Colloquium

The Department of Statistics
presents

Paul Brace
Department of Political Sciences
Rice University

Formal approaches to politics are sometimes criticized for an over-emphasis on logical taughtness and an under-emphasis on empirical support. Empirical complexity makes it difficult to distinguish causal relationships posited by theory from the confounding influences arising in real world settings. Too often this forces formal theorists to retreat to a ceteris paribus world. This talk starts from the premise that the way to deal with complexity is through the systematic comparison of individual behavior in institutions over time. Behavior (assumed purposive) is believed conditioned by the incentives and disincentives embodied in institutional rules and procedures. The best way to control for complexity is to treat institutions as variables, thus allowing control for confounding influences and alternative initial conditions that frustrate those studying behavior in single institutions. Typical methodological approaches are insufficient for this task. On the one hand, the need to examine alternative institutional settings would lead us to consider cross sectional or unit variations. Convincing causal inference requires concern for temporal change which would normally lead to the use of time series methods. Yet time series analysis typically focuses on a single unit (nation, city, state) and doesn't allow consideration of institutional variation on outcomes.

Over the last dozen years there has been growing interest in models that can pool units over time. Pooled Cross Sectional Time Series analysis presents special advantages for examining the effects of institutions on behavior over time. The emergence of these relatively new methodologies and the desire for greater empirical connection and generalizability in formal modeling produces a tremendous opportunity for advancing our understanding of how rules shape behavior and outcomes. At this point we have an abundance of un or under-tested theoretical propositions and a wealth of appropriate methods for evaluating them. What we lack are reliable data on elites in institutions over time.

The talk will conclude with a demonstration of a data collection project currently being undertaken by the speaker with funding from the United States National Science Foundation. Coding procedures will be discussed and reliability results will be presented.