¹ See, for example, Miller, Warren and Stokes, Donald, ‘Party Government and the Salience of Congress’, Public Opinion Quarterly, XXVI (1962), 531–46Google Scholar; Kostroski, Warren, ‘Party and Incumbency in Postwar Senate Elections’, American Political Science Review, LXVII (1973), 1213–34CrossRefGoogle ScholarHinckley, Barbara, Hofstetter, R. and Kessel, J., ‘Information and the Vote: A Comparative Election Study’, American Politics Quarterly, 11 (1974), 131–58CrossRefGoogle Scholar and Nelson, Candice, ‘The Effect of Incumbency on Voting in Congressional Elections’, Political Science Quarterly, XCIII (1978–1979), 655–78.Google Scholar

² Mayhew, David, Congress: The Electoral Connection (New Haven, Conn.: Yale University Press, 1974)Google Scholar, and ‘Congressional Elections: The Case of the Vanishing Marginals’, Polity, VI (1976), 295–317Google Scholar and Fiorina, Morris, ‘The Case of the Vanishing Marginals: The Bureaucracy Did It’, American Political Science Review, LXXI (1977), 177–81.CrossRefGoogle Scholar

³ Fenno, Richard, Home Style (Boston, Mass.: Little, Brown, 1978).Google Scholar

⁴ Jones, Charles, ‘The Role of the Campaign in Congressional Polities’, in Jennings, M. Kent and Zeigler, Harmon, eds., The Electoral Process (Englewood Cliffs, N.J.: Prentice-Hall, 1966), p. 29.Google Scholar

⁵ Two exceptions are the 1974 National Election Study and the 1968 Comparative State Election Project. Both include questions about Senate elections but do not permit comparisons with House elections.

⁶ For House and Senate re-election rates from 1960 to the present, see Congressional Quarterly Weekly Report, 7 07 1979, p. 1351.Google Scholar

⁷ Since the Senate, unlike the House, is proportioned by states and not by population, a disproportionate number of Senate respondents in the sample will fall in states of large population: in 1978, the states of Michigan, Illinois and Texas. However, with only one exception, dichotomizing Senate respondents into those from the three largest-population states and those from all other states showed no difference in results. The exception is reported in fn. 9 below.

⁸ The traditional recall questions asks, ‘Do you happen to remember the names of the candidates for Congress – that is, for the House of Representatives in Washington – that ran in the district this November?’ It is asked in a post-election survey, typically conducted in December and January following the election.

⁹ The only potentially interesting difference in Senate results between those states (Michigan, Illinois and Texas) largest in population and all others occurs on the question of recognition of the challenger. While 91 per cent of the voters in the three large-population states could recognize and rate the Senate challenger, 81 percent could do so in all other states. Recognition of the incumbent was virtually the same. This fact does not change the results of Table 3 of course: even at 81 per cent rates, Senate challengers are quite competitive with other Senate candidates and quite distinct from House challengers. Nevertheless, the impact of a state's size of population on a Senate candidate's visibility (whether from increased media attention, more money-raising capacity, the greater stakes perceived in the contest) might merit further attention.

¹⁰ The questions asked about seven different kinds of contact: whether respondents had met the candidates, attended meetings where they spoke, talked to their staff, received mail, read about them in newspapers or magazines, heard them on the radio or seen them on television. The same pattern for Senate and House incumbents, challengers, and non-incumbents shown for the three kinds of contacts reported in Table 3 can be shown for the other questions. These three were selected to indicate the range of contact found for candidates and to differentiate personal contact from two different kinds of ‘mediated’ contact.

¹¹ This small effect of party identification on the visibility measures may explain the fact that Democratic non-incumbents in the table are slightly but consistently more visible than Republican non-incumbents. There are more Democratic identifiers than Republicans in both the full sample and among voters in the sample.

¹² Of all respondents asked the ‘anything special’ question, 458 said there was something special and 1,837 said there was not.

¹³ If one removes the small number of voters not able to recognize or rate the Senate candidates or House incumbents, the distributions for the first three columns in Table 9 would be very similar to those reported. Because of the large number not able to recognize or rate the House challenger, both distributions are reported in the table. It should also be noted that respondents placed in the 50–59 degree range are, with only a few exceptions, actually giving 50 degree ratings.

¹⁴ The zero-order correlations (Pearson's r) between affect and party identification and affect and the vote for incumbent or challenger are as follows, where affect is measured as thermometer degrees for the incumbent minus degrees for the challenger, and party identification is measured as in accord with, against, or in no relation to the incumbent: for the House, affect and party = + ·31; affect and vote = + ·57; for the Senate, affect and party = + ·39, affect and vote = + ·64. For the independent effects of affect, controlling for party identification, see Table 10 and the corresponding discussion.

¹⁵ Formally, a dichotomous dependent variable violates the assumption of regression analysis that the error terms are normally distributed, and so may reduce confidence in the results. Therefore the regressions for all contests (see lower portion of Table 10) were checked by use of a probit procedure with no major difference in significance levels found for the coefficients for either Senate or House elections.

¹⁶ Alternative measures of affect were tried including incumbent affect alone, challenger affect, and a ‘collapsed’ relative affect score which assigned incumbents and challengers high, medium, and low ratings and measured the difference. Incumbent and challenger affect alone were each strongly correlated with the vote, although as might be expected less strongly than the relative affect measure selected for the model. The collapsed measure, it should be noted, did considerably less well, possibly owing to the concentration of cases in the middle category. Alternative measures of contact were also tried, including a relative measure of some contact with the incumbent only, with both or with neither candidates, or with the challenger only. For both affect and contact, the decision was to select the relative measure giving the most detail, which was also the one most strongly correlated with the vote.

¹⁷ See fn. 15.

¹⁸ Of the four models reported in Table 10, only the House all-contest model might be suspected of multicollinearity, with some independent variables intercorrelated at levels of ·5 and ·6. (By other rules of thumb, however, multicollinearity would not be said to be present: i.e., the multiple R for the regression very considerably exceeds the intercorrelations between any of the independent variables and the multiple R for the regression exceeds the multiple R when the other independent variables are regressed against incumbency, which is the most highly intercorrelated variable. See also Lemieux, Peter, ‘A Note on the Detection of Multicollinearity’, American Journal of Political Science, XXII (1978), 183–96)CrossRefGoogle Scholar. Nevertheless, for those following the first rule of thumb, it can be reported that all of the coefficients had very low standard errors and yielded extremely large F statistics with four of the coefficients significantly different from zero at the 0·001 level and the fifth just missing significance at 0·001.

¹⁹ Jacobson, Gary, ‘The Effects of Campaign Spending on Congressional Elections’, American Political Science Review, LXXII (1978), 469–91CrossRefGoogle Scholar and Glantz, Stanton A., Abraraowitz, Alan I. and Burkart, Michael P., ‘Election Outcomes: Whose Money Matters’, Journal of Politics, XXXVIII (1976), 1033–41.CrossRefGoogle Scholar