al fang

cv      google scholar

I am a Behavioral Scientist at Zillow Group.

My research is in behavioral and psychological science with a focus on applications in markets, politics, and public policy. I also specialize in the design and analysis of experiments, causal inference, machine learning, applied statistics, and survey methodology.

Prior to Zillow, I was a postdoctoral associate at the Yale Center for the Study of American Politics. I received my Ph.D. in political science and a master's in quantitative methods from Columbia University, and I completed my undergraduate studies at the University of California, Berkeley. I also worked in policy evaluation analyzing the impact of large-scale policy interventions for the federal government and philanthropic foundations.

This page contains my academic research and writing.


  • Perceptions of Deservingness and the Politicization of Social Insurance: Evidence from Disability Insurance in the United States
    with Gregory Huber. Accepted at American Politics Research, 2019.

    local pdf | online appendix

    Concerns about the deservingness of policy beneficiaries appear to explain skepticism about redistributive social assistance programs. Many social insurance programs, despite requiring beneficiaries to pay in ahead of time, require discretionary evaluations of the merits of claims for benefits. Do perceptions of deservingness also affect attitudes toward these discretionary social insurance programs? Examining the politics of Social Security Disability Insurance (SSDI), a program whose size and beneficiaries have been increasingly politicized, we investigate these questions by analyzing novel survey data and two experiments conducted on national surveys. We show that people use information about a beneficiary's eligibility-determining impairment -- but not their race, which prior work argues is a key heuristic -- to infer their deservingness. Moreover, support for SSDI is responsive to policy arguments emphasizing the program's social insurance features and potential abuse. Our findings demonstrate important psychological processes relevant to the contemporary politicization of social insurance programs involving discretionary eligibility rules.
  • Can the Government Deter Discrimination? Evidence from a Randomized Intervention in New York City
    with Andrew Guess and Macartan Humphreys. Journal of Politics, 2019.

    online appendix | replication files | replication output

    Racial discrimination persists despite established antidiscrimination laws. A common government strategy to deter discrimination is to publicize the law and communicate potential penalties for violations. We study this strategy by coupling an audit experiment with a randomized intervention involving nearly 700 landlords in New York City and report the first causal estimates of the effect on rental discrimination against blacks and Hispanics of a targeted government messaging campaign. We uncover discrimination levels higher than prior estimates indicate, especially against Hispanics, who are approximately 6 percentage points less likely to receive callbacks and offers than whites. We find suggestive evidence that government messaging can reduce discrimination against Hispanics but not against blacks. The findings confirm discrimination's persistence and suggest that government messaging can address it in some settings, but more work is needed to understand the conditions under which such appeals are most effective.
  • The Comparative Effectiveness on Turnout of Positively versus Negatively Framed Descriptive Norms in Mobilization Campaigns
    with Alan Gerber, Gregory Huber, and Catlan Reardon. American Politics Research, 2018.

    local pdf | online appendix

    Are mobilization appeals that include information about descriptive voting norms more effective at increasing turnout when the descriptive norm is framed positively (by highlighting a referent's desirable behavior and encouraging consistent behavior) instead of negatively (by highlighting a referent's lack of desirable behavior as problematic but also encouraging that behavior)? Few published studies have experimentally assessed this question and yield mixed results. We address the need for additional replication by designing and analyzing data from two field experiments conducted across four states in the 2014 primary and general elections. We find no differential effects on turnout of framing descriptive voting norms positively or negatively. The results are not sensitive to election context, the mode of treatment delivery, or whether the descriptive norm appeals involve a group or self referent. Additional research is needed to understand the conditions under which positive versus negative descriptive norm framing has distinguishable effects on turnout.
  • Do Subtle Linguistic Interventions Priming a Social Identity as a Voter Have Outsized Effects on Voter Turnout? Evidence from a New Replication Experiment
    with Alan Gerber and Gregory Huber. Political Psychology, 2018.

    local pdf | online appendix

    An ongoing debate in political psychology is about whether small wording differences have outsized behavioral effects. A leading example is whether subtle linguistic cues embedded in voter mobilization messages dramatically increase turnout. An initial study analyzing two small-scale field experiments argued that describing someone as a voter (noun) instead of one who votes (verb) increases turnout rates 11 to 14 points because the noun activates a person's social identity as a voter. A subsequent study analyzing a large-scale field experiment challenged this claim and found no effect. But questions about the initial claim's domain of applicability persist. The subsequent study may not have reproduced the conditions necessary for the psychological phenomenon to occur, specifically the electoral contexts were not competitive or important enough for the social identity to matter. To address the first of these critiques, as well as other potential explanations for different results between the first two studies, we conduct a large-scale replication field experiment. We find no evidence that this minor wording change increases turnout levels. This research provides new evidence that the strategy of invoking the self does not appear to consistently increase turnout and calls into question whether subtle linguistic cues have outsized behavioral effects.
  • Nongovernmental Campaign Communication Providing Ballot Secrecy Assurances Increases Turnout: Results from Two Large-Scale Experiments
    with Alan Gerber, Gregory Huber, and Andrew Gooch. Political Science Research and Methods, 2018.

    local pdf | online appendix

    Doubts about the integrity of ballot secrecy persist and depress political participation among the American public. Prior experiments have shown that official communications directly addressing these doubts increase turnout among recently registered voters who had not previously voted, but evaluations of similar messages sent by non-governmental campaigns have yielded only suggestive effects. We build on past research and analyze two large-scale field experiments where a private non-partisan non-profit group sought to increase turnout by emphasizing ballot secrecy assurances alongside a reminder to vote in a direct mail voter mobilization campaign during the 2014 midterm election. Our main finding is that a private group's mailing increases turnout by about 1 percentage point among recently registered nonvoters. This finding is precisely estimated and robust across state political contexts, but is not statistically distinguishable from the effect of a standard voter mobilization appeal. Implications and directions for future research are discussed.
  • The Effect on Turnout of Campaign Mobilization Messages Addressing Ballot Secrecy Concerns: A Replication Experiment
    with Alan Gerber, Gregory Huber, and Catlan Reardon. PLoS ONE, 2017.

    local pdf | online appendix

    Given the persistence of public doubts about the integrity of ballot secrecy, which depress turnout, two prior experiments have shown precise evidence that both official governmental and unofficial mobilization campaigns providing assurances about ballot secrecy increases turnout among recently registered nonvoters. To assess whether these findings replicate in other political settings, we describe a replication experiment where a non-governmental, nonpartisan mobilization campaign sent similar treatment mailings to recently registered nonvoters during the 2014 general election in Mississippi. We find that sending mobilization mailings containing assurances about ballot secrecy protections has no effect on turnout rates in this setting, which is characterized by an unusually low baseline turnout rate. These results are consistent with past research concluding that nonpartisan GOTV mail has very weak effects among very low turnout propensity registrants, and suggest that there are heterogeneous effects of ballot secrecy treatments associated with subjects' characteristics and the electoral context.
  • The Generalizability of Social Pressure Effects on Turnout Across High-Salience Electoral Contexts: Field Experimental Evidence from 1.96 Million Citizens in 17 States
    with Alan Gerber, Gregory Huber, and Andrew Gooch. American Politics Research, 2017.

    local pdf | online appendix

    Prior experiments show that campaign communications revealing subjects past turnout and applying social pressure to vote (the "Self" treatment) increase turnout. However, nearly all existing studies are conducted in low salience elections, raising concerns that published findings are not generalizable and are an artifact of sample selection and publication bias. Addressing the need for further replication in high salience elections, we analyze a field experiment involving 1.96 million subjects where a nonpartisan campaign randomly sent Self treatment mailers, containing a subjects vote history and a comparison of each subjects history to their state median registrants turnout behavior, in high salience elections across 17 states in 2014. Sending the Self mailer increases turnout by 0.7 points, or 2.2%. This effect is largely consistent across states, with somewhat larger effects observed in states with lower ex ante election salience. Our study provides precise evidence that social pressure effects on turnout are generalizable.

other writing