Is more information always better? An experimental study of charitable giving and Hurricane Katrina.

1. Introduction

The devastation resulting from Hurricane Katrina has elicited unprecedented levels of charitable giving on the part of the general public. In just the first 11 weeks following Katrina, private charities donated approximately $2.7 billion, and $62 billion was appropriated by Congress (Frank 2005). Though motivations for giving to charitable organizations are varied, there are at least two reasons that people may increase donations to charitable organizations when a disaster occurs. First, the disaster may change their perceptions of the likelihood and cost of a disaster. Second, donors may be more sympathetic to the plight of others hurt by the disaster and so may increase donations.

When a disaster occurs, images of individuals' suffering are broadcast widely. The pain and suffering that these victims are feeling enters the everyday lives of observers all over the country, even all over the world. Contributions then flow into charities involved in relief efforts. Why does the immediacy of victims' plight increase giving? Research shows that donors are often much more willing to give to a specific, identified victim of a particular event than to disaster relief in the abstract or to efforts to prevent or lessen the potential damage from disasters. Schelling (1968) identified this phenomenon as the "identifiable victim" effect. (1) When a potential problem turns into a real one, its victims are transformed from "statistical victims," probabilities of injury and death, to real live victims. When an event actually occurs and a particular person or group is hurt, this evokes greater sympathy for the victims, and thus greater giving. Schelling phrases it this way: "The more we know, the more we care."

Jenni and Loewenstein (1997) and Small and Loewenstein (2003) focus on the psychological mechanisms that may contribute to the identifiable victim effect. They find evidence that the immediacy and salience of a real victim, which seem intuitively to be responsible for the change in perceptions, are not the most important factors. Their research indicates that the most important factor may be the relative size of the group of victims who can be helped relative to the number of people at risk. In a disaster, the identified victims are their own reference group: 100% of them have been affected and can be helped by assistance.

Another possibility is that information about victims and their suffering, whether provided through the media or direct experience with victims of a disaster, may simply decrease social distance between the donor and the victim, which then causes other-regarding behavior (including charitable contributions) to increase (Bohnet and Frey 1999). In this study, we examine the impact of perceptions, attitudes, and information, as well as the endowment and price of giving, on donations to charity in a laboratory setting. Subjects complete a set of decisions involving the opportunity to donate to a charity that was active in disaster relief in conjunction with Hurricane Katrina. In addition, we collect data on disaster experience, sympathy for others, and perceptions of the likelihood and cost of various events (weather and non-weather disasters, accidents, etc.). We report analysis of the relationship between these psychological factors and actual giving behavior.

2. Design

The design of the study incorporates an experiment and a survey. The experiment consists of a set of experimental measures of altruism and charitable giving that have been successfully employed by the researchers to study charitable giving (see Eckel and Grossman 2003, 2006a, b). Because these measures involve real trade-offs between a subject's own earnings in the experiment and the amount sent to the charities, these measures are likely to be more accurate and informative than are survey-based measures of altruism and charitable donations, where little is at stake. The survey component of the study was completed after the experiment and involves collection of information on demographics, other giving behavior, previous disaster experience, sympathy, risk perceptions, and perceptions of Katrina victims.

Experimental Measure of Altruism

We measure preferences for giving using a modified dictator allocation task (see Eckel and Grossman 2003, 2006a, b). For each decision, subjects are provided with an endowment and are offered the opportunity to donate any part of their endowment (in private and anonymously) to a given charitable organization. We use the strategy method: Subjects make a set of allocation decisions with different parameters, one of which is chosen for payment. The set of decisions varies the target charitable organization, the endowment, and the extent to which contributions are subsidized by the experimenter.

Three different charities the American Red Cross (ARC), the Salvation Army (SA), and Oxfam International (Oxfam)--were selected because all were active in providing aid to Hurricane Katrina victims and because they are likely to vary in terms of how they are perceived by subjects. The ARC is a prominent national and international disaster relief agency and is familiar as such to participating subjects. The SA is another large charity that helped with Hurricane Katrina victims, but one that is likely to be perceived as "more local" in scope and religious in orientation. Oxfam also provides disaster relief, but since its role is less focused on immediate aid to victims and more focused on long-term rebuilding, it is likely to be less familiar to the subjects and is therefore less likely to be associated with stranded, helpless Hurricane Katrina victims.

The experimental design has subjects make 12 separate contribution decisions for each of the three charities. There are four decisions for each of three endowment levels, $10, $20, and $50. The four decisions vary the subsidy level, with matching rates of 0%, 25%, 50%, and 100%. This gives a total of 36 decisions for each subject. Instructions and forms are available upon request.

Survey

The survey consists of several components. In order to look at the empathy and distress levels of our subjects, we employ the Interpersonal Reactivity Scale (Davis 1994). For an instrument to measure risk perceptions regarding catastrophic events, we adapted questions from Halpern-Felsher et al. (2001) and Kunreuther (1996) and added similar items of our own. In addition, we developed a set of questions to elicit the perceptions that our subjects had of the Katrina victims. The survey also collects information on standard demographics, religion, charitable giving, and experience with natural disasters (either direct or indirect).

Locations

Two locations were chosen for the study, one of which was more affected than the other by Katrina. While every part of the country has in some way been influenced by the hurricane, the magnitude of the impact varies. Texas has been host to nearly 135,000 displaced persons, with all of the attending financial costs and disruption. In addition, as a Gulf Coast state, Texas is at risk for future hurricane devastation, and its residents have experienced such catastrophes in the past. Minnesota, on the other hand, experienced relatively little immediate impact from Katrina (for example, only 1444 victims applied for assistance from that state). (2) However, the two states have in common the risk from other smaller and more localized natural disasters, such as flooding and tornadoes.

Procedures

A total of 10 sessions were conducted, six at the University of Texas at Dallas (UTD) and four at Saint Cloud State University (SCSU), with a total of 265 student participants. Nine sessions were conducted in undergraduate classes in economics and social science, and the 10th session (with eight subjects) was conducted in a master's-level introductory statistics class. Sessions were run in May and November of 2006, about 8 and 15 months following Hurricane Katrina, respectively. Subjects were recruited from, and the sessions were run in, classrooms. (3) Participation was voluntary, and the subjects' professors were not present during the experiment.

The researchers entered the classroom and introduced themselves and their assistants. Subjects were told that participation was voluntary and that six subjects would be chosen at random and paid in cash. Students in the classroom were then given the opportunity to leave: On average about one third of students at UTD and one fourth of students at SCSU opted out. We then distributed consent forms, after which subjects were again given an opportunity to opt out.

At the start of the experiment, we asked for a volunteer to be the monitor for the group. The monitor was paid $20 and was responsible for randomly selecting the subjects for payment, verifying the payment to the subjects, and verifying the payment to the charities. We then distributed and read the instructions aloud. The instruction phase included sample allocation problems (one for each charity) and a quiz to test for understanding of the matching procedure, as well as an explanation of the payment procedure. After the experimenter and assistants verified that everyone understood the task, identification numbers (used to maintain anonymity and for payment), decision sheets, and envelopes were distributed.