Research

Paper available here. [Appendix] [Replication Materials]

Abstract. We link over- and under-confidence in math at ages 8-11 to education and employment outcomes 22 years later among the children of PSID households. About twenty percent of children have markedly biased beliefs about their math ability, and beliefs are strongly gendered. Conditional on measured ability, childhood over- and under-confidence predict adolescent test scores, high school and college graduation, majoring or working in STEM, earnings, and unemployment. Across all metrics, higher confidence predicts better outcomes. These biased beliefs persist into adulthood and could continue to affect outcomes as respondents age, since intermediate outcomes do not fully explain these long-run correlations.

Paper available here. [Appendix]

Abstract. Governments often rely on schools to implement multiple programs targeting child outcomes. How to improve the implementation of these programs is an important, open question. As part of a randomized controlled trial in Odisha, India, we measured the impacts of a nutrition program and a monitoring intervention on the implementation of a pre-existing school-based nutrition program, specifically the Indian government’s iron and folic acid supplementation (IFA) program. The new nutrition intervention distributed a micronutrient mix (MNM) to be added to school meals while the monitoring intervention varied the intensity of monitoring activities. We find that high intensity monitoring improved implementation of the government’s IFA program, while the MNM intervention crowded it out. The net effect is that high intensity monitoring improved child health, while the MNM intervention did not. Both crowd-out of the IFA program and sensitivity to monitoring were predominantly found among schools that were resource or capacity constrained.

Paper available here. [Appendix] 

Abstract. Reducing the rate of anemia is a primary public health concern in many developing countries. This paper studies the Indian government's school-based Iron and Folic Acid (IFA) supplementation program. We provide a descriptive analysis of program implementation patterns across 377 schools in the state of Odisha. We document that the more advantaged blocks had more consistent tablet distribution, and that distribution within the less advantaged blocks was less predictable, and plausibly random. We use this quasi-random variation to estimate the causal effect of the policy using a difference-in-differences strategy. The IFA program had no effect on hemoglobin levels at the mean, but there is a significantly larger effect for moderately anemic students in schools that were more recently distributing tablets, relative to schools that had run out of tablets. These results suggest that school-based supplementation has the potential to improve hemoglobin levels, but that breaks in supplementation – either due to inconsistent tablet distribution or the constraints of a school calendar – limit the long-term efficacy of school-based supplementation programs.

Draft available here as SSRN working paper. 

Abstract. Minority-group workers in the US understand that discrimination is widespread, and when they experience rejection at work, observe incomplete information about the cause. Accurate or not, perceived discrimination may generate distrust, retaliation, or anticipated further discrimination. It is therefore important to assess how workers form beliefs about experiencing discrimination and how those beliefs affect job performance and labor supply. In an online experiment (N=5,000), I randomly assign workers to be evaluated by promotion procedures with varied potential to discriminate and the information workers have about past promotions. Learning that a manager knew workers' race and gender and previously promoted mostly white men increases the share of workers who perceive discrimination from 3 to 33 percent, lowers retention by 3-6 percent, and increases reservation wages by 9 percent. Thus, employers could improve equity and efficiency by reducing perceived discrimination. I find, however, that increasingly-common anti-bias hiring policies—blinding managers to demographics or using unbiased algorithms—are unlikely to alone eradicate perceptions of discrimination when minority groups remain under-represented.

Draft available here as SSRN working paper.

Abstract. Can fictional narratives contribute to building political momentum? In an online experiment (N=6,000), learning about the Inflation Reduction Act (IRA) strengthens beliefs about government responsiveness to citizen action by only 0.07sd. Watching a short, fictional story about political climate advocacy as a loose backstory to the IRA yields much larger effects on beliefs (0.5sd). While IRA information alone does not affect climate advocacy, the story increases information-gathering about climate marches by 54 percent and donations to lobbying organizations by 19 percent. We show evidence that beliefs and emotions may drive this effect.

Draft available here.

Abstract. Political action spreads through social networks, so citizens may have power to shape policy both through their own advocacy and by recruiting others to act. Do citizens try to spread grassroots action? If so, do they work to build broad, bipartisan coalitions or to recruit others like them? We focus on the climate movement, where most citizen advocates are Democrats. Mobilizing bipartisan action could more effectively promote climate policy in Congress, but record-high affective polarization—animosity towards counter-partisans—may impede cross-party grassroots cooperation. In online experiments with 25,000 participants, we connect Democrats with other Americans across the political spectrum (all of whom believe climate change is human-caused) to understand whether and how they try to recruit others to push for climate policy. Democrats are motivated to recruit others—they are 10% more likely to email Congress when doing so allows them to invite others to act. Even while Democrats say that a bipartisan climate movement would be more effective, however, they are 27% more likely to invite other liberals than conservatives to email Congress. This gap does not arise from Democrats’ own distaste for engaging with counter-partisans, but rather can be explained by their correct beliefs that their invitation will have about half as much impact on conservatives’ action. Anticipated affective polarization drives these beliefs: Democrats estimate that conservatives would respond three times more to invitations that did not identify them as liberals.