As with any long-term change, there are several possible explanations of why the correlation between presidential and House results declined and then returned to a higher level. There is another possible alternative that is statistical in nature. A plausible statistical explanation is that the changes shown in Figures 8.3 and 8.4 may have changed the dispersion of presidential and House votes such that the resulting correlation declined. That is, the decline may just reflect trends in variable variance.

The technical explanation involves the consequences of change. For the first half of the last century, the differences in partisan support between regions were very large. The Democrat's dominance of the South and the high vote percentages for Democrats created considerable variance in presidential and House votes. A correlation is a calculation of the covariance of two variables with the variance of each variable affecting the results. The greater the covariance and the greater the variance of the variables involved, the greater the correlation. It is possible that the variance of one or both variables also might decline as partisan changes occur. As regional differences declined from the 1950s through the 1980s, the results for one or both offices might cluster more in the 40–60 percent range, resulting in reduced national variance for scores for the office. This decline in the standard deviation of presidential results up until 1980 (Figure 8.5) occurred, whereas the House standard deviation has not changed.