Madeleine Gleave and Matthew Brigham, Dr. Daniel Nielson, Political Science
I. INTRODUCTION
Helping behavior in a given society provides an interesting measure of social trust and prosociality. Whether individuals will assist strangers in need should indicate baseline levels of empathy, perceived social obligation, and generalized trust. But to date few studies have explored experimental interventions that might affect the rate of helping in a developing-country context. In particular, we sought to explore the effects of social status on pro-sociality in a poor country; i.e., would individuals respond differently to those of higher status? To men or women? To foreigners? We explored these questions through a field experiment where we randomly assigned the status, nationality, and gender of requesters of assistance in urban Kampala, Uganda, when stopping 1,200 Ugandans on the street and asking for assistance.
II. METHODOLOGY
Joined by a trained team of Ugandans, we gathered data in areas throughout the city of Kampala, Uganda. On a given day, the predetermined area of research was chosen based on convenience; however, we randomly assigned each of the treatment conditions to three or four locations within each area. The location assignment determined which of the three main treatment conditions subjects would receive, as each researcher was dressed and scripted to signal the following socio-economic profile: student, NGO worker, or businessperson. Each treatment was also divided into gender and nationality subcategories. Local Ugandan citizens of both genders were hired to be researchers and recorders along with American researchers of both genders.
Upon arriving at a location, after two minutes a recorder stationed nearby (but hidden to subjects) gave a signal for the researcher to contact the closest person to the researcher at that moment. Following a strictly observed script, the researcher would then explain to the subject that they were a student, an NGO worker, or a businessperson, saying:
“I need to get to a meeting and my driver was supposed to come here and pick me up. I need to call him, but I am out of airtime. I would go buy some, but I need to stay here because he is coming. Can please you go buy 1,000 shillings of MTN airtime for me? Thank you.”
Upon hearing this request, the subject could then agree to do the favor, refuse to do the favor, or walk away. The subject could also agree but not return, and thus steal the money. If the subject returned and completed the task, the research would offer a nominal amount of money (500 shillings, or approximately $0.20) as compensation for the subject’s time. Upon the return of a subject, or upon refusal, the researcher would then record the outcome and wait for two minutes before contacting the next subject. We also recorded unsolicited additional outcomes, including refusal of compensation and attempts to network with the researcher (requesting contact information or job opportunities).
Researchers rotated through locations in each area, each day to increase comparability. Through our use of a full factorial design, we gathered data for each combination of treatment and subcategory, creating twelve possible combinations of researcher type. Thus, we are able to compare each condition and subcategory combination against all of the other condition and subcategory combinations in our analysis through pooling.
III. RESULTS
Our sample included 1,166 subjects drawn from locations in ten different areas of the urban center of Kampala, Uganda. Using the data gathered from the researcher-subject interactions, we tested several comparative hypotheses based on our theory of self-interested and altruistic motivators. We use a logistic regression model for each in order to add the subject gender as a covariate. We also use robust standard errors for each model to address heteroskedasticity issues.
As the table above shows, we found that Ugandans were most likely to help women, Americans, and higher-status individuals over Ugandans, men, and lower-status requesters. As supporting measures, we also find subjects network more with higher-status professionals and foreigners, We also tested for heterogeneous effects based on subject gender when interacted with the researcher gender. Our results follow our assumptions about social norms considering women’s vulnerability, thus cueing altruism: men are 9.4% more likely to help women than men. Reinforcing the chivalry role for men, men also help women more than women help other women. Rates of refusal of compensation also follow these patterns.
IV. DISCUSSION & CONCLUSION
We argue that variations in gender, nationality, and socio-economic status influence helping behavior by activating altruistic and/or self-interested motives within subjects. These are not necessarily mutually exclusive; they can work together or in opposite directions. Although the effects of altruism and self-interest on behavior are almost impossible to fully disentangle from each other, our theory parses out the expected independent effects of each factor on helping behavior. Social norms dictate women and potentially foreigners are more vulnerable, cueing altruism and more willingness to help; socio-economic disparities signal opportunities for selfinterest when helping professionals and foreigners, also increasing willingness to help.
These findings provide evidence that status, nationality, and gender cues influence helping behavior by activating intrinsic altruistic and/or self-interested motivations. Although this varies between groups of subjects, there is an average willingness to help that shifts substantively in response to our treatments. From these results we uncover relationships that begin to suggest answers to broader questions about generalized trust and social capital.