92
create a model including all the macro-factors specified above, I would hardly be able to
do the analysis for die whole sequence. The argument is practical: data about economic
performance may be available for the whole 1983-2007 sequence, but no comparable
series with indicators of presidential popularity, perceptions about performance of die
economy, state-level government or candidate evaluation, and even partisanship or
ideological positions exist. Thus, I would have to either limit my analysis to an
extremely short time period, or drop most of my observations due to missing values.
That situation given, I prefer to take a different risk and evaluate how Congress-related
covariates affect the chances of success in subnational executive races. Many observers
might immediately point out that an omitted variable bias might be tainting results; <⅛,'
nonetheless, nothing prevents the experiment of appraising whether even a small part of
the variance can be explained by candidate-level strategic behavior. Ultimately, this t
project's goal is not to capture how the whole world performs, but simply to understand
the relationship between political careers and legislative production in multilevel
systems.
Empirical Strategy
Empirical analyses of the effects of legislative submission over subnational
success require some adjustments in the data. Given that success can only be evaluated
over individuals that did run for an executive position, it is nonsense to include subjects
without immediate subnational expectation in the sample. Thus, I decided to drop those
observations. As a consequence, I created two separate samples, one of 198
gubernatorial candidates and another of 97 mayoral contenders. Since the goal is to
evaluate how bill submission affects executive success, I decided to work with