Kim Sarnoff

Abstract: Many real-world inference problems unfold over time: employers learn about ability across tasks, consumers evaluate products through repeated use, and policymakers revise beliefs as new data arrive. Yet despite its ubiquity, research on dynamic updating has largely focused on a single implication of Bayesian reasoning: order independence. This paper experimentally tests a broader set of restrictions implied by Bayes' rule, emphasizing both order independence and the previously unexamined property of prior sufficiency: the principle that the most recent posterior should serve as a sufficient statistic for past information. In a multi-period updating experiment with a rich set of parameters, participants repeatedly revise beliefs after receiving signals of varying strength and structure. Three main results emerge. First, only roughly a third display order dependence, overreacting to conflicting signals. Second, violations of prior sufficiency are widespread: beliefs formed sequentially tend to grow more extreme, and models assuming prior sufficiency, such as Grether (1980), fit poorly beyond the first update. Finally, the data indicate that participants process signals in aggregate, explaining prior sufficiency violations.

Abstract: Commerce requires trust, but trust is difficult when one group can expropriate another due to differences in power. This can lead the weaker group to self-segregate into industries and activities; female-led businesses, for example, tend to be small and clustered in a small number of industries where collaborators are also female. We present a model which relates this economic segregation to rule of law, and predicts that female trust depends on the protective preferences of adjudicators in weak rule of law environments. We then show that effective dispute negotiation in Lusaka, Zambia, especially as administered by "market chiefs," enables trusting behavior by female entrepreneurs, both in cross-section correlations and in two artefactual field experiments. Such trust generates increased economic returns. We find considerable heterogeneity across market chiefs in their preferences for protecting women, and that female entrepreneurs are more likely to want to reveal their gender with chiefs who are more likely to favor women. 

Abstract: We study how policies that disincentivize misconduct in organizations can generate changes in abusers' behaviors that negatively impact victims. We describe a model where "managers" choose to commit harmful actions of varying intensities against "employees," who can report these actions as "misconduct." We show that when the marginal disutility from managers' actions is particularly small, increasing the ease of reporting misconduct, the severity of punishment for managers, or the efficacy of investigation technology may in fact harm employees. These policies may motivate managers to commit harmful actions that employees do not want to report or induce managers to opt out of interacting with employees altogether. We provide a dynamic extension where reports generate precedents for the organization and employees, showing that the model converges to a steady state where employees are worse off than initially and harmful actions are never punished.

Abstract: In many environments, updating occurs in small increments over time and agents juggle belief revisions across a menu of options: for example, teachers assess multiple students over a long period of interaction. This paper uses an array of online experiments to explore how people respond to uni-dimensional and multi-dimensional information over various horizons. I develop an econometric approach for estimating the weights individuals place on their prior beliefs and on the signals they receive over time. I use this approach to document several patterns. First, participants underweigh the prior but overweigh signals. Second, the weight placed on signals changes over time. That time dependence leads to substantial sequencing effects and an apparent recency bias that becomes more pronounced over time. Third, receiving simultaneous information on multiple uncertain outcomes leads to important "grouping" effects: even when the initial priors over outcomes differ, posteriors converge over time.

Abstract: Over the past several decades, lab experiments have offered economists a rich source of evidence on incentivized behavior. In this article, we use detailed data on experimental papers to describe recent trends in the literature. We also discuss various experimentation platforms and new approaches to the design and analysis of the data they generate.