Not all forecasts of an eventual Bush victory amount to grasping at straws. In particular, one very highly regarded statistical model that predicts presidential elections -- that developed by Yale economist Ray Fair -- calls for a Bush victory by a substantial margin.
Mr. Fair has developed and tested a variety of formulations since the mid-1970s. The version he now uses tries to predict the Democratic share of the two-party vote based on the growth rate of the national product in the second and third quarters of the year of the election, the inflation rate in the preceding two years, and whether or not an incumbent is running for re-election.
His model's track record is impressive: its average error for the last six elections is a mere 1.1 percentage points (though in close elections, such as 1976 or 1968, this does not suffice to prevent mistaken predictions).
This year, the model speaks in stentorian tones: Though the weakness of the American economy is a major election issue and many economic indicators are truely alarming, the numbers that are relevant for Mr. Fair's model are far from fatal -- indeed, they suggest that the president should win by a fairly hefty margin.
Why then is the Bush campaign stalled? Let us first accept the findings of what I consider the best accounts of why people vote as they do. As formulated by Stanley Kelley of Princeton (and extended by his student John Geer, now at Arizona State University) this suggests that almost no one votes on the basis of a single issue. People instead make up their minds by summing up sets of considerations for and against particular parties and candidates, and voting for the one they like the most or dislike the least.
Many voters include considerations about the economy in their lists of issues they care about. As a consequence, while few voters cast ballots simply on the basis of the economy, broad changes in the economy nevertheless affect many voters at the margin. Because it correctly focuses on the marginal issue for many voters, Mr. Fair's model usually works well.
This year, however, things look different in many senses. Because of the intense popular discussions of American economic decline, and perhaps some peculiar features of the current economic recovery (in which, to exaggerate for clarity, profits, but not jobs, are recovering, and everyone is worried about debts), many voters are "reading" the usual economic facts differently. (This is perhaps clearest in the many voters who supported Paul Tsongas or Ross Perot, but these candidates surely have no monopoly on anxiety in the 1992 race.) As a consequence, citizens are going to vote differently from what the model -- which in its present form, does not make explicit use of, say, consumer confidence measures -- predicts. At best, the president can hope for a squeaker, rather than the relatively easy time the model forecasts.