate event, do you think government agencies would help you? 1: No -> 2:Yes 56% 44% - - -
In case of an unfortunate event, do you think private agencies would help you? 1: No -> 2:Yes 64% 36% - - -
Interpersonal trust
In the event of an emergency, who can you rely to borrow money 1: Nobody -> 2: Family -> 3: Friends-
Neighbours -> 4: Combination 14% 68% 16% 3% -
In the event of an emergency, who you can rely for help to childcare 1: Nobody -> 2: Family -> 3: Friends-
Neighbours -> 4: Combination 9% 85% 5% 1% -
In the event of an emergency, who you can rely to for temporal lodging 1: Nobody -> 2: Family -> 3: Friends-
Neighbours -> 4: Combination 6% 81% 12% 1% -
Institutional trust
Trust level in the government 1: None -> 5: Very high 77% 16% 4% 2% 1%
Trust level in the police 1: None -> 5: Very high 56% 24% 14% 4% 1%
Trust level in the church 1: None -> 5: Very high 23% 15% 22% 22% 18%
Cooperation
Proportion of the community that cooperates to improve the neighbourhood 1: Nobody -> 5: All 36% 35% 15% 7% 7%
Proportion of the community that cooperates to solve a problem in the neighbourhood
1: Nobody -> 5: All 30% 21% 11% 19% 19%
Empowerment
If there is a problem in the neighbourhood, how do you think the situation will be dealt with?
1: Each person deals with it independently
33% 25% 31% 7% 4% -> 5: The neighbourhood jointly
mobilizes
If some public transport lines are removed, how do you think the situation will be dealt with?
1: Each person deals with it independently
20% 17% 47% 17% - -> 4: The neighbourhood jointly
mobilizes
At what degree do you think that the locality takes into account the problems that you have raised?
1: None -> 5: Very high 47% 31% 15% 4% 4%
Social capital increases →
5.2. Discrete choice models with latent variables
Discrete choice models are grounded in the random utility theory, which states that
consumers take the economic decisions that maximize their benefit or utility, subject to legal,
financial, physical and other constraints (McFadden, 2001). In a discrete choice modelling,
choice maker I faces a set of available alternatives j each of which is associated to a utility Uij
with the structure of Equation 4, where Vij is the systematic or measurable utility, and 𝜀𝑖𝑗 is a
random error component. The systematic utility is generally specified as a linear combination
of a vector of parameters ij to estimate and a set of observed attributes Xij.,plus an alternative
specific constant (ASCij), that represents the net influence of all unobserved or not explicitly
included characteristics of the individual an the alternative that affect the utility function
(Ortúzar & Willumsen, 2011, chapter 7).
𝑈𝑖𝑗 = 𝑉𝑖𝑗 + 𝜀𝑖𝑗 (4)
Depending on the structure of 𝜀𝑖𝑗 different models are derived. For instance, if it is assumed to
be independent and identically Gumbel distributed, the well-known multinomial logit (MNL) is specified,
where the probability of selecting a determined alternative is given by Equation 5, where is a scale
parameter normalised to one for identification purposes (Ortúzar & Willumsen, 2011, chapter 7).
Other common specifications of discrete choice models such as the nested logit and mixed
logit allow to relax some critical assumptions of the MNL such as the independence of
alternatives and homoscedasticity in preferences.
𝑃𝑗𝑞 =exp(𝑉𝑖𝑗)
∑ exp (𝑉𝑖𝑗𝑗 ) (5)
Latent variables can be accounted in discrete choice models through the integrated choice
and latent variable (ICLV) framework, allowing to explicitly model the cognitive process that
underlies the choice process (Vij & Walker, 2016). A ICLV model is confirmed by two
components, a multinomial choice model and a latent variable model. The latter is commonly
modelled with a MIMIC as explained in Section 5.1, allowing to evaluate the direct and indirect
effects of observed and latent variables in the utility function. The latent variable is included in
the utility function as shown in (6), being necessary to estimate the associated parameters ,
resulting the model structure of Figure 2. Even though the notation and indices used suggest
that the utility function depends on the individual i, alternative j, and latent variable p, it is
important to mention that it could also depend on the choice situation. However, this is not the
case for the current analysis.
𝑈𝑖𝑗 =∑ 𝛽𝑖𝑗𝑋𝑖𝑗 +∑ 𝜃𝑖𝑝𝑗𝜂𝑖𝑝𝑗 + 𝜀𝑖𝑗 (6)
Figure 2 ICLV model framework
Given this model structure, the unconditional choice probability is given by (7) (Walker &
Ben-Akiva, 2002), where 𝑦𝑖𝑗 is the vector of observed choices, f(*) is the density function of
the indicators used to measure the latent variables and g(*) is de density function of the latent
variables. If the choice component is specified as a MNL, then the term
Pr(𝑦𝑖𝑗|𝑋𝑖𝑗 , 𝜂𝑖𝑝𝑗 , 𝛽𝑖𝑗 , 𝜃𝑖𝑝𝑗) follows the structure of (5). This functional form can be estimated
using simulated maximum likelihood methods (Train, 2009), allowing for the simultaneous
estimation of all ICLV components.
Pr(𝑦𝑖𝑗|𝑉𝑖𝑗) = ∫ Pr(𝑦𝑖𝑗|𝑋𝑖𝑗 , 𝜂𝑖𝑝𝑗 , 𝛽𝑖𝑗 , 𝜃𝑖𝑝𝑗) 𝑓(𝐶𝑖𝑝𝑗|𝜂𝑖𝑝𝑗 , 𝛾𝑖𝑝𝑗) 𝑔(𝜂𝑖𝑝𝑗|𝑆𝑖𝑝𝑗 , ∝𝑖𝑝𝑗) 𝑑𝜂𝑖𝑝𝑗𝜂𝑖𝑝𝑗
(7)
6. Results and discussion
6.1. Modelling approach
Initially, we validated the latent variable component of the ICLV framework. This required
to test the conceptual structure proposed in Figure 1. For the purpose, the social capital
measurement model was assessed with variance based SEM using individual data from the
general survey, allowing to empirically test the causal relations defined, as well as the
consistency, reliability, validity and goodness of fit of the structure (Van Riel et al., 2017). The
hierarchical structure was also assessed following the paradigm by Koufteros et al. (2009).
Then, the ICLV modelling framework was implemented to evaluate the effects of social
capital on the willingness to use TransMiCable. After testing different specifications for the
utility function of the choice component (estimated with the SP data of stated choices), and
debugging non-significant observed causes of the social capital domains, two final models (M1
and M2) were selected. In M1, the latent construct social capital is introduced in the utility
function interacting with the ASC (Figure 3), capturing the mean effect on the probability of
choosing the cable. In M2, social capital interacts with the cost savings and travel time savings,
determining the effect that the construct has on the valuation of travel time and cost (Figure
4). In both cases the latent variable component of social capital is measured following the
model conceptualized in Section 5.1 and particularly, in Figure 1.
Figure 3 ICLV M1model structure
Figure 4 ICLV M2 model structure
The systematic utility function for the alternative TransMiCable (TMC) for models M1 and
M2 is presented in (8) and (9) respectively, including social capital (SC), travel time savings
(TT), additional waiting time (WTT), additional walking time (WKT) and cost savings (C). Since
the choice exercise consists of two alternatives, and the attributes of the cable are referenced
to the current typical mode, the systematic utility for the later was normalized to zero. The
attribute cost was transformed to hundreds of COP. To conclude, the two components of each
ICLV model were estimated simultaneously using the Apollo package (Hess & Palma, 2019),
available in the R software.
𝑈𝑇𝑀𝐶,𝑀1 = 𝐴𝑆𝐶𝑇𝑀𝐶 + 𝜃𝑆𝐶 ∗ 𝑆𝐶 + 𝛽𝑇𝑇 ∗ 𝑇𝑇 + 𝛽𝑊𝑇𝑇 ∗ 𝑊𝑇𝑇 + 𝛽𝑊𝐾𝑇 ∗ 𝑊𝐾𝑇 + 𝛽𝐶 ∗ 𝐶 (8)
𝑈𝑇𝑀𝐶,𝑀2 = 𝐴𝑆𝐶𝑇𝑀𝐶 + (𝛽𝑇𝑇 + 𝜃𝑇𝑇,𝑆𝐶 ∗ 𝑆𝐶) ∗ 𝑇𝑇 + 𝛽𝑊𝑇𝑇 ∗ 𝑊𝑇𝑇 + 𝛽𝑊𝐾𝑇 ∗ 𝑊𝐾𝑇+(𝛽𝐶
+ 𝜃𝐶,𝑆𝐶 ∗ 𝑆𝐶) ∗ 𝐶 (9)
6.2. Modelling results
Table 5 presents the estimation results for the structural equations of the latent variable
component for each ICLV. Measurement equations (2) and respective thresholds (3), are not
presented for the sake of the simplicity of the manuscript, but all parameters were positive and
statistically significant different than zero at the 95% confidence, so the positive relation
between indicators and domains was confirmed.
Table 5 Estimation results latent variable component ICLV models
Domain / Cause / M1 M2
construct domain Estimate t-test Pr(>|t|) Estimate t-test Pr(>|t|)
Groups
Time in house < 8 years -0.39 -4.84 0.00 -0.55 -9.82 0.00
Owns house 0.28 4.31 0.00 0.27 5.27 0.00
Higher education 0.35 4.57 0.00 0.44 6.95 0.00
Networks
Female 0.19 2.23 0.03 0.13 1.47 0.14
Age 18-28 -0.56 -5.30 0.00 -0.51 -4.69 0.00
Age 28-41 -0.38 -3.48 0.00 -0.30 -2.78 0.01
Interpersonal trust
Female 0.36 2.91 0.00 0.34 2.77 0.01
Higher education 0.47 2.92 0.00 0.52 3.24 0.00
Unemployed -0.36 -2.92 0.00 -0.35 -2.92 0.00
Owns motorcycle 0.20 1.28 0.20 0.22 1.27 0.20
Institutional trust
Household size < 4 0.18 2.89 0.00 0.17 3.38 0.00
Age 18-28 -0.65 -6.99 0.00 -0.52 -5.89 0.00
Age 28-41 -0.69 -7.37 0.00 -0.56 -7.67 0.00
Age 41-58 -0.45 -5.40 0.00 -0.43 -5.96 0.00
Cooperation
Female 0.15 1.73 0.08 0.08 0.88 0.38
Age 18-28 -0.29 -2.79 0.01 -0.26 -2.47 0.01
Divorced, separated -0.15 -1.69 0.09 -0.16 -1.91 0.06
Empowerment
Female -0.33 -3.02 0.00 -0.38 -3.68 0.00
Age 41-58 0.34 2.89 0.00 0.28 2.49 0.01
Zone 1 -0.58 -5.67 0.00 -0.48 -5.11 0.00
Social capital
Groups 12.18 2.28 0.02 9.19 3.13 0.00
Networks 6.67 2.30 0.02 4.37 3.09 0.00
Interpersonal trust 2.23 1.99 0.05 1.49 3.01 0.00
Institutional trust 12.43 2.28 0.02 10.92 3.11 0.00
Cooperation 7.53 2.27 0.02 5.73 3.47 0.00
Empowerment 4.92 2.22 0.03 4.67 3.01 0.00
There are some small differences in the structural estimates of M1 and M2, because the
calibration of the combined probability function (7), given the differences in the structure of the
utility, leads to different solutions. However, the relations and values in both models are similar.
The heterogeneity of the domains of social capital is mainly captured by the individual’s age,
education and time living in the neighbourhood. If the total effect of this observed variables is
calculated, one can determine that individual stocks of social capital are significantly lower for
persons younger than 41 years, separated, divorced and for those living in the neighbourhood
for less than 8 years. In contrast, social capital levels are higher for people with superior
education, females, households that own a house and those with less than 4 residents.
Surprisingly, social capital levels are not sensitive to the income, private vehicle ownership and
economic occupation, which might be explained by the uniformity of the sample in these
aspects, given that most of the respondents could be categorized in low-income population
segments. Also, as expected, all the domains have a positive significant effect on the
measurement of social capital.
It is curious that residents of the lower zones of the cable hillside (zone 1), that is, closer to
the interchange base station Portal el Tunal, seem to have lower level of the domain
empowerment, even though the average income and urban physical conditions are relatively
better. This situation might be related to the sense of community and cooperation developed
in the upper zones, which was demonstrated in que civic participation and community
leadership that was crucial for the development of the TransMiCable project, considering that
these were the most affected zones by the low accessibility and transport disadvantages that
characterized the study area before the cable. If this is correct, it demonstrates that individual
social capital might develop even in adverse situations, where the solution of collective
solutions requires social organization.
Table 6 Estimation results choice component ICLV models
Parameter Description M1 M2
Estimate t-test Pr(>|t|) Estimate t-test Pr(>|t|)
𝐴𝑆𝐶𝑇𝑀𝐶 ASC: TransMiCable 6.15 7.08 0.00 0.89 2.69 0.01
𝜃𝑆𝐶 Interaction social capital with ASC
0.41 2.25 0.02 - - -
𝛽𝑇𝑇 Travel time savings 0.39 4.08 0.00 3.31 7.14 0.00
𝜃𝑇𝑇,𝑆𝐶 Interaction social capital with travel time savings
- - - 0.25 2.96 0.00
𝛽𝑊𝑇𝑇 Waiting time -1.39 -3.61 0.00 -2.16 -5.11 0.00
𝛽𝑊𝑇𝑇 Walking time -0.59 -1.24 0.22 -0.49 -1.01 0.31
𝛽𝐶 Cost savings 1.03 7.66 0.00 4.44 4.95 0.00
𝜃𝐶,𝑆𝐶 Interaction social capital with cost savings
- - - 0.28 2.47 0.01
Log-Likelihood (whole ICLV model) -31,734.6 -31,743.5
Log-Likelihood (choice component -842.0 -853.8
Rho2 (choice component) 0.603 0.597
Likewise, Table 6 presents the estimation results for the parameters of the choice
components. Most parameters were significant at the 95% confidence level, except for the
parameter of walking time, which seems to be of little relevance. The ASC for TransMiCable
is significantly higher than zero, suggesting a higher preference for the cable – ceteris paribus
– over the current typical model of the individual. This might evidence some policy bias in the
responses of SP experiment, given that the cable is a project that generates a lot of expectative
and is desired by the community. The probability of using TransMiCable is negatively affected
by the additional walking and waiting times.
M1 model indicates that higher stocks of social capital are associated with a higher
probability of choosing the new cable car line. This is according to out hypothesis that the use
of the cable might related with an appropriation process of the project, which should be more
important for those with a more solid sense of community. In this line, dense networks of
interaction broaden the participants’ sense of self, developing the sense of community, and
enhancing the participants taste for collective benefits. In fact, historical analysis suggests that
networks of organized reciprocity an civic solidarity, that is, high social capital levels, are a
precondition for socioeconomic modernization, instead of being an epiphenomenon for it
(Putnam, 2000). Given this theory and the results of the present study, the evaluation of social
capital is relevant for the evaluation of transport related investments, as the case of
TransMiCable. The success of this kind of projects and demand forecasts might be related to
the social capital levels of the individuals interacting in the influence zone, especially when
vulnerable communities are in place. On the other hand, M2 results suggest that individuals
with higher social capital have a higher valuation of the travel time and cost savings, so the
benefits of the project are perceived differently.
Finally, if the goodness of fit of the choice components in the models M1 and M2 is
compared with the correspondent reduced model (Log-Likelihood -1,525; Rho2 0.287), it can
be concluded that the introduction of the concept social capital improves the forecasting
capabilities of the mode choice model. The reduced model refers to the case where all the
causal variables for social capital and its domains are explicitly included in the utility function
and no latent construct is considered. Therefore, the latent variable seems to successfully
account for the relations between the characteristics and attitudes that determine a large part
of the mode choice process of the individuals.
7. Conclusion
The research aimed to evaluate the influence that social capital has on the willingness to
use a new cable car line in Bogotá. To do this, integrated choice and latent variable models
were specified, where social capital was explicitly incorporated in the utility function of the
alternative as a second order latent variable formed by six first order domains: groups,
networks, interpersonal trust, institutional trust, cooperation and empowerment. Major findings
could be summarized into three points. First, higher social capital levels are associated with a
greater willingness to use the new cable. Second, the valuation of travel time and cost savings
increases with social capital levels. Third, differences in individual social capital stock might be
explained by sociodemographic attributes, principally age, education level and time living in
the neighbourhood.
Furthermore, the incorporation of the latent construct social capital significantly increases
the predictive capabilities of the mode choice model. This is in line with theory and findings
established in related studies, where it is hypothesized that accounting for social interactions,
social influence, attitudes and perceptions could help to improve the understanding of the
complex relations that make up the economic behaviour of individuals (i.e. the process of mode
choice) (Dugundji et al., 2008, 2011).
Stocks of social capital might be a factor to use for decision-makers in the definition of risks,
demand forecast or financial evaluation of transport investments. However, the notion that the
success of a project is linked to higher levels of social capital could be used as a focalization
tool with great care, since it could transform into a perverse incentive that prioritizes public
investments in zones with better social conditions, broadening inequality gaps in Global South
countries characterized by high levels of poverty and vulnerability. Instead, we encourage
decision and policy makers to consider that demand of transport infrastructure and services,
especially public transport related, is conditioned to the characteristics of the social interactions
of the community, so it is a factor that should be considered in viability assessments and
operation phases in order to guarantee a higher rate of success of the outcomes proposed.
The study presents a series of limitations that must be addressed in future research. To
start, the study population belongs to a very specific context, so evaluations for more general
samples are required for the evaluation of the implications in different population segments
and circumstances. Also, the hypothetical nature of the stated choice experiment might
produce bias results, so further model specifications with mixed revealed and stated
preferences is prudent. In particular, the expectations of the community toward TransMiCable
could induce some policy bias in the response of the survey, so further exercises incorporating
a wider range of public and private mode alternatives is necessary. Ultimately, MIMIC based
ICLV framework assumes the latent variable as continuous, so the specification of discrete
choice models with latent classes could permit to evaluate different types or classes of social
capital, rather than a unidimensional measure.
References
Arellana, J., Oviedo, D., Guzman, L. A., & Alvarez, V. (2020). Urban transport planning and access inequalities: A tale of two Colombian cities. Research in Transportation Business and Management, September, 100554. https://doi.org/10.1016/j.rtbm.2020.100554
Ben-Akiva, M., de Palma, A., McFadden, D., Abou-Zeid, M., Chiappori, P. A., de Lapparent, M., Durlauf, S. N., Fosgerau, M., Fukuda, D., Hess, S., Manski, C., Pakes, A., Picard, N., & Walker, J. (2012). Process and context in choice models. Marketing Letters, 23(2), 439–456. https://doi.org/10.1007/s11002-012-9180-7
Bocarejo, J. P., Portilla, I. J., Velásquez, J. M., Cruz, M. N., Peña, A., & Oviedo, D. R. (2014). An innovative transit system and its impact on low income users: The case of the Metrocable in Medellín. Journal of Transport Geography, 39, 49–61. https://doi.org/10.1016/j.jtrangeo.2014.06.018
Bollen, K. (1989). Structural equations with latent variables. Wiley-Interscience. Bourdieu, P. (1986). The forms of social capital. Handbook of Theory and Research for the Sociology
of Education (Greenwood, Westport, CT) pp 241–258 Brehm, J., & Rahn, W. (1997). Individual-Level Evidence for the Causes and Consequences of Social
Capital. American Journal of Political Science, 41(3), 999–1023. Brock, W. A., & Durlauf, S. N. (2001). Discrete choice with social interactions. Review of Economic
Studies, 68(2), 235–260. https://doi.org/10.1111/1467-937X.00168 Cantillo-García, V., Guzman, L. A., & Arellana, J. (2019). Socioeconomic strata as proxy variable for
household income in transportation research. Evaluation for Bogotá, Medellín, Cali and Barranquilla. DYNA (Colombia), 86(211), 258–267. https://doi.org/10.15446/dyna.v86n211.81821
Carrasco, J. A., & Cid-Aguayo, B. (2012). Network capital, social networks, and travel: An empirical illustration from Concepción, Chile. Environment and Planning A, 44(5), 1066–1084. https://doi.org/10.1068/a43222
Carrasco, J. A., & Miller, E. J. (2006). Exploring the propensity to perform social activities: A social network approach. Transportation, 33(5), 463–480. https://doi.org/10.1007/s11116-006-8074-z
Church, A., Frost, M., & Sullivan, K. (2000). Transport and social exclusion in London. Transport Policy, 7, 195–205.
Congdon, P. (2010). A multiple indicator, multiple cause method for representing social capital with an application to psychological distress. Journal of Geographical Systems, 12(1), 1–23. https://doi.org/10.1007/s10109-009-0097-5
Currie, Graham, & Stanley, J. (2008). Investigating links between social capital and public transport. Transport Reviews, 28(4), 529–547. https://doi.org/10.1080/01441640701817197
Currie, Grahama, Richardson, T., Smyth, P., Vella-Brodrick, D., Hine, J., Lucas, K., Stanley, J., Morris, J., Kinnear, R., & Stanley, J. (2010). Investigating links between transport disadvantage, social exclusion and well-being in Melbourne - Updated results. Research in Transportation Economics, 29(1), 287–295. https://doi.org/10.1016/j.retrec.2010.07.036
Di Ciommo, F., Comendador, J., López-Lambas, M. E., Cherchi, E., & Ortúzar, J. de D. (2014). Exploring the role of social capital influence variables on travel behaviour. Transportation Research Part A: Policy and Practice, 68, 46–55. https://doi.org/10.1016/j.tra.2014.08.018
Dugundji, E. R., & Gulyás, L. (2008). Sociodynamic discrete choice on networks in space: Impacts of agent heterogeneity on emergent outcomes. Environment and Planning B: Planning and Design, 35(6), 1028–1054. https://doi.org/10.1068/b33021t
Dugundji, E. R., Páez, A., & Arentze, T. (2008). Social networks, choices, mobility, and travel. Environment and Planning B: Planning and Design, 35(6), 956–960. https://doi.org/10.1068/b3506ged
Dugundji, E. R., Páez, A., Arentze, T. A., Walker, J. L., Carrasco, J. A., Marchal, F., & Nakanishi, H. (2011). Transportation and social interactions. Transportation Research Part A: Policy and Practice, 45(4), 239–247. https://doi.org/10.1016/j.tra.2011.01.001
Forero, J., , Molano, F. (2015). El paro cívico de octubre de 1993 en Ciudad Bolívar (Bogotá): la formación de un campo de protesta urbana. Anuario Colombiano de Historia Social y de la Cultura, 42(1), 115-143. https://dx.doi.org/10.15446/achsc.v42n1.51347
Guzman, L. A., & Oviedo, D. (2018). Accessibility, affordability and equity: Assessing ‘pro-poor’ public transport subsidies in Bogotá. Transport Policy, 68(June 2017), 37–51. https://doi.org/10.1016/j.tranpol.2018.04.012
Guzman, L. A., Oviedo, D., & Rivera, C. (2017). Assessing equity in transport accessibility to work and study: The Bogotá region. Journal of Transport Geography, 58, 236–246. https://doi.org/10.1016/j.jtrangeo.2016.12.016
Hess, S., & Palma, D. (2019). Apollo: A flexible, powerful and customisable freeware package for choice model estimation and application. Journal of Choice Modelling, 32, 1–43. https://doi.org/10.1016/j.jocm.2019.100170
Kamruzzaman, M., Wood, L., Hine, J., Currie, G., Giles-Corti, B., & Turrell, G. (2014). Patterns of social capital associated with transit oriented development. Journal of Transport Geography, 35, 144–155. https://doi.org/10.1016/j.jtrangeo.2014.02.003
Koufteros, X., Babbar, S., & Kaighobadi, M. (2009). A paradigm for examining second-order factor models employing structural equation modeling. International Journal of Production Economics, 120(2), 633–652. https://doi.org/10.1016/j.ijpe.2009.04.010
Li, Y., Pickles, A., & Savage, M. (2005). Social capital and social trust in Britain. European Sociological Review, 21(2), 109–123. https://doi.org/10.1093/esr/jci007
Lu, X., & Pas, E. I. (1999). Socio-demographics, activity participation and travel behavior. Transportation Research Part A: Policy and Practice, 33A(1), 1–18. https://doi.org/10.1016/s0965-8564(98)00020-2
Lucas, K. (2012). Transport and social exclusion: Where are we now? Transport Policy, 20, 105–113. https://doi.org/10.1016/j.tranpol.2012.01.013
Lucas, K., Bates, J., Moore, J., & Carrasco, J. A. (2016). Modelling the relationship between travel behaviours and social disadvantage. Transportation Research Part A: Policy and Practice, 85, 157–173. https://doi.org/10.1016/j.tra.2016.01.008
Maness, M., Cirillo, C., & Dugundji, E. R. (2015). Generalized behavioral framework for choice models of social influence: Behavioral and data concerns in travel behavior. Journal of Transport Geography, 46, 137–150. https://doi.org/10.1016/j.jtrangeo.2015.06.005
Madrigal, A., & Sánchez, Y. (2014). Las memorias del conflicto armado y la violencia en Colombia: Ciudad Bolívar como referente de mantenimiento de memoria colectiva significativa en Bogotá. Ciudad Paz-Ando, 5(2), 71-86. https://doi.org/10.14483/2422278X.5356
McFadden, D. (2001). Economic Choices. American Economic Review, 91(3), 351–378. https://pubs.aeaweb.org/doi/pdf/10.1257/aer.91.3.351%0Ahttp://elsa.berkeley.edu/
Munford, L. A. (2017). The impact of congestion charging on social capital. Transportation Research Part A: Policy and Practice, 97, 192–208. https://doi.org/10.1016/j.tra.2017.01.018
Narayan, D., & Cassidy, M. F. (2001). A Dimensional Approach To Measuring Social Capital: Development And Validation Of A Social Capital Inventory. Current Sociology, 49(2), 59–102. https://doi.org/10.1177/0011392101049002006
Neves, B. B., & Fonseca, J. R. S. (2015). Latent class models in action: Bridging social capital & internet usage. Social Science Research, 50, 15–30. https://doi.org/10.1016/j.ssresearch.2014.11.002
Ortúzar, J. de D., & Willumsen, L. G. (2011). Modelling Transport (4th Editio). Wiley. Östh, J., Dolciotti, M., Reggiani, A., & Nijkamp, P. (2018). Social Capital, Resilience and Accessibility in
Urban Systems: a Study on Sweden. Networks and Spatial Economics, 18(2), 313–336. https://doi.org/10.1007/s11067-017-9375-9
Oviedo Hernandez, D., & Titheridge, H. (2016). Mobilities of the periphery: Informality, access and social exclusion in the urban fringe in Colombia. Journal of Transport Geography, 55, 152–164. https://doi.org/10.1016/j.jtrangeo.2015.12.004
Owen, A. L., & Videras, J. (2009). Reconsidering social capital: A latent class approach. Empirical Economics, 37(3), 555–582. https://doi.org/10.1007/s00181-008-0246-6
Páez, A., Scott, D. M., & Volz, E. (2008). A discrete-choice approach to modeling social influence on individual decision making. Environment and Planning B: Planning and Design, 35(6), 1055–1069. https://doi.org/10.1068/b3320t
Paldam, M. (2000). Social capital: One or many? Definition and measurement. Journal of Economic Surveys, 14(5), 629–653. https://doi.org/10.1111/1467-6419.00127
Putnam, R. D. (2000). Bowling Alone: America’s Declining Social Capital. Culture and Politics, 6(1), 223–234. https://doi.org/10.1007/978-1-349-62965-7_12
Robison, L. J., Shupp, R. S., Jin, S., Siles, M. E., & Ferrarini, T. H. (2012). The relative importance of selfishness and social capital motives. Journal of Socio-Economics, 41(1), 118–127. https://doi.org/10.1016/j.socec.2011.10.008
Sarmiento, O. L., Higuera-Mendieta, D., Wilches-Mogollón, M. A., Guzman, L. A., Rodriguez, D. A., Morales, R., Méndez, D., Bedoya, C., Linares-Vásquez, M., Arévalo, M. I., Martínez-Herrera, E., Montes, F., Meisel, J. D., Useche, A. F., García, E., Triana, C. A., Medaglia, A. L., Hessel, P., Arellana, J., … Diez Roux, A. V. (2020). Urban Transformations and Health: Methods for TrUST--a Natural Experiment Evaluating the Impacts of a Mass Transit Cable Car in Bogotá, Colombia. Frontiers in Public Health, 8, 64. https://doi.org/10.3389/fpubh.2020.00064
Savage, M., Devine, F., Cunningham, N., Taylor, M., Li, Y., Hjellbrekke, J., Le Roux, B., Friedman, S., & Miles, A. (2013). A New Model of Social Class? Findings from the BBC’s Great British Class Survey Experiment. Sociology, 47(2), 219–250. https://doi.org/10.1177/0038038513481128
Schwanen, T., Lucas, K., Akyelken, N., Cisternas Solsona, D., Carrasco, J. A., & Neutens, T. (2015). Rethinking the links between social exclusion and transport disadvantage through the lens of social capital. Transportation Research Part A: Policy and Practice, 74, 123–135. https://doi.org/10.1016/j.tra.2015.02.012
Sharmeen, F., Arentze, T., & Timmermans, H. (2014). An analysis of the dynamics of activity and travel needs in response to social network evolution and life-cycle events: A structural equation model. Transportation Research Part A: Policy and Practice, 59, 159–171.
https://doi.org/10.1016/j.tra.2013.11.006 Stanley, Janet, & Lucas, K. (2008). Social exclusion: What can public transport offer? Research in
Transportation Economics, 22(1), 36–40. https://doi.org/10.1016/j.retrec.2008.05.009 Stanley, Janet, Stanley, J., Vella-Brodrick, D., & Currie, G. (2010). The place of transport in facilitating
social inclusion via the mediating influence of social capital. Research in Transportation Economics, 29(1), 280–286. https://doi.org/10.1016/j.retrec.2010.07.035
Stanley, Janet, & Vella-Brodrick, D. (2009). The usefulness of social exclusion to inform social policy in transport. Transport Policy, 16(3), 90–96. https://doi.org/10.1016/j.tranpol.2009.02.003
Stanley, John, Stanley, J., & Hensher, D. (2012). Mobility, Social Capital and Sense of Community: What Value? Urban Studies, 49(16), 3595–3609. https://doi.org/10.1177/0042098012447002
Train, K. E. (2009). Discrete Choice Methods with Simulation. In Discrete Choice Methods with Simulation, Second Edition. Cambridge University Press. https://doi.org/10.1017/CBO9780511805271
Urry, J. (2012). Social networks, mobile lives and social inequalities. Journal of Transport Geography, 21, 24–30. https://doi.org/10.1016/j.jtrangeo.2011.10.003
Utsunomiya, K. (2016). Social capital and local public transportation in Japan. Research in Transportation Economics, 59, 434–440. https://doi.org/10.1016/j.retrec.2016.02.001
Van Riel, A. C. R., Henseler, J., Kemény, I., & Sasovova, Z. (2017). Estimating hierarchical constructs using consistent partial least squares: The case of second-order composites of common factors. Industrial Management and Data Systems, 117(3), 459–477. https://doi.org/10.1108/IMDS-07-2016-0286
Vij, A., & Walker, J. L. (2016). How, when and why integrated choice and latent variable models are latently useful. Transportation Research Part B: Methodological, 90, 192–217. https://doi.org/10.1016/j.trb.2016.04.021
Walker, J., & Ben-Akiva, M. (2002). Generalized random utility model. Mathematical Social Sciences, 43(3), 303–343. https://doi.org/10.1016/S0165-4896(02)00023-9
