Ray Fair has constructed a model explaining the incumbent share of the two-party presidential popular vote for the elections of 1916-1984. The economic measures of the growth rate of real per capita GNP and the absolute value of the rate of inflation are singled out as important. In addition, a time trend favoring Democrats and whether an incumbent is seeking reelection are included.
I argue that the Fair model is misspecified. Including the rate of growth in the Dow Jones Industrial Averages significantly improves the fit of the model, eliminates the trend as significant, reduces the income parameter, and strengthens the inflation parameter. In addition, the model remains stable when including the 1908 and 1912 elections, and up-dating to include the 1988 election strengthens the results.
The explanatory power of the model is impressive with only one missed election outcome (1976) between 1908 and 1988, and this miss is easily explained by the impact of Watergate on the Republican candidate, President Ford.