Presidential Popularity and Negative Voting: An Alternative Explanation of the Midterm Congressional Decline of the President's Party* | American Political Science Review | Cambridge Core (original) (raw)

Abstract

Midterm congressional elections have been generally viewed as relatively sterile affairs marked by reduced turnout, party voting, and the play of politically idiosyncratic forces such as friends-and-neighbors voting. The usual reduction in the number of seats controlled by the President's party, according to the “surge and decline” thesis, simply reflects the departure of short-term forces which presumably benefited the president's party two years earlier. In this study an alternative thesis is proposed which considers midterm election outcomes within the context of the current political environment. Evaluations of the President's performance are found to be directly associated with congressional preferences over a series of midterm elections from 1946 through 1966. Moreover, controlling for party identification, persons who disapprove of the President's performance were generally more likely to vote and to cast their ballot against the President's party than were his admirers to support it. This “negative voting” bias helps to explain why the Democratic and Republican parties have performed more poorly in those midterm elections during which they occupy the White House.

References

1 There is little need to repeat here the long and well known list of election-year studies. Several are particularly important, however, for providing the theoretical grounding for many of our ideas on midterm voting. Principally important, of course, is Campbell, Angus, Converse, Philip E., Miller, Warren E., and Stokes, Donald E., The American Voter (New York: Wiley, 1960)Google Scholar. Also valuable is the collection of essays contained in their subsequent work, Elections and the Political Order (New York: Wiley, 1966)Google Scholar. For an excellent summary of this body of conventional wisdom see Greenstein, Fred I., The American Party System and the American People, 2nd edition (Engle-wood Cliffs: Prentice-Hall, 1970)Google Scholar, chapter 3.

2 This literature, on the other hand, is somewhat less prestigious and conclusive in its findings. Important among the research examining coattails with individual level data is Miller, Warren E., “Presidential Coattails: A Study in Political Myth and Methodology,” Public Opinion Quarterly, 19 (1955–1956), 353–368 CrossRefGoogle Scholar Also, McPhee, William N. and Glaser, William A., Public Opinion and Congressional Elections (New York: Free Press, 1962)Google Scholar.

3 Campbell, Angus, “Surge and Decline: A Study of Electoral Change,” in Elections and the Political Order, pp. 41–43 Google Scholar. Also, Arseneau, Robert B. and Wolfinger, Raymond E., “Voting Behavior in Congressional Elections” (paper delivered at the APSA convention, 1973, Figure 1)Google Scholar.

4 Campbell, et al., Elections and the Political Order, Table 11-1, p. 197 Google Scholar.

5 Silva, Ruth C., “A Look Into a Crystal Election Ball,” New York Times Magazine, 10 October 1954; p. 13 Google Scholar. Key, V. O. Jr., Politics, Parties and Pressure Groups, 5th edition (New York: Thomas Y. Crowell, 1964), pp. 567–574 Google Scholar; and Hinckley, Barbara, “Interpreting House Midterm Elections: Toward a Measurement of the In-Party's ‘Expected’ Loss of Seats,” American Political Science Review, 61 (September, 1967), 694–700 CrossRefGoogle Scholar.

6 Stokes, Donald E. and Miller, Warren E., “Party Government and the Salience of Congress,” in Elections and the Political Order, pp. 204–209 Google Scholar.

7 Ibid., pp. 202–203. Elsewhere Stokes estimates that the proportion of variance in congressional district voting which is explained by national forces is 32 per cent as compared to 49 per cent for district forces. See “A Variance Components Model of Political Effects,” in Mathematical Applications in Political Science, ed. Claunch, John M. (Dallas: Southern Methodist University Press, 1965), pp. 61–85 Google Scholar. Richard S. Katz persuasively argues that Stokes's estimate for the influence of national-level forces is too low; Katz in his correlation analysis revises the estimate upward to 54.6 per cent. See Katz, “The Attribution of Variance in Electoral Returns: An Alternative Measurement Technique,” American Political Science Review, 67 (September, 1973), 817–828 CrossRefGoogle Scholar. This estimate has been statistically corrected to 45 per cent. See communication by Achen, Christopher, American Political Science Review, 68 (March, 1974), 1272–1273 Google Scholar. It should be remembered that these estimates of the sources of the congressional vote include presidential election years.

8 Campbell, , Surge and Decline …, pp. 55–57 Google Scholar; a similar theme is also given in his “Voters and Elections: Past and Present” in Journal of Politics, 26 (November, 1964), 745–757 CrossRefGoogle Scholar. For both Campbell and Hinckley (whose measurement is the loss and gain of seats) the systematic partisan shift becomes the residual difference between the predicted normal vote (a seat loss) and the actual results. Defining the error term as the partisan shift prevents us from testing the accuracy of surge and decline and ignores the causes of a national partisan movement.

10 Arseneau and Wolfinger, passim.

12 “The Election,” Time Magazine, 13 November 1950, p. 19 Google Scholar.

13 Moley, Raymond, “What Happened,” Newsweek, 10 November 1958, p. 61 Google Scholar.

14 Burns, James MacGregor, Roosevelt: The Lion and the Fox (New York: Harcourt, Brace and World, 1956), pp. 198–203 Google Scholar; and Leuchtenburg, William E., Franklin D. Roosevelt and the New Deal (New York: Harper and Row, 1963), pp. 822–823 Google Scholar.

15 Schlesinger, Arthur M. Jr., A Thousand Days (Boston: Houghton Mifflin, 1965), pp. 822–823 Google Scholar. Kennedy, Robert F. also mentions this consideration in his account of the crisis, Thirteen Days (New York: Norton, 1969)Google Scholar. Also see Allison, Graham, “Conceptual Models of the Cuban Missile Crisis,” The American Political Science Review, 63 (September, 1969), 712–714 CrossRefGoogle Scholar.

16 Barber, James D., Presidential Character (Engle-wood Cliffs: Prentice-Hall, 1973), pp. 278–280 Google Scholar.

17 Moley, Raymond, “The Presidency,” Newsweek, 11 November 1946, p. 116 Google Scholar.

18 “Correspondents' Election Preview,” Newsweek, 6 November 1950, p. 25 Google Scholar. Also, Key, pp. 563–568.

19 Whatever misfortune or shortcomings Eisenhower may have displayed in his dealings with Congress or his own cabinet, it is difficult to gainsay the General's adroitness as a campaigner, and the 1954 congressional elections are no exception. The following passage reveals that he recognized that his public prestige was important to others, and he intended to exchange it for his own profit. Joseph Meek, the Republican senatorial candidate running against the incumbent Illinois Democrat Paul Douglas, had publicly opposed certain features of Eisenhower's foreign aid program during the campaign. Eisenhower describes the affair in his memoirs:

In April, I had told him I would not back him until he announced that he would support my program. Consequently, I remained silent on his candidacy until he gave public assurance that “you can count on my loyalties and my support as the junior senator from Illinois.” On August 12 1 was photographed with seventeen candidates, not yet members of the Congress, who because they promised to support my program had won the approval of the Citizens for Eisenhower, who were again readying themselves for action in the campaign. I wanted and needed a Republican Congress; but I could see no sense in working for office-seekers who were ready to object to every proposal I made. Eisenhower, Dwight D., Mandate for Change (Garden City, N.Y.: Doubleday, 1963), p. 433 Google Scholar.

To the degree that politicians believe the first premise that the President's popularity is important to their own success, the incumbent President possesses a “bargaining advantage” which, as we see here, can be used to purchase future congressional loyalty.

20 Sundquist, James L. provides an interesting analysis and persuasive discussion on this point in Politics and Policy: The Eisenhower, Kennedy and Johnson Years (Washington, D.C.: Brookings Institution, 1968), pp. 452–466 Google Scholar. During the campaign Republican senatorial and congressional candidates publicly disavowed allegiance to President Eisenhower's economic programs. On this point see Key, pp. 569–570.

21 Nixon, Richard M., Six Crises (New York: Doubleday, 1962), p. 310 Google Scholar. Campbell et al. in The American Voter find that 40 per cent of the families felt a direct impact of the recession and that their responses to the economy correlated with evaluations of the Eisenhower administration and 1958 congressional vote intentions (pp. 386–391).

22 The only systematic investigation of the effects of the strategies of congressional candidates seeking to dissociate themselves from their party's national leadership was performed by Robert A. Schoenberger for the 1964 presidential election. Republican candidates who severed any connection with the Goldwater candidacy received on the whole a larger share of the congressional vote than did those congressional candidates who maintained party loyalty. “Campaign Strategy and Party Loyalty: The Electoral Relevance of Candidate Decision-Making in the 1964 Congressional Elections,” American Political Science Review, 63 (June, 1969), 515–520 CrossRefGoogle Scholar. For a more general analysis of politicians' beliefs about the importance of campaign strategies see Kingdon, John W., Candidates for Office: Beliefs and Strategies (New York: Random House, 1969)Google Scholar.

23 The author is cunently investigating the degree to which career decisions of politicians are based on the President's public standing. Entry onto a number of rungs in the career ladder may be affected. The number of candidates seeking each party's congressional nomination, the percentage of congressmen from each party trying to become a senator or governor, and the percentage who decide to retire from the House are currently being examined. If presidential popularity is an important consideration in career movement we should find an inverse correlation between the Democratic and Republican percentages on the above measures. If a low presidential popularity is found to discourage viable candidates from the President's party from attempting to get on or move up the office ladder, while at the same time encouraging candidacies from the opposition, the overall congressional vote could, in part, reflect systematic party differences in the quality of the candidacies.

24 Bryce, James, The American Commonwealth (Norwood, Mass.: Macmillan and Company, 1913), p. 128 Google Scholar. This “parliamentary” perspective of British observers continues today. See Polsby, Nelson W., “Review Article: The British Science of American Politics,” British Journal of Political Science, 2 (Oct., 1972), 492 CrossRefGoogle Scholar.

25 They continue, “The crux of our theory is that changes in the party balance are induced primarily by negative rather than positive attitudes toward the party controlling the executive branch of federal government” (The American Voter, p. 554).

26 Kanouse, David E. and Hanson, L. Reid Jr., Negativity in Evaluations (Morristown, N.J.: General Learning Press, 1972)Google Scholar. The authors summarize the experimental research: “It seems that negativity biases occur against a backdrop of perceived bliss–indeed perhaps because of it. Given that most people perceive the world as a predominantly positive place, there are a number of reasons why one might expect them to weigh positive information rather more lightly than negative. First, there is the well-known judgmental anchoring, or contrast, effect. In a world of ointment the fly seems bad indeed. Second, if most choices and behavior-relevant evaluations are made from a range of general positive alternatives, it is simpler and less effortful to sort the alternatives on the basis of their few negative aspects rather than the many positive ones,” p. 10. Another review of a negativity bias is Jordan's, Nehemiah, “The ‘Asymmetry’ of ‘Liking’ and ‘Disliking’: A Phenomenon Meriting Further Reflection and Research,” Public Opinion Quarterly, 29 (Summer, 1965), 315–322 CrossRefGoogle Scholar.

27 Perhaps no better recent illustration of this phenomenon exists than the case of Hubert Humphrey in 1968. Throughout the 1950s and mid-1960s Senator Humphrey, co-founder of the Americans for Democratic Action, was the chief originator and advocate of liberal social policies in the Senate, yet because of his association with President Johnson's Vietnam war policies, Humphrey suffered a severe drop in popularity within the liberal community during his 1968 bid for the presidency.

28 Although negative evaluations may be more instrumental, positive evaluations are generally preferred and more pervasive. Thus, incumbents may be able to overcome negative voting largely as a result of their generally positive evaluation by the citizenry. See Sears, David O. and Whitney, Richard E., Political Persuasion (Morristown, N.J.: General Learning Corporation, 1973), pp. 10–17 Google Scholar.

29 Bloom, Howard S. and Price, H. Douglas in a time-series study of the effects of short-run economic change on congressional voting find a negativity bias operating. “Political Parties are ‘punished’ by the voters for economic downturns but are not ‘rewarded’ accordingly for prosperity.” American Political Science Review, 69 (December, 1975), 1240–1253 CrossRefGoogle Scholar. Other studies use the party of the administration as the benchmark for testing the effects of economy on short-run political change. See Kramer, Gerald H., “Short-term Fluctuations in U.S. Voting Behavior, 1896–1964,” American Political Science Review, 65 (March, 1971), 131–143 CrossRefGoogle Scholar.

30 The survey question asks which party the respondent would like to see “win his state in the next congressional election.” Responses to this item correlate at .80 (Pearson r) with reported actual vote..

31 Nixon's 1970 congressional election campaigning has been described by Evans and Novak as a “white heat.” See Evans, Rowland Jr. and Novak, Robert D., Nixon in the White House: The Frustration of Power (New York: Random House, 1971)Google Scholar.

32 At the presidential level, Richard W. Boyd in his analysis of the 1968 presidential election found such a transference from President Johnson to his party's nominee, Hubert Humphrey. See Boyd, , “Popular Control of Public Policy: A Normal Vote Analysis of the 1968 Election,” American Political Science Review, 66 (June, 1972), 440 CrossRefGoogle Scholar.

33 Harding, John, “The 1942 Congressional Elections,” American Political Science Review, 38 (February, 1944), 41–58 CrossRefGoogle Scholar.

34 Each poll was taken within one month of the election. Unfortunately, in 1970 A.I.P.O. failed to ask the presidential popularity item in its postelection survey.

35 Norval D. Glenn has thoroughly examined this issue and determined that any systematic sampling bias differences are relatively minor. Polls taken before 1950 used quota sampling and as a result slightly under-represented the lower income and poorly educated citizenry. “Problems of Comparability in Trend Studies with Opinion Poll Data,” Public Opinion Quarterly, 34 (Spring, 1970), 82–91 CrossRefGoogle Scholar.

36 It was necessary to follow A.I.P.O.'s classification of southern states. They are Virginia, North Carolina, South Carolina, Georgia, Florida, Kentucky, Tennessee, Alabama, Mississippi, Arkansas, Louisiana, Oklahoma, and Texas. All others are classified as non-Southern. In the table below we can see that the South contained the bulk of non-competitive congressional districts.

37 Within the South, the percentage point differences were –18 points for 1966. For 1954 and 1958 there were too few southern Republicans who disapproved to permit computation of percentages. Note that the higher turnout among disapproves in both regions was particularly strong during the last two midterm elections for which there is evidence. If this indicates a forming trend, it may reflect another symptom of the much belabored and recurrent theme of weakening party ties. With party identification becoming less important, attitude conflict involving party identification is less likely to be resolved in its favor. For a sampling of this literature see Burnham's, Walter Dean “The Changing Shape of the American Political Universe,” American Political Science Review, 59 (March, 1965), 7–28 CrossRefGoogle Scholar. This thesis is thoroughly explored in Converse, Philip E., “Change in the American Electorate,” in The Human Meaning of Social Change, ed. Campbell, Angus and Converse, Philip E. (New York: Russell Sage Foundation, 1972), pp. 263–337 Google Scholar.

38 Independents in the present paper include respondents who when questioned further acknowledged leaning toward one of the two parties. Thus some of the “effects” of presidential popularity reported are open to the suspicion of partial spuriousness from concealed partisanship. Therefore, the analysis was replicated where possible using this purer group of independents; except for added random variation due to the reduced sample size, the same conclusions hold. Including leaners among identifiers would have merely transferred this issue of hidden spuriousness to the party identifiers.

39 yThe relationships between turnout and popularity are similar for the South and have been included in Table 4 to increase the sample size. The reported presidential vote for 1964 was unavailable, and therefore, the 1966 data had to be excluded. Of course, evidence based on recall two years earlier must be looked upon with circumspection. In this instance a systematic recall bias–as compared to fading memory which distributes randomly–does not appear to be a problem.

40 At this point some bemused readers may be entertaining the idea that “approval” and “disapproval” do not measure sentiments of equal intensity. Perhaps disapproval represents an extreme response volunteered only when the respondent feels strongly on the issue. After all, when a new president enters office without a track record a large majority of the public prefer to approve his job performance rather than withhold judgment. If approval frequently substitutes for no opinion then it is not too surprising to find it relatively uninfluential in guiding voting choices. Two pieces of evidence challenge this argument: First, from all of the available surveys, when respondents were asked their opinion on a ten-point “strongly like” to “strongly dislike” scale, a majority who disapproved of the President's job performance, nonetheless, placed themselves on the liking end of the scale. This suggests that disapproval should not be viewed as registering extreme negative evaluation. Second, Richard Boyd, using a refined five-point job performance index from the 1968 SRC survey, finds a very similar relationship, with the benefits accruing from strong approval (“very good”) not nearly so great as the harm caused by strong disapproval (“very poor”). Boyd, , “Popular Control of Public Policy…,” Figure 9, p. 440 Google Scholar.

41 Mueller, John E., War, Presidents and Public Opinion (New York: John Wiley and Sons, 1973)Google Scholar, chapter 10.

42 During Republican administrations this figure is slightly lower at 1.3 percentage points.

43 Tufte, Edward, “Determinants of the Outcome of Midterm Congressional Elections,” American Political Science Review, 69 (September, 1975), 812–826 CrossRefGoogle Scholar.

44 Lydon, Christopher, “The Awful Arithmetic,” New York Times, April 21, 1974 Google Scholar; Apple, R. W. Jr., “Election Problem of G.O.P. Essayed,” New York Times, February 9,1974 Google Scholar.

45 Regression analysis further reveals that a ten-percentage point decline in President Nixon's popularity reduced preference for his party's candidates by 2.67 percentage points. Because the percentages in Figure 4 refer to the preferences for the entire public–with nonvoters included–they may not in fact reflect the potential vote. Probably inclusion of nonvoters who disproportionately favor the Democratic party significantly underestimates Republican strength, despite the fact the negative evaluations are more stimulating. None of the estimates appear to be contaminated by serial correlation according to the Durbin-Watson test.

46 To see if the Watergate affair has had any special influence on Republican preferences other than through Nixon's popularity, the analysis was repeated with the data since the election of 1972 excluded. The projection of the slope for the pre-Watergate data estimates closely the proportion favoring a Republican victory given the reduced support for President Nixon.

47 There is some evidence which argues this view, however. In a late August Gallup survey, respondents were asked, “Would you say that your choice for President is more a vote FOR your candidate or AGAINST the other candidate?” Of those with an opinion, only 28 per cent viewed their vote as against the other candidate ( The Gallup Opinion Index, October, 1972, p. 7 Google Scholar).

49 There is some evidence of a close association between voting preferences and evaluations of the chief executive. For Great Britain, see Goodhart, C. A. E. and Bhansali, R. J., “Political Economy,” in Political Studies, 18 (March, 1970), 43–106 CrossRefGoogle Scholar. For France, which is more presidential than parliamentary, see Charlot, Jean, Les François et De Gaulle (Paris: Plon, 1971), pp. 215–238, 245 Google Scholar.

50 Susan E. Howell compares legislator visibility for five systems in “System Effects on Legislator Visibility in Five Democratic Countries” (paper presented at the 1974 APSA meetings).

51 Greenstein, Fred I., “Popular Images of the President,” in The Presidency, ed. Wildavsky, Aaron (Boston: Little, Brown, and Co., 1969), pp. 287–295 Google Scholar.

52 Kernell, Samuel, Sperlich, Peter W., and Wildavsky, Aaron, “Public Support for Presidents” in The Presidency, 2nd ed. (Boston: Little, Brown and Co., 1975), pp. 148–181 Google Scholar.

53 Collapsing categories of education and checking for curvilinearity failed to uncover any systematic relationships hidden by the statistics.

54 I wish to thank Professor Richard Brody for allowing analysis of restricted data.

55 Controlling for party–Strong Democrat, Weak Democrat, Independent, Republican (the strong and weak categories had to be combined because of a small N)–the relationships reported in Table A-2 remain essentially unchanged.