The 2004 U.S. Presidential Forecast:

Figures in parenthesis =  t-scores.
* = statistical significance .05, one tail.
R-sq. = .96 Adj. R-sq. = .94 SEE = 1.52 D-W = 1.28   n = 13 ( presidential elections 1952-2000)

Using most recent figures where GNP change from 4th Quarter 99 through 2^nd Quarter 2004 = 1.32, Bush popularity = 47 (Gallup, July 2004), Incumbent Advantage = 1 as Bush is the elected incumbent president, and Jobs = 3.67 (Job growth from Current Population Survey), we get the following prediction:

2004 vote = 31.16 + .26*Popularity + 1.58*GNPxElect +2.30*IncAdv + .59*Jobs
2004 vote = 31.16 + .26*47 + 1.58*1.32 +.59*3.67
2004 vote = 31.16 + 12.22 + 2.09 + 2.16
2004 vote = 49.93 for George W. Bush.

Our margin of predicted victory for John Kerry is razor thin. The narrow difference makes the race too close to call. This race could go either way.

The model:

Our Jobs Model holds the two-party popular vote share to be a function of July presidential popularity, six-month GNP change (interacted with whether or not the president is running), incumbent party advantage, and job growth over the administration.  All measures are taken mid-summer of the election year. The regression model is estimated over the 1952-2000 period, yielding an R-squared of .96.  With current values of the independent variables, respectively of P = 47, G =1.32, I = 1, and J = 3.67, the forecast for President Bush is 49.93 percent of the two-party popular vote.  Given the Standard Error of Estimate of the model – 1.52 – the race is clearly too close to call.  It could go easily go either way, for Bush or Kerry.

The data.

Dependent variable: Incumbent party's share of the two-party popular vote.

Independent variables:
 

Presidential popularity in July of the election year as measured by Gallup.

GNP x Elect Percent change (nonannulaized) in GNP (constant dollars) from fourth quarter of the year before the election to the 2nd quarter of the election year interacted with whether an elected president is running (scored 1) or not running (scored 0.5).

Jobs = Growth in jobs over first 3.5 years of president’s term. Data (not seasonally adjusted) are the number of employed in Civilian Labor Force (16 years and older) from Bureau of Labor Statistic’s Current Population Survey of Households. Entries calculated by taking (# employed in June of election year - # employed in January of inauguration year) / # employed in January of inauguration year. Incumbent Party Advantage = 1 if incumbent party candidate is the elected president (1956, 1972, 1980, 1984, 1992, 1996) or following a president who died in office (1948, 1964), 0 if incumbent party candidate has a lukewarm association with the previous president (1952, 1976, 1988), - 1 if incumbent party candidate and the president are not united (1960, 1968, 2000).