“Serenity Now, Save Later: Evidence on Retirement Savings Puzzles from a 401(k) Field Experiment” (with Saurabh Bhargava)

Economists have identified several psychological frictions that could explain why many 401(k)-eligible employees undersave for retirement despite generous matching incentives. We investigate four of these frictions through a field experiment that randomized 1,137 low-saving employees at a large US firm to information- and incentive-based treatments at the end of a survey assessing each friction’s baseline prevalence. We describe four findings: (1) We corroborate previous research by showing that low retirement literacy is pervasive and correlated with saving, but present causal evidence that reducing such deficits (by providing specific recommendations) does not affect saving. (2) In an analysis of plan confusion, we estimate that one-fifth to one-third of 401(k) non-participants mistakenly believed themselves to be enrolled, and these employees enrolled at high rates upon discovering their actual status. (3) We find no evidence that enrollment complexity impedes saving: few employees perceived enrollment as prohibitively time-consuming and simplifying enrollment further did not increase saving. (4) Finally, we present new evidence directly implicating present focus as a cause of undersaving by showing that a significant share of employees increased saving in response to an immediate small reward ($10 gift card) but not to a treatment clarifying the dramatically larger, but delayed, plan match. Calibrations suggest that beta-delta models of present-biased preferences cannot plausibly account for the documented behavior and stated beliefs of employees. We propose an alternative hedonic model of present focus that does explain our findings—and possibly other retirement savings puzzles—and offers a psychological rationale for encouraging long-run savings by linking existing 401(k) accounts to a more liquid account designed to relieve near-term financial anxiety.

“Save(d) by Design” (with Saurabh Bhargava, Rick Mason, and Shlomo Benartzi).  SSRN Working Paper, October 2018. Featured in Harvard Business Review, Feb. 2020: How digital design drives user behavior.

Online 401(k) enrollment interface design varies extensively along non-economic dimensions such as how options are presented and plan information is displayed. Yet there is little evidence on how these factors affect behavior. We describe field experiments showing that randomized design variation can be very influential, with one design increasing average contributions among over 8,000 employees from 501 different firms by a magnitude equivalent to that predicted by increasing matching incentives by over 60% of the typical match limit. This design also made decisions more responsive to cross-plan variation in match incentives, highlighting complementarities between design and incentives.


“The heterogeneous effect of affirmative action on performance.” 2019. (with Anat Bracha and Alma Cohen). Journal of Economic Behavior & Organization, 158, 173-218.

This paper experimentally investigates the effect of gender-based affirmative action (AA) on performance, focusing on a tournament environment. The tournament is based on GRE math questions commonly used in graduate school admission, and at which women are known to perform worse on average than men. We find that the effect of AA on female participants is heterogeneous: AA lowers the performance of high-ability women and increases the performance of low-ability women. Our results are consistent with two possible mechanisms: One is that AA changes incentives differentially for low- and high-ability women, and the second is that AA triggers stereotype threat.

“The slider task: An example of restricted inference on incentive effects.” 2016. (with Felipe Augusto de Araujo, Erin Carbone, Marli Dunietz, Ania Jaroscewicz, Rachel Landsman, Diego Lam, Lise Vesterlund, Stephanie Wang & Alistair Wilson). Journal of the Economic Science Association, 2(1), 1-12.

Real-effort experiments are frequently used when examining a response to incentives. For a real-effort task to be well suited for this exercise its measurable output must be sufficiently elastic over the incentives considered. The popular slider task in Gill and Prowse (Am Econ Rev 102(1):469–503, 2012) has been characterized as satisfying this requirement, and is increasingly used to investigate responses to incentives. However, a between-subject examination of the slider task’s response to incentives has not been conducted. We provide such an examination with three different piece-rate incentives: half a cent, two cents, and eight cents per slider completed. We find only a small increase in performance: despite a 1500% increase in the incentives, output only increases by 5 %. With such an inelastic response we caution that for typical experimental sample sizes and incentives the slider task is unlikely to demonstrate a meaningful and statistically significant performance response.

“Predicting health behaviors with economic preferences and locus of control.” 2015. (with Julian Jamison). Journal of Behavioral and Experimental Economics, 54: 1-9.

We present new results on the relationship between health behaviors and experimental measures of time and risk preferences. In contrast to recent findings in the economics literature, we find no evidence linking time preference and self-reported health behaviors and outcomes such as smoking and BMI. We also introduce evidence that internal locus of control—a psychological construct that refers to the tendency to attribute to oneself control over outcomes—explains significant variation in health behaviors, in models including traditional measures of risk and time preference

Current Projects

“The Psychological Costs of Job Search” (with George Loewenstein)

Conventional economic models of job search struggle to explain several behavioral patterns observed among unemployed job seekers such as limited time spent on search. We propose a novel theoretical framework to explain search intensity decisions that incorporates psychological costs associated with time spent on search during unemployment. The model provides a new perspective on why job seekers spend little time actively searching, and why active search declines over the course of unemployment. It also produces novel predictions on search intensity across different search activities and relationships between search behavior and emotions. We test predictions of the model in a series of surveys with several hundred unemployed job seekers and find support for predictions on dynamics of psychological costs over the course of unemployment spells and relationship to time spent on search and subjective judgments of sufficiency of search time.

“The Behavioral Economics of Job Search: A Field Experiment Testing for Suboptimality of Search Volume & Strategies” (with Saurabh Bhargava)

This project aims to contribute to economists’ understanding of job search behavior by testing basic assumptions about the optimality of search effort and strategy and the assumed irrelevance of psychological factors such as discouragement. One reason that behavioral insights have been slow to affect labor economics is the difficulty of capturing labor market dynamics realistically in the laboratory. This research will address this gap by leveraging access to real-world data on the online job search behavior of several hundred thousand job seekers and a field experiment with unemployed job seekers. These will provide new evidence on (i) how individuals search, including effort and search wages, (ii) the optimality of such search relative to a benchmark of the standard model, and (iii) the psychological consequences of unemployment and their effect on search. Further, it will test psychologically informed interventions intended to improve job outcomes. Evidence on whether unemployed job seekers search suboptimally for work and, if so, what specific interventions can improve search will have implications for how economic theory should model job search decisions and unemployment.

“Anger and Polarization in Policy Preferences” (with Lea Heursen)

We test whether experimentally activating existing anger at the government causes people to support more extremist policies and political candidates. We define extremism in terms of distance of preferred policy outcomes from the status quo. We randomized
2000 U.S. voting-eligible adults to a condition in which anger at the government was activated with a thought-listing prompt, or to a condition in which participants listed their thoughts on a neutral topic. A manipulation check confirms that the treatment induces anger in the predicted direction. We find that this minimal prompt activating existing anger at the government leads to an increase in expressed extremism across different policy areas. Interventions designed to give participants the possibility to vent their anger by communicating to their congressional representatives or to other participants, failed to attenuate the impact of anger on policy preferences. This result suggests the possibility that even temporarily heightened anger may drive changes in policy preferences expressed by the electorate.