* This study is a revised draft of a paper presented at the Annual Meeting of the American Political Science Association, New York, September, 1969. I would like to thank John Fenton, Frank Munger, Allan Sindler, and Ronald Weber who read and commented on this or earlier drafts. My thanks also to the University of Massachusetts for a Faculty Research Grant which facilitated compilation and analysis of data.

¹ See V. O. Key's analysis of Senate and House elections in relation to the presidential vote, Parties, Politics, and Pressure Groups, 5th ed. (New York: Thomas Y. Crowell Co., 1964)Google ScholarPubMed; Cummings, Milton C. Jr., Congressmen and the Electorate (New York: The Free Press, 1966)Google Scholar; Hinckley, Barbara, “Interpreting House Midterm Elections: Toward a Measurement of the In-Party's ‘Expected’ Loss of Seats,” this Review (September, 1967), 694–700 Google Scholar; Press, Charles, “Voting Statistics and Presidential Coattails,” this Review (December, 1958), 1041–1050 Google Scholar, and “Presidential Coattails and Party Cohesion,” Midwest Journal of Political Science (November, 1963), 320–335 Google Scholar; Miller, Warren E., “Presidential Coattails: A Study in Political Myth and Methodology,” Public Opinion Quarterly (Winter, 1955–1956), 353–368 Google Scholar; and Campbell, Angus and Miller, Warren E., “The Motivational Basis of Straight and Split Ticket Voting,” this Review (June, 1957), 293–312 Google Scholar.

One study of House elections on their own terms is to be found in Jones, Charles, Every Second Year (Washington, D.C.: The Brookings Institute, 1967), pp. 45–71 Google Scholar. And see Jones, , The Republican Party in American Politics (New York: Macmillan, 1965), pp. 76–90 Google Scholar. See also Schoenberger, Robert A., “Campaign Strategy and Party Loyalty: The Electoral Relevance of Candidate Decision-making in the 1964 Congressional Elections,” this Review (June, 1969), 515–520 Google Scholar.

² See citations, note 1.

³ Stokes, Donald E., “A Variance Components Model of Political Effects,” in Mathematical Applications in Political Science, ed. Claunch, John M. (Dallas: Southern Methodist University, 1965)Google Scholar.

⁴ Kabaker, Harvey M., “Estimating the Normal Vote in Congressional Elections,” Midwest Journal of Political Science (February, 1969) 58–83 CrossRefGoogle Scholar.

⁵ For initial studies of voter attitudes in congressional elections, based on nationwide survey studies, see Miller, Warren E. and Stokes, Donald E., “Constituency Influences in Congress,” this Review (March, 1963)Google Scholar; “Party Government and the Saliency of Congress,” Public Opinion Quarterly (Winter, 1962)Google ScholarPubMed; and Representation in the American Congress, forthcoming.

⁶ For utilization of this scheme and the concept of a “normal vote,” see Campbell, Angus et al., Elections and the Political Order (New York: John Wiley & Sons, 1966)Google Scholar.

⁷ Weber, Ronald E. and Munger, Frank J., “Party Identification and the Classification of State Party Systems,” paper represented at the annual meeting of the American Political Science Association, Washington, D.C., September, 1968 Google Scholar. The technique for making these calculations—generally analagous to the procedures employed in the MIT Simulmatics project—calls for synthesizing state electorates by subdividing them into voter-types whose party identification scores are then determined through an additive formula from national survey data. A detailed description of the procedures can be found in Weber, Ronald E., Public Opinion in the States: A Simulation Approach (unpublished doctoral dissertation, Syracuse University, 1969)Google Scholar.

⁸ Percentages of Democratic Party identifiers are as follows:

Party identification results for 1968 from the 13 states of the University of North Carolina's Comparative State Elections Project should soon be available.

⁹ Campbell, Angus, “Voters and Elections: Past and Present,” Journal of Politics (November, 1964), 745–757 CrossRefGoogle Scholar; and Converse, Philip, “The Concept of a Normal Vote,” in Elections and the Political Order, p. 19 Google Scholar.

¹⁰ Froman, Lewis A. Jr., The Congressional Process (Boston: Little, Brown & Co., 1967), p. 170 Google Scholar. For data on House incumbents, see Jones, , Every Second Year, pp. 63–68 Google Scholar.

Such a similarity may surprise some who hold to the traditional view that the Senate—representing whole states rather than small homogeneous districts—is a much more electorally competitive body. One explanation may be that the incidents of electoral mortality commonly cited for the Senate to show its comparative “competitiveness” are drawn from a very few large urban states marked by highly competitive, active parties. The incumbent success rate for New York, California, Pennsylvania, Illinois, Michigan, and Ohio is substantially lower than the overall rate. Of the 24 Senate contests in those states, incumbents won 12 of 19 or 63%—still a definite edge, but a less strong performance than the overall pattern. But these are only 6 of 50 states. The general situation for Senate incumbents is one of electoral safeness.

¹¹ The percentages are based on the 167 simultaneous contests between 1948 and 1966 in the 34 states sufficiently competitive to have elected at least one Senator or Governor from both parties. The turnout differential between Senate and gubernatorial contests, by state for each section, was found to be virtually nil, supporting the assumption that the election results are produced by substantially the same electorate.

The Senate incumbency advantage is not merely a “second-term” advantage, as it might be thought since after two full terms or 12 years of Senate service, the advanced age of incumbents compared to more youthful opponents might depress their success rate. But the data show no decline in success rate for third- or fourth-term tries. Based on all Senate contests in the period in the 34 states, incumbents trying for a second term won 77%; for a third term, 78%; and for a fourth term (37 tried), 81%.

¹² As reported in the text, x² = 7.9; significant at .01, 1df. For presidential years, x² = 7.31, corrected for continuity, significant at .05, 1df; for off-years, x² = 4.10, significant at .05, 1df.

For this study, an incumbent is defined as any Senator in office at the time of the election. Thus the category would include some small number of Senators previously appointed and not previously elected.

¹³ Incumbency was quantified by assigning a “3” to Democratic incumbent victories, a “2” to non-incumbent victories, and a “1” to Republican incumbent victories. The partials for incumbency and Senate deviations controlling for presidential-vote deviations are .42 and for presidential-vote deviations and Senate deviations controlling for in-cumbency, .58.

¹⁴ In order to conduct correlation analyses for all elections, “presidential-vote deviations” for the off-years was quantified as the mean presidential vote deviation for presidential years (+4.7) rather than 0, which would have indicated a pro-Republican deviation. Thus a +4.7 deviation for off-year elections should indicate neither a pro-Republican nor pro-Democratic pull of the presidential vote operating on the Senate election results. The partial for presidential vote and Senate vote deviations controlling for incumbency is .35.

¹⁵ All 34 states were somewhat competitive by definition during the time period, but a check was carried out by ordering the states from high to low Democratic base party vote (thus in order of increasing two-party competition) following the Weber-Munger scheme, with no relationship observable between this ordering and the amount of deviation.

¹⁶ Schlesinger, Joseph, Ambition and Politics: Political Careers in the United States (Chicago: Rand McNally & Co., 1966), pp. 68, 69 Google Scholar.