Individual Differences in Voting Behavior: Further Investigations of Involvement

Michael L. Rothschild, University of Wisconsin
Michael J. Houston, University of Wisconsin
ABSTRACT - Involvement is becoming an important variable in explaining differences in information processing and decision making styles. In this paper, these differences are examined across three political races and across subjects with varying levels of interest in the political process.
[ to cite ]:
Michael L. Rothschild and Michael J. Houston (1980) ,"Individual Differences in Voting Behavior: Further Investigations of Involvement", in NA - Advances in Consumer Research Volume 07, eds. Jerry C. Olson, Ann Abor, MI : Association for Consumer Research, Pages: 655-658.

Advances in Consumer Research Volume 7, 1980     Pages 655-658


Michael L. Rothschild, University of Wisconsin

Michael J. Houston, University of Wisconsin


Involvement is becoming an important variable in explaining differences in information processing and decision making styles. In this paper, these differences are examined across three political races and across subjects with varying levels of interest in the political process.

There are significant differences in cognitive complexity within subjects across races, with high level races eliciting the highest levels of complexity. There is no difference in complexity across individuals within high level races but the difference increases and becomes significant as lower level races are examined.


The search for and explanation of individual differences permeates the entire spectrum of research on human behavior. Social psychologists examine individual differences in response to social influences. Industrial psychologists focus on individual differences in work behavior. Consumer psychologists are interested in individual differences in buying behavior.

Increasingly, many of these consumer researchers are turning to non-traditional contexts in which to examine individual differences in "buying" behavior (e.g., charitable contributions, birth control). Typically, these nontraditional contexts include areas in which persuasive attempts similar to those employed in the marketing of goods occur. Thus, those knowledgeable with the framework and procedures for studying the effects of marketing strategies, i.e., consumer researchers, have become appropriate parties to pursue empirical issues in these areas.

A nontraditional area that has realized a recent influx of work by consumer researchers (e.g., Rothschild, 1978; Rothschild and Ray, 1974; Palda, 1975; Kline, 1972; Omura, 1979) is voting behavior. Such an influx is not surprising given the dramatic increases in campaign expenditures and the usage of media similar to those employed by traditional marketers.

Individual differences in voting behavior would seem to be a fruitful area in which to test some of the concepts that have been useful in more traditional realms of consumer behavior. This is particularly the case when the well-defined and relatively simple nature of the choice environment of a voting decision is considered.

The purpose of this paper is to examine the role of "involvement" as a moderator of individual differences in voting behavior. A recent model of involvement as it applies in consumer behavior (Houston and Rothschild, 1978) is employed to suggest and test hypotheses concerning situational and individual influences on differences in voting behavior.


While, at a general level, involvement refers to the overall level of interest in some issue or object, a variety of imprecise conceptualizations have appeared in the behavioral science literature. They have recently been categorized into three distinct types. Situational involvement (SI) refers to the tendency of a situation or issue to generate across a preponderance of individuals an equivalent level of concern for their behavior when confronted by that situation or issue. Situational involvement is generally high when most people perceive the consequences of less-than-optimal behavior in the situation as rather severe. It also tends to be high when an issue is so prevalent across society that it becomes difficult for an individual to avoid the issue, regardless of his or her personal proclivities toward the issue. In other words, the presence of the issue is so overwhelming that it affects learning and behavior in individuals in spite of the absence of personal importance being attached to the issue. It is this latter form of situational involvement that is relevant to this paper.

Enduring involvement (EI) deals with the ongoing personal concern with an issue that is exhibited by an individual. High levels of enduring involvement result from considerable prior experience in dealing with an issue and/or a strong linkage of the issue to the individual's unique structure of values. Thus, the nature of enduring involvement is such that it reflects a between-individuals within-issue perspective, while situational involvement assumes a between-issues within-individual perspective.

Levels of situational and enduring involvement interact to influence response involvement (RI). Response involvement refers to the complexity or extensiveness of cognitive or behavioral processes characterizing an individual's relationship to an issue. Levels of response involvement are reflected in a sequence of stages that comprise the process by which an individual reaches a decision regarding the issue. While response involvement can be examined at each stage of the sequence, the concern of this paper is with response involvement at what is labeled the presearch stage, i.e., the complexity of the cognitive framework used by the individual to relate to an issue.

The effects of situational and enduring involvement on response involvement can be likened to the mean and variance of a statistical distribution. A decision area high in situational involvement elicits a mean level of response involvement across a sample of individuals that is higher than the mean level of response involvement elicited by a decision area low in situational involvement. The variance about each mean is determined by differences in enduring involvement towards the decision area exhibited by the individuals.

A measure of response involvement at the presearch stage has been developed based upon social judgment theory (Sherif, Sherif, and Nebergall, 1965). According to social judgment theory, for any issue which can be defined as having a spectrum of positions, an individual would have a range of acceptable positions (i.e., a latitude of acceptance), a range of objectionable positions (i.e., a latitude of rejection), and a range of neither acceptable nor objectionable positions (i.e., a latitude of non-commitment). The notion of a latitude of acceptance is incorporated into the measure of response involvement examined here. However, while social judgment theory examined latitudes of acceptance in terms of an overall issue, a disaggregated approach is taken here. In the present approach the latitude of acceptance is examined for each salient dimension of the issue. For each salient dimension there is, then a range of possible values, with a portion of these deemed acceptable by the individual (i.e., the latitude of acceptance for the dimension). High response involvement is indicated when generally narrow latitudes of acceptance exist across many salient dimensions (few positions are acceptable), while low response involvement occurs when the latitudes of acceptance are wide (many positions are acceptable). This focus on latitudes of acceptance, while contrary to social judgment theory's focus on latitudes of rejection, has been successfully assumed by others (Atkins and Bieri 1968; Rhine and Severance, 1970).

A second component of response involvement in the pre-search stage concerns the absolute number of salient dimensions used by an individual in developing a stance towards an issue. High and low levels of response involvement are reflected by large and small numbers of salient dimensions, respectively.

Combining the notions of the latitude of acceptance and the size of the set of evaluative dimensions used by an individual, an overall matrix-like measure of response involvement can be derived. The horizontal axis of the matrix represents the width of the latitudes of acceptance across attributes, while the vertical axis represents the number of dimensions of the issue that are salient. High response involvement is therefore indicated by a rather long, narrow matrix (see Figure la) where several dimensions with generally narrow latitudes of acceptance are salient. Low response involvement is indicated by a short, wide matrix (Figure 1b) where a small number of salient dimensions with generally wide latitudes of acceptance exist. Note that it is the shape, and not the size, of the matrix that reflects complexity. A large/small number of salient dimensions is offset by wide/narrow latitudes of acceptance. Thus, the ratio of the vertical axis to the horizontal axis provides an overall direct measure of response involvement, i.e., the higher the ratio, the greater the level of response involvement.



Furthermore, as indicated earlier, levels of situational and enduring involvement interact to effect a level of response involvement. The nature of the interactions and their effects can vary considerably. When both situational and enduring involvement are high, response involvement in the presearch stage will tend to be high. However, even in the absence of high enduring involvement, high situational involvement can generate high response involvement when the issue is so pervasive that it overcomes the individual differences reflected in enduring involvement. In such a case individual differences in response involvement will be minimal, i.e., complex cognitive structures will tend to occur across all individuals, regardless of the level of enduring involvement. Based on the above, the following hypotheses were tested in this investigation:

H1: As situational involvement increases, the mean level of response involvement increases;

H2: There is an interaction effect of situational and enduring involvement on response involvement such that at low levels of situational involvement there will be a direct effect of enduring involvement on mean response involvement while at high levels this effect will disappear.

H3: Greater variance in response involvement, as exhibited by standard deviations will occur at lower levels of situational involvement.


Concern for individual differences in voting behavior is relevant to political candidates and organizations who want to influence the decision to vote and the outcome of that vote. The increased sophistication of campaign techniques allow persuasive messages to be directed at specific segments of the electorate in a manner similar to marketing a product to specific groupings of buyers. Individual differences can provide the basis for forming such segments and an understanding of the behavioral patterns occurring within them. Campaigns, in turn, can be individually tailored to each segment.

Historically, the presidential race has been the one most studied by political science researchers. This tradition was begun in the early work by Lazarsfeld (1948) and Campbell (1966), and was continued until quite recently. Most of this work found "limited effects" of mass media on voter behavior. More recent work (Patterson and McClure, 1976; Palda, 1975; Kline, 1972; Rothschild, 1978) has in part been based on nonpresidential races and has found communications effects on voter behavior.

This paper continues in the more recent vein with the basic premise that if researchers are interested in studying voting behavior they cannot restrict themselves to the presidential election. While the presidential election is perhaps the most interesting due to its prominence in the media and the interest that it generates, it is for these very reasons that it can prove fruitless as a behavioral context within which to examine individual differences in voting processes. Certainly, individual differences in the direction of a vote occur in a presidential campaign. Such knowledge, however, is useless to the candidate or party wanting to influence that outcome prior to its occurrence. More useful evidence regarding individual differences would relate to the processes leading up to a voting decision, e.g., criteria used for evaluating candidates, information sources consulted, receipt of persuasive messages, etc. The authors contend that the nature of a presidential election is such that it will tend not to reveal useful individual differences in processes leading up to voting decisions.

During the period 1952-1974, the voting age population increased 40% while presidential campaign expenditures increased 690% (Rothschild, 1978). Also, during this period television developed into perhaps the most pervasive medium to which individuals are exposed. Political advertising and news broadcasts concerning the presidential election occur on virtually a daily basis. Coupled with the attention the issue receives in other media (e.g., newspapers, magazines, radio), the presidential election is a topic that is virtually unavoidable by the large majority of society. Thus, as an issue, it represents a high level of situational involvement of the type whose effect will overcome individual differences in enduring involvement and tend to generate equally complex cognitive structures across individuals. Such an effort is not expected from election races of lower situational involvement, e.g., congressional, state assembly, judicial, and local elections. In these types of elections individual differences in enduring involvement will tend to cause individual differences in response involvement.


The hypotheses were tested with data collected from a convenience sample of 103 subjects of voting age in September, 1976. Subjects responded to three electoral races as representations of varying levels of situational involvement: the Carter-Ford presidential race, a local race for the U.S. House of Representatives, and a local race for the State Assembly. As a measure of enduring involvement for each election, subjects were asked to indicate the extent of personal concern each felt for the outcome of the race. Subjects indicating they "care very much" were placed in one category to represent a high degree of enduring involvement, while those indicating they "care somewhat" or "don't care at all" were placed in a second category to represent a lower level of enduring involvement. The two lesser degrees of concern were collapsed into one category in order to generate sufficient sample sizes for low levels of enduring involvement.

Cognitive structures (response involvement) with respect to each election were measured for each subject using the matrix approach described above. Latitudes of acceptance were measured for each of eleven dimensions using a nine point scale (see Figure 2) for each dimension. Subjects indicated latitudes of acceptance for each dimension by circling each value on the scale which corresponded to an acceptable level of the dimension.



On a separate set of scales, subjects indicated for each electoral race whether they would use a particular dimension in evaluating candidates. As an overall measure of the complexity of cognitive structure for each election, the following ratio was computed to quantitatively represent the shape of the response involvement matrix for each subject:


A more detailed description of the derivation and development of this measure and evidence of the validity of its components can be found in Rothschild and Houston (1977, 1979).


Data analysis involved a comparison of values of cognitive complexity across races and across individuals. Table 1 shows the overall difference in means across races. The differences in each comparison are significant and the differences are in the hypothesized direction. The standard deviations for the races are virtually equal, thus lending no support to H3.



Table 2 shows differences between individuals with high and low levels of enduring involvement for each race. The differences decrease as the situational involvement of the races increases. The difference is insignificant in the presidential race (t = 1.04, n.s.), marginally significant in the congressional race (t = 1.63, p < .06) and highly significant in the state assembly race (t = 3.11, p < .001). The differences are in the expected direction and H2 receives strong support. In addition, as expected, the number of individuals reporting high enduring involvement decreases and the number reporting low EI increases as the level of the race declines. Finally, one should note that for high EI, RI scores tend to increase as the level of race diminishes but for low EI, RI scores tend to decrease as level of race diminishes. Figure 3 shows these results graphically.


The findings from the study described above provide tentative support for the contention that the presidential election is a less-than-optimal context for examining the effects of individual differences in voting behavior. The construct of involvement was used as a conceptual basis for generating and testing the hypothesis that the presidential election represents such a high level of situational involvement that it overwhelms the effects of personal involvement and results in equally complex cognitive structures regarding the election across all individuals. Elections at local levels, however, do not conform to this tendency and are better suited for the examination of individual differences. Their salience is more a function of the individual; the salience of the presidential election seems to derive primarily from the attention and publicity that it receives in the mass media.





The implications of these findings for the emerging process-orientation in consumer research are substantial. To the extent that individual differences in consumer decision processes are of interest to marketing strategists for segmentation purposes, such a basis of segmentation would seem less useful in cases of high-involvement products. The usefulness of process-oriented segmentation variables would be restricted to products of a low-involvement nature. However, the extent to which these findings on voting behavior apply to product purchase behavior remains an empirical question.


Atkins, A. L., and Bieri, J. (1968), "Effects of Involvement Level and Contextual Stimuli on Social Judgment," Journal of Personality and Social Psychology, 9, 197-204.

Campbell, A. (1966), Elections and the Political Order, New York: Wiley.

Churchill, Gilbert A. Jr. (1976), Marketing Research: Methodological Foundation, Hinsdale, Illinois: The Dryden Press.

Houston, Michael J., and Rothschild, Michael L., "Conceptual and Methodological Perspectives on Involvement." Proceedings of the American Marketing Association, Series #43, 1978, 184-187.

Kline, F. G. (1972), "Mass Media and the General Election Process: Evidence and Speculation," paper presented at the Syracuse University Conference on Mass Media and American Politics, Syracuse, New York.

Lazarsfeld, P. F., Berelson, B. R., and Gaudet, H. (1948), The People's Choice, New York: Columbia University Press.

Omura, Glenn S., "Role of Attribute Generality in Cognition of Political Candidates," Proceedings, Tenth Annual Convention, The Association for Consumer Research, 1979.

Palda, K. S. (1975), "The Effect of Expenditure on Political Success," Journal of Law and Economics, 18, 745-771.

Patterson, T. W., and McClure, R. D. (1976), "Television and the Less Interested Voter: The Costs of an Informed Electorate," The Annals of the American Academy of Political and Social Science, 425, 88-97.

Rhine, R. J. and Severence, L. J. (1970), "Ego Involvement, Discrepancy, Source Credibility, and Attitude Change," Journal of Personality and Social Psychology, 16, 175-90.

Rothschild, M. L. (1978), "Political Advertising: A Neglected Policy Issue in Marketing," Journal of Marketing Research, 15, 58-71.

Rothschild, M. L. and Houston, M. (1977), "The Consumer Involvement Matrix: Some Preliminary Findings," in Contemporary Marketing Thought, ed. B. Greenberg and D. Bellenger, Chicago: American Marketing Association.

Rothschild, M. L. (1979), "The Consumer Involvement Matrix: A Strategic Marketing Tool," Working Paper, Graduate School of Business, University of Wisconsin-Madison.

Rothschild, M. L., and Ray, M. L. (1974), "Involvement and Political Advertising Effect: An Exploratory Experiment,'' Communication Research, 1, 264-285.

Sherif, C., Sherif, M., and Nebergall, R. (1965), Attitude and Attitude Change, Philadelphia: Saunders.