This dissertation presents work on gerrymandering in American legislative districts and on school competition and school choice. The work on gerrymandering analyzes how to measure gerrymandering and investigates some of its causal effects. The analysis of how to measure gerrymandering is presented in Chapter 1 and in the first half of Chapter 2. The context is the following. Legislative maps are often evaluated along dimensions of proportionality (the alignment between parties' seat shares and their state- or nation-wide vote shares) and competitiveness (the fraction of contests with uncertain winners). Since a map is intended to be used for multiple elections, policy-makers want to accurately predict how it will perform on these dimensions in the future. Doing this is difficult because future elections will differ from past ones due to changes in the demographic composition of the electorate and as a result of electoral shocks to preferences and turnout costs. Citing this uncertainty, the U.S. Supreme Court recently ruled that the judicial system is incapable of adjudicating claims of partisan gerrymandering.
The first contribution of the dissertation is to develop a method for predicting the uncertainty in a map's performance due to electoral shocks and changes in demographics. The method relies on a structural voting model, which describes the preference and turnout decisions of a potential voter. The model decomposes an election into (i) a set of candidate qualities and (ii) individual-level utility parameters. I assess map performance in two steps. First, I examine the effect of electoral shocks by simulating alternative values of the candidate qualities and utility parameters. Second, I investigate the influence of demographic changes by re-running the simulations using different electorates. I apply the method to rich data from the 2008 to 2018 general elections in North Carolina and show that it allows credible and precise evaluations of maps. I also show that the method is better than existing approaches at predicting gerrymandering outcomes in excluded elections.
The remainder of Chapter 2 concerns the causal effects of gerrymandering. Specifically, I examine whether the probability that someone turns out to vote is influenced by the competitiveness of his or her legislative districts. I do this by comparing outcomes over time for individuals in North Carolina who were placed into more or less competitive districts in 2011 as part of the decadal ``redistricting" process. I compare individuals who shared the same districts in each legislative chamber (U.S. House, NC Senate, NC House) before redistricting and who differed in districts for only one chamber after redistricting. Within these comparison groups, I match individuals on demographics and history of turnout and party registration. I find that being placed into a less competitive district reduces turnout. Effects grow over time and exist in both midterm and presidential elections. By 2018, having been placed in a district in which one party is always predicted to win versus one in which the parties have an even chance of winning reduces turnout by 1.9 percentage points for U.S. House districts and 1.4 percentage points for NC House and NC Senate districts. These results highlight the importance of considering district competitiveness when drawing legislative maps.
Chapter 3 is work that is joint with Rajeev Dehejia, Cristian Pop-Eleches, and Miguel Urquiola. It examines how schools' incentives are influenced by the way in which households make school choice decisions. A summary is as follows. Recent work examines whether households choose schools based on school value added (Abdulkadiroglu et al. 2020; Beuermann et al. 2019). Given that value added is difficult to observe, households' choices are likely to depend on both (i) how much they care about value added and (ii) how well informed they are about which schools have high or low value added. We examine this concern using administrative data, a survey, and an experiment in Romanian high school markets. Using the survey, we can explain households' preferences based on their beliefs about school traits, rather than on the values of these traits that are measured by researchers. In the administrative data, we find that households' choices are better explained by measured values of peer quality than by measured values of value added. By contrast, in the survey data, we find that households' beliefs about value added and peer quality have equal explanatory power for their choices. This motivates an experiment in which we provide households with information on school value added. We find that the information has a positive but heterogenous effect on the extent to which households prioritize value added in their school choices. Effects are largest for households who were initially less certain of their choices and for households with low-scoring students.
| Identifer | oai:union.ndltd.org:columbia.edu/oai:academiccommons.columbia.edu:10.7916/d8-7gcp-yn07 |
| Date | January 2020 |
| Creators | Ainsworth, Robert M. |
| Source Sets | Columbia University |
| Language | English |
| Detected Language | English |
| Type | Theses |
Page generated in 0.0017 seconds