Economic conditions and forecasting of Australian elections revisited

In a recent paper in this journal, Jackman and Marks (1994) examine the relationship between economic conditions and election outcomes in Australia using data from 1949 onwards. Their analysis, however, is in my opinion technically inadequate because they seem to have overlooked the possible existence of autocorrelation in their models. In this paper, I present a re‐examination of some of their models which takes autocorrelation into account and which results in improvements in both the fit of the models and of their overall predictive accuracy. Inflation is shown to be a more definite influence on vote shares than are changes in unemployment rates. Also, I show that Jackman and Marks’ conclusion that the 1993 election was an exceptional one in the way in which economic conditions affected aggregate vote shares is incorrect, and that the election result was actually quite well within the normal range in the extent to which it was accurately predicted by economic conditions.