The Los Angeles Times, in trying to predict the 2004 election, rolls out this fact: “In every election since 1960, the party in the White House lost when the unemployment rate deteriorated during the first half of the year. If the rate improved, the party in the White House won.” Matthew Yglesias takes this prediction apart:
The generic proposition “rising unemployment is bad for incumbents” is so plausible, that one’s inclined to give this claim a pass on first reading, but the specific claim that the causal factor here is the unemployment rate in the first six months of the year before the election smacks of specification-searching.
All too true — predictive models of elections are far from perfect. It’s worth remembering that every election model worth its salt predicted Al Gore clearing 53% in the popular vote in 2000. The one thing everyone could agree on after the 2000 election was that these models obviously needed some rejiggering.
To be fair, it’s the intersection of these models and journalists looking for hooks that produce junk predictions like the one above. The relationship between electoral victory and the six-month unemployment trend in the LAT story sounds good, but it’s the political equivalent of an ESPN commentator saying, “The Patriots have never lost a home playoff game the week after winning on the road by more than two touchdowns.” It’s largely meaningless.
Clearly, unemployment will be a factor in the 2004 election — but it won’t be the only factor (posted by Daniel Drezner).
The Dish 