This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution license (http://creativecommons.org/licenses/by/3.0/).
The relationship between risk in the environment, risk aversion and inequality aversion is not well understood. Theories of fairness have typically assumed that pie sizes are known exante. Pie sizes are, however, rarely known ex ante. Using two simple allocation problems—the Dictator and Ultimatum game—we explore whether, and how exactly, unknown pie sizes with varying degrees of risk (“endowment risk”) influence individual behavior. We derive theoretical predictions for these games using utility functions that capture additively separable constant relative risk aversion and inequity aversion. We experimentally test the theoretical predictions using two subject pools: students of Czech Technical University and employees of Prague City Hall. We find that: (1) Those who are more riskaverse are also more inequalityaverse in the Dictator game (and also in the Ultimatum game but there not statistically significantly so) in that they give more; (2) Using the withinsubject feature of our design, and in line with our theoretical prediction, varying risk does not influence behavior in the Dictator game, but does so in the Ultimatum game (contradicting our theoretical prediction for that game); (3) Using the withinsubject feature of our design, subjects tend to make inconsistent decisions across games; this is true on the level of individuals as well as in the aggregate. This latter finding contradicts the evidence in Blanco
The relationship between risk aversion and inequality aversion is not well understood. It has been noted, however, that they are closely related in certain choice situations.
The Dictator game and the Ultimatum game are standard decision situations that feature this tradeoff. The theoretic prediction for the Dictator game, under the standard assumption of selfish preferences, is zero giving. For the Ultimatum game, the theoretic prediction is zero giving and an acceptance threshold of zero or the smallest monetary unit above zero. The experimental evidence, however, shows that there is significant giving in both games, and that thresholds are greater than the smallest monetary unit in the Ultimatum game. Specifically, transfers to the recipient are about twenty percent of the pie size for Dictator games (but see Cherry
Bolton & Ockenfels [
Blanco
This paper contains three contributions. First, we study the relationship between risk aversion and inequality aversion using properly incentivized Dictator and Ultimatum games, with varying degrees of risk (“endowment risk”).
Our predictions are presented in
In the Dictator game, a dictator decides on the proportion
The relationship between risk aversion and inequality aversion can also be tested with an Ultimatum game. As in the Dictator game, a proposer decides on the proportion
For Ultimatum giving decisions we cannot derive predictions from theory as we do not know the expectations that givers had about the inequality aversion of the receiver they would be paired with.
We extend our analysis by exploiting the withinsubject feature of our design (which is explained in more detail in
Let the motivation function (
In Equation (1),
To increase the precision of our predictions, we introduced additional assumptions about our ERC specification. The literature suggests that a utility function in the constantrelativeriskaversion form is a suitable approximation for human behavior.
We indicate relative risk aversion by
As
In the Ultimatum game, a responder evaluates the offer
The responder rejects the offer when her motivation function is negative; that is, when her expected utility is lower than the disutility from having an unequal distribution. At the threshold, the lowest offer the responder accepts requires the motivation function to equal zero. Rearranging (2’), the condition is:
Equations (4) and (4’) imply that the level of risk may have an effect on choices in the Dictator and Ultimatum games. For subjects with a typical degree of risk aversion, 0 <
Numerical simulations, however, show that the effect of even a considerable increase in risk should result in minor differences in Dictator and Ultimatum game choices.
We also extend our analysis and exploit the withinsubject feature of our design by deriving predictions for withinsubject consistency across games (but keeping the endowment risk constant). Following Blanco
Rearranging (4’) gives:
Equation (5) derives
We can then derive prediction intervals for giving choices in the Dictator game from the observed threshold choice in the Ultimatum game, and vice versa. As shown in Equation (6), the relationship depends on the degree of risk aversion,
Predictions across games based on the ERC formula.
The relation between Ultimatum thresholds and Dictator giving is nondecreasing, as increases in inequality aversion will increase both Dictator giving and threshold in the Ultimatum game. We thus formulate:
To make hypotheses 1D and 1U experimentally testable, we specified endowments with increasing risk (“scenarios”) for Dictator and Ultimatum games. In particular, we provided endowments in the form of lottery tickets
Operationalization of endowment risk.
Possible realizations of the pie sizes S^{i} (Dictator and Ultimatum)  

Games  Norisk condition

Lowrisk condition

Highrisk condition

Dictator    900 or 1100  300 or 1700 
Ultimatum  1000  900 or 1100  300 or 1700 
We employed two different subject pools: students of the Czech Technical University (CTU) and employees of Prague City Hall (CH). The subject pool consisted of 44 CH employees and 116 CTU students.
The experiment was conducted at CERGEEI on a portable experimental laboratory. We conducted 14 experimental sessions.
Overview of experimental sessions.^{15}
Session  1  2  3  4  5  6  7  8  9  10  11  12  13  14  Total 

Type of subjects  CTU  CTU  CTU  CTU  CTU  CTU  CTU  CTU  CTU^{R}  CTU^{R}  CH  CH  CH  CH  
















The experiment was programmed in zTree [
Each round random rematching was used. As mentioned, subjects were informed of the outcomes of their decisions in the Dictator and Ultimatum games at the end of a session.
Since the model predicts that risk aversion influences individual decision making, we categorized—as mentioned—all subjects according to their risk aversion through an additional scenario (see Scenario Four in the Instructions, in Appendix A.5). Similar to the procedure in Holt & Laury [
Elicitation of risk aversion.
Safer option  EV  Riskier option  EV  


1000  if n>  40  ,  1250  otherwise  1100  60  if n>  40  ,  2400  Otherwise  996 

1000  if n>  50  ,  1250  otherwise  1125  60  if n>  50  ,  2400  Otherwise  1230 

1000  if n>  60  ,  1250  otherwise  1150  60  if n>  60  ,  2400  Otherwise  1464 

1000  if n>  70  ,  1250  otherwise  1175  60  if n>  70  ,  2400  Otherwise  1698 

1000  if n>  80  ,  1250  otherwise  1200  60  if n>  80  ,  2400  Otherwise  1932 
Implied ranges of risk aversion r.
Number of safe choices  Range of Relative Risk Aversion

Risk Preference Classification  


1.15  ∞  Riskloving 

0.86  1.15  Riskneutral 

0.60  0.86  Somewhat Riskaverse 

0.33  0.60  Intermediately Riskaverse 

0.04  0.33  Very Riskaverse 

–∞  0.04  Highly Riskaverse 
As in Holt & Laury [
To test whether inconsistent subjects are different from consistent subjects, we ran twosample KolmogorovSmirnov tests on sociodemographic variables. The distribution of inconsistent subjects does not differ for the sex and income variables, but differ for the variables
We note that, on average, Dictator giving is 19%, Ultimatum offers are 43%, and Ultimatum thresholds are 26%, which is in line with previous findings in the literature (see Camerer [
Before analyzing our game data, we characterize the determinants of risk aversion for “consistent” subjects through regression analyses.
Linear regression of risk aversion (defined as the safe choices) on sociodemographic variables.
Variables  Effects 




0.05 (0.03) 



0.06 (0.04) 

–0.63 (0.82) 

0.07 (0.50) 


Both students and City Hall employees are, on average, soundly riskaverse.
We use linear regression models clustered at the individual level to analyze the game data. We also use the robust Huber/White/sandwich estimator for the variance (Froot [
Linear regression of giving on risk aversion and risk in the Dictator and Ultimatum games (with clustered errors).
Dictator  Ultimatum  

Model 1D1  Model 1D2  Model 1D3  Model 1D4  Model 1U1  Model 1U2  Model 1U3  Model 1U4  




5.81  4.68  6.73  5.59  




(3.96)  (4.08)  (4.10)  (4.24)  
−0.97  −0.97  0.30  0.30 





(1.81)  (1.83)  (3.58)  (3.62) 





1.39  1.39  1.63  1.63  
(0.86)  (0.86)  (1.41)  (1.42)  
−1.80  −1.80  −2.40  −2.40  
(4.15)  (4.19)  (3.19)  (3.21)  
−0.34  −0.34  
(1.77)  (1.78)  
−0.00  −0.00  −0.64  −0.64  
(0.49)  (0.50)  (0.51)  (0.51)  
−0.32  −0.32  −0.21  −0.21  
(0.32)  (0.32)  (0.27)  (0.27)  
3.81  3.81  −1.95  −1.95  
(3.65)  (3.66)  (3.52)  (3.53)  











3.98 

3.98  −2.93  2.67  −2.93  2.67  

(6.36) 

(6.38)  (4.08)  (7.08)  (4.09)  (7.10)  
−3.87  −6.46  −3.87  −6.46  −4.38  −5.26  −4.38  −5.26  
(5.16)  (4.57)  (5.18)  (4.58)  (4.94)  (4.95)  (4.96)  (4.97)  



















204  204  306  306  

(102)  (102)  204  204  (102)  (102)  306  306 

0.08  0.21  0.08  0.21  0.04  0.07  0.04  0.08 
Robust standard errors in parentheses ***
The models 1D1 and 1U1 test the hypotheses 1D and 1U with the simplest specifications. In addition, we run regressions including the sociodemographic variables (models 1D2 and 1U2), and we also run—as robustness tests—regressions with interaction dummies to test if there is a significant interaction between endowment risk and risk aversion (models 1D3, 1D4, 1U3 and 1U4).
Model 1D1 provides support for hypothesis 1D: subjects with relatively high risk aversion give significantly (
Model 1U1 does not provide support for hypothesis 1U: while subjects with relatively high risk aversion set, in line with hypothesis 1U, higher thresholds in the Ultimatum game compared to those with relatively low risk aversion, the effect is not statistically significant. Surprisingly, endowment risk has a highly significant effect (
At first glance, there appears to be a subjectpool effect in Model 1D1, as City Hall employees tend to give significantly (
Even though, as mentioned, we cannot derive theoretical predictions for giving in the Ultimatum game, we ran a regression of Ultimatum giving on the same variables as in
Since we can only test hypotheses 2D and 2U by including somewhat riskaverse (but not riskloving) subjects and very riskaverse (but not highly riskaverse) subjects, the number of observations is relatively small. In particular, we had to exclude not only inconsistent subjects, whose risk preference cannot be measured with sufficient precision to test hypothesis 1 and 2, but also highly riskaverse (5 safe choices) and riskloving (0 safe choices) subjects, as they are not within the domain of hypothesis 2.
To measure the effect of endowment risk, we include a dummy variable
Model 2D1 and 2U1 test the hypotheses with the simplest specifications. We also run regressions including the sociodemographic variables (models 2D2 and 2U2), and for hypothesis 2D we also run—as robustness tests—Tobit regressions (models 2D3 and 2D4) since many subjects made a choice of zero in the Dictator game.
Supporting hypothesis 2D, model 2D1 shows that the dummy for highendowment risk is not statistically significant. The dummy for highendowment risk is also insignificant in Model 2D2 (including the sociodemographic variables) and in models 2D3 and 2D4 (the robustness tests using Tobit regressions). Hypothesis 2D is thus supported by our data.
Contradicting hypothesis 2U, model 2U1 shows that both dummies for highendowment risk (
Linear regression of giving on risk aversion and endowment risk in the Dictator and Ultimatum games (with clustered errors).
Dictator  Ultimatum  

Model 2D1  Model 2D2  Model 2D3(Tobit)  Model 2D4(Tobit)  Model2U1  Model2U2  

−0.83  −0.83  −0.92  −0.93 


(1.96)  (2.02)  (2.52)  (2.50) 











1.16  2.09  −2.05  
(1.17)  (1.42)  (1.55)  

−0.49  −0.86  −0.08  
(0.43)  (0.59)  (0.54)  

5.32  9.48  −1.20  
(5.88)  (7.71)  (6.69)  



0.55  


(1.36)  


2.19 

−3.74  −4.21  16.33 

(12.07) 

(15.63)  (8.58)  (17.12)  

−3.53  −10.71  −8.78  −20.01  −1.88  −1.23 
(7.14)  (7.88)  (12.71)  (13.97)  (6.80)  (7.80)  


15.67 

12.45 



(10.53) 

(13.50) 




80  80  80  80  120  120 

(40)  (40)  (40)  (40)  60  60 

0.11  0.22  (0.02)  (0.04)  0.02  0.09 
Robust standard errors in parentheses ***
As in the regressions for testing hypothesis 1, there appears to be a subjectpool effect in Model 2D1 and 2D3, as City Hall employees tend to set significantly higher thresholds (
We used Equation (6) and subjects’ threshold choices in the Ultimatum game to derive prediction intervals from the observed threshold choices in the Ultimatum game for their giving choices in the Dictator game, and vice versa. This is a test of consistency between the two games on the individual level.
Observed and predicted data on the individual level.
Successful  Unsuccessful  Total  

Predict D from U and visa versa  8 (10%)  75 (90%)  83 (100%) 
Observed and predicted data on the aggregate level.
We also examined how well the ERC formulation predicts aggregate behavior.
Summarizing the results, Hypothesis 1D and 2D are supported. Specifically, those who are more riskaverse are also more inequalityaverse in the Dictator game (1D). Though the sign is correct for the Ultimatum game responder (IU), it is not significant, possibly because the number of observations was too low. This tentatively supports our first hypothesis, that those with higher risk aversion are more inequalityaverse (and thus possess a stronger preference for fairness) than those with lower risk aversion. Our finding that more riskaverse people tend to be more inequalityaverse is roughly in line with the results in FerreriCarbonell & Ramos [
Using the predictions derived from ERC theory, as implemented by a simple constantrelativeriskaversion utility function (2), we find that endowment risk has no effect on giving in the Dictator game, confirming hypothesis 2D. Endowment risk does have a significant positive effect on the acceptance threshold in the Ultimatum game, contradicting hypothesis 2U. The effect, an increase in endowment risk leading to an increase in threshold, is in line with the prediction, but the size of the effect is much larger than predicted. This indicates that risk, which is not an ingredient of ERC theory, may affect acceptance thresholds in Ultimatum games.
We can thus conclude that the data corroborate the predictions from specification of the extended ERC formula for the Dictator game (1D and 2D), but not for the Ultimatum game (1U and 2U). We find that subjects are not consistent across games at the individual level, which contradicts hypothesis 3 but is consistent with the results of Blanco
It is important to recall that people make inconsistent choices as viewed from the perspective of the theory that we use. One might conjecture that the functional form we used to incorporate risk preferences in the ERC formula may not be appropriate for the Dictator or Ultimatum game. The fact that our test of the effect of endowment risk on subject behavior, Hypothesis 2, is confirmed for the Dictator game, but contradicted for the Ultimatum game, suggests that the functional form we used may be appropriate for the Dictator game, but not for the Ultimatum game. We note that our test predicting the responses in one game from those in the other game is rather demanding test. These predictions are conditional on the risk preferences, which have been derived from our variant of the Holt & Laury [
Our data thus give tentative support to the claim that there is a positive relationship between risk aversion and fairness considerations. Our data also suggests that ERC theory, as formulated, does not seem particularly well suited to account for the effects of risk in the environment and risk preferences across the Dictator and Ultimatum games. The ERC formulation seems to fare better for the Dictator game.
We tested our theoretical predictions experimentally on two different subject pools: students of Czech Technical University—a subject pool we have drawn on previously that produced behavior in line with the behavior of student subjects elsewhere [
To summarize, we find that: (1) Those who are more riskaverse are also more inequalityaverse in the Dictator game in that they give more. We believe this finding is novel. We find a similar result for the Ultimatum game but that result is statistically not significant; (2) Using the withinsubject feature of our design, and in line with our theoretical prediction, varying risk does not influence behavior in the Dictator game, but does so in the Ultimatum game; (3) Using the withinsubject feature of our design, subjects tend to make inconsistent decisions across games; this is true on the level of individuals (confirming the findings in Blanco
A reasonable intuition for that finding could be this: encountering states of inequality in the world makes one aware of the danger that one self may end up in some such state. This is likely to be the more threatening for a person the more s/he is riskaverse. This threat might activate, or intensify, inequality aversion.
Cherry
Engelmann & Strobel [
Recent experimental work conducted in parallel by Güth
In addition, we also studied the socalled Trust game in such an environment (see Appendices A.5, A.6, and A.8). Assuming selfregarding preferences, the situation for responders in the Trust game is theoretically equivalent to that of dictators in a standard Dictator scenario. However, responders in our design and implementation of the Trust game cannot infer precisely the proposer’s initial decision (because the amount sent is multiplied by an unknown random factor X), and proposers cannot foresee the responders’ reaction. Proposers probably developed beliefs about responders’ behavior, but we were not able, given time constraints, to control for these during the experiment. We are therefore not able to theoretically derive the effects of risk preferences for proposing and responding in the Trust game and therefore decided not to include the Trust game in our analysis. We note that none of the relationships turned out significant for the Trust game.
In principle, the recipient’s proposed share can also be determined in absolute terms. There are three reasons why we did not use absolute terms. First, in theories of fairness and reciprocity only relative terms matter. Second, an exante allocation in absolute terms could result in a negative outcome for the decision maker (when the realized pie is small), which might trigger loss aversion and confound our results. Third, in the present paper we are not interested in optimal contract design (this could solve the preceding problem, but would also complicate our design beyond what seems feasible to implement.)
Note that we have not used the ERC formulation to derive our hypotheses regarding the relation between risk aversion and inequality aversion in the Dictator and Ultimatum game, as this formulation is moot on the possible effects of risk aversion on inequality aversion. Predictions could be derived for people with different risk preferences, if the inequality aversion parameter
This can easily be reformulated in terms of Fehr & Schmidt [
For standard stakes (such as the ones in our experiment) the constantrelativeriskaversion form of the utility function can be rationalized experimentally by the results of Holt & Laury [
Note that always
For riskloving (
See Appendix A.3.
Note that for certain Ultimatum threshold levels no prediction of positive Dictator giving exists. For example, for Highly Riskaverse subjects, an Ultimatum threshold of 0.23 predicts zero Dictator giving. Ultimatum thresholds
We ran, at the tail end of the CTU sessions, two control sessions with an additional 32 CTU student subjects (sessions 9 and 10, indicated by CTU^{R} in
Each session contained an even number of participants, and was constrained by the maximum lab capacity of 16 people. We attempted to have at least 12 participants in each session but scheduling the four CH sessions turned out to be difficult. There is no apriori reason that we can think of that would suggest this was more than a logistic inconvenience.
Subjects were thus paid twice as recipients in the Dictator game (once for the lowrisk and once for the highrisk condition).
In the Trust game, subjects were informed of the amount they received once they were asked at the end of a session to make a decision as responder.
When the experimental sessions were conducted, the exchange rate was about 23 CZK/1 USD, implying that our subjects—not counting appearance fees—earned on average 17–18 USD. The local purchasing power of these payoffs was about twice as much. Thus, it seems fair to say that the stakes were considerable for both students and City Hall employees. Since CH employees (and students) were told ex ante what average earning they could expect, we believe that only subjects that thought the money was worth their troubles signed up forthe experiment.
Sessions lasted from 60 to 100 minutes, with student sessions typically being in the lower half and the CH sessions in the upper half of the interval.
As in Holt & Laury [
We were aware ex ante (based, for example, on evidence reported in Hey & Orme [
The average number of safe choices is 3.5 for students and 3.2 for City Hall employees. This result is not out of line with other evidence (e.g., [
We ran, as a robustness test, ordered logistic regressions with the number of safe choices as the independent variable: Signs are unaffected and the significance levels are roughly the same. The effect of being female, however, is no longer significant (
As robustness tests, we rerun the regressions using, to capture the difference in risk preferences, the variable
Running a Tobit regression, to account for the leftcensoring of Dictator giving, gives the same significance levels for Models 1D1 and 1D3, and slightly higher ones for Models 1D2 and 1D4.
Note that subjects with one safe choice include subjects who lean towards being somewhat riskaverse (0.86 <
An Ftest shows that the dummies for high and low endowment risk are not significantly different (
The success percentages are symmetric by design. The analysis in
For each of the levels of risk aversion we studied, we created a grid of values for the inequality aversion parameter
Funds for the experiment were provided by GAUK (V. Babicky) and GACR (A. Ortmann). S. van Koten is grateful for the JeanMonnet Fellowship and the financial support of the Robert Schuman Center for Advanced Studies at the EUI in Florence. We are grateful to Prague City Hall for helping us contact municipal employees and we thank our colleague Petra Brhlikova for her help in the organization of the experimental sessions. We benefited from comments on early versions of this article by Martin Dufwenberg, Werner Güth, Glenn Harrison, David K. Levine, Ondřej Rydval, Jakub Steiner, and participants of GfeW, FUR, and ESA meetings. We are also grateful for the very constructive contributions of the two (anonymous) referees and the guest editor, Dorothea Herreiner, of GAMES. The usual caveat applies.
The authors declare no conflict of interest.
The patterns of answers on the HoltLaury test and the riskaversion interval.
1^{st}  2^{nd}  3^{rd}  4^{th}  5^{th}  Interval  Risk aversion  Occurrence 



RISKY  RISKY  RISKY  RISKY  RISKY  [1.15; +∞]  Riskloving  9% 
SAFE  RISKY  RISKY  RISKY  RISKY  [0.86; 1.15]  Riskneutral  4% 
SAFE  SAFE  RISKY  RISKY  RISKY  [0.60; 0.86]  Somewhat Riskaverse  4% 
SAFE  SAFE  SAFE  RISKY  RISKY  [0.33; 0.60]  Intermediately Riskaverse  4% 
SAFE  SAFE  SAFE  SAFE  RISKY  [0.04; 0.33]  Very Riskaverse  9% 
SAFE  SAFE  SAFE  SAFE  SAFE  [−∞; 0.04]  Highly Riskaverse  24% 


SAFE  RISKY  RISKY  RISKY  SAFE  [−∞; 1.15]  Undeterminable  2% 
SAFE  RISKY  RISKY  SAFE  RISKY  [0.04; 1.15]  Undeterminable  2% 
SAFE  RISKY  RISKY  SAFE  SAFE  [−∞; 1.15]  Undeterminable  3% 
SAFE  RISKY  SAFE  RISKY  RISKY  [0.33; 1.15]  Undeterminable  2% 
SAFE  RISKY  SAFE  RISKY  SAFE  [−∞; 1.15]  Undeterminable  1% 
SAFE  RISKY  SAFE  SAFE  RISKY  [0.04; 1.15]  Undeterminable  2% 
SAFE  RISKY  SAFE  SAFE  SAFE  [−∞; 1.15]  Undeterminable  3% 
SAFE  SAFE  RISKY  SAFE  RISKY  [0.04; 0.86]  Undeterminable  3% 
SAFE  SAFE  RISKY  SAFE  SAFE  [−∞; 0.86]  Undeterminable  1% 
SAFE  SAFE  SAFE  RISKY  SAFE  [−∞; 0.60]  Undeterminable  2% 
RISKY  RISKY  RISKY  SAFE  RISKY  [0.04; +∞]  Undeterminable  4% 
RISKY  RISKY  SAFE  RISKY  RISKY  [0.33; +∞]  Undeterminable  3% 
RISKY  SAFE  RISKY  SAFE  RISKY  [0.04; +∞]  Undeterminable  4% 
RISKY  SAFE  SAFE  RISKY  RISKY  [0.33 ; +∞]  Undeterminable  1% 


RISKY  RISKY  RISKY  RISKY  SAFE  [−∞; +∞]  Undeterminable  3% 
RISKY  RISKY  RISKY  SAFE  SAFE  [−∞; +∞]  Undeterminable  4% 
RISKY  RISKY  SAFE  SAFE  SAFE  [−∞; +∞]  Undeterminable  1% 
RISKY  SAFE  RISKY  RISKY  SAFE  [−∞; +∞]  Undeterminable  1% 
RISKY  SAFE  RISKY  SAFE  SAFE  [−∞; +∞]  Undeterminable  1% 
RISKY  SAFE  SAFE  RISKY  SAFE  [−∞; +∞]  Undeterminable  1% 
RISKY  SAFE  SAFE  SAFE  SAFE  [−∞; +∞]  Undeterminable  7% 
Sociodemographic characteristics of City Hall employees and students.
As shown in the histograms, the sociodemographic characteristics differ markedly across the two subject pools, though there is also considerable overlap. In
Effect of risk on giving in the Dictator game.
Effect of risk on acceptance threshold in the Ultimatum game.
Taking into account that subjects may have made errors in their choices, we look at the size of the error between the observed choices and the prediction interval.
Prediction error.
Welcome! You are about to participate in an economics experiment. You will be asked to make a series of decisions. Your decisions will have payoff consequences that will also depend on other participants’ decisions. You will be paid privately in cash immediately after the experiment is over. You will get 1 CZK for each 20 ECU (experimental currency units) that you earn during the experiment.
We ask that from now on you refrain from any communication, whether verbal or nonverbal, with other participants. If you have a question, raise your hand and an experimenter will come to assist you.
Throughout the experiment you will, for every single decision (where applicable), be matched randomly with one other participant. The probability that you will be matched with the same participant for more decisions is therefore rather low.
All in all you will be asked to make 17 decisions. You will be informed about the payoff consequences of any of these decisions only after your have made your last decision.
[Any questions?]
During the experiment we will use the following
[Any questions?]
[Any questions?]
[Any questions?]
Example: choice +: 1000 ECU if N>40, 1250 ECU otherwise
or *: 60 ECU if N>40, 2400 ECU otherwise
(note that numbers will vary across decisions)
[Any questions?]
[Please turn your attention now to the computer screen but keep these hard copy instructions readily accessible.]
Thank you for participating in the experiment.
The sequencing of the decisions was the same for all participants, except those in sessions 9 and 10; see footnote 6 for an explanation:
Decision 1: Ultimatum proposal with no risk (pie size 1000)
Decision 2: Risk attitude measurement (n>40,
Decision 3: Dictator with low risk (pie size 900 or 1100)
Decision 4: Trust game sending with high risk (factor 1.2 or 2.8)
Decision 5: Ultimatum proposal with high risk (pie size 300 or 1700)
Decision 6: Risk attitude measurement (n>50,
Decision 7: Ultimatum threshold with high risk (pie size 300 or 1700)
Decision 8: Trust game sending with low risk (factor 1.8 or 2.2)
Decision 9: Ultimatum proposal with low risk (pie size 900 or 1100)
Decision 10: Risk attitude measurement (n>60,
Decision 11: Ultimatum threshold with low risk (pie size 900 or 1100)
Decision 12: Risk attitude measurement (n>70,
Decision 13: Dictator with high risk (pie size 300 or 1700)
Decision 14: Ultimatum threshold with no risk (pie size 1000)
Decision 15: Trust game return with high risk (factor 1.2 or 2.8)
Decision 16: Risk attitude measurement (n>80,
Decision 17: Trust game return with low risk (factor 1.8 or 2.2)
Consider scenario One with the total amount of S=200 ECU. You have the role of participant A and you made a choice of 84.
Question 1: What are your earnings from this scenario?
Question 2: What are the earnings of participant B, who has been selected randomly and assigned to you for this scenario. Please fill out your answer in the space above and confirm.
In the zTree program, participants were given instructions and referred to the printed instructions to guide their understanding of the decision tasks. In
Dictator Scenario
Proposer role
norisk condition
Ultimatum Game
Respondent role
lowrisk condition
Trust Game
highrisk condition
HoltLaury task
Instructions in the zTree program.
Original (Czech)  English translation 

[Questionnaire] 
[Questionnaire] 
[U_Pnone] 
[U_Pnone] 
[HL40] 
[HL40] 
[D] 
[D] 
[T_P] 
[T_P] 
[U_Phigh] 
[U_Phigh] 
[HL50] 
[HL50] 
[U_Rhigh] 
[U_Rhigh] 
[T_Phigh] 
[T_Phigh] 
[U_Plow] 
[U_Plow] 
[HL60] 
[HL60] 
[U_Rlow] 
[U_Rlow] 
[HL70] 
[HL70] 
[Dhigh] 
[Dhigh] 
[U_Rnone] 
[U_Rnone] 
[T_Rhigh] 
[T_Rhigh] 
[HL80] 
[HL80] 
[T_Rlow] 
[T_Rlow] 
[Demographics] 
[Demographics] 
[Results] 
[Results] 