A Bayesian Multilevel Modeling Approach to Time Series Cross-Sectional Data

(with Joseph Bafumi, Luke Keele, and David Park)
Published in Political Analysis, Spring 2007

The analysis of time series cross-sectional (TSCS) data has become increasingly popular in political science. Meanwhile, political scientists are also becoming more interested in the use of multilevel models. However, little work exists to understand the benefits of multilevel modeling when applied to TSCS data. We employ Monte Carlo simulations to benchmark the performance of a Bayesian multilevel model (BML) for TSCS data against popular models used in the literature. We use various diagnostics to analyze the performance of our approach relative to these techniques. Compared to the most commonly employed estimators for such data, we find that the Bayesian multilevel model is (1) equally unbiased on average, (2) considerably more efficient, and (3) reports higher quality standard errors. Moreover, the BML is more general and flexible, which offers researchers additional advantages for TSCS data.


Presidential Power and Distributive Politics: Federal Grants Expenditures In the 50 States, 1983-2001

Abstract: What explains the differences in federal spending in individual states across time? The current empirical literature on distributive politics has focused on Congress and neglected the presidency in explaining the distribution of federal spending, and it has looked at limited periods of time. I address these flaws by using Bayesian multilevel techniques to model of spending on grants over an extended period (1983-2001). Presidential variables are quite significant in influencing the distribution of grant expenditures.

The Weak Leviathan? Testing Partisan Theories of Political Influence on Defense Procurement in Congressional Districts: 98th-102nd Congress

Abstract: Analyses of federal spending across congressional districts have tended to ignore the institutional, geographical, and longitudinal context in which these districts are located. This neglects other sources of political influence, such as Senators, other representatives, party leaders, and Presidents. A decade-long data set (98th-102nd Congress, or 1983-1992) of defense procurement awards was collected and fitted to a Bayesian multilevel time series cross-sectional model. Partisan hypotheses from the modern literature on political determinants of federal spending are tested. There is little evidence that the Democratic party has been successful in steering defense spending towards Democrat districts, voters, members of the Armed Services committee, and Democratic state Congressional delegations–at a time period when Democrats were the Congressional majority and would have been expected to be able to do so.


Get every new post delivered to your Inbox.