Public goods in the field: Katrina evacuees in Houston.

1. Introduction

Crises and disasters, whether natural or man-made, are defined by conditions of uncertainty, disorder, and stress. This raises the question of whether social cooperation is sustainable in the aftermath of a disaster. In this research, we focus on cooperation with a public goods game played by individuals evacuated from New Orleans to Houston after Hurricane Katrina. We ask whether stress affects how individuals contribute to a public good in the aftermath of a natural disaster.

The study predominantly focuses on African-American evacuees from New Orleans who were relocated to Houston-area shelters in the weeks immediately after Hurricane Katrina. In this study, 352 evacuees participated in small groups across six different Houston evacuation shelters from September 10 through 19, 2005. The study consisted of several standard experiments and a questionnaire. The experiments reported here are adaptations of "dictator" and "public goods" experiments that measure cooperation among Katrina evacuees. We find strong evidence of group cooperation in the Houston area shelters. We also find an independent negative effect of stress on cooperation.

2. Motivation

On the morning of August 29, 2005, after the passage of Hurricane Katrina over the Gulf Coast of Louisiana, Mississippi, and Alabama, levees designed to protect the city of New Orleans from flood surges gave way and parts of New Orleans began to fill rapidly with water. Many of those who stayed behind sought shelter by moving to the Superdome (and later to the New Orleans Convention Center), although many remained trapped on rooftops, attics, and upper-level apartment and office buildings. Although there are many conflicting reports and viewpoints as to the handling of the crisis by the city of New Orleans, the state of Louisiana, and the Federal Emergency Management Agency, as well as regarding the behavior of the citizens of New Orleans in response to the flooding, there was consensus that the city had experienced a major disaster.

The conventional wisdom is that a natural disaster throws victims into a state of anarchy in which social conventions are abandoned and replaced by self-interested survival instincts. Certainly observers of the post-Katrina landfall were not surprised by reports of widespread looting. Media reports of gang behavior, rapes of children, and murders in both the Superdome and Convention Center were readily accepted. It turned out these reports were exaggerated or inaccurate.

Substantial research suggests that cooperative norms are commonplace in the aftermath of a natural disaster. In a synthesis of a multitude of studies, Drabek (1986, p. 179) notes that in the aftermath of a disaster there will be "heightened levels of internal solidarity." In effect, the in-group of victims becomes well defined, and group identification becomes highly salient. This is understandable in that everyone is in the same position and they have shared experiences. This leads us to expect high levels of group cooperation in response to a disaster, especially for people who identify with one another as victims or survivors.

By contrast, many psychologists point to disasters as heightening stress for victims. It is well known that stress is part of an emotional state that leads to a decrement in the quality or speed of performance in most tasks. For example, Lazarus (1999) argues that stress should be considered a subset of emotions, with important physiological and cognitive effects. Much of what is of interest are the coping strategies of humans to stress. Janis (1993) contends that decision making under stress results in "hypervigilance" to some stimuli and disregard of others. This leads to "stereotyped thinking in terms of oversimplified categories and reliance on simpleminded decision rules" (p. 65). Weisaeth (1993) points to "post-traumatic stress disorder" as common following a natural disaster, although the long-term effects for individuals are highly variable.

Evidence for cooperative or noncooperative behavior after a natural disaster is practically nonexistent. Although there are numerous attitudinal studies and anecdotes about behavior, to our knowledge no one has used measurement tools taken from experimental economics and applied them to the aftermath of a disaster. Those working in the field of disasters and hazards agree with Bolin's (1989, p. 71) point that there has been little study of "sheltering populations evacuating a disaster site...." It is well known that victims of lower socioeconomic status are more likely to use such shelters. Staying in a large shelter is also known to be a stressor if people are removed from their family or witness interpersonal violence (Bolin 1989, p. 71). Beyond this, there is little understanding of the behavioral strategies that evacuees from a large-scale disaster might employ.

We use a behavioral experiment to shed light on responses to crisis and disaster, and we raise the question of whether cooperation, as measured by a public goods game, exists among very poor people who were involuntarily evacuated from a natural disaster.

3. Subject Recruitment and Sample Characteristics

We conducted these experiments "in the field" under unusual conditions. We selected subjects recuperating from highly traumatic conditions while housed in evacuation shelters. In the early stages of the Hurricane Katrina evacuation, Houston, Texas, was the primary evacuation point for people from New Orleans and greater Louisiana. Within the course of a week, the evacuee population in the Houston area swelled to an estimated 200,000 people. Many of those people evacuated before the landfall of Hurricane Katrina. However, a significant group of people stranded in New Orleans were bused to Houston and were then temporarily housed in the Houston Astrodome, the neighboring Reliant Center, and the George R. Brown Convention Center in downtown Houston. As those shelters overflowed, other small and mid-sized shelters were opened by a wide range of organizations, including the Salvation Army, the American Red Cross, and church and community organizations. Our focus is on the individuals who were stranded in New Orleans, were bused to Houston, and were housed in the shelters.

The shelters varied in terms of size, living conditions, and security. In the larger shelters, evacuees often slept in large arena-sized areas, whereas in some smaller shelters, they were accommodated in spaces ranging from indoor gymnasiums to small school rooms and even semiprivate "family" rooms in some church community buildings. Living conditions within the shelters varied. Both large and small shelters were cramped and noisy, and many evacuees showed signs of both physical exhaustion and emotional fatigue. Security was heavy within the large mass-accommodation shelters of the Astrodome, Reliant Center, and the Convention Center. Evacuees were given picture IDs and were required to wear them. Relief workers were also required to wear identification badges, and National Guard members and other security officers screened everyone coming into the shelters. Security in the smaller shelters varied considerably, although none were as restrictive as in the mass shelters. These factors played an important role in recruiting subjects and the conditions under which we conducted the experiments.

Our sampling began from the City of Houston's list of all recognized evacuation shelters within the greater Houston area. We selected locations to conduct the study on the basis of size, accessibility, and feasibility of the location for conducting the experiments. Generally, we required a space large enough to accommodate 20 to 25 people at a time. In each shelter that we visited, we obtained prior permission from the shelter administrators, local government officials, or both to conduct our study. In no case were we turned down by a shelter official to conduct the study.

The first author, with the help of an assistant, recruited all subjects for this study. Participants should be regarded as a convenience sample. A systematic random sample of evacuees was deemed impossible given the diversity of conditions and size of each shelter. We often recruited evacuees from common areas within the shelters such as dining halls and TV rooms, but also in the general sleeping quarters. We were concerned that having family members in the same session might bias our results in favor of cooperation; therefore, we tried to restrict participation to only one family member per session. When individuals agreed to participate, the assistant took them to the room or area where we were conducting the study. The size of each group session varied depending on the facilities available to us. If the room was small, we stopped recruiting after 16-18 subjects.

In most cases, we used a room or private area in the shelter to conduct the study. In the best circumstances, we were given a room with chairs and tables and closed doors. In other cases, we ran the experiments in common areas, which were more prone to noise and distractions. Overall, however, distractions were relatively minor once we began the session and usually were caused by small children, a ringing cell phone, or the occasional late-comer wanting to participate. However, general research conditions were satisfactory given the difficulty of conducting experiments in the field and given the overall conditions within the shelters that we visited.

Overview of the Sample