Finance education and social preferences: Experimental evidence

Journal of Behavioral and Experimental Finance 4 (2014) 57–62

Contents lists available at ScienceDirect

Journal of Behavioral and Experimental Finance journal homepage: www.elsevier.com/locate/jbef

Full length article

Finance education and social preferences: Experimental evidence Bryan C. McCannon ∗ Department of Finance, Saint Bonaventure University, P.O. Box 42, Saint Bonaventure, NY 14778, USA

article

info

Article history: Received 20 December 2013 Received in revised form 16 October 2014 Accepted 18 October 2014 Available online 10 November 2014 JEL classification: C9 G1 H4

abstract What impact does a finance education have on the social preferences and the resulting behaviors of individuals? Experiments of a free riding game are conducted where a wealthcreating investment decision is made. The contribution benefits the group, but the incentives are such that an individual, lacking social preferences, would rather make no contribution and free ride off others. It is shown that as one’s education in finance increases, less free riding occurs and more wealth is generated. Thus, education provided in finance promotes pro-social choices that generate wealth even when external incentives are absent. © 2014 Elsevier B.V. All rights reserved.

Keywords: Experiment Finance education Free ride Social preference

1. Introduction Financial transactions can, typically, be characterized by investments being made in the expectation of wealth creation. The returns, though, are uncertain. Oftentimes, the uncertainty of an investment is the risk associated with the behavior of the recipient. A self-interested individual may choose actions that benefit him, but are detrimental to the investor. Social preferences, where a person cares not only about individual gain, but also the well-being of others, can conceptually enhance aggregate wealth. While there are numerous types of market failures that exhibit these general features, an example explored here is what is known as the free rider problem.1 In it, a

∗

Tel.: +1 716 375 2145. E-mail address: [email protected].

1 By the free riding problem I mean the general incentive problem of not contributing to a public good (non-excludable and non-rival in consumption), but rather benefit from it relying on others to use their resources to provide it. This stands in contrast to the specific practice of purchasing shares without paying for them. http://dx.doi.org/10.1016/j.jbef.2014.10.001 2214-6350/© 2014 Elsevier B.V. All rights reserved.

group of individuals or organizations are to work together to achieve a goal. All benefit when the goal is achieved, but do not necessarily have the incentive to expend their own resources to achieve it. For example, a network of investment banks may collectively finance a development project. Oversight of the project, including proper use of the funds, competitive bidding by suppliers, etc., thrives if all members of the network participate. The incentives of each individual organization, though, are to reduce expenses and free ride off of the efforts of the others. As another example, a brokerage firm may, rather than invest the time and resources to conduct independent market analysis, simply rely on non-independent sources of information without providing appropriate research and investigation. Consequently, the CFA Institute includes a diligence and reasonable basis clause in its Standards of Professional Conduct (CFA Institute, 2010). While these are just two examples of free riding in finance, in general, free riding leads to an underprovision of wealth-generating activities and, potentially, market failure. Consequently, it is worthwhile to investigate what influences the preferences of future financial professionals.

58

B.C. McCannon / Journal of Behavioral and Experimental Finance 4 (2014) 57–62

While pro-social preferences can improve upon the free riding problem, the question arises as to what factors lead to higher levels of these improvements? Previous research has indicated that diverse factors such as gender (Sell, 1997), culture, and even brain functioning using fMRI measurements (Krajbich et al., 2009) are associated with differences in free riding behavior. Here, I explore the hypothesis that education can affect these preferences and resulting behaviors. Specifically, I investigate whether an education in finance encourages or discourages wealthcreating investments when the financial environment does not provide adequate institutional incentives to do so. Previous research suggests that education is important, but does not provide a clear picture of the potential effect. For example, research in economics has investigated whether an economics education distorts individual behaviors outside of the classroom. Marwell and Ames (1981) conduct experiments on the free rider game, as is investigated here, and show that economics students free ride more than others. Carter and Irons (1991) provide results of bargaining games showing that economists make lower offers. Frank et al. (1993) report experiments of the prisoner’s dilemma and show that economics students cooperate less. Frank and Schulze (2000) conduct corruption experiments and illustrate that they are more likely to take corrupt bribes. Using empirical data of charitable donations at a university, Frey and Meier (2003) give evidence that economics students contribute less. Research in business education, in general, has shown that it also correlates with lower charitable donations (Meier and Frey, 2004). Taken together and given that finance and economics education share much in common, this body of research casts doubt on the potential effect of a finance education on social preferences.2 The dilemma of research such as this is to disentangle selection effects from learning effects. Does the education itself change behavior or are those who choose to study the field that is predisposed to act differently? Thus, to understand the impact of finance education on social preferences; one must be able to isolate the effect of learning. One attempt has been made to separate the two drivers of outcomes. McCannon and Peterson (forthcoming) explore the selection versus learning issue for a finance education, but in a different institutional environment. The environment considered is an investment game where contract enforcement does not exist. Thus, it studies investing behavior. They show that those who choose to study finance make lower investments and return less, but provide evidence suggesting that a finance education reverses these preferences. The econometric method used interacts major and age to separate selection and learning. A direct link between behavior and coursework taken is not done and, consequently, the marginal impact cannot be assessed. Also, it does not contrast personal gain from benefit to others and, therefore, does not fully explore the influence of ‘‘other-regarding’’ preferences. Thus, the

2 An exception is Yezer et al. (1996) who conduct ‘‘lost letter’’ experiments to measure moral behavior. Evidence suggests that students of economics are more willing to engage in such moral behaviors.

work presented here clarifies the issue by studying a social setting where personal and other’s gain is in conflict and directly investigates the marginal impact of finance course. The objective here, then, is to investigate whether enhanced coursework in finance discourages free riding, as may be suggested by the work McCannon and Peterson (forthcoming), or does it promote selfish gain at the expense of others, as shown to be the case in economics education by Marwell and Ames (1981). Previous research in finance tends to focus on the related issue of financial literacy. For example, Wang (2009) and Sjöberg and Engelberg (2009) consider the relationship between financial literacy, education, and risk taking. Peng et al. (2007) presents survey evidence that personal investment education correlates with investment knowledge and savings behavior. Bernheim and Garrett (2003) study information on employer-based financial education and find that these programs improve savings. Likewise, Chira et al. (2012) find that educational attainment correlates with student loan choices. Hence, the work presented here contributes to the understanding of the link between financial education and outcomes. Experimental research analyzing the free rider game has a long history. See Ledyard (1995), Zellmer (2003), and Chaudhuri (2011) for comprehensive literature reviews. In the free rider game subjects play in groups. Each has an endowment and chooses how much to invest in a common fund, keeping the residual as personal gain. Contributions to the common fund grow and are shared equally amongst the group. Growth of the fund is such that while aggregate wealth expands as more is contributed, the division is such that an individual receives back less from every dollar invested than by retaining it. Thus, a guaranteed negative return arises. Here each subject is endowed with five ‘‘experimental dollars’’. The common fund triples and is evenly shared amongst the four members of the group. Hence, each dollar contributed returns only seventy-five cents. Hence, absent social preferences, the incentives are such that it is optimal to make no contribution and free ride on the donations of others. Dominating the research on the free riding game is an investigation of how institutional features, such as the endowment, group size, information, or the growth rate, affect contributions. The tactic here is to use the free rider game as an instrument to assess how external factors affect preferences and, thus, behaviors. The experimental methods are presented in Section 2. Section 3 provides the econometric results, while Section 4 concludes. 2. Experimental design To address this question experiments were conducted with undergraduate students at a small, private university in upstate New York. Subjects were recruited from general education classes. Additionally, individuals were recruited from classes within the business school.3 An online reservation manager was used to schedule the

3 Economics is within the school of business and the faculty are joined with those in finance into one department.

B.C. McCannon / Journal of Behavioral and Experimental Finance 4 (2014) 57–62

sessions. The recruitment strategy targeted students in classes taken by underclassmen along with classes taken by upperclassmen in both the general education courses and those within the business school. This provided the opportunity to study choices of those who had taken a significant amount of coursework in finance and business, those who had selected fields where this will be an important part of their education (but the coursework has not been completed), along with students who did not choose and will not have much exposure to finance. Five experimental sessions were conducted in November 2012 and another four were conducted in April 2013. The number of participants in each session ranged from twelve to seventeen. There were a total of 147 experimental subjects. Each session lasted approximately one hour. Subjects completed a background information questionnaire and engaged in the experiment. The experimental subjects played the Free Rider Game.4 In this game, the subjects were randomly selected into groups of four. Each person in the group is endowed with five ‘‘experimental dollars’’ and chose how much to contribute to a ‘‘common pot’’. The subjects were informed that the pot tripled and then was evenly shared amongst the four group members. In the first two sessions the subjects played five rounds of the game. In three sessions they played six rounds, while in the final four sessions three rounds were conducted. In each round new random groupings were made. Subjects were informed of their earnings from one round before making their selections in the next. It was explained that the more experimental dollars they were able to earn in the rounds, the more real dollars they obtained. Specifically, the total number of experimental dollars earned in all rounds of play would be aggregated. The total experimental dollars earned, then, would be converted into ‘‘real’’ dollars. The subjects were instructed that the amount they earned would be determined not only by the choices they made, but also were going to be affected by the choices of others. They were also informed that in previous, typical sessions (based on a pilot study) subjects earned on average over $20, but the amounts ranged between $10 and $40. The payouts provided a $10 ‘‘show-up’’ payment. A scale was adopted where if the maximum payoff was achieved in each round $50 could be earned, down to a minimum of just the show-up fee. Payments were rounded to five dollar increments (or rather, a step-scale was developed). The average monetary payment received by a subject in the experiment was $22.50. Along with selecting their contribution, subjects were asked to provide their expectation regarding others’ choices. Specifically, they were asked, ‘‘other than your contribution, what do you think will be the total contribution of the other three members of the group?’’ In sessions with a number of subjects not divisible by four, responses were selected at random to complete a four-person group to score the remainder’s outcome. Thus, for example, in a session with fifteen subjects, twelve are

4 While some in the field refer to it as the Free Rider Game, others refer to it as the Public Goods Game.

59

Table 1 Summary statistics. Outcomes:

Mean

Median

Min

Max

St. dev.

Contribution Guess Outcome FIN

2.504 8.400 10.039 2.320

3 8.5 10 1

0 0 4.75 0

5 16 16 12

1.77 3.50 2.39 2.72

Controls:

Mean

New York Pennsylvania Ohio New Jersey Other state US citizen Vote

0.725 0.059 0.019 0.026 0.089 0.923 0.538

Mean Freshman Sophomore Junior Senior MBA student Male

0.355 0.115 0.162 0.364 0.004 0.668

put into three groupings. Hence, this leaves three remainders. One of the twelve would be selected at random to provide the fourth contribution to the remaining three subjects.5 PowerPoint slides presenting the rules of the game, along with printed instruction were provided. Subjects were given the opportunity to ask clarifying questions between play began. With regard to the background information solicited, common demographic control variables were collected. They include gender, nationality, state of residence, major, and year in school. Furthermore, subjects were asked how many economics and finances courses they had taken. At the university where the experiments took place, economics is not a separate program, but only provide service courses within the finance department.6 Finally, given that the experiments occurred in November of 2012 and April 2013, each subject was asked whether or not s/he had voted in the recent election. Table 1 provides descriptive statistics of the sample of the resulting 530 observations. First, the results from the experiments are presented. The variable Contribution is the amount a subject contributed to the common fund (out of his/her endowment of five). One could earn as low as 3.25 if he gives all five and the other three in the group contribute zero, and could earn as much as 16.25 if each of the other three give all five and he provides nothing. Furthermore, subjects were asked how much they expected the other three members of the group will contribute in total. Their response constitutes the variable Guess. Thus, one who expects complete free riding of the others in the group has Guess = 0, while a belief of full participation by the other three results in Guess = 15. Additionally, when Guess is more than three times Contribution, then a subject is giving less than what he or she expects of others. The variable Outcome is the total amount earned in a round by a subject, which is the sum of the amount retained and the return from the tripled fund. The primary variable of interest, FIN, records the number of finance courses taken by the subject. Background controls include dummy variables for state of

5 Variation in group size is an artifact of students signing up for a session, but not showing up for it! 6 Thus, classes such as Money and Banking and Econometrics are listed as finance course codes.

60

B.C. McCannon / Journal of Behavioral and Experimental Finance 4 (2014) 57–62 Table 2 Relationship between finance education and contributions. Description

Advanced fin. educ. Significant fin. educ. Some fin. educ. Little/no fin. educ.

Definition

FIN FIN FIN FIN

≥9 = 7, 8 = 4, 5, 6 ≤3

Contribution

3.56 2.68 1.80 2.60

residence, citizenship, year in school, gender, and whether the subject voted. The typical subject contributed almost exactly one-half of his/her endowment to the common fund. Interestingly, subjects expected more contributions from others than they gave themselves, since Guess exceeds 3 × Contribution.7 Thus, subjects free rode more than they expected out of others. The typical student has taken 2.3 finance courses. Since all business majors are required to take an introductory microeconomics class, along with an introductory macroeconomics course, the bulk of the subjects have no finance education. There are subjects who have a substantial amount of finance education (up to twelve courses) and given that the median number of courses is one, but the mean is over two, the pool contains a skewed, right-tail to the distribution. Thus, I am able to exploit the differing levels of finance education taken and its impact on free riding. The subject pool is dominated by US citizens from the state of New York. There are more males than females, which is a result of oversampling from business classes. The sample contains both underclassmen and upperclassmen, again allowing for a differentiation in educational backgrounds to be explored. First, in an exploration of the data set, the sample of observations is portioned into groupings based on the number of finance courses taken. Table 2 presents the mean value of Contribution and Guess for the subsamples of those with advanced, significant, and some finance education. Additionally, the percentage of observations within each sample in which contributions were full and zero are recorded. The evidence suggests that there is a positive relationship between wealth-creating contributions and finance education. There is a 98% increase in the amount given by subjects with advanced financial educations, as compared to those with just some training. Relatedly, those who have completed a lot of finance coursework are much more likely to contribute all five experimental dollars and less likely to fully free ride. Interestingly, there is also a relationship between finance education and expectations on others. While the average value of Guess in each subsamples are similar, since the actual contributions made are differing, then it seems as though subjects’ expectations of others relative to their own behavior is substantial. Those with less finance education not only free ride

7 The maximum value of Guess observed is 16, which must be a misunderstanding since the contributions cannot exceed 15. This value is observed only once in the data. Excluding this observation from the analysis has no qualitative effects on the main results.

Guess

8.55 8.80 8.80 8.29

Contribution %=0

%=5

N

16.7% 22.0% 37.7% 16.9%

50.0% 24.0% 15.9% 17.8%

18 50 69 393

more, but are expecting that their choices are far less than their fellow subjects. Thus, there is an optimism of the behavior of others. Those with advanced finance education actually behave as if they expect others to free ride more than they plan to do. In other words, they are contributing at a level above what they expect others to do. Again, this is evidence of more developed social preferences. Finally, the subset with little to no finance education behaves differently. They make larger contributions and, relatedly, are less willing to not make a contribution. Thus, the selection effect found in McCannon and Peterson (forthcoming) findings is evidenced here. Further econometric analysis is needed to establish the statistical significance of these results. 3. Econometric results If the results of McCannon and Peterson (forthcoming) hold in general, we would expect those who are not intending to study finance (and consequently have no finance education) to contribute substantially due to a selection effect. Finance students beginning their education should have low contributions, but as education increases so too should giving. Hence, a nonlinear, U-shaped relationship between contributions and education is expected. To assess the impact of a finance education on free riding, regression analysis is conducted where contributions to the common fund is used as the dependent variable. A constant term and all background control variables where included in each specification (but not reported). The number of finance courses taken, along with subject’s expectations, is included as well. The results are presented in Table 3. Robust standard errors are calculated. Both heteroskedastic-robust standard errors along with standard errors clustered by round of play are reported in Table 2.8 Column I considers only a linear relationship and finds a positive, but insignificant, effect of education on behavior.9 Column II, though, illustrates that there is a strong nonlinear relationship between the number of finance courses taken and free riding. The final row of Table 2 provides the value of FIN where the U-shaped relationship

8 Similarly, standard errors clustered by session were calculated. The statistical significance of the primary variables of interest is unchanged and, hence, is not reported. 9 Evidence of a selection effect driving the U-shape exists. If one considers only the subsample of subjects who are finance majors, the linear relationship between finance education and public goods contributions is statistically significant (coefficient = 0.155; p-value < 0.01) with state, gender, round, session controls and Guess and Vote as additional explanatory variables.

B.C. McCannon / Journal of Behavioral and Experimental Finance 4 (2014) 57–62

61

Table 3 Econometric results (dependent variable = Contribution).

FIN

I

II

III

IV

V

0.0061 (0.0358)

−0.2969

−0.3610

−0.3405

−0.3543

FIN 2

(0.0907)***

(0.0816)*** [0.1153]***

(0.0837)*** [0.1094]***

(0.0870)***

0.0356 (0.0096)***

0.0410 (0.0087)***

0.0314 (0.0097)*** [0.0128]**

0.0316 (0.0096)*** [0.0126]**

0.2147 (0.0200)***

0.2195 (0.0208)*** [0.0063]***

0.2204 (0.0210)*** [0.0063]***

GUESS

FINANCE

0.1711 (0.2558) [0.2332]

Controls: Background? Round? Session?

YES NO NO

YES NO NO

YES NO NO

YES YES YES

YES YES YES

adj R2 AIC

0.0161 2107.5

0.0388 2096.2

0.2113 1985.1

0.2320 1983.5

0.2312 1985.0

4.17

4.40

5.42

5.61

Bottom

Heteroskedastic-robust standard errors are reported in parentheses. Standard errors clustered by round of play are reported in brackets. Background control variables include all those reported in Table 1. ** 5% level of significance. *** 1% level of significance.

hits a minimum, or rather, its ‘‘bottom’’. Contributions are minimized at a value just over four classes. Recall, all finance students must first take two economics courses and all business students must take, in addition, a class in corporate finance. Thus, the reduction in contributions with fewer finance coursework completed can be explained by non-finance, business majors contributing less, i.e. a selection effect. The upswing in the relationship, then, is the effect of a finance education. As students, who choose to study finance, take their coursework, they contribute more and free ride less. Thus, a finance education promotes wealth-creating investments and discourages free riding. Column III includes a control for the subject’s expectations of others’ contributions. Its positive relationship and statistical significance indicate further pro-social behavior. When a subject expects others to free ride, the he himself is more likely to free ride, but if his expectations are that others are contributing, then he contributes more as well. The impact of a finance education, though, remains unchanged. Column IV includes round and session fixed effects. Sessions naturally differ in composition and vary in the number of rounds played. Round fixed effects control for the possibility of adaptation in play (i.e. learning). The relationship and statistical significance is unaffected by the inclusion of these controls. Hence, composition of the experimental session and adaptation in play do not explain the results. Column IV includes a dummy variable for whether the subject is a finance major, Finance. Including field of study does not explain behavior. Thus, it is the coursework and not the selection into the field of study in finance that is driving behavior. Additionally, robustness checks are conducted, but not presented. Since Contribution is a discrete variable, ordered

probit and count data models can be alternatively estimated. Doing so provides the same results and, therefore, is not presented. Similarly, one may be concerned that variation in the number of rounds may be affecting the results. Truncating the data set to only three rounds of play, which occur in all sessions, again provides the same main result. Advanced finance education leads to more wealthcreating contributions and less free riding. Finally, binary probit models can be estimated where the dependent variables are dummy variables equal to one if no contribution is made and a full contribution is made. Again, similar results arise. Thus, the results presented in Table 2 are robust. 4. Conclusion Does an education in finance affect social preference? Specifically, does it promote wealth-improving behaviors even when there is no external enforcement mechanism? Experiments in the free rider game is conducted and analyzed to address this issue. It is shown that free riding decreases as the number of finance courses completed increases. Thus, a finance education promotes wealth creation. By investigating the marginal impact of finance coursework on behavior, I am able to directly identify the marginal impact of additional coursework. Thus, the learning effect is isolated from the selection effect, which has proven to be an obstacle in previous research. One limitation of the study is that without controlling students’ entire menu of coursework completed, along with other life developments, such as for example internships, it is still difficult to guarantee that it is indeed the work done in the classrooms that is changing individuals behavior outside of the classroom. The results are able to conclude, though,

62

B.C. McCannon / Journal of Behavioral and Experimental Finance 4 (2014) 57–62

that selection effects, while also important, are not driving the results. The work, given the design and methods used, is unable to explore the mental mechanism that drives changes in individual preferences. Potentially either neuroscientific research or, at the very least, extensive survey methods need to be employed to identify exactly how finance education changes individual’s preferences. What the results show, though, is that the education is associated with changes in outside-the-classroom behavior. The results are consistent with changes in behavior, but the experimental methods cannot directly measure preferences. A remaining issue is a full explanation of the nonlinear relationship. Evidence is presented that selection effects drive the initial downward relationship, while the finance education reverses the trend. Further investigation is needed to fully clarify the impacts. Future research should investigate other social preferences and the impact of finance education and training on them. Only one specific behavior is investigated here, free riding. While an important issue in banking and trading, other pro-social tendencies, such as altruism, reciprocity, hold-up problem, trust, and corruption also potentially merit investigation. The noteworthy finding here is that a finance education encourages otherregarding behaviors, which facilitate the creation of wealth. This stands in contrast to the related literature in behavioral economics. Additionally, the results presented here suggest that it is, in fact, the learning from the education that drives the outcome, rather than the selection effects as typically arises in research on the effects of an economics education. Thus, future investigations may want to identify the topics/material which encourages the good behaviors observed here. Acknowledgments Author would like to thank Joe Coate, Rich MacDonald, Jim Mahar, Jeff Peterson, and Mark Wilson for helpful discussions about the project. Author also thank Dave DiMattio for helping in recruiting and Kim McCannon for being a lovely assistant during the experiments. Financial support was provided by the Koch Foundation and author greatly appreciate their contribution to this public good.

Appendix A. Supplementary data Supplementary material related to this article can be found online at http://dx.doi.org/10.1016/j.jbef.2014.10. 001. References Bernheim, B. Douglas, Garrett, Daniel M., 2003. The effects of financial education in the workplace: evidence from a survey of households. J. Public Econ. 87 (7–8), 1487–1519. Carter, John R., Irons, Michael D., 1991. Are economists different, and if so, why? J. Econ. Perspect. 5 (2), 171–177. CFA Institute, 2010. Code of ethics and standards of professional conduct. Codes Stand. Position Pap. 14 (1), 1. Chaudhuri, Ananish, 2011. Sustaining cooperation in laboratory public goods experiments: a selective survey of the literature. Exp. Econ. 14 (1), 47–83. Chira, Inga, Chiang, Eric, Houmes, Robert, 2012. Can undergraduate students make optimal decisions about student loans? J. Econ. Finance Educ. 11 (1), 1–14. Frank, Robert H., Gilovich, Thomas, Regan, Dennis T., 1993. Does studying economics inhibit cooperation? J. Econ. Perspect. 7 (2), 159–171. Frank, Björn, Schulze, Günther G., 2000. Does economics make citizens corrupt? J. Econ. Behav. Organ. 43 (1), 101–113. Frey, Bruno S., Meier, Stephan, 2003. Are political economists selfish and indoctrinated? Evidence from a natural experiment. Econ. Inquiry 41 (3), 448–462. Krajbich, Ian, Camerer, Colin, Ledyard, John, Rangel, Antonio, 2009. Using neural measures of economic value to solve public goods free-rider problem. Science 326, 596–599. Ledyard, John O., 1995. Public goods: a survey of experimental research. In: Roth, Alvin E., Kagel, John H. (Eds.), Handbook of Experimental Economics. Princeton University Press, pp. 111–181. Marwell, Gerald, Ames, Ruth E., 1981. Economists free ride, does anybody else? J. Public Econ. 15 (3), 295–310. McCannon, Bryan C., Peterson, Jeffrey, 2014. Born for finance? Experimental evidence of the impact of finance education. J. Behav. Finance (forthcoming). Meier, Stephan, Frey, Bruno S., 2004. Do business students make good citizens? Int. J. Econ. Bus. 11 (2), 141–163. Peng, Tzu-Chin Martina, Bartholomae, Suzanne, Fox, Jonathan J., Cravener, Garrett, 2007. The impact of personal finance education delivered in high school and college courses. J. Family Econ. Issues 28 (2), 265–284. Sell, Jane, 1997. Gender, strategies, and contributions to public goods. Soc. Psychol. Q. 60 (3), 252–265. Sjöberg, Lennart, Engelberg, Elisabeth, 2009. Attitudes to economic risk taking, sensation seeking and values of business students majoring in finance. J. Behav. Finance 10 (1), 33–43. Wang, Alex, 2009. Interplay of investors financial knowledge and risk taking. J. Behav. Finance 10 (4), 204–213. Yezer, Anthony M., Goldfarb, Robert S., Poppen, Paul J., 1996. Does studying economics discourage cooperation? Watch what we do, now what we say of how we play. J. Econ. Perspect. 10 (1), 177–186. Zellmer, Jennifer, 2003. Linear public goods experiments: a metaanalysis. Exp. Econ. 6 (3), 299–310.