Imagine being a nice guy: A note on hypothetical vs. incentivized social preferences

Abstract

We conducted an experimental study on social preferences using dictator games similar to Fehr et al. (2008). Our results show that social preferences differ between subjects who receive low-stakes monetary rewards for their decisions and subjects who consider hypothetical stakes. Our findings indicate that, apart from incentives, gender plays an important role for the categorization of different social preferences.

1 Introduction

The fields of economics and other social sciences, such as psychology, often differ in their views on the use of monetary incentives in experiments. Economists usually argue that financial rewards create a more realistic environment within the lab (Rosenboim & Shavit, 2012), causing subjects to consider their decisions more carefully (Carpenter et al., 2005). Psychologists, on the other hand, tend to believe that experimental subjects are generally intrinsically motivated and need no financial reward for decision-making (Camerer & Hogarth, 1999). Of course, this distinction is somehow artificial and far from building a clear-cut dividing line between those professions, but the role of incentives in experiments remains an important field, given that previous research provides empirical evidence that different incentive mechanisms usually, but not always, induce different behavioral responses from experimental subjects.¹

In this study, we examined the effects of the presence or absence of monetary incentives on other-regarding behavior, that is, social preferences. Social preferences, such as egalitarianism or generosity, are argued to be highly relevant to decision-making in a variety of economic and social contexts such as philanthropy and charitable giving, organ donations, or family transfers (see Kolm & Ythier, 2006, for a comprehensive overview). However, exactly how monetary rewards affect those social preferences remains unclear. Compared to a hypothetical setting without financial rewards, we found that even low-stakes monetary incentives 1) decrease (strongly) egalitarian choices, 2) increase spiteful choices, but 3) also increase generous choices.

A common way to elicit social preferences is to use the dictator game (DG) in which a sender (dictator) decides how to allocate a sum of money to himself and a receiver.² There are only few studies on the effect of introducing financial incentive mechanisms in DGs and these report mixed results: Sefton (1992) found significantly more self-interested offers in a DG with a low-stakes financial reward compared with Forsythe et al.’s (1994) results for an equally designed hypothetical setting. In Dana et al. (2007), receivers in a binary DG were instructed to choose hypothetically between an equal and an unequal distribution, while the choices of dictators were incentivized: Compared with the incentivized treatment, a larger share of subjects picked the egalitarian option in the hypothetical treatment. Amir et al. (2012) reported that 1$ incentives in an online DG significantly decreased average offers compared to a no-stakes DG. On the other hand, Ben-Ner et al. (2008) showed that dictators facing decisions involving real money were slightly more generous compared with subjects considering hypothetical money, but this difference was not significant in statistical and economic terms, even after controlling for subject-specific characteristics.

Various models to describe different types of social preferences. In economics, the most popular ones are Fehr and Schmidt (1999), who incorporate envy and altruism in the utility function, the theory of equity, reciprocity, and competition (ERC) by Bolten and Ockenfels (2000), in which deviations from egalitarian distribution result in disutility, and the Quasi-maximin model by Charness and Rabin (2002), which takes into account the lowest payoff of a distribution. However, it is not easy to distinguish these different models in standard DGs (Daruvala, 2010, e.g., uses evaluations of money distributions in groups of 11).

In our study, we used DGs similar to those of Fehr et al. (2008) which provided us with a simple way to categorize different types of social preferences. Subjects were presented with three sets of dichotomous choices to allocate money to themselves and another person. In the prosocial game, the dictator chose between two different allocations, (0.5,0.5) and (0.5,0). The dictator could increase his partner’s payoff at no cost to achieve an egalitarian distribution. In the envy game, the dictator faced a choice between (0.5,0.5) and (0.5,1). An increase in the partner’s payoff was possible only by deviating from the egalitarian distribution. In the sharing game, the feasible allocations were (0.5,0.5) and (1,0). Choosing the egalitarian option in the prosocial or the envy game indicates inequality aversion: In the former case, the decision maker does not want the other person to earn less than himself, and in the latter case, he does not want his partner to earn more. The sharing game can be regarded as a strong form of the prosocial game—the fundamental difference is that taking the egalitarian option in the sharing game is costly for the dictator. A particular advantage of choosing the three dichotomous choice sets instead of a single (continuous) DG is that subjects’ choices can be directly translated to five (weak and strong) types of social preferences, namely spite, egalitarianism, and generosity (Table 1).

Table 1: Subcategories of social preferences.

Source: Own compilation based on Fehr et al. (2008).

These types of social preferences differ with respect to their underlying motives. In particular, Table 1 shows that spiteful subjects always choose options making their counterparts worse off, even if they could increase their partners’ payoff at no cost in the prosocial game. Weakly egalitarian subjects choose the egalitarian option whenever they are not disadvantaged; they are not spiteful but they also do not allow their counterparts to earn more than them. Strongly egalitarian subjects always choose the egalitarian option—no matter if it “hurts” others or them. Weakly generous subjects grant their counterparts a higher payoff if this is at no cost for them. Finally, strongly generous subjects always choose the best option for their counterparts in the three DGs, even if this means a costly transfer in the sharing gameand a deviation from the (0.5,0.5) distribution in the envy game,with the latter choice representing the fundamental difference to the egalitarian category.

The dictator games of Fehr et al. (2008) have been applied in a variety of settings (e.g., Svensson, 2009; Bauer et al., 2011a; Bauer et al., 2011b; Fehr et al., 2011; Zaleskiewicz and Helka, 2011; House et al., 2012); experimental subjects were usually children or adolescents, incentives were usually sweets. Fehr et al. (2011) provided small monetary incentives to a group of adolescents. In our experiment, we used an adult subject pool that was randomly assigned to an incentivized treatment and a hypothetical treatment. To our knowledge, we are the first to study hypothetical vs. incentivized decisions in the Fehr et al. (2008) DGs in an adult subject pool.

Having a look on the DG results mentioned above, the introduction of financial incentives is likely to lead to different motives. The results of Dana et al. (2007) speak for more hypothetical egalitarism, whereas those of Ben-Ner et al. (2008) indicate slightly more incentivized generosity. Following Amir et al. (2012), we expect incentivized subjects to be less nice than subjects of our hypothetical treatment, that is, we hypothesize them to be more spiteful and less strongly egalitarian or generous.

2 Experiment

The social preferences experiment involved three allocation decisions similar to the Fehr et al. (2008) DGs. Since the experiment lasted less than five minutes, it was preceded by two other unrelated experiments. ³ Combining short experiments this way is a common practice (see, e.g., Fischbacher and Föllmi-Heusi, 2013).

In total, six sessions were carried out at the experimental lab of the University of Hamburg in November 2012 with 150 students participating. Subjects were invited via the recruitment software hroot (Bock et al., 2012). They received a 5.00 EUR show up fee that was announced in the invitation.⁴ 90% came from Germany, 50% were male, and the average age was 25.14 ( SD = 4.73).

Subjects were randomly divided into two subgroups: 80 made decisions involving real money ( incentivized treatment), 70 made hypothetical choices ( hypothetical treatment) in the DGs. ⁵ The assignment to the experimental conditions was independent of the treatments in the preceding experiments. The only differences across treatments was that respondents in the hypothetical treatment were instructed to imagine they could choose between two allocations within the DGs, whereas subjects in the incentivized treatment were informed that one of their decisions and the decision of another experimental subject would be paid out to them (see Appendix A).

The experiment was computerized using z-Tree (Fischbacher, 2008). Possible allocations in the three DGs were presented to the subjects as outlined in section one, with payoffs of 0.00 EUR, 0.50 EUR, and 1.00 EUR (see Appendix A).⁶ To every dictator’s decision one receiver was randomly matched. In order to avoid reciprocity, the matching procedure ensured that the roles of senders and receivers remained independent and anonymous: a dictator A sending a transfer to a receiver B received a transfer from another, unrelated dictator C. Subjects were aware of this type of matching. The exercise was repeated for the three DGs; following Fehr et al. (2008), we kept the ordering of the DGs identical across subjects. Subjects received no feedback in the one-shot DGs and transactions were kept anonymous in order to rule out that social preferences resulted from strategic behavior or that they were affected by selfish motives (Fehr et al., 2008).

In the incentivized treatment, one DG was randomly chosen and we paid out the money the subjects allocated to themselves and received from another subject in that game. Average earnings were 1.15 EUR ( SD = 0.39 EUR).

3 Results

Table 2 illustrates the number of subjects by treatment that fall into the five categories of social preferences based on the aggregated decisions in the DGs. The categorization of subjects is significantly different across treatments (Pearson’s χ²(4) = 17.9, p < 0.01, two-sided).

Table 2: Results of the behavioural subcategories (by treatments)

Notes: One subject could not be categorised based on her/his choices in the DGs.

Taking a closer look at the individual types of social preferences, the differences in the egalitarian categories are most obvious: 34 subjects (48.6%) in the hypothetical treatment could be categorized as egalitarian (weakly or strongly), but only 22 (27.5%) behave egalitarian in the incentivized treatment. The difference of 12 subjects (21.1 percentage points) is highly significant (Fisher’s exact test, p = 0.01, two-sided). Choices in the sharing game revealed that 15 subjects (21.4%) in the hypothetical treatment picked the egalitarian distribution (0.5,0.5) in all DGs ( strongly egalitarian), whereas the number dropped to 2 (2.5%) in the incentivized treatment (Fisher’s exact test, p < 0.01, two-sided). Unlike in the envyand prosocial game, choosing (0.5,0.5) in the sharing gameinvolved a costly transfer and thus represents a strong form of other-regarding behavior in terms of inequality aversion and altruism as defined by evolutionary biology (Fehr et al., 2008). Taken together, considering real money seriously influenced the equality motive, even for relatively low stakes. Similar results for DGs with low stakes were also reported by Sefton (1992) and Amir et al. (2012).

In contrast, we found that incentivized subjects were slightly more generous. Pooling generous and strongly generous subjects resulted in a total number of 50 subjects (62%) in the incentivized, and 34 (49%) in the hypothetical treatment; yet this difference is not significant (Fisher’s exact test, p = 0.10, two-sided). Also the strongly generous choices do not differ significantly by treatment. This pattern is in line with Ben-Ner et al. (2008), who found non-significantly larger generosity for real as compared with hypothetical choices in DGs.

Finally, examining the frequency of spiteful choices, we found a significant difference between the two treatments (Fisher’s exact test, p = 0.04, two-sided); 9 (10%) of the incentivized subjects chose the option that minimized their anonymous partner’s payoffs in all DGs. In contrast, only one subject out of 70 (1%) made this kind of decision when subjects only imagined being a dictator in the experimental setting.

With respect to gender, the majority of 51 men (69%) in our sample could be categorized as generous, whereas the largest fraction of 38 female subjects fell into the egalitarian category, with even 14 of them (19%) in the strongly egalitarian class (see Appendix B for an overview and Croson & Gneezy, 2009, for comparable results). In this respect, choices of women ( men) are on average comparable to choices in our hypothetical ( incentivized) treatment. According to two-sided Fisher exact tests, the treatment effect (incentivized vs. hypothetical) in the weakly generous as well as the strongly egalitarian category is significant for women ( p= 0.096 and p< 0.01, respectively) but not for men. These findings suggest that women respond more to the introduction of monetary incentives. Similar to our result, Eckel and Grossman (1996) found in their “punishment game” that women are more responsive to changes in the incentive structure than men.

4 Discussion and conclusion

In this paper, we presented the results of an experiment on social preferences elicited by using DGs similar to Fehr et al. (2008). We showed that incentivizing subjects affects their choices in DGs and the categorization into different social preference classes.

(Almost) nobody wanted to be spiteful (only 1 of our subjects) when choices had no monetary consequences. Furthermore, strongly egalitarian choices that indicate an aversion to disadvantageous andadvantageous distributions almost disappeared when people were incentivized (3%). The majority of incentivized subjects (62%) displayed generous choices. In the hypothetical treatment, egalitarian choices were slightly more frequent than generous choices (48.6 % vs. 48%). The increase of spiteful and generous choices and the decrease of strongly egalitarian choices in the face of low monetary consequences indicate that the elicitation of subjects’ true preferences might be complicated when using a hypothetical treatment. Ultimately, generous as well as spiteful choices only have material consequences if monetary rewards are at stake.

Within the context of the existing literature, our results in the incentivized treatment are very much in accordance with the categorization of 16- and 17-year old adolescents presented in Fehr et al. (2011). In their experiment, Fehr et al. (2011) used a comparably higher stake of 6 EUR for this subsample and found that the majority could be categorized as generous (60%), while only 26% fell into the egalitarian category and 14% in the spiteful category. The results presented in Fehr et al. (2011) also confirm the gender effect we observed in our data: while female subjects between 16- and 17 years were more frequently categorized as egalitarian, male subjects turn out to be in the generous category more often.

Our results imply that experimental findings for social preferences depend crucially on the underlying earning mechanism; even low stakes are able to systematically change the categorization into different classes of social preferences. The effects of stakes on decision-making have also been reported in a number of other, partly comparable, experimental environments (see Camerer & Hogarth, 1999, for an overview). It would be interesting to see if the categorization of the incentivized treatment changes with the integration of large stakes (Carpenter et al. 2005); however, the results provided by Fehr et al. (2011) suggest that the size of stakes (in absolute and relative terms) do not necessarily lead to systematically different categories.

According to our results, hypothetical and incentivized decisions reflect fundamentally different situations. Experimenters have to evaluate in which cases intrinsic motivation in hypothetical scenarios vs. motivation caused by monetary rewards are better able to predict real behavior. We do not want to judge whether relying on subjects’ intrinsic motivation or incentivizing subjects leads to more external valid experimental results. Monetary incentives inside (and outside) the experimental lab might crowd out intrinsic motivation (Frey & Oberholzer-Gee, 1997), or they might reveal the true face of a hypothetically nice guy.