Friday, October 22, 2004
I just got my copy of P.S. in the mail. This month the primary focus in the Presidential election, and it includes the official predictions of the most prominent election forecasters. Here are the results, with key variables included:
Abramowitz: approval ratings, GDP growth, length of time party has held the white house. Result = 53.7% for Bush. (Boo!)
Campbell: Labor Day trial heat poll, GDP growth. Result = 53.8% Bush. (Boo!)
Wlezien & Erikson: leading economic indicators, presidential approval, trial heats ( all of these are modified over the course of the year- and the LEI's, approval, and trial heats are all going downhill right now as far as Bush is concerned). Result = 51.7% to 52.9% for Bush. (Boo!)
Holbrook: perception of personal finance (weighted by economic news), presidential approval. Result = 54.5% for Bush. (Boo!)
Lewis-Beck & Tien: presidential approval, GNP growth interacted with incumbency, incumbency advantage, employment growth. Result = 50.1% for Kerry. (Yeah!)
Lockerbie: consumer perceptions of future economy, incumbent party has held white house 2 terms or more. Result = 57.6% for Bush. (Boo!)
Norpoth: primary vote for 2 candidates in New Hampshire, long-term partisan balance, electoral cycle (similar to whether then incumbent party has held the white house for 2 terms or more). Result = 54.7% for Bush. (Boo!)
So of 7 predictions, all but one thinks Bush will win comfortably. This is a little disconcerting (Boo!), so let's dig a little deeper, shall we? If you look carefully you should notice some things.
Almost all of these models use presidential approval and economic numbers. Both of these measurements are suspect. First, everyone uses the Gallup poll because it has been around forever and allows for intercomparability over time. Anyone who has paid attention to the last 2 cycles or read Ruy Teixeira's blog should know all about the problems with Gallup. As for the economic data, GDP is very unreliable these days because of the change in distribution. It used to be that a rising tide lifted all boats. Now GDP growth has a much weaker relationship to personal income or job growth- 3/4 or corporate revenue growth has gone to profits, not wage increases or new hires. So as a political indicator, GDP growth is dead, dead, dead. And you should note that the only model which takes employment into account shows a very narrow Bush defeat.
Finally, forecasting models are notoriously bad at capturing non-systematic events. Economics and popularity play a role in every election. But war, scandal, etc. are difficult to operationalize as numbers, and are not consistent influences from one election to the next. Political scientists hope these sorts of phenomena are captured by approval ratings, but now we are stuck with an unreliable measure of presidential approval.
So take these predictions with a pound of salt. Let's give these guys a break, though. There are all good people tackling a very difficult job with daring. It's just that I don't know anyone who really believes Bush is going to win 54-46 except maybe Bush.
P.S. This is disgusting. These people are desperate and pathetic and have no place in a democracy. They should all be in jail. I have had it.