Algorithm is Experiment: Machine Learning, Market Design, and Policy Eligibility Rules

Faculty Sponsor(s): 
Yusuke Narita

Project Description(s):

Professor Narita is recruiting one Fellow to work with him and his co-authors on the following projects:

Today’s society increasingly resorts to data and algorithms for decision-making and resource allocation. For example, judges and police officers make legal judgments using predictions from past data and machine learning algorithms. Similar algorithms are also used by financial institutions (such as banks and insurance companies), retailers, tech companies, and governments. School districts, college admissions systems, and labor markets use computational algorithms for position allocations, producing useful data. Data-driven, algorithmic decision-making thus impacts the lives of numerous people.

This observation motivates me to work on “data mechanism design,” with applications to public policy and business. By “data mechanisms,” I mean data-generating decision/allocation systems that impact individuals’ welfare and incentives.

I am seeking a full-time research assistant to help in related on-going projects. Below are a few examples:

Project 1: Algorithm is Experiment: Machine Learning, Market Design, and Policy Eligibility Rules

Machine learning, market design, and other algorithms produce a growing portion of decisions and recommendations. Such algorithmic decisions are natural experiments (conditionally quasi-randomly assigned instruments) since the algorithms make decisions based only on observable input variables. We use this observation to characterize the sources of causal-effect identification for a class of stochastic and deterministic algorithms. Data from almost every algorithm is shown to identify some causal effect. This identification result translates into a treatment-effect estimator. We prove that our estimator is consistent and asymptotically normal for well-defined causal effects. The estimator is easily implemented even with high-dimensional data and complex algorithms. Our estimator induces a high-dimensional regression discontinuity design as a key special case. We are applying this method to a variety of policy and business problems.

Project 2: Hearing the Voice of the Future: US Presidential Elections

Many countries face growing concerns that population aging may make voting and policy-making myopic. This concern begs for electoral reform to better reflect voices of the youth, such as weighting votes by voters’ life expectancy. This paper predicts the effect of the counterfactual electoral reform on the 2016 U.S. presidential election. Using the American National Election Studies (ANES) data, I find that Hillary Clinton would have won the election if votes were weighted by life expectancy. I plan to extend this analysis to use better data and measure the effects on the welfare of different generations

Requisite Skills and Qualifications:

This project involves econometric/statistical and empirical aspects. An ideal candidate is someone who has done coursework in econometrics, statistics, machine learning, and empirical microeconomics.

Special Application Instructions:

None.

Sponsor Name: 
Narita