The Magazine

The Economy and the Election

If the past is any guide, Bush looks formidable.

Apr 12, 2004, Vol. 9, No. 30 • By JAMES PIERESON
Widget tooltip
Single Page Print Larger Text Smaller Text Alerts

Where, then, does the economy stand as we move into the 2004 campaign? The recent "Blue Chip" forecasts that came up with the 4.5 percent real growth figure for 2004 is a survey of more than 50 independent economic forecasters. They expect slightly over 3 percent growth for the first three quarters. Inflation is expected to hold steady at around 1.9 percent. The prime rate of interest, currently 4 percent, is predicted to increase slightly to 4.5 percent by the end of 2004. The unemployment rate, 5.7 percent as of March, is expected to fall nearer to 5 percent by November. The economy is growing at a healthy clip, and is expected to continue to do so into 2005--at least according to the forecasters.

Will this be good enough for Bush to win? As it happens, Ray C. Fair, a professor of economics at Yale University, has developed a statistical model that employs just a few economic variables to make accurate predictions of the presidential vote. His model, first outlined in 1978 in The Review of Economics and Statistics, was initially tested against all presidential elections since 1916. Fair has adjusted and refined the model over the years, and has offered his own predictions (based on his model) for impending presidential races.

His original model was designed to predict the two-party division of the popular vote based on two variables: the percentage change in real per capita Gross Domestic Product (GDP) during the first three quarters of the election year and the annualized change in the rate of inflation (as measured by percentage change in the Consumer Price Index) during the first 15 quarters of the incumbent's term (that is, from the incumbent's inauguration to September 30 of the election year). Though Fair experimented with other variables, such as the unemployment rate, he found that these two provided the greatest predictive power.

In applying his model to successive elections, he observed that while it was accurate in most cases, it was very wide of the mark in predicting the outcome of the Clinton-Bush-Perot three-way race in 1992. The model predicted, prior to the election and based on the economic conditions then prevailing, that Bush would win by a wide margin, with around 54 percent of the two-party vote. The economy during 1992, while not growing rapidly, still expanded at a rate of 2.2 percent (real GDP per capita) with very low inflation. These conditions should have been sufficient to give Bush a comfortable victory. Instead, he lost by about the same margin by which he was expected to win.

Why such a large error for this particular election? Fair noted that, while the economy was expanding during the election year, overall growth had been very slow throughout Bush's four-year term. There were few quarters of robust growth; and there was a mild recession in 1990 and 1991. The lack of good economic news during Bush's term may have generated a sense of gloom or pessimism about the economy that could not be dispelled by modest growth during the reelection campaign. Perhaps voters had formed an assessment of the economy over the entire term and were less influenced by the short-run news than assumed by the original model.

On the basis of such reasoning, Fair added a "good news" factor to his model which he defined as the number of quarters during an incumbent's term in which real GDP expanded by an annual rate of more than 3.2 percent. Such growth would certainly qualify as robust, and would be highlighted to the public by the president and his economic team. On the other hand, in the absence of such good news, an incumbent will be placed on the defensive by a challenger who will link a weak economy to his opponent's ineffective policies. Fair suggests this is what happened to George Bush in 1992, since he could point to only two quarters of good economic news during his entire term (the first quarters of 1989 and 1990), while he presided over several quarters of weak or negative economic reports.

Fair's model thus predicts the incumbent party's share of the two-party vote using three economic variables: (1) growth in real GDP per capita during the first three quarters of the election year; (2) the average increase in the rate of inflation over the 15 quarters of the presidential term up to the election; and (3) the number of quarters during the term in which growth in real GDP per capita exceeded 3.2 percent.

Testing his model against election results from 1916 through 2000 (22 elections), he reports that all of these factors have an effect on the outcome. For every 1 percentage-point gain in real GDP, the incumbent gains .69 percent of the vote, and for every "good news" quarter, he gains .84 percent of the vote. Inflation has the opposite effect: For every 1 percentage-point increase in inflation, the incumbent party loses about .78 percent of the vote.