Stock Market Predicts Kerry Win!

Sort of. The sample sizes on things like this are necessarily small, equalling at most the number of presidents. And each election is unique. Still, this CNN story notes that

Presidential elections are usually won by the incumbent party; in the past 104 years, in 26 elections, the party holding the White House has lost it only 10 times.

But those incumbent losses have been heralded every time by poor stock-market performance, either in the four years leading up to the election or in the last year before the election, according to recent research by Ken Tower, chief market strategist at CyberTrader, a unit of Charles Schwab

If, and this is a big if, this reflects some true fact rather than historical accident or spurious causality, then Bush is in trouble — barring a big rally over the next four months:

The stock market under Bush is somewhat historically unique in that the big drop ocurred just as he was entering office, so it’s hard to really blame Bush for that. Sure, some would say that expectations of tax-cut fueled massive deficits and consistently bad policies contributed to the drop, but that’s a bit of a stretch. In fact, the market has done rather nicely since early 2003:

(This is a graph of SPY, a passive exchange traded fund that tracks the performance of the S&P 500)

So, if memories are short, people will vaguely recall that the first half of Bush’s administration was bad news, investment-wise, the next year was rather nice, and the year of the election was neither great nor terrible. Viewed that way, it’s not entirely clear that the stock market under Bush really fits into the historical bin described in the CNN story.

Also, be suspicious of social science results that are phrased along the lines of “Y always follows either X1 or X2 ocurring,” where X1 and X2 are related but distinct concepts. Here, X1 and X2 are 4-year declines in the stock market and weak growth in the year leading up to the election. The problem with this is that X1 and X2 are not chosen at random, nor on any a priori grounds. Instead, someone likely looked an array of stock-related measures and found two for which one or the other always occured in the 10/26 times the incumbent party lost.

Of course, if I look at enough such measures, I can always find a set that will explain anything. For example, suppose I had data on annual average temperature and rainfall in every state over the last 104 years (26 elections). I’m not going to do it, but I guarantee you that I could eventually find a small set of states, possibly two, whose combinations of either above or below average scores on either the rainfall or temperature metric also predicts when the incumbent party will lose the White House. Then I could write headlines like “Kerry Campaign Closely Monitoring Rainfall in Wyoming; Bush Hopes for Heat Wave in New Hampshire.”