I ride my bike a lot. A couple of months ago, I had an unusual pain in my left knee. When it didn’t go away for a couple of weeks, I decided to see an orthopedist. I had never met this man before in my life, but I went anyway because I trusted that his training and experience in pathologies of the human body would enable him to see things that I or someone else in my family would probably overlook or misunderstand.

I thought about this decision process again today when I read a tweet from someone complaining about how North Korea experts seemed to be coming out of the woodwork after Kim Jong Il’s death. That tweet stung a bit because I’d publicly offered my own quick thoughts on prospects for democratic transition in that country in 2012, based on a global statistical analysis I’d done over the weekend. In my 140-character take, I said explicitly that I was no Korea expert, but I still felt confident that my statistical analysis–and the many years of experience watching and thinking about processes of democratization and state collapse that informed it– gave me something useful to add to the conversation on what happens next in that specific country.

The tweet got me thinking again about whether or not that confidence is warranted. I believe it is. When my knee was hurting, I could have asked my wife or my brother or someone else who knows me really, really well to tell me what was wrong and what to do about it.  I didn’t, though, because as deeply as they know me, they don’t know a lot about knee injuries. My wife can accurately predict what time I’ll wake up in the morning, what I’ll eat for breakfast, and which of the movies at the multiplex I’d most like to see, but she doesn’t have the expertise in orthopedics to know what was causing my knee to hurt or how to make it better.

Similarly, I’ve encountered many people in my professional life who know a lot about the history, culture, and social and political dynamics of specific places, but I remain unconvinced that the deep “local” knowledge they possess automatically positions them to give more reliable diagnoses of, and forecasts about, every situation that arises there.  Sometimes–maybe even usually–a little distance and a lot of comparative perspective can be a good thing.

I can’t tell you the names of anyone in the inner circle of North Korean politics, and Almaty is the closest I’ve ever come to Pyongyang. What I do know, however, is that only one of the nearly 100 transitions to democracy in the past forty years happened in a country that had no previous democratic experience and, according to Freedom House, was as repressive in the preceding year as North Korea is now (Mongolia in 1990). I know that states very rarely collapse, and that when they do, those implosions usually occur in countries where the central government has never established real control over much of its territory (think Chad, not China).

Even very unlikely events can still happen, and North Korea could turn out to be an exception to these striking patterns. If I were betting my house on the outcome, though, I’d be greatly relieved to have that comparative perspective at my fingertips. When it comes to diagnosing and forecasting the outcome of rare political crises, I think deep knowledge of specific cases is overrated. Specialists in what Charles Tilly called recurrent “mechanisms” and “processes” can bring something valuable and important to the party, too.

PS. On Twitter and Google+, I’ve heard from a few people who read this post and thought I was arguing that local knowledge is not helpful. I probably should have made clearer that there are lots of analytical questions for which “local” knowledge is indispensable (e.g., who is likely to succeed a dead leader, or how a population is likely to react to a policy pronouncement). Here, I’m talking narrowly about anticipating the occurrence of rare events (e.g., state collapse) or the dynamics of specific recurrent process (e.g., transitions from authoritarian rule). These are phenomena that are rare or even unique in a single case but recurrent across many. On those, I think comparative perspective is generally more valuable. But that’s me.

PPS. I also wrote a follow-up post on how we’re programmed to give statistical forecasts short shrift.