Research

Refereed publications


Working papers

Abstract. Every year many colleges provide housing for admitted students. However, there is no college admissions process that considers applicants’ housing needs, which often results in college housing shortages. In this chapter, I formally introduce housing quotas to the college admissions problem and solve it for centralized admissions with common dormitories. The proposed setting is inspired by college admissions where applicants apply directly to college departments, and colleges are endowed with common residence halls. Such setting has many real-life applications: hospital/residents matching in Japan (Kamada and Kojima, 2011, 2012, 2015), college admissions with scholarships in Hungary (Biró, 2012), etc.

A simple example shows that there may not be a stable allocation for the proposed setting. Therefore, I construct two mechanisms that always produce some weakened versions of a stable matching: a Take-House-from-Applicant-stable and incentive compatible cumulative offer mechanism that respects improvements, and a Not-Compromised-Request-from-One-Agent-stable (stronger version of stability) cutoff minimising mechanism. Finally, I propose an integer programming solution for detecting a blocking-undominated Not-Compromised-Request-from-One-Agent-stable matching. Building on these results, I argue that presented procedures could serve as a helpful tool for solving the college housing crisis.

Presented at: Conference on Mechanism and Institution Design at the National University of Singapore (online, Summer 2022) & 12th Conference on Economic Design at the University of Padova (in person, Summer 2022)


Abstract. This study proposes a number of solutions to resource allocation problems in an affirmative action agenda. Quotas are introduced as a way to promote members of minority groups. In addition, reserves may overlap: any candidate can belong to many minority groups, or, in other words, have more than one trait. Moreover, once selected, each candidate can fill one reserve position for each of her traits, rather than just one position for one of her traits. This makes the entire decision process more transparent for applicants and allows them to potentially utilize all their traits. I extend the approach of Sönmez and Yenmez (2019) who proposed a paired-admissions choice correspondence that works under no more than two traits. In turn, I allow for any number of traits focusing on extracting the best possible agents, such that the chosen set is non-wasteful, the most diverse, and eliminates collective justified envy. Two new, lower- and upper-dominant choice rules and a class of sum-minimizing choice correspondences are introduced and characterized.


Work in progress

Abstract. In this paper I implement optimization techniques for detecting the efficient trade off between ex-post Pareto efficiency (for one side of a two-sided matching market) and ex-ante stability for small one-to-one matching markets. Neat example (Roth, 1982) proves that there is no matching mechanism that achieves both efficiency (for one side of the one-to-one matching market) and stability. As representative mechanisms I choose deferred-acceptance for stability, and top trading cycles combined with random serial dictatorship for the last stage for Pareto efficiency (both of them are strategy-proof for one side of the market). I compare performances of a randomized matching mechanism that efficiently simultaneously relaxes efficiency and stability, and a convex combination of two representative mechanisms. Results show that the constructed mechanism significantly improves efficiency and stability in comparison to mentioned convex combination of representative mechanisms. In addition, I propose an alternative deep learning approach as in Ravindranath et al. (2021).


Abstract. This paper studies the current college admissions procedure in France, called Parcoursup. On the one side of the corresponding two-sided matching market there are candidates possessing none, one or two of the following traits (targets of an affirmative action): a scholarship-holder (la boursier), a resident. On the other side there are regular institutions (la formation) and boarding schools (l'internat), i.e. institutions with private dormitories. Parcoursup mechanism is based on the college-proposing deferred acceptance algorithm (CDA). I identify a shortcoming of this mechanism that prevents it from being stable even with a single regular institution and no boarding schools. Then I propose two alternatives to Parcoursup. The first one is a stable CDA (SCDA) constructed with stable choice rules  satisfying substitutes condition of Hatfield and Milgrom (2005) for regular institutions and boarding schools. However, in order to obtain such choice rules some features of the model should be neglected. Thus, I also design the second alternative: a modified CDA (MCDA) constructed with stable choice rules that consider the full model.