Journal of Behavioral and Experimental Finance 17 (2018) 83–96

Contents lists available at ScienceDirect

Journal of Behavioral and Experimental Finance

journal homepage: www.elsevier.com/locate/jbef

Full length article

Examining socially responsible investment preferences: A discrete

choice conjoint experiment
George Apostolakis a,b, *, Gert van Dijk a,c,d , Frido Kraanen c , Robert J. Blomme e
a
Nyenrode Business Universiteit, Breukelen, The Netherlands
b
Department of Economics, University of Crete, Rethymno Campus, 74100, Rethymno, Greece
c
TIAS School for Business and Society, Tilburg University, Tilburg, The Netherlands
d
Department of Social Sciences, Wageningen University, Wageningen, The Netherlands
e
Centre for Leadership and Management Development, Nyenrode Business Universiteit, Breukelen, The Netherlands

article info a b s t r a c t
Article history: Using a hierarchical Bayesian estimation of a discrete choice conjoint model, this research examines
Received 25 November 2017 preferences for different responsible investment characteristics from the perspective of pension ben-
Accepted 11 January 2018 eficiaries. We conducted a choice-based conjoint experiment in which participants were invited to
Available online 6 February 2018
select a preferred investment portfolio among different options by combining attributes such as socially
responsible investments and impact investments. Based on a sample of 334 respondents, the results show
Keywords:
the utility and relative importance that members of the administrative organization of a Dutch pension
Conjoint analysis
Impact investing fund with a cooperative structure attach to the socially responsible portfolio. Latent class analysis yielded
SRI three segments of pension beneficiaries with different levels of psychological distance toward socially
Institutional investors responsible investments.
Pension funds © 2018 Elsevier B.V. All rights reserved.
Psychological distance

1. Introduction investments to satisfy responsibility criteria have not been suf-

The conventional form of investing is vulnerable to high-
volatility events and crises in the financial markets. Most im-
portantly, conventional investing has proven insufficient for ad-
dressing important social issues. A newly introduced investment
culture, known as ‘‘impact investing’’, strives for long-term social
gains by addressing social problems rather than attempting to
maximize financial returns. Additionally, pension funds and other
institutional investors are increasingly examining investments
through the lens of social responsibility and are pursuing more
active roles in corporate governance. This study explores prefer-
ences for a socially responsible portfolio by employing a discrete
choice conjoint experiment. Our sample consists of members of
the administrative organization of a Dutch pension fund with a
cooperative structure in the healthcare sector.
Responsible investments have increasingly gained momentum
over the past few years. Institutional investors, such as pension
funds, are proactively integrating such practices into their in-
vestment strategies as they acknowledge sustainability and social
responsibility as part of their fiduciary duty (Koedijk and Slager,
2011). Nevertheless, pension fund beneficiaries’ preferences for
*
Correspondence to: Nyenrode Business Universiteit, Straatweg 25, 3621 BG
Breukelen, P.O. Box 130, 3620 AC Breukelen, The Netherlands.
E-mail address: g.apostolakis@nyenrode.nl (G. Apostolakis).
ficiently investigated. Moreover, the literature provides contra-
dictory findings regarding the long-term financial performance
of responsible funds compared with that of conventional funds;
therefore, concrete conclusions have not been made.
With respect to member inclusion in the design of the invest-
ment policies of pension funds, particularly with respect to their
impact and socially responsible investing in the long term, we
must first address the effect of psychological distance on individual
preferences. The phenomenon of psychological distance can be
illustrated by the following example. In general, when people are
asked for their preferences regarding sustainable animal produc-
tion, they can rather easily identify the criteria with which farmers
should comply. In this case, people cite criteria such as ‘‘animal
welfare’’, ‘‘environmental impact’’, ‘‘impact on children’s health’’,
and ‘‘impact on health in the long run’’. Generally, people indicate
that they believe these conditions are important. In addition, they
understand that the end user should pay the extra cost of sustain-
able production. However, when the same people must make the
same decisions while standing in front of shelves in a supermarket,
they address concrete and short-run architectures of choice (van
Trijp, 2013). Thus, discussing long-term effects in relation to the
next generation and in connection with people elsewhere in the
world results in different choices from those made in supermarkets
or grocery stores.

https://doi.org/10.1016/j.jbef.2018.01.001
2214-6350/© 2018 Elsevier B.V. All rights reserved.
84 G. Apostolakis et al. / Journal of Behavioral and Experimental Finance 17 (2018) 83–96
Construal level theory (CLT) posits that distal and abstract
thoughts are formulated at a higher construal level and that they
therefore are placed at a further psychological distance with re-
spect to the reference point (Liberman and Trope, 2008; Trope
and Liberman, 2000, 2003). In contrast, thoughts closer to the
reference point of the self, here and now, are more concrete and
formulated at a lower construal level and can thus be placed at
a closer distance. Furthermore, high-level (low-level) construals
are characterized by desirability (feasibility). Moreover, the shift
from high- to low-level construals and from desirability to feasibil-
ity may result in somewhat inconsistent intertemporal decisions
(Trope and Liberman, 2003; Zhao et al., 2007). Whether the per-
ceived immediate costs outweigh future benefits can also explain
the inconsistency in intertemporal choices. When viewed from a
distance, the benefits outweigh the costs, but when the moment to
make a choice arrives, the costs appear larger than the anticipated
costs, resulting in changed preferences (Eyal et al., 2004). Typically,
future benefits loom larger than future costs, but in the present, the
reverse is true (Eyal et al., 2004), implying that savings in the short
run represent a cost – rather than a benefit – for the future (Lynch
and Zauberman, 2006).
CLT holds that greater psychological distance is associated with
people’s abstract beliefs and core values. The theory predicts that
concrete construals have a higher probability of implementation
because they are considered more feasible and attractive than
abstract construals (Liberman et al., 2007). Using our previous
example of sustainable production, we note that people often
state their support for sustainable products. However, they do not
purchase such products when they are confronted with short-term
choice architectures because feasibility prevails over desirability.
This example illustrates the inconsistency between stated prefer-
ences and real choices.
Similarly, we can argue that long-term investment prefer-
ences characterized by social responsibility and social impact traits
formed at a higher construal level are considered desirable and
less feasible; therefore, they are located further from and contrast
with the financial aspects of pensions. In Fig. 1, we show the
psychological distance of several investment characteristics in a
two-dimensional map: social and temporal. Impact investment
criteria, such as healthy aging, and socially responsible criteria,
such as sustainability, are depicted at a higher psychological dis-
tance than financial return criteria because they are assumed to be
more abstract construals. Choice inconsistency will emerge when
people who favor socially responsible criteria choose an invest-
ment portfolio based on financial criteria shifting, which itself is
based on their initial preference. In such a case, our selfish short-
term interests seem to be at odds with our collective long-term
interests.
In this study, which is based on a discrete-choice modeling
framework, respondents consider trade-offs between investment
criteria such as impact investment (ImpI), socially responsible
investment (SRI), and the additional cost in the selection pro-
cess of a socially responsible investment portfolio, given that the
opportunity to make such a decision. In the research design of
the present study, we argue that even if responsible investments
achieve worse, the same, or better financial returns as conventional
investments, pension beneficiaries must pay the extra cost stem-
ming from the screening process, ceteris paribus (i.e., maintaining
the portfolio’s risk/return profile). The framework to elicit indi-
vidual preferences is based on random utility theory (McFadden,
1986; Louviere et al., 2000). Empirical data for this study were
collected from individuals who work in the care and welfare sector
in the Netherlands by using a web-based discrete choice survey.
Each survey instrument contained 10 choice profiles and compared
3 attributes of 2 responsible investment portfolios. The attributes
varied from 3 to 10 levels according to a full profile choice-conjoint
Fig. 1. Investment levels and psychological distance.
design. The results of the analysis of the Hierarchical Bayesian (HB)
regression models show the relative effect of the ImpI criteria, SRI
criteria, and costs on the selection of a responsible investment
portfolio.
Decker and Trusov (2010) postulate that conjoint analysis rep-
resents a widespread class of methods for eliciting preferences—
particularly for identifying and evaluating new product concepts.
Rietjens (2011) employs a conjoint model to determine the pen-
sion system preferences of pension fund participants. The results
of her analysis indicate that pension fund participants place the
investment performance attribute second in importance to the
coverage ratio index. Parient (2011) uses a choice-based conjoint
(CBC) framework and a conditional logit model to estimate the
demand for microcredit in Serbia. Accordingly, we followed a
discrete CBC framework using Sawtooth Software. We use this
method because it closely emulates the real circumstances under
which the decision-making process occurs. Furthermore, a discrete
choice-based design results in less noisy data than a traditional
conjoint design that use ratings, and because of a few attribute
combinations, this design was preferred to more advanced designs
such the adaptive conjoint analysis. Note that because this study
utilized the Sawtooth Software’s CBC analysis, each respondent
did not see the same combination of attributes and levels. In
this computerized experiment, 300 versions of the questionnaire
were created. Our sample consisted of 3600 members working
in the Dutch healthcare sector, and we received 334 completed
questionnaires in return.
Our study contributes to the literature of socially responsible
investment (SRI) and provides insight into the literature on impact
investments. We build a hypothetical investment portfolio that
comprises investments that fulfill certain SRI and impact crite-
ria, and we investigate the utility level that pension beneficiaries
gain from each level and attribute to this socially responsible
investment product. We know little about pension beneficiaries’
preferences related to responsible and impact investments. We
explore this issue by calculating utility coefficients that reflect
the relative influence of the 17 attribute levels of this study with
regard for the decision to invest in a SRI portfolio. Furthermore, we
derived importance scores that reflect the sensitivity of pension
beneficiaries’ choices to variations in the levels of each attribute.
Finally, this study examines whether pension beneficiaries would
sacrifice financial efficiency by contributing a small part of their
expected pension income in exchange for investments that add
social value and whether they are capable of making consistent
 G. Apostolakis et al. / Journal of Behavioral and Experimental Finance 17 (2018) 83–96
choices with regard to investment decisions that affect their well-
being.
Hence, in this experimental empirical study we assess the
tradeoffs for ImpI criteria, SRI criteria, and costs on the selection of
a responsible investment portfolio process using a market-utility-
based approach. Specifically, we address the following research
issues:
Research Objective 1: How do pension fund beneficiaries, trade-
off between socially responsible, impact investment and cost di-
mensions when choosing an investment portfolio, given the avail-
ability of such a choice?
Research Objective 2: What is the relative importance of impact
and socially responsible investment criteria when choosing an
investment portfolio, given the availability of such a choice?
Research Objective 3: How do trade-offs in the investment
portfolio selection process vary across diverse segments of the
pension fund beneficiaries?
The remainder of the study is organized as follows. After the
introduction, we introduce the method of choice-based conjoint
analysis and our sample. Our findings regarding preferences for
a responsible portfolio follow. In the final section, we discuss our
findings and conclude.
2. Research methodology
To explore the above research objectives, we employ qualitative
(focus groups) and quantitative (CBC analysis) methods.
2.1. Focus groups
First, as a pre-study (before the CBC analysis), we conducted fo-
cus group discussions to identify the levels and attributes of the SRI
portfolio. Focus groups are important for understanding people’s
perceptions of product attributes and levels. We conducted three
semi-structured focus group sessions stratified by age (above 45
and below 45) and occupation (members and pension managers)
and consisting of 24 individuals overall. The focus group discus-
sions elaborated on pension beneficiaries’ increased involvement
through the selection of an investment portfolio and its appropri-
ate characteristics. In the discussions, 4 attributes (ImpI, SRI, social
initiatives, and costs) were identified as important for inclusion
in the SRI portfolio. The attribute of social initiative was excluded
after the pilot study to increase simplicity and reduce the difficulty
of the tasks.
2.2. CBC experimental design
We use CBC analysis because it overcomes several of the limi-
tations of traditional conjoint analysis. First, CBC analysis mimics
human behavior in a choice process. Second, we can include a none
option to indicate a decrease in demand if a respondent finds the
presented choice combination unattractive. Third, we can calculate
interactions. Fourth, we can calculate part-worth utilities at the
individual level by using the Hierarchical Bayes (HB) estimation.
In this experimental choice-based conjoint model, we showed
numerous cards/tasks. Each card contained two responsible in-
vestment profiles from which participants must choose the most
preferable (or the alternative of none). The participants were pre-
sented with 10 comparisons and two hypothetical profiles of a
SRI portfolio and were then asked to choose the preferred profile
between the two alternative profiles. Including a none alternative
option also helped us identify the point at which the investment
portfolio becomes indifferent to members (Bakken and Frazier,
2006). Other scholars advocate for the use of a none option because
it renders the experiment more realistic (Vermeulen et al., 2008).
Finally, to determine the reliability and validity of our model, we
85
used two holdout tasks in the design of the CBC model. In Fig. 2,
we present an example of the CBC card used. The first line includes
impact investment characteristics, the second line presents SRI
characteristics, and the third line presents the extra cost of a SRI
portfolio.
The design of the CBC model is based on a randomized de-
sign method. We distributed versions of the survey based on an
algorithm that maximized the level balance, orthogonality and
efficiency, and each participant randomly received a version of the
CBC questionnaire out of a pool of 300 versions of the question-
naire included in the design file (Sawtooth Software, 2008). Thus,
we created 300 different cards containing different combinations
of attribute levels, as shown in Fig. 2, and participants typically
received different versions of CBC tasks.
The discrete choice-conjoint experimental design involved a
full-profile design. We implemented a modest overlap design
(balanced overlap), which balances between efficiency and actual
choice and leads to more precise estimates than a minimal overlap
(complete enumeration) if interaction effects arise. Furthermore,
such a design better simulates the choice process (Orme, 2009).
The relative D-efficiency measure indicates that a balanced overlap
design is 96% efficient compared to complete enumeration.1
2.3. Analysis
For the CBC analysis, we used the HB method, Sawtooth Soft-
ware SSI v.8.3.10 and Sawtooth Smart v.4.23. For the descriptive
analysis, we used STATA v.13.
2.4. CBC measure
2.4.1. Responsible investment attributes
We employ CBC/HB analysis to compute the utility coefficients
to examine the relative influence of different socially responsible
investment characteristics on pension beneficiaries’ utilization de-
cisions. These utilities reflect the relative influence of the study’s
attributes on the decision to invest in SRI portfolios. Furthermore,
we investigate which investment attribute is the most preferable
and which investment asset allocation maximizes participants’
utility. Finally, we use latent class analysis to identify different
segments of pension beneficiaries whose utilization choices were
influenced by different attributes.
Participants select their portfolio of preferences based on 3
attribute combinations. In turn, the attributes consist of 4 ImpIs
(i.e., healthy aging, working conditions, living conditions, and med-
ical innovation and research), 10 responsible investments accord-
ing to SRI criteria of positive (e.g., sustainable investments, envi-
ronmental friendly investments, good corporate governance) and
negative screening (exclusion of non-social ethical investments
such as, for example, the exclusion of investments in firms related
to the arm industry), and 3 levels of cost (the willingness of respon-
dents to accept a somewhat lower pension per month, i.e., −e21,
−e26, −e31). Table 1 lists the selected investment attributes, their
levels, and their classification into 3 broad conceptual categories:
ImpI, SRI, and costs. We include the following 3 attributes, which
are discussed in detail below:
Impact investments. The literature on impact investing is still
nascent. Using the following definitions, we attempt to clarify
the distinction between ImpI and SRI. According to a Rockefeller
Foundation report, ImpI is defined as an investment intended to
create a positive impact beyond financial returns (O’Donohoe et
al., 2010). The GIIN network gives the following definition: ‘‘Impact
1 The D-efficiency measure was 917.585 and 877.547 for the complete enumer-
ation and balanced overlap designs, indicating a 4% loss in efficiency.
86 G. Apostolakis et al. / Journal of Behavioral and Experimental Finance 17 (2018) 83–96

Fig. 2. Example of a pair-wise choice-based conjoint card used in the final study. Note: Here are two socially responsible investment packages you may like. Choose by
clicking one of the buttons.

Table 1
Choice-based conjoint attributes and levels.
Attributes Levels (Alternatives)
1. Invest in healthy aging, healthy living and wellness
2. Invest in improved working conditions
1. Impact investments
3. Invest in improved living conditions
4. Invest in medical innovation and in medical research
1. Invest in firms committed to sustainability
2. Invest in firms that have environmental concerns, a strong
commitment to recycling and waste reduction practices
3. Invest in firms that have strong employee relationships
4. Invest in firms that seek to be active in the local community by
2. SRI
sponsoring charitable donations and employee volunteerism
5. Invest in firms that promote standards for human rights
6. Exclude investments in firms related to the weapon, alcohol, tobacco
and gambling industries
7. Exclude investments in firms that experiment on animals
8. Exclude investments in firms that are involved in human rights
violations
9. Exclude investments in pornographic magazines and in related
material
10. Exclude investments in firms related to nuclear power
1. −e21 of your occupational pension per month
3. Cost 2. −e26 of your occupational pension per month
3. −e31 of your occupational pension per month

investments are investments made into companies, organizations,
and funds with the intention to generate measurable social and
environmental impact alongside a financial return’’.2 According
to novel reports, ImpI can be differentiated from SRI (Martin,
2013; Social Finance Finethic, 2012). Impact investors are more
proactive in implementing ImpIs to achieve certain social goals.
They are interested in prioritizing social returns to maximize both
financial efficiency and financial return. In our design, investments
in alternative asset classes include targeted investments related
to healthy aging, medical innovation and improved working and
living conditions.
Socially responsible investments. Capelle-Blancard and Monjon
(2012, p. 239) briefly define SRI, stating that ‘‘Socially responsi-
ble investing (SRI), also known as ethical investing, refers to the
integration of environmental, social, and corporate governance
considerations (ESG) into the investment process’’. In recent years,
institutional investors such as pension funds and large investment
banks have begun to consider the criteria of social responsibility
and sustainability in examining their investments. The design of
their pension fund policy and investment strategy is based on
2 http://www.thegiin.org/cgi-bin/iowa/resources/about/index.html.
investment beliefs that usually include such criteria. Additionally,
investing according to SRI principles is inherent to the core value
of cooperatives. The goal of social value creation is contained in
cooperative principles, meaning that cooperatives can internalize
value added and member benefits when they adopt social invest-
ment beliefs (Novkovic, 2008).
Nilsson (2009, p. 8) argues that although the financial per-
formance of SRI mutual funds is ambiguous, ‘‘the most common
result when comparing SRI and ‘regular’ investments is that there
is no significant difference in financial performance’’. Two studies
indicate that investors also care about non-financial attributes
because of the variation in money flow in the funds. Bollen (2007)
shows that the flow of money is less sensitive to lagged negative
returns in SRI funds than that in conventional funds. Renneboog et
al. (2008b) find that SRI funds do not underperform conventional
funds, except in some countries. Renneboog et al. (2011) argue
that past performance sensitivity depends on funds’ SRI screen-
ing activities. Friede et al. (2015) provide recent evidence from
a literature review analysis that there is a non-negative relation
between both ESG factors and ESG criteria and corporate financial
performance.
Furthermore, empirical research examining SRI preferences us-
ing conjoint analysis shows that financial attributes are most often
 G. Apostolakis et al. / Journal of Behavioral and Experimental Finance 17 (2018) 83–96 87

preferred to SRI attributes (Brimble et al., 2013; Rietjens, 2011; Table 2

Vyvyan et al., 2007). Vyvyan et al. (2007) find that all investors, Sample demographics.

regardless of whether they are described as environmentalists, Variable Obs. Mean Stand. dev.
ranked financial performance-related criteria as most important. Age 334 58.177 11.205
Additionally, their results provide evidence to support an attitude– Gender (Men–Women) 334 1.623 0.485
behavior gap in relation to environmentalist attitudes and invest- Education (1–6) base = Primary education
ment decision making. VMBO/MAVO/LBO 334 0.066 0.248
MBO 334 0.254 0.436
SRI commonly uses two types of investment screening: nega- HAVO/VWO 334 0.066 0.248
tive and positive screening. According to Renneboog et al. (2008a), HBO 334 0.476 0.500
negative screening refers to the practice of excluding stocks or University 334 0.132 0.339
industries from SRI portfolios based on social, environmental and Marital status (1–5) base = Married
ethical criteria. In contrast, positive screening targets the best firms Divorced 334 0.054 0.226
according to both market share and criteria related to good corpo- Living together 334 0.06 0.238
Widowed 334 0.063 0.243
rate governance, labor relations, the environment, the sustainabil- Single 334 0.129 0.335
ity of investments, and the stimulation of cultural diversity. In our Number of children 334 1.422 1.348
CBC model, we include one attribute for responsible investments Living condition (1–5) base = Own house
that consists of ten levels: five positive screenings and five negative Rental house 334 0.165 0.371
screenings. To make the selection more convenient, we provide a Free accommodation/parents’ home 334 0.012 0.109
detailed description for each type of screening. Other 334 0.009 0.094
Retired (Yes–No) 334 1.644 0.480
Additional cost of the SRI portfolio. The final attribute of our CBC Working years 334 32.769 11.434
Working hours 334 31.305 12.015
model is the additional administrative cost of providing a SRI
Net HH income (1–11) 334 5.093 2.586
portfolio. Introducing individual choices such as socially respon- Intention to invest in SRI (1–7) 334 3.781 1.935
sible portfolios to the pension system will increase administrative Risk tolerance (1–7) 334 2.805 1.136
costs. Furthermore, we admit that impact and socially responsible Perceived retirement age 334 64.117 4.077
investments, although they can yield returns equal to those of Perceived health (1–5) 334 3.766 0.890
conventional investments, have greater costs because of the ad- Note: Marital status: 1 = married, 2 = divorced, 3 = living together, 4 = wid-
ditional cost of screening (Borgers and Pownall, 2014; Boudt et al., owed, 5 = single. Education from the primary to university level (1 = primary,
2 = VMBO/MAVO/LBO, 3 = MBO, 4 = HAVO/VWO, 5 = HBO, 6 = university),
2013; Hong and Kacperczyk, 2009; Statman and Glushkov, 2009).
living condition, 1 = own house, 2 = rental house, 3 = subletting house, 4 =
Under such considerations, and although we note that impact and free accommodation, with parents, 5 = other. Net household (HH) income (1–
socially responsible investments’ long-term financial returns may 11) from 1 < 600 to 11 > 6000 (≤e600, e601–e1200, e1201–e1800, e1801–
be even higher than those of conventional investments, to measure e2400, e2401–e3000, e3001–e3600, e3601–e4200, e4201–e4800, e4801–
individuals’ utility gain from including non-financial attributes in e5400, e5401–e6000, >e6000). Perceived health from 1 = poor, 2 = not so good,
3 = fair, 4 = good, 5 = very good. Risk tolerance from 1 = lower to 7 = higher risk
an investment portfolio, the additional administration costs for tolerance (Scale of 5 items, Cronbach’s alpha = 0.7276).
the responsible portfolio are translated into lower pensions for
beneficiaries.
To make it easier for respondents to comprehend this trade-off, The survey sample consists of 3600 members of the cooper-
financial returns are translated into average net pension income ative. To reduce non-response rates, 3600 invitations were sent
per month. Therefore, the respondents must state their preference three times. Overall, 334 web-based questionnaires were obtained.
for a specific pension-income combination. We consider that the
The respondents were 62% female, and the age of the respon-
net pension in the health care sector is an average of e390 per
dents ranged from 23 to 80 years old, with an average age of
month, with the AOW state pension excluded. One-third of the
58.2. Therefore, although the respondents are not representative
net pension is the paid premium (e130), whereas two-thirds (up
to e260) come from investment revenue. For the purpose of this of the general population, the sample is more representative of
experiment, we argue that investing in a responsible portfolio the pension fund’s population. In Table 2 we present the study’s
while providing choices for pension beneficiaries will increase sample demographics. We employ house ownership as a proxy for
the transaction costs and reduce the investment revenue by 10% wealth. The majority of respondents are homeowners. Measuring
(−e26), resulting in a reduction of 0.024% in the expected pen- risk tolerance in a scale of 1 to 7, with higher values to indicate
sion, including the state pension (e1100 − e26 = e1074), ceteris more risk appetite, we find that on average, respondents wish to
paribus. Complementary to this anchor price, we include 8% and avoid risky choices (2.805).
12% deductions in investment revenue, representing 5 euros below We measure psychological distance in three dimensions (social,
and above the anchor price, respectively. The discrete choices are temporal, and spatial) for impact investments, SRI, and finan-
thus −e21, −e26, and −e31. cial return. Higher values of psychological distance indicate that
respondents perceive investments to be further in distance. For
3. Sample and data example, a semantic differential question – i.e., if I invest only in
medical innovation and research, this will have mainly a positive
In this study, we examine investment preferences in the context
effect for myself vs. mainly a positive effect for others – measures
of PGGM, a Dutch cooperative in the financial sector and the asset
the social distance of this impact investment. In Table 3, temporal
manager of the second-largest Dutch pension fund for employees
distance of medical innovation and SRI positive investments have
working in the field of care and well-being (PFZW). PFZW has the
legal status of a non-profit foundation and is part of the Dutch the greatest average psychological distance. In contrast, social
collective pension system. Following a division of responsibilities financial return has the lowest average psychological distance. In
that occurred in late 2008, PFZW addresses pension policy and general, the financial variables of pension and financial return have
PGGM manages the fund’s assets. PGGM members work in nursing, lower psychological distance than the impact investments and SRI
home care, hospitals, special care, and mental care. As a result, 73% in all corresponding psychological distance (PD) dimensions.
of the company is composed of women, 47% of whom are between Next, we construct an aggregate measure of PD for impact
51 and 65 years old. investment, SRI, and financial return to further explore the PD.3
88 G. Apostolakis et al. / Journal of Behavioral and Experimental Finance 17 (2018) 83–96

Table 3
Psychological distance characteristics.
Variable Obs. Mean Stand. dev.
Impact investments
PD →Social distance healthy aging 334 4.126 1.413
PD →Temporal distance healthy aging 334 4.784 1.440
PD →Spatial distance healthy aging 334 4.093 1.356
PD →Social distance medical innovation 334 4.716 1.418
PD →Temporal distance medical innovation 334 4.973 1.511
PD →Spatial distance medical innovation 334 4.509 1.360
PD →Social distance working conditions 334 4.775 1.721
PD →Temporal distance Working conditions 334 4.620 1.541
PD →Spatial distance working conditions 334 4.566 1.609
PD →Social distance living conditions 334 3.967 1.657
PD →Temporal distance living conditions 334 4.506 1.551
PD →Spatial distance living conditions 334 4.171 1.466
SRI
PD →Social SRI positive 334 4.769 1.466
PD →Temporal SRI positive 334 4.967 1.537
PD →Spatial SRI positive 334 4.677 1.430
PD →Social SRI negative 334 4.440 1.854
PD →Temporal SRI negative 334 4.620 1.793
PD →Spatial SRI negative 334 4.719 1.658
Financial return
PD →Social pension 334 3.808 1.620
PD →Temporal pension 334 4.269 1.557
PD →Spatial pension 334 4.060 1.502
PD →Social financial return 334 3.183 1.680
PD →Temporal financial return 334 3.952 1.765
PD →Spatial financial return 334 3.617 1.488
PD aggregate 3-Factor (Social, Temporal, and Spatial)
PD →Impact investments 334 4.484 1.007
PD →SRI 334 4.699 1.232
PD →Financial return 334 3.815 1.206
Note: PD = psychological distance from 1 = lower distance to 7 = higher distance; see questionnaire in the appendix.

However, this does not mean that we perceive PD as a one-
dimensional notion. Instead, PD aims to help us to explore and
better describe its relation to different investment attributes. The
SRI has the highest average distance (4.699), followed by impact
investing (4.484) and financial return (3.815), which has the lowest
PD. That means that respondents perceive SRI as closer to their
inherent ethical values than financial return.
4. Analysis and results
4.1. Discrete choice conjoint analysis
4.1.1. Validity and reliability
First, we examine the reliability and validity of our model.
The median time to complete the survey was 15.9 min. Of the
334 respondents, 15.5% chose the none option for all 10 tasks,
indicating an absolute indifference to the SRI product.
In this study, as in other studies in the field, (Cerjak et al., 2010;
Klein et al., 2010), we account for the face validity, content validity,
convergent validity, and predictive validity of the conjoint choice
tasks.
Face and convergent validity focus on how the tasks are op-
erationalized and whether they sufficiently reflect the attribute.
Face validity measures whether the results of the conjoint analysis
(part-worths) correspond with experts’ expectations, for example,
whether the results make sense in view of the available theoretical
and empirical knowledge about the market place. We expect our
respondents to prefer a lower instead of a higher reduction in
their pension. However, we cannot establish expectations for the
SRI or ImpI attributes. Content validity examines the plausibility,
completeness and adequateness of an analysis (Klein et al., 2010).
3 Cronbach’s alphas for the aggregate measure of PD of impact investment, SRI
and financial return are 0.8873, 0.8493, and 0.8455, respectively.
One way to analyze the content validity of a conjoint analysis is
to calculate the average relative importance of each attribute and
to validate whether the results meet a priori expectations. Ac-
cordingly, we expect that investment product one will be rated as
more important than product two because we expect the selected
attributes to be more desirable for product 1.4
The criterion-related validity is usually split into concurrent
validity and predictive validity. Concurrent validity measures how
well the estimated values reflect the input data. The correlation
coefficients of Pearson’s r and Kendall’s tau are used to measure
concurrent validity (Klein et al., 2010). Each participant completed
two ‘‘holdout’’ choice tasks that were not included in the utility
estimates (Orme, 2009). Data from the 8 remaining choice tasks
were used to predict the responses to the holdout tasks. We mea-
sured the concurrent validity by assessing whether the Pearson’s
correlation coefficient and Kendall’s tau between the two tasks are
positive and significant (0.74, 0.71). The t-test of the difference in
means revealed a p-value of 0.063, indicating that the difference
between the two means is significant at a 10% level of significance.
Predictive validity measures the extent to which the estimated
results overlap with participants’ real purchase intentions. The
first-choice-hit rate employed in this research is one of the criteria
that are very often used to measure predictive validity (Brusch
et al., 2002; Cerjak et al., 2010; Vriens et al., 1998). First-choice
models are available for disaggregate data and follow the maxi-
mum utility rule. In other words, if three products are included in
a scenario, each individual is assumed to choose the product that
maximizes utility.
The hit rate examines whether the real purchasing decision
matches the predicted purchasing decision. Hit rates are calculated
by comparing the model’s prediction of an individual respondent’s
4 The two products included in the holdout tasks are as follows: product 1 =
invest in healthy living + sustainability − e21; and product 2 = invest in improved
working conditions + good corporate governance − e31.
 G. Apostolakis et al. / Journal of Behavioral and Experimental Finance 17 (2018) 83–96
choice (based on the maximum utility rule) with the respondent’s
actual choice (McCullough, 2002). The hit rate equals the total
number of hits divided by the total sample size. The hit rate is given
by
HR = /
Dcorrect D
total . illustrate the results of the HB analysis. The part-worth estimates
The hit rates for holdout cards are 65.3% and 61.3% for the first
and second holdout cards, respectively. We thus observe that the
correct predicted choices are lower in the second holdout task.
The average hit rate considering both holdout tasks is 63.3%. Initial
hit rates can be compared with hit rates from a random model
to measure how successfully the model has been able to capture
respondents’ choices. In our case, there are three alternatives in
each holdout task. Therefore, a random model could be expected
to have a hit rate of 33%, which is lower than the 63% hit rate of our
model.
Another measure of predictive validity is the mean absolute
error (MAE). MAE is defined as the sum of the differences between
the predicted share of preferences and the actual share of prefer-
ences for all products in a holdout task divided by the number of
products in the holdout task.
∑ ⏐⏐ ⏐
MAE = i ⏐Di −D̂i ⏐/VS
⏐ engaging in impact investing (Kruif and Ipenburg, 2015). According
where Di is the respondent proportion of the validation stimulus
I, D̂i is the estimated proportion of the validation stimulus I, and
VS is the number of validation stimuli. Lower MAE values indicate
higher predictive validity. For the first and second holdout cards,
the MAE is 2.15 and 4.24, respectively, whereas across the 2 con-
cepts, the estimated MAE is 3.19.
We measure reliability (internal consistency) by repeating the
same card but in reverse order. The sixth- and tenth-choice tasks
in the web-based survey were identical (but the concepts were
rotated). The two holdout cards were used both to test the pre-
dictive ability of the models and to calculate the test–retest reli-
ability. The test–retest reliability rate is 81%, indicating that the
respondents answered these holdouts in the same way 8 of 10
times on average; i.e., given that their choices were completely
reliable, this measure indicates the proportion of respondents who
gave the same response in the two holdout tasks. Another measure
of internal consistency, Cronbach’s alpha, is 0.85. We follow Bryan
et al. (2000) and use another measure of reliability, the kappa
coefficient, which measures the reliability of nominal variables.
It is generally agreed that a kappa coefficient above 0.6 indicates
good agreement. The number of responses showing agreement is
80.5%, whereas the number of responses showing agreement that
would have been expected by chance is 35.4%. Thus, the kappa
statistic is approximately 0.7. The amount of agreement indicates
that we can reject the hypothesis that respondents made their
determinations randomly (see Table 4).
4.1.2. CBC/HB conjoint results
Calculating the HB model yielded a root likelihood (RLH) of
0.715, which is close to the best possible value of 1. RLH is an
indication of how well the data fit the model. Table 5 shows
the utilities that members attribute to our socially responsible
portfolio. The percentage certainty is 0.694, whereas the average
variance is 9.121. Moreover, the hit rate is 63% and the MAE is 3.19.
We provide the average utilities (zero-centered differences) in our
tables. We use this normalized scale to place each respondent on
the same scale so that each respondent is given essentially equal
weight when averaging the scores for groups or for the sample
(see Appendix). To test whether the main effect of an attribute
had any significant (different from zero) levels, we provide 1%, 5%
and 10% levels of significance using a Bayesian approach (see in
the Appendix). For example, investing at the healthy aging and
89
retirement attribute-level, we are 99% confident that this level for
this main effect is significantly larger than zero.
Part-worth utility values (β ) represent the relative influence of
each criterion level on respondents’ choices, where higher values
indicate a stronger influence on choices. In Table 5, Panel A, we
show that for the ImpI and SRI attributes, both the utility maximiz-
ing levels and the most preferred levels are investing in educating
people in healthy aging and investing in firms with environmen-
tal concerns (β = 12.321, SD = 29.645; β = 28.608, SD =
34.327). The least preferred levels are improving living conditions
and excluding investment in nuclear power (β = −3.969, SD =
44.915; β = −40.099, SD = 39.825). After retirement, investing
in healthy aging is 21.056 utiles higher than investing in improved
living conditions. The attribute of cost is negative and significant,
indicating that higher prices reduce the utility gained from the SRI
product.
Dutch institutional investors follow exclusion and ESG-
integration strategies more often than other investors to facilitate
positive selection and affect investing strategies (Wagemans et al.,
2014). However, impact investing has gradually piqued the interest
of pension funds. Moreover, Dutch pension funds are increasingly
to another recent report, 1.8% of the total assets under manage-
ment of Dutch pension funds are allocated to impact investments
(Duiker et al., 2016). This report suggests that the majority of im-
pact investments made by Dutch pension funds belongs to public
equity. In addition, Dutch institutional investors allocate almost
half of their impact investments to green technology. Improving
access to finance, such as investing in microfinance, constitutes
almost one-third of their impact investments, and improving ac-
cess to basic services such as health and water represents a 15%
share of the total impact investments made by Dutch institutional
investors.
PGGM employs exclusions, ESG integration, and impact invest-
ing as instruments for its responsible investment strategy. Ac-
cording to a SRI report by PGGM, the organization has invested
approximately e8.9 billion, or approximately 4.9% of its total assets
under management, to investments that aim to add social value.
The largest share of PGGM’s impact investments has climate and
environmental themes, such as climate impact, solar and wind
energy, and CO2 -efficient buildings and production. Investing in
solutions related to food (efficient food production, food wastage),
health (medicine, treatment, care homes) and water (water purifi-
cation, clean water conservation) are less common than climate
and environmental investments. Such an outcome, results from
the available responsible investment opportunities that satisfy
the fund’s risk/return profile. PGGM also affects Dutch society by
investing locally, which has already had a measurable impact on
society, for example, improving access to good healthcare and
improving the production of renewable energy. Regarding SRI
criteria, PGGM excludes mainly companies that are involved in
controversial weapons, human rights violations, environmental
pollution and tobacco production.
From our results and the pension fund’s socially responsible
report, we observe that the pension funds and its members share
common views on responsible investment criteria such as environ-
mental technology, environmental concerns, working conditions,
human rights violations and controversial weapons. PGGM invests
in the themes of food and health impact investments, for example,
in firms devoted to healthy nutrition and healthcare solutions,
respectively. Although these investments are in line with member
preferences, this policy needs to be strengthened because healthy
aging and retirement are found to maximize members’ utility
regarding impact investments.
Panel B shows the importance ratio of the portfolio attributes.
The highest importance is afforded to the SRI attribute at 52.6%,
90 G. Apostolakis et al. / Journal of Behavioral and Experimental Finance 17 (2018) 83–96

Table 4
Validity and reliability check. Portfolio shares of preference holdout cards.
Simulation Share Standard error
Portfolio 1 25.97 1.78
Portfolio 2 44.76 2.10
None 29.27 2.27
Actual Holdout card 1 Holdout card 2
Portfolio 1 28.44 32.34
Portfolio 2 45.51 44.01
None 26.05 23.65
Holdout card 1 Holdout card 2 Total
Predictive validity
Hit rate 64.7% 61.7% 63.2%
MAE 2.15 4.24 3.19
Internal (concurrent) validity
Test–retest reliability 81%
T -test = −2.166 Difference = −0.063**
Interclass correlation rho 0.744*** Kendall tau-b 0.711***
Cronbach’s alpha 0.853
Agreement Expected agreement Kappa Standard error
80.54% 35.39% 0.699*** 0.039
Note: Estimated method: Randomized first choice. Portfolio 1 comprises healthy aging, sustainability and e21. Portfolio 2
comprises working conditions, good corporate governance and e31 euros. For the 2 holdout cards, portfolio 1 comprises
educating people, sustainable investments and e21, and portfolio 2 comprises improved working conditions, good
corporate governance and e31. * p < 0.10, ** p < 0.05, *** p < 0.01.

Table 5
Choice-based conjoint/HB.
Panel A: Average utilities (zero-centered diffs) Average utilities Stand. dev.
Impact investments
Invest in educating people to remain healthy and active for a healthy 12.321*** 29.645
aging
Invest in better working conditions in the care and welfare sector 0.382 33.669
Invest in medical innovation and in research for new cures and −3.969 44.915
treatments
Invest in improved living conditions after retirement −8.735*** 28.797
SRI
Environmental concerns inclusion 28.608*** 34.327
Weapon industry exclusion 24.499*** 44.236
Sustainable investments inclusion 23.073*** 33.541
Community involvement inclusion 21.813*** 51.19
Bad HR treatment exclusion 10.265* 26.342
Good HR treatment inclusion 5.251 34.622
Good corporate governance inclusion 2.168 42.978
Ethical consideration exclusion −26.957*** 57.682
Nuclear power exclusion −40.099*** 39.825
Animal experiment exclusion −48.621*** 44.795
Cost −3.010*** 6.163
None −20.837*** 268.084
Panel B: Average importance Average importance Stand. dev.
SRI 52.608 12.61
Impact investments 28.744 10.1
Cost 18.648 13.199
Note: Preliminary iterations = 10.000, Draws used per respondent = 1.000, Skip factor for draws used = 5, Total
number of iterations = 15.000, n = 334. Perc. certainty = 0.694 Aver. perc. certainty = 0.700, RLH = 0.715, Aver.
RLH = 0.719, Average variance = 9.121, Average aver. variance = 8.534, Parameter RMS = 3.037, Aver. parameter
RMS = 3.064, Hit rate = 0.63, MAE = 3.19. Including age as covariate. * , ** , *** indicate that the parameter of the
corresponding variable significantly influences respondents investing preferences and give the probability of 90%,
95%, or 99% of utilities, respectively, out of 1000 iterations being above (or below) zero.

followed by ImpI at 28.7% and cost at 18.7%. The results of our
HB analysis can be interpreted to indicate both that members are
willing to sacrifice some financial efficiency for a new way of in-
vesting that is more socially oriented and that the most important
consideration in their choice of SRI product is the SRI attribute,
followed by the ImpI attribute, and then by the cost.
4.1.3. Latent CBC segmentation analysis
We estimate the discrete-choice conjoint model by using the
latent CBC model. To identify the appropriate number of segments,
we based the model on the CAIC and BIC. The CAIC is used to
determine the optimal number of segments, and lower values
are preferred. According to Table 6, the number of segments that
minimizes the CAIC is 3. These three segments were labeled ‘‘On
the fence’’ (20.6%), ‘‘Materialistic’’ (18.7%) and ‘‘Advocates’’ (60.7%).
In Tables 7 and 8, we present the segment characteristics
according to the latent class analysis. A Pearson’s χ 2 test or an
F -test using analysis of variance (ANOVA) indicate that gender,
retired status and income level vary significantly among the three
segments. Membership in the first segment is more characterized
by female gender and non-retired status. This segment is also
characterized by younger age and lower average income than the
other segments. Membership in the second segment is charac-
terized by higher level of income and more working hours on
 G. Apostolakis et al. / Journal of Behavioral and Experimental Finance 17 (2018) 83–96 91

Table 6
Summary of best replications with a total of 5.
Groups Repl. Log-lik. Pct. cert AIC CAIC BIC ABIC χ2 Rel. χ 2
2 5 −2191.48 25.35 4440.96 4640.78 4611.78 4519.64 1488.03 51.31
3 4 −2106.77 28.23 4301.54 4604.73 4560.73 4420.93 1657.44 37.67
4 1 −2080.83 29.11 4279.66 4686.20 4627.20 4439.74 1709.33 28.97
5 5 −2061.51 29.77 4271.02 4780.92 4706.92 4471.80 1747.97 23.62
Note: Summary of best replications.

Table 7
Sample characteristics.
Variable Group 1 n = 67 Group 2 n = 64 Group 3 n = 203 Pearson’s χ 2 /F -test
Mean Std. dev. Mean Std. dev. Mean Std. dev.
Age 57.060 10.220 60.344 8.934 57.862 12.081 1.62
Gender (Men–Women) 1.687 0.467 1.453 0.502 1.655 0.476 9.91***
Education (1–6) base = Primary education 0.015 0.122 0.016 0.125 – – 9.913
VMBO/MAVO/LBO 0.119 0.327 0.047 0.213 0.542 0.227
MBO 0.209 0.410 0.203 0.406 0.286 0.453
HAVO/VWO 0.060 0.239 0.063 0.244 0.069 0.254
HBO 0.493 0.504 0.516 0.504 0.458 0.499
University 0.104 0.308 0.156 0.366 0.133 0.340
Marital status (1–5) base = Married 0.731 0.447 0.797 0.406 0.650 0.478 12.630
Divorced 0.060 0.239 0.016 0.125 0.064 0.245
Living together 0.104 0.308 0.031 0.175 0.054 0.227
Widowed 0.045 0.208 0.031 0.175 0.079 0.270
Single 0.060 0.239 0.125 0.333 0.153 0.361
Number of children 1.657 1.513 1.438 1.308 1.340 1.300 11.54
Living condition (1–5) base = Own house 0.806 0.398 0.875 0.333 0.798 0.402 2.983
Rental house 0.164 0.373 0.109 0.315 0.182 0.387
Free accommodation/parents’ home 0.015 0.122 0.016 0.125 0.010 0.099
Other 0.015 0.122 – – 0.010 0.099
Retired (Yes–No) 1.776 0.420 1.531 0.503 1.635 0.482 8.71**
Working years 32.015 10.374 34.859 10.046 32.360 12.130 1.35
Working hours 30.104 13.787 35.172 13.764 30.483 10.530 4.2**
Net HH income (1–11) 4.449 2.593 5.359 2.967 5.207 2.435 2.36*
Intention to invest (1–7) 3.657 1.675 1.906 1.620 4.414 1.708 54.11***
Risk tolerance (1–7) 2.788 1.124 2.928 1.389 2.772 1.052 0.47
Perceived retirement age 64.388 3.746 63.828 3.799 64.118 4.274 0.31
Perceived health (1–5) 3.761 0.854 3.625 1.106 3.813 0.823 1.09
Note: Segment percentages, 20.6%, 18.7%, 60.7% for groups 1, 2 and 3, respectively. The highest values for each variable among the three segments presented in bold. F -stat
in bold.
*
p < 0.10.
**
p < 0.05.
***
p < 0.01.

average, whereas those belonging in the third segment have a
greater intention of investing in the SRI portfolio.
Second, to provide further insight into the characteristics of the
three identified segments, an MNL regression was estimated. To
extract the marginal effects from the estimated binary and MNL
models, the ‘‘MFX2’’ STATA command was used. Thus, all regres-
sion tables provide the results of the logistic regression analyses.
They report the marginal effects at the means (MEMs), i.e., the
effects when the rest of the independent variables are kept at their
mean values. We estimated robust standard errors, and z-values
are presented in parentheses (STATA command VCE).
The importance of the concomitant variables provides insight
into the factors affecting the differences in preferences among the
three segments. The results from the multinomial logistic analysis
suggest that positive coefficients indicate that a higher likelihood
of not being a pensioner increases the probability that a member
belongs to segment 1 (Table 9).
Having a higher level of education and being married decrease
the chance of belonging to the third segment. Increases in income
increase the probability of being in segment 3 but negatively affect
the probability of being in segment 1. Further, the estimated values
of the regression coefficients of the intention to invest in the SRI
portfolio variable are significant and positive for segment 3 but
negative for segment 2. We do not observe any wealth effect from
our proxy variable of house ownership.
Increased psychological distance for ImpI increases the proba-
bility of being in segment 3 and decreases the probability of being
in segment 1 because the aggregate psychological distance for
ImpI is negative for segment 1 and positive for segment 3. Finally,
a greater level of psychological distance from SRI increases the
chance of being in segment 1.
The examination of the importance scores and utilities across
the three segments yielded significant results. We compute multi-
variate analyses of variance (MANOVAs) across utility values and
importance scores of each segment. The MANOVAs across the
utility values F (28, 636) = 50.90, p < 0.001, Wilk’s Λ = 0.095,
partial η2 = 0.69, and importance scores, F (4660) = 3.99, p <
0.005, Wilk’s Λ = 0.953, partial η2 = 0.024 of the three segments
were significant.
The utility values in Table 10 show the relative influence of each
attribute level on participants’ choices, where higher values reflect
stronger preferences. The importance scores, ANOVAs, and F -tests
show the relative sensitivity choices to variations in the levels of
each attribute.
The utility values show that the first two segments preferred
investing in healthy aging, whereas the third segment preferred
investing in improved working conditions. Moreover, ethical con-
siderations were important for the second segment, whereas sus-
tainability and environmental concerns were important for the
first and third segments, respectively.
92 G. Apostolakis et al. / Journal of Behavioral and Experimental Finance 17 (2018) 83–96

Table 8
Sample characteristics: psychological distance/groups.
Variable Group 1 n = 67 Group 2 n = 64 Group 3 n = 203 F -test
Mean Std. dev. Mean Std. dev. Mean Std. dev.
Impact investments
PD →Social distance healthy aging 3.925 1.385 3.922 1.703 4.256 1.310 2.22
PD →Temporal distance healthy aging 4.687 1.520 4.094 1.779 5.034 1.208 11.23***
PD →Spatial distance healthy aging 3.791 1.188 4.109 1.827 4.187 1.220 2.170
PD →Social distance medical innovation 4.567 1.438 4.516 1.869 4.828 1.233 1.64
PD →Temporal distance medical innovation 4.791 1.562 4.469 2.031 5.192 1.242 6.38***
PD →Spatial distance medical innovation 4.373 1.423 4.547 1.868 4.542 1.135 0.42
PD →Social distance working conditions 4.388 1.642 4.641 2.118 4.946 1.586 2.92*
PD →Temporal distance working conditions 4.134 1.546 4.344 1.945 4.867 1.338 7.22
PD →Spatial distance working conditions 4.209 1.591 4.641 2.003 4.660 1.462 2.08
PD →Social distance living conditions 3.463 1.480 4.141 1.999 4.079 1.568 3.99**
PD →Temporal distance living conditions 4.075 1.531 4.313 1.967 4.709 1.371 4.95***
PD →Spatial distance living conditions 3.821 1.302 4.391 1.981 4.217 1.306 2.76*
SRI
PD →Social SRI positive 4.896 1.383 4.594 1.982 4.783 1.294 0.72
PD →Temporal SRI positive 5.030 1.435 4.688 1.999 5.034 1.394 1.31
PD →Spatial SRI positive 4.791 1.332 4.563 1.934 4.675 1.271 0.42
PD →Social SRI negative 4.433 1.869 4.344 2.041 4.473 1.795 0.12
PD →Temporal SRI negative 4.403 1.818 4.469 1.952 4.739 1.731 1.17
PD →Spatial SRI negative 4.627 1.603 4.594 1.966 4.788 1.573 0.46
Financial return
PD →Social pension 3.716 1.433 3.500 1.799 3.936 1.611 1.91
PD →Temporal pension 4.209 1.483 3.703 1.857 4.468 1.436 6.11***
PD →Spatial pension 4.000 1.243 3.625 1.795 4.217 1.456 3.91**
PD →Social financial return 3.134 1.424 3.094 1.900 3.227 1.691 0.19
PD →Temporal financial return 3.940 1.660 3.188 1.885 4.197 1.698 8.31
PD →Spatial financial return 3.687 1.258 3.141 1.708 3.744 1.460 4.17**
PD aggregate 3-Factor (Social, Temporal, and Spatial)
PD →Impact investments 4.185 0.957 4.344 1.578 4.626 0.735 5.75***
PD →SRI 4.697 1.107 4.542 1.769 4.749 1.055 0.69
PD →Financial return 3.781 1.119 3.375 1.427 3.965 1.126 6.03***
Note: PD = Psychological distance. The highest values for each variable among the three segments presented in bold.
*
p < 0.10.
**
p < 0.05.
***
p < 0.01.

The none option was less likely to be selected by the third
segment than by the first two segments. Furthermore, the second
segment was more sensitive to price changes than the other two
segments.
The importance scores show that variations in the degree of
SRI exerts the strongest influence on participants’ choices. In line
with our results in Table 5 the highest importance is indicated by
the SRI attribute, followed by ImpI and cost. Interestingly, the cost
attribute is the least important attribute in selecting a SRI portfolio
in all segments; i.e., the participants’ choices are less sensitive to
fluctuations in the amount that pension participants must sacrifice.
The respondents in the second segment were also more sensitive
to variations in SRI. The ImpI attribute exerts a relatively greater
influence on the participants’ choices in the first and third segment
than on the choices of those in the second segment. However, dif-
ferences in the ImpI importance scores among the 3 segments were
not significant. Finally, ‘‘On the Fence’’ and ‘‘Advocate’’ participants
were more sensitive to cost variations, whereas ‘‘Materialistic’’
participants were the least sensitive to cost variations.
In Table 11, we provide the results of the simulations for the
SRI product that maximizes the utility of the three segments. In
simulation 1, we show the response of the three segments to
the investment product with maximum utility (healthy aging +
environmental concerns −e21) compared to the none option. We
find that 70.46% of the participants would prefer portfolio 1 to
a none alternative. Although we find an increase in the share of
preferences for the investment product in the third segment, the
predicted utilization of the investment product declined in the
other two segments. In simulation 2, we predict the utilization of
the investment portfolio that minimized utility (improved living
conditions + exclusion of animal experiments +e31). When port-
folio 2 is offered, it is preferred by 56.3% of the participants. Overall,
we find a 14% increase in the selection of the none option from
29.541 for portfolio 1 to 43.661 for portfolio 2. For those in seg-
ment 3, this simulation predicted that the utilization of portfolio 2
instead of portfolio 1 would decrease the share of preferences by
9.2 percentage points from 98.248 to 89.062. The demand of both
portfolios is inelastic and is less than one. Market simulations show
that a e1 increase of portfolio 1 will decrease its market share by
0.38 points, with the greatest reduction found for segment 1 (1.59),
and the least reduction for segment 3 (0.012).
Next, we estimate pension beneficiaries’ willingness to pay
(WTP) for attribute levels of the SRI portfolio. To estimate their
WTP, we use a simulation-based approach rather than estimate
WTP as a linear coefficient because the linear approach does not
reflect true WTP, and thus, it is sometimes misleading (Orme,
2001). We find that the respondents are willing to pay an extra
e5.7 to switch from investing in improving living conditions to
investing in healthy aging. Furthermore, to switch from excluding
animal experiments to addressing environmental concerns, the
respondents are willing to pay an extra e21.7. These findings
verify that the trade-offs between the ImpI levels are smaller in
magnitude than the trade-offs in SRI.
5. Discussion and conclusion
In this study, we investigated pension beneficiaries’ preferences
in a portfolio characterized by ImpI and SRI in the healthcare sector
and their willingness to sacrifice some of the financial efficiency
in their pension management. We used a discrete-choice conjoint
 G. Apostolakis et al. / Journal of Behavioral and Experimental Finance 17 (2018) 83–96 93

Table 9
Multinomial logistic regression.
Variables Group 1 (n = 67) Group 2 (n = 64) Group 3 (n = 203)
Age 0.003 0.000 −0.003
(0.718) (0.113) (−0.634)
Gender = Male 0.017 −0.068 0.051
(0.265) (−1.532) (0.650)
Education = Higher 0.084 0.053 −0.137**
(1.618) (1.618) (−2.226)
Marital status = Married 0.091 0.055* −0.146**
(1.638) (1.710) (−2.287)
Number of children 0.018 0.001 −0.019
(0.919) (0.069) (−0.816)
Living condition = House owner 0.020 0.042 −0.062
(0.300) (1.189) (−0.837)
No retired 0.184** −0.052 −0.132
(2.535) (−1.264) (−1.587)
Working years 0.001 −0.001 0.000
(0.228) (−0.509) (0.089)
Working hours 0.001 0.001 −0.002
(0.172) (1.021) (−0.606)
Net HH income (1–11) −0.026** −0.005 0.031**
(−2.400) (−0.829) (2.539)
Risk tolerance (1–7) 0.010 0.017 −0.027
(0.415) (1.356) (−1.000)
Perceived health (1–5) −0.013 −0.031 0.045
(−0.416) (−1.516) (1.283)
Intention to invest in SRI (1–7) −0.022 −0.081*** 0.103***
(−1.470) (−6.652) (5.713)
PD →Impact investments −0.092*** 0.001 0.091**
(−2.924) (0.059) (2.520)
PD →SRI 0.051** −0.012 −0.040
(2.139) (−0.750) (−1.392)
PD →Financial return −0.007 −0.004 0.011
(−0.352) (−0.304) (0.445)
Pseudo R2 0.2389
Observations 334
Note: Robust z-statistics in parentheses. The outcome of the logistic regressions is the marginal effect at the means
(MEM); dy/dx is the discrete change of the dummy variable from 0 to 1.
*
p < 0.10.
**
p < 0.05.
***
p < 0.01.

experiment to elicit their preferences, and HB analysis was used to
analyze the choice data. Although the majority of the respondents
were willing to invest in a responsible portfolio, 15.5% of them did
not want to have any socially responsible attribute combination
and sacrificed a small amount of pension income.
The results demonstrated the utility gained for different level
of each of the attributes and revealed the importance of the ImpI,
SRI and cost attributes (research objective 1). Firms committed to
sustainability, firms with environmental concerns and firms active
in the local community were the SRI criteria that characterized
positive screenings and were the most preferred attribute levels.
Regarding SRI belonging to negative screenings, the respondents
most preferred excluding investments in the weapon industry.
Regarding ImpI, investments in healthy aging, healthy living and
wellness were the investments from which the respondents gained
the most utility, whereas investments in improved living condi-
tions were associated with the least utility gain. Moreover, the
most important criteria proved to be related to SRI, followed by
ImpI and cost (research objective 2). This study illustrates the
application of discrete-choice conjoint analysis to the study of
responsible investment for pension funds. We consider this study
to contribute to the development of an investment strategy aimed
at increasing the involvement of pension beneficiaries in the de-
sign of pension investment policy. Therefore, to ensure that the
members could consistently perform the choice-based conjoint
experiment, we selected an investment portfolio that was closer to
their inherent values and did not require complex choices involv-
ing risk/return portfolio characteristics. The results indicated that
almost one-fifth of the respondents provided inconsistent answers.
Considering this finding, although the majority of respondents can
choose effectively, interventional policies should be implemented
to increase human efficiency. Improving financial literacy can help
pension fund beneficiaries make more consistent choices and feel
more secure about the decision-making process involved in invest-
ment.
A latent segmentation analysis revealed the existence of 3 seg-
ments (research objective 3). The first segment, which included
younger participants and was labeled ‘‘on the fence’’, was the one
in which respondents felt more insecure and likely less confident
about the true outcome of SRI. Promoting SRI attributes would
help to increase awareness of the benefits of such investment poli-
cies. Future research should investigate the influence of consumer
confidence in SRI portfolios and how to increase the engagement
and involvement of pension beneficiaries. Furthermore, we should
address the problem of the insufficient interest of young adults
in pension issues, which may limit their involvement. A likely
solution would be to implement communication strategies – such
as piquing the interest of young adults with respect to healthy
aging – aiming to increase their awareness of better well-being
after retirement.
94 G. Apostolakis et al. / Journal of Behavioral and Experimental Finance 17 (2018) 83–96

Table 10
Segmentation analysis.
Group 1 Stand. dev. Group 2 Stand. dev. Group 3 Stand. dev. F -test
Panel A: Average utilities
Impact
Invest in educating people 14.518 29.355 38.280 14.424 3.412 26.511 42.292***
Invest in better working conditions −0.963 31.891 −23.627 15.635 8.396 33.294 25.297***
Invest in improved living conditions 0.150 30.021 16.686 14.925 −19.681 23.629 57.249***
Invest in medical innovation −13.706 48.641 −31.340 22.333 7.874 44.581 23.257***
SRI
Sustainable investments 32.541 36.817 46.761 22.237 12.480 30.290 34.549***
Environmental concerns 27.759 45.833 28.974 16.680 28.773 36.793 0.026
Good corporate governance 16.408 44.243 39.321 26.451 −14.245 31.630 56.531***
Community involvement 27.077 63.367 3.708 21.372 25.783 48.680 5.09***
Good HR treatment −4.984 30.184 −34.321 22.712 21.105 27.121 108.728***
Weapon industry exclusion 12.347 41.144 −2.760 23.599 37.104 46.000 26.42***
Animal experiment exclusion −55.783 32.608 −102.427 23.691 −29.294 39.680 108.468***
Bad HR treatment exclusion 0.114 30.706 0.185 18.019 16.794 22.658 17.468***
Ethical consideration exclusion −11.843 37.719 60.345 22.439 −59.469 27.200 304.324***
Nuclear power exclusion −43.636 46.757 −39.786 26.194 −39.030 41.336 0.338
Cost −2.567 6.760 −3.960 3.724 −2.857 6.137 0.995
None 52.618 47.099 457.400 136.210 −195.855 84.420 1293.24***
Panel B: Average importance
SRI 51.52 14.623 58.17 8.696 51.21 13.111 8.048***
Impact investments 29.22 11.3705 26.91 5.108 29.16 10.616 1.305
Cost 19.25 14.356 14.92 10.251 19.62 12.983 3.227**

Note: The highest value for each variable among the three segments is presented in bold.* p < 0.10,** p < 0.05,*** p < 0.01.

Table 11 psychological distance, we must bring the attributes of the SRI

Latent class simulation analysis. portfolio to a more concrete mentality by transforming them from
Portfolio shares of preferences by group segment intangible to tangible through the use of behavioral techniques
Segment 1 Segment 2 Segment 3 Total such as nudging policies, stemming from carefully designed choice
Portfolio 1 52.201 1.429 98.248 70.459 architectures.
None 47.779 98.799 1.752 29.541 Behavioral economists stress that because the choices of indi-
Stand. error (4.459) (0.811) (0.575) (2.299) viduals are anchored by behavioral biases and bounded rationality,
(4.459) (0.811) (0.575) (2.299)
they are not always rational decision makers and optimal plan-
Portfolio 2 11.008 0.003 89.062 56.339 ners. Inconsistent behavior is frequently attributed to self-control
None 88.992 99.997 10.938 43.661
Stand. error (2.958) (0.003) (1.778) (2.554)
problems, bounded rationality, and myopic views. We suggest
(2.958) (0.003) (1.778) (2.554) that intervention strategies be implemented to overcome these
Note: Standard errors are given in parentheses, randomized first choice. Portfolio
behavioral biases. For instance, Lynch and Zauberman (2007) argue
1 comprises those levels of the portfolio that give the maximum of utility: healthy that policy makers who encourage people to take actions that
aging + environmental concerns for e21. Portfolio 2 comprises the portfolio with seem costly in the short run but that are beneficial in the long
the minimum utility gained: improved living conditions + animal experiment run (i.e., saving for retirement) should frame these individual deci-
exclusion. Both portfolios are estimated at a cost of e31. The price elasticity of
sions as though they were to occur in the distant future. Although
demand in absolute value is 0.133 for portfolio 1 and 0.096 for product 2. WTP for
improved living conditions to healthy aging = e5.7 for all respondents and e3.7, providing people with the ability to make their own investment
e9, and e10.5 for segments 1, 2, and 3, respectively. WTP for switching from animal decisions and to become involved and participate in the design of a
experiment exclusion to environmental concerns = e21.7 for all respondents. pension fund’s investment strategy is subject to human behavioral
biases, a suitable choice architecture focused on shifting people
away from acting irrational will help their decision making. This
The second segment displayed the highest average risk toler- choice architecture aims to moderate the effects of psychological
ance and was labeled ‘‘materialistic’’ because the utility from con- distance by making desirable choices feasible and abstract notions
tinuing to invest in a conventional manner was very high among more concrete. Furthermore, a suitable choice architecture can
this segment: they typically preferred the none option, indicating incorporate framing and nudging strategies to transcend pension
their indifference to SRI choices. Furthermore, this segment had the fund beneficiaries’ human behavioral biases.
highest level of income on average; thus, there may be a reverse
relationship between SRI portfolio selection and income. For this 5.1. Limitations
segment, an increased level of pension autonomy would preferably
be realized from an investment portfolio intended to maximize fi- Several limitations of this study stem from the research setting
nancial efficiency through the selection of an investment portfolio and the methodology used to elicit preferences based on a hypo-
that is aligned with their risk/return profile rather than a social thetical investment product. Ding et al. (2005) show that partici-
attribute profile. pants in conventional conjoint tasks exhibit lower price sensitivity
Most of the respondents classified into the third segment were than participants in conjoint tasks with incentive structures that
characterized by greater psychological distance than the other align with actual purchase behavior, because the product is not
two segments and were described as ‘‘advocates’’. For this third immediately available for purchase in the hypothetical research
segment, although pension beneficiaries are more willing to con- setting of conventional conjoint studies. Bornemann and Hom-
tribute to investing in a socially responsible manner, their greater burg (2011) posit that because psychological distance increases
psychological distance may result in more inconsistent decision the salience of desirability relative to feasibility concerns, con-
making when real choices must be made. To reduce this gap in sumers perceive high prices more favorably when the purchase
 G. Apostolakis et al. / Journal of Behavioral and Experimental Finance 17 (2018) 83–96
is temporally distant than when the purchase is temporally close.
Furthermore, Bornemann and Homburg (2011) argue that psy-
chologically distant settings may decrease consumers’ emphasis
on the sacrifice-related cost implications in conjoint tasks. Thus,
particularly in our setting, we should expect younger people and
people at a greater psychological distance from retirement to show
a higher preference for SRI products because the pecuniary sacri-
fice of their pension is perceived to occur later.
A second limitation is that our study concerns a limited group
of people in the Dutch healthcare sector. We built a responsible
investment portfolio that includes impact investments oriented to
this sector. Accordingly, our findings may not be representative
of the general population. Future studies can expand our research
setting to other segments of the population and can examine sev-
eral impact investment criteria related to different sectors. Finally,
because we examined preferences for a hypothetical investment
product, we cannot compare choices in a real-life scenario and es-
timate respondents’ actual behavior. However, previous research
has shown that stated preference methods are one of the best
methods to elicit and accurately predict actual purchase behavior
(Louviere et al., 2000).
6. Conclusion
Disclosing individual investment preferences is important for
pension funds that are examining investment solutions for better
well-being. Investing based on impact and ethical criteria does
not contradict pension fund boards’ mandate when it is aligned
with the interests of pension beneficiaries. This study fills a gap
in the literature regarding pension beneficiaries’ preferences for
responsible investments. The findings reveal which responsible in-
vestment attributes and which levels of such attributes maximize
non-financial utility. Based on our findings and suggestions, pen-
sion funds can better align their socially responsible investments
to the interests of their beneficiaries. Most pension funds prefer a
combination of SRI and impact investments in their portfolio, even
when a small pension deduction is involved.
In the Dutch pension system, pension beneficiaries primarily
have the role of contributors, with limited involvement in decision-
making processes regarding their pension capital. Richardson
(2011) posits that in general, pension funds neither provide forums
for active dialogue among investors nor encourage their partic-
ipation in investment policymaking. With this study, we show
that although a substantial number of respondents do not make
systematic choices, most pension participants were able to con-
sistently select their preferred investment portfolio. Accordingly,
pension beneficiaries can be further engaged in pension policy
affairs by increasing their involvement in the design of pension
funds’ investment policy. Because pension funds will examine
their investments with respect to socially responsible and impact
criteria, not solely return-maximizing criteria, such increased in-
volvement will strengthen the democratic governance of pension
funds and the legitimacy of pension fund trustees in pursuing their
social goals.
Funds that are more democratically governed appear to be at
the forefront of SRI. In credit unions and cooperative financial
institutions, SRI investments are incorporated based on democratic
decision making. People participating in these decisions bolster the
legitimacy and public accountability of such decisions (Richardson,
2011). Although risks in the Dutch collective pension schemes
are currently shared, there is pressure for a shift in the future
from defined benefit (DB) to defined contribution (DC) plans, in
which individuals would bear more of the risk associated with
their pensions. However, Richardson (2011) argues that this shift in
occupational pension plans also would give a voice to pension fund
participants, specifically to SRI, and democratic decision making
95
could become more supportive of pension beneficiaries. However,
the shift to more individualistic plans that consider the voice of
pension participants regarding investment policy and that intro-
duce individual choices would eventually require an appropriate
choice architecture to assist decision making. This choice architec-
ture should be designed according to the collective mandate and
pension beneficiaries’ preferences and desires for the direction of
investment policy. Such an architecture can integrate responsible-
investment policies that are currently followed by many pension
funds.
Disclaimer
Socially responsible investing does not necessarily lead to lower
financial returns than conventional investing, as indicated by this
hypothetical scenario.
Appendix A. Supplementary data
Supplementary material related to this article can be found
online at https://doi.org/10.1016/j.jbef.2018.01.001.
References
Bakken, D., Frazier, C.L., 2006. Conjoint analysis: Understanding consumer decision
making. In: Grover, R., Vriens, M. (Eds.), The Handbook of Marketing Research.
(Chapter. 15).
Bollen, N.P.B., 2007. Mutual fund attributes and investor behavior. J. Finan. Quant.
Anal. 42 (03), 683–708. http://dx.doi.org/10.1017/S0022109000004142.
Borgers, A.C.T., Pownall, R.A.J., 2014. Attitudes towards socially and environmen-
tally responsible investment. J. Behav. Exp. Finance 1, 27–44. http://dx.doi.org/
10.1016/j.jbef.2014.01.005.
Bornemann, T., Homburg, C., 2011. Psychological distance and the dual role of price.
J. Consum. Res. 38 (3), 490–504.
Boudt, K., Cornelissen, J., Croux, C., 2013. The impact of a sustainability constraint
on the mean-tracking error efficient frontier. Econom. Lett. 119 (3), 255–260.
http://dx.doi.org/10.1016/j.econlet.2013.03.020.
Brimble, M., Vyvyan, V., Ng, C., 2013. Belief and investing: Preferences and attitudes
of the faithful. Aust. Account. Bus. and Finance J. 7 (1), 23–41.
Brusch, M., Baier, D., Treppa, A., 2002. Conjoint analysis and stimulus presentation —
A comparison of alternative methods. In: Jajuga, K., Sokołowski, A., Bock, H.-H.
(Eds.), Classification, Clustering, and Data Analysis. Springer Berlin Heidelberg,
Berlin, Heidelberg, pp. 203–210. Retrieved from http://link.springer.com/10.
1007/978-3-642-56181-8_22.
Bryan, S., Gold, L., Sheldon, R., Buxton, M., 2000. Preference measurement using
conjoint methods: An empirical investigation of reliability. Health Econ. 9 (5),
385–395.
Capelle-Blancard, G., Monjon, S., 2012. Trends in the literature on socially responsi-
ble investment: Looking for the keys under the lamppost. Bus. Ethics: Eur. Rev.
21 (3), 239–250. http://dx.doi.org/10.1111/j.1467-8608.2012.01658.x.
Cerjak, M., Haas, R., Kovačić, D., 2010. Brand familiarity and tasting in con-
joint analysis. British Food J. 112 (6), 561–579. http://dx.doi.org/10.1108/
00070701011052664.
Decker, R., Trusov, M., 2010. Estimating aggregate consumer preferences from
online product reviews. Int. J. Res. Mark. 27 (4), 293–307. http://dx.doi.org/10.
1016/j.ijresmar.2010.09.001.
Ding, M., Grewal, R., Liechty, J., 2005. Incentive-aligned conjoint analysis. J. Mark.
Res. 42 (1), 67–82.
Duiker, J., Bijleveld, B., Ipenburg, V.P., 2016. Impact investing: from niche to
mainstream. Utrecht: VBDO. Retrieved from http://www.vbdo.nl/files/news/
ImpactInvestment_Final.pdf.
Eyal, T., Liberman, N., Trope, Y., Walther, E., 2004. The pros and cons of temporally
near and distant action. J. Pers. Soc. Psychol. 86 (6), 781–795. http://dx.doi.org/
10.1037/0022-3514.86.6.781.
Friede, G., Busch, T., Bassen, A., 2015. ESG and financial performance: Aggregated
evidence from more than 2000 empirical studies. J. Sustain. Finance Invest. 5 (4),
210–233. http://dx.doi.org/10.1080/20430795.2015.1118917.
Hong, H., Kacperczyk, M., 2009. The price of sin: The effects of social norms on
markets. J. Financ. Econ. 93 (1), 15–36. http://dx.doi.org/10.1016/j.jfineco.2008.
09.001.
Klein, A., Nihalani, K., Krishnan, K.S., 2010. A comparison of the validity of
interviewer-based and online-conjoint analyses. J. Manage. Mark. Res. 4.
Koedijk, K., Slager, A., 2011. Investment Beliefs: A Positive Approach to Institutional
Investing. Palgrave Macmillan.
Kruif, D.D., Ipenburg, P.V., 2015. Benchmark responsible investment by pension
funds in the Netherlands 2015 bridging the gap. Utrecht: VBDO. Retrieved
from http://www.vbdo.nl/files/report/VBDOResponsibleInvestmentDutchPens
ionFunds.pdf.
96 G. Apostolakis et al. / Journal of Behavioral and Experimental Finance 17 (2018) 83–96
Liberman, N., Trope, Y., 2008. The psychology of transcending the here and now.
Science 322 (5905), 1201–1205. http://dx.doi.org/10.1126/science.1161958.
Liberman, N., Trope, Y., McCrea, S.M., Sherman, S.J., 2007. The effect of level of
construal on the temporal distance of activity enactment. J. Exp. Soc. Psychol.
43 (1), 143–149. http://dx.doi.org/10.1016/j.jesp.2005.12.009.
Louviere, J.J., Hensher, D.A., Swait, J.D., 2000. Stated Choice Methods: Analysis and
Applications. Cambridge University Press.
Lynch, J.G., Zauberman, G., 2006. When do you want it? Time, decisions, and public
policy. J. Publ. Policy Mark. 25 (1), 67–78. http://dx.doi.org/10.1509/jppm.25.1.
67.
Lynch, J.G., Zauberman, G., 2007. Construing consumer decision making. J. Consum.
Psychol. 17 (2), 107–112. http://dx.doi.org/10.1016/S1057-7408(07)70016-5.
Martin, M., 2013. Making impact investible (Impact Economy Working Papers
No. Vol. 4). Impact economy. Retrieved from http://www.impacteconomy.com/
making_impact_investible.
McCullough, D., 2002. A user’s guide to conjoint analysis. Mark. Res. 14 (2), 18.
McFadden, D., 1986. The choice theory approach to market research. Mark. Sci. 5 (4),
275–297.
Nilsson, J., 2009. Segmenting socially responsible mutual fund investors: The influ-
ence of financial return and social responsibility. Int. J. Bank Mark. 27 (1), 5–31.
http://dx.doi.org/10.1108/02652320910928218.
Novkovic, S., 2008. Defining the co-operative difference. J. Socio-Econ. 37 (6), 2168–
2177. http://dx.doi.org/10.1016/j.socec.2008.02.009.
O’Donohoe, N., Leijonhufvud, C., Saltuk, Y., Bugg-Levine, A., Brandenburg, M., 2010.
Impact investments: An emerging asset class. JP Morgan, Rockefeller Founda-
tion, and the Global Impact Investing Network. Retrieved from http://ventureatl
anta.org/wp-content/uploads/2011/11/JP-Morgan-impact_investments_nov20
10.pdf.
Orme, B., 2001. Assessing the monetary value of attribute levels with conjoint
analysis: Warnings and suggestions (RESEARCH PAPER SERIES) Sequim, WA
98382: Sawtooth Software, Inc. p. 6.
Orme, B., 2009. Fine-tuning CBC and Adaptive CBC questionnaires. Sequim, WA
98382: Sawtooth Software, Inc.
Parient, W., 2011. What is the demand for microcredit? The case of rural areas
in Serbia. In: Armendariz, B., Labie, M. (Eds.), The Handbook of Microfinance.
World Scientific, Singapore.
Renneboog, L., Ter Horst, J., Zhang, C., 2008a. Socially responsible investments:
Institutional aspects, performance, and investor behavior. J. Banking Finance
32 (9), 1723–1742. http://dx.doi.org/10.1016/j.jbankfin.2007.12.039.
Renneboog, L., Ter Horst, J., Zhang, C., 2008b. The price of ethics and stakeholder
governance: The performance of socially responsible mutual funds. J. Corp.
Finance 14 (3), 302–322. http://dx.doi.org/10.1016/j.jcorpfin.2008.03.009.
Renneboog, L., Ter Horst, J., Zhang, C., 2011. Is ethical money financially smart?
Nonfinancial attributes and money flows of socially responsible investment
funds. J. Financ. Intermed. 20 (4), 562–588. http://dx.doi.org/10.1016/j.jfi.2010.
12.003.
Richardson, B.J., 2011. From fiduciary duties to fiduciary relationships for socially
responsible investing: Responding to the will of beneficiaries. J. Sustain. Finance
Invest. 1 (1), 5–19. http://dx.doi.org/10.3763/jsfi.2010.0002.
Rietjens, M., 2011. Young people’s preferences for financial versus non-financial
attributes of pension funds (MCs). Netspar. (054).
Sawtooth. 2008. Sawtooth software, technical paper series, CBC v6.0, technical
paper. Sawtooth software inc.
Social Finance, & Finethic. 2012. Microfinance, impact investing, and pension fund
investment policy survey. Retrieved from http://www.socialfinance.org.uk/res
ources/social-finance/microfinance-impact-investing-and-pension-fund-inves
tment-policy-survey.
Statman, M., Glushkov, D., 2009. The wages of social responsibility. Financ. Anal. J.
65 (4), 33–46. http://dx.doi.org/10.2469/faj.v65.n4.5.
Trope, Y., Liberman, N., 2000. Temporal construal and time-dependent changes in
preference. J. Pers. Soc. Psychol. 79, 876–889.
Trope, Y., Liberman, N., 2003. Temporal construal. Psychol. Rev. 110 (3), 403.
van Trijp, H.C.M., 2013. Encouraging Sustainable Behavior: Psychology and the
Environment. Psychology Press.
Vermeulen, B., Goos, P., Vandebroek, M., 2008. Models and optimal designs for
conjoint choice experiments including a no-choice option. Int. J. Res. Mark.
25 (2), 94–103. http://dx.doi.org/10.1016/j.ijresmar.2007.12.004.
Vriens, M., Loosschilder, G.H., Rosbergen, E., Wittink, D.R., 1998. Verbal versus
realistic pictorial representations in conjoint analysis with design attributes.
J. Prod. Innov. Manage. 15 (5), 455–467. http://dx.doi.org/10.1111/1540-5885.
1550455.
Vyvyan, V., Ng, C., Brimble, M., 2007. Socially responsible investing: The green
attitudes and grey choices of Australian investors. Corp. Governance Int. Rev.
15 (2), 370–381. http://dx.doi.org/10.1111/j.1467-8683.2007.00567.x.
Wagemans, F., Duiker, J., Fijt, E., Kruif, D.D., 2014. Benchmark responsible invest-
ment by insurance companies in the Netherlands 2014. Utrecht: VBDO.
Zhao, M., Hoeffler, S., Zauberman, G., 2007. Mental simulation and preference
consistency over time: The role of process-versus outcome-focused thoughts.
J. Mark. Res. 44 (3), 379–388. http://dx.doi.org/10.1509/jmkr.44.3.379.