Man Without Qualities

Thursday, June 03, 2004

Smackdown In The Battle Of The Predictors: Yale Wins

Scott Elliott's attempts to predict the probable 2004 Presidential election results with a formula that adjusts each state's 2000 margin based on current national opinion polls. With all due respect to Mr. Elliott, to point out that such a methodology is "weak" would be to speak in high euphemism. Can even he take seriously his "prediction" that Kerry will beat Bush, 337 electoral votes to 201, with a popular vote total of 52.87% to Bush's 45.3% - with Kerry picking up every state Gore won plus several of 2000's red states: Ohio, Arizona, Nevada, Florida and Missouri? His formula also awards New Hampshire to Kerry, who is, after all, from the state next door, which may mean something.

But the rest is ridiculous. The Elliott prediction is a dance in a hall of mirrors: current media fussing tends to disturb polling to make it less reliable. Poll users and taker know that - which may have something to do with the fact that they're not taking many serious polls right now. In any event, polls show at best a tiny edge for Kerry even after all the fussing - which is not good news for him.

Does any sensible person seriously believe that the Abu Graid prisoner abuse scandal - which is essentially what is responsible for the President's drop in the polls - is really going to count for more than the overall performance of the economy or is not going to dissipate by November? If that's right, then it will be the first time a prisoner abuse scandal has meant anything much as an election issue, although such scandals come up all the time - as Bob Herbert ably points out. Hopeful Democrats fixating on current polls have offered no explanation why things should be so different this time around. It probably won't happen. It never has.

Some Republicans are worried and some Democrats are hopeful that voters aren't giving enough attention to the positive economy. But how much of a naif does one have to be not to recall that it takes about six months of strong employment numbers before voters seriously get the message that the economy is doing really well - and that six months has not yet run, but the numbers are still very good. It probably will happen. It always has.

Perhaps Mr. Elliott and others could help themselves to come back to earth by taking a quick review of the well-known Fair Model Presidential Vote Equation, which keys off fundamental economic patterns. The Yale professor (Roy Fair) responsible for the model last applied it to information as of April 29, 2004, when it showed Mr. Bush prevailing with 58% of the two-party vote. Is a prisoner abuse fuss 5,000 miles away going to undo that margin? Worse for Democratic fantasists, this is what Professor Fair had to say at that time:

The predictions of GROWTH, INFLATION, and GOODNEWS for the previous forecast from the US model (February 5, 2004) were 3.0 percent, 1.9 percent, and 3, respectively. The current predictions from the US model (April 29, 2004) are 3.2 percent, 2.0 percent, and 3, respectively. In the previous forecast 2004:1 was predicted to be a GOODNEWS quarter, but it turned out not to be. For the current forecast 2004:2 is predicted to be a GOODNEWS quarter, so the total number of GOODNEWS quarters is the same at 3. The economic predictions thus changed very little. The prediction of GROWTH, the per capita growth rate in the first three quarters of 2004 at an annual rate, has increased to 3.2 from 3.0 for the previous forecast, and the prediction of INFLATION has increased from 1.9 to 2.0. These new economic values give a prediction of 58.74 percent of the two-party vote for President Bush rather than 58.68 percent before. The main message that the equation has been making from the beginning is thus not changed, namely that President Bush is predicted to win by a sizable margin.

Note that both quarters 2003:4 and 2004:1 are close to being good news quarters. The non per capita growth rates are 4.1 and 4.2 percent respectively, and with population growing at about 1 percent, these are per capita growth rates of 3.1 and 3.2 percent respectively. The trigger for a good news quarter is 3.2 percent per capita growth. In the above discussion I have not counted 2004:1 as a good news quarter because the growth rate to two decimal places is slightly less than 4.2 percent, but this is a very close call. If both of these quarters are counted as good news quarters, the vote prediction rises from 58.74 percent to 60.42 percent, since each good news quarter contributes 0.837 percentage points to the vote prediction.

If anything, every one of these factors has improved (or, in the case of inflation, not deteriorated materially) since April 29. The reader is invited to calculate his or her own up-to-date prediction using this page. Note to hopeful Democratic fantasists: "GOODNEWS" means the number of quarters in the first 15 quarters of the Bush administration in which the growth rate of real per capita GDP is greater than 3.2 percent at an annual rate - not the amount of positive coverage the Administration is able to squeeze out of the mainstream media. Indeed, growth rates for past quarters have recently been revised upwards.

Absent some very striking new developments, this election will not be decided by Iraq or even national security issues generally - contrary to way too many short-sighted pundits. But, to the extent national security matters in the election, the issue generally favors Mr. Bush in the large.

And, with respect to Iraq in particular, John Kerry has no particular advantage over the President.


Comments: Post a Comment