OUP user menu


Darren Dahl (Editor in Chief)Eileen FischerGita JoharVicki Morwitz

16 out of 115

Friends of Victims: Personal Experience and Prosocial Behavior

Deborah A. Small, Uri Simonsohn
DOI: http://dx.doi.org/10.1086/527268 532-542 First published online: 1 October 2008


Why do different people give to different causes? We show that the sympathy inherent to a close relationship with a victim extends to other victims suffering from the same misfortunes that have afflicted their friends and loved ones. Both sympathy and donations are greater among those related to a victim, and they are greater among those in a communal relationship as compared to those in an exchange relationship. Experiments that control for information support causality and rule out the alternative explanation that any effect is driven by the information advantage possessed by friends of victims.

  • Affect/Emotions/Mood
  • Charity and Gift Giving
  • Judgment and Decision Making
  • Experimental Design and Analysis (ANOVA)

Charitable giving in 2005 reached more than $260 billion, or around 1.9% of personal income (Giving USA Foundation 2005). Thousands of charitable organizations in the United States and elsewhere support a wide array of causes. What drives people to give to charity, and how do they choose among competing causes? Why do some people donate money to researching Alzheimer's disease and others to nourishing children in Africa? Why do some people Race for the Cure® of breast cancer, while others walk to cure multiple sclerosis? The current research sheds light on these general questions. More specifically, we examine one powerful reason behind people's donation choices. It is expected that personally knowing someone with a particular misfortune influences prosocial behavior toward others with the same misfortune.

To systematically establish the association between relationships with victims and prosocial behavior toward other victims of the same misfortune, we surveyed volunteers at several local charitable organizations and asked them about their relationships with victims prior to volunteering. For each of several misfortunes, they indicated if they had an acquaintance, colleague, casual friend, close friend, distant relative, parent, sibling, spouse, and/or other who has or had suffered from it. We obtained 116 responses:26 from Action AIDS, 75 from the Alzheimer's Association, and 15 from the Special Olympics.

As predicted, a significantly higher proportion of volunteers, prior to volunteering, knew a victim of the misfortune targeted by the charity for which they volunteered (M = 81.6%) than knew a victim of a misfortune targeted by the other two charities (M = 46.2%; χ2(1) = 40.6, p < .0001). For each of the three organizations, the pairwise comparisons were significant at the 5% level. We obtained similar and also significant results when comparing the average number of victims that volunteers knew.

Based on this preliminary evidence, we sought to understand the psychological processes supporting this pattern of charitable decisions. We theorized that a personal relationship with a victim predisposes an individual to sympathize with other victims of the same misfortune and that this greater sympathy moves them to behave prosocially. This occurs because the feelings inherent to close relationships carry over to others. This carryover is hypothesized to occur only for victims of the same misfortune, and thus it is different from the well-documented affect carryover effect (Forgas 1995; Loewenstein and Lerner 2002; Schwarz and Clore 1996) because it is specific to those to whom the target feels connected; that is, their ailment matches that of a loved one or friend.


In this section, we develop our theoretical framework in three steps. We begin by discussing the effects of reducing social distance on sympathy and giving. Then we examine why friendship with a victim increases sympathy and giving specifically to other victims of the same misfortune. Finally, we consider the role of relationship closeness in moderating the core effects.

Sympathy and Social Distance

The existing literature on sympathy and social distance consistently shows that factors that reduce feelings of distance between victims and potential benefactors tend to promote sympathy and helpful behavior (Loewenstein and Small 2007). We use the terms sympathy to refer to “emotional concern for others” (Wispe 1986) and social distance to refer to “feelings of closeness (or lack thereof) between individuals” (Loewenstein and Small 2007). Consistent with the link between social distance and sympathy, research shows that people give more to identifiable victims than to unidentifiable or statistically presented victims (Kogut and Ritov 2005a, 2005b; Small, Loewenstein, and Slovic 2007). This effect is demonstrated even when no meaningful information is provided about the identified victim (Small and Loewenstein 2003). Other identifying factors, such as showing a victim's face or being in the presence of a victim, also increase prosocial behavior (Bohnet and Frey 1999; Charness and Gneezy 2006). Additionally, individuals are more sympathetic toward victims who belong to their in-group rather than their out-group (e.g., Dovidio et al. 1997; Flippen et al. 1996; Levine et al. 2002), who are similar to them (Krebs 1975; Stotland and Walsh 1963), or who have a vested interest in a cause (Miller and Ratner 1998; Ratner and Miller 2001). Research by Batson and colleagues has shown that reducing social distance by asking or priming people to take the victim's perspective also enhances sympathetic feelings and altruistic behaviors (Batson, Early, and Salvarani 1997; Batson et al. 2003; Coke, Batson, and McDavis 1978). In sum, these findings illuminate several factors that reduce social distance between victims and potential benefactors and hence promote sympathy and giving. In the following section, we describe how misfortunes themselves, rather than victim characteristics, can reduce social distance.

Personal Experience

It is widely believed that personal experience with misfortune is consequential. Real-world correlational data support this notion. For example, Barnett et al. (1986), found that participants who had been raped reported greater sympathy when watching a video about a rape victim than did those who had never been raped. Christy and Voigt (1994) found that those who reported being abused as a child indicated that they would be more likely than those who had never been abused to intervene if they saw a child being abused. A recent experimental study by Simonsohn et al. (forthcoming) showed that even when informational content, relevance, and format are held constant, experienced information causally influences behavior more strongly than information that strangers are observed experiencing.

Other research examines the impact of being close to someone else's personal experience with misfortune. For instance, research in risk perception finds that knowing a victim of a misfortune increases perceived risk and propensity to take precautionary action (see, e.g., Manheimer, Mellinger, and Crossley 1966; Schiff 1977). Similarly, in a recent article, Washington (2006) finds that male legislators tend to vote more liberally with regard to women's issues if they have daughters rather than sons.

Although the literature on personal experience does not address giving behavior per se, the findings suggest that people are especially affected by the experiences of people close to them. Given the relationship between social distance and sympathy, it is likely that the personal experience of a close other, by reducing social distance to other victims, also enhances sympathy and giving.

Intergroup Contact

The findings about personal experience dovetail with Intergroup Contact Theory, a prominent theory in the domain of prejudice and intergroup relations. Since the 1930s, social scientists have proposed that interracial experiences reduce prejudice (Baker 1934) and that separation promotes it. This idea was formalized by Allport (1954), who inspired hundreds of studies on the general topic (see Pettigrew and Tropp [2006] for a meta-analysis that includes 515 studies). In virtually all reported studies, personal contact with one out-group member increases positive feelings toward other members of that group, presumably because it reduces social distance to the group. Thus, if victims of a particular misfortune are construed as a group, contact with a member of the misfortune-group should also reduce social distance and increase positive feelings toward the rest of the group.

Implicit in this literature, and also in the literature on personal experience, is the notion that such effects are directed toward categorically similar individuals and experiences. For example, contact with a citizen from Afghanistan should improve feelings toward Afghans but is not expected to change feelings toward African Americans. Similarly, knowing someone who has been through a hurricane should increase hurricane risk avoidance but is not expected to change heart disease risk avoidance. Based on these implicit assumptions, we expect that a personal relationship with a victim increases sympathy only to victims who share the same misfortune.

Thus, combining the insight that personal experience and intergroup contact reduce social distance with the notion that reduced social distance promotes sympathy and prosocial behavior, we come to our first hypothesis:

  • H1: A personal relationship with a victim of a given misfortune increases sympathy and hence prosocial behavior toward victims who suffer from the same misfortune and does not increase sympathy and prosocial behavior toward victims who suffer from other misfortunes.

Relationship Closeness and Orientation

Relationships between people are not all equal. Most important for our purpose, they vary in social distance. Distance in relationships is emphasized in the Aron et al. (1991) theory of self-other overlap. Aron et al. argue that the closeness in relationships affects resource-allocation strategy because outcomes are perceived to be interdependent. This is related to the distinction made between communal and exchange relationship orientation (Clark et al. 1987). Relationships among friends, family members, and significant others tend to have a communal orientation. Members of communal relationships feel close to one another and are willing to provide benefits for one another even if doing so brings no personal gain. In contrast, relationships among acquaintances or people who do business with one another tend to have an exchange orientation. In these more utilitarian relationships, feelings are more distant and benefits are only offered to obtain some personal gain.

Although closeness and relationship orientation research focuses on behavior within the relationship, we propose that closeness within a relationship also predicts sympathy and prosocial behaviors toward people outside such relationships. Specifically, we propose that sympathy should vary as a function of relationship type because of inherent differences in closeness. For example, just as prosocial behaviors are more likely in communal than in exchange relationships, we expect greater prosocial behavior toward other victims among those in a communal relationship with a victim of the same misfortune than among those in an exchange relationship.

  • H2: The closer the relationship with a victim, the greater the degree of sympathy and giving toward other victims of the same misfortune.

We report the results from three studies designed to test our hypothesis. In study 1, we examine, using a survey, the relationship between knowing a victim of various real misfortunes (Alzheimer's disease, breast cancer, and serious job layoff) and self-reported sympathy toward other hypothetical victims of the same or different misfortunes. In study 2, we assess the causal nature of this relationship in an experiment that randomly assigns (i) friendship, (ii) victimhood, and (iii) the identity of the potential aid recipient. Finally, in study 3, we examine how the nature of the relationship with a victim influences sympathy and prosocial behavior toward other victims by inducing pairs of participants to be acquaintances, colleagues, or friends with a victim. We close by discussing both basic research and practical marketing implications of our results.

Study 1

Although the survey of volunteers described in the introduction is suggestive of our hypotheses, there are several possible mechanisms for the association. Charities, for example, may locate in areas with high concentrations of victims, and hence volunteers would naturally (and spuriously) tend to know victims targeted by charities for which they volunteer. Another possibility is that people may volunteer because they seek to directly help their loved ones rather than others suffering from the same misfortune. To rule out these possibilities, we conducted a survey asking participants about their sympathy toward hypothetical victims of various misfortunes in different scenarios. Therefore, the survey does not rely on actual charitable behavior, which could result from the alternative explanations just described. The sample consisted of 100 travelers at Thirtieth Street Station in Philadelphia (49% males; ages ranged from 17 to 69, with a mean of 39 years).


Participants were approached in the waiting area of the train station and asked to fill out a short survey for academic research in exchange for a lottery ticket. The survey took approximately 5 minutes. Participants read three fictional scenarios about individuals experiencing various misfortunes: Alzheimer's disease, breast cancer, or a job layoff with disastrous financial consequences (app. A contains the original stimuli). After reading each scenario, participants were asked to report their feelings toward the person in the story using Batson and colleagues' Emotional Response Scale (Coke et al. 1978), on which respondents indicate the extent to which they feel upset, distressed, sympathetic, alarmed, grieved, troubled, compassionate, perturbed, softhearted, worried, tender, and disturbed, on a 1 (Not at all) to 7 (Extremely) scale. Prior research documents that the scale consists of two factors, which we refer to as sympathy and personal distress. The sympathy factor consists of sympathetic, compassionate, tender, and softhearted and predicts altruistic motives, whereas the personal distress factor consists of the remaining items and typically predicts egoistic motives (Batson, Fultz, and Schoenrade 1987). Factor analyses of our data also led to these two factors. We use the average of the four in the sympathy factor as a measure of sympathy (Cronbach's α = .82).

After completing the scale for all three scenarios, respondents indicated how close they were to a person who has suffered from each of the three misfortunes described in the scenarios, on a scale ranging from 1 (Never known anyone) to 10 (Person closest to me in the world). We refer to this 1–10 variable as friend. Finally, participants reported their age and gender.


We asked each of the 100 respondents about three scenarios, resulting in a total of 300 respondent-scenario observations. One respondent did not answer any of the friend questions, leaving 297 observations. Eleven respondents left blank at least one of the four variables that were used to measure sympathy, leaving 266 observations for the analyses that follow. Table 1 reports means and standard deviations for friend and sympathy and demonstrates ample variation within each scenario.

View this table:
Table 1

Means and Standard Deviations for Friend and Sympathy: Study 1

Alzheimer's diseaseCancerLayoff
Friend (1–10)4.575.976.80
Sympathy (1–7)4.985.384.34
No. observations898889

Note.—Friend is self-reported closeness to a real victim suffering from the misfortune. Sympathy is self-reported sympathy toward the victim in the scenario (average of four variables from Batson et al.'s Emotional Response Scale). Standard deviations are in parentheses.

    To assess the impact of friend on sympathy, controlling for differences across scenarios, we estimated a regression with respondent-scenario as the unit of observation (i.e., three observations per respondent), sympathy as the dependent variable, and friend as the key predictor. The regression included dummies for each scenario and controls for respondents' average answer to friend and sympathy in the other two scenarios (average-friend and average-sympathy). Intuitively, this specification assesses whether participants are relatively more sympathetic toward victims of misfortunes with whom they have a relatively closer connection. We also estimated a specification in which, instead of these averages, we included respondent fixed effects (one dummy per respondent). The results are reported in table 2.

    View this table:
    Table 2

    Friendship with a Victim and Sympathy toward Victim in Scenarios (OLS): Study 1

    Controls for average of friend and sympathy in other two scenarios

    Individual fixed effects

    Friend (1–10).059**.060**
    Alzheimer's disease scenario dummy1.019**.717**
    Cancer scenario dummy1.499**1.057**
    Average friend in other two scenarios−.054*
    Average sympathy in other two scenarios.837**
    Individual fixed effects (df = 89)NoYes
    No. observations266266

    Note.—Dependent variable is sympathy. Boldface font highlights the critical variable in the regression. Entries in table correspond to point estimates in OLS regression. Standard errors are in parentheses. The unit of observation is a person-scenario (i.e., participants without missing observations appear three times in the regression). Friend is self-reported closeness to a real victim suffering from the misfortune. Sympathy is self-reported sympathy toward victim in scenario (average of four variables in Batson et al.'s Emotional Response Scale). Average friend and average sympathy are the average of these variables, for that respondent, in the other two scenarios.

    • * Significant at the 5% level.

    • ** Significant at the 1% level.

    Consistent with hypothesis 1, the point estimate for friend is positive and significant in columns 1 and 2 (β = .06, p < .01), indicating that respondents who were closer to someone with the misfortune depicted in the scenario expressed more sympathy for the victim in that scenario.

    The negative coefficient on the average-friend variable indicates that the closer respondents were to people suffering from misfortunes in the other scenarios, the less relative sympathy they felt for the victim in this scenario. This further supports hypothesis 1 by establishing that friendship with a victim has a positive effect only on other victims of the same misfortune; friendship does not make people more sympathetic toward those suffering from another misfortune. The positive coefficient for average-sympathy, in turn, implies that self-reported sympathy was positively correlated across scenarios at the individual level—an effect probably driven by individual differences (e.g., in use of scales or in overall sympathy).

    Although these results are consistent with hypothesis 1, the correlational nature of the data limits our ability to draw inferences about causality. To this end, we conducted two additional experiments. The goal of study 2 was to assess whether such effects are causal, and the goal of study 3 was to examine hypothesis 2, that sympathy for a victim varies as a function of relationship type/closeness to other victims of the same misfortune.

    Study 2

    In this study, we randomly assigned participants to become friends with either a victim or a nonvictim and then examined their generosity toward other victims of the same or different misfortune. In the real world, friendship with a victim is confounded by information: friends likely know more than do others about their friend's ailment—such as its severity, risk, and opportunities to aid those suffering from it. By providing identical information to friends of both victims and nonvictims, we can isolate the causal impact of friendship with a victim on generosity toward other victims of the same misfortune.

    In this experiment, we first induced friendship between randomly matched participants and then turned some friends into victims by having them give up a $10.00 endowment. We then compared the amount of money allocated by those keeping the $10.00 to another participant-turned-victim, as a function of friendship with a victim.

    We also varied the allocation recipient (participant-turned-victim or scholarship fund) to examine if the effect is limited to victims of the same misfortune. This also allows us to rule out the possibility that participants are using their friend's outcome as a reference point and thus are more likely to part with their own money when their friend has $0.00 than when their friend has $10.00. If either the carryover effect of sympathy was general, rather than limited to victims of the same misfortune, or if participants used their friend's outcome as a reference point, we would expect friends of victims to give more than friends of nonvictims, both to another victim and to the scholarship fund. Thus, the experiment consisted of a 2 (friend status: friend is victim/friend is not victim) × 2 (recipient: participant who lost money/scholarship fund) between-subject design.


    We conducted the experiment in large classrooms. Sessions ranged in size from having 24 to 97 University of Pennsylvania undergraduate participants. A total of 280 individuals (62.6% female) participated in exchange for the amount they earned as a result of the Allocation Task.


    All participants began the experiment with $10.00. The experimenters assigned dyads of two people sitting directly in front of/behind each other. Each dyad was assigned a number, and within each dyad one participant was assigned an A role and the other a B role. Participants knew their own number/letter and that of the other member of their dyad, but they did not know the number/letter of any other participant.

    Friendship Induction

    All dyads engaged in an abridged version of the Relationship Closeness Induction Task (RCIT), commonly used to generate relationships in a laboratory setting (for a review, see Sedikides et al. [1999]). The relationship induction was labeled Communication Task and included two sets of questions. The first set consisted of introductory questions such as “What is your first name?” and the second consisted of more personal questions such as “What is one recent accomplishment that you are proud of?” (see app. B for the full set of questions).

    Creating Victims

    All A's got to keep their $10.00. B's, however, could lose their money. The experimenter flipped a coin in front of everyone; the outcome of the coin flip determined if odd-numbered B's or even-numbered B's lost their money. Therefore, half of the A's were friends with a victim of money loss, and the other half of A's were friends with a nonvictim.

    Allocation Task

    The A's then drew one slip of paper from a bag to determine the recipient to whom they could donate in the Allocation Task. If an A drew her own number, she was instructed to return it and draw again, so that no one was allocating to their own friend. Half of A's drew a number representing a B who lost money; the other half drew the name of a scholarship fund.

    In sum, A's were randomly assigned into one of four conditions: their B friend either became a victim or not, and their potential recipient in the Allocation Task was either another B who lost money or the scholarship fund. Hypothesis 1 predicts that friend-of-victim A's will be more generous to B's than friend-of-nonvictim A's but that A's will be equally generous to the scholarship fund regardless of what happens to their friend.


    Overall there were 142 observations. The average allocation was $2.65, with a standard deviation of $2.52. Allocations were slightly larger for the scholarship fund (M = $2.78, SD = 2.85) than for fellow participants (M = $2.54, SD = 2.17) but not significantly so (t(140) = .56, p = .573). A total of 137 participants made allocations between $0.00 and $5.00, but five allocated the full $10.00 (four of these were given to the scholarship fund). The results reported below do not change qualitatively if we remove these five participants. More important, they do not change if we just remove the single $10.00 allocation given to a B participant (in the friend of victim condition).

    Figure 1 shows the average allocation in each of the four conditions. We conducted a 2 (friend's state) × 2 (recipient) ANOVA on allocations. It revealed no main effect of either friend's state (F(1,138) = .339, p = .56) or recipient (F(1,138) = .394, p = .53) but a significant interaction between the two factors (F(1,138) = 4.31, p < .05). This interaction results from the fact that friends of victims gave more to their B recipient (M = $3.06, SD = 2.36) than did friends of nonvictims (M = $1.95, SD = 1.80, t(71) = −2.25, p < .03) but that friends of victims gave less to the scholarship fund (M = $2.45, SD = 2.86 vs. M = $3.08, SD = 2.31), though not significantly so (t(67) = .913, p = .364).

    Figure 1

    Mean Allocation by Condition in Study 2

    Note.—Lines above bars represent standard errors.

    In sum, the results support hypotheses 1: friends of victims were more generous toward same-misfortune victims, but they were no more generous than friends of nonvictims toward different-misfortune victims.

    Study 3

    Study 2 discretely manipulated friendship with a victim. Study 3 tests hypothesis 2 by examining whether relationships of varying closeness to a victim lead to different degrees of sympathy and prosocial behavior toward other victims. The study consists of an experiment where we induce three different types of relationships—acquaintances, colleagues, or friends—that vary in closeness. According to Relationship Orientation Theory (Clark and Isen 1982; Clark and Mills 1979), exchange relationships (acquaintances and colleagues) are more distant than communal relationships (friends). By creating both acquaintances, those who just see each other, and colleagues, those who interact but in an inherently less communal/close way than friends, we can study two levels of closeness within the exchange relationship category and further shed light on how the nature of relationships influences sympathy for others.

    We made two additional changes in this study as compared with the procedures of study 2. First, to examine mediation, we measured both sympathy and prosocial behavior. Second, we eliminated a possible confound of minimal groupness (Tajfel 1970) from study 2, where it was possible that friends of victims gave more because of the parity of the number identifying dyads (even vs. odd) with the recipient. Here, the only unique shared feature between friends of victims and potential recipients is the misfortune of the friend.


    We recruited 96 University of Pennsylvania student participants for a number of studies conducted in a university lab in exchange for $10.00. Participants were randomly paired and accompanied to a private room by an experimenter, one dyad at a time.

    Relationship Induction

    Each pair was randomly assigned to one of three conditions. In the friend condition, participants completed the same Relationship Closeness Induction Task (RCIT) employed in study 2.

    For the colleague condition, we created a modified RCIT designed to induce a working relationship rather than a friendship, using questions such as “Are you relatively better with Excel or with PowerPoint?” (see app. C for the full set of questions). The instructions stated that the purpose of the task was “to figure out how best to work together.” The task had the same number of questions and time allotted to discuss answers as the traditional RCIT. In the acquaintance condition, participants entered the room together and could see each other but did not interact; instead they moved directly to the task described in the next paragraph.

    Creating Victims

    Following the relationship induction, each participant received an envelope containing $10.00 in paper tokens exchangeable for real money at the end of the study. Each participant also received a sheet of paper labeled Allocation Task. The sheet contained their unique pair number and instructions stating that each participant would randomly receive one of two cards labeled KEEP or LOSE and that the participant who draws KEEP can keep the $10.00 and the participant who draws LOSE must forfeit the tokens to the experimenter.

    Allocation Task

    All participants returned to individual cubicles in a large laboratory room for the allocation task. Participants who drew KEEP in the previous task received a survey entitled Part II. The survey instructed that each KEEP participant would draw a number that represented a LOSE participant to whom they could give any share of their $10.00. The instructions were otherwise identical to those of study 2. Participants who drew LOSE did not complete this portion of the study.


    Following the allocation decision, each KEEP participant completed the same emotional response scale (Coke et al. 1978) with respect to the person whose number they drew. As in study 1, we created an average of the four variables associated with sympathy (α = .89).

    Sympathy for Scenario Victims

    Because this study required more between-subjects conditions than study 2, we did not include a condition in which KEEP participants could give to victims of a different misfortune (as we did in study 2 with the scholarship fund). Instead, we asked participants to read scenarios similar to those used in study 1 and to complete the same Emotional Response Scale regarding the victim in each scenario.

    Follow-up Questions

    To control for preexisting relationships between pairs of participants, participants rated how well they knew the person with whom they were in the room prior to the study, on a scale ranging from 1 (Never met this person before) to 7 (This person is my best friend). Finally, as a manipulation check of relationship closeness, participants rated how close they felt to the person with whom they were matched for the relationship induction task, on a scale ranging from 1 (Not at all close) to 7 (Extremely close).


    Of the 96 participants who drew KEEP, 91 stated they had never before met the person with whom they entered the room. The results presented below were obtained analyzing only these 91 observations, but they remain qualitatively unchanged if participants with preexisting relationships are included.

    Results of the manipulation check indicate that the friend manipulation led to closer relationships (M = 2.96, SD = 1.25) than the colleague manipulation (M = 2.17, SD = 1.39), which, in turn, led to closer relationships than the acquaintance manipulation (M = 1.22, SD = .50; all three pairwise differences are p < .01). Given that the scenario victims suffered more severe misfortunes than losing $10.00, it is not surprising that participants expressed less sympathy for the fellow participant losing $10.00 (M = 3.14, SD = 1.67) than for the scenario victims (M = 4.82, SD = 1.33, t(179) = 7.5, p < .0001).

    An Omnibus test rejects the null that sympathy toward the person they could give money was the same across conditions (F(2,88) = 6.29, p = .003). Consistent with hypothesis 2, sympathy in the friends condition (M = 3.91, SD = 1.55) was higher than in both the colleagues condition (M = 2.98, SD = 1.73, t(58) = 2.28, p = .025) and the acquaintances condition (M = 2.52, SD = 1.45, t(60) = 3.49, p < .001). Although the difference between colleagues and acquaintances was in the expected direction, it did not prove significant (t(58) = 1.15, p = .255).

    Consistent with the prediction that sympathy is only affected toward victims of the same misfortune, an Omnibus test does not reject the null that sympathy toward the scenario victims was the same across conditions (F(2,88) = .32, p = .728), and none of the pairwise comparisons led to statistically significant differences (p = .53, .45, and .60, respectively).

    Turning to allocations, the average allocation across conditions was $1.84, with a standard deviation of $2.17. Only one participant, assigned to the colleague condition, gave more than $5.00 (the full $10.00).

    An Omnibus test of relationship type on allocation amount shows that giving varies significantly across conditions (F(2,88) = 3.49, p = .035). Figure 2 plots both the raw mean allocations by condition and those net of controlling for sympathy; the sympathy-controlled means are relevant for our subsequent discussion about mediation.

    Figure 2

    Mean Allocation by Condition in Study 3

    The pattern of allocation amount mirrors that of sympathy: friends of victims gave more than colleagues of victims (t(58) = 1.75, p = .083) and more than acquaintances (t(60) = 2.58, p = .011). If we exclude the single $10.00 allocation, the difference between friends and colleagues of victims is significant at the 5% level (t(57) = 2.51, p = .014).

    Similar to the results with sympathy as the dependent variable, allocations by colleagues and acquaintances were in the predicted direction but not significantly different (t(58) = .79, p = .434). Excluding the $10.00 allocation outlier does not qualitatively change this result.

    Mediation Analysis

    We next examine whether the effect of relationship type on allocation amount is mediated by sympathy, following the four steps suggested by Baron and Kenny (1986). The first step was reported previously: allocation is influenced by the relationship manipulation (see fig. 2). The second step consists of establishing that sympathy is associated with allocation amount. To this end we estimated a linear regression with allocation amount as the dependent variable and sympathy as the predictor. The point estimate is positive and significant (β = .64, p < .0001): the more they felt sympathy toward the person to whom they could give, the more people gave.

    An interesting question with respect to step 2 is whether the association between sympathy and allocation amount is driven by sympathy per se or by a more general emotional reaction. To address this issue, we estimated a multivariate regression with allocation amount as the dependent variable and both sympathy and personal distress (the average of the other eight variables in the emotional response scale) as predictors. Consistent with Batson et al. (1987), we found that sympathy is positively associated with allocation amount (β = .91, p < .0001) and personal distress is not (β = −.354, p = .061). This is consistent with the idea that the impact on giving is not a general emotional but rather a specifically sympathetic response.

    Step 3 consists of estimating jointly the effect of the relationship type and of sympathy on allocation amount, and step 4 assesses whether the impact of relationship type is significantly attenuated with respect to step 1. Step 3 is depicted graphically in figure 2, which shows the differences across relationship types with and without controlling for sympathy. Consistent with sympathy mediating the effect, the differences are greatly attenuated when controlling for sympathy. For example, controlling for sympathy causes a drop in the difference between friends and colleagues from $.95 to $.42. Since amount allocated was not significantly different between the colleague and acquaintance conditions, we estimate a Sobel test (1982) collapsing across these two conditions, which assesses whether the difference between allocation amount in these conditions and allocation amount in the friends condition was mediated by sympathy. The Sobel test does support mediation (Z = 2.72, p = .007). Estimating the Sobel test separately for the two pairwise comparisons leads to significant mediation for the difference between the friend versus acquaintance condition (Z = 3.11, p = .002) and marginally significant mediation for the friend versus colleague condition (Z = 1.84, p = .067).

    In sum, we find that closer relationships with victims lead to more sympathetic feelings toward other victims of the same misfortune and that this effect in turn mediates greater generosity to other such victims.

    General Discussion

    Despite the social and economic value of charitable giving to our economy, there is a dearth of research in this domain (Bendapudi, Singh, and Bendapudi 1996). Furthermore, this literature has emphasized solicitation strategies, focusing for example on the role of framing (Gourville 1998), anchor points (Fraser, Hite, and Sauer 1998), mortality salience (Ferraro, Shiv, and Bettman 2005), and the bundling of products with promised donations by companies (Strahilevitz and Myers 1998). Although this knowledge is useful, it does not address what is unique about charitable giving: that it concerns the well-being of others rather than the self and that sympathy, therefore, is of central importance (for an exception, see Bagozzi and Moore 1994).

    In contrast, there is a great deal of research on behavior with respect to others' welfare in psychology and economics. However, this literature typically focuses on the identity of victims but has not considered the identity of the misfortunes, independent of the individuals involved. This is a notable exclusion, because much real-world charity outside the laboratory goes to organizations that target specific misfortunes rather than individuals. Understanding prosocial behavior toward misfortunes is therefore fundamental to understanding real-world charitable giving.

    Here, we argue that a misfortune can be a shared characteristic that facilitates prosocial behavior toward individuals whose misfortunes match that of a friend or loved one. Consistent with hypothesis 1, we find that closeness with a victim increases sympathy and prosocial behavior toward other victims of that misfortune. Of importance, no such effect is observed for victims of other misfortunes, which distinguishes our effect from general affective carryover effects. Finally, in support of hypothesis 2, we find that closer relationships give rise to greater sympathy for others of the same misfortune than do more distant relationships, which is analogous to theory and evidence about differential behavior within relationships of varying closeness.

    An important feature of the studies we present is that, through experimental design, we isolate the impact of friendship with a victim on sympathy and generosity while ruling out alternative explanations. In particular, an association between knowing people with a given misfortune and prosocial behavior toward others with the same misfortune could be explained by a spurious association due to unobservables. For example, charities may locate in areas of high incidence of the misfortune they target, and if volunteers choose organizations near their homes, then a spurious correlation would arise. The association could also be driven by a self-interested motive on the part of the donor. Specifically, people whose friends have a specific misfortune may act prosocially toward organizations that target that misfortune because they think that they themselves may get it or be affected by it in the future. Finally, differences in informational content or salience could explain such a relationship because friends of victims typically know more about their friend's disease and about opportunities to help. Our evidence strongly suggests that independent of such causes, friendship with a victim affects sympathy and social preferences directly.

    To be sure, there is much anecdotal evidence supporting this notion. Celebrities such as former First Lady Nancy Reagan (Alzheimer's disease), soccer star Mia Hamm (bone marrow disease), and actor Rob Lowe (breast cancer) actively participate in efforts to help others who share their loved ones' misfortunes. Similarly, all national presidents of Mothers Against Drunk Driving (MADD) joined the organization after a family member was killed by a drunk driver.

    Limitations and Future Research

    As with all empirical research, there are trade-offs between examining behavior inside versus outside the laboratory. In the lab we can manipulate friendship and victimhood, eliminating the potential confounds associated with knowing a real person with a particular misfortune. However, we sacrifice the profundity of personal experience with real tragedy. Studies 2 and 3 reflect this compromise. Alternatively, we can study people whose actual close friends and relatives have experienced misfortunes of much greater severity than a loss of $10.00, as we do in study 1, but the correlational nature of the studies limits our ability to make inferences about causality. The studies we present combine both sources of data to provide converging evidence that friends of victims sympathize more with other people with the same misfortune.

    Future research should examine more precisely the relationship between friendship with a victim and sympathy. In particular, it would be useful to distinguish between two possible paths to such an effect. The first is that friendship changes the perceived severity of the misfortune. The second possible path is that friendship increases the weight placed on the well-being of other victims suffering from a friend's misfortune. The former explanation implies that friends of victims give more because they infer that the cause is more deserving, whereas the latter implies that friends give more because they care more about the well-being of victims of that particular misfortune. Future research should investigate the extent to which the effect we document here is explained by changes in caring and/or in perceived suffering.

    Moreover, although we focused on sympathy because of its established link to prosocial behavior (Loewenstein and Small 2007), it is possible that other emotions, such as guilt, are induced by friendship with a victim and that these other feelings also play a role in charitable choices. Future research could more thoroughly measure and manipulate emotion to test these possibilities.

    Marketing Implications

    In addition to advancing the understanding of consumer behavior in this domain, these findings have important practical implications. As we explain below, charities can employ the insights of these findings in at least two ways to aid their fundraising efforts:(i) to increase the expected yield of a given donor and (ii) to identify high-yield donors.

    In terms of the first of these two, charities could strive to create personal relationships between victims and potential benefactors. Charities do often describe or show images of specific victims in their advertising campaigns, but such attempts seem designed to benefit from the identifiable victim effect (Kogut and Ritov 2005a, 2005b; Small et al. 2007) rather than to create friendship between donors and victims.

    Our findings suggest that creating a personal link with a victim could also be an effective intervention. If, as our results show, a relationship created in the lab in a few minutes can significantly increase giving, then surely a charity can inspire a connection between a victim and a benefactor through its solicitations. This may explain why charities often employ celebrities such as Michael J. Fox and Magic Johnson to successfully raise money. It is plausible that people feel that they are socially close to such celebrities and hence may exhibit more sympathy toward other victims of those misfortunes.

    Although charities do seem to be aware of the technique just described, they seem less aware, if at all, of the potential gains associated with targeting, rather than producing, friends of victims. The challenge lies, of course, in identifying these high-yield donors. We expect that events such as the Race for the Cure® are successful in great part because they have found one way of doing so. When victims participate in the fundraising efforts, they naturally tap their social networks (friends and families) for donations. Other charities could similarly activate a victim as a social node, initiating a special form of viral marketing that identifies and attracts high-yield donors.

    Our results confirm the necessity of conducting research on charitable giving to understand the ways that it is different from, not the same as, other forms of consumer behavior. Unlike many consumer decisions, charitable giving is motivated by sympathy, an emotion engendered (as we show) by personal relationships with victims. Charities can improve their fundraising efforts by nurturing and building on this unique dynamic. Through insight into the relationship between sympathy and giving, charities can better fulfill their unique missions to respond to human need and suffering.

    Appendix A Scenarios Used in Study 1

    Situation 1

    Susan is a college instructor and a poet. She is married to Sam and they are grandparents. Recently, she began drifting away from herself and from everyone around her. She suffers from Alzheimer's, which is a progressively degenerative form of dementia, involving loss of memory and language skills. Her first symptoms were forgetting simple things, such as what she did with the shopping list, losing her temper easily. For a while, she can fool herself and others. Over time, she loses her grip and fails to recognize her closest loved ones.

    Situation 2

    Sarah works as secretary for a law firm in center-city. She was 37 years old and the mother of two children, Jerry and Samantha, when she learned that she had breast cancer. After a radical mastectomy (i.e., removal of a breast), she found that it had spread to her lymph nodes, which is much more difficult to treat than breast cancer cases where the cancer is contained. She is starting an intense schedule of chemotherapy.

    Situation 3

    Joe lost his job at the post office right after the holiday season. He had worked there since graduating from high school 10 years ago. His confidence has been shaken, and he is under intense financial pressure since he has a mortgage and three young children. His wife is a substitute teacher and cares for their kids so they all depend on his income.

    Appendix B Relationship Closeness Induction Task Instructions and Questions Used in Studies 2 and 3

    In today's study, we're interested in how people get to know each other. For the next part of the experiment you will be doing a communication task which will help you get to know another student in the class.

    Here's how it works: you will both be given two lists of questions. We would like you to engage in as natural a conversation as possible using these questions. There is a time limit on each of the two lists. Try to get to as many of the questions on the list as you can, but don't worry if you don't get to all of them. I will keep time and tell you when to go on to the second list.

    When this occurs, take a couple of seconds to finish what you were discussing and then move on to the second set of questions. You are not required to answer any questions or talk about anything that might make you uncomfortable. After you have completed this task you will take part in an allocation task.

    Communication Task

    List I

    1. What is your first name?

    2. Where are you from?

    3. What year are you at Penn?

    4. What are your hobbies?

    5. What would you like to do after graduating from Penn?

    List II

    1. If you could travel anywhere in the world, where would you go and why?

    2. What is one thing happening in your life that makes you stressed out?

    3. If you could have one wish granted, what would that be?

    4. What is one recent accomplishment that you are proud of?

    Appendix C “Colleague” Induction Questions Used in Study 3

    Communication Task

    List I

    1. What is your first name?

    2. What is your major?

    3. What days of the week do you have more open blocks during school time?

    4. Are you relatively better with Excel or with Powerpoint?

    5. Do you find meetings in casual spaces (e.g., coffee shops) to be more or less productive than those in more formal places (e.g., a study room)?

    List II

    1. What do you dislike most about working on projects with other people?

    2. What is your best skill that can be used in course projects?

    3. What skill are you least competent in?

    4. What kinds of group projects have you done really well on?


    View Abstract