6 - Testing the Models

Summary

In the previous chapter I asked the models to explain many patterns they were not designed to explain. These repeated confrontations with data provided tough tests of the models. However, the patterns analyzed in Chapter 5 were all drawn from the statistical portrait of vetoes in Chapter 2. In this chapter I reverse the procedure: I use the models to generate predictions that lie outside the initial data. Then I test the predictions using new data. The predictions deal with three distinctively different phenomena: concessions, deadlines, and the legislative productivity of Congress.

In Chapter 2, I reported information gleaned from a content analysis of legislative histories. The legislative histories often reveal concessions in successors to vetoed bills, though sometimes the record is ambiguous and hard to interpret. The prevalence of concessions supplied the starting place for the model of sequential veto bargaining. Is there additional support for the fundamental “stylized fact” targeted by the model? Beyond this, do patterns in concessions conform to the predictions of the models? In this chapter I combine data on roll call votes on initial and successor bills with estimates of legislators’ preferences, derived from all roll call votes, to derive estimates of the direction and magnitude of policy concessions between successive bills in veto chains. These new data on concessions confirm the importance of concessions in veto bargaining and provide several tests of the game theoretic models.

A strong test for a model comes from predictions about a hitherto unstudied phenomenon. The bargaining models make several striking predictions about a phenomenon that political scientists have not previously studied: changes in the probability of vetoes immediately before presidential elections, a “deadline effect.”