1 Abramowitz Alan I. and Segal Jeffery A., ‘Determinants of the Outcomes of US Senate Elections’, Journal of Politics, 48 (1986), 433–9; Campbell James E.. ‘Forecasting the 1986 Midterm Elections to the House of Representatives’, PS, 19 (1986), 83–7; Campbell James E., ‘Evaluating the 1986 Congressional Election Forecasts’, PS, 20 (1987), 37–41; Hibbs Douglas A. Jr., ‘President Reagan's Mandate from the 1980 Elections: A Shift to the Right?’, American Politics Quarterly, 10 (1982), 387–420; Jacobson Gary C. and Kernell Samuel, Strategy and Choice in Congressional Elections (New Haven, Conn.: Yale University Press, 1983); Lewis-Beck Michael S., ‘Election Forecasts in 1984: How Accurate Were They?’, PS, 18 (1985), 53–62; Lewis-Beck Michael S., ‘A Model Performance’, Public Opinion, 9 (03–04 1987), 57–8; Lewis-Beck Michael S. and Rice Tom, ‘Forecasting US House Elections’, Legislative Studies Quarterly, 9 (1984), 475–86; Lewis-Beck Michael S. and Rice Tom, ‘Forecasting Presidential Elections: A Comparison of Naive Models’, Political behavior, 6 (1984), 9–21; Lewis-Beck Michael S. and Rice Tom, ‘Are Senate Election Outcomes Predictable?’, PS. 18 (1985), 745–54; Oppenheimer Bruce I., Stimson James A. and Waterman Richard W., ‘Inter- 227–47; Rosenstone S. J., Forecasting Presidential Elections (New Haven, Conn.: Yale University Press, 1983); Tufte Edward R., Political Control of the Economy (Princeton, NJ: Princeton University Press, 1978); Witt Evans, ‘A Model Election?’, Public Opinion, 5 (12–01 1983), 46–9.
2 Campbell , ‘Forecasting the 1985 Midterm Elections to the House of Representatives’, p. 85; Lewis-Beck , ‘Election Forecasts in 1984: How Accurate Were They?’, pp. 60–2.
3 Another possible variable to predict forecast accuracy would be whether a voter had been following the polls. Unfortunately, this question was never asked in these NES data (with the exception of a 1980 panel). However, we do have plentiful data on media usage, which would be the vehicle for poll-watching. The media variable (see Table 1) is significant in only two of the eight contests. Thus, while information about polls is available almost exclusively through the media, the tendency to be a heavy media consumer makes little difference in our model. Furthermore, and perhaps more important, there is evidence that, although poll stories have become a much more common news event over time (e.g., the number of poll stories on American network television news broadcasts more than doubled between 1972 and 1980), the overall ability of the electorate to predict the winner of presidential elections follows no such time trend (see Figure 1).
5 The results given in Equation 1, although based on a rather small sample, are quite robust; that is, they are not dependent upon one or two influential observations. Deleting each observation in turn, then re-estimating, never lowers the goodness-of-fit statistic (R 2) more than a few percentage points.
6 Kramer Gerald H., ‘The Ecological Fallacy Revisited: Aggregate- versus Individual-Level Findings on Economics and Elections and Sociotropic Voting’, American Political Science Review, 77 (1983), 92–111.