Images and Voters' Decision-Making Processes

Citation:

Dan Nimmo (1975) ,"Images and Voters' Decision-Making Processes", in NA - Advances in Consumer Research Volume 02, eds. Mary Jane Schlinger, Ann Abor, MI : Association for Consumer Research, Pages: 771-782.

Advances in Consumer Research Volume 2, 1975      Pages 771-782

IMAGES AND VOTERS' DECISION-MAKING PROCESSES

Dan Nimmo, University of Tennessee

The language of contemporary political campaigns frequently likens the voter to a consumer in the political market place. This paper explores various facets of voters' political images as components of their decision-making processes as consumers. Using data from a nationwide survey conducted in the 1972 presidential election, it relates cognitive, affective, and conative aspects of voters' self-images, candidate images, and party images to voting behavior. Correlation, factor analytic, and stepwise regression techniques indicate that voters' affective responses to the major party candidates, rather than cognitive and conative considerations, were the key components in the electorate's 1972 decision-making processes.

During the last century the language of political campaigns possessed a martial tone. "Cadres" of party workers, obeying party "captains" organized in a chain of command, "drilled" phalanxes of "loyal" voters and scorned "traitors" and "turncoats." Shortly after the turn of the century, according to historian Richard Jensen (1968), a merchantilist style supplanted older political ways. In the new argot what was once a battlefield became a "market place," elections "sales campaigns," advertising the "pitch," and voters "consumers." Accepting more recent lexicon, this paper explores what considerations enter voters' decision-making processes when they act as consumers in the political market place.

To bridge the gap between what political scientists know about electoral behavior and the notion that voters are consumers, we use a concept in vogue among professional campaigners and journalists as well as political scientists and economists--i.e., that of image. Published popular and scholarly commentaries point to the importance of the images of political candidates to the electorate's voting decisions: In his widely read account of the 1968 presidential election McGinniss (1969) offered the gist of image politics, that "style becomes substance" where "the Medium is the massage and the masseur gets the votes [p. 302." A specialist in electronic politiking added to claims made for candidate images by a portrayal of the techniques of televised packaging of images (Wyckoff, 1968). A former Chairman of the Federal Communications Commission, Newton N. Minow (1973), has declared that "television's most significant political characteristic probably is its ability to present an image of a politician--providing an indication of his character and personality [p. 6]." And political scientists Natchez and Bupp (1968), drawing upon voting in presidential elections from 1952-1964, concluded that the "best single predictor of voting behavior . . . is candidate image [p. 411]."

An understanding of the relation between images and voters' decision-making processes requires, first, a definition of image and, second, a model of voting decisions. For the definition it is convenient to turn to Boulding's discussion in his provocative monograph (1956). Following Boulding we say that each person possesses an image, or series of images, of the world. Each image consists of the person's subjective understanding of things--i.e., of what he or she believes to be true about something, likes or dislikes about it, and proposes to do about it (what social psychologists such as Smith, 1947, refer to as a person's cognitions, affects, and conations). This use of image parallels the definition of "brand image" in advertising and market research. Thus, Downing (1964) defines a brand image as "a constellation of feelings, ideas and beliefs associated with a brand by its users and non-users mainly as a result of experience of its advertising and performance [p. 14]." As with brand images, political images do not exist apart from the political objects (or their symbolic surrogates) that stimulate political thoughts, feelings, and inclinations. In sum, an image is a human construct imposed upon an array of perceived attributes projected by an object, event, or person. Thus, for instance, a candidate's "image" consists of how voters perceive him/her,perceptions based upon both the subjective appraisals made by the voters and the messages (utterances, attributes, qualities, etc.) transmitted by the candidate.

Assuming such a definition of image will serve us, what model of voting decisions will suffice? Numerous voting studies in recent decades suggest such a model (Kelley and Mirer, 1974) especially those of presidential elections carried out by the Survey Research Center/Center for Political Studies (SRC/CPS) of the University of Michigan. The first major volume in the Center's voting studies, prepared by Campbell, Gurin, and Miller (1954), was an extensive analysis of the 1952 presidential election. It identified major components of the voter's "motivation" to select one candidate over another. The authors pointed to three components of that motivation--the voter's party identification (whether he thought of himself as a Democrat, Republican, or Independent), orientation toward election issues, and orientation toward the competing candidates. Campbell, Gurin, and Miller demonstrated that these three forces directly related to candidate preference so that, for example, the stronger a person considered himself to be a Republican, and/or liked the Republican candidate, and/or preferred Republican issue postures, the more likely he would vote for Eisenhower. The less distinct his stand on any of these three factors, the more likely his choice of the opposition's party candidate (or, in many instances, the less likely he would vote at all).

Here, then, was an early statement of three key variables--party, candidate, and issue orientation--that enter the voting decision. In a second volume Campbell, Converse, Miller, and Stokes (1960) raised party identification to a prime position. Party identification--or,more aptly, the partisan self-image of the voter (Butler and Stokes, 1969)--acts as a predisposing attitude, usually formed early in life, and a perceptual screen influencing how each voter sees and evaluates candidates, issues, and political parties. In short, partisan self-images influence other political images. Moreover, partisan self-images are long-term influences on the voter farther removed from the voting act than other short-term influences such as the voter's images of the candidates and parties or issues of domestic and foreign policy (Campbell, Converse, Miller, & Stokes, 1966).

The question that concerns us within this context of long-term and short-term components is what is the relative impact of voters' images in making electoral decisions? Generally, as Natchez points out in his thoughtful review and critique (1970) published studies of the SRC indicate that "the most powerful components of the electoral decision process are those which capture attitudes toward the competing candidates [p. 575]." In the early Presidential elections following World War II domestic nor foreign policy issues had a great influence on election outcomes; also, attitudes toward the political parties and other political groups were not generally associated with shifts in voters' preferences between elections. Rather, candidate image was the primary motivator of voting decisions. Although the pattern has changed some, candidate images retain considerable potency. For example, using SRC/CPS data from the 1960 and 1964 presidential elections, Natchez and Bupp (1968) found that issues clearly had come to be of considerable influence on voting, but candidate image remained more important.

Along with Natchez and Bupp, studies of more recent presidential elections also indicate an increased role for issues in shaping voting behavior, yet these investigations confirm that candidate image remains a principal short-term force. In 1968, for instance, the Comparative State Election Project (Kovenock, Beardsley, & Prothro) focused upon voting behavior in Southern states (combining several statewide surveys with a nationwide sampling). By comparing respondents' positions on selected issues with their perceptions of the candidates' stands on the same issues, the analysis explained a considerable portion of the variation in voters' preferences for Nixon, Humphrey, and Wallace. Although issues were clearly important in 1968, however, a key component of the voting decision was the voters' images of the candidates' issue stands along with their perceptions of other candidate qualities. And, an analysis of the 1972 presidential election (Miller, Miller, Raine, & Brown, 1973) also found that how warmly or coldly voters evaluated the candidates and the perceptions of voters of the candidates' issue stands compared with their own positions were major factors in helping voters make up their minds.

Natchez and Bupp observed (1968) that the long-term component of voting-partisan identification--was, to be sure, "causally" associated with candidate image so that "people do not perceive candidates through neutral eyes" but "seem to take their focus from previously established identifications [p. 446]." Yet, partisans sometimes have more positive images of the opposition party's candidate than they do of their own. Boyd (1969) demonstrated, at least for the presidential elections of 1956-1964, that when partisans have images of the candidates that are inconsistent with their partisan self-images, this constitutes the major factor explaining their defection from party loyalty in voting. Using SRC/CPS data Boyd derived measures of consistency between the long-term component of a person's voting decision, party identification, and each of three short-term components--i.e., the voter's image of the candidates, of the political parties, and of selected policy issues. Boyd also calculated a measure of the degree to which voters of varying intensities of partisan self-image voted in each election in keeping with their partisanship or defected to the opposition. Comparing the defection rates of voters with the consistency of their partisan self-images and images of candidates, parties, and issues he found that images of the candidates was the principal statistical explanation for voting defection.

Not only does contemporary research indicate that candidate image is the most important short-term force contributing to partisan preference and defection, but it also suggests that how people perceive candidates is the principal determinant of whether they will split their tickets and how. In 1972 Miller and Jackson (1973) conducted a probability sampling of registered voters in three counties of southern Illinois. Their focus was upon the factors related to voters splitting their tickets between the two major political parties in contests for the presidency (Richard Nixon vs. George McGovern) and the governorship (Richard Ogilvie vs. Daniel Walker). Respondents evaluated each of the four candidates using various seven-point semantic differential scales. With the same scales respondents also rated traits of their Ideal President and Ideal Governor. The researchers derived measures of the proximity between respondents' evaluations of each presidential candidate and their image of the Ideal President; they also calculated proximity measures for the gubernatorial candidates. Miller and Jackson found that candidate images (that is, how closely each voter's image of any candidate approximated the voter's image of the "ideal" office-holder) enabled them to predict straight and split-ticket voting with considerable success.

Running through the bulk of major recent research relating various components of voting decisions to electoral behavior, then, has been a view that "candidate images" (or, more accurately, voters' images of candidates) are powerful short-term forces in the electorate's decision-making. The remainder of this paper examines that possibility. Specifically it seeks to identify what dimensions in the relationship between voters' issue, candidate, and party images influence voting decisions--i.e., are beliefs about candidates/party/ issues (cognitions), feelings (affects) toward these political objects, or inclinations (conations) regarding them most influential? Using data from the 1972 presidential election study of the CPS (based upon a nationwide probability sample) it is possible to measure five selected components of the voting decision--partisan self-images and ideological self-images of voters, issue orientations of the electorate, and voters' images of the candidates and of the political parties. By correlating and factor analyzing such measures and conducting a regression analysis we can describe the relative impact of each component upon voting.

Of the five components, three pertain to a voter's self-image; i.e., what kind of partisan the voter perceives he is, what kind of liberal or conservative he thinks he is, and what stands he takes on specific policy questions. Our interest in these is in the voter's perceptions of self, not of the candidates or the political parties. The SRC/CPS has been measuring the distribution of partisan self-images in America for more than two decades by asking cross-sections of American citizens a series of questions regarding whether they "usually think" of themselves as Republicans, Democrats, Independents, or something else. This practice in 1972 produced a seven-point scale of partisanship (ranging from a score of "l" for strong Democrat through "3" for a strict Independent to "7" for strong Republican). Also in the 1972 election survey respondents placed themselves on a seven-point scale measuring their ideological self-images: viz., extremely liberal, liberal, slightly liberal, moderate/middle of the road, slightly conservative, conservative, and extremely conservative. Finally, we use a measure for the voter's issue stands. Respondents in the 1972 survey, again using a seven-point scale, revealed their positions on several policy issues. We employ nine issues in this analysis covering a diverse gamut of social, economic, foreign policy, civil rights, and civil liberties matters: (l) whether the federal government should guarantee that every person has a job and a good standard of living, (2) whether income taxes should be progressive or everyone pay the same portion of their income regardless of how much they make, (3) the question of legalization of the use of marijuana, (4) busing to achieve racial integration in public schools, (5) government vs. private coverage of medical and hospital expenses, (6) immediate withdrawal from Vietnam vs. efforts to achieve complete military victory, (7) total vs. no government action to battle inflation, (8) protection of the legal rights of those accused of committing crimes, and (9) whether the federal government should make every effort to improve the social and economic position of blacks and other minority groups. On each seven-point scale for the respondent's self-placement on an issue, the score of "l" represented a position favorable to governmental action, progressive taxation, legalization of marijuana, busing, withdrawal from Vietnam, or Protection of the rights of the accused. A score of "7" indicated a position at the opposite extreme. By combining respondents' self-placements on each of the nine issues, we have a composite seven-point index of each voter's issue orientations.

In employing 1972 survey data to examine the relative impact of candidate images we use several measures of cognitive, affective, and conative aspects of images. For the cognitive aspect, i.e., belief about each candidate regardless of how one feels about or tends to react toward him, there are two sets of measures. First, in addition to being asked to identify what kind of liberal or conservative each respondent considered himself, the respondent-using again a seven-point scale--designated what he believed to be the respective ideological leanings of Richard Nixon and of George McGovern. Second, on each of the nine issues listed above, each respondent designated where he believed each candidate's stand was on the same seven-point scale he used to declare his own position. As in the case of respondents' self-placement on issues, we use a composite index to summarize their perceptions of each candidate's stands across all nine selected issues.

The affective aspect of candidate images consists of two measures of voters' feelings about contenders and likes and dislikes. First, we use the respondents' ratings of Nixon and McGovern on a "feeling thermometer" that taps the coolness or warmness of feelings on a scale of 0-100 degrees. Second, using the responses of people to an open-end standard SRC/CPS question asking for likes and dislikes about each candidate, we have a candidate index, a 13-point scale on which the higher the score the more pro-McGovern the respondent's affect. The index was calculated by taking, for each respondent, the number of things he/she liked or disliked about each candidate (up to three of each were coded in the survey); pro-Republican plus anti-Democratic responses were subtracted from pro-Democratic plus anti-Republican responses and the range of scores (-6 to +6) converted to a thirteen-point scale.

The conative aspect of images is more complex and warrants detailed consideration. Three decades ago Smith (1947) differentiated between three aspects of attitudes, or images--cognitions, affects, and conations. He stressed that a person's beliefs and feelings about any political object have much to do with what he proposes to have done about it. What one wants done Smith treated as conative and labeled "policy orientation." Following Smith and others, Harding, Proshansky, Kutner, and Chein (1969) examined the conative aspect of prejudice but broadened the notion to include people's "action orientations" as well as their ideas about "what should be done." They reported various studies measuring conation as the social distance perceived by people between themselves and members of various ethnic groups. This idea of distance, or proximity, we incorporate into our measurements as the conative aspect of images. Thus, with respect to a political candidate, the conative aspect is more than the voter's policy orientation in the sense of his intention to vote for or against the candidate. It refers also to how close a voter perceives himself to the political object, i.e., his proximity along specific dimensions (or with reference to specific matters) of sufficient salience to trigger for him an idea of what he proposes to do about the object. The 1972 survey provides two kinds of data to construct measures of proximity. First, respondents placed both themselves and each candidate on a seven-point ideological scale; it is possible to compare the absolute difference between the respondent's self-location and where he places a candidate and thereby derive a scale along which the higher a person's score, the more distant he considers his ideological position from that of Nixon or McGovern. Similarly, respondents rated themselves and each candidate on nine salient policy issues; we calculate an issue proximity index for each candidate indicating, as the score increases, a growing distance between the respondent's issue stand and his perception of the candidate's stand.

TABLE 1

SUMMARY MEASURES OF SELF, CANDIDATE, AND PARTY IMAGES: 1972

In measuring images of the Republican and Democratic political parties we employed procedures parallel to those for measuring candidate images. The cognitive aspect of each party's image consists of the respondent's (l) perception of each party's position on the seven-point liberal/conservative scale and (2) perception of each party's stand on each of the nine policy issues (again combined into composite scales for each party). The affective aspects of party images we tap with a party index derived from respondent's stated likes and dislikes about each party. And, we employ ideological and issue proximity measures similar to those used for candidates to measure the conative aspect by relying upon respondents' placements of each party on the liberal/ conservative scale and on the nine policy issue scales. Table l summarizes the 23 measures and the labels used in reporting findings.

An initial question is whether any of the 23 variables relate to the vote. One way of answering this question is to correlate each.image measure with respondents' voting behavior (using dummy variables of "O" for a Nixon vote and "l" for a McGovern vote). Space limitations make it impossible to present the full correlation matrix of each image variable with the vote and with one another. However, of the simple correlation coefficients between each of the image variables and the vote all but two proved statistically significant. Those two were respondents' perceptions of Nixon's issue stands (r=.05) and of the issue stands of the Republican party (r=.08). Of the significant coefficients, only a few are notably high. These include the ratings respondents gave to each candidate on the feeling thermometer (-.64 and .64 for Nixon and McGovern respectively), the respondents' proximities to each candidate's perceived ideological position (.56 and -.50 for Nixon and McGovern), the respondents' partisan (-.50) and ideological (-.47) self-images, and the respondents' proximities to what they perceive as the Republican party's ideological position (.46).

In sum, simple correlations indicate that almost all of the 23 image measures relate to the vote and that selected political self-images, candidate images, and party images are especially noteworthy. Yet, simple correlations will not reveal which of these components are uppermost in-the voters' decision-making processes. Nor does such analysis suggest the relative importance of the cognitive, affective, and conative dimensions of voters' images in helping them to make up their minds. To get at these matters factor analysis is a particularly useful tool.

Table 2 displays the results of a varimax analysis explaining 50% of total variance among all variables excluding the vote. The four-factor solution vividly refutes the original three-component model of voters' self, candidate, and party images. Factor I consists of nine variables, each pertaining to issues. Respondents' self-locations on issues, their perceptions of both candidate and party issue stands, and their proximities to candidates and parties on issues--all load on this factor and it thus exhausts all measures that in any way deal with issues. The second factor has ten variables and is bipolar, with five measures at the positive end and five at the negative. It is the most difficult factor to interpret for two of the items measure voters' self-images (partisan and ideological), three measure the affect aspect of candidate images (the two thermometers and the candidate index), the party index relates to affect toward parties, and the remaining variables are the respondents' proximities to the candidates' and parties' perceived ideological positions. If anything, this factor is a candidate-party factor. The remaining two factors offer no such interpretative problems. Factor III is a conservative factor consisting of respondent perceptions of the ideological positions of Nixon and the Republicans; the fourth factor is a liberal factor made up of perceived ideologies of McGovern and the Democrats. A second factor analysis with the vote added to the other 23 variables rounds out the picture of the principal components of the 1972 vote decision. With the exception of the vote, which enters as the highest loading variable on the second factor, the same four factors emerge as in the first analysis.

TABLE 2

FACTOR LOADINGS OF MEASURES OF IMAGE VARIABLES EXCLUDING VOTE BEHAVIOR

The results of factor analyses warrant two assertions. First, the preconceived categories of the components of electoral decisions summarized in Table l are not distinct dimensions in the factor analyses. Nor are the electoral components derived from earlier voting studies (especially party, candidate, and issue orientation) clear-cut. To be sure, issue orientations stand out. However party and candidate orientations meld, and conservative and liberal orientations join the list as identified dimensions of the voting decision. Second, the vote appears on the candidate-party factor. Combining most strongly with it are three affective measures--i.e., the ratings of McGovern and Nixon on the candidate thermometers and the candidate index. This suggests a strong relationship between affective orientations toward the candidates and the vote in 1972, perhaps even that these affects were the overriding components of the electoral decision. Further to unravel some of the intricacies in the complex relationship between these electoral components and voting, we turn to the third and final stage of analysis, multivariate regression.

The results of a stepwise regression analysis using vote behavior as a dependent variable and the 23 image measures as independent variables appears in Table 3.

TABLE 3

REGRESSION ANALYSIS OF IMAGE VARIABLES EXPLAINING 1972 PRESIDENTIAL VOTE

The Beta coefficients indicate that numerous variables had some independent effect upon the vote, but the relative importance of most (such as the respondent's proximity to the Democratic party's perceived ideological position) was virtually nonexistent. The key component of the vote was the affective evaluation of the candidates on the feeling thermometers. The affective orientation to Nixon explained 41% of the variance in the vote and affect toward McGovern added another 18%; the Beta coefficients of both are strong: -.29 for Nixon and .30 for McGovern. (Since the vote was coded as "0" for Nixon and "1" for McGovern, the negative Beta for Nixon indicates the negative correlation between his thermometer ratings and the vote.) Following affective orientations toward the candidates in importance are two variables originally conceived of as conative aspects of the candidate images and which factor analyses indicated as belonging to a candidate-party factor--that is, respondents' proximities to each of the candidate's perceived ideological predilections. Finally, the only other variable explaining an additional one per cent variance is the voter's partisan self-image (Beta = -.11). Thus, in a regression solution in which all independent variables explain a total of slightly less than two-thirds of the variance in the 1972 vote, five variables provide 64%. All five of these were on the candidate-party factor and four pertain directly to preconceived categories of affective and conative aspects of candidate images.

Composite indices of the perceived issue stands of the candidates and of the parties probably obscure some of the differences that voters detect between contestants on specific issues. This may account for the fact that issues did not appear in this analysis as relatively important explanations of the 1972 vote. Another regression analysis helped test for this possibility. This regression maintained the separation of five key issues (thus, entering respondents' self-locations, perceived candidate stands, and proximities to candidates on each issue as separate variables in the analysis). Only the voters' proximities to the perceived Republican position on the Vietnam issue (addition of four percent variance and Beta .20) and proximities to McGovern on Vietnam (one percent variance and Beta = -.16) had relatively important effects. Thus, the farther respondents deemed themselves from perceived Republican policy on Vietnam and the closer to McGovern's perceived stand, the more likely they voted for McGovern. Again the thermometer ratings emerged as the two most important variables, in a regression explaining 728 of the variance in vote the feeling ratings explained 59%. Since regression analysis permits a probe of only the direct relationships between selected variables and the vote, we remain uncertain as to what indirect effects such influences as partisan and ideological self-images have as they condition voters' perceptions of candidates. The moderate correlations between these variables and affective orientations toward the candidates lead us to suspect such an indirect, conditioning effect. Moreover, other analyses comparing the direct and indirect effects of partisan self-images on the 1972 vote suggest considerable indirect effects (Miller, 1972).

In sum, despite the restrictions inherent in available measures and techniques there are at least partial answers to the questions posed at the outset of this presentation. First, most of the 23 image measures exhibit a direct relationship to the vote. Second, there are discrepancies between earlier studies' conceptions of the components of the vote as being the voter's self-image, candidate images, and party images and the principal components derived from reported factor analyses. Instead, the components appear to be issue, candidate-party, and liberal-conservative orientations. Yet, step-wise regression indicates that candidate image concerns on the candidate-party factor, especially in their affective and conative aspects, are the most important explanations of the 1972 vote. Unanswered, of course, are such questions as the relative magnitude of indirect effects on candidate images and voting behavior of other influences such as Party identification.

REFERENCES

Boulding, K. E. The image. Ann Arbor: University of Michigan Press, 1956.

Boyd, R. W. Presidential elections: an explanation of voting defection. American Political Science Review, 1969, 63, 498-514.

Butler, D. i Stokes, D. Political change in Britain. New York: St. Martin's Press. 1969.

Campbell, A., Gurin, G. & Miller, W. E. The voter decides. New York: Harper, Row and Co., 1954.

Campbell, At Converse, P., Miller, W. & Stokes, D. The American voter. New York: John Wiley & Sons, 1960.

Campbell, A., Converse, P., Miller, W. & Stokes, D. Elections and the political order. New York: John Wiley 6 Sons, 1966.

Downing, J. What is a brand image? Advertising Quarterly, 1964-65, 1, 13-19.

Harding, J., Proshansky, H., Kutner, B. & Chein, I. Prejudice and ethnic relations. In Lindzey, G. & Aronson, E. (Eds.),The handbook of social psychology. Vol. 5. Reading, Mass. Addison-Wesley Publishing Co., 1959.

Jensen, R. American election campaigns: a theoretical and historical typology. Paper presented at the meeting of the Midwest Political Science Association, Bloomington, Indiana, 1968.

Kelley, S. & Mirer, T. The simple act of voting. American Political Science Review, 1974. 68, 572-91.

Kovenock, D., Beardsley, P. & Prothro, J. Status, party, ideology, issues, and candidate choice: A preliminary, theory-relevant analysis of the 1968 American presidential election. Paper presented at the meeting of the International Political Science Association. Munich, 1970.

McGinniss, J. The selling of the president 1968- New York: Trident Press, 1969.

Miller, A. H., Miller, W. E., Raine, A. S. & Brown, T. A. A majority party in disarray: Policy polarization in the 1972 election. Paper presented at the meeting of the American Political Science Association, New Orleans, 1973.

Miller, R. E. & Jackson, J. S. Split ticket voting: An empirical approach derived from psychological field theory. Paper presented at the Southern Political Science Association, Atlanta, 1973.

Minow, N. M., Martin, J. B. g Mitchell, L. M. Presidential television. New York: Basic Books, Inc., 1973.

Natchez, P. B. Images of voting: The social psychologists. Public Policy, 1970, 18, 553-87.

Natchez, P. B. & Bupp, I. C. Candidates, issues, and voters. Public Policy, 1968, 17, 409-37.

Smith, M. B. The personal setting of public opinion. Public Opinion Quarterly, 1947, 11. 507-23.

Wyckoff, G. The image candidates. New York: The Macmillan Co., 1968.

----------------------------------------