As children, most of us delighted in Rudyard Kipling's “just-so” stories. Kipling began with a simple fact, for example, elephants have long noses. Then he made up a fanciful story to explain it: perhaps a crocodile stretched an elephant's nose and the trait was passed down to other elephants. Just so!

In the previous chapter, I began with some simple facts, and then made up a causal mechanism to explain them. The three facts were: vetoes actually occur, veto chains are relatively common, and concessions often occur over the course of a chain. The first model failed to generate any of these patterns, but it supplied a useful framework for additional thought. The override model can generate the first and second patterns, but not the third. The sequential veto bargaining (SVB) model can generate all three.

Are the override and SVB models merely just-so stories? Or do they capture something real about the dynamics of interbranch bargaining? How can we tell? To answer this question I rely on the criteria outlined toward the end of Chapter 3. As indicated there, a sine qua non for a good model is an ability to explain empirical puzzles it was not designed to explain. Table 2.12, which I reproduce in separate tables throughout this chapter, is filled with empirical puzzles that go well beyond the three “stylized facts.” For example, why does the probability of a veto increase with legislative significance during divided government, but not during unified government? Why are veto chains short? Why don't any of the covariates predict the success of an override attempt, once the decision to attempt an override is made? Can the models explain these facts?