Wednesday, November 28, 2012

Nikhil Agarwal investigates the medical match (and school choice)

The matching literature has been short on theoretically sophisticated investigators who are simultaneously tooled-up for the most serious kinds of empirical work and curious about the rules of the game that make markets work.  Nikhil Agarwal should put your mind at rest on that score. His dissertation committee consists of Ariel Pakes, Susan Athey, Parag Pathak and me. And his job market paper reports an investigation of a market close to my heart:  An Empirical Model of the Medical Match

"Abstract: This paper develops a framework for estimating preferences in two-sided matching markets with non-transferable utility using only data on observed matches. Unlike single-agent choices, matches depend on the preferences of other agents in the market. I use pairwise stability together with a vertical preference restriction on one side of the market to identify preference parameters for both sides of the market. Recovering the distribution of preferences is only possible in an environment with many-to-one matching. These methods allow me to investigate two issues concerning the centralized market for medical residents. First, I examine the antitrust allegation that the clearinghouse restrains competition, resulting in salaries below the marginal product of labor. Counterfactual simulations of a competitive wage equilibrium show that residents' willingness to pay for desirable programs results in estimated salary markdowns ranging from $23,000 to $43,000 below the marginal product of labor, with larger markdowns at more desirable programs. Therefore, a limited number of positions at high quality programs, not the design of the match, is the likely cause of low salaries. Second, I analyze wage and supply policies aimed at increasing the number of residents training in rural areas while accounting for general equilibrium effects from the matching market. I find that financial incentives increase the quality, but not the number of rural residents. Quantity regulations increase the number of rural trainees, but the impact on resident quality depends on the design of the intervention."

Nikhil is also doing exciting empirical work on school choice: here's the abstract from a forthcoming (and largely completed) working paper;


Sorting and Welfare Consequences of Coordinated Admissions: Evidence from New York City
with Atila Abdulkadiroglu and Parag PathakComing Soon.
Centralized and coordinated application systems are a growing part of recent school choice reforms. This paper estimates preferences for schools using rank order lists from New York City's new high school assignment system launched in Fall 2003 to study the consequences of coordinating school admissions in a mechanism based on the student-proposing deferred acceptance algorithm. Compared to the prior mechanism with multiple offers and a limited number of choices, there is a 40% increase in enrollment at assigned school. The old mechanism restricted choices and placed many students close to home, while the new mechanism assigns students to schools 0.7 miles further from home on average. Student preferences trade off proximity and school quality, but are substantially heterogeneous. Even though students prefer closer schools, the new mechanism is more likely to assign students to schools they prefer and this more than compensates for the distance increase. The average welfare increases by the equivalent of 0.25 miles from the new mechanism. Students from all boroughs, demographic groups, and baseline achievement categories obtain a more preferred assignment on average from the new mechanism, suggesting that allocative changes involving assignment mechanisms need not be zero-sum.


Nikhil is on the market. You could hire him this year.

1 comment:

  1. Nice post !! Marriage.....Its an important and beautiful part of our life.Because it contain amazing rituals.Marriage make two strange persons spouse to each other.
    Shubh Sanjog

    ReplyDelete

Note: Only a member of this blog may post a comment.