Trusting and trustworthiness: What are they, how to measure them, and what affects them

Abstract

The study examines trust by investigating potential determinants (factors determined at birth and childhood) and correlates (various views and attitudes) of trusting and trustworthiness. We examine behavioral and survey measures and conclude that the amount sent in the trust game is a good, albeit partial and overreaching, measure of one facet of trust. Common survey measures of trusting capture well other facets of trusting. The proportion sent back in the trust game represents well trustworthiness in the specific context in which it was developed, an investment situation, but falls short in capturing other facets of trustworthiness. The Machiavellian scale appears to be a weak measure of any facets of trustworthiness. Gender is the primary determinant of investment-related facet of trusting, and personality is a strong determinant of other facets. One personality trait - agreeableness - explains investment-related trustworthiness.

Full text

Trusting and trustworthiness: What are they, how to measure them,
and what affects them
Avner Ben-Ner
a,*
, Freyr Halldorsson
b
a
Carlson School of Management, University of Minnesota, MN, United States
b
City of Reykjavik, Office of Human Resources, Radhus Reykjavikur, 101 Reykjavik, Iceland
article info
Article history:
Received 6 September 2007
Received in revised form 7 September 2009
Accepted 16 October 2009
Available online 22 October 2009
JEL classification:
C72
C91
D63
PsycINFO classification:
2260
3020
Keywords:
Trusting
Trustworthiness
Trust game
Measurement
abstract
The study examines trust by investigating potential determinants (factors determined at
birth and childhood) and correlates (various views and attitudes) of trusting and trustwor-
thiness. We examine behavioral and survey measures and conclude that the amount sent
in the trust game is a good, albeit partial and overreaching, measure of one facet of trust.
Common survey measures of trusting capture well other facets of trusting. The proportion
sent back in the trust game represents well trustworthiness in the specific context in which
it was developed, an investment situation, but falls short in capturing other facets of trust-
worthiness. The Machiavellian scale appears to be a weak measure of any facets of trust-
worthiness. Gender is the primary determinant of investment-related facet of trusting,
and personality is a strong determinant of other facets. One personality trait – agreeable-
ness – explains investment-related trustworthiness.
Ó2009 Published by Elsevier B.V.
1. Introduction
Trust is regarded as a central concept for understanding important economic, social and political behaviors. It is said to
aid economic, business, social and many other types of interactions; its absence thwarts many relations and causes much
harm. Trust has been defined as the ‘‘voluntary transfer of a good or favor to someone else, with future reciprocation ex-
pected but not guaranteed” (Gunnthorsdottir, McCabe, & Smith, 2002, p. 50). This representative definition invokes not only
a trustor but also a trustee. An understanding of the concept of trust requires therefore understanding of the two related yet
distinct concepts of trusting and trustworthiness (Ben-Ner & Putterman, 2001;Hardin, 2004). The goal of this paper is to ex-
plore conceptually and empirically the two sides of trust. What are they? What demographic, cognitive, personality, attitu-
dinal and other factors correlate with them? What are reasonable measures of trust?
Trust has been commonly measured through a survey instrument (used by the General Social Survey and the World
Values Survey) that requires the endorsement of ‘‘Most people can be trusted” or ‘‘You can’t be too careful in dealing with
0167-4870/$ - see front matter Ó2009 Published by Elsevier B.V.
doi:10.1016/j.joep.2009.10.001
*Corresponding author. Tel.: +16126240867.
E-mail address: benne001@umn.edu (A. Ben-Ner).
Journal of Economic Psychology 31 (2010) 64–79
Contents lists available at ScienceDirect
Journal of Economic Psychology
journal homepage: www.elsevier.com/locate/joep

people.” This measure forces the respondent to endorse a favorable view of the trustworthiness of most other people or
express general fearfulness of others. Since the publication of Berg, Dickhaut, and McCabe (1995) [BDM], economists have
used the amount sent in the investment or trust game as a behavioral measure of trusting, and the amount or proportion
returned as a measure of trustworthiness. Nevertheless, it has been understood that the amounts sent and returned may
reflect more than just trusting and trustworthiness, respectively, and that these have many layers of meaning and possible
interpretations (see, for example, Cox, 2004; Fehr, forthcoming).
Given the importance of and scholarly interest in trust, these measures have come under increasing scrutiny, and several
attempts have been made to relate them to each other and to other measures. Most of these attempts have been part of the
broader agenda of understanding what trust is and to what might it be related (see, for example, Ashraf, Bohnet, & Piankov,
2006; Fehr, forthcoming; Fehr, Fischbacher, von Rosenbladt, Schupp, & Wagner, 2003; Glaeser, Laibson, Scheinkman, & Sout-
ter, 2000; Gunnthorsdottir et al., 2002; Sapienza, Toldra, & Zingales, 2007). We contribute to this literature by taking a broad
sweep at the question of what is trust. We emphasize the distinction between trusting and trustworthiness,
1
and investigate
separately their meaning, measurement, and determinants. Following a strategy similar to that pursued by the authors cited
above, we evaluate the relationships between amount sent, amount returned in the trust game, and diverse survey measures
related to trusting and trustworthiness on the one hand and risk attitudes, views regarding probabilities of good and bad events,
altruism, optimism, faith in God, personality, general mental ability, family demographics, and other measures, on the other
hand. We interpret the meaning of trusting and trustworthiness on the basis of their relationship to various views and attitudes,
and relate them to background factors that individuals were born with or were shaped in early childhood.
The main thrust of this paper is empirical and is based on an analysis of detailed survey and experimental information for
approximately 200 subjects, students at a large American public university. We develop a conceptual framework to guide
and constrain our empirical investigation, with the primary goal of informing empirically about possible relationships be-
tween different variables, rather than testing detailed theoretical hypotheses about such relationships.
The rest of this paper is organized as follows. In Section 2we define trusting and trustworthiness, relate them to other
social preferences and to factors that affect their strength, and explore behavioral and survey measures relative to our def-
initions. The methods and data are described in Section 3. Section 4presents the results, and Section 5provides a discussion
of our results and compares them with findings in the literature. Section 6offers conclusions.
2. Trusting and trustworthiness: conceptual issues
2.1. Definitions of trusting and trustworthiness
Trusting is the inclination of a person ‘‘A” to believe that other persons ‘‘B” who are involved with a certain action will
cooperate for A’s benefit and will not take advantage of Aif an opportunity to do so arises. A– the trustor – must therefore
be willing to show his or her vulnerability by taking the risk that B– the trustee – may act in a way that does not benefit A.
The concept of trusting requires that the action hold the potential of a loss by the trustor; it does not require but is compat-
ible with a potential gain and/or loss by the trustee. Trusting can vary from complete distrusting to complete or ‘blind’
trusting.
A common example of trusting is when a business person Aengages ‘‘on trust” rather than through a binding contract in a
transaction that will be profitable to Aif another business person Bcooperates but will entail a loss to Aif Btakes advantage
of A. A different example is represented by the situation when a father lets his child enter a swimming pool knowing that if
he were at some distance from the pool and his child were at risk of drowning, then unknown people would jump to the
rescue, sacrificing at least comfort and dry clothes. Yet another example of daily manifestation of trusting is when one
crosses the street without fearing that a speeding car will crush her (even in the absence of a law punishing this, one
may safely assume), or walk in the streets among high rise buildings despite the fact that someone could drop an object ran-
domly from a high-floor window and hit her on the head. In these examples trusting reflects the belief that others are
responsible and competent individuals (Hill & O’Hara, 2006).
Trusting thus reflects a multifaceted expectation or belief by Athat: (1) Bwill not take advantage of the situation to make
a gain while imposing a loss on A, (2) Bwill not act maliciously towards A, (3) Bwill be willing to make small sacrifices for A,
and (4) Bis competent to act favorably towards A. These facets vary potentially in degree rather than being dichotomous.
Their importance is contingent on situation and context, as the examples discussed above suggest, and may vary across indi-
viduals. When not complete or absolute, trusting may be used in combination with other means: a business person may risk
a small amount of money in a transaction based on trust only, but will require a contract in a larger transaction; and a father
may not leave a child unattended by the pool for more than a few minutes.
Trustworthiness is the willingness of a person Bto act favorably towards a person A, when Ahas placed an implicit or ex-
plicit demand or expectation for action on B. The implicit demand may entail a situation in which a child is drowning and Bis
expected to do something to save the child, that Bdrive with care and stop at pedestrian crossing when Ais on the road, or
1
Hardin (2004, p. 29) writes: ‘‘Surprisingly, much of the literature on trust hardly mentions trustworthiness, even though implicitly much of it is primarily
about trustworthiness, not about trust.” More recently, several authors started addressing this distinction: Ashraf et al. (2006),Buchan, Croson, and Solnik
(2003), and Sapienza et al. (2007).
A. Ben-Ner, F. Halldorsson / Journal of Economic Psychology 31 (2010) 64–79 65

that Breward an investment made by A, like in the trust game. In most situations the expectations are based on social norms
that do not entail a precise course of action in a variety of situations, providing Bconsiderable room for discretion as to what
constitutes trustworthy behavior. Thus a reasonable Amay trust that a person Bstanding by a pool will be trustworthy in the
specific way of jumping into the water to come to the rescue of a drowning child, although this is likely to be considerably
muted when the child is in a rapidly moving river and there is a risk to the life of the rescuer. The meaning of trustworthiness
is very specific in the case of driving through a quiet intersection, but it is likely to be muddled when driving conditions are
difficult because of traffic and weather. In the case of situations represented by the trust game, trustworthiness means not
taking advantage of the vulnerability that Ahas created; since anything that Bretains is a windfall, the more Breturns the
more trustworthiness he or she exhibits towards Awho sought to maximize returns on investment.
2
Trustworthiness has
thus four parallel facets to trusting. As is the case with trusting, these facets are variable and vary across contexts and
individuals.
2.2. Determinants and correlates of trusting and trustworthiness
In this section we are taking a pragmatic approach to the development of a conceptual framework that relates trusting
and trustworthiness to various factors to inform the exploratory empirical investigation, which is the main goal of this paper.
We do not propose here a comprehensive biological, social or psychological theory of the formation and evolution of such
complex attitudes (Fehr, forthcoming;Hardin, 2004, make bold steps in these directions).
What might explain the level of trusting and trustworthiness of an individual, especially a young adult like most of the
participants in our experiments? We distinguish between factors that influence or determine trusting and trustworthiness,
and factors that are correlated or merely associated with them but are not likely to be causally related. The first category in-
cludes factors determined at birth and during childhood, including gender, age, ethnicity, and birth order, as well as personality
and mental ability (mostly determined at birth and through childhood experiences).
3
The list is obviously not exhaustive, but
includes the main demographic factors that, along with personality and cognitive ability, have been identified in the literature
as sources of differences in numerous attitudes and behaviors, including those of direct or indirect relationship to trusting and
trustworthiness.
4
The second category consists of views, attitudes, social preferences, values and beliefs (views and attitudes, for brevity) that
collectively describe an individual’s understanding of the world and other people. Views and attitudes include trusting and
trustworthiness and determined by similar factors. The list of possible views and attitudes is very long; here we concentrate
on a few relatively basic views and attitudes that are plausible correlates of facets of trusting and trustworthiness and there-
fore can help understand them better: optimism and pessimism, faith in God, altruistic and reciprocal tendencies, beliefs
about the honesty of others, attitude towards fairness, and risk attitudes.
5
Greater optimism – a comparatively positive out-
look on the likelihood of good outcomes – is likely to be related to a trusting attitude. Similarly, an exaggerated expectation of
good things happening is likely to be positively associated with trusting, whereas an exaggeration of negative outcomes is likely
to undermine trusting. It is not clear in what way faith in God as compared to lack of such faith would be associated with trust-
ing of people; it is possible that faith in God is reflective of trust in general, or alternatively, a lack of faith in people.
2.3. Measurement of trusting and trustworthiness
Trusting and trustworthiness are multifaceted concepts, and their various facets are unlikely to be perfectly correlated.
This suggests that single measures of trusting and trustworthiness cannot capture the true extent of their meaning. So
how to measure them? In principle, one could ask well-targeted questions that focus on the facets of trustworthiness and
trusting. However, such questions are not commonly asked, so we focus on questions that are asked in surveys aimed at
understanding trust. The three principal items are: (1) ‘‘Generally speaking, would you say that most people can be trusted
2
The specific meaning of trustworthiness depends on B’s sense of fairness regarding the distribution of a pie over which Bhas the ultimate decision-making
power, and the sense of what is the appropriate degree of reciprocity. Bmay regard the amount he or she receives from Aas his or hers – after all, Bmakes the
ultimate decision about its division, just like a dictator – so Bmay think of the amount as an endowment he or she needs to divide.
3
The predominant view among personality scholars is that personality traits are shaped by genetic inheritance and childhood experiences, and that by early
adulthood they become essentially fixed (Allik, Laidra, Realo, & Pullmann, 2004; McCrae & Costa, 1999). There is a vast literature that relates personality to
various attitudes and behaviors (e.g., Ajzen, 2005). Personality traits correlate with a variety of behaviors relevant to trusting and trustworthiness, including
cooperation and reciprocity (e.g., Ben-Ner, Kong, Magan, & Putterman, 2004; Ben-Ner, Kong, & Putterman, 2004).
4
Boys and girls may be born with different predilections to trust and to be trustworthy derived from the long evolutionary process in which men and women
may have played different roles; in addition, boys and girls have different experiences related to their gender, which may affect how much they trust people
they do not know, or the degree of trustworthiness they manifest towards unknown people (on gender differences see Croson and Gneezy (2008)). Birth order is
probably associated with different experiences within the family; for example, only children and first born have to compete less for attention and other
resources, which may have an impact on their social preferences, including trusting and trustworthiness (Sulloway (1996) and others have argued for
pronounced birth order effects). Different experiences people may have as members of different ethnic groups may affect their trusting and trustworthiness
(e.g., Bahry, Kosolapov, Kozyreva, & Wilson, 2005). For age and birth order, see Harbaugh and Steven G. Liday Jr. (2003) and their references. We are not aware
of a comprehensive theory of the determinants of trusting and trustworthiness (or of other dispositions and preferences), let alone one that is empirically
testable. It is beyond the cope of this paper to offer specific hypotheses concerning the determinants of trusting and trustworthiness.
5
We are not aware of any comprehensive theories that correlate various views, attitudes and dispositions, or of empirical work that does so. We therefore do
not offer specific hypotheses about the relationship between various views and attitudes and trusting and trustworthiness.
66 A. Ben-Ner, F. Halldorsson / Journal of Economic Psychology 31 (2010) 64–79

or that you can’t be too careful in dealing with people?” (General Social Survey and World Values Survey), (2) ‘‘On a scale
from 1 to 6 where 1 is ‘Relatively cautious’ and 6 is ‘Relatively trusting,’ how would you describe your interactions with other
people?”, and (3) ‘‘Which of the following statements reflects best your view? I will trust a person until I have clear evidence
that he/she can’t be trusted or I will not trust a person until I have clear evidence that he/she can be trusted” (see Rotter,
1980).
These items do not focus on any single facet of trusting; rather, they reflect alternative formulations of the basic question,
whether the respondent is trusting of others or not, without specifying any context, staying away entirely from the concept
of reciprocity or any insinuation that other people may be reacting to A’s actions. The differences among the three items are
much nuanced, with the first item stating the alternatives in the starkest way and forcing a choice between two extremes.
The third item likewise (although perhaps a little more mildly) forces a choice between strongly-stated alternatives, whereas
the second item presents the alternatives more mildly and allows a finer grading of the answer on a scale of 1–6. This item
focuses on the respondent’s attitudes, unlike the first and third items, which emphasize the trustworthiness of others. The
responses to these questions may well depend on the theory of trust that respondents have or their associations based on
their recent or formative experiences. Hence it is difficult to predict which of these better reflects a person’s tendency to be
trusting.
Survey items aimed at capturing trustworthiness are uncommon, probably as a result of the concentration of trust re-
search on trusting. Gunnthorsdottir et al. (2002) and Burks, Carpenter, and Verhoogne (2003) use the Machiavellian scale
(Christie & Geis, 1970) to capture key personality elements that may predict reciprocity in the context of trustworthiness.
In the absence of other survey-based measures, we use the Machiavellian scale not as a personality instrument but as a po-
tential measure of trustworthiness. However, we also use two items from the scale that may be viewed as components of
trustworthiness: ‘‘Anyone who completely trusts anyone else is asking for trouble,” and ‘‘P.T. Barnum was wrong when
he said that there’s a sucker born every minute.” Endorsement of the first item may suggest that the respondent would take
advantage of others’ trust, signifying personal lack of trustworthiness, or may reflect the respondent’s lack of trusting of oth-
ers, irrespective of his or her own trustworthiness. A similar issue arises in the context of the second item.
The amount sent in the BDM trust game has been used as a measure of the degree of trusting by A, the sender. It is obvi-
ously in B’s interest to take advantage of Aby keeping all that Asends. If a rational and self-interested Aexpects, with some
probability, that Bwill not take advantage of A’s vulnerability, then Atrusts Bto a degree. Ahas to figure out the probability
that Bwill return some money, and what amount, to evaluate A’s utility function and compare it with the sure bet of not
sending any amount. So the amount sent is a good reflection of the first facet of trusting in Bby A. However, as Cox
(2004) and others have pointed out, the amount sent may also reflect A’s altruism towards B, independent of A’s trusting.
The amount sent by Ain the trust game and similar situations reflects his or her attitude towards risk entailed by the gamble
of sending money to an unknown B, as well as his assumptions about the social preferences of B, reciprocity as well as the
sense of what is fair in the division of a windfall over which Bhas greater power, in addition to any altruistic motivation.
Greater risk aversion is thus likely to temper trusting, as the economics literature has suggested, while altruism is likely
to be associated with more sending by Ain the trust game.
The commonly-employed behavioral measure of trustworthiness is the amount or proportion returned by Bin the trust
game. This amount may reflect trustworthiness in the sense of reciprocity, and also altruism or unconditional kindness to-
wards A. The trust game captures important aspects of trust (but as just argued, not only trusting) in many economic and
social contexts but not where there is no potential gain for Bso Ahas to trust the absence of malice on B’s part, B’s compe-
tence, or B’s willingness to make a small sacrifice.
3. Method
We collected survey and experimental data pertaining to the trusting and trustworthiness measures described earlier, as
well as additional information about subjects, including demographics, personality, cognitive ability, and attitudes. Our
empirical exploration proceeds in several steps, investigating: (1) the association between individual characteristics deter-
mined at birth and in childhood on the one hand, and behavioral and survey measures of trusting and trustworthiness on the
other hand, (2) the association between individual attitudes and views, and behavioral and survey measures of trusting and
trustworthiness, and (3) the relationship between survey and behavioral measures of trusting and trustworthiness, control-
ling for individual characteristics determined at birth and in childhood. We present first results for trusting and then for
trustworthiness; the two sets of analyses have substantially the same structure, with the main difference arising from the
need to account for the possibility of reciprocity in the analysis of trustworthiness.
3.1. Participants and procedure
The study uses a sample of 204 participants. Two-thirds of the participants were female and consisted of undergraduate
and graduate university students that were recruited through email solicitation from a large cross section of the University of
Minnesota student population to take part in a study on decision-making. Half the participants were in the role of senders in
the trust game (trustors), and half in the role of receivers (trustees).
A. Ben-Ner, F. Halldorsson / Journal of Economic Psychology 31 (2010) 64–79 67

The study procedure consisted of two parts: an online survey, and an experimental session that participants attended sev-
eral days later. The first part involved the sign-up process for ‘‘a study in interactive decision-making,” followed by a survey
that included demographic and other background information, various attitudinal survey items, measures of trust, risk-re-
lated attitudes, and a personality inventory. This part was done online at participants’ leisure, several days before they par-
ticipated in the trust game (registration closed two days prior to the experiment), and took about 30–45 min. This separation
should lower the probability that a participant’s mood, state of mind or unknown priming would generate artificial corre-
lations between survey responses and behavior in the trust game.
The second part consisted of trust game experiments, followed by a cognitive test. Participants were randomly assigned to
play as a trustor (called person in role Ain the experiment) or trustee (role B) in the trust game; participants in the two roles
were seated in separate rooms and all interactions took place over a computer network. This setup ensured that participants
were not aware of the identity of their counterparts. During the experimental sessions, after introduction and on-screen
instructions and questions testing their comprehension of the experiment, each participant played seven rounds of the
experiment, each time paired with a different person. In each round, participants in both roles were each credited with
10 ‘‘experiment dollars” (E$), which converted to real dollars at the rate E$1 = $0.14 at the end of the session. Subject A
was asked to choose a whole number of experiment dollars, E$X
a
2ð0;1;...;10Þ, to send to subject B, knowing that Bwould
receive triple the amount sent and could send back a proportion (restricted to sixths of the amount received e.g., 0, 1/6, 1/3,
etc.) of that amount, including nothing. Thus, Awould earn any part of the E$10 kept and could earn an additional amount
between E$0 and E$30, depending on B’s choice. Bwould earn between E$10 and E$(10 + 3X
a
X
b
), where X
b
63X
a
is the
amount Bsends to A. The decisions were made on a grid where rows indicate A’s possible choices of how much to send
and columns B’s potential choices of how much to return. The actual choices were carried out by highlighting a row or a col-
umn. The grid is replicated in Appendix A. In this paper we use only data from the first round.
6
After the last round participants were asked to complete the Wonderlic Personnel Test, a measure of general mental abil-
ity. Then the $10 show-up fee was paid, along with any earnings (about $15 from the trust game plus about $2 as a random
payout for lottery choices, described later). This part took approximately 45 min.
3.2. Measures
3.2.1. Survey measures of trusting and trustworthiness
We use the three aforementioned survey measures to capture trusting.
Trusting GSS – This is the trust item from the General Social Survey and the World Values Survey. The item is: ‘‘Generally
speaking, would you say that most people can be trusted, or that you cannot be too careful when dealing with people?” Re-
sponses are made on a binary response scale, where 0 = ‘‘You cannot be too careful when dealing with people” and, 1 = ‘‘Most
people can be trusted.”
Trusting interactions – Another measure of general trusting was captured with the item: ‘‘Which of the following state-
ments reflects best your view?” The first statement (coded 0) is ‘‘I will not trust a person until there is clear evidence that
he or she can be trusted,” while the second statement (coded 1) is ‘‘I will trust a person until I have clear evidence that he or
she can’t be trusted.”
Trusting evidence – A third general measure of trusting is: ‘‘On a scale from 1 to 6 where 1 is ‘‘Relatively cautious” and 6 is
‘‘Relatively trusting,” how would you describe your interactions with other people?”
Trustworthiness is measured by the Machiavellian scale and two items from the scale.
Machiavellianism – A scale that asks participants to state their level of agreement or disagreement to 20 statements on a
five-point scale ranging from ‘‘Strongly disagree” to ‘‘Strongly agree.”
Trusting is trouble – ‘‘Anyone who completely trusts anyone else is asking for trouble.”
People no suckers – ‘‘P.T. Barnum was wrong when he said that there’s a sucker born every minute.”
3.2.2. Behavioral measures of trusting and trustworthiness – trust game
Amount sent –X
a
, the number of experimental dollars, ranging from 0 to 10, sent by participants in role A.
Proportion sent back –X
b
, the proportion sent back by participants in role B, with options restricted to sixths of the amount
received i.e., 0% (0/6th), 17% (1/6th), 33% (2/6th), 50% (3/6th), 67% (4/6th), 83% (5/6th), and 100% (6/6th).
3.2.3. Explanatory variables
3.2.3.1. Individual characteristics determined at birth and in childhood
.
Gender, ethnicity and age were obtained from the online
survey through direct questions.
Birth order was inferred from information supplied in the online survey (number of siblings younger and older than
respondent).
General mental (cognitive) ability was measured with the Wonderlic Personnel Test, which is timed (12 min) and consists
of 50 items.
6
The instructions and sample screen shots are at https://netfiles.umn.edu/users/benne001/www/papers/ep/ep-01.pdf. Further description of the
experiments can be found in Ben-Ner and Putterman (2009).
68 A. Ben-Ner, F. Halldorsson / Journal of Economic Psychology 31 (2010) 64–79

The Big Five personality traits were captured by the NEO-FFI, which uses a total of 60 items to capture neuroticism, extra-
version, openness, agreeableness, and conscientiousness.
7
3.2.3.2. Individual attitudes and views. We experimented with two alternative measures of risk preferences, one based on a
survey item, and another based on incentivized choices of risky vs. riskless outcomes.
Risk aversion-survey – A single item measured willingness to take risks: ‘‘Some people like taking risks, or are quick to
assume risks, such as starting a business or gambling, while other people are more cautious about taking risks, for example
they prefer a secure job with fixed pay to an uncertain venture. On a scale from 1 to 6 where 1 is ‘‘Very willing” and 6 is ‘‘Very
reluctant,” how would you describe your attitude towards taking risks?” (This scale was reversed from the original to show
increasing risk aversion.)
Risk aversion-lottery – A risk attitude score, increasing in risk aversion, was derived from responses to a series of choices
for an amount with certainty, or a 50/50 chance of getting a similar, higher or lower amount in the manner of Eckel and Wil-
son (2004); participants were paid randomly for one of their choices.
8
Faith – A single item: ‘‘On a scale from 1 to 6 where 1 is ‘‘Do not believe/Agnostic” and 6 is ‘‘Very strong belief” how would
you describe your faith in God?”
Optimism – A single item asked participants to rate their own level of optimism or pessimism: ‘‘On a scale from 1 to 6
where 1 is ‘‘Very pessimistic” and 6 is ‘‘Very optimistic,” how would you describe yourself?”
Cheating by others – A single item measured if people think that others are honest if given the opportunity to cheat: ‘‘In
your opinion, when presented with the opportunity, would most students cheat on an exam?” Responded to on a yes/no
scale (0 = No, 1 = Yes).
Probability good – A single item based on the familiar ‘‘lost wallet” experiment, provided in response to the following
statement: ‘‘We would like your opinion about the probability of the occurrence of certain events. Please state your answer
in form of a percent from 0 to 100 (0 means something will never happen, 100 means it certainly will happen). What is the
probability of: Having your lost wallet and its contents ($100 and an ID with your phone number on it) returned to you.”
Probability bad – A single item in response to the following statement: ‘‘We would like your opinion about the probability
of the occurrence of certain events. Please state your answer in form of a percent from 0 to 100 (0 means something will
never happen, 100 means it certainly will happen). What is the probability of: Being in a plane crash while flying a commercial
flight from Minneapolis to New York.”
The next two variables are derived from simulated dictator game games, played with hypothetical money and hypothet-
ical receivers during the online survey portion of the study. The first variable (unconditional kindness) is from standard dic-
tator games where the gender of the receiver is specified; subjects in simulated dictator game experiments give similar
amounts of hypothetical money to what subjects in dictator game experiments with real money give (see Ben-Ner, Kramer,
& Levy, 2008; Ben-Ner, McCall, Stephane, & Wang, 2009). The second variable (reciprocity) is derived from a dictator game in
which the sender was the receiver in a previous dictator game, and received $5; this is a simplified version of the reciprocity
game in Ben-Ner et al. (2004) where senders in the second dictator game previously received any amount between $0 and
$10 in increments of $1.
Unconditional kindness – Average of amount sent in a hypothetical dictator game to a female and to a male, in each case
starting with a $10 endowment.
9
Reciprocity – Amount sent in a hypothetical dictator game to a person who had sent the participant $5 in a previous hypo-
thetical dictator game.
4. Results
Tables 1 and 2 present variable descriptions and descriptive statistics. Table 1 focuses on the 102 subjects who were in
role Ain the trust game, and Table 2 on the 102 subjects who were in role Bin the same trust game. We provide the statistics
separately for the two groups of subjects because we do all analyses within the two groups separately. For the common vari-
ables there are no statistically significant differences between the two groups.
7
For references to the literature that describes and uses the Wonderlic and the NEO-FFI instruments see Ben-Ner et al. (2004).
8
The risk instrument is a series of five decisions, each one a choice between earning $1 with certainty and engaging in a lottery with a 50% chance of earning
nothing and a 50% chance of earning a positive amount: $1.80, $2.00, $2.33, $2.67 and $3.00 in the first, second, third, fourth and fifth decisions, respectively.
Subjects were to make all five decisions and were told that one would be selected at random to be paid off on at their experiment session. The risk aversion
score was only created for subjects who answered the questions in a consistent manner, resulting in 12 subjects not having a score.
9
The hypothetical experiments were introduced as follows. ‘‘The next four questions ask you to consider a simple and completely anonymous experiment in which
you are paired with another participant. The experimenter gives you $10 and asks you to consider what to do with it: keep it all yourself, or send some or all of it to
another participant with whom you have been paired and who is not aware of your identity. You are in the role of sender and your counterpart in the role of receiver.”
The first hypothetical experiment was as follows. ‘‘Imagine that the experimenter has given you $10. The experimenter has also paired you randomly with another
participant in the experiment; in this case, it turns out to be a male student but that is all that you know about that individual. You will remain anonymous to one
another during and after the experiment. You, as the sender, are asked to consider what to do with your $10: keep it all, or send some or all of it to the other participant
with whom you have been paired. You give ___ You keep ___.” The second hypothetical experiment was identical except that the gender of the receiver was
female.
A. Ben-Ner, F. Halldorsson / Journal of Economic Psychology 31 (2010) 64–79 69

Starting with trusting measures, we note that the average amount sent was E$5.47, a little more than half of the initial
endowment, comparable to the average in similar trust games. The survey measure ‘‘trusting GSS” suggests that 62% of sub-
jects state that most people can be trusted (as compared to the 45% Glaeser et al. (2000) found in their sample); on the
‘‘trusting evidence” measure 75% say that they tend to trust a person until there is evidence they cannot be trusted. The mean
of the ‘‘trusting interactions” measure is 3.89, indicating a relatively trusting average tendency. Note that in the case of all
four measures larger values indicate greater trusting.
The behavioral trustworthiness measure is the proportion sent back in the trust game, averaging just under 40%, again
similar to other trust games. The primary survey measure of trustworthiness, the Machiavellianism scale, averages 56.24
in our sample (lower than the average of 73.0
10
reported by Gunnthorsdottir et al. (2002)). The second survey measure, ‘‘trust-
ing is trouble,” averages 2.72 on a scale of 1–5. The larger these two measures are the less likely the degree of trustworthiness.
The ‘‘people no suckers” survey measure of trustworthiness averages 2.56 and is likely to be associated with greater trustwor-
thiness the larger values it assumes, as is the larger the proportion returned.
Table 3 reports the results of the analysis of the association between the two sets of factors and the behavioral measure
and three survey measures of trusting. The number of observations (in this and other tables) is often less than 102 because of
missing observations for one or more variables. For each dependent variable we have three models, one with the birth and
early-childhood variables, another with views and attitudes variables, and a third with the two sets of variables combined. As
noted earlier, the two sets of factors are not independent, the second arguably being determined by the first, so the estimates
in the third model may suffer from multicollinearity and need to be interpreted with care. We used OLS for ‘‘amount sent,”
which is measured in experimental dollars and can assume values between 0 and 10 in increments of one, as well as for
‘‘trusting interactions,” which can assume ordinal values between 1 and 6. Ordered logit analysis for both variables generates
substantially similar results (not reported here). For the ‘‘trusting GSS” and ‘‘trusting evidence,” which are dichotomous vari-
ables, we used logit analysis.
The patterns in Table 3 may be characterized as evidencing a relatively strong relationship between factors determined at
birth and in childhood and our four measures. The relationship between personality and trust measures is notable for survey
measures and to a lesser extent for the ‘‘amount sent.” More agreeable, more extroverted, less conscientious and less
Table 1
Definitions and descriptive statistics of variables for trustors (subjects in role A).
Variable name Question/description Minimum
value
Maximum
value
Mean SD
Amount sent Amount sent in trust game: 0 to 10 experimental dollars 0 10 5.47 3.16
Trusting GSS Generally speaking, would you say that: 1 = most people can be trusted, or
0 = you cannot be too careful when dealing with people
0 1 0.62 –
Trusting
interactions
Interactions with other people: 1 = Relatively cautious, 6 = Relatively trusting 1 6 3.89 1.23
Trusting
evidence
0 = Will not trust until I have clear evidence that a person can be trusted,
1 = Will trust until I have clear evidence that a person can’t be trusted
0 1 0.75 –
Gender Participant gender: 0 = Male, 1 = Female 0 1 0.68 –
White Ethnic background: 0 = non-white, 1 = white 0 1 0.76 –
Age Age of participants 18 43 20.56 3.71
Birth order 0 = first born or only child, 1 = latter born 0 1 0.64 –
GMA General mental ability (Wonderlic Personnel Test). Higher score reflects greater
ability
9 42 30.11 6.03
Neo N Big Five personality trait: neuroticism. Higher score reflects more neuroticism 4 40 20.12 9.03
Neo E Big Five personality trait: extraversion. Higher score reflects more extraversion 11 46 29.83 6.70
Neo O Big Five personality trait: openness. Higher score reflects more openness 16 43 30.68 6.30
Neo A Big Five personality trait: agreeableness. Higher score reflects more
agreeableness
18 45 32.06 5.53
Neo C Big Five personality trait: conscientiousness. Higher score reflects more
conscientiousness
17 46 33.17 6.53
Others not
cheating
When given the opportunity, would most students cheat on an exam? 0 = Yes,
1=No
0 1 0.47 –
Faith in God Faith in God: 1 = Do not believe/Agnostic, 6 = Very strong belief 1 6 3.92 1.87
Risk aversion-
lottery
Choice of lottery over certain amount, with 1 as most risk loving and 6 most risk
averse
1 6 3.14 1.52
Risk aversion-
survey
How would you describe your attitude towards taking risks: 1 = Very wiling,
6 = Very reluctant
1 6 3.48 1.17
Optimism How would you describe yourself: 1 = Very pessimistic, 6 = Very optimistic 2 6 4.35 1.14
Unconditional
kindness
Average amount sent in two hypothetical dictator games with $10 endowment,
one with female recipient and one with male recipient
0 10 3.59 2.40
Probability
good
Probability of good thing happening (having your lost wallet and its contents
($100 and an ID with your phone number on it) returned to you)
0 90 27.65 22.24
Probability bad Probability of bad thing happening (being in a plane crash while flying a
commercial flight from Minneapolis to New York)
0 80 7.49 14.70
10
Gunnthorsdottir et al. (2002) add a constant of 20 to their raw Machiavellianism scores so their actual reported average value is 93.0.
70 A. Ben-Ner, F. Halldorsson / Journal of Economic Psychology 31 (2010) 64–79

neurotic individuals seem to be more trusting in our sample. Female subjects send less money, but this finding does not carry
to other measures, and the female subjects exhibit greater ‘‘trusting GSS.” The views and attitudes included in the estima-
tions show a less consistent effect across behavioral and survey measures. Optimism is positively associated with the survey
measures, but not with the ‘‘amount sent.” ‘‘Unconditional kindness” is positively associated with ‘‘amount sent” but only
insignificantly so with the survey measures. The ‘‘probability good” variable is also positively associated (but not signifi-
cantly) with survey measures. ‘‘Risk aversion-survey” bears no statistically significant relationship to any of our measures,
and the same is true of ‘‘risk aversion-lottery” when substituted for the survey measure (not shown). ‘‘Faith is God” has a
small positive (but statistically insignificant) association with two survey measures but not with ‘‘amount sent.” The belief
that students would not cheat even if they could (‘‘Others not cheating”) has a positive association with survey measures but
not with ‘‘amount sent.” In conclusion, Table 3 suggests that our two sets of factors, each separately and together, explain
fairly well the variation in the measures of trusting.
The definition of trusting in economics links it strongly to risk attitudes, so that the more risk averse are considered less
trusting. We have seen already that neither the behavioral nor the survey risk aversion measures are related to the behav-
ioral or survey measures of trusting in Table 3.InTable 4 we examine, just like in Table 3, the factors that are associated with
‘‘risk aversion-survey” and ‘‘risk aversion-lottery” as dependent variables. The two measures have generally different asso-
ciations with the two sets of factors, and many of the associations differ from those documented in Table 3. This adds evi-
dence to the lack of an empirical relationship between risk attitudes and the various measures of trusting in our sample.
The finding in Table 3 of similarity in the determinants of the various measures suggests the possibility that they are re-
lated to each other. The simple correlations among the various measures presented in Table 5 provide further suggestive evi-
dence of this: ‘‘amount sent” is positively correlated with the survey measures, and the three survey measures are
substantially correlated with each other. In Table 6 we examine the relationship between ‘‘amount sent” and each of the
three survey measures, controlling for the factors determined at birth an in childhood. Table 6 shows that all three survey
measures are positively related to ‘‘amount sent” but only ‘‘trusting evidence” reaches statistically significant levels.
Table 2
Definitions and descriptive statistics of variables for trustees (subjects in role B).
Variable name Question/description Minimum
value
Maximum
value
Mean SD
Proportion
sent back
Proportion sent back in trust game: seven possible values: 0, 1/6th, 2/6th, 3/6th,
4/6th, 5/6th, 6/6th
0 100 39.87 27.47
Machiavellian
scale
Score on Machiavellianism scale. Higher score reflects more Machiavellianism 40 75 56.24 7.34
Trusting is
trouble
Anyone who completely trusts anyone else is asking for trouble: 1 = Strongly
disagree, 5 = Strongly agree
1 5 2.72 1.00
People no
suckers
P.T. Barnum was wrong when he said that there’s a sucker born every minute:
1 = Strongly disagree, 5 = Strongly agree
1 5 2.56 0.90
Gender Participant gender – 0 = Male, 1 = Female 0 1 0.62 –
White Ethnic background: 0 = non-European, 1 = European 0 1 0.80 –
Age Age of participants 17 41 21.31 4.58
Birth order First born or only child vs. latter born: 0 = first born or only child, 1 = latter born 0 1 0.55 –
GMA General mental ability (Wonderlic Personnel Test). Higher score reflects greater
ability
14 41 29.93 5.58
Neo N Big Five personality trait: neuroticism. Higher score reflects more neuroticism 2 40 22.17 8.60
Neo E Big Five personality trait: extraversion. Higher score reflects more extraversion 14 43 29.13 6.88
Neo O Big Five personality trait: openness. Higher score reflects more openness 13 46 30.71 7.06
Neo A Big Five personality trait: agreeableness. Higher score reflects more
agreeableness
10 44 30.49 6.01
Others not
cheating
When given the opportunity, would most students cheat on an exam? 0 = Yes,
1=No
0 1 0.49 –
Faith in God Faith in God: 1 = Do not believe/Agnostic, 6 = Very strong belief 1 6 3.75 1.96
Risk aversion-
lottery
Choice of lottery over certain amount, with 1 as most risk loving and 6 most risk
averse
1 6 3.11 1.58
Risk aversion-
survey
How would you describe your attitude towards taking risks: 1 = Very willing,
6 = Very reluctant
1 6 3.41 1.17
Optimism How would you describe yourself: 1 = Very pessimistic, 6 = Very optimistic 1 6 4.08 1.22
Unconditional
kindness
Average amount sent in two hypothetical dictator games 0 10 3.36 2.49
Probability bad Probability of bad thing happening (being in a plane crash while flying a
commercial flight from Minneapolis to New York)
0 60 6.23 10.55
Probability
good
Probability of good thing happening (having your lost wallet and its contents
($100 and an ID with your phone number on it) returned to you)
0 95 26.54 22.28
Reciprocity Amount sent in hypothetical dictator game to the sender of $5 in a previous
hypothetical dictator game
0 10 3.88 2.56
Trusting GSS Generally speaking, would you say that: 1 = most people can be trusted, or
0 = you cannot be too careful when dealing with people
0 1 0.68 –
Neo C Big Five personality trait: conscientiousness. Higher score reflects more
conscientiousness
16 46 31.24 6.64
A. Ben-Ner, F. Halldorsson / Journal of Economic Psychology 31 (2010) 64–79 71

Table 7 examines the determinants of trustworthiness in a parallel fashion to Table 3. The estimation method is OLS for
the proportion sent back by subjects in role Bin the trust game as well as for the three survey measures. The explanatory
variables are the same as in Table 3, with the addition of ‘‘reciprocity” and ‘‘trusting GSS” in the estimation of the model that
includes views and attitudes for all four dependent variables. ‘‘Reciprocity” is included in order to check whether trustwor-
thiness is indeed related (or the same as) reciprocity. ‘‘Trusting GSS” is included in order to provide comparability with other
studies; its exclusion makes practically no difference to the estimates on the other variables. ‘‘Amount sent” in the estima-
tion of ‘‘proportion sent back” is included to capture reciprocation by subjects in role Bto greater amounts sent (or trusting
exhibited) by subjects in role A.
11
The patterns in Table 7 show a moderate relationship between factors determined at birth and in childhood and our four
measures, especially ‘‘proportion sent back” and ‘‘Machiavellian scale.” The relationship between personality and the Machi-
avellian scale is notable: more agreeable, more extroverted, more open, and more conscientious individuals are less Machi-
avellian in our sample. The signs on the estimates on these variables in the estimation of ‘‘proportion sent back” are
consistent with those in the estimation of the Machiavellian scale, but the estimates for ‘‘proportion sent back” are less pre-
cise. The same is generally true of the other two survey items, ‘‘trusting is trouble” and ‘‘people no suckers.” Female subjects
return a smaller proportion of the money received and show a slightly greater degree of Machiavellianism, but both effects
are statistically insignificant. Higher GMA is associated with a somewhat higher degree of Machiavellianism and ‘‘trusting is
trouble” measure (and a lower but statistically insignificant proportion sent back).
Table 3
Determinants of trusting – amount sent in trust game, and of survey measures of trusting.
*
p< 0.10.
**
p< 0.05.
***
p< 0.01.
11
We excluded from analysis of proportion sent back the six cases in which participants in role Bwere sent nothing, and therefore were forced to send back
nothing.
72 A. Ben-Ner, F. Halldorsson / Journal of Economic Psychology 31 (2010) 64–79

The views and attitudes included in the estimations show an inconsistent effect across behavioral and survey measures.
‘‘Faith in God” has negative but statistically insignificant effect on the proportion sent back, it is associated with lesser
Machiavellianism. Optimism is associated with slightly lower Machiavellianism but no the proportion returned in the trust
game. ‘‘Trusting GSS” is negatively associated with Machiavellianism and positively with ‘‘people no suckers” but is not asso-
ciated with the proportion sent back. Other variables have weaker and generally insignificant effects.
Next we turn to assess the relationship among the four measures we used for trustworthiness. In Table 8 we present
simple correlations among the measures, and note that ‘‘proportion sent back” is mildly and negatively correlated with
the Machiavellian scale and with ‘‘trusting is trouble” but not with ‘‘people no suckers.” The survey measures are mildly
Table 4
Determinants of risk aversion-survey and lottery-based measures.
Standard errors are in parentheses.
*p< 0.10.
**p< 0.05.
***p< 0.01.
Table 5
Correlations among trust measures.
Number of observations are in parentheses.
*p< 0.05.
**p< 0.01.
A. Ben-Ner, F. Halldorsson / Journal of Economic Psychology 31 (2010) 64–79 73

correlated among themselves. In Table 9, which parallels Table 6, we examine the relationship between ‘‘proportion sent
back” and each of the survey measures while controlling for factors determined at birth and in childhood.
12
We find that
there is no significant association between each of the three measures.
5. Discussion
The goal of our empirical work has been to gain a better understanding of the concept of trust, by investigating the deter-
minants and correlates of different measures of trusting and trustworthiness. There is no objective criterion by which to
evaluate which are the right measures of trusting and trustworthiness: the behavioral measures derived from the trust game
and the survey measures capture more than just trusting and trustworthiness and at the same time they do not capture some
of their important facets. In this section we use our findings in conjunction with our conceptual framework to assess the
different measures, and then to evaluate the factors that affect trusting and trustworthiness. We also compare our findings
with those of others, but note the limitations of such comparisons because of the substantial differences in the empirical
frameworks from which they are derived.
13
Our strategy for interpreting the trusting and trustworthiness measures relies on relating them to variables that reflect
subjects’ views and attitudes on matters related to trust, but are not directly focused on trust. We found (in Table 3) that
‘‘amount sent” is related only to ‘‘unconditional kindness” – not to risk attitudes (either lottery-based or survey-based),
nor to optimism, nor to exaggeration of the likelihood that good or bad things happen, nor to the view that others may cheat,
nor to faith in God.
14
About 40 cents of a dollar sent by Aare explained by the amount he or she sent in the dictator game.
15
What
Table 6
Relationship between amount sent by Ain the trust game and survey measures of trusting (OLS).
Standard errors are in parentheses.
*p< 0.10.
12
As in Table 7, we omit the participants who received nothing.
13
Different studies use widely different sets of explanatory variables (and that are often defined in different ways), often rely on small numbers of
observations, and employ different trust game parameters (e.g., sometimes the trustee does not receive an endowment, or the amount sent is doubled rather
than tripled).
14
Tan and Vogel (2008) find a positive association between religious belief and amount sent.
15
Ashraf et al. (2006) and Sapienza et al. (2007) find a strong effect from ‘‘unconditional kindness,” also derived from a dictator game. Ashraf et al. (2006),
Sapienza et al. (2007), and Buchan, Croson, and Solnik (2003) are among those who include a variable reflecting the amount expected to be returned by subjects
in role B. Ashraf et al. interpret a large regression constant as indication of an intrinsic component for sending; our constants are mostly statistically
insignificant.
74 A. Ben-Ner, F. Halldorsson / Journal of Economic Psychology 31 (2010) 64–79

Table 8
Correlations among trustworthiness measures.
Standard errors are in parentheses.
*p< 0.10.
**p< 0.05.
***p< 0.01.
Table 7
Determinants of trustworthiness – proportion returned in trust game, and of survey measures of trustworthiness.
*
p< 0.10.
**
p< 0.05.
***
p< 0.01.
A. Ben-Ner, F. Halldorsson / Journal of Economic Psychology 31 (2010) 64–79 75

is the rest? It is possible that some of the rest of the amount sent is explained by an investment motive felt by people who want to
make money and are willing to trust their counterparts. In addition to altruism, the amount sent is explained by gender – men
send more
16
– and perhaps by higher extraversion and lower conscientiousness. This is compatible with the view that a portion of
the amount sent is for investment or gambling reasons, more so by men and extravert individuals who are lower on the consci-
entiousness trait than others.
17
Why this is not being picked up by measures of risk attitudes is not clear, but the strong motive
for sending money in the expectation of a high return is found in all studies that control for that. The fact that risk aversion is
determined by factors that are entirely different from those that determine amount sent further suggests that these measure dif-
ferent things (compare estimates Tables 3 and 4).
18
The survey measures of trusting are explained mainly by optimism, and in the case of ‘‘trusting GSS,” also by the view that
others are not cheating. There is no association between the survey measures of trusting and unconditional kindness, or faith
in God, or estimation of probabilities of good and bad events or risk attitudes. It makes sense that an optimistic attitude is
associated with trusting in others not for investment purposes but for their willingness and competence to act favorably
rather than maliciously or even make small sacrifices (the second, third and fourth facets of the definition of trusting in Sec-
tion 2.1). These facets of trusting need not be related to an individual’s unconditional kindness towards others; sharing an
endowment with others in the dictator game – or in the trust game – may be a mark of fairness or kindness, but not nec-
essarily of trusting in the sense of these facets. The determinants of the survey measures – high agreeableness and low extra-
version, and to some degree low neuroticism – make sense in this interpretation of the facets of trusting not motivated by
the desire to make a profit.
19
Table 9
Relationship between proportion returned by Bin the trust game and survey measures of trustworthiness (OLS).
Standard errors are in parentheses.
*p< 0.10.
**p< 0.05.
16
This is also found by Buchan et al. (2003) and Burks et al. (2003), Ashraf et al. (2006), and Swope, Cadigan, Schmitt, and Shupp (2008) find no gender effect.
17
Swope et al. (2008) also find that extraversion (using a different personality instrument than ours) is associated with more sending in the trust game. Burks
et al. (2003) find a positive relationship between greater Machiavellianism and amount sent.
18
A clear connection between sending in the trust game and risk attitudes has not been established, and the results are mixed. Schechter (2007) finds the
expected relationship, but Eckel and Wilson (2004) and Ashraf et al. (2006) do not. Fehr (forthcoming) summarizes neurobiological literature that supports,
indirectly, the connection between risk attitudes and trusting.
19
Agreeableness is frequently characterized as a tendency to be compassionate, trusting and cooperative rather than suspicious, inconsiderate and
pessimistic. Extraversion is often described as the tendency to have positive emotions and show optimism. See, for example, Briggs (1992).
76 A. Ben-Ner, F. Halldorsson / Journal of Economic Psychology 31 (2010) 64–79

The behavioral measure and survey measures of trusting show positive but weak relationships (in Table 6): statistically
insignificant relationship between amount sent and ‘‘trusting GSS” and ‘‘trusting interactions” and statistically significant
with ‘‘trusting evidence.” This is in line with what other researchers have found.
20
To sum up the discussion of trusting measures, we interpret our findings as confirmation of the views we expressed in
Section 2. The behavioral measure, amount sent, measures both less and more than trusting. It captures altruism or uncon-
ditional kindness towards others, and probably trusting in the sense that it is commonly interpreted in the context of the
investment game (willingness to risk one’s investment in the expectation to make a return on it). The survey measures cap-
ture trusting in more general situations, when others are expected to behave generously and competently when one is in
need of assistance. The different facets of trusting captured by the behavioral and survey measures are related to each other
and reflect mildly overlapping attitudes, but they are clearly not measuring the same thing.
Consider now the measures of trustworthiness. ‘‘Proportion sent back” (in Table 7) is not related to the amount sent by
trustors
21
or to the ‘‘reciprocity” variable. This seems to eliminate reciprocation as an explanation of trustworthiness. The pro-
portion returned is not related to other views and attitudes, including ‘‘trusting GSS”
22
or ‘‘unconditional kindness,” eliminating
altruism or unconditional kindness as a possible explanatory variable for the proportion returned in the trust game.
23 ,24
It
therefore appears that the proportion returned reflects a sense of obligation to reward trustor’s investment (rather than recip-
rocation of trusting), that is, trustworthiness. This interpretation is supported by the positive effect of the agreeableness per-
sonality trait, which suggests that individuals who are more agreeable will respond more favorably to an obligation imposed
on them.
Of the survey measures, the Machiavellian scale shows a strong relationship to the views and attitudes included in the
regression. More Machiavellianism is associated with less optimistic, reciprocal and trusting attitudes; whether such atti-
tudes are linked to the probability that a person will not respond favorably and competently to implied demands for assis-
tance is hard to assess. Viewed from the perspective of determinants of Machiavellianism, the profile of a high scorer may be
partly associated with lower trustworthiness. The profile is that of a person low on agreeableness, openness, extraversion
and conscientiousness, but high on cognitive skills, which fits well with a tendency to cheat (Gunnthorsdottir et al.,
2002) and thus may be thought of as a reflection of one’s trustworthiness. Given the clear opportunity to ‘‘cheat” one’s coun-
terpart in the investment game, Machiavellianism could be expected to be more strongly related to proportion sent back
than is observed here. Although it is positively correlated with proportion sent back (in Table 8), once we control for birth
and childhood factors (in Table 9) the correlation essentially goes away,
25
and what remains is agreeableness as the single best
predictor of the proportion sent back. The other two survey measures, which are single items from the Machiavellianism scale,
do not differ much from the full scale.
We interpret our findings to suggest that the proportion sent back in the trust game is a good measure of trustworthiness,
at least of the facet that reflects on one’s willingness not to take advantage of the vulnerability another person created in the
expectation of a financial gain. The Machiavellian scale appears to capture some aspects of trustworthiness, but it is a poor
candidate for a survey measure. There is a clear research opportunity to develop a better understanding of the concept of
trustworthiness and of simple survey ways to measure it.
6. Conclusions
Trusting and trustworthiness have multiple facets. The commonly-used behavioral measure of trusting, the amount sent
in the Berg et al. (1995) trust or investment game, represents a good measure of an important facet of trusting, the willing-
ness to take a chance that an investment will be rewarded by a stranger. But a substantial portion of this measure (perhaps
40% of it) captures more than just trusting, as it reflects unconditional kindness towards others, and this measure does not
seem to represent well other facets of trusting. Such facets, representing what most people probably think of trusting – will
you be there for me should I need you? – are captured better by simple survey measures, such as the commonly-used Gen-
eral Social Survey one. The various facets of trusting are determined by different inherited or childhood-influenced factors.
Most of the variation in the amount sent in the trust game is explained by gender (females send less), with weak influence
from personality traits. In contrast, the facets captured by surveys are heavily influenced by personality in the way probably
most researchers would intuit. However, risk attitudes are not related to trusting (or to trustworthiness).
20
Glaeser et al. (2000) found ‘‘trusting GSS” not to be related to amount sent, but other measures, similar to ours, were related; Fehr et al. (2003) found similar
results, Lazzarini, Madalozzo, Artes, and Siqueira (2003) obtained mixed results, whereas Holm and Danielson (2005) found some differences in these
relationships between Tanzania and Sweden. Results in the same spirit (comparing survey trust measures with behavioral measure from a public goods game)
were reported by Gächter, Herrmann, and Thöni (2004).Burks et al. (2003) find no relationship between amount sent and ‘‘trust in strangers” measure.
21
Buchan et al. (2003), Ashraf et al. (2006), Swope et al. (2008), and Sapienza et al. (2007) find mild positive effects. However, there is no clear indication that
these authors eliminated observations in which trustees received nothing, therefore returned nothing, which may contribute to a positive relationship between
‘‘amount sent” and ‘‘proportion sent back” (as it would be the case if we included these cases in our analysis; see also footnote 11, above).
22
Glaeser et al. (2000) and Sapienza et al. (2007) find such relationships, and Ashraf et al. (2006) find a similar relationship to a related trusting measure.
23
Ashraf et al. (2006) find a positive effect.
24
We find no significant gender effect; Burks et al. (2003), Ashraf et al. (2006), Sapienza et al. (2007), and Swope et al. (2008) find the same, whereas Buchan
et al. (2003) find that females return a little more than men. We also find no significant correlation with faith in God; Tan and Vogel (2008) find a positive
association with religious belief. Birth order approaches significance: latter born individuals return a higher proportion of what they received.
25
Burks et al. (2003) do not find a relationship between Machiavellianism and proportion returned, but Gunnthorsdottir et al. (2002) do.
A. Ben-Ner, F. Halldorsson / Journal of Economic Psychology 31 (2010) 64–79 77

The behavioral measure for trustworthiness is the proportion returned by trustees in the trust game. We conclude that
this is good measure of the facet of trustworthiness that reflects the sense of obligation implicitly imposed on a person by
another who expects a monetary gain and not to be exploited. The personality trait agreeableness is the main determinant of
this facet of trustworthiness. The Machiavellian scale does not seem to be a good measure of any particular facet of trust-
worthiness, although it is probably mildly (negatively) correlated with it.
We reached these conclusions by undertaking an empirical approach that was prescribed by a theoretical framework in
which we distinguished between birth and childhood-influenced factors that determine attitudes, including trusting and
trustworthiness, and views and attitudes that are likely to be linked to trusting and trustworthiness by association rather
than causality. This distinction allowed us to attach meaning to various measures, and to assess the factors that determine
trusting and trustworthiness.
Our study has familiar limitations. The student sample is not representative of the diverse people around us, and it is pos-
sible that our findings are not representative of other groups. A useful research direction would be to explore the relation-
ships we examined here in other samples. More importantly, a deeper understanding of the different facets of trusting and
trustworthiness, how they are related to risk aversion, what experiences account for them and how can they be enhanced
would be a valuable research contribution. The continued search for usable and reliable measures is an important step in
this direction.
Acknowledgements
We thank the Russell Sage Foundation for a grant to Ben-Ner and Louis Putterman. We are grateful to the editor and ref-
erees for extremely helpful comments, and to comments received at presentations at Cornell University, University of Min-
nesota Law School, and a SABE annual conference.
Appendix A. Decision (interaction) grid
78 A. Ben-Ner, F. Halldorsson / Journal of Economic Psychology 31 (2010) 64–79

References
Ajzen, I. (2005). Attitudes, personality and behavior (2nd ed.). Milton-Keynes, England: Open University Press/McGraw-Hill.
Allik, J., Laidra, K., Realo, A., & Pullmann, H. (2004). Personality development from 12 to 18 years of age: Changes in mean levels and structure of traits.
European Journal of Personality, 18, 445–462.
Ashraf, N., Bohnet, I., & Piankov, N. (2006). Decomposing trust and trustworthiness. Experimental Economics, 9, 193–208.
Bahry, D., Kosolapov, M., Kozyreva, P., & Wilson, R. W. (2005). Ethnicity and trust: Evidence from Russia. American Political Science Review, 99, 521–532.
Ben-Ner, A., McCall, B. P., Stephane, M., & Wang, H. (2009). Identity and in-group and out-group differentiation in work and giving behaviors: Experimental
evidence. Journal of Economic Behavior and Organization, 72, 153–170.
Ben-Ner, A., Kong, F., Magan, D., & Putterman, L. (2004). Reciprocity in a two-part dictator game. Journal of Economic Behavior and Organization, 53, 333–352.
Ben-Ner, A., Kong, F., & Putterman, L. (2004). Share and share alike? Gender-pairing, personality, and cognitive ability as determinants of giving. Journal of
Economic Psychology, 25, 581–589.
Ben-Ner, A., Kramer, A., & Levy, O. (2008). Economic and hypothetical dictator game experiments: Incentive effects at the individual level. Journal of Socio-
Economics.
Ben-Ner, A., & Putterman, L. (2001). Trusting and trustworthiness. Boston Law Review, 81, 523–551.
Ben-Ner, A., & Putterman, L. (2009). Trust, communication, and contracts: An experiment. Journal of Economic Behavior and Organization, 70, 106–121.
Berg, J., Dickhaut, J., & McCabe, K. (1995). Trust, reciprocity, and social history. Games and Economic Behavior, 10, 122–142.
Briggs, S. R. (1992). Assessing the five-factor model of personality description. Journal of Personality, 60, 253–293.
Buchan, N. R., Croson, R., & Solnik, S. (2003). Trust and gender: An examination of behavior, biases, and beliefs in the investment game. Journal of Economic
Behavior and Organization, 68, 466–476.
Burks, S., Carpenter, J., & Verhoogne, E. (2003). Playing both roles in the trust game. Journal of Economic Behavior and Organization, 51, 195–216.
Christie, R., & Geis, F. (1970). Studies in machiavellianism. New York: Academic Press.
Cox, J. (2004). How to identify trust and reciprocity. Games and Economic Behavior, 46, 260–281.
Croson, R., & Gneezy, U. (2008). Gender differences in preferences. Journal of Economic Literature, 47, 1–27.
Eckel, C. C., & Wilson, R. K. (2004). Is trust a risky decision? Journal of Economic Behavior and Organization, 55, 447–465.
Fehr, E. (forthcoming). On the economics and biology of trust. Journal of the European Economic Association.
Fehr, E., Fischbacher, U., von Rosenbladt, B., Schupp, J., & Wagner, G. G. (2003). A nation-wide laboratory. Examining trust and trustworthiness by integrating
behavioral experiments into representative surveys. Institute for Empirical Research in Economics, University of Zürich, Working Paper No. 141.
Gächter, S., Herrmann, B., & Thöni, C. (2004). Trust, voluntary cooperation, and socio-economic background: Survey and experimental evidence. Journal of
Economic Behavior and Organization, 55, 505–531.
Glaeser, E. L., Laibson, D. I., Scheinkman, J. A., & Soutter, C. L. (2000). Measuring trust. The Quarterly Journal of Economics, 115, 811–846.
Gunnthorsdottir, A., McCabe, K., & Smith, V. (2002). Using the Machiavellianism instrument to predict trustworthiness in a bargaining game. Journal of
Economic Psychology, 23, 49–66.
Harbaugh, W. T., Krause, K., Liday Jr., S. G., & Vesterlund, L. (2003). Trust in children. In: E. Ostrom & J. Walker (Eds.), Trust and reciprocity, interdisciplinary
lessons from experimental research. NY: Russell Sage Foundation.
Hardin, Russell (2004). Trust and trustworthiness. NY: Russell Sage Foundation.
Hill, C. A., & O’Hara, E. A. (2006). A cognitive theory of trust. Washington University Law Review, 87, 1717–1796.
Holm, H. J., & Danielson, A. (2005). Tropic trust versus Nordic trust: Experimental evidence from Tanzania and Sweden. The Economic Journal, 115, 505–532.
Lazzarini, S. G., Madalozzo, R. C., Artes, R., & Siqueira, J. O. (2003). Measuring trust: An experiment in Brazil. Ibmec Working Paper No. 40, Ibmec Business
School of São Paulo.
McCrae, R. R., & Costa, P. T. Jr., (1999). A five-factor theory of personality. In L. A. Pervin & O. P. John (Eds.), Handbook of personality: Theory and research (2nd
ed., pp. 139–153). New York: Guilford Press.
Rotter, J. B. (1980). Interpersonal trust, trustworthiness, and gullibility. American Psychologist, 35, 1–7.
Sapienza, P., Toldra, A., & Zingales, L. (2007). Understanding trust. CRSP Working Paper No. 621.
Schechter, L. (2007). Traditional trust measurement and the risk confound: An experiment in rural Paraguay. Journal of Economic Behavior and Organization,
62, 272–292.
Sulloway, F. (1996). Born to Rebel: Birth order, family dynamics, and creative lives. New York: Pantheon.
Swope, K. J., Cadigan, J., Schmitt, O. M., & Shupp, R. (2008). Personality preferences in laboratory economics experiments. Journal of Socio-Economics, 37,
998–1009.
Tan, J. H. W., & Vogel, C. (2008). Religion and trust: An experimental study. Journal of Economic Psychology, 29, 832–848.
A. Ben-Ner, F. Halldorsson / Journal of Economic Psychology 31 (2010) 64–79 79