Leeat Yariv, Ph.D. - US grants

Major decisions in business, economics, and politics often result from some collective choice process. Shareholders approve mergers, the Federal Open Market Committee sets interest rates, and citizens elect their president. Recent theoretical work has added much desired realism to models of collective decision making by including important features such as deliberation, abstention, and costly information acquisition. This proposal describes an array of experiments to test these newly proposed theories. The experiments are designed to be sensitive to possible biases in decision making (e.g., conformist behavior) and information processing (e.g., selective exposure). As such, the results will suggest ways to enrich the standard rational choice paradigm by admitting behavioral aspects into the analysis of collective choice problems.

This study is broadly divided into two parts. When information is exogenously determined the proposed experiments include several novel features, e.g., the comparison of a wide range of voting institutions (including majority and unanimity) in the presence of possibly conflicting preferences. In addition, this study evaluates the impact of deliberation and abstention on collective outcomes. The project also initiates a sequence of laboratory explorations when information is endogenously collected. This part will generate empirical insights about the interaction between institutions, the incentives to acquire information and the quality of the resulting collective decision process.

The intellectual merit of the proposed activity is manifold. The experiments are founded on theoretical models at the frontier of economic analysis of voting behavior, awaiting a thorough empirical investigation. Using a revealed preference approach allows the researchers to identify important behavioral elements germane to collective decision making. The empirical methods employed center around the Quantal Response Equilibrium, a novel framework allowing for some degree of "noise" in choices. This approach is being increasingly used to analyze experimental data and permits structural estimation of unobserved preferences and individual heterogeneity.

The broader impacts resulting from the proposed activity include the opportunity for improving collective choice procedures when participation and information collection is costly. The findings about how deliberation affects institutional outcomes could provide foundations for the political and philosophical theories of deliberative democracy. Furthermore, the experimental methodology used to study communication could spill over to other contexts such as auctions and mechanism design at large. The broader impacts include two additional contributions: the development of jVote, an open-source Java program for running voting experiments, and a sequence of mini-conferences to be held at Caltech. Web-based jVote will facilitate disseminating the experimental methodology to a wide audience of economists and political scientists, including those without access to a physical lab. The mini-conferences will serve to strengthen the dialogue between theory and empirics, and between economics and political science.

Matching markets have an important role in the economy. They determine such varied outcomes as which student goes to which college, who gets what job, and who buys what house. Recent work on the theory of matching markets has been useful to the design of markets as well as for the repair of market failures. Much of this work, however, has focused on the design and analysis of centralized matching markets. Many matching markets are not fully centralized; for example, college admissions decisions in the United States are made by each individual institution on its own. Furthermore, almost all of the centralized matching marekts are preceeded by decentralized opportunities for participants to match. This award funds research on the game theory of these decentralized markets as well as lab experiments and a field study to identify the crucial elements that determine decentralized market outcomes.

The field study uses the newly developed theory to analyze a unique data set tracing the way a specific adoption agency matched adopting parents to domestic birth mothers. The goal is to use these data to better understand the preferences of adoptive parents and birth mothers towards potential matches.

Broader Impacts: The theory and experiments will provide useful tools for improving the design of matching markets. The field study gives crucial information for child advocates who are seeking to improve state-level policy towards adoption and foster care.

The outcomes of higher education are a consequence of individuals' personality traits and many
dimensions of the university experience. The Caltech Cohort Study (CCS) will track the entire
undergraduate student body of the California Institute of Technology (Caltech) over several years. It
uses repeated, incentivized surveys to elicit behavioral proxies: risk aversion, strategic sophistication,
competitiveness, social behavior, etc. The results will be married with objective outcomes: academic,
social, and financial. The project will examine the evolution of attributes over college years, the
formation of social interactions, and ultimately how certain attributes interact to produce educational
outcomes. The CCS also serves as an ideal environment for answering fundamental questions about
the external validity of incentivized surveys and laboratory experiments.

The CCS has four unique features. First, Caltech's small size makes it possible to track nearly
the entire student body over time (90% of students completed the first three surveys). Second, the
survey design allows measurement of a large set of behavioral traits for each individual. Third,
the repeated nature of the study makes it possible to track the stability of these traits across time
with minimal selection bias. Fourth, the array of behavioral traits elicited can be tied to real-world
behaviors and outcomes. The combination of these elements should be invaluable for a large battery
of basic social science questions.

The proposed study has the potential to improve the educational process in US higher education.
Understanding what impacts social and scholastic outcomes will help design certain aspects of college
education aimed at increasing well-being and academic performance, as well as furthering gender and
racial equality.

The proposal focuses on three areas. First, the dataset's longitudinal nature will allow for a careful
inspection of the stability (or lack thereof) of a wide array of behavioral attributes. Second, the rich
data on different layers of interactions over time (social, academic, and geographic) combined with
the comprehensive set of elicited attributes will allow us to study network formation processes. Last,
the data can be cross-checked against Caltech's experimental laboratory data to examine the impacts
of selection into experiments as well as the differences in responses occurring in the lab and outside.
The survey will also allow for a comparison of responses across other platforms and populations.
Ultimately, the detailed longitudinal data that will be collected have the potential to uncover
predictors of academic outcomes (chosen major, grades, and graduation), and the channels affecting
the evolution of different traits.