Moral distance in dictator games

Abstract

We perform an experimental investigation using a dictator game in which individuals must make a moral decision — to give or not to give an amount of money to poor people in the Third World. A questionnaire in which the subjects are asked about the reasons for their decision shows that, at least in this case, moral motivations carry a heavy weight in the decision: the majority of dictators give the money for reasons of a consequentialist nature. Based on the results presented here and of other analogous experiments, we conclude that dicator behavior can be understood in terms of moral distance rather than social distance and that it systematically deviates from the egoism assumption in economic models and game theory.

1 Introduction

In standard dictator game (DG) experiments, one of the players, the dictator, is given a fixed amount of money and a set of instructions. The instructions explain that the money has been assigned to two players — the dictator and the recipient — but that only the dictator is entitled to decide how much money to keep and how much money to give to the recipient. In this simple situation, there are at least three sources of information relevant to the decision: i) information related to the dictator herself, i.e., whether or not the decision is observed by others; ii) information the dictator receives about the recipient; and iii) information derived from the game framing and the language used in the instructions.

Previous papers have shown that the outcome of a DG changes considerably depending on the type of information provided by the experimenters. For instance, when the dictator makes the decision under conditions of absolute privacy and anonymity and has no information about the recipient, nearly no one donates anything. Donations tend to be very low, around 10% of the pie on average (Hoffman et al., 1994; Hoffman et al., 1996). However, when the dictator receives reliable information regarding the recipient, even while keeping anonymity constant, donations increase. When no information is available about the recipient, the dictator may have doubts as to the recipient’s existence and therefore have no reason to share the money. In contrast, when dictators were shown pictures of the recipients in a particular experiment, up to 25% of the subjects gave as much as half of the total amount, although 58% of them kept all the money for themselves (Burnham, 2003). In another experiment, the dictators were told that their donation would be given to the Red Cross (Eckel & Grossman, 1996). Thirty-one percent of the dictators in this experiment gave part of the money (17% gave half of the amount), while 10% gave the full amount. Offers also increased when the dictators were told the recipient’s surname (Charness & Gneezy, 2003), when they received proof of the actual existence of recipients, or when they were told that the recipient was a fellow classmate (Frohlich & Oppenheimer, 2001). These experiments have shown that if conditions of anonymity are relaxed, e.g., dictator’s behavior or relevant information about the recipients are revealed, donations will be higher, although the full amount of money will never be donated — with the exception of the Red Cross experiment.

The language used in the instructions is yet another factor that may determine the outcome of the experiment. For example, the fact that subjects are told that they are participating in an experiment may make them lose interest in their partner, or may cause subjects to feel that they are playing a game — that it is a competition and they must therefore win, that is, keep all or part of the money for themselves (Frohlich et al., 2004). The nature of the payments, the presence or absence of the experimenters, the possibility of taking part in future experiments — and thus winning more money — may also determine the outcome (Bolton et al., 1998).

The aim of these experiments is to measure the prosocial conduct or other-regarding preferences revealed by the subjects, thus challenging the assumption of self-interested behavior. The supposition that individuals will attempt to maximize their own benefits, a standard assumption in economics, could lead to the conclusion that no one will give anything in the DG since giving nothing is the behavior that maximizes the dictator material’s payoff. However, as we have seen, subjects often deviate from this prediction in the laboratory setting. In light of these experimental results, several hypotheses have been developed to explain these deviations.

Hoffman et al. (1996) show that as anonymity is relaxed, donations increase. In the opinion of these authors, this is due to the fact that anonymity generates “social distance,” which they define as “the degree of reciprocity that subjects believe exists within a social interaction” (Hoffman et al., 1996, p. 654). The social isolation or lack of sense of community produced by anonymity leads to the predominance of self-interested behavior. Nonetheless, the hypothesis of social distance has been widely criticized. The concept of reciprocity, for example, is employed in a very broad sense since in reality the DG does not pose so much a problem of reciprocity — in that the recipient can do nothing- as one of subject identifiability. Indeed, it is identifiability that determines behavior (Bohnet & Frey, 1999).

Many experiments have shown that it is not so much the degree of dictator anonymity that determines donations, but rather the information that the dictator has about the recipient. Thus, for example, in an extensive article in which several hypotheses are reviewed to explain donations in dictator games and others are proposed, Bolton et al. (1998) argue that, if I know nothing about the other person and if I give her the money, I can err in her favor, but I prefer to err in my favor because I am not a saint (the “I’m-no-saint hypothesis”). According to these authors, however, the hypotheses that explain DG donations are unclear when the dictator has information about the recipient. It should be noted that Bolton et al. (1998) do not find evidence for what they call the “kindness hypothesis,” according to which “dictators might be motivated to give a gift for one or more of many reasons: e.g., a sense of moral obligation, pleasure derived from giving, to demonstrate kindness” (283). This hypothesis, however, encompasses very distinct motivations that must be analyzed separately. Although the pleasures of giving or demonstrating kindness are greatly affected by anonymity — in that the recipient not knowing if the dictator gives or not takes away the pleasure of giving and does not allow the dictator to demonstrate that she is kind — moral obligation is not affected by anonymity but rather by the lack of information about the recipient’s actual situation. In the classic Hoffman et al. (1994) experiment, and in most experiments along these lines, the complete lack of information about the recipient removes the necessary conditions to donate out of a sense of moral obligation:

By imposing double anonymity, HMSS [Hoffman et al., 1994] have removed virtually all motivation for donating money to one’s partner. The decision makers cannot identify each other, nor do they have enough information to know if their partner is poor or otherwise deserving of their generosity; thus there is little or no basis for altruism to play a part in the decision. We assert that fairness and altruism require context: the circumstances of the recipient determine what is the fair or appropriate charitable action to take. In the Hoffman et al. (1994) study, the information available to the Proposer is unlikely to motivate charitable behavior (Eckel & Grossman, 1996, p. 183-184).

If, as Eckel & Grossman have shown, altruistic behavior requires an appropriate context, which is inexistent in most experiments, what would happen if the dictator makes a decision in an easily recognizable moral context; a context in which dictators have information, for example, about their partner’s poverty or illness? In Eckel & Grossman’s experiment, in which the partner is the Red Cross, donations increase, but not in a spectacular way. Donations increase much more when the money that the dictator can distribute is the result of the joint effort of both the dictator and the partner, given that, in this case, “social pressures, moral imperatives, and the warm glow of giving are likely to magnify the reward to skillful or deserving recipients” (Ruffle, 1998, p. 259). As Ruffle points out, the results of this experiment contradict Hoffman et al.’s claim that other-regarding preferences have to do with “an expectations phenomenon” and not with “an autonomous private preference for equity” (Hoffman et al., 1994, p. 348).

Moral preferences, the sense of moral obligation, moral imperatives or equity appear repeatedly in discussions on the outcomes of a diverse range of DG experiments. Yet, in spite of the fact that donations occur in these experiments, we cannot speak with any certainty about dictators’ reasons for giving, as the only thing we know for sure is the preference that has been revealed in the decision, that is, whether the dictator gave or did not give. Indeed, aside from the theoretical speculations of experimenters, we do not know why dictators give. With a view to bridging this gap, in the following section we present two experiments (a hypothetical one and a real-payoffs one) in which the moral context is easily recognizable and in which dictators are expressly asked about the reasons for their action. We can then determine if the “kindness hypothesis” is to some degree certain: either because the dictators take pleasure in giving, give to demonstrate kindness, or give out of a sense of moral obligation. In the third section, we discuss these results and defend the hypothesis that dictators’ responses are determined by the “moral distance” between the dictator and the recipient, a distance that permits us to classify the behaviors that may arise in the dictator game.

2 Two moral dictator games: Design and results

To analyze how much the number of donors and the size of donations increase in a moral context, we need to study a DG situation in which the moral structure of the problem is set out in a clear manner. This requires dictators having relevant moral information about the recipient, that is, a context in which they are able to identify the moral nature of their decision. The instructions must prevent dictators from thinking that they are just playing a game, or that the recipients do not actually exist. The two experiments reported in Brañas-Garza (2006) and presented below fulfil these conditions.

2.1 Design and quantitative results of the experiments¹

The first experiment was conducted with 77 students from the University of Córdoba (Spain) in November 2002. Subjects were asked to collaborate in a research study on the problems that arise when trying to distribute non-divisible and finite goods. We avoided using the term “experiment” in order to dispel the possible negative effects of the word. Once the study was explained, the subjects received a large envelope containing the following items: a small envelope, three 5€ bills, a questionnaire, instructions and a sheet with 341 numerical codes, three of which were marked with a circle. The size of the envelope permitted the subjects to carry out the operations in complete privacy. The money and the questionnaire were placed in the small envelope where the centers were marked. The subjects then stood up and placed the envelope in a box. The subjects’ names did not appear anywhere and they kept the large envelope. After they were given the instructions they were told the following: “Only you will know what is in the large envelope” (the money that remained). After allocating the money, the subjects stood up, placed the small, sealed envelope (containing their choice, the money and the questionnaire) in the box and left the room. The subjects were not paid for participating in the experiment. None of the researchers (experimenters) that ran the experiment were professors at the universities where the experiments were performed and had not had any previous contact with the students.

In the instructions, the students were informed that the 341 codes appearing on the sheet referred to centers that receive medicines in Asia, Africa and South America, that they had been assigned three of these centers, and that for every 5€ they donated, a box of medicines would be sent to one of “their” three centers. All of the subjects in the experiment had to write the three codes on their small envelope and circle the centers — a maximum of three — to which they allotted the 5€ in medicines. The subjects were also told that they were entitled to keep 5€ for participating in the research study on the distribution of goods. Logically, if they kept 5€ for themselves, one of the centers would not receive any medicines (in this case they were told to circle only two of the three centers that had been assigned to them).² The procedure was performed under conditions of complete anonymity by means of a double-blind mechanism similar to that described by Hoffman et al. (1996).

The questionnaire included questions regarding sex, age, the weekly allowance given to them by their parents (none of the experimental subjects — all of whom were university students — had a paying job), and the reasons for their decision. This last question reads: “Could you please tell me the reasons for your decision in the problem posed here?”

Using the same instructions and under identical conditions of anonymity, a second experiment was performed in March 2003 with 98 students from the University of Jaén (Spain). The subjects were placed in four groups. This experiment was hypothetical as the subjects were not given real money and had to decide how to divide the amount hypothetically by responding to a series of questions. Furthermore, the decision was made under three distinct scenarios which were presented to them in a random manner:

1. a) No information about the recipients.

2. b) Information revealing that the recipients were poor people from the Third World.

3. c) Similar to b, but informing subjects that the money would be used to buy medicines.

The main results of the experiments are shown in Table 1. As can be observed, the increase in the number of donors and donations is spectacular compared to other experiments (see Section 1). As occurs in all the experiments, when subjects lack information about the recipient they hardly give anything. However, as in this case they knew that the recipients were poor, or poor and in need of medicines, the majority of dictators gave all the money — in spite of the fact that they were told in the instructions that they could legitimately keep one 5€ bill for participating and donate the other two. As we will see below, some of them did in fact keep the money, but only a few. Thus, information about partners clearly has an effect on donations to the same or to a larger degree than dictator anonymity. But what drives dictators to donate? What motives do they give for doing so — tastes, customs, traditions, religious motives, or moral reasons? The next subsection explores the main reasons given by the participants in both experiments.

Table 1: Two moral dictator game experiments.

¹ There are not significant differences in donations between the real-payoff experiment and the “Poverty+Drugs” treatment of the hypothetical experiment (see Brañas-Garza, 2006).

2.2 Qualitative results

As we have pointed out in the introduction, in this paper we are interested in how subjects explain their behavior after taking part in a dictator game. To this end, the experiments considered here are particularly appropriate since contrary to standard experimental dictator games, our experiments trigger an amazing amount of altruistic behavior. But before analyzing the claims made by the subjects, let us recall two features of the experimental design that have contributed to the prominence of unselfish behaviors. Whereas the first is a just a methodological question, the second leads us to the kind of philosophical claim that we shall make in this paper.

Firstly, the results are somewhat special because participants did not face the standard continuum choice of how to divide a pie, but instead had to decide how to distribute three 5€ bills among different potential recipients (see Brañas-Garza 2006). As we can observe in Table 1, the majority of participants (75%) allocate the full endowment to the recipients in the real-payoffs experiment. This is quite interesting for our analysis as it permits us to explore the reasons of a large number of participants who decided to give the money. Secondly, and more interestingly, the experiment used a “moral frame,” the kind of design that allows moral motivations such as fairness and altruism to play a role (Eckel & Grossman, 1996). Thus, although these are rather atypical dictator results, they are the kind of results that permit us to explore moral motivations in depth. Having clarified this, we will now concentrate on the main arguments provided by the participants.

2.2.1 The prominence of consequentialist arguments

Consequentialist reasons predominate among the answers given by participants in both the real-payoffs and the hypothetical experiment. The dictators give all the money because giving it has the most beneficial outcomes in terms of satisfying needs. In the real-payoffs experiment, 60% of the participants provide a consequentialist argument (see Table 2). The complete list of answers given by participants in the real-payoffs experiment can be found in the Appendix.³

Table 2: Reasons given by subjects in both experiments.

¹ The reasons from the hypothetical experiment correspond to the treatment in which the participants have full information. We report only this treatment to make it comparable with the real-payoffs experiment.

² The number of answers is restricted to 69 and 86 in the real-payoffs and the hypothetical experiment respectively since the rest of participants did not provide any explanation for their action.

As we will see below, the immense majority of those who gave all the money appealed to moral, consequentialist or deontological reasons.⁴ The former are more concerned about what would have happened had they not given the money (or what happens if they do), while the latter are more concerned about what they should do. Responses based on reasons of solidarity, equity, generosity or a sense of humanity are labelled as deontological. Although we believe that these labels are not particularly precise — dictators do not theorize, but express the reasons for their decision — a clear difference can be seen between those who underline the beneficial consequences of the action and those who appeal to the duty of giving.

The remaining labels reflect the various justifications provided by the dictators for giving the full amount or not giving anything. Thus, for example, some consider it legitimate to not give all the money (“legitimacy”) because the experimental instructions permit them to do so: “You are entitled to keep five euros, although this amount will depend on your decision [remember that everything is absolutely anonymous].” It is interesting to note that a large majority of the experimental subjects did not take this rule or “right” into account and instead gave all the money, explaining their action in moral terms of a consequentialist or deontological nature. Others, however, made their decisions in a random manner, did not believe in the experiment, did not believe that the money would reach its destination and thus did not give the money, or justified their decision for clearly self-interested reasons. Nonetheless, the justifications given for keeping part or all of the money are so varied and disperse that they become quite insignificant in comparison to the consequentialist or deontological reasons revealed by those who gave the full amount.

These results are confirmed by the findings obtained in the hypothetical experiment. Similar to what occurs in the first experiment, the majority of participants (79%) in the second experiment provide a consequentialist reason similar to that of the real-payoffs experiment (see Table 2). Again the majority provided a moral reason to justify giving all the money, although on this occasion the cases of self-interest increase, while those appealing to deontological reasons decrease. This may be due to the hypothetical nature of the experiment as it is more likely that dictators will keep all or part of the money when the decision does not have real consequences: as no money is involved, no one who is needy will lose out. The data, however, do not permit us to reach definitive conclusions on this question.

2.2.2 The importance of having a moral reason

Having a (moral) reason is decisive when facing a DG decision. In Table 3 the behavior of those who provide a moral reason — either consequentialist or deontological — is compared with the behavior of those who do not provide this kind of reason.

Table 3: Moral reasons and behavior.

Almost all the participants who give the three 5€ bills provide a moral reason for their behavior. Of course, this reveals nothing about the direction of the causality and we cannot claim that their behavior is grounded in moral reasons. But at least we can say that when they have to justify their actual behavior they use a moral argument. Moreover, the difference in giving behavior between those who provide a moral reason and those who do not is statistically significant (Mann-Whitney U-test: z=−6.65, p<0.001).

2.2.3 Behavioral differences between moral reasons

Having a moral reason to act is important in a dictator game, but are different reasons related to different types of behavior? In other words, does the giving behavior of those who provide a consequentialist reason differs from the behavior of those who provide a deontological reason? Since all the participants that provide a moral reason a posteriori give at least two bills, in Table 4 we study if there are differences between consequentialist and deontological subjects when deciding whether to allocate two or three bills to the recipient.

Table 4: Differences in giving between consequentialists and deontologists.

These results clearly show that the behavior of participants who provide a consequentialist reason do not differ from the behavior of those who provide a deontological reason (Mann-Whitney U-test: z=−0.319, p=0.750).

When carefully analyzing all of the answers given by consequentialists and deontological subjects, we find only a minor difference, yet it is one that can be interesting from a philosophical point of view. There is a positive correlation between considering that the dictator experiment constitutes a “cheap decision”⁵ and giving a consequentialist argument (see Table 5). This correlation is statistically significant (χ ²=7.359, p=0.007).⁶ Although we did not have a priori a theory about why these two variables should be related, and therefore it is a post hoc correlation, this result allows as to go further in our speculation about subjects’ moral reasons.

Table 5: Considering the DG a cheap decision by moral reason.

This can be explained by the fact that here we find two distinct types of consequentialism (Mulgan, 2001): almost half of those who give all the money in the real-payoffs experiment (49%) justify their decision in terms of a non-neutral consequentialism and one-third (32%) justify the reasons for their choice in terms of a neutral or impersonal consequentialism restricted to impersonal values. Neutral consequentialism is clearly presented by the subjects as the maximization of the impersonal value in statements such as “the more people that have medicines the better” or “this money is more useful in these countries” or “their scarcity.” Non-neutral consequentialism is demonstrated in the choice of an action that produces the best outcome, including one which will benefit the agent. The experimental subjects manifested this in statements such as “I’m not losing anything,” “the money isn’t mine, I didn’t earn it,” “they need it more than me,” “it’s of more use to them than to me,” “it’s not going to do me any good,” “it benefits them more than me,” or “I didn’t have the money when I came and I still don’t.” When ranking the possible outcomes from best to worst, non-neutral consequentialists consider that keeping all or part of the money does not benefit them in any way whatsoever — donating, then, is a cheap decision, even a morally cheap one.

2.2.4 Moral framing and consequentialist reasons

The “moral framing” used in our experiment is a crucial part of the experimental design. Above we described how moral framing increases donations in a standard dictator game. In addition, we explained how this moral framing triggers participants’ moral reasons, especially consequentialist ones. In this last part of the results we use the data from the hypothetical experiment to show how consequentialist arguments are more common, the more morally loaded the framing.

Figure 1 plots the reasons given by subjects to rationalize their behavior in the hypothetical experiment. In this case, we have categorized the reasons given by subjects when facing the three different treatments: without moral framing, knowing that recipients are poor and knowing this fact and additionally knowing that the money will be used to buy medicines. Figure 1 chiefly reveals three facts. First, when subjects do not receive any information about the recipients, almost 70% of them use this “lack of information” argument to justify their (mainly self-interested) behavior. Second, when they know that the money will be sent to poor countries, but do not know exactly where the money will end up, 20% of them do not trust the system. In other words, they are not sure if the money will be used for a good purpose. Third, the more information they have, the more frequent the consequentialist reasons. Thus, whereas almost no subject gives a consequentialist reason when having no information, this proportion rises to 50% when they know the recipients are poor and to 70% when they know the money will be spent to buy medicines. Consequentialist arguments are backed here by efficiency arguments, because morality is not a matter of wasting resources.

Figure 1: Reasons by treatment in the hypothetical experiment.

Donating money for medicines reduces the uncertainty of the dictator’s decision, improving the probability that her decision would have good consequences. Therefore, consequentialist arguments increase when the money is spent in medicines.

3 Discussion: Social distance or moral distance?

Eckel & Grossman’s claim that if dictators are provided the appropriate context they will have reasons to donate is more than fulfilled in Brañas-Garza (2006), albeit the result would have to be supported by further “moral” experiments. On the other hand, it may be too hasty to reject the kindness hypothesis if we focus on one aspect of this hypothesis, that is, dictators’ sense of moral obligation. Let us explore this aspect in greater detail.

First, like in many experiments, Brañas-Garza (2006) finds that dictator anonymity is not a crucial variable to understanding how much and why dictators donate. The key informational variable is the knowledge that dictators have about their partners. Similar to what occurs in the majority of experiments with DGs, in our hypothetical experiment, dictators hardly donate anything when the dictator lacks information about the recipient. But this is not due to a problem of social distance, lack of reciprocity or isolation. Those who lacked information explained their decision with statements such as “I don’t know who is going to get it, they might have more than me,” “I might need it for something more important than the recipients,” “if I don’t have any information about who the money is going to or what it is going to be used for, I won’t donate it” or “it could be someone who is richer than me,” statements which were repeated time and again to justify their action. When information about the recipient is lacking, it is perfectly legitimate to give priority to oneself: if the money goes to someone who needs it less than I do, I would be doing an injustice to myself. Thus it has less to do with a question of social distance than the fact that dictators make and analyze their decision whilst bearing in mind that they live in an unfair society in which it is quite likely that the unknown recipient will be better off economically than they are. In such a case there is no moral obligation to give anything.⁷

Drawing on Hoffman, McCabe and Smith’s concept of social distance, we propose a definition of moral distance as the degree of moral obligation that the dictator has towards the recipient. If the experiment is designed in such a way that donating is an actual obligation, the moral distance will be null and the dictator ought to donate all or part of her resources according to her level of well-being. In such a situation the dictator has the actual obligation to donate since it “is an obligation which, in a particular situation, is not superseded by any other obligation” (Crawford, 1969, p. 316; Zimmerman, 1996: 26). If the particular situation is such that donating cannot be said to be an obligation at all, the moral distance will be maximum.

The distance between dictator and recipient is not greater or lesser only as a function of the reciprocity that subjects belief exist within a social interaction, as in Hoffman, McCabe and Smith’s definition of social distance (Hoffman et al., 1996, p. 654). In fact, the concept of reciprocity depends on the concept of obligation — cases of reciprocity are those in which if you give me, I should give you; if I help you, you should help me, etc. But then social distance is only a case of moral distance, a special case in which anonymity plays a crucial, negative role, for the dictator must know what the recipient has done for him before to donate or not donate — as in Ruffle experiment (Ruffle, 1998). This negative relation — more anonymity less reciprocity — is not present in most instances of moral distance, for it is not true that more anonymity leads to a lesser obligation to donate when moral distance is null, and it is not true that less anonymity leads to more obligation to donate when the moral distance is maxim. However, anonymity must be ensured in order to exclude any possible influence of reputation (I give so that others see that I am giving).

Nor is moral distance related to the personal closeness between the dictator and recipient: the dictator can donate more money to a stranger than to an acquaintance if the stranger needs it, however close they may be. Following upon Jonathan Glover, Abelson (2005) refers to moral distance as “the emotional closeness between agent and beneficiary” and states that “our moral obligations to aid vary in strength with the emotional closeness to us of our fellows. When we know whom we are aiding, that very fact brings our beneficiaries closer to us than unknown strangers, and we feel more responsible for their well being” (Abelson, 2005, p. 35). In the experiments analyzed here, there is no emotional closeness between the dictator and the recipient. It has nothing to do with knowing who the recipient is, but the recipient’s state of well being. Given that the recipient is badly off, there is an obligation to help; an obligation which some fulfill — the majority in this case — and others do not. It is this obligation, and not emotional ties, what establishes the moral distance between dictator and recipient. Moral distance depends on the relevant information that the dictator has regarding the recipient’s situation and regarding her own situation.

The freedom that the dictator has to give or not give and the greater or lesser moral distance between the dictator and the recipient permits the behavior of the dictators to be labelled in the following way.

1. a) If the moral distance between the dictator and the recipient is null and the donations are also null or very low, the dictator behaves in a morally egotistical manner.

2. b) If the moral distance between the dictator and the recipient is maximum and the dictator donates nothing at all, the dictator behaves as a reasonably self-interested individual.

3. c) If the moral distance between the dictator and the recipient is maximum and the dictator gives part or all of her money, she performs an act of generosity.

4. d) If the moral distance between the dictator and the recipient is null and the dictator gives part or all of her money, her behavior is morally right.

Note that we are speaking about labelling dictator behavior, not about the reasons for action. We judge behavior from the viewpoint of the observer (the experimenter), from the outside, from the point of view of one who observes the moral distance between the dictator and the recipient and if the dictator gives money or not. The reasons for action are unknown: perhaps the dictator gave all the money — when, for example, moral distance is null — because the experiment was conducted on a Monday, or perhaps she gives part of the money when the moral distance is at its maximum because she was happy. Whatever the reasons may be for action, behavior can be labelled from the viewpoint of the experimenter, as we did so above. Given the manner in which DG experiments are usually conducted, it is not possible to determine the reasons for action, that is, the internal perspective of the dictator. It is possible to observe only the dictator’s conduct and speculate on her motives. What we can assert, however, is that first, the behavior of many dictators deviates from money maximization; second, that social distance does not explain why dictators donate and third, that the labels proposed here seem to fit the range of behaviors revealed in all the DG experiments (moral behavior, generosity, reasonable self-interest, selfish behavior). These labels respond to moral distance and donations; two objective categories that are independent of motivations.

It is interesting to note that all of these behaviors figure in the vast literature on DGs, albeit they appear separately in a variety of experiments. We have seen, for example, how dictators act with reasonable self-interest when lacking information given that in a situation of this type they are morally distant from the recipient. Self-interest works here as a social norm (Miller, 1999) that permits the dictators to take care of themselves without being labelled as selfish, that is, behaving in a morally egotistical manner. Even so, generosity is possible as some dictators donate part of the money to an unknown recipient in a double-blind DG. Likewise, when dictators have to divide legitimately earned money with the recipient they hardly donate anything — 95% of the dictators keep the full amount of money (Cherry et al., 2002, p. 1218). In such a case, there is no actual obligation to donate since the dictator obtained the money through her own personal effort. Here the moral distance with the partner is maximum, therefore making it legitimate to defend one’s own interests.⁸ For whatever the reason, moral conduct arises in experiments in which either the dictator has to divide the money that she has produced with her partner (Ruffle, 1998) or if she has to give money to alleviate poverty and buy medicines (Brañas-Garza 2006). On these occasions moral distance is null and there is an actual obligation to donate; an obligation that the dictators usually fulfill.⁹ In these cases we also find self-interested behavior. Given that the moral distance is null, there are those who keep all the money. When we modify the circumstances surrounding the decision, we will have both different moral distances and different outcomes.

Unlike what occurs in other experiments, here we have attempted to approach the internal perspective of the dictator, that is, the reasons that drive the dictator’s behavior, to see if behavior and motivations merge, in other words, if dictators donate money for moral reasons when moral distance is null. While the quantitative analysis by Brañas-Garza (2006) found a spectacular increase in donations when reducing the moral distance between the dictator and the recipient, the analysis of the qualitative data presented here shows that the dictators — without previous agreement, without communicating with one another and under conditions of absolute anonymity- appeal to moral motivations such as solidarity or helping the needy to justify their decision. To put it another way, fulfilling one’s obligation to give does not seem to be a random choice nor the result of non-moral reasons, but of moral reasons. Dictators seem to “measure” moral distance well as they refuse to give money when they have no information about the recipient, but do give it when they know the recipient is poor or is poor and needs medicine. In our experiments the response (to give or not to give) to the greater or lesser moral distance is generally tied to moral motivations.

Because the justification is given following the DG, we can argue that this is a mere post-hoc verbal justification that does not imply that the behavior we call moral is actually the result of a moral judgement (Haidt, 2001). As we stated above that, the experiment analyzed here does not allow us to establish a causal link between such behavior and the post hoc explanation for this behavior. But perhaps this is not the proper way to approach this issue. As Scanlon argues, judgements on what is correct or incorrect are not normal empirical judgements, but “claims about what we have reasons to do” (Scanlon, 1998, p. 2). The fact that the experimental subjects who donated all the money coincide in their reasons for action; a coincidence that occurs without previous agreement or group negotiation and pressure, seems to support the reasons the dictators gave for donating in the specific context proposed to them. Although these are undoubtedly post hoc reasons and we cannot affirm that they cause the action without committing a fallacy, there can be no doubt that it is quite a different ball game when dictators say they give money because their partner needs it than when they make their decision by tossing a coin in the air. The partner’s poverty seems to be sufficient grounds for the dictator to believe that she should donate in a specific situation and for her to indeed donate.

There may be other circumstances in which the dictators give and yet their behavior could be labelled as egotistical, since as we have seen the decision comes cheap: the money they donate has come at no cost to themselves and as many state, “won’t do them much good.” In such circumstances morally correct behavior is cheap. Perhaps this explains the high number of non-neutral consequentialist responses that we detected in both experiments. Whatever the case may be, it is interesting to confirm the coherency of the labels we have chosen to use since those who appeal to reasons of a deontological nature do not consider if their decision is cheap or not, but rather the duty of giving. This is yet another indication that these are true reasons for action; a fact that would in turn support the hypothesis that it is moral distance — the sense of obligation of the kindness hypothesis — and not social distance which marks the difference in experimental DGs. What seems to be unquestionable is that the moral framing determines dictators’ decisions. However, it is necessary to investigate further to understand better the influence of moral distance in the dictator’s behavior, and the relationship of this concept with other concepts and hypothesis used to explain donations in DGs.