2026-09-19T23:46:56Zhttps://dspace.mit.edu/server/oai/requestoai:dspace.mit.edu:1721.1/1073192026-06-17T14:42:36Zcom_1721.1_7582com_1721.1_7581col_1721.1_131022
Jerry A. Hausman and Glenn Ellison.
Hou, J. Mark (Jie Mark)
Sodomka, Eric
Stier Moses, Nicolás E
Massachusetts Institute of Technology. Department of Economics.
Massachusetts Institute of Technology. Department of Economics
2017-03-10T15:05:08Z
2017-03-10T15:05:08Z
2016
2016
http://hdl.handle.net/1721.1/107319
972738467
Thesis: Ph. D., Massachusetts Institute of Technology, Department of Economics, 2016.
Cataloged from PDF version of thesis. "with Eric Sodomka and Nicolas E. Stier-Moses"--Page 6 [Below title of Chapter 1].
Includes bibliographical references.
Chapter 1 focuses on the problem of predicting equilibrium outcomes in large online auction markets. For online retailers, content publishers, and search engines, predicting how the behavior of their auction markets might respond to policy changes is an important business problem. However, this problem is challenging due to both the size and the complexity of such real-world markets. We introduce a method for predicting how various statistics of such markets adjust to changes in supply and demand by: (1) modeling the auction market mechanism as a Walrasian mechanism, (2) coarsening the resulting Walrasian market via a stochastic block model, (3) computing the Walrasian equilibrium of this coarsened market through sampling, and (4) using the resulting equilibrium, together with some reduced-form adjustments, to approximate the equilibrium of the initial auction market. We demonstrate the internal consistency of this method through formal proofs and synthetic experiments, and demonstrates its accuracy by comparison with the equilibrium outcomes of a more realistic pacing-based model of auction markets. Chapter 2 introduces a model of consumer choice in which consumers simplify their latent high-dimensional preference vector into a low-dimensional one used for choosing products. This assumption induces a particular population structure over consumers' simplified preferences, which allows for tractable estimation in high dimensional settings. Estimation is performed via a stochastic gradient descent-based algorithm, and we evaluate its performance through a variety synthetic benchmarks. We also estimate the model on consumer consideration data, finding that the average consumer uses only 6 of 16 product attributes when forming their consideration set, and that this leads to a utility of loss of 2 - 3% on average. Chapter 3 uses admissions data from the University of Bologna's medical school to analyze how students' entrance exam rankings affect their subsequent academic performance. We find that: (1) worse rankings lead to worse academic performance, (2) this impact is more negative for worse-ranked students, (3) this impact on academic performance operates mostly through courseload rather than through GPA, and (4) male and female students' academic performance do not respond differentially to rank.
by J. Mark Hou.
Ph.D.
151 pages
eng
Massachusetts Institute of Technology
MIT theses are protected by copyright. They may be viewed, downloaded, or printed from this source but further reproduction or distribution in any format is prohibited without written permission.
http://dspace.mit.edu/handle/1721.1/7582
Economics.
Topics in applied econometrics
Thesis
application/pdf
false
Publication
unknown
unknown
<Publication xmlns="https://www.openaire.eu/cerif-profile/1.1/" id="69c6191b-fdbc-46c0-a2ca-d73bb02518d7">
<Type xmlns="https://www.openaire.eu/cerif-profile/vocab/COAR_Publication_Types">http://purl.org/coar/resource_type/c_1843</Type>
<Language>eng</Language>
<Title>Topics in applied econometrics</Title>
<PublishedIn>
<Publication>
</Publication>
</PublishedIn>
<PublicationDate>2016</PublicationDate>
<Authors>
<Author>
<DisplayName>Hou, J. Mark (Jie Mark)</DisplayName>
<Affiliation>
<OrgUnit>
</OrgUnit>
</Affiliation>
</Author>
<Author>
<DisplayName>Sodomka, Eric</DisplayName>
<Affiliation>
<OrgUnit>
</OrgUnit>
</Affiliation>
</Author>
<Author>
<DisplayName>Stier Moses, Nicolás E</DisplayName>
<Affiliation>
<OrgUnit>
</OrgUnit>
</Affiliation>
</Author>
</Authors>
<Editors>
</Editors>
<Publishers>
<Publisher>
<DisplayName>Massachusetts Institute of Technology</DisplayName>
<OrgUnit />
</Publisher>
</Publishers>
<License>http://dspace.mit.edu/handle/1721.1/7582</License>
<Keyword>Economics.</Keyword>
<Abstract>Chapter 1 focuses on the problem of predicting equilibrium outcomes in large online auction markets. For online retailers, content publishers, and search engines, predicting how the behavior of their auction markets might respond to policy changes is an important business problem. However, this problem is challenging due to both the size and the complexity of such real-world markets. We introduce a method for predicting how various statistics of such markets adjust to changes in supply and demand by: (1) modeling the auction market mechanism as a Walrasian mechanism, (2) coarsening the resulting Walrasian market via a stochastic block model, (3) computing the Walrasian equilibrium of this coarsened market through sampling, and (4) using the resulting equilibrium, together with some reduced-form adjustments, to approximate the equilibrium of the initial auction market. We demonstrate the internal consistency of this method through formal proofs and synthetic experiments, and demonstrates its accuracy by comparison with the equilibrium outcomes of a more realistic pacing-based model of auction markets. Chapter 2 introduces a model of consumer choice in which consumers simplify their latent high-dimensional preference vector into a low-dimensional one used for choosing products. This assumption induces a particular population structure over consumers' simplified preferences, which allows for tractable estimation in high dimensional settings. Estimation is performed via a stochastic gradient descent-based algorithm, and we evaluate its performance through a variety synthetic benchmarks. We also estimate the model on consumer consideration data, finding that the average consumer uses only 6 of 16 product attributes when forming their consideration set, and that this leads to a utility of loss of 2 - 3% on average. Chapter 3 uses admissions data from the University of Bologna's medical school to analyze how students' entrance exam rankings affect their subsequent academic performance. We find that: (1) worse rankings lead to worse academic performance, (2) this impact is more negative for worse-ranked students, (3) this impact on academic performance operates mostly through courseload rather than through GPA, and (4) male and female students' academic performance do not respond differentially to rank.</Abstract>
<Access xmlns="http://purl.org/coar/access_right"
>
</Access>
</Publication>