Вы находитесь на странице: 1из 12

Journal of Choice Modelling 38 (2021) 100270

Contents lists available at ScienceDirect

Journal of Choice Modelling


journal homepage: http://www.elsevier.com/locate/jocm

Mapping potentials and challenges of choice modelling for social


science research
Ulf Liebe a, *, Jürgen Meyerhoff b
a
Department of Sociology, University of Warwick, Coventry, United Kingdom
b
Institute for Landscape Architecture and Environmental Planning, Technische Universität Berlin, Berlin, Germany

A R T I C L E I N F O A B S T R A C T

Keywords: This paper argues that choice modelling is a useful approach for all social sciences, while at the
Causal analysis same time disciplines such as sociology and political science can contribute significantly to the
Context effects future development of choice modelling. So far choice modelling has mainly been applied in
Decision rules
disciplines that investigate types of consumption choices, be it marketing to investigate prefer­
Machine learning
Network analysis
ences for new products, transportation to analyse mode choices, or environmental economics to
Rational choice elicit preferences for public goods. However, using the information that can be gained from in­
Social sciences dividual choices among mutually exclusive alternatives has gained increasing popularity in other
disciplines as a powerful tool to test theoretical hypotheses and generate insights into individual
behaviour. Examples are the acceptance of refugee shelters in peoples’ neighbourhood, the choice
of where to commit a crime or the evolution of social networks. A good point of departure for an
expansion of choice modelling within the social sciences is the common foundation that many
disciplines share that are gathered under the umbrella of social sciences. Research traditions and
theoretical models include rational choice concepts, and choice modelling can be linked to cross-
cutting methods, including agent-based models, network analysis, and machine learning. At the
same time, disciplines can complement each other in studying choice behaviour, as they can
contribute concepts and tools less familiar to the other disciplines. Finally, all social science
disciplines face challenges when it comes to issues such as causal analysis, heterogeneity in de­
cision rules, joint decision making, or big data. Choice modelling and a cross-disciplinary dia­
logue can contribute to meeting these challenges.

1. Choice as being fundamental to social science

Choices are a ubiquitous part of everyday life and thus are the subject to scientific analysis in many disciplines gathered under the
umbrella term social sciences. Until today, however, their investigation has mainly been a topic in disciplines that deal with consumer
decision making. What type of products do consumers prefer? How much are they willing to pay for new products or product attri­
butes? What kind of transportation mode do people favour for commuting? How do they value travel time in different means of
transportation? What are the citizens’ preferences for public goods in an environmental context? Choice modelling has been applied
more and more often during the last two decades to answer such questions in areas such as marketing, transportation, health eco­
nomics or environmental economics.

* Corresponding author.
E-mail addresses: ulf.liebe@warwick.ac.uk (U. Liebe), juergen.meyerhoff@tu-berlin.de (J. Meyerhoff).

https://doi.org/10.1016/j.jocm.2021.100270
Received 1 August 2020; Received in revised form 16 November 2020; Accepted 6 January 2021
Available online 20 January 2021
1755-5345/© 2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
U. Liebe and J. Meyerhoff Journal of Choice Modelling 38 (2021) 100270

By pointing out the common foundations of the social sciences, presenting a broad range of applications of choice modelling in
social science research and by identifying opportunities for cross-disciplinary research, this paper provides a broader context for the
Special Issue “Choice Modelling in Social Science Research”. We address readers from the social sciences who are familiar with or
interested in choice modelling as well as choice modellers who are interested in integrating techniques and findings from social science
research to enrich their research. Furthermore, we try to identify areas where social scientists and choice modellers could collaborate.
The analysis of discrete choices, for example, related to voter participation (Rivers, 1988) and home-leaving (Aquilino, 1991), has
always been a part of social science disciplines such as sociology and political science (see Manski, 1981: 60 for further examples).
Recently, however, choice modelling – advanced discrete choice models – have gained increasing popularity in these disciplines (see
Glasgow, 2011 for political science). Examples of applications are the acceptance of refugee shelters close to peoples’ place of resi­
dence (Liebe et al., 2018), the choice of where to commit a crime (Bernasco and Block, 2009) or the evolution of social networks (see
Snijders, 1996; Pink et al., 2020). A reason why choice modelling is attractive for many social science disciplines is given by Bernasco
and Block (2009: 123), who apply it in criminology, and write that “[…] because the [choice] model is based on the explicit
formulation of choice criteria that individual offenders are expected to use, the link among theory, model formulation, and hypothesis
testing is short.” Therefore, compared to other approaches in social science research, in which theoretical assumptions are loosely
related to the statistical analysis, choice modelling provides a framework in which the statistical model is closely related to theoretical
assumptions (Bruch and Feinberg, 2017).
On the other hand, to enhance the behavioural realism of choice models, choice modelling research has incorporated concepts of
social science and psychological research, for example, related to moral decision making (Chorus, 2014). Against this background, this
paper wants to bring attention to the fact that choice modelling is a useful approach for social sciences in general. At the same time,
choice modelling can benefit from a broad involvement of all social science disciplines into its future development. Starting point for
this argument is that there are at least three features that build a common foundation.
The first feature refers to the fact that in social science the explanation of decisions is often of interest: consumers choose products;
individuals’ choose occupations; adolescents choose friends; offenders choose offence locations; voters choose parties; families choose
residential locations; and so on. Social science researchers aim at better understanding and explaining the corresponding decision-
making processes as well as the choice outcomes. In this regard, many studies still employ standard survey research, including sim­
ple questions on self-reported behaviour as well as survey items on the importance of crucial behavioural determinants (Kroneberg and
Kalter, 2012). Yet, such survey items make it difficult to disentangle the importance of different choice determinants. It is not un­
common that all aspects of a decision-making situation are perceived as somehow important and behavioural responses can be prone to
socially desirable response behaviour (e.g., Mutz, 2011; Auspurg and Hinz, 2015; Liebig et al., 2015 for factorial surveys). In stated
choice experiments, which are often employed in choice modelling research, respondents repeatedly choose between choice alter­
natives that vary in attribute levels (e.g., Louviere et al., 2000; Auspurg and Hinz, 2015; Liebe et al., 2021). Based on an experimental
design of attribute-level combinations, repeated choices and trade-offs between choice attributes, these experiments offer a way to
experimentally disentangle the importance of decision-making attributes. They could also reduce a social desirability bias as socially
desirable choice options are less obvious (see Auspurg et al., 2015 and Liebig et al., 2015 for factorial surveys).
The second feature is that the many disciplines in the social sciences have a common theoretical background. In each discipline,
there is a tradition of applying some type of rational choice theory to explain outcomes of interest, that is, economic, social, and
political phenomena. Citizens who participate in social movements, voters, politicians but also criminals are often described as rational
actors who in a broad definition of rationality adhere to a specific decision rule (Tutić and Liebe, 2017 for an overview), where utility
maximisation is the still most commonly assumed rule (Kroneberg and Kalter, 2012). Even if the rational choice theory has been
criticised and critically discussed in various contexts, including psychological, sociological and political science research (Coleman and
Fararo, 1992; Green and Shapiro, 1994; Kahneman and Tversky, 2000), it is an established theoretical framework across the social
sciences, and choice modelling can be closely linked to many variants of rational choice theory employed in sociology, political sci­
ence, criminology, etc. (Chorus, 2014; Liebe et al., 2021).
The third feature points to the research methods such as multifactorial survey experiments, agent-based modelling, network
analysis, and machine learning. Especially multifactorial survey experiments are increasingly popular in sociology and political sci­
ence research. As already mentioned above, such experiments have the advantage of uncovering the importance of key attributes, for
example, in the evaluation of social, political and economic situations. They allow for estimating causal effects of situational attributes
and lower socially desirable response behaviour. Factorial surveys are multifactorial experiments, in which respondents evaluate a
description of a situation. They are an established method in sociological research, which has increasingly gained popularity in the last
decades (Wallander, 2009; Auspurg and Hinz, 2015). Discrete choice experiments, in which respondents choose between mutually
exclusive alternatives (Louviere et al., 2000) have been applied less frequently, for example, in sociological and political science
research. Their advantage is, however, that they are based on a close link to action theories (Liebe et al., 2021) and can help to identify
key factors of decision making, the acceptance of policies and heterogeneity in decision rules (Chorus, 2014). Agent-based modelling is
another approach that is applied in several social science disciplines and can be fruitfully combined with choice modelling in order to
study dynamics in decision making and their macro-level outcomes (Bruch and Atwell, 2015; Elsenbroich and Payette, 2020). Social
network analysis is prevalent in many social sciences and can be linked to choice modelling (e.g., Calastri et al., 2018; Feinberg et al.,
2020; Pink et al., 2020). Furthermore, big data offer new possibilities, and the combination of machine learning and choice modelling
will play an essential role in this context (Brathwaite et al., 2017).
Despite the three common features addressed above, there are also noteworthy differences among the disciplines. For example,
many applications of choice modelling are concerned with predicting behaviour and estimating policy effects as well as welfare effects
such as compensation variations and marginal willingness-to-pay values for product or policy attributes. In contrast, in sociology,

2
U. Liebe and J. Meyerhoff Journal of Choice Modelling 38 (2021) 100270

researchers are more interested in examining social context effects on decision making as well as testing sociological theories (see
Bruch and Feinberg, 2017). Therefore, at the analysis stage, sociological applications place a stronger focus on the importance of
theory-oriented main and interaction effects of choice attributes. Compared with foundations of choice experiment research in random
utility theory, stated choice analysis in political science research has been motivated in the context of causal analysis (see the influ­
ential paper by Hainmueller et al., 2014). Their starting point is the potential outcomes framework of causal inference (Neyman, 1923;
Rubin, 1974), and conjoint experiments help to uncover the effects of situational dimensions on behavioural outcomes. For example,
they are related to the acceptance of immigrants (Hainmueller and Hopkins, 2015) and asylum seekers (Bansak et al., 2016), pref­
erences for climate-change agreements (Bechtel and Scheve, 2013), voting behaviour (Kirkland and Coppock, 2018), and media choice
(Mummolo, 2016) (see Leeper et al., 2020 for more examples).
While conjoint experiments are not choice experiments, as they lack the corresponding theoretical foundation (McFadden, 1974;
Louviere et al., 2010), many applications in political science aim/claim to measure preferences using conjoint experiments including
paired choice tasks. Here exists some inconsistency in the use of terms such as stated preferences, conjoint experiments and choice

Table 1
Examples of applications in different social science areas.
Area Example Data type Some insights

Employment Logan (1996): Opportunity and Choice in Socially Survey data on individuals’ occupation, Stability of matching processes over
Structured Labor Markets age, education, race, employer’s time and discrimination towards non-
prestige, etc.; development of a two- whites in managerial occupations for
sided logit model men
Vote choice De Vries (2007): Influence of attitudes towards Individual-level survey data for The effect of EU attitudes on national
European integration on national elections Denmark, Germany, Netherlands, elections depends on the country
United Kingdom; conditional logit context – the salience of EU issues for
model including information about voters and partisan conflicts
party positions
Intergroup social Zeng and Xie (2008): A Survey data on adolescents’ friend Strong preferences for homophily
contacts Preference-Opportunity-Choice Framework with nominations, grades 7 to 12; conditional regarding age, GPA, socio-economic
Applications to Intergroup logit model considering preferences and status, as well as for status asymmetry;
Friendship opportunity structures no effects of opportunity structures
Criminology Bernasco and Block (2009): Crime location choice Revealed preference data on incidents, Spatial choice models show that the
based on Chicago Police Department presence of illegal drug markets
records increases and ethnic segregation in
potential target areas decreases
mobility of offenders
Ethnic and Hainmueller et al. (2014): acceptance of immigrants Stated choice data on pairs of US citizens with positive preferences
migration in the US immigrants who vary in their attributes for immigrants with higher education,
studies higher-status professions, language
skills etc., country of origin less
relevant
Refugee studies Liebe et al. (2018): From welcome culture to Repeated stated choice experiment on Country of origin and religion of
welcome limits? Uncovering preference changes preferences for refugee and migrant refugees are considered less important
over time for sheltering refugees in Germany homes in Germany than decent housing conditions and
whether refugees arrive as families or
single persons. Rather high preference
stability
Residential Bruch and Swait (2019): Choice Set Formation in Panel study of households living in 64 Different decision criteria are
mobility and Residential Mobility and Its Implications for neighbourhoods in Los Angeles County; important at each stage of
segregation Segregation Dynamics development of a choice set formation neighbourhood selection processes;
model; agent-based model choice sets differ between race and
income groups, which affects
segregation outcomes
Social policy Stadelmann-Steffen and Dermont (2019): Citizens’ Conjoint experiment data on paired Higher support of basic income
Opinions About Basic Income Proposals Compared – policy variants for basic income in schemes in Finland than in
A Conjoint Analysis of Finland and Switzerland Finland and Switzerland Switzerland; evaluation of basic
income attributes similar in both
countries; preference for restricted
access to basic income for non-
nationals.
Discrimination Liebe and Beyer (2020): Examining discrimination in Stated choice experiment data on Discrimination towards presumed
everyday life: a stated choice experiment on racism carpooling in Germany; choice “Turkish” drivers, which is driven by
in the sharing economy attributes included price, car type, and Xenophobic attitudes
rating but also to the perceived ethnic
background of the driver
Morality Chorus et al. (2020): Diabolic dilemmas of Stated choice experiment data on Covid- High proportion of respondents with
COVID-19: An empirical study into Dutch society’s 19 policy scenarios; representative aversion against making taboo trade-
trade-offs between health impacts and other effects sample of the Dutch population offs, e.g. between taxes and fatalities;
of the lockdown yet population is willing to sacrifice
certain degree of educational disad-
vantage and income loss in society.

3
U. Liebe and J. Meyerhoff Journal of Choice Modelling 38 (2021) 100270

experiments. Based on the theoretical foundations of causal analysis, in political science applications, it is common practice to
construct the full factorial of attribute-level combinations across alternatives and to select randomly choice sets from this full factorial.
Further, average marginal component effects are the main outcome of interest, and they are estimated by non-parametric models,
which is a clear difference to using parametric choice models.

2. Applications of choice modelling in social science

In the following, we briefly present a selection of studies that demonstrate and exemplify the use of choice modelling in various
contexts across different social sciences, sorting them by publication date (see Table 1, further example studies are mentioned in the
text). These applications cover a wide range of topics, including labour market processes, criminal behaviour, social networks, resi­
dential mobility, discrimination, and morality.

2.1. Employment and labour market processes

Mirroring labour market processes, in which preferences and resources of employers and workers are matched, Logan (1996)
developed a two-sided logit model and applied it to General Social Survey data from the US. The model estimates jointly preferences of
both workers and employers (the conditional logit model is a special case of the two-sided logit model). The data analysis revealed, for
example, the stability of matching processes over time and discrimination towards non-whites in managerial occupations for men.
Also, while the model assumes consistency in choice behaviour, it is not restricted to utility maximisation and hence compatible with,
for example, different decision rules and habits. Another example of choice modelling research on employment is the study by
Humburg and van der Velden (2015). They use stated choice experiments to measure employers’ preferences for curriculum vitae
characteristics and skills when hiring university graduates. In a first experiment, they mirrored the selection processes at the job
interview stage and in a second experiment the hiring stage. The participants from nine European countries evaluated choice sets with
two hypothetical job candidates. The data were analysed using mixed logit models which showed that occupation-specific human
capital (interview stage), as well as interpersonal skills (hiring stage), matter most for employers. Looking at the demand side of the
labour market and testing assumptions of sociological rational choice theory, Möser et al. (2019) used stated choice experiments to
examine preference heterogeneity of Swiss school students regarding apprenticeship positions as well as first employment positions.
Demel et al. (2019) employed choice experiments to elicit the job preferences of students in different European countries. They find
two main differences in students’ preferences, one due to cultural differences and another due to economic differences. Only for
students from the Spanish universities commuting time is essential. Most likely, this is due to the close familial relationships inherent to
the Spanish culture or the lower willingness to change their place of residence. At the same time, they find a stronger preference for a
permanent contract in Spain, a fact that might be driven by the very high unemployment rate in this country.

2.2. Vote choice

De Vries (2007) examines the effects of attitudes towards the EU on national elections in Denmark, Germany, Netherlands, and the
UK. She uses individual-level panel data and employs a conditional logit model on voter choice that takes the differences between a
respondent’s self-placement and a party’s position on EU integration into account. The results show that the relevance of EU issues for
national elections depends on the country context. In a study on candidate preferences, Teele et al. (2018) used a paired conjoint design
to examine US voters’ preferences for political office candidates depending on attributes such as gender, age, number of children,
occupation, and number of years in politics. While at first sight, respondents evaluate women and men candidates similarly, there is a
strong preference for women with a “traditional household profile” (married and having children), indicating double standards which
in turn can help to explain the underrepresentation of women in politics.

2.3. Intergroup social contacts

Zeng and Xie (2008) modify the conditional logit model and propose a discrete choice model analysis to separate effects of
preferences and opportunity structures. They demonstrate this type of analysis by using Add Health survey data on friendship choices
to investigate the effects of homophily and status asymmetry, as well as opportunity structure in adolescents’ friendship nominations.
They find strong support for the former and no effect for the latter. Adolescents are more likely to nominate a friend with similar age,
grade point average (GPA), and socioeconomic status (SES); there is also high racial homophily. While such homophily effects are
stronger in this study, status asymmetry is also supported. It implies that adolescents are more likely to nominate someone with a
higher status (GPA, SES) compared with a lower status, that is, the effect of status distance is more negative if a potential friend has a
lower status. On the other hand, opportunity structures as functions of grade levels (e.g. opportunities to spend time together) do not
affect the effects of educational performance, socioeconomic status and race. Focusing on network-related choice behaviour, Pink et al.
(2020) elaborate on the relationship between the stochastic actor-oriented model for the co-evolution of networks and behaviour
(SAOM) and choice modelling. They apply SAOM to jointly estimate selection and influence effects in friendship networks in German
schools, based on longitudinal data. Their findings indicate, for example, that next to a selection effect regarding academic perfor­
mance, friends adapt to their friends’ academic achievement over time, pointing to an influence effect.

4
U. Liebe and J. Meyerhoff Journal of Choice Modelling 38 (2021) 100270

2.4. Crime location choices

Bernasco and Block (2009) investigate offenders’ crime location choices based on data from the Chicago Police Department. They
apply spatial choice models and test a set of eleven hypotheses derived from theory and the literature. Conditional logit models on the
likelihood to choose a target area reveal some insights confirming research from ethnographic studies, for example, related to gang
territoriality and ethnic segregation as restrictions for offender mobility. Frith (2019) extends applications of conditional logit models
on crime location choices by using mixed logit and latent class choice models to uncover offenders’ preference heterogeneity regarding
offence location choices of serious acquisitive crime offenders in York, UK.

2.5. Neighbourhood choice

Bruch and Swait (2019) analyse residential mobility and racial segregation by developing a choice set formation (CSF) model.
Contrary to standard models in which individuals consider all alternatives and attributes, the CSF model assumes a multiple stages
process in which boundedly rational actors first select a reduced set of choice alternatives and then select the alternative they prefer
most. Based on L.A.FANS panel data of households living in 64 sampled neighbourhoods in Los Angeles County, that is, revealed
neighbourhood preferences data, they find that the CFS models fit the residential mobility data better than a standard conditional logit
model. Choice sets – neighbourhood alternatives considered by individuals – differ between racial and income groups. Blacks and
Hispanics and low-income groups are more likely to consider neighbourhoods with a disproportionate share of their own groups. Using
an agent-based model, Bruch and Swait (2019) further demonstrate that racially stratified choice sets can boost racial segregation. A
similar agent-based model, which also follows the highly influential work by Schelling (1971), has been provided by Bruch and Mare
(2006). They compare (Schelling’s) threshold and continuous preference functions and show that the former leads to higher and the
latter to lower levels of racial segregation. Based on stated preference data – vignettes on neighbourhoods which vary in their race
composition – they find that continuous functions are empirically more plausible than (Schelling’s) threshold functions. Ibraimovic
and Hess (2017) examine individuals’ preferences for neighbourhoods in Lugano, Switzerland, using a stated choice experiment. In
this experiment, respondents repeatedly chose between three hypothetical neighbourhood alternatives the one they prefer most. These
alternatives varied, among others, in the concentration of co-nationals and share of foreigners. Based on latent class choice models,
Ibraimovic and Hess (2017) found that the sample can be separated into three classes differing in their preferences. There exist two
classes with a positive preference for co-nationals and one class which does not significantly differnciate between co-nationals and
foreigners.

2.6. Immigration and refugee studies

Hainmueller et al. (2014) implemented a conjoint experiment of the acceptance of immigrants in the US. Respondents from the
general population were confronted with pairs of immigrants who varied in characteristics such as gender, education, language, origin,
profession, and job experience. Based on a fully randomized design and random selection from all possible attribute-level combinations
across alternatives, they can single out the effects of these attributes and their corresponding levels. Hainmueller et al. (2014) esti­
mated average marginal component effects and found, for example, that immigrants with higher education are strongly preferred over
lower-education immigrants, while cultural differences indicated by country of origin matter less. Sobolewska et al. (2017) use a
conjoint experiment to investigate preferences for successful immigrant integration in the UK and the Netherlands. Based on evalu­
ations of paired immigrant profiles, the study shows that integration is perceived as a multidimensional process including cultural,
social and civic aspects. Dimensions such as religiosity are, in contrast, comparatively unimportant.
Against the background of the so-called refugee crisis in Europe, Liebe et al. (2018) employ a latent class choice model and a 2-state
Markov model to examine preferences for refugee and migrant homes in people’s vicinity as well as preference changes over time.
Based on stated choice experiment data collected in Germany in 2015 and 2016 (test-retest), they find that initially, a majority of the
respondents rather disapproved refugee homes in their vicinity. Their preferences were stable over time. A minority of the respondents,
who were initially rather approving of refugee and migrant homes, were more likely to change their preferences towards a somewhat
disapproving position in 2016. Experience of contact with refugees and migrants, higher education, and general pro-immigration
attitudes explain the acceptance of refugee and migrant homes as well as preference stability over time. Country of origin and reli­
gion of refugees and migrants are considered less critical than decent housing conditions and whether refugee and migrants arrive as
families or single persons. In this regard, the results indicated the importance of humanitarian aspects of sheltering and integration of
refugees and other migrants into society.

2.7. Preferences for (social) policy options

Stadelmann-Steffen and Dermont (2019) conducted a conjoint experiment in which respondents in Finland and Switzerland
evaluated paired policy variants for basic income schemes. Their experiment varied in several attribute levels, for example, related to
the level of a basic income (for adults and children), the degree of universality for non-citizens and the interplay with other welfare
schemes. Compared to a stated choice design, in which respondents choose between alternatives, respondents in this experiment had to
indicate the likelihood that they would approve each of the paired alternatives in a referendum on a scale from 0 to 100 per cent. While
support for a basic income scheme is higher in Finland than in Switzerland, the evaluation of basic income attributes is similar in both
countries. For example, in both countries, respondents prefer a scheme that complements other social benefits instead of replacing

5
U. Liebe and J. Meyerhoff Journal of Choice Modelling 38 (2021) 100270

them. Further, in both country contexts, the acceptance of basic income schemes increases if the access for non-nationals is restricted.
The usefulness of stated choice experiments to study policy preference is also illustrated in Lu et al. (2020), who, in the context of
Brexit, study Britons’ preferences for different options for the relationship between Britain and the EU (i.e. “what sort of Brexit the
British people want”).

2.8. Discrimination

Examining discriminatory preferences in an everyday life context, Liebe and Beyer (2020) applied conditional and random
parameter logit models to stated choice experiment data on carpooling. In this study, German respondents were asked to choose
between various carpooling offers varying not only regarding price, car type, and rating but also to the perceived ethnic background of
the driver. Respondents had a lower likelihood to ride with a person presumed to be “Turkish” and “Italian”, compared to be “German”.
This discrimination is driven by xenophobic attitudes in line with a prejudice-theoretical explanation of discrimination. In another
choice experiment with a small sample of German respondents, Beyer and Liebe (2015) addressed a problem of research on
anti-Semitism, namely to uncover how ‘‘criticism’’ of and ‘‘boycotts’’ against Israel are related to anti-Semitic attitudes. In a stated
choice experiment on the purchase of orange juice in a supermarket, they varied the attributes fruit content, the origin of oranges and
price; origin of oranges referred to South Africa, USA and Israel. The survey also included questions on anti-Semitic attitudes and
perceived primary group norms. Conditional and random parameter logit models, including interactions with respondents’ charac­
teristics, indicate an association between anti-Semitic motives and boycott of Israeli products. Individuals with anti-Semitic friends or
family and individuals who deny Israel the right to defend itself were more likely to not choose the product from Israel.

2.9. Moral choice

In the context of the COVID-19 pandemic, Chorus et al. (2020) conducted a stated choice experiment study to investigate Dutch
citizens’ preferences for different policies to end the lockdown. The data are representative for the Dutch adult population, and the
experiment included several choice attributes such as (direct or indirect) increases in the number of deaths, of people with lasting
physical injuries, and of people with lasting mental injuries due to the coronavirus crisis, as well as an increase in the number of
children with lasting educational disadvantages and of households with a net income loss. The experiment also included a payment
vehicle of a one-off coronavirus tax per household. The results imply for example that the average willingness to sacrifice in order to
avoid one fatality directly or indirectly related to COVID-19 equals 15 cases of lasting mental health problems, 18 children with lasting
educational disadvantage and 77 households with long term decline in net income. A latent class choice analysis indicates preference
heterogeneity in the population and that the willingness to sacrifice cases with health problems and educational disadvantages varies
across educational groups, for instance. Further, a high proportion of respondents has an aversion against making taboo trade-offs (also
see Chorus et al., 2018) between taxes and fatalities, for instance. A strong tendency towards taboo trade-off aversion can also be found
in a choice experiment study on Americans’ preferences towards paying organ donors, in which, for example, many respondents reject
trade-offs between payments and the number of kidney transplants (Elías et al., 2019).

3. Opportunities for cross-disciplinary research

There are a couple of issues that are relevant across social science disciplines and where different disciplines could benefit from
each other. In the following, we briefly discuss (some of) these issues (see Table 2). We devote more space to the issue of taste het­
erogeneity and construct validity as the explanation of (choice) behaviour is an integral part of social science research.

3.1. Taste heterogeneity and construct validity

An area with great potential and a need for collaborations across disciplines is the examination of taste heterogeneity and construct
validity related to concepts such as values, attitudes, norms, and perceptions. These are latent constructs, which are integrated into
choice models in what is often called hybrid choice models (HCM). McFadden (1986) pioneered the development of choice models that
integrate latent psychological and sociological characteristics already over 30 years ago. The objective of HCMs, as latter presented by
Ben-Akiva et al. (2002), is to go beyond the random utility model (RUM) in its narrowest formulation integrating a behavioural

Table 2
Cross-disciplinary issues for choice modelling research.
Issue Explanation

Taste heterogeneity and construct Preferences can vary in a population and such heterogeneity can be explained by considering values, attitudes, norms
validity and perceptions.
Causality Definition of the treatment variable; reverse causality of choice behaviour and its determinants
Decision rules and processes Heterogeneity of decision rules, e.g. utility maximisation, loss aversion, choice set formation
Joint decision making Choices are made jointly in a partnership or at the household level and not by a single decision maker
Sensitive topics Social desirability bias is a problem in surveys on sensitive topics; stated choice experiments can lower this bias.
Social networks Individual choice behaviour depends on social embeddedness.
Big data and machine learning Combining machine learning techniques with theory-informed choice models

6
U. Liebe and J. Meyerhoff Journal of Choice Modelling 38 (2021) 100270

approach to develop predictive choice models further.


In the last years, there has been an increasing interest in applications of hybrid choice models to explain taste heterogeneity, with
applications in different disciplines including sociology (e.g., Czajkowski et al., 2017; Mariel et al., 2015; Mariel and Meyerhoff, 2016;
Liebe et al., 2021; Mariel and Meyerhoff, 2016). Hybrid choice models are an excellent example where expertise from different dis­
ciplines can be combined to improve choice modelling applications. Constructs such as values, attitudes, norms and perceptions are an
integral part of sociological and psychological research, including the development of appropriate survey-based measurement in­
struments. For example, models of decision making in social science research such as Schwartz’s norm-activation model (Schwartz,
1977) could be combined with choice modelling to better model the actual decision-making process and outcome. On the other hand,
appropriate statistical choice models that can integrate these sociological and psychological constructs are mostly developed in
transportation and different fields of applied economics such as environmental economics; see, for example, Czajkowski et al. (2017)
for a study on social norms, and Mehdizadeh et al. (2019) or Franceschinis et al. (2020) for studies on norm activation.
Nevertheless, constructs such as values, attitudes, social and personal norms, and perceptions are often not clearly defined in the
literature, and various definitions and corresponding theories exist across the social sciences (e.g. Ajzen, 1991; Eagly and Chaiken,
1993; Schwartz, 1994). The definitions provided in the box below indicate some important conceptual differences.

Definition of latent constructs to examine construct validity


Values: “There is widespread agreement in the literature regarding five features of the conceptual definition of values: A value is a (1) belief (2) pertaining to
desirable end states or modes of conduct, that (3) transcends specific situations, (4) guides selection or evaluation of behaviour, people, and events, and (5) is
ordered by importance relative to other values to form a system of value priorities […]. These are the formal features that distinguish values from such related
concepts as needs and attitudes.” (Schwartz 1994: 20)
Attitudes: Attitudes are less “fundamental” than values. An “[a]ttitude is a psychological tendency that is expressed by evaluating a particular entity with some
degree of favour or disfavor” (Eagly and Chaiken, 1993: 1).
Norms: In surveys, researchers do not measure a social norm as such, but rather, they measure “subjective norms”, which can be defined as “the perceived social
pressure to perform or not to perform the behaviour” (Ajzen, 1991, p. 188), and – more specifically – normative beliefs, defined as “the likelihood that important
referent individuals or groups approve or disapprove of performing a given behaviour” (Ajzen, 1991, p. 195). Compared to social norms, personal norms (also
moral norms or internalised norms) are not supported by positive and negative sanctions of third parties; they rely on self-sanctioning such as bad consciousness
in case of non-compliance with a norm. In surveys, researchers often measure “feelings of moral obligations” regarding a specific behaviour (Schwartz, 1977) as
an indicator of personal norms.
Perceptions: “Different people may experience the same sensory information in radically different ways, because perception is an active, creative process in which
raw sensory data are organized and given meaning” (Passer et al., 2008: 200). They can be created in bottom-up and top-down processing. While bottom-up
processing means that single elements of a stimulus are combined into a “unified perception”, top-down processing means that “sensory information is
interpreted in light of existing knowledge, concepts, ideas and expectations.” (Passer et al., 2008: 200).

The social science literature suggests a hierarchy from more stable/manifest constructs, such as values, to less stable/latent con­
structs, such as perceptions. While all constructs can have direct effects on preferences and corresponding behaviour, a hierarchy
begins with fundamental values (e.g. universalism) affecting general attitudes (e.g. environmental concern). In turn, they affect
specific attitudes (e.g. attitudes towards recycling), as well as subjective norms (e.g. perceived social approval and moral obligation to
recycle) towards a specific (choice) behaviour (e.g. choice between different recycling options).
However, so far, most applications of HCMs in the context of choice behaviour included a direct effect of more general concepts,
such as values and attitudes, on stated preferences (see Mariel et al., 2020). These effects are expected to be stronger if indirect effects
are also taken into account, that is, effects of general attitudes on specific attitudes towards a behaviour, which in turn directly affect
stated preferences. Strong attitude-behaviour as well as norm-behaviour links follow from Ajzen’s correspondence rule that attitudes
and behaviour should be measured at the same level of specificity (Ajzen, 1988, 1991). However, somewhat counterintuitively, this
comes at the cost of low explanatory power. The correspondence rule implies that the indicators for attitudes and behaviour become
very similar (i.e., a positive effect of an attitude towards recycling – being in favour of recycling – on the stated intention to recycle).
Thus, while there might be high predictive power, that is, an association between specific attitudes/norms and stated preferences, the
explanatory power might be rather low.
There are arguments in favour of considering general concepts such as attitudes and their direct and indirect effects via specific
concepts on stated choice behaviour. Borriello and Rose (2019) analysed the effects of general and specific attitudes in a stated choice
experiment study on train services. They argue, and empirically support, that both general and specific (in their terms “localised”)
attitudes should be included as latent variables in the hybrid choice model. Omitting one of these constructs, especially specific at­
titudes, would not fully reflect the psychological processes involved in choice-making and might lead to inconsistent estimates.
On the other hand, it is also suggested in the literature that more fundamental and general constructs, like values and general
attitudes, might be rather stable over time and create less need for a complex modelling approach, such as hybrid choice models,
because endogeneity bias should be low or non-existent: “… not all latent variables are equally endogenous. Indeed, more mutable
constructs like attitudes and perceptions may be more likely to be endogenous to the choice behaviour of interest, due to learning
effects, cognitive dissonance, etc., but constructs such as social norms and values that are stable over longer periods of time are much
less likely to be influenced by short-term behaviour” (Vij and Walker, 2016: 212).
Perceptions are another construct that is discussed in the context of taste heterogeneity in and construct validity of choice data.
McFadden (1986) already explicitly considered perceptions, conceptualised as “beliefs regarding the products”, as a useful determi­
nant of preferences and behaviour. He provides an example that illustrates the differences between attitudes and perceptions: “For
example, the purchaser of a room air conditioner may have perceptions of the durability of alternative brands, attitudes regarding the
importance of durability, preferences among specific brands and models, a protocol to maximize preference taking into account the

7
U. Liebe and J. Meyerhoff Journal of Choice Modelling 38 (2021) 100270

opportunity cost of the outlay for the product, and a behavioural intention to choose a specific brand” (McFadden, 1986: 276).
In line with the reasoning by McFadden (1986), Bahamonde-Birke et al. (2017) point out that attitudes are “individual-specific”
latent attributes which are constant across choice alternatives. Perceptions, however, are “alternative-specific” latent attributes, and
hence, they vary across choice alternatives. On the one hand, perceptions can be influenced by individuals’ characteristics, such as
socio-economic variables or attitudes; on the other hand, perceptions are also affected by stimuli and information present in decision
environments. Bahamonde-Birke et al. (2017: 490) find effects of perceptions beyond attitudinal effects in a choice model. They argue:
“perceptions may explain a significant portion of the variability normally captured by alternative specific constants, offering signif­
icant improvements in model goodness-of-fit”.
Besides estimating separate effects of attitudes and perceptions in choice models, (social-) psychological theories also suggest
potential relationships between attitudes and perceptions. In top-down processing in the creation of perceptions (see Passer et al.,
2008: 200) concepts such as attitudes are employed to interpret information referring to different choice alternatives and hence at­
titudes can be a determinant of perceptions. This is also proposed in dual-process theories, such as Fazio’s MODE model, for example,
regarding automatic processes: “In effect, the automatically-activated attitude serves as a filter through which the object is viewed, just
as suggested by the field’s longstanding emphasis on the constructive nature of perception […]. Once activated, attitudes color
perceptions of the object, and ensuing judgments or behaviors” (Fazio and Olson, 2014: 155).
Another area which is suggested in the (social-)psychological literature and has important implications for choice models is the
relationship between perception and attention (see Passer et al., 2008: 201). In this regard, perceptions might help to shed more light
on attribute- and alternative-non-attendance in choice experiments (Hess and Hensher, 2013; Weller et al., 2014).

3.2. Causality

Implicitly or explicitly, as in the case of political science applications of conjoint experiments, researchers assume that they esti­
mate causal effects of policy interventions or attributes of a choice alternative. However, there is a disconnect between choice
modelling and causal analysis in social science research (similar to transportation research, as discussed by Brathwaite and Walker,
2018). In choice modelling research it is not always clear what the exact causal treatment is (e.g., a whole policy or attributes of a
policy) and this has implications for assumptions about causality (Brathwaite and Walker, 2018). Here, political science applications of
conjoint analysis (e.g., Hainmueller et al., 2014) are very explicit and specific about causal assumptions. Apart from identifying causal
effects within a conjoint or choice experiment, that is, the attribute effects, questions emerge such as to what extent reverse causality
regarding determinants of choice behaviour is present. Kroesen et al. (2017) show in a two-wave panel study that travel choice
behaviour can affect attitudes stronger than vice versa. Such issues of reverse causality are also discussed in the sociological literature
(e.g., Leszczensky and Wolbring, 2019). Further, as mentioned in section 3.1, it is important to think about the causal relationships
between behavioural determinants such as attitudes, norms, and perceptions, and a specific behavioural choice. Hybrid choice models
can help to uncover the proposed causal effects (Vij and Walker, 2016; Liebe et al., 2021).

3.3. Decision rules

Another challenge for applying choice modelling in social science is that not only preference heterogeneity exists but also decision-
rule heterogeneity. Albeit utility maximisation might often be a plausible assumption, it is only one rule among others (Hess et al.,
2012; Chorus, 2014). In economics, the assumption of utility maximisation is often seen as a prerequisite to estimate welfare effects
(see Hess et al., 2018); yet, ways for calculating welfare measures under other decision rules have been suggested (Dekker and Chorus,
2018). In other disciplines, however, calculating welfare measures is not the critical objective and hence, depending on the appli­
cation, respondents’ application of alternative decision rules itself might be of greater interest. The decision rules provide a common
ground for collaborations regarding the study of decision-rule heterogeneity across social science disciplines. For instance, research in
hiring decision making shows that employers might apply shortcuts such as focusing on specific attributes or systematically reducing
the candidate pool by using specific screening rules (Bills et al., 2017), which can be examined with revealed or stated choice data.
Random regret minimisation has gained some popularity in choice modelling research (Chorus, 2010; Van Cranenburgh et al., 2015),
and it could be fruitfully implemented in choice models in sociological and political science research. This decision rule is closely
related to loss aversion (Tversky and Kahneman, 1991). For example, when analysing neighbourhood choices, it could take into ac­
count that losses in neighbourhood attributes such as environmental conditions and share of same ethnic group loom larger than gains
of equal size. Also, models of choice set formation (see Bruch and Swait, 2019) can take into account that, depending on the type of
decision, individuals might not consider all alternatives and attributes. Instead, they can be described as boundedly rational actors first
selecting a reduced set of choice alternatives and then selecting the alternative they prefer most.

3.4. Joint decision making

Most applications of choice modelling assume a single decision-maker, for example, a single consumer choosing between different
product alternatives. Yet, many decisions regarding housing, jobs, family planning are joint decisions made in a partnership or at the
household level, more generally. The application of choice modelling in social science research would therefore be more realistic if
choice models can capture these joint decision-making processes. Indeed, there are different approaches to how this can be done. For
example, Timmermans et al. (1992) study residential choice behaviour in dual-earner households. They develop a decompositional
model of joint decision making taking hierarchical information integration into account. The final model outcomes were based on

8
U. Liebe and J. Meyerhoff Journal of Choice Modelling 38 (2021) 100270

integrating separate evaluations of the household partners. An alternative would be that partners answer choice tasks jointly. In
principle, studies can employ an individual-based approach by classifying intra-household interactions and modelling them sequen­
tially, or a group-based approach by using group utility functions (see Ho and Mulley, 2015). Other studies that examine
intra-household decision making include Dosman and Adamowicz (2006) who investigate decisions on family vacations based on
revealed and stated preference data; they also consider sociological concepts/measures of power as well as bargaining behaviour.
Scarpa et al. (2012) investigate couples’ preferences for tap water attributes based on stated preference data of both household
members.

3.5. Sensitive topis

Surveying sensitive topics is another challenge in empirical social science research (Krumpal, 2011). Similar to the randomized
response technique and list experiments that have been proposed to lower socially desirable response behaviour, stated choice ex­
periments also have the potential to lower social desirability. Due to its multifactorial experimental design, the sensitive attribute is
only one among other attributes in a choice set, and respondents need to make a trade-off between attributes, which can make a
socially desirable response less obvious. This effect has already been demonstrated in factorial survey experiments (Mutz, 2011;
Auspurg et al., 2015; Beyer and Liebe, 2015; Liebe et al., 2020). For example, regarding just earnings, a comparison between direct
questioning and a factorial survey indicates that gender is evaluated less relevant using the former but much more relevant using the
latter (Liebig et al., 2015). Therefore, gender gaps in just earnings can be uncovered using multifactorial survey experiments, whereas
direct questions lead to socially desirable responses; that is, there is no (just) gender difference. Applications of stated choice ex­
periments to sensitive issues include a study on wildlife crime in Denmark (Højberg et al., 2017) and ethnic discrimination in Germany
(Liebe and Beyer, 2020, also see Table 1). Hojberg et al. (2017) uncovered strong preferences of landowners for illegal lethal actions in
the context of wolf re-immigration; sensitive issues included a choice attribute on attempts to shoot wolves. Liebe and Beyer (2020)
examine preferences for carpooling options in Germany and reveal ethnic/racial discrimination by varying the origin of drivers as one
attribute in the stated choice experiment. The sensitive attribute was the drivers’ name. The study results show discrimination towards
drivers with a presumed Turkish origin, which can be explained by xenophobic attitudes and lack of regular contact with perceived
‘foreigners’. Yet, multifactorial survey experiments can still be prone to a social desirability bias, especially if they contain a small
number of attributes as well as repeated decisions by respondents. Therefore, depending on the concrete topic and application, it might
be beneficial to apply a between-subject design and to present only one vignette or choice set per respondent (Beyer and Liebe, 2015;
Walzenbach, 2019).

3.6. Social context

Behaviour and decision making is influenced by social contexts (Bruch and Feinberg, 2017). Social norms and social capital shape
micro- and macro-level outcomes in a society and hence social ties and networks play a crucial role in explaining economic, social and
political outcomes (Bourdieu, 1986; Portes, 1998; Coleman, 1990; Putnam, 1993). Combining choice modelling with network analysis
(Calastri et al., 2018; Feinberg et al., 2020; Neilson and Wichmann, 2014; Pink et al., 2020; Wichmann et al., 2016) is a powerful
approach to investigate the choice processes involved in social tie formation and the causal determinants of partner choice as well as
the effects of social networks on individuals’ choice behaviour. Dynamic network analyses – studying social tie formation over time –
allow for examining non-static choice behaviour. It can easily be combined with a choice modelling framework, as both share sta­
tistical modelling techniques that can be interpreted in terms of random utility maximisation (Snijders, 1996; Pink et al., 2020). Also,
stated choice experiments could be employed to study social context effects. For example, it is possible to include context as a choice
attribute (see Beyer and Liebe, 2015 for a study on anti-Semitism) or to combine framing and stated choice experiments (see Carlsson
et al., 2010 for a choice experiment on preferences for organic and fair trade coffee, in which respondents were randomly assigned to
treatments with information about choices by other consumers).

3.7. Big data and machine learning

Big data and machine learning offer many new opportunities for social science research. Machine learning is widely perceived as
data-driven and this has led to discussions about its relationship with causal analysis and theory testing in social science research (see
Pearl, 2019). Nevertheless, supervised machine learning can help to understand decision-making processes better and to uncover
heterogeneous treatment effects (Molina and Garip, 2019). Unsupervised machine learning is related to latent class analysis and latent
variable modelling, which are commonly employed in choice modelling research (Molina and Garip, 2019). Also, decision trees, a
frequently used machine learning technique, can be linked to rational choice and microeconomic theory (Brathwaite et al., 2017).
Furthermore, artificial neural networks have been applied to choice data in order to analyse, for example, automobile and travel mode
choices (Mohammadian and Miller, 2002; Hagenauer and Helbich, 2017). While the potential of artificial neural networks is still to be
exploited for social science research, an important factor to consider is the sample size that is needed for this type of analysis
(Alwosheel et al., 2018). Machine learning studies typically work with large datasets, and this is not always the case for choice
modelling research in the social sciences. More generally, as indicated in a review by Hillel et al. (2020), referring to datasets, model
validation and model optimisation, there are several methodological pitfalls and areas of improvement of using machine learning as an
alternative approach to random utility theory.
Besides the common issues and opportunities for cross-disciplinary research across the social sciences mentioned in Table 2, there is

9
U. Liebe and J. Meyerhoff Journal of Choice Modelling 38 (2021) 100270

also methodological research in different disciplines, for example, regarding the validity and reliability of stated choice data. Topics
include choice task complexity (e.g., , Swait and Adamowicz; Bansak et al., 2019), fatigue by respondents (e.g., Meyerhoff et al., 2015;
Bansak et al., 2018), and test-retest reliability (e.g., Liebe et al., 2016). Studies also employ methods such as eye-tracking to explore
issues such as attribute attention and choice consistency in choice and conjoint tasks (e.g., Uggeldahl et al., 2016; Jenke et al., 2020).
However, it seems as if much of this literature in different disciplines is unconnected and hence could benefit from a cross-disciplinary
dialogue.

4. Conclusion and potential for future research

While choice modelling is established in some areas of economics, so far it has received less attention in other social science
disciplines such as sociology or political science. Currently, this is changing. Especially the number of stated choice experiments as well
as conjoint experiments increases in sociological and political science research. Further, choice modelling is linked to agent-based
modelling, network analysis, and machine learning; all of these methods gain popularity across social science disciplines. It is pre­
cisely the common methodological focus, interest in the explanation of decision making in economic, social and political contexts, as
well as the traditions of rational choice theory in various social science disciplines, which indicate the usefulness of choice modelling
across the social sciences. In this regard, our contribution only provides a snapshot of the potential of choice modelling for social
science research.
An essential advantage of choice models is their close link to (micro-)theories in different social sciences. Social scientists are
interested in explaining educational choices, vote choice, migration decision making, for instance, and therefore using a modelling
framework that mirrors these processes more accurately than “standard” survey research would be more appropriate. In recent years
substantial progress has been made in improving the flexibility of choice models and incorporating different decision rules as well as
preference heterogeneity. Also, compared with standard survey research in the social sciences, theoretical assumptions about pref­
erences can be tested, especially when combined with experimental data or agent-based modelling. A common critique of choice
modelling in the social sciences is that these models are too simple in the sense that they cannot capture complex social context effects,
sequential decision-making processes, as well as joint decision-making processes. There is undoubtedly still much work to do on all of
these aspects, but it has to be stressed that contributions already exist that address these problems offering avenues for collaborations
across disciplines.
In this paper, we tried to make the point that choice modelling provides an overarching methodological framework for the social
sciences. Much of social science research is concerned with individual and household decision making, which in turn results in macro-
level social, political and economic outcomes. While choice modelling is related to the often citicised rational choice assumptions, it
again is vital to note that choice modelling is a flexible tool which can accommodate different decision rules and important aspects
such as joint and sequential decision making. This flexibility opens up great potential for social science research as well as bridging
choice modelling and social science.

CRediT authorship contribution statement

Ulf Liebe: Conceptualization, Formal analysis, Writing - original draft, Writing - review & editing. Jürgen Meyerhoff: Concep­
tualization, Formal analysis, Writing - original draft, Writing - review & editing.

Declaration of competing interest

There are no conflicts of interest.

References

Ajzen, I., 1988. Attitudes, Personality, and Behavior. Open University Press.
Ajzen, I., 1991. The theory of planned behavior. Organ. Behav. Hum. Decis. Process. 50 (2), 179–211.
Alwosheel, A., van Cranenburgh, S., Chorus, C.G., 2018. Is your dataset big enough? Sample size requirements when using artificial neural networks for discrete
choice analysis. Journal of Choice Modelling 28, 167–182.
Aquilino, W.S., 1991. Family Structure and Home-Leaving: a further specification of the relationship. J. Marriage Fam. 53 (4), 999–1010.
Auspurg, K., Hinz, T., 2015. Multifactorial experiments in surveys: conjoint analysis, choice experiments, and factorial surveys. In: Wolbring, T., Keuschnigg, M.
(Eds.), Experimente in Den Sozialwissenschaften. Soziale Welt Sonderband 22. Baden-Baden, Germany, Nomos, pp. 291–315.
Auspurg, K., Hinz, T., Sauer, C., Liebig, S., 2015. The factorial survey as a method for measuring sensitive issues. In: Engel, U., Jann, B., Lynn, P., Scherpenzeel, A.,
Sturgis, P. (Eds.), Improving Survey Methods: Lessons from Recent Research. Routledge, New York, pp. 137–152.
Bahamonde-Birke, F.J., Kunert, U., Link, H., Ortúzar, J. de D., 2017. About attitudes and perceptions: finding the proper way to consider latent variables in discrete
choice models. Transportation 44 (3), 475–493.
Bansak, K., Hainmueller, J., Hangartner, D., 2016. How economic, humanitarian, and religious concerns shape European attitudes toward asylum seekers. Science
354, 217–222.
Bansak, K., Hainmueller, J., Hopkins, D.J., Yamamoto, T., 2018. The number of choice tasks and survey satisficing in conjoint experiments. Polit. Anal. 26, 112–119.
Bansak, K., Hainmueller, J., Hopkins, D.J., Yamamoto, T., 2019. Beyond the breaking point? Survey satisficing in conjoint experiments. Political Science Research and
Methods (online first).
Bechtel, M., Scheve, K.F., 2013. Mass support for global climate agreements depends on institutional design. Proc. Natl. Acad. Sci. Unit. States Am. 110, 13763–13768.
Ben-Akiva, M., Mcfadden, D., Train, K., Walker, J., Bhat, C., Bierlaire, M., Bolduc, D., Boersch-Supan, A., Brownstone, D., Bunch, D.S., Daly, A., De Palma, A.,
Gopinath, D., Karlstrom, A., Munizaga, M.A., 2002. Hybrid choice models: progress and challenges. Market. Lett. 13, 163–175.
Bernasco, W., Block, R., 2009. Where offenders choose to attack: a discrete choice model of robberies in Chicago. Criminology 47 (1), 93–130.

10
U. Liebe and J. Meyerhoff Journal of Choice Modelling 38 (2021) 100270

Beyer, H., Liebe, U., 2015. Three experimental approaches to measure the social context dependence of prejudice communication and discriminatory behavior. Soc.
Sci. Res. 49, 343–355.
Bills, D.B., Di Stasio, V., Gërxhani, K., 2017. The demand side of hiring: employers in the labor market. Annu. Rev. Sociol. 43, 291–310.
Bourdieu, P., 1986. The forms of capital. In: Richardson, J. (Ed.), Handbook of Theory and Research for the Sociology of Education. Greenwood Press, New York,
pp. 241–258.
Borriello, A., Rose, J.M., 2019. Global versus Localised Attitudinal Responses in Discrete Choice. Transportation.
Brathwaite, T., Vij, A., Walker, J.L., 2017. Machine Learning Meets Microeconomics: the Case of Decision Trees and Discrete Choice arXiv:1711.04826.
Brathwaite, T., Walker, J.L., 2018. Causal inference in travel demand modeling (and the lack thereof). Journal of Choice Modelling 26, 1–18.
Bruch, E., Atwell, J., 2015. Agent-based models in empirical social research. Socio. Methods Res. 44 (2), 186–221.
Bruch, E., Feinberg, F., 2017. Decision-making processes in social contexts. Annu. Rev. Sociol. 43, 207–227.
Bruch, E.E., Mare, R.D., 2006. Neighborhood Choice and Neighborhood Change. Am. J. Sociol. 112 (3), 667–709.
Bruch, E., Swait, J., 2019. Choice set formation in residential mobility and its implications for segregation dynamics. Demography 56, 1665–1692.
Calastri, C., Hess, S., Daly, A., Carrasco, J.A., Choudhury, C., 2018. Modelling the loss and retention of contacts in social networks: the role of dyad-level heterogeneity
and tie strength. Journal of Choice Modelling 29, 63–77.
Carlsson, F., García, J.H., Löfgren, A., 2010. Conformity and the demand for environmental goods. Environ. Resour. Econ. 47, 407–421.
Chorus, C.G., 2010. A new model of random regret minimization. Eur. J. Transport Infrastruct. Res. 10 (2), 181–196.
Chorus, C.G., 2014. Capturing alternative decision rules in travel choice models: a critical discussion. In: Hess, S., Daly, A. (Eds.), Handbook of Choice Modelling.
Edward Elgar, Northampton, MA, pp. 290–310.
Chorus, C.G., Pudāne, B., Mouter, N., Campbell, D., 2018. Taboo trade-off aversion: a discrete choice model and empirical analysis. Journal of Choice Modelling 27,
37–49.
Chorus, C., Sandorf, E.D., Mouter, Niek, 2020. Diabolic dilemmas of COVID-19: an empirical study into Dutch society’s trade-offs between health impacts and other
effects of the lockdown. PloS One 15 (9), e0238683.
Coleman, J.S., 1990. Foundations of Social Theory. Belknap Press, Cambridge, Ma.
Coleman, J.S., Fararo, T.J. (Eds.), 1992. Rational Choice Theory: Advocacy and Critique. Sage, Newbury Park, CA.
Czajkowski, M., Hanley, N., Nyborg, K., 2017. Social norms, morals and self-interest as determinants of pro-environment behaviours: the case of household recycling.
Environ. Resour. Econ. 66 (4), 647–670.
De Vries, C.E., 2007. Sleeping giant: Fact or fairytale? How European integration affects national elections. Eur. Union Polit. 8 (3), 363–385.
Dekker, T., Chorus, C.G., 2018. Consumer surplus for random regret minimisation models. Journal of Environmental Economics and Policy 7 (3), 269–286.
Demel, S., Mariel, P., Meyerhoff, J., 2019. Job preferences of business and economics students. Int. J. Manpow. 40 (3), 473–499.
Dosman, D., Adamowicz, W., 2006. Combining stated and revealed preference data to construct an empirical examination of intra household bargaining. Rev. Econ.
Househ. 4 (1), 15–34.
Eagly, A.H., Chaiken, S., 1993. The Psychology of Attitudes. Harcourt Brace Jovanovich College Publishers, Fort Worth, Tex.
Elías, J.J., Lacetera, N., Macis, M., 2019. Paying for kidneys? a randomized survey and choice experiment. Am. Econ. Rev. 109, 2855–2888.
Elsenbroich, C., Payette, N., 2020. Choosing to cooperate: modelling public goods games with team reasoning. Journal of Choice Modelling 34, 100203.
Fazio, R.H., Olson, M.A., 2014. The MODE model: attitude-behavior processes as a function of motivation and opportunity. In: Gawronski, B., Sherman, J.W.,
Trope, Y. (Eds.), Dual-process Theories of the Social Mind. The Guilford Press, New York, pp. 155–171.
Feinberg, F., Bruch, E., Braun, M., Hemenway Falk, B., Fefferman, N., McDonnell Feit, E., Helveston, J., Larremore, D., McShane, B.B., Patania, A., Small, M.L., 2020.
Choices in networks: a research framework. Market. Lett. 31, 349–359.
Franceschinis, C., Liebe, U., Thiene, M., Meyerhoff, J., McBratney, A., Field, D., 2020. Norm Activation in Discrete Choice Experiments (Manuscript).
Frith, M.J., 2019. Modelling taste heterogeneity regarding offence location choices. Journal of Choice Modelling 33. Article 100187.
Glasgow, G., 2011. Introduction to the virtual issue: recent advances in discrete choice methods in Political Science. Polit. Anal. 19, 1–3.
Green, D.P., Shapiro, I., 1994. Pathologies of Rational Choice Theory: A Critique of Applications in Political Science. Yale University Press, New Haven.
Hagenauer, J., Helbich, M., 2017. A comparative study of machine learning classifiers for modeling travel mode choice. Expert Syst. Appl. 78, 273–282.
Hainmueller, J., Hopkins, D.J., 2015. The hidden American immigration consensus: a conjoint analysis of attitudes toward immigrants. Am. J. Polit. Sci. 59, 529–548.
Hainmueller, J., Hopkins, D.J., Yamamoto, T., 2014. Causal inference in conjoint analysis: understanding multidimensional choices via stated preference experiments.
Polit. Anal. 22, 1–30.
Hess, S., Hensher, D.A., 2013. Making use of respondent reported processing information to understand attribute importance: a latent variable scaling approach.
Transportation 40, 397–412.
Hess, S., Daly, A., Batley, R., 2018. Revisiting consistency with random utility maximisation: theory and implications for practical work. Theor. Decis. 84, 181–204.
Hess, S., Stathopoulos, A., Daly, A., 2012. Allowing for heterogeneous decision rules in discrete choice models: an approach and four case studies. Transportation 39,
565–591.
Hillel, T., Bierlaire, M., Elshafie, M.Z.E.B., Jin, Y., 2020. A systematic review of machine learning classification methodologies for modelling passenger mode choice.
Journal of Choice Modelling 100221 online 6 August 2020.
Ho, C., Mulley, C., 2015. Intra-household interactions in transport research: a review. Transport Rev. 35, 33–55.
Højberg, P.L., Nielsen, M.R., Jacobsen, J.B., 2017. Fear, economic consequences, hunting competition, and distrust of authorities determine preferences for illegal
lethal actions against gray wolves (Canis lupus): a choice experiment among landowners in Jutland, Denmark. Crime, Law and Social Change 67, 461–480.
Humburg, M., van der Velden, R., 2015. Skills and the graduate recruitment process: evidence from two discrete choice experiments. Econ. Educ. Rev. 49, 24–41.
Ibraimovic, T., Hess, S., 2017. Changes in the ethnic composition of neighbourhoods: analysis of household’s response and asymmetric preference structures. Pap.
Reg. Sci. 96 (4), 759–784.
Jenke, L., Bansak, K., Hainmueller, J., Hangartner, D., 2020. Using eye-tracking to understand decision-making in conjoint experiments. Polit. Anal. 1–27.
Kahneman, D., Tversky, A. (Eds.), 2000. Choices, Values and Frames. Cambridge University Press, Cambridge, UK.
Kirkland, P.A., Coppock, A., 2018. Candidate choice without party labels: new insights from conjoint survey experiments. Polit. Behav. 40, 571–591.
Kroesen, M., Handy, S., Chorus, C.G., 2017. Do attitudes cause behavior or vice versa? An alternative conceptualization of the attitude–behavior relationship in travel
behavior modeling. Transport. Res. Part A 101, 190–202.
Kroneberg, C., Kalter, F., 2012. Rational choice theory and empirical research: methodological and theoretical contributions in Europe. Annu. Rev. Sociol. 38, 73–92.
Krumpal, I., 2011. Determinants of social desirability bias in sensitive surveys: a literature review. Qual. Quantity 47, 2025–2047.
Leeper, T.J., Hobolt, S.B., Tilley, J., 2020. Measuring subgroup preferences in conjoint experiments. Polit. Anal. 28, 207–221.
Leszczensky, L., Wolbring, T., 2019. How to deal with reverse causality using panel data? Recommendations for researchers based on a simulation study. Socio.
Methods Res. online first.
Liebe, U., Beyer, H., 2020. Examining discrimination in everyday life: a stated choice experiment on racism in the sharing economy. J. Ethnic Migrat. Stud. (online
first).
Liebe, U., Hundeshagen, C., Beyer, H., von Cramon-Taubadel, S., 2016. Context effects and the temporal stability of stated preferences. Soc. Sci. Res. 60, 135–147.
Liebe, U., Mariel, P., Beyer, H., Meyerhoff, J., 2021. Uncovering the nexus between attitudes, preferences and behavior in sociological applications of stated
choiceexperiments. Socio. Methods Res. 50 (1), 310–347.
Liebe, U., Meyerhoff, J., Kroesen, M., Chorus, C., Glenk, K., 2018. From welcome culture to welcome limits? Uncovering preference changes over time for sheltering
refugees in Germany. PloS One 13 (8), e0199923.
Liebe, U., Moumouni, I.M., Bigler, C., Ingabire, C., Bieri, S., 2020. Using factorial survey experiments to measure attitudes, social norms, and fairness concerns in
developing countries. Socio. Methods Res. 49, 161–192.

11
U. Liebe and J. Meyerhoff Journal of Choice Modelling 38 (2021) 100270

Liebig, S., Sauer, C., Friedhoff, S., 2015. Using factorial surveys to study justice perceptions: five methodological problems of attitudinal justice research. Soc. Justice
Res. 28, 415–434.
Logan, J.L., 1996. Opportunity and choice in socially structured labor markets. Am. J. Sociol. 102, 114–160.
Louviere, J.J., Hensher, D.A., Swait, J.D., 2000. Stated Choice Methods: Analysis and Application. Cambridge University Press, Cambridge, England.
Louviere, J., Flynn, T.N., Carson, R.T., 2010. Discrete choice experiments are not conjoint analysis. Journal of Choice Modelling 3 (3), 57–72.
Lu, H., Rohr, C., Howarth, D., Pollitt, A., 2020. Jonathan Grant What sort of Brexit do the British people want? A longitudinal study examining the ‘trade-offs’ people
would be willing to make in reaching a Brexit deal. Journal of Choice Modelling 37, 100233.
Manski, C.F., 1981. Structural models for discrete data: the analysis of discrete choice. Socio. Methodol. 12, 58–109.
Mariel, P., Meyerhoff, J., 2016. Hybrid discrete choice models: gained insights versus increasing effort. Sci. Total Environ. 568, 433–443.
Mariel, P., Liebe, U., Meyerhoff, J., 2020. A Critical Assessment of the Current Use of Hybrid Choice Models in Environmental Economics. Manuscript.
Mariel, P., Meyerhoff, J., Hess, S., 2015. Heterogeneous preferences toward landscape externalities of wind turbines—combining choices and attitudes in a hybrid
model. Renew. Sustain. Energy Rev. 41, 647–657.
McFadden, D., 1974. Conditional logit analysis of qualitative choice behavior. In: Zarembka, P. (Ed.), Frontiers in Econometrics. Academic Press, New York,
pp. 105–142.
McFadden, D., 1986. The choice theory approach to market research. Market. Sci. 5 (4), 275–297.
Mehdizadeh, M., Zavareh, M.F., Nordfjaern, T., 2019. Mono-and multimodal green transport use on university trips during winter and summer: hybrid choice models
on the norm-activation theory. Transport. Res. Pol. Pract. 130, 317–332.
Meyerhoff, J., Oehlmann, M., Weller, P., 2015. The influence of design dimensions on stated choices in an environmental context. Environ. Resour. Econ. 61,
385–407.
Mohammadian, A., Miller, E., 2002. Nested logit models and artificial neural networks for predicting household automobile choices: comparison of performance.
Transport. Res. Rec.: Journal of the Transportation Research Board 92–100, 1807.
Molina, M., Garip, F., 2019. Machine learning for sociology. Annu. Rev. Sociol. 45, 27–45.
Möser, S., Glauser, D., Becker, R., 2019. Valuation of labour market entrance positions among (future) apprentices - Results from two discrete choice experiments.
J. Choice Model. 33, 100180.
Mummolo, J., 2016. News from the other side: how topic relevance limits the prevalence of partisan selective exposure. J. Polit. 78, 763–773.
Mutz, D.C., 2011. Population-Based Survey Experiments. Princeton University Press, Princeton, NJ.
Neilson, W., Wichmann, B., 2014. Social networks and non-market valuations. J. Environ. Econ. Manag. 67, 155–170.
Neyman, J., 1923. On the application of probability theory to agricultural experiments: essay on principles. Section 9. Stat. Sci. 5, 465–472.
Passer, M., Smith, R., Holt, N., Bremner, A., Sutherland, E., Vliek, M., 2008. Psychology: the Science of Mind and Behaviour. McGraw-Hill Education/Europe, Middle
East & Africa.
Pearl, J., 2019. The seven tools of causal inference with reflections on machine learning. Commun. ACM 62 (3), 54–60.
Pink, S., Kretschmer, D., Leszczensky, L., 2020. Choice modelling in social networks using stochastic actor-oriented models. Journal of Choice Modelling 34. Article
100202.
Portes, A., 1998. Social capital: its origins and applications in modern sociology. Annu. Rev. Sociol. 24, 1–24.
Putnam, R.D., 1993. Making Democracy Work. Princeton University Press, Princeton, New Jersey.
Rivers, D., 1988. Heterogeneity in models of electoral choice. Am. J. Polit. Sci. 32, 737–757.
Rubin, D.B., 1974. Estimating causal effects of treatments in randomized and nonrandomized studies. J. Educ. Psychol. 66, 688–701.
Scarpa, R., Thiene, M., Hensher, D.A., 2012. Preferences for tap water attributes within couples: an exploration of alternative mixed logit parameterizations. Water
Resour. Res. 48, W01520.
Schelling, T.C., 1971. Dynamic models of segregation. J. Math. Sociol. 1, 143–186.
Schwartz, S.H., 1977. Normative influences on altruism. In: Berkowitz, L. (Ed.), Advances in Experimental Social Psychology, vol. 10, pp. 221–279.
Schwartz, S.H., 1994. Are there universal aspects in the structure and contents of human values? J. Soc. Issues 50, 19–45.
Snijders, T.A.B., 1996. Stochastic actor-oriented models for network change. J. Math. Sociol. 21, 149–172.
Sobolewska, M., Galandini, S., Lessard-Phillips, L., 2017. The public view of immigrant integration: multidimensional and consensual. Evidence from survey
experiments in the UK and The Netherlands. J. Ethnic Migrat. Stud. 43, 58–79.
Stadelmann-Steffen, I., Dermont, C., 2019. Citizens’ opinions about basic income proposals compared – a conjoint analysis of Finland and Switzerland. J. Soc. Psychol.
online first.
Swait, J., Adamowicz, W., 200. Choice environment, market complexity and consumer behavior: a theoretical and empirical approach for incorporating decision
complexity in models of consumer choice. Organ. Behav. Hum. Decis. Process. 86, 141–167.
Teele, D.L., Kalla, J., Rosenbluth, F., 2018. The ties that double bind: social roles and women’s underrepresentation in politics. Am. Polit. Sci. Rev. 112, 525–541.
Timmermans, H.J.P., Borgers, A.W.J., van Dijk, J., Oppewal, H., 1992. Residential choice behaviour of dual earner households: a decompositional joint choice model.
Environ. Plann. 24 (4), 517–533.
Tutić, A., Liebe, U., 2017. Rational choice. In: Outhwaite, W., Turner, S. (Eds.), Sage Handbook of Political Sociology. Sage, London (Chapter 13).
Tversky, A., Kahneman, D., 1991. Loss aversion in riskless choice: a reference dependent model. Q. J. Econ. 107 (4), 1039–1061.
Uggeldahl, K., Jacobsen, C., Lundhede, T., Olsen, S., 2016. Choice certainty in Discrete Choice Experiments: will eye tracking provide useful measures? Journal of
Choice Modelling 20, 35–48.
Van Cranenburgh, S., Guevara, C.A., Chorus, C.G., 2015. New insights on random regret minimization models. Transport. Res. Pol. Pract. 94, 91–109.
Vij, A., Walker, J.L., 2016. How, when and why integrated choice and latent variable models are latently useful. Transp. Res. Part B Methodol. 90, 192–217.
Wallander, L., 2009. 25 Years of factorial surveys in sociology: a review. Soc. Sci. Res. 38, 505–520.
Walzenbach, S., 2019. Hiding sensitive topics by design? An experiment on the reduction of social desirability bias in factorial surveys. Survey Research Methods 13,
103–121.
Weller, P., Mariel, P., Oehlmann, M., Meyerhoff, J., 2014. Stated and inferred attribute non-attendance in a design of designs approach. Journal of Choice Modelling
14, 43–56.
Wichmann, B., Chen, M., Adamowicz, W., 2016. Social networks and choice set formation in discrete choice models. Econometrics 4, 42.
Zeng, Z., Xie, Y., 2008. A preference-opportunity-choice framework with applications to intergroup friendship. Am. J. Sociol. 114, 615–648.

12

Вам также может понравиться