One response to the previous chapters might be to agree that theory is inconclusive and look to empirical research to determine whether the positive or negative effects of decentralization tend to dominate. “Where theory leaves us completely stranded,” advise Dahl and Tufte (1973, p. 44), “we can be rescued only by data.” It may not be possible to say in the abstract whether a decentralized constitution is better or worse than a more centralized one. But the proof of the pudding is in the eating.

There are several reasons why turning to empirical observations at this point might seem dangerous, if not perverse. First, even if we found the empirical “pudding” to be tasty, we would not know what it was we were consuming. If theory tells us not to expect a consistent relationship between decentralization and most political and economic outcomes, finding such a relationship in one data set gives us no reason to expect similar results in others. For instance, suppose one found a correlation between decentralization and individual liberty in a sample of countries.