Essays in political economy and applied econometrics
Garofalo, Pablo Javier
MetadataShow full item record
This dissertation is comprised of two essays in political economy and one in micro-econometrics. Each of them proposes an alternative methodology to improve on the estimation of a specific economic phenomenon. The first essay studies the political allocation of US federal resources to localities taking into consideration that the states are also actively involved in allocating resources to localities. I found that federal funds are biased towards localities within states that are not represented by the same party as the one that represents the federal government. This finding implies that a strategic federal government takes into account that non-aligned states have different spending priorities. These results suggest that past research on the allocation of federal resources to localities has shown biased estimates when the political allocation of resources is not studied in the context of a multi-layered government environment. The second essay exploits the existence of extended interlude periods (i.e., time between elections and government change date) from Latin American countries to identify a causal effect of a change in the probability of electoral defeat on a change in the budget deficit. Theoretical studies on the strategic use of debt argue that governments issue more debt when facing a higher probability of electoral defeat. Testing this hypothesis has proven challenging since measures of that probability are potentially endogenous. Since my identification strategy is focused on identifying the effects of electoral surprises, I provide a plausible source of exogenous variation. I find that the higher the increase in the probability of electoral defeat (victory), the larger the increase (decrease) in the deficit. The third essay studies the properties of a maximum likelihood estimator (MLE) of dynamic panel data models with fixed effect when difference GMM methods suffer from weak identification. While previous studies propose moments to solve the weak identification under difference GMM for stationary processes only, this study shows that MLE solves the weak identification issue not only when the process is stationary, but also when it is not.