Debnam, Jakina. 2017. “Selection Effects and Heterogeneous Demand Responses to the Berkeley Soda Tax Vote.” American Journal of Agricultural Economics 99 (5): 1172–87. doi:10.1093/ajae/aax056.  (Abstract)(Author's Copy)(Web Appendix)

Working Papers
Identity During a Crisis:  COVID-19 and Ethnic Divisions in the United States 
with Marie Christelle Mabeu and Roland Pongou. (Working Paper)
During a crisis, does a community's ethnic composition influence policy efficiency? How do the effects of ethnic divisions differ from those of ethnic diversity? Despite the large body of work which considers ethnic composition, little attention has been given to how it matters for crisis-response policy. Using the lens of the COVID-19 pandemic in the United States, we show that ethnic divisions, rather than ethnic diversity, significantly reduce the efficacy of crisis response. U.S. counties with high levels of ethnic divisions fared worse than their less-divided counterparts after lockdowns in both COVID-19 cases and related deaths. Ethnic diversity had little effect, except in areas with high racial segregation. Crisis-response policies led to smaller mobility reductions and less mask-wearing in ethnically divided counties. These results are not driven by a lack of physical public goods, socioeconomic differences, or by the prevalence of high-risk populations. Findings are robust to various strategies of causal identification and falsification checks. Our results suggest that policies promoting ethnic and racial integration can allay the negative social and economic impacts of crises.

Endogenous Responses to Paternalism: Examining Psychological Reactance in the Lab and the Field",  
with David R. Just. (Working Paper)(Web Appendix)
By accounting for limited human computational ability, willpower, and rationality within economic models, work in behavioral economics has highlighted the ways in which individuals' choices may systematically deviate from their own best interest. As a result, policymakers have considered any number of paternalistic policies (both overt taxes and restrictions, or more subtle "nudges") to move individuals closer to optimal outcomes. Much work, however, remains to characterize optimal design within this new class of policy instruments and to understand their aggregate impact. We present a theoretical framework of individual response to paternalistic interventions which considers, in addition to the set of behavioral responses explicitly incentivized by the policy, an additional behavioral outcome - the agent's impulse to re-establish whatever perceived choice set he had before the intervention occurred. We refer to this behavioral outcome as psychological reactance, a concept introduced by Brehm (1966). In support of this framework, we first provide evidence on the nature and magnitude of reactance responses from a laboratory experiment designed to measure response to paternalistic advertisements. We then present evidence of consumption responses to paternalistic advertisement in and around New York City during the policy debate surrounding then Mayor Bloomberg's proposed restrictions on sugary drink consumption within city limits (popularly referred to as a "soda ban"). Our findings support the existence of real interaction effects of paternalistic public policies.

Peers and Persuasion Across Collegiate Social Networks
Using a unique set of text and network data from a social network, this paper measures the persuasiveness of peers’ communications among college undergraduates’ course selection at Cornell University. I use idiosyncratic shocks to students’ information sets to create an instrumental variable and find that while in general, the effect of receiving an additional piece of information about a course is a decrease in the likelihood that a student enrolls in the course, if the message-giver is a peer, the effect of this additional message is up to a 7.4% increase in the likelihood that a student enrolls in that course. This finding is consistent with theories of information aggregation where individuals ‘tag’ information with sources as they incorporate these sources into their final decisions. I support key assumptions using exponential random graph models and in-person survey data which I collected from 112 undergraduate students. To the best of my knowledge, this work is the first in economics to empirically investigate theories of social influence using non-experimental field data.

What Do Happiness Data Mean? Evidence from a Survey of the Respondents",
with Daniel J. Benjamin, Marc Fleurbaey, Ori Heffetz, and Miles Kimball
With a specially designed survey, we examine how respondents understand the meaning of subjective well-being (SWB) survey questions, including commonly used measures of life satisfaction and happiness. In particular, we study how respondents identify the time frame of the questions and the components of their life that fall within the scope of the questions. We also study how respondents come up with a number on a bounded scale for rating their own SWB, and we investigate the reference points and reference distributions to which they compare their own situation. We devote particular attention to heterogeneity of these various aspects across respondents. Our results have implications for interpreting responses to SWB questions; in particular, our results shed light on the extent to which responses are interpersonally comparable.

Works in Progress
Correcting Subjective Well-Being Measures for Cross-Sectional Difference in Scale Use,
with Daniel J. Benjamin , Marc Fleurbaey, Ori Heffetz , Miles S. Kimball 
Subjective well-being (SWB) measures are measured on numerical or verbal scales that may be interpreted differently by different respondents. This paper addresses how to correct SWB measures for cross-sectional differences in use of numerical scales when a numerical scale is also used for other questions for which cross-sectional differences in answers can be assumed to arise primarily (aside from i.i.d. differences) from differences in scale use. Regression results using scale-use-corrected SWB measures as the dependent variable are contrasted with results when regressing raw SWB measures on the same set of regressors.