Research Seminar: Micro-, Macro- and Meso-Level Influences on Generosity in America

Several alarming statistics about a potential decline in American generosity were released during the first months of 2023. If the latest data, collected during the COVID-19 pandemic period, suggest that generosity in America has declined, what can the recent past teach us about why we are where we are today? The Do Good Institute is working on two separate reports for the Generosity Commission, which was formed in 2020 to study the apparent decline in generosity, that address this question. The first report, which will be published in October, attempts to measure the individual-level and community-level influences on giving money and volunteering time - two of the most basic and common forms of generosity - in the United States. After estimating multilevel models of both behaviors, the study finds that individual-level (micro-level) factors have much more influence than state-level (macro-level) factors.

This presentation will feature results from the second report, which investigates the degree to which social connectedness influences giving and volunteering. These measures of social connectedness can be considered meso-level variables, which measure the influences of groups,organizations, social networks or other institutions into the analysis. One of the primary goals for this report is to develop models that allow us to measure the influence of one type of charitable behavior on the likelihood of acting charitably in other ways. The study starts by analyzing the relationship between giving and volunteering. Most scholars agree that these activities are related, but economists have tended to view them as substitutes: if you volunteer, you are less likely to give and vice versa. However, empirically, people who give are also more likely to volunteer – that is, that giving and volunteering are complementary behaviors.

To learn more about this relationship, we need to confront the problem of endogeneity: volunteering affects giving and vice versa. Finding instruments is a real challenge, since all the micro-level variables belong in both models, but the macro-level variables in our multilevel models offer some possibilities for variables to exclude - and the problem of overidentification is hard to solve. This presentation will show what we’ve done to overcome this problem and how our proposed solution can be applied to models where meso-level variables are added to the models.