¹ For comment and footnote citations, see Key, V. O. Jr., Politics, Parties, and Pressure Groups, 5th ed. (New York, 1964), p. 574.Google Scholar Key's chapter on congressional elections, pp. 545–574, includes an analysis of midterm Senate and House elections. Cummings, Milton C. Jr., Congressmen and the Electorate (New York, 1966)Google Scholar, limits his extensive analysis of congressional elections to those held in presidential election years. Cummings is concerned mainly with relating presidential-year House elections to the vote for President in the same or preceding elections. Where midterm elections are mentioned, they are treated as one of a number of factors influencing subsequent presidential-year outcomes.

² Op. cit., pp. 553–571.

³ Campbell, Angus, Converse, Philip E., Miller, Warren E., and Stokes, Donald E., The American Voter (New York, 1960), especially pp. 124, 127.Google Scholar

⁴ Bean, Louis H., The Mid-Term Battle (Washington, D.C., 1950)Google Scholar; Moos, Malcolm, Politics, Presidents and Coattails (Baltimore, 1952)Google Scholar; Silva, Ruth, “A Look into a Crystal Election Ball,” New York Times Magazine, Oct. 10, 1954.Google Scholar

⁵ Campbell, Angus, “Voters and Elections: Past and Present,” Journal of Politics, 26 (November, 1964), 745–757.CrossRefGoogle Scholar See also Key, op. cit., pp. 568, 569.

⁶ The New York Times heading, “November May Tell What's Ahead in Vietnam,” July 17, 1966, IV, 3, is illustrative of a widespread view of the midterm election as a referendum on the conduct of government by the President and his party. See Key, p. 565. Key, p. 569, also remarks that “some [midterm] loss is to be expected…. A sufficiently large loss may indicate a genuine withdrawal of popular favor from the President's party.” The problem lies in interpreting what a “sufficiently large loss” is.

⁷ The figures are taken from CQ, Congress and the Nation, 1945–1964 (Washington, D.C., 1965), p. 63.

⁸ Bean, op. cit., pp. 18–20, averages the difference between the number of House seats won in a presidential election and the number won at midterm in each election sequence from 1928 to 1946. He arrives at the figure of “26 or 30.”

⁹ Key, op. cit., pp. 554, 569–571. Key's evidence supports the larger point made throughout that chapter that congressional elections should be viewed as part of national party voting trends. Cummings, op. cit., pp. 6–27, offers a longer span of similar evidence to support Key's point of the similarity in party voting for presidential and House candidates within presidential-year elections.

¹⁰ Of the 1952 Republican districts in the first quartile (lowest percentage vote for Eisenhower), 18 were lost in 1954; in the second quartile, 3; in the third and fourth, 0. 1956–58 showed the same general tendency with 15 lost in the first quartile; in the second, 17; third, 11; and fourth, 5. Democratic districts for 1960–62 repeated the pattern: first quartile, 5; second, 4; third, 1; fourth, 0.

¹¹ CQ seems to be the only data source available that gives the presidential vote for all congressional districts. Data used for this paper have been taken from “What Happened in the 1956 Elections,” CQ Weekly Report, May 10, 1957, Special Supplement, the first presentation of this material; from “Complete Returns of the 1960 Elections by Congressional District,” CQ Special Report, March 10, 1961; and from “Complete Returns of the 1964 Elections by Congressional District,” CQ Special Report, March 26, 1965. Absolute vote figures are available from 1956 on and percentages from 1952 on.

¹² See CQ Weekly Report, November 11, 1966. The first three sequences, presented in the tables, form the basis for the paper's analysis and conclusions. The 1964–66 data are based on unofficial voting returns without benefit of CQ's expected 1967 identificiation of those districts whose lines have been redrawn since 1964. Nevertheless, because the 1964–66 sequence so clearly supported the paper's original conclusions, it has been included in the text. The support it offers may also serve to demonstrate the predictive possibilities of the analysis.

¹³ See Campbell, Angus and Miller, Warren E., “The Motivational Basis of Straight and Split Ticket Voting,” this Review, 51 (June, 1957), 309.Google Scholar Press, Charles, “Voting Statistics and Presidential Coattails,” this Review, 52 (Dec, 1958), 1041–1050 Google Scholar, on the other hand, assumes that aggregate data can get to an identification of coattails. See also Press, “Presidential Coattails and Party Cohesion,” Midwest Journal of Political Science, 7 (Nov., 1963), 320–325 CrossRefGoogle Scholar, and CQ's analysis of Eisenhower's coattails, CQ Weekly Report, February 3, 1956.

¹⁴ For 1956, 1960, and 1964, a comparison of absolute vote figures has been used to determine whether the President ran ahead or behind. Since the number of votes cast for President is almost invariably larger than the congressional vote, the presidential percentage may be lower than the congressional percentage even though the absolute vote is greater. Since absolute vote figures are not available for 1952, the 1952 data are derived from a comparison of the percentages of the vote that President and House candidates received in a district.

¹⁵ It may be of interest to note that switched districts are more vulnerable to midterm loss than those marginal districts where candidates received under 55% of the vote. In 1954, 60% of switched districts were lost as compared with 39% of marginal districts: in 1958, 67% of switched districts were lost and 55% of marginal districts. As would be expected, those marginal districts which were also switched districts seem even more vulnerable. In 1952, all marginal, switched districts were districts where the President ran ahead. Of the 12, 11 (or 92%) were lost in 1954. In 1958, 100% of the 6 marginal, switched districts were lost.

¹⁶ In 1956 a similarly high percentage of switched districts where the President ran ahead were lost—67%, but no comparison can be made since there were no switched districts where the President ran behind. Insufficient data prohibited analysis of 1960's switched districts.

¹⁷ The breakdown for the other 11 seats lost is as follows. In districts between 60% and 69.9% Democratic congressional vote: where the President ran ahead, 4 were lost out of 57; where the President ran behind, 3 were lost out of 33. In districts between 70% and 100% Democratic congressional vote, 0 were lost where the President ran ahead, 4 were lost where the President ran behind. Three of these last four were involved in Florida redistricting.

¹⁸ See Miller, Warren E., “Presidential Coattails: A Study in Political Myth and Methodology,” Public Opinion Quarterly (Fall, 1955), 353–368 CrossRefGoogle Scholar, for an argument against a large vote gap as a sign of a coattail effect. Campbell and Miller, “The Motivational Basis of Straight and Split Ticket Voting,” op. cit., argue that neither a large nor small vote gap may be taken as a certain sign of a coattail influence.

¹⁹ The question may be raised whether the Republican midterm losses under analysis are merely products of a decreasing GOP trend in those districts, extending beyond the two sets of elections examined. Midterm loss could thus be explained as much by a long-run trend factor as by the short-run coattail factor. However, 52% of 1954's lost districts and 65% of 1958's lost districts increased their congressional GOP percentages in the subsequent presidential election, thus following what Press, “Presidential Coattails and Party Cohesion,” op. cit., p. 325, calls the “expected national trend.”

Some data from other elections can be cited in corroboration of this section's evidence. Silva, op. cit., p. 76, reports that in the 5 midterm elections from 1934–50, of 281 seats lost, only 32 were lost in districts where the congressional candidate ran at least 5 percentage points ahead of the presidential candidate two years before.

²⁰ This explanation seems to contradict the widely held opinion that the results of the 1956 election showed the absence of effective presidential coattails. But the point should be made that we cannot tell how many seats which were carried in the Republican column in 1956 might well have been lost without Eisenhower's presence on the ticket, given the lack of enthusiasm for the GOP as a party in 1956.

²¹ Of those 9 “safe” districts where Eisenhower ran ahead in 1956, which were destined to be lost in 1958, 4 had registered quite “competitive” congressional vote percentages in 1954—50.9%, 52.7%, 52.8%, and 57.6%.

²² Calculation of the correlation coefficient yields r = .55, which is not a very strong relationship.

²³ Cummings, op. cit., p. 55.

²⁴ It should be noticed that the three elections where absolute vote figures are available to derive the number of districts where the President ran ahead suggest a similar rate of loss. In 1956–58, 26% of districts where the President ran ahead were lost. In 1960–62, 25%. And in 1964–66, according to preliminary analysis, 23%.