, ,

An op-ed on political science forecasting in the New York Times by Jacqueline Stevens drew quite the response from bloggers.  One such response appeared on the dartthrowingchimp blog and made a few points worth mentioning.

The premise of the original article was that political scientists waste their time making predictions due to their poor track record of accuracy.  Stevens makes the assertion that attempts to predict political shifts are doomed to fail and do not exceed random chance.

However, the blog argues several points that have merit.  The first is that political scientists do not, in fact, make predictions often enough.  Political scientists are often more focused on describing and explaining the current situation rather than looking ahead to possible changes in the future landscape.  Second, and consequently, the track record is not so much poor as nonexistent.  When Stevens claims no one predicted some event, Chimp replies (a) that in many cases no one was even asked, and (b) in other cases like the Soviet Union, in fact many people did forecast the event, though usually with vague timelines.

Perhaps the most important point made in the post is that there will almost always be a distribution of predictions.  Not all forecasters will agree on the paths of future events and central tendencies should not be confused with a consensus.  Divergence in predictions may uncover important variables for collection.  Additionally, aggregating diverse opinions tends to lead to more accurate forecasts than taking any single one.  So please give us your informed estimate now, particularly on questions where you disagree strongly with the current estimate.