Wednesday, March 10, 2010

causes and correlations

I think there's something in western culture that demands causes for major events. In the years before the Enlightenment, I suppose people were willing to attribute outcomes to God, or the devil. In our more secular and scientific era, we want theories and formulas that demonstrate causation.

I've always been puzzled by news reports attempting to explain why the stock market went up or down on a particular day. Too many factors are at work most of the time.

I hope that my political science colleagues will not de-frock me for saying that we do a poor job of explaining election outcomes, too. Sometimes there are multiple indicators that point the same way and can be combined to show why Jones won and Smith lost. But most of the time the only sure conclusion is that Jones did a better job of getting his supporters to the polls than Smith did.

The media, even more than scholars, want instant analyses, and are willing to ascribe changes in opinion polls to whatever they are focusing on at the moment. Presidential popularity tracks with economic data much of the time, as Brendan Nyhan points out. He makes this telling point: "If/when the economy picks up, Obama's speeches will start "connecting" and everyone will marvel at how effective the White House political team has become."

What's more useful than single point apparent causes are the trends of multiple indicators.

No comments:

Post a Comment