Essays in Education and Voting
Name
carlstein-aecarls-phd-economics-thesis.pdf
Size
12.29 MB
Format
Adobe PDF
Checksum (MD5)
3113e017e538227a58e8ccd1799d5639
Author(s)
Carlstein, Anne E.
Advisor(s)
Pathak, Parag
Morris, Stephen
Date Issued
May 2026
Publisher
Massachusetts Institute of Technology
Abstract
This thesis studies the design of allocation and choice mechanisms, focusing on the economics of education and voting theory.
The first chapter addresses the following question: What happens if students are randomized to elite schools? I study the causal effects of attending several highly sought-after public high schools in New York City (termed the “screened select”) under two starkly different admissions regimes: a traditional screening process based on test scores and grades, and a partial lottery introduced in 2021. Leveraging this policy reform, I compare the impacts of screened select attendance before and after the admissions change using an IV strategy. While the screened select schools boosted overall SAT scores by approximately 30 points under traditional screened admissions, they had no positive impact in the post-reform lottery era. However, these effects are heterogeneous: students who would not have been admitted under full screening saw positive SAT Math impacts, and students who would have been admitted under full screening saw negative SAT Math impacts. I provide evidence that these effects are primarily driven by changes in the school peer composition rather than shifts in school resources or curricula. Using a stylized theoretical model of admissions, I formalize the tradeoff between equity and testing outcomes and derive conditions for when a partial lottery would be optimal.
The second chapter explores ranked-choice voting. This election method has sparked intense debate in political circles as it has gained traction as an alternative to standard plurality rule (e.g., the New York City mayoral contest). Proponents of ranked-choice voting argue that it prevents spoiler effects and selects candidates that are widely accepted by the overall electorate (such as centrist candidates or the Condorcet winner, i.e., the candidate who would pairwise defeat all competitors, if such a candidate exists). Restricting attention primarily to the three candidate case, I derive theoretical results which describe the performance of ranked-choice voting according to several metrics. These include the selection of consensus candidates, fairness to certain types of parties, and performance under strategic manipulation by voters. Under sincere voting, while ranked-choice voting dominates standard plurality rule in terms of Condorcet winner selection, it could further be strictly improved with a modified elimination rule. In comparison with a stylized sequential primary-then-general election setup, ranked-choice voting does not favor specific types of parties. Under strategic voting, ranked-choice voting rules out certain undesirable equilibria unfavorable to consensus candidates that plurality rule does not. I highlight examples of mayoral and congressional elections that show relevant ranked-choice voting phenomena.
The third chapter examines monotonicity failures in ranked-choice voting caused by endorsements. An endorsement-induced monotonicity failure occurs when a candidate who would have won an election loses after receiving an endorsement, specifically, after being moved upward in some subset of voter rankings. I consider a probabilistic function that captures the effect of an endorsement, and show conditions under which such failures arise. I show that endorsement-induced monotonicity failures cannot occur in three candidate elections, in contrast to standard monotonicity failures. With four candidate elections and general preferences, endorsement-induced monotonicity failures may occur under certain necessary conditions. These cases occur when the order of elimination of candidates is changed by an endorsement. Under single-peaked preferences, endorsement-induced monotonicity failures are more limited: impossible with four candidates, and possible for only one class of profiles in the five candidate case.
MIT Department
Massachusetts Institute of Technology. Department of Economics
Terms of Use
In Copyright - Educational Use Permitted
Copyright retained by author(s)
Persistent DSpace Link