The Impact of Descriptive Norms on Altruistic Behavior in Dictator Games

The Impact of Descriptive Norms on Altruistic Behavior in Dictator Games

Master’s Thesis

Name: Ting Lu Student Number: 11570512

Track: Behavioral Economics & Game Theory Supervisor: NilsKöbis

Statement of Originality

This document is written by Ting Lu who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

Summary

This thesis studies the impact of descriptive norms on altruistic behavior. Two kinds of descriptive norms are discussed in this study – (1) _{the belief of the frequency of} altruistic behavior and (2) _{the actual frequency of altruistic behavior. We conduct} simultaneous and sequential dictator games to test these two norms separately. Results show that the belief of the frequency of altruistic behavior influences altruistic behavior through the norm activation model while the actual frequency of altruistic behavior does not have significant effect on behavior after eliminating the effect of the recipient’s endowment. In order to test whether the recipient’s endowment influences people’s belief, we conduct another experiment. The result proves that a significantly negative relationship exists which means people believe in altruistic behavior is more when the recipient’s endowment is low. Since people are likely to behave in compliance with what they think the majority does, we conclude that informing people about the low endowment or the low progress of the recipient increases the frequency of altruistic behavior in their beliefs which in turn leads them to be altruistic. In a word, different descriptive norm impacts altruistic behavior differently.

Introduction

Will you be more willing to donate €2 to Wikipedia if they tell you that less than 1% of readers have donated money compared to they just ask for €2 without informing you about others’ behavior? And will you give a hand when you see a car accident and everyone around you is going to help compared to the situation that no one helps? The belief of the frequency of a behavior (the belief about others’ behavior) and the actual frequency of a behavior (others’ actual behavior) are defined as descriptive norms. Descriptive norms are proved to be powerful at persuasion in some situations (Cialdini, 2003) Thus, descriptive norms are often used as persuasion strategies to lead people to do the socially desired behavior, such as pro-environmental behavior (Goldstein et al., 2004). In this paper, we focus on the effect of descriptive norms on

altruistic behavior. The reason to study this kind of behavior is that altruistic behavior is extremely important in our life. Altruistic behavior which is defined as “the behavior that benefits another organism, not closely related” (Trivers 1971, P. 35), is highly correlated with human cooperation (Fehr and Gachter, 2002) and efficiency improvement (Fehr et al., 1997). If every individual is selfish, the world would not operate so efficiently as it does now. Imagine a society, no one will give you a hand when you are in trouble. Employees would pay the least effort and employers would pay the least amount of wages, public goods would not be constructed and the word “volunteers” might not have been existed. To some extent the world could be indifferent. But this is just an imagination, altruistic behavior is the innate behavior which controlled by DNA, even our hunter-gatherer ancestors shared the food with others.

Two studies are conducted in this paper. In study 1, we use online experiment to test the impact of two kinds of descriptive norms (the belief of the frequency of altruistic behavior and the actual frequency of altruistic behavior) on the activation of altruistic behavior. Previous studies use the norm activation model to present the activation of altruistic behavior. In the norm activation model, the activation of altruistic behavior relates to perceived responsibility and personal norms (moral obligations people hold for themselves). Two different views hold about how this model is defined. Some studies such as Black et al. (1985) believe that the model is a moderator model, in which the influence of personal norm on behavior is moderated by perceived responsibility. But recent studies such as De Groot and Steg (2009) find that the model is more like a mediator model, which means perceived responsibility is an antecedent of personal norm, and personal norm influences behavior. In study 1, firstly, we test whether the norm activation model is a moderator model or a mediator model, then the first question will be answered. The question is how does descriptive

norms relate to altruistic behavior. Then we conduct the second study to test how we can encourage people to behave more altruistically by influencing the descriptive

norms. In order to answer this question, we use another online experiment to test

whether the recipient’s endowment influences the descriptive norms. Four hypotheses are tested in our studies. The first hypothesis is that the norm activation model should be a mediator model in study 1. The second hypothesis is that descriptive norms should correlate with altruistic behavior (study 1). The third hypothesis is to test whether people are more altruistic when descriptive norms inform them that all of the other respondents are altruistic and when all of the others are egoist (study 1). As for the last hypothesis, the recipient’s endowment is supposed to influence the belief about the frequency of altruistic behavior (study 2).

The paper consists of 4 parts. The first part is literature review. In this part, we first explain why altruistic behavior is important and why people have altruistic behaviors. We focus on the normative explanation of altruistic behavior and then review papers on descriptive norms influencing people’s behavior like voting and donating. The second part of the paper is the introduction of study 1. Results are analyzed in the next part and we then explain our second study. Finally, we give our discussion and conclusion of the paper.

Literature review

Altruistic behavior which is defined as “the behavior that benefits another organism, not closely related” (Trivers 1971, P. 35) can be divided into various forms. Depending on how the utility gained by altruistic behavior, altruism can be divided into pure altruism and impure altruism (Wilhelm et al., 2017). Pure altruism assumes that the motive for altruistic behavior is the utility gained by the recipient’s benefit (Becker, 1974). While the impure altruism assumes people get utility from the action of giving, which is considered as warm-glow (Andreoni, 1989). From the perspective of the purpose of altruism, reciprocal altruism stands for altruistic behavior with the

expectation that the other organism will reciprocate in the similar way later (Trivers, 1971). The concept is close to the “tit for tat” strategy - be nice to the person who are nice to you and be bad to the person who are bad to you. As for the party which receives the benefit from altruism, parochial altruism is defined as the behavior which is “shaped by parochialism - a preference for favoring the members of one’s ethnic, racial or language group” (Bernhard et al. 2006, p. 912). Previous research (Rabbie, 1989) has found that the in-group favoritism of the in-group members relates to the expectation that in-group members will reciprocate altruism in the future, indicting correlation and overlap between different kinds of definitions.

The reason to study this kind of behavior is that altruistic behavior is very important in our life. Altruistic behavior enhances cooperation (Fehr and Gachter, 2002), even with strangers, benefiting the species in survival competitions. In the experiment Fehr and Gachter (2002), the mean cooperation level is significantly higher when punishment is allowed. That means altruistic punishment increases cooperation level as well as group efficiency. And Fehr et al. (1997) also show a big improvement of efficiency amongst employers and employees when reciprocal altruism exist compared to the game theory prediction (employers pay the minimum wage and employees pay the minimum effort). Moreover, from the perspective of evolution, group selection favors those groups with more altruistic individuals, these groups are more likely to survive in competition (Wilson, 1975).

Motivations for altruistic behavior vary a lot. One important reason for altruistic behavior is strong reciprocity, which is defined as rewarding kind actions as well as punishing unkind actions. Altruistic behavior with the purpose of reciprocity is defined as reciprocal altruism (Trivers, 1971). Fehr et al. (1997) and Gintis et al. (2003) show that reciprocity is the reason of the altruistic behavior among employers and employees. In Fehr’s study, employers’ payoff is defined as 100 times the effort

of employee minus the wage (1100e w). And the employee’s payoff is defined as wage minus the cost of effort ( 2 wc(e) ). The probability of being caught shirking is zero. If employers and employees are all self-interested, the result should be the same as the prediction of game theory. That is to say, employees always pay the minimum effort no matter what the wage received and employers always choose the lowest wage to pay. However, the result shows that employers prefer to choose higher wages than the lowest one, and employees are willing to pay more effort when they receive higher wages. This is considered as positive reciprocity. As for the negative reciprocity situation in the third-party punishment dictator games, Fehr and Fischbacher (2004) find that most third parties punish dictators who transferred less than half of their endowment. The explanation in their study is that third parties are enforcing the cooperative norms and then punish those who violated social norms.

Besides reciprocity, emotion is also considered as a crucial factor in causing altruistic behavior. If subjects punish free riders in the repeated public good games for the consideration of group efficiency, Fehr and Gachter (2002) find that negative emotions trigger punishment in the one shot public good game even when this is costly. The experiment in their paper shows shat people punish more when free riders deviate more to the average level. Their study supports their idea because cooperators become angry when free riders deviate more. On the other hand, positive emotions gained by altruistic behavior can be seen as extra rewarding which increases utility in the paper Fehr and Camerer (2007). And Batson et al. (1988) finds that empathy leads to altruistic motivation to help others.

From the perspective of psychology, studies have proved that the activation of helping behavior relates to ascription of responsibility, personal norms (moral obligations people hold for themselves) and awareness of consequences. Based on data of bone

marrow donating questionnaire, Schwartz (1973) finds that personal norms only impact volunteering when ascription of responsibility is not very low. This means the norm activation model is a moderator model in Schwartz’s study. However, a lot of studies find that the model is more like a mediator model, which means personal norms mediate the relationship between ascription of responsibility and altruistic behavior. For instance, De Groot and Steg (2009) test the norm activation model in five different studies and find consistent support for the mediator model. And Steg and De Groot (2010) further reveal that personal norms can be strengthened by stressing subjects’ responsibility for the problem. For this reason, we expect that the norm activation model in our studies would be a mediator model.

In this paper, we focus on the impact of social norms on altruistic behavior. Social norms are behavioral regularities and rules in our daily life, such as dropping garbage into the trash bin rather than on the ground. People obey social norms because they are afraid of embarrassment or sanctions by others (Bicchieri and Mercier, 2014). Social norms are applied within the group, i.e., only group members obey them and out-group members don’t have to enforce them (Hogg and Reid, 2006). However, Bernhard (2006) finds that out-group members punish norm violators in the other group in the third-party punishment game.

Social norms can be divided into descriptive norms and injunctive norms. Descriptive norm deals with the expected frequency of a specific behavior, for example, how many other people would take the action (Köbis et al., 2018). And injunctive norm is about whether this behavior is approved by others, that is, what others think you should do. We focus on the impact of descriptive norms on behavior. Various studies have shown that descriptive norm influences people’s behavior in a variety of fields and this persuasion strategy has been already used to lead people to do the socially desired behavior. For example, Goldstein et al. (2004) use a field study to find that

social norms are more successful in making guests reuse towels compared to the traditional appeal by using environmental protection as an argument to force behavior. Reno et al. (1993) show that social norms have impact on littering. But descriptive norms appear to be more situation-specific since they only successfully reduce littering when environment is clean. Besides environment protection, Köbis et al. (2015) finds that descriptive norms are highly correlated with corrupt behavior, that is if participants perceive corruption is a common behavior, then they are likely to corrupt themselves. Similar findings have proved in voting, Gerber and Rogers (2006) shows that it is more effective to motivate people to vote by informing them that a lot of people have voted than a few people have voted.

However, there is rare study that focuses on the influence of descriptive norms on altruistic behavior. For example, will people be more willing to donate if they know that everyone is donating? And will people give a hand if they know that everyone is helping? Latane and Darley (1968) finds that descriptive norms influence helping behavior in emergency situations. In their study, the room where the participants are made to wait for an interview is gradually filled with smoke. They find participants are more likely to interfere in when he or she is the only witness compared to the situation that other witnesses exist. In order to explain the influence of group size on helping behavior in an emergency situation, Latane and Nida (1981) conclude that social influence, audience inhibition, and diffusion of responsibility are three important factors causing the phenomenon of the social inhibition of helping. The audience inhibition in that paper means that the presence of others could inhibit helping behavior when people think their behavior could be seen by others and might be seen negatively. Social influence means a subject might feel the situation is not so emergent and the inaction could be the expected behavior if all of others do not take action (injunctive norm). Diffusion of responsibility means that other persons’ existence could make the subject feel less responsible because the subject might believe others will help (descriptive norm) in the emergency situation. Besides,

Croson et al. (2009) use a field experiment to show that donors tend to make higher contributions when they think that others might make high(-er) contributions. And they point out that the desire to conform to the perceived descriptive norms might be the crucial reason to change people’s behavior and donate more. Compared to Croson’s study, our experiment controls more variables. In Croson’s study, the recipient’s payoff is different when a subject thinks that others will contribute at a high level and at a low level. Then, the difference of the recipient’s payoff could influence donors’ contribution. In our experiment, we eliminate the effect of the recipient’s payoff on behavior. By doing this, the real effect of descriptive norms could be seen. And based on Croson’s study, we expect that people would be more altruistic when they perceive a high frequency of altruistic behavior than a low frequency of altruistic behavior.

In study 1, we use experiment to test how do the descriptive norms influence altruistic behavior in dictator games. Dictator games are suitable to test altruism, if participants are selfish, then they will always choose the dominant strategy - don’t transfer money. If our hypotheses hold true in dictator games, to some extent they could be applied to other situations. Our study includes two kinds of descriptive norms (the belief about the frequency of altruistic behavior and the actual frequency of altruistic behavior). The impact of the belief about the frequency of altruistic behavior is tested in a simultaneous 5 players dictator game (4 dictators with 1 recipient). In this situation, 4 dictators make decisions at the same time. Before making decisions, they have to make a belief of the other players’ behavior. On the other hand, the impact of the actual frequency of altruistic behavior is tested in a sequential 5 players dictator games (4 dictators with 1 recipient), where dictators make decision one by one. Data in these two situations are compared with the data in the control group - no descriptive norm situation (1 dictator with 1 recipient) in order to eliminate the effect of B’s endowment on A’s behavior. We combine our study with the norm activation model to see what roles do perceived responsibility and perceived normative pressure

play in the relationship of descriptive norms and altruistic behavior. In order to further study the factor that influences descriptive norms, we conduct study 2 to test the impact of B’s endowment on A’s belief about the frequency of altruistic behavior.

Study 1

Procedure

In the first study, we investigate whether descriptive norms correlate with altruistic behavior. For that reason, we analyze altruistic behavior under different situations with various perceived frequencies of altruistic behavior. In total 127 participants (102 player A and 25 player B) did our online questionnaire. 50 player A were from the University of Amsterdam, and 52 player A were from China, player B were all from China. Players are told at the beginning of the experiment that they have the chance to win up to €10 depending on their decisions.

Participants are randomly divided into player A and player B. They fill out an online questionnaire consisting of three parts (as shown in figureⅠand figure II). The first control part consists of four rounds. In every round, each player A is paired with a player B. Player A is endowed with €10 in each round while player B is endowed with €0 or €2 or €4 or €6 in different round. Player A decides to transfer €2 to player B or not to transfer in every round. Part 2 is a simultaneous dictator game which consists of one round. In this part, every four player A and one player B are divided into a group. Player A is endowed with €10 and player B is endowed with €0. First, player A predicts how many other A players in the group will transfer €2 to player B. After predicting, every player A decides to transfer €2 or not to transfer €2 to player B simultaneously. Part 3 are sequential dictator games which consists of 10 rounds. Compared to part 2, in part 3 four player A make decisions one by one in different situations. For each player A, he or she makes decision as the first one in one round, as the second one in two rounds, as the third one in three rounds, and as the last one in four rounds. Decisions of previous player A are told and each player A is asked to predict the decisions of other players afterwards. Then player A decides to transfer €2 to player B or not to. Besides deciding to transfer money or not to in every round, player A choose their perceived responsibility value and perceived normative pressure value in a self-reporting way (details will be described below). Player B does not need to do any decision in all rounds. At the end of the experiment, one part is selected randomly and one group of participants are selected randomly to get paid.

Measures

One kind of descriptive norm (the belief of the frequency of altruistic behavior) is induced by one item. The item is “please predict how many other players will transfer money” in the simultaneous dictator game and “please predict how many players will transfer money after you” in sequential dictator games. The other kind of descriptive norm (the actual frequency of altruistic behavior) is induced by informing players the previous players’ decisions in sequential dictator games. In order to test the impact of descriptive norms on altruistic behavior, perceived normative pressure and perceived personal responsibility, we measure the three elements in every round. Perceived normative pressure is measured by one item, which is “I ought to transfer €2”. Perceived personal responsibility is measured by the other one item, which is “I feel responsible for the earnings of player B.” Participants rate their agreements on each item on a 7-point scale from 0 (totally disagree) to 6 (totally agree). Participants rate these two items in every round. Altruism is measured by the decision of choosing to transfer €2 or not to transfer. Transferring €2 is seen as altruistic behavior while not transferring is regarded as egoistic behavior.

Results

Hypothesis 1: The norm activation model should be a mediator model in our study Result 1: Mediation has been found in the norm activation model, which means perceived responsibility is an antecedent of personal norm, and personal norm influences behavior

Following the method by Baron and Kenny (1986) to test mediating in the model, mediation is revealed in the norm activation model in our study. With the intention of ruling out the effect of B’s endowment on A’s behavior, 102 objects in one round of the no descriptive norm part (B is endowed with €2) are used to test the hypothesis. In

the OLS regression of altruistic behavior on perceived responsibility (p<0.001, coefficient of perceived responsibility =0.171), the regression of perceived normative pressure on perceived responsibility (p<0.001, coefficient of perceived responsibility =0.812) and the regression of altruistic behavior on perceived normative pressure (p<0.001, coefficient of perceived normative pressure =0.167), the relationships are all significant. The coefficient of perceived responsibility to perceived normative pressure is 0.812, indicating highly correlated and positive relationship. Moreover, in the regression of altruistic behavior on perceived responsibility and perceived normative pressure, the effect of perceived responsibility (p=0.005, coefficient of perceived responsibility=0.086) and perceived normative pressure (p<0.001, coefficient of perceived normative pressure=0.105) are both significant, supporting the partial mediation. Same results are also found in the situations when B is endowed with €4 or €6. But in the situation when B is endowed with €0, full mediation is revealed since perceived responsibility is no longer significant after controlling perceived normative pressure.

Hypothesis 2: Descriptive norms are expected to correlate with altruistic behavior Result 2: After eliminating the effect of B’s endowment, the belief of the frequency of altruistic behavior impacts altruistic behavior by influencing perceived responsibility while the actual frequency of altruistic behavior does not have significant effect on the behavior

Firstly, we test whether B’s endowment impacts A’s behavior. The OLS regression of transfer behavior on B’s endowment in the no descriptive norm situation shows significantly negative relationship (p<0.001, coefficient of B’s endowment=-0.04). In order to rule out the effect of B’s endowment on altruistic behavior, we analyze the difference of perceived responsibility and the difference of perceived normative

pressure in simultaneous dictator games compared to those in control part. And the difference of perceived responsibility and the difference of perceived normative pressure in sequential dictator games compared to those in control part are also analyzed. For example, if a subject is of the view that the other player A will transfer money in the simultaneous dictator game, then in the subject’s mind, the current payoff of the player B should be €2, which is the same as that in the control situation that B is endowed with €2. Then the difference of perceived responsibility and the difference of perceived normative pressure in the simultaneous and control situations are caused by the belief of the frequency of altruistic behavior. In our study, ∆responsibility and ∆pressure stand for the difference of perceived responsibility and the difference of perceived normative pressure separately when B is believed to own a specifically current payoff in the simultaneous dictator game compared to the control situation that B is endowed with the same payoff. Belief stands for the number of other A players the subject thinks will transfer money. The table below summarizes detailed descriptions.

We still follow the method by Baron and Kenny (1986) to test mediating in the model. In the regression of ∆pressure on belief (p=0.031, coefficient of belief=0.480), the regression of ∆responsibility on belief (p=0.002, coefficient of belief=0.698), and the regression of ∆pressure on ∆responsibility (p<0.001, coefficient of ∆responsibility

=0.624), the relationships are all significant. The coefficient of belief to ∆responsibility is 0.698, indicating highly correlated relationship. And in the regression of ∆pressure on belief and ∆responsibility, the effect of belief is insignificant (p=0.789), but the effect of ∆responsibility is significant (p<0.001, coefficient of ∆responsibility=0.618), supporting the full mediation.

The belief of the frequency of altruistic behavior influences altruistic behavior as hypothesis, however, the actual frequency of altruistic behavior does not make any significant difference after eliminating the effect of B’s endowment. In order to test the mediating in the model of the actual frequency of altruistic behavior, data in the four rounds that the subject is the last player A to make decision in the sequential dictator games are compared to data in the rounds that B is endowed with the same current payoff in the control situation (details are in table II). For example, the round that 1 other player A before the subject transferred money (B’s current payoff is €2) in the sequential dictator games is compared to the round that B is initially endowed with €2 in the control situation. And the difference of perceived responsibility and the difference of perceived normative pressure in these two situations are caused by the actual frequency of altruistic behavior. ∆responsibility ∆pressure and ∆behavior stand

for the difference of perceived responsibility, the difference of perceived normative pressure and the difference of behavior in the round of sequential dictator games compared to the control situation that B is endowed with the same current payoff. Actual frequency stands for the number of other players who transferred money.

In the regression of ∆pressure on actual frequency of altruistic behavior (p=0.163), the regression of ∆responsibility on actual frequency of altruistic behavior (p=0.298), and the regression of ∆pressure on ∆responsibility (p<0.001, coefficient of ∆responsibility=0.720), relationships are not all significant, indicating that the actual frequency of altruistic behavior does not have significant effect on perceived responsibility and perceived normative pressure, but perceived responsibility has significantly positive effect on perceived normative pressure. Besides, in the regression of ∆behavior on actual frequency of altruistic behavior, the relationship is still insignificant. That is to say, the actual frequency of altruistic behavior does not have significant effect neither on perceived responsibility and perceived normative pressure, nor on altruistic behavior.

Result 3: Compared to the no descriptive norm situation, transfer level is significantly lower when participants believe that 0 or 1 of other 3 players will transfer money, and

significantly higher when participants believe that 2 or 3 of other 3 players will transfer money

We compare the behavior of those participants who believe no one else will transfer money (B’s current payoff is €0) in the simultaneous dictator game to the behavior in the control dictator game when B is endowed with €0. In these two situations, B currently owns €0 when A make decisions. We find that 72.55% of player A in control situation chose to transfer money while only 8.33% of player A transferred in the simultaneous situation. Then we use one-tailed T test to see whether the level of transfer is significantly higher in control situation compared to the simultaneous situation. Results show significantly higher transfer level in the control situation (P<0.001, t=6.801). Then we compare the situation that B currently owns €2 to the situation that B is believed to own €2, 62.75% of player A in the former situation and 36.84% of that in the latter situation transferred money. Significantly higher transfer level is also found in the control situation (P=0.003, t=2.793). As for the situation that B is endowed with €4 and the situation that B is believed to own €4, 51.96% of player A in the former situation and 73.81% of player A in the latter situation transferred money. And the level of transfer is significantly higher in the latter situation (P=0.006, t=-2.577). Besides, comparing the situation that B is endowed with €6 to the situation that B is believed to own €6, we find that 49.02% of player A in the former situation and 90.00% of player A in the latter situation transferred money. The transfer level is also significantly higher in the simultaneous situation (P=0.001, t=-3.669). To summarize, compared to the no descriptive norm situation, transfer level is significantly lower when participants believe that 0 or 1 of other 3 players will transfer money, and significantly higher when participants believe that 2 or 3 of other 3 players will transfer money.

inform them that all of others are altruistic than all of others are egoist

Result 4: In the simultaneous dictator game, people are more likely to transfer money when they believe a high frequency of altruistic behavior than a low frequency of altruistic behavior

We analyze the data in the simultaneous dictator game. In this part, 4 dictators decide to transfer €2 or not to at the same time after predicting how many other players will transfer money. The result shows in total 102 subjects, 42 subjects (41%) believe that 2 of other 3 players will transfer money, and 73.8% of them choose to transfer €2 to player B. Besides, 38 subjects (37%) believe that 1 of the other 3 players will transfer money, and around 37% of them transferred money. 10 subjects (10%) believe that all of other 3 players will transfer money and 90% of them transferred money themselves. The rest 12 subjects (about 12%) believe that no one else will transfer money, and only one of those subjects transferred money, that’s about 8%. We then use one-tailed T test to check whether the level of altruistic behavior is significantly higher when subjects believe that all other 3 players will transfer money compared to when they believe that no one will transfer money. Result shows significantly higher transfer level in the high frequency situation (P<0.001, t= -6.274). The result is consistent with the result in Croson et.al (2009), that is to say, if people believe more of the others will transfer money, then they will transfer money more likely.

Result 5: Informing people that a low actual frequency of altruistic behavior is better at activating altruistic behavior than not informing them about others’ behavior

We have proved that the actual frequency of altruistic behavior does not have an effect on transfer actions after removing the effect of B’s endowment (result 2). So the information of the actual frequency influences people’s behavior by conveying the

information about B’s endowment. That is to say, the information that nobody transferred money has similar effect of the information that B is endowed with €0 on behavior. The average transfer level in the situation that A is the last one to make decision and no other players transferred money is 66%. While in the simultaneous dictator game, the average transfer level is around 54%. The one-tailed T test shows that the transfer level is significantly higher in the former situation than in the latter situation (t=1.718, p=0.044). We explain this lower average transfer level in the simultaneous dictator game by showing that about half of the player As believe a low frequency of altruistic behavior when they lack other information.

We find that the belief of the frequency of altruistic behavior influences people’s behavior by a specific mode, but the other kind of descriptive norm (the actual frequency of altruistic behavior) does not have significant effect on people’s altruistic behavior itself. But why do these two descriptive norms work differently? One explanation is that when people are told others’ certain actions, they pay more emphasis on B’s endowment. They make decisions by considering B’s endowment rather than other players’ decisions. But why people do not behave like this way in the simultaneous dictator game is not clear. Maybe in the sequential dictator games, we show B’s endowment on the graph but we do not show that in the simultaneous situation. It could make people consider less of B’s endowment and consider more of the belief when making decisions if they could not see the certain payoffs on the graph and have to calculate it by themselves.

Study 2

Study 1 proves that belief has significant effect on altruistic behavior, so influencing people’s belief about the frequency of altruistic behavior could be an efficient way to

lead people to behave altruistically. But what factor impacts people’s belief? We make a hypothesis that in the sequential dictator games, the information about other players’ behavior before the subject could influence the subject’s belief about other players’ behavior afterwards.

In order to test our hypothesis, we regress the data in the sequential dictator games. When the subject is the third player to make decision, the previous two players’ decisions are made known to the subject and the subject predicts the last player’s behavior. The OLS regression of the belief about the last player’s behavior on the first two players’ behavior does not show significant relationship (p>0.05). Then we regress the belief about the last two players’ behavior on the first player’s behavior when the subject is the second player to make decision, the result also shows insignificant relationship. In fact, the information of other players’ behavior relates to two effects. One is the effect of B’s payoff to A’s belief. That is to say, when player A is the third to make decision, if none of the 2 previous players transferred money, then B has €0 when A makes his/her belief about the last player’s behavior. And if both of the 2 previous players transferred money, then B has €4 when A makes his/her belief. The difference of B’s payoff could influence the belief about the last player’s behavior, that is the effect of B’s payoff. The other effect on this belief is the pure effect by others’ behavior after eliminating the effect of B’s payoff. Because the relationship is insignificant when the two kinds of effects exist, we suppose that these two effects influence the belief from two opposite directions. As a result, the two effects cancels out, showing insignificant relationship on the belief when the information of others’ behavior is told.

In order to test the hypothesis, we firstly test the effect of B’s endowment on player A’s belief about others’ behavior. If the relationship is significant, then the hypothesis holds.

Procedure

In the second study, we conduct another simultaneous dictator game to test whether B’s endowment has significant effect on player A’s belief about the frequency of altruistic behavior. We did a priori sample size calculation before data collection. The result shows that the minimum required sample size is 28 (effect size = 0.30, statistical power level = 0.80, number of predictors = 1, probability level = 0.05). 50 participants (40 player A and 10 player B) who were from China did our online questionnaire. In this experiment, participants are randomly divided into player A and player B. Player A is endowed with €16, player B is endowed with €0 or €5 or €10 in three different rounds. The group consists of 4 player A and 1 player B. 4 player A decide to transfer €2 to player B or not to transfer at the same time. Before making the decision, they have to predict how many other player A will transfer money. For example, the question in one round is “you are endowed with €16, player B is endowed with €5, please predict how many other player A will transfer money?” And player B does not need to do any decision in the experiment. Participants are told at the beginning of the experiment that one round will be selected randomly at the end of the experiment, and one group of participants will be selected in that round. Participants who made the right prediction in the selected group will be paid.

Measures:

The belief of the frequency of altruistic behavior is measured by one item, which is “Please predict how many other player A will transfer money?”

Result

B’s endowment, the relationship is significantly negative (p<0.001, coefficient of B’s endowment=-0.14). That means the increase of B’s endowment decreases the number of other players who will transfer money in player A’s belief. The result is consistent with the hypothesis. Then we can explain why the information of previous players’ behavior does not have significant effect on people’s belief in the sequential dictator games. We would like to answer this question by first clarifying it. When A is the third player to make decision and the two previous players did not transfer money, then B’s current payoff is €0. Because B’s payoff has negative effect on the belief of the frequency of altruistic behavior, then the prediction of the last player’s behavior should be significantly different compared to the situation that previous two players transferred money (people should be more likely to think that the last player will transfer money when B has €0 than B has €4). However, data shows that there is no significant difference between these two situations. So we infer that others’ behavior (the actual frequency of altruistic behavior) has positive effect on the belief of the frequency of altruistic behavior, offsetting the negative effect of B’s endowment on behavior. But the conclusion still has limitations. In the experiment, each group consists of 5 players, but the number of people who are engaged in a similar situation in real life, such as donation, could be huge. It is natural to think that the information about others’ behavior impacts belief differently when group size changes. Besides, B’s payoff and the group size are shown clearly on the graph in the experiment, and player A can calculate B’s final payoff easily under a specific belief. But in real life, the recipient’s payoff and the group size are usually not told, making it difficult to make an accurate belief about the frequency of altruistic behavior. Because of the limitations in the experiment, it is possible that the information of how many others have done impacts people’s belief about how many others will do in real life.

Discussion

as pro-environmental behavior (Reno et al., 1993), voting behavior (Gerber and Rogers, 2006), corrupt behavior (Köbis et al., 2015) and so on. But studies about descriptive norms and altruistic behavior are limited. Previous studies prove that descriptive norms correlate with donor behavior (Croson et al., 2009) and helping behavior (Latane and Darley, 1968). But as far as we know, there is no study which rules out the effect of the recipient’s payoff on altruistic behavior when studying the impact of descriptive norms on altruistic behavior. The other innovative point in this study is that we induce two kinds of descriptive norms (the belief of the frequency of altruistic behavior and the actual frequency of altruistic behavior) and compare their influences on altruistic behavior. The comparison of the impact of these two descriptive norms on behavior are rare in previous studies. The main finding of this thesis is that these two descriptive norms work differently. The belief of the frequency of altruistic behavior influences altruistic behavior significantly while the actual frequency does not have significant effect after ruling out the impact of the recipient’s payoff on behavior (study 1). To be specific, compared to the no descriptive norm situation, the belief about the low frequency of altruistic behavior decreases the willingness to be altruistic while the belief about the high frequency of altruistic behavior works oppositely.

Our research also makes a number of other findings. Firstly, the norm activation model of altruistic behavior in our study is a mediator model. That means perceived responsibility is an antecedent of personal norm, and personal norm influences behavior. This is consistent with the study in De Groot and Steg (2009). Nevertheless, the belief about the frequency of altruistic behavior influences behavior via the norm activation model which means belief has an impact on altruistic behavior by influencing perceived responsibility. Then we can conclude that it is an efficient way to activate altruistic behavior by firstly strengthening the responsibility or increasing the frequency of altruistic behavior in people’s belief. Another finding is that the descriptive norm (the belief of the frequency of altruistic behavior) seems to

guide people’s behavior toward a specific direction. People are more likely to be altruistic when they believe a high frequency of altruistic behavior compared to a low frequency of altruistic behavior. The belief of a high frequency of altruistic behavior is an altruism norm while the belief of a low frequency of altruistic behavior can be seen as a anti-altruism norm. People seem to take the norm as a guideline when making altruistic decisions. On the other hand, the information of the actual frequency of altruistic behavior does not have significant effect on altruistic behavior itself as it influences behavior by conveying the message of B’s endowment.

Clearly, making people believe that more the others are altruistic could be an efficient way to lead people to behave altruistically. We conduct another experiment to test whether the recipient’s endowment influences player A’s belief about other A players’ behavior. The result shows that people believe a high frequency of altruistic behavior when the recipient’s endowment is low. In conclusion, focusing on the disadvantage of the recipient’s situation is an ideal way to activate altruistic behavior. On the one hand, the low endowment of the recipient increases people’s willingness to be altruistic directly. On the other hand, recipient’s low endowment also makes people believe that more other people are altruistic, they activate the altruistic behavior themselves. As for why the information about the actual frequency of altruistic behavior does not have significant effect on the belief, our best guess is that the effect of the recipient’s payoff offsetting the effect of others’ behavior on the belief.

There are also many limitations in our research. Firstly, due to data limitations, we do not have access to participants’ age, gender and education background. Taking these variables into account could make the analyses more accurate. Besides, in these two experiments we give players some initial endowment. This house money could make players be more generous in experiments than in real life. And in experiments, the endowments of player A are the same, the difference in A’s endowment could make

the result be different in real life. In a word, the experiment which lacks external validity to some extent, may not be applicable to some real life situations. For instance, the experiment in Croson et al. (2009) concludes that mentioning the percentage of listeners who donate can influence the belief about others’ behavior, and then encourage people to donate. However, in our study, informing participants others’ behavior does not make any difference on belief. Our explanation for this conflict between Croson’s study and our study is that the effect of the recipient’s endowment and the effect of others’ behavior could be different in different situations. In some situations, the former is stronger while in some situations the latter wins. The trade-off between these two elements decides whether the information of others’ behavior influences the belief. Moreover, the way framing the information could also make people focus on different perspectives. What is more, we conduct dictator games in this research, future work remains to test descriptive norms in other settings, such as public good games. And we only test the effect of recipients’ endowment on belief, other factors could also impact belief efficiently.

Conclusion

In a word, descriptive norms impact altruistic behavior differently. The belief about the frequency of altruistic behavior influences altruistic behavior by impacting perceived responsibility. And people are more likely to be altruistic when they believe a high frequency of altruistic behavior. Informing people about the disadvantageous situations of the recipient, such as the low endowment, could be an ideal way to arouse altruistic behavior. Because the low endowment of the recipient not only stimulates altruistic behavior directly but also increases the frequency of altruistic behavior in people’s belief. However, the information about the actual frequency of altruistic behavior does not have effect on altruistic behavior itself. It influences behavior by conveying the message about the recipient’s current payoff. And the information of the actual frequency also does not have effect on belief. So, telling

people we only have €1 rather than less than 1% of readers donated when asking for donations, and telling others that we are lacking help instead of no one gave us a hand when asking for help.

Reference

Andreoni, J. (1989). Giving with Impure Altruism: Applications to Charity and Ricardian Equivalence. Journal of Political Economy,97(6), 1447-1458.

Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical

considerations. Journal of Personality and Social Psychology,51(6), 1173-1182. Batson, C. D., Dyck, J. L., Brandt, J. R., Batson, J. G., & Al, E. (1988). Five studies testing two new egoistic alternatives to the empathy^altruism hypothesis. Journal of Personality and Social Psychology,55(1), 52-77.

Becker, G. (1974). A Theory of Social Interactions.

Bernhard, H., Fischbacher, U., & Fehr, E. (2006). Parochial altruism in humans. Nature,442(7105), 912-915.

Bicchieri, C., & Mercier, H. (2014). Norms and Beliefs: How Change Occurs. The Complexity of Social Norms,37-54.

Black, J. S., Stern, P. C., & Elworth, J. T. (1985). Personal and contextual influences on househould energy adaptations. Journal of Applied Psychology,70(1), 3-21. Cialdini, R. B. (2003). Crafting Normative Messages to Protect the

Environment. Current Directions in Psychological Science,12(4), 105-109. Croson, R., Handy, F., & Shang, J. (2009). Keeping up with the Joneses: The

relationship of perceived descriptive social norms, social information, and charitable giving. Nonprofit Management and Leadership,19(4), 467-489.

Fehr, E., Gachter, S., & Kirchsteiger, G. (1997). Reciprocity as a Contract Enforcement Device: Experimental Evidence. Econometrica,65(4), 833. Fehr, E., & Gächter, S. (2002). Erratum: Altruistic punishment in humans. Nature,415(6868), 137-140.

Fehr, E., & Fischbacher, U. (2004). Third Party Punishment and Social Norms. SSRN Electronic Journal.

Fehr, E., & Camerer, C. F. (2007). Social neuroeconomics: The neural circuitry of social preferences. Trends in Cognitive Sciences,11(10), 419-427.

Gintis, H., Bowles, S., Boyd, R., & Fehr, E. (2003). Explaining altruistic behavior in humans. Evolution and Human Behavior,24(3), 153-172.

Goldstein, N. J., Cialdini, R. B., & Griskevicius, V. (2004). A room with a viewpoint: Using norm-based appeals to motivate conservation behaviors in a hotel

setting. PsycEXTRA Dataset.

Groot, J. I., & Steg, L. (2009). Morality and Prosocial Behavior: The Role of

Awareness, Responsibility, and Norms in the Norm Activation Model. The Journal of Social Psychology,149(4), 425-449.

Hogg, M. A., & Reid, S. A. (2006). Social Identity, Self-Categorization, and the Communication of Group Norms. Communication Theory,16(1), 7-30.

Köbis, N. C., Prooijen, J. V., Righetti, F., & Lange, P. A. (2015). “Who Doesn’t?”—The Impact of Descriptive Norms on Corruption. Plos One,10(6). Köbis, N. C., Iragorri-Carter, D., & Starke, C. (2017). A Social Psychological View on the Social Norms of Corruption. Corruption and Norms,31-52.

Latane, B., & Nida, S. (1981). Ten years of research on group size and helping. Psychological Bulletin,89(2), 308-324.

Latane, B., & Darley, J. M. (1968). Group inhibition of bystander intervention in emergencies. Journal of Personality and Social Psychology,10(3), 215-221.

Ottoni-Wilhelm, M., Vesterlund, L., & Xie, H. (2014). Why Do People Give? Testing Pure and Impure Altruism.

Rabbie, J. M., Schot, J. C., & Visser, L. (1989). Social identity theory: A conceptual and empirical critique from the perspective of a behavioural interaction

model. European Journal of Social Psychology,19(3), 171-202.

Reno, R. R., Cialdini, R. B., & Kallgren, C. A. (1993). The transsituational influence of social norms. Journal of Personality and Social Psychology,64(1), 104-112. Rogers, T., & Gerber, A. (2006). Everyones voting and so should you: How descriptive social norms affect intention to vote. PsycEXTRA Dataset.

Robert L. Trivers. (1971). The Evolution of Reciprocal Altruism. The Quarterly Review of Biology,46(1), 35-57.

Schwartz, S. H. (1973). Normative explanations of helping behavior: A critique, proposal, and empirical test. Journal of Experimental Social Psychology,9(4), 349-364.

Steg, L., & Groot, J. (2010). Explaining prosocial intentions: Testing causal

relationships in the norm activation model. British Journal of Social Psychology,49(4), 725-743.

Wilson, D. S. (1975). A theory of group selection. Proceedings of the National Academy of Sciences,72(1), 143-146.