Working Papers

Pluralistic Ignorance and Group Size: A Model of Dynamic Impression Management

I develop a theory of group interaction in which individuals who act sequentially are "impression managers" concerned with signaling what they believe is the majority opinion. Equilibrium dynamics may result in most individuals acting according to what they mistakenly believe is the majority opinion, a situation known as "pluralistic ignorance". Strong beliefs over others' views increases pluralistic ignorance in small groups, but decreases it in large groups. Pluralistic ignorance may lead to a perverse failure of collective action, where individuals don't act selfishly because they incorrectly believe they are acting in a way that benefits others, which results in an inefficient drop in social interactions. A policymaker who wants to maximize social welfare will reveal information about the distribution of opinions in large groups, but will manipulate information to make the minority seems larger if the group is small. I critically review empirical evidence in light of the model and vice versa, and I develop tests for the novel theoretical predictions.


Lead if Others Expect It: Laboratory Evidence on What Motivates Pro-Social Influence (with Michael J. Hiscox)

Individuals act more pro-socially when they are in a position to lead others, and their behavior is motivated by social expectations -- they lead when they think others expect it. Social expectations have received little attention and often are a confound of altruistic motivations or reputational concerns. Our experiment is anonymous, so we abstract from reputational concerns. Our design measures (i) pro-social actions when an individual can and cannot influence others, (ii) revealed preferences over the decisions in (i), and (iii) beliefs and beliefs over beliefs over the averages of (i) and of (ii). We develop theory to show that teasing apart altruism and social expectations requires all these measures. We find little evidence for pure or impure altruistic leadership, while social expectations play a significant role. Social expectations therefore provide an underexplored but potentially fruitful avenue for increasing pro-social influence.

Social Identity As Hierarchies Of Selves: Theory and Experimental Evidence From Regional Contributions in Mexico

I develop a model that nests the two main approaches of how primes affect behavior with a  `hierarchies of selves' approach, which assumes that individuals want to signal which of possibly several norms drive their decisions. While a prime may nudge behavior towards a cognitively accessible norm, it will nudge it away from the norm if it confounds the signal that another norm drives behavior. This gives a novel account of how the impact of a prime is affected by situation-specific norms that are not primed. I asked subjects to rank their region and nation. Two weeks later, I randomly assigned them to a national prime and to situations with a national norm. I find support for the unique predictions of the hierarchies of selves approach: the impact of a prime depends on whether there is an un-primed regional norm. The results have implications for priming experiments designed to infer norms, problems of replicability and recommendations of when to use norms to affect behavior.



Work in Progress

Regional Identity and Nationalist Appeals: Evidence From A Large Scale Field Experiment (with Sam Asher) 

In an experiment where we varied the content of donation petitions across 560 supermarkets in nine states in Mexico, we found no evidence of regional bias in donations nor of an impact of a nationalist prime. These results differ from those commonly cited in the experimental literature on in-group bias and the impact of a common-group prime. However, a subsequent analysis shows that regionally poor municipalities responded more positively to a nationalist prime, while regionally rich municipalities had a regional bias that diminished with a nationalist prime. One plausible explanation is that a municipality's relative wealth proxies for regionalism. An online experiment from a companion paper provides supportive evidence for this hypothesis. The results suggest that the problems in replication of priming experiments may be due to unobserved heterogeneity, and highlight the need for more evidence from the field.


History Revision as Updating Social Expectations

I develop a model in which a principal may `revise history' by changing how followers interpret a group's current intentions based on past events. Revisions of history have been a political tool of persuasion used by historians, political leaders and popular mobilizations. In the model, Milosevic has private information about Croat's type-dependent preferences. Serbs form beliefs about Croat's intentions based on their past interactions. Milosevic may change Serbs' interpretation of Croat's past actions by  convincing them that many types of Croats would have acted in the same way. By changing how Serbs interpret history, Milosevic is able to change Serbs' behavior towards Croats that would otherwise not be justified.

Endogenous Interaction and Spillovers in Peer Groups (with Ben Golub and Evan Sadler)

Policy is often concerned with the optimal composition of peer groups -- e.g., for production or education -- in the presence of peer effects. There is evidence, however, that endogenous link formation decisions determine which peer effects operate: sub-sets of individuals choose whether to interact productively, in addition to making other choices (e.g., effort in learning). Their decisions depend on all their attributes, their prior relationships, and on what is going on in the rest of the group -- a set of endogenous decisions omitted from standard peer effects models. We propose a tractable model that incorporates endogenous link formation. To test the model, we have partnered with an NGO focused on early childhood development in rural Mexico. The experimental design varies the composition of playdate groups that meet during bi-weekly events organized by the NGO. The design proceeds in two stages. In the first stage, we estimate parameter values of our model based on group performance.The second stage features three arms: status quo, the standard model that ignores endogenous link formation, and our model. We hypothesize that explicitly incorporating endogenous linking decisions will significantly improve our outcomes of interest: caregiving practices and children's neurodevelopment.

Followership in Public Good Provision: The Role of Social Expectations. (with Michael J. Hiscox)

We provide experimental evidence that individuals follow pro-social actions in order to meet social expectations. While growing evidence suggests individuals respond positively to information about other's contribution to public goods, we don't fully understand the motivation that drives individuals to follow. We present evidence that distinguishes between different motivations. We abstract from material and social image motivations by making this decision anonymous and having no direct benefits - individuals decide how much to contribute to a charity. We're interested in distinguishing between two hypotheses. The first is that individuals are influenced for altruistic reasons -- they follow because they've learned that the public good is productive and want to contribute more. The second is that they do so reluctantly, out of a social expectation to do so -- they follow because they've learned that it is costlier to avoid disappointing others. We find  evidence for reluctance.