Robert Ainsworth

Publications

District competitiveness increases voter turnout: evidence from repeated redistricting in North Carolina

Quarterly Journal of Political Science, 2024, 19(4): 387-432
With Emanuel Garcia Munoz and Andres Munoz Gomez
QPJS article. Online appendix. Data and code archive.

We study whether competitive legislative districts cause higher voter turnout. To do so, we employ rich data on the 2006 to 2020 elections in North Carolina. We make use of variation in district competitiveness due to repeated bouts of redistricting, a process in which district boundaries are redrawn. Specifically, we compare people who share the same districts in each legislative chamber (U.S. House, NC Senate, NC House) before redistricting but who differ in districts after redistricting. We match these people on demographics, party registration, and pre-redistricting turnout. We then track their turnout behavior in post-redistricting elections. For the U.S. House, switching from an uncompetitive “80-20” district to a competitive “55-45” district increases turnout by a rate of 1 percentage point per election of exposure. For the state chambers, the magnitude is 0.6. Effects are highly persistent and sum across chambers. They appear to be explained in part by a learning channel, where living in a competitive district induces people to believe that races can be competitive.

Why do households leave school value added on the table? The roles of information and preferences

American Economic Review, 2023, 113(4): 1049-82
With Rajeev Dehejia, Cristian Pop-Eleches, and Miguel Urquiola
NBER Working Paper No. 28267. VoxEU summary. AER article. Data and code archive.

Romanian households could choose schools with 1 standard deviation worth of additional value added. Why do households leave value added “on the table”? We study two possibilities: (i) information and (ii) preferences for other school traits. In an experiment, we inform randomly selected households about schools’ value added. These households choose schools with up to 0.2 standard deviations of additional value added. We then estimate a discrete choice model and show that households have preferences for a variety of school traits. As a result, fully correcting households’ beliefs would eliminate at most a quarter of the value added that households leave unexploited.

Teaching global public health in the undergraduate liberal arts: a survey of 50 colleges

American Journal of Tropical Medicine and Hygiene, 2012, 87(1): 11-15
With David R. Hill and Uttara Partap

Working papers

District partisan slant shifts voters’ party affiliation: evidence from repeated redistricting in North Carolina

With Carlos Estrada and Emanuel Garcia Munoz. Slides

Does the slant of a legislative district merely reflect the district’s voters, or can it also affect them? To investigate this, we collect rich data on districts and voters from the 2006-2022 elections in North Carolina. We exploit variation in slant due to redistricting, the process in which district boundaries are redrawn. We show that living in a district where one party is powerful causes people to shift their party affiliation toward that party. Effects increase with the number of elections of exposure and sum across legislative chambers. They also seem to stem from changes in preferences, not strategic behavior. An implication of the results is that uncompetitive districts contribute to polarization: under uncompetitive districts, people and places who lean Democratic get put into Democratic-controlled districts and become more Democratic; vice versa for those who lean Republican. As an illustration, we assess the impacts of the districts that were used in North Carolina during the 2010s. Relative to a counterfactual of competitive districts, the 2010s districts led to a sizable increase in geographic polarization—though only a slight increase in polarization at the individual level.

Malleable minds: The effects of STEM v. humanities-focused curricula

With Rajeev Dehejia, Andrei Munteanu, Cristian Pop-Eleches, and Miguel Urquiola. Slides

We examine the impact of assignment to STEM v. humanities-focused curricula in Romania’s high school system. We apply a regression discontinuity design to administrative and survey data to estimate effects on educational pathways, career plans, and non-cognitive outcomes. An overarching theme of our findings is the malleability of students to what they study. Assignment to STEM increases STEM college enrollment by 23 pp and STEM career intentions by 17 pp, with corresponding negative effects on humanities-related enrollment and career plans. Exploring mechanisms, we find that STEM assignment changes students’ self-perceived academic abilities and their preferences over academic subjects and job tasks. We find less evidence for stories related to sunk costs, peer and teacher influences, or beliefs about earnings and work conditions. Importantly, STEM assignment involves a risk: it improves wellbeing, high school satisfaction, and career prospects on average, but reduces performance on a high school exit exam, leading to lower college enrollment and retrospective satisfaction among low-achieving students. Similarly, it makes female students report negative experiences with college professors and peers. A final finding is that male students become more conservative. We provide suggestive explanations for our results on wellbeing and political preferences by measuring effects on time use, friendships, and expectations.

Measuring residential sorting responses to local taxes for school spending

With Damon Clark

Many studies have attempted to estimate the impacts on student outcomes of increased K-12 education spending. These studies are potentially confounded by effects of spending on household migration and thus on student composition within localities and schools. This paper estimates “residential sorting responses” using a regression discontinuity design applied to data from California on elections for parcel taxes—lump-sum taxes levied by a school district that can be used for operational spending. A key innovation is to construct rich measures of an area’s residents and in- and out-migrants using longitudinal income tax records for the population of California taxpayers over a twenty-two year period. In line with research on local taxes for operational spending in other states, we find that passing a parcel tax increases operational spending by around 6% and leads to a small decline in the share of students eligible for free or reduced-price meals. However, we find null effects on our sorting measures from the tax data; hence we conclude that any geographic-based sorting responses are small.

Measuring gerrymandering by recovering preferences and turnout costs

In this paper, I develop a new method for how to measure whether a legislative map is gerrymandered. My method allows evaluating a map along two key dimensions of map quality. These are proportionality (the alignment between a party’s seat share and its statewide vote share) and competitiveness (the fraction of legislative contests with uncertain winners). The method is designed to account for the main criticism of existing approaches for measuring gerrymandering. In particular, it is commonly argued that existing approaches cannot accurately predict how a map will perform in future elections. This is because future elections will be subject to an unknown electorate and an unknown set of electoral shocks. Importantly, the U.S. Supreme Court recently ruled that this uncertainty is so intractable that the judicial system is unqualified to determine whether a map is gerrymandered. My method responds to this criticism by directly assessing the degree of uncertainty in a map’s quality. The method uses a simulation procedure that is built on top of a structural voting model. The model describes the preference and turnout decisions of a potential voter and decomposes an election into a small number of utility parameters. I fit the model in multiple elections and measure how much the utility parameters vary over time. I then simulate counterfactual elections by drawing from the across-election distribution of these utility parameters. I also allow for demographic changes in the electorate by manually altering the set of potential voters. As an empirical example, I apply the method to rich data from the 2008 to 2020 general elections in North Carolina. I show that the method allows credible and precise evaluations of maps. I also show that it has stronger predictive power than existing approaches.