In a recent post, I mentioned a blog post by Nate Silver which I hadn’t managed to google. It is here
Models Based on ‘Fundamentals’ Have Failed at Predicting Presidential Elections
I think this is a serious effort worthy of publication in a top political science journal (not that anyone asked me). Silver attempted to find all election forecasts in a well defined population (there is a long borin definition in the post) and see how they performed. Notably all of the models showed excellent fit within the sample used to estimate parameters. He is particularly hard on models which use only (mostly economic) fundamentals not supplemented with data from polls.
The “fundamentals” models, in fact, have had almost no predictive power at all. Over this 16-year period, there has been no relationship between the vote they forecast for the incumbent candidate and how well he actually did — even though some of them claimed to explain as much as 90 percent of voting results.
The R-squared of the regression of outcome on forecast is 0.0371.
It is certainly true that Silver has limited respect for prominent academics who attempt to forecast presidential elections using fundamentals. But it seems to me that his lack of deference is justified by solid data analysis.