Tyler Cowen directs my attention to Chris Coynes' new book After War: The Political Economy of Exporting Democracy. The prognosis isn't good:

What do the data indicate regarding the effectiveness of reconstruction as a means of achieving liberal democracy? In short, the historical record indicates that efforts to export liberal democracy at gunpoint are more likely to fail than succeed. Of the twenty-five reconstruction efforts, where five years have passed since the end of occupation, seven have achieved the stated benchmark, resulting in a 28 percent success rate. The rate of success stays the same for those cases where ten years have passed. For those efforts where at least fifteen years have passed, nine out of twenty-three have achieved the benchmark for success, resulting in a 39 percent success rate. Finally, of the twenty-two reconstruction efforts where twenty years have passed since the exit of occupiers eight have reached the benchmark, resulting in a 36 percent success rate.

It's worth saying, of course, that you're unlikely to ever find the United States actually invading other countries in order to turn them into democracies. Rather, it so happens to be the case that pretty much all of the good candidates for "enemy" status are dubiously democratic regimes, so that rhetorical invocation of democratic values becomes an attractive strategy. The poor record, in practice, of armed democratization is just a further reason to think that such rhetoric should be basically ignored. Sometimes situations may arise where using military force to topple a foreign government is the right thing to do (Germany during World War II and Afghanistan after 9/11 come to mind) and then I think we have an obligation to do our best to bequeath a decent new regime to the place we've conquered. But the prospects for success aren't nearly good enough to make this the reason for launching a war.