Attribution Theory and Voter Choice

Richard F. Yalch, University of Washington
[ to cite ]:
Richard F. Yalch (1975) ,"Attribution Theory and Voter Choice", in NA - Advances in Consumer Research Volume 02, eds. Mary Jane Schlinger, Ann Abor, MI : Association for Consumer Research, Pages: 783-792.

Advances in Consumer Research Volume 2, 1975      Pages 783-792


Richard F. Yalch, University of Washington

[The author wishes to thank Professor H. Woods Bowman, University of Illinois at Circle Campus and Northwestern University doctoral candidates Alice Tybout and Richard Bagozzi for assistance in the data collection.]

Voters frequently possess little campaign information but still satisfy their citizenship obligation by selecting candidates for various offices. Self-attribution concepts have been suggested as the appropriate internal information processing mechanism in situations in which attitudes have low saliency and are products of behavior rather than its antecedent. The assertion that voter behavior and attitudes reflect self-attributions was tested in a field experiment conducted during a local election. False feedback after a pre-election interview was hypothesized to be internalized as the subject's attitude toward voting and to influence actual turnout. The findings support the predictions and suggest campaign techniques to influence voter choice.

Theorists have tended to adopt one of two approaches when conceptualizing the voter decision process. The "rational" model portrays the voter as a recipient of a stream of communications describing the candidates' positions on the issues. He contrasts this with his own position and on the basis of the matching of desires and expectations decides if and for whom to vote. Perhaps the best known advocate of this approach is Downs (1957) in his influential Economic Theory of Democracy. On the other hand, the "emotional" voter model assumes that the voter is exposed more to nonissue information from which he develops a perception of the candidates' images. His ultimate decision is based on his affection for any of these images. Nimmo (1970) provides a thorough discussion of the art of image politics.

This characterization probably overemphasizes the differences in the approaches. Clearly, image can be defined to include the candidate's position on important issues and the political economist could include the candidate's personality as one of his attributes. Moreover, there is the common assumption that the voter is exposed to a considerable amount of campaign information, pays attention to it, understands it, and remembers enough of it so that it influences his ultimate choice. Although this assumption is reasonable for presidential and major state elections, there are reasons to doubt that it is appropriate for local elections. Anyone who has conducted interviews prior to a minor election or engaged in door-to-door canvassing is quickly impressed by the few voters knowing more than a candidate's name and party identification and the many not able to recall even that much.

Other evidence also suggests that local elections should be considered as different behavioral situations than the more often researched and analyzed national elections. Books on campaigning are quick to cite instances in which a candidate won an election because the voters had mistaken his name for that of a prominent person (see Nimmo, 1970, pp. 11-12). Realizing the limited information guiding the voters, candidates worry about their ballot position much the way brand managers are critically concerned about shelf space. In primary elections, the voters do not have the help of party identification to guide them and poor ballot positions have been known to cause the defeat of competent incumbents. To further support the contention that issues are not the decisive factor in local elections. one can cite Stokes and Miller's (1966) conclusion that only about seven per cent of a national sample had "any discernible issue content" in the reasons they gave for voting Democratic or Republican in the 1958 congressional elections. Therefore, it is likely that issues are even less important in a city council race. A campaign manager trying to elect a candidate by promoting his positions or image would have an extremely difficult time convincing a majority of the voters. Consequently, most local campaigns concentrate on building name identification in the hope that the "mere exposure" hypothesis is valid. [The mere exposure hypothesis states that the repeated presentation of a neutral stimuli is a sufficient condition for increased affection for it. Thus, just seeing a candidate's name frequently should cause one to gradually favor him (Zajonc, 1968).]

In a frequently cited article, Krugman (1965) distinguishes between the possible effects of a mass communication campaign on the basis of the target behavior. If it is a high involvement task, such as the selection of a president, the campaign's effectiveness can be determined by its immediate effect on the audience's attitudes which will determine behavior. This is the appropriate situation for applying a rational, emotional or combined voter decision model as the vehicle to identify the critical determinants of choice. However, for less involving behaviors, which the typical minor election vote decision is, messages may not have a noticeable effect on attitudes until the point of purchase or in usage. Krugman (1965) was not entirely certain how this occurred and, in fact, concluded his article by pointing out the need for a low involvement model to explain the relationship between attitudes and behavior in these situations.

One possible framework for analyzing low involvement behavior is attribution theory since it describes how individuals assign attitudes to behaviors when they are not sure of the exact causes (Kelley, 1967). Bem (1972) has extended this theory to explain the self-assignment of an attitude after a behavior. Specifically, the first proposition of his self-perception theory states that:

Individuals come to "know" their own attitudes, emotions, and other internal states partially by inferring them from observations of their own overt behavior and/or the circumstances in which this behavior occurs.

This proposition has received a compelling amount of support from research on many diverse behaviors and situations. For example, Lepper (1973) demonstrated that children's perception of their own honesty was a function of their appearing to behave honestly in a prior task; Ross, Insko and Ross (1971) found that subjects used the physical evidence of an initial questionnaire response as a guide in a second response despite the fact that the evidence had been altered so that it reflected an opposite attitude; McArthur, Kiesler and Cook (1969) used a false personality feedback to induce greater volunteering for a post-experimental task; and Schachter and Singer (1962) reported that subjects determined their emotional state after being drug-aroused by observing the state of a confederate in the same situation.

Although this research is impressive in its robustness and consistency, it has been rarely utilized in consumer behavior research. This may only reflect the general time lag in the "borrowing" process that has occurred between the behavioral sciences. However, it also might reflect a justified caution about adopting theories that have been primarily tested in laboratory settings, using student samples, self-report dependent measures, and inconsequential behaviors. One might rightfully question the relevance of the theory. Can it be used to alter "meaningful" behavior in a natural environment with a reasonable expenditure of resources?

This research study had several objectives. First, it attempted to test self-perception theory in a natural setting and to demonstrate that it suggests ways to alter both attitudes and behavior. Inducing a change in both variables is nontrivial task given the evidence that an attitude change does not lead to a behavior change (Festinger, 1964). Second, it explored the perplexing finding that attitudes toward the candidates in an election do not account for much of the variance in the decision to vote or abstain in an election (Yalch, in press). This involved determining whether an individual's perception of himself as a likely or unlikely voter was a major determinant of voting behavior. Lastly, a possible strategy for increasing voter participation in the electoral process was to be evaluated.


The experiment was conducted during two special aldermanic elections in Chicago during the summer of 1973. The preliminary election featured five candidates and was used to pilot test the induction procedure. The main study occurred during the run-off election as part of a pre-election survey. The 190 subjects were selected from official lists of registered voters and contacted door-to-door. After the subjects had completed the survey, the interviewers pretended to score the responses on a bogus voter profile sheet. This was given to the respondents accompanied by a brief verbal explanation of its interpretation. The subjects' voting behavior on election day was assessed from party records and their self-perception as frequent or infrequent voters was determined in a post-election interview.


There were two treatment groups and one control group. The 77 subjects randomly assigned to the above average treatment received the voter profile sheet illustrated in Figure l. Their responses were supposedly represented by the red line drawn in after the interview. The position of the line to the right of the center or normal line was explained as indicating that their responses reflected a greater interest and concern about the election and politics than the average citizen. In addition, it was pointed out that this suggested that they would be more likely to vote in an election. The profile was left with them to further reinforce the information that they were above average citizens.

The average group consisted of 85 subjects receiving the profile sheet illustrated in Figure 2. It was identical to the other profile except that their response line was drawn near the center of the scale. This was interpreted to them as evidence that they were about the same in their political orientation and interest as the average person in the community. They were not told that they would be very likely to vote in the election, but they were also allowed to keep the profile sheet.

The 28 persons in the control group were interviewed door-to-door, answering the same questions as the treatment groups, but were not told anything about their disposition toward politics. This group's behavior enables one to estimate the absolute effect of being given feedback after an interview

Dependent measures

The treatment effects were assessed in two ways. First, the voting behavior of each subject was determined from records maintained by a campaign organization. The ability to have an unobtrusive dependent measure was a major reason why voter turnout rather than candidate choice was selected as the experimental behavior. This is advantageous in minimizing threats to the construct validity of the experiment. The second measure was the subjects' self-perception as a frequent or infrequent voter. This was determined in a telephone survey conducted within weeks after the election. Several general questions were asked in addition to the self-perception question to minimize the possibility that the subjects would think that this was a test of their ability to recall the profile information.






1. Subjects given above average feedback will be more likely to report themselves as above average in the post-election interview than those given average feedback.

2. Subjects given above average feedback will be more likely to vote in the election than those given average feedback.

These two hypotheses were generated from self-perception theory's proposition than an individual's attitudes and behavioral dispositions reflect his past behavior. By altering his interpretation of this experience, it should be possible to alter his beliefs about himself. This in turn should influence his future behavior. No hypotheses were prespecified for the comparisons between the control group and the treatment groups because there was no information available as to how the residents of this community perceived themselves. Since this district usually has one of the highest turnout rates in the city, it was likely that many voters would naturally perceive themselves as above average. A tentative prediction was that the control group's attitudes and behavior would fall somewhere in-between the two treatment groups's.


Before the treatment effects were analyzed, a test of subject randomization was made. This involved comparing the preliminary election voting behavior of the groups (assessed from party records), and their intention to vote prior to receiving the feedback (assessed in the interview). There was a slight but statistically insignificant higher voter turnout in the above average and control groups compared to the average group. However, since there was only a weak association between voting in the preliminary election and voting in the run-off, this was not judged to be problematic. A more critical test was the difference in the voting intentions of the groups since this had a high correlation with subsequent voting. There was virtually no difference between the groups and it was concluded that the randomization had been successful.

Table 1 presents the post-election self-perceptions of the subjects. The difference between the above average and average feedback groups is statistically significant (p=.016), using a test for the difference in proportions (Ferguson, 1973, pp. 160-162). The average group also had a significantly lower probability of saying that they were above average voters than the control group (p=.035). Apparently, the control group did have a high self-perception of themselves as politically active citizens as their responses are equivalent to those from the above average treatment group. [A comparison of the voting behavior of the control group with another sample interviewed by telephone suggests that this group may have contained more politically involved citizens than the other groups. The above average feedback, therefore, probably did increase the group's self-perception more than is indicated by comparisons with the control group.]

The behavioral effects of the treatments are presented in Table 2. The second hypothesis was supported as the above average group had a significantly higher turnout than the average feedback group (p=.041). All comparisons with the control group were insignificant, partially as a consequence of the small size of the group. The finding that the average feedback group had a much lower self-perception and likelihood of voting relative to the control group suggests the possibility that telling persons that they are only average might undermine their intrinsic interest because one is inclined to think of himself as above average. Being told that this view is incorrect may have caused many in the average feedback group to question whether they really cared that much about voting in the election.



[Several subjects could not be recontacted after the election so the numbers in this table are smaller than the number participating in the experiment.]




The experimental findings demonstrate the validity of self-perception theory and its applicability to local election voting behavior. There was clear evidence that voters used their past behavior and the interviewers' interpretation of it to determine their disposition toward voting in the election. A local election is a low involvement activity and persons usually have uncertain and unstable attitudes. Therefore, attitudes anchored to a prior behavior should have a more significant role in determining an individual's course of action. However, the behavior-attitude relationship appears to be more ambiguous to the individual than previously thought. The external information provided by the interviewer caused many persons who thought of themselves as only average to perceive themselves as above average, and conversely, many above average persons began to consider themselves as only average. The previous experiences which lead to this original self-perception were virtually negated by a single experience and the false information about it. The post-behavior attitude reassessment process represents a fruitful area for future consumer behavior research and a potentially very powerful method to influence behavior.

The study also provides further evidence for the belief that the decision to vote in an election is based on factors in addition to one's attitudes toward the candidates in the election. An individual's perception of himself as the type of person who votes frequently seems to be one of these factors. How this self-perception develops is not yet clear, but childhood socialization and initial voting experiences probably play an important role. This finding helps to explain why many citizens continue to vote even though their candidates always lose. The act of voting in the absence of a tangible reward probably causes the individual to assume that he has a very strong citizenship attitude.

Since behavior appears to be a major determinant of one's attitudes and future behavior, political campaigns might benefit from a reallocation of resources from "information" dissemination to techniques designed to induce voters to make an early behavioral commitment to the candidate. The recent congressional elections featured a promotional strategy labelled "Dollars for Democrats". The ostensible purpose of the activity was to raise money for the campaigns, but it probably had an additional benefit in having voters make a small commitment to the Democratic party. On election day, a recall of this commitment could be influential in determining how one votes. Freedman and Fraser (1966) have demonstrated the importance of yielding to a small request on the likelihood of yielding to larger requests at a future date. The methods for inducing a favorable attitude without resorting to a flood of campaign propaganda are many and should be the object of future research and practice.


Bem, D. J. Self-perception ,theory. In L. Berkowitz (Ed.), Advances in experimental social psychologY. New York: Academic Press, 1972.

Downs, A. An economic theory of democracy. New York: Harper and Row, 1957.

Ferguson, G. A. Statistical analysis in psychology and education. New York: McGraw-Hill. 1975.

Festinger, L. Behavioral support for public opinion change. Public Opinion Quarterly, 1964, 28, 404-417.

Freedman, J. L. & Fraser, S. C. Compliance without pressure: the foot-in-the door technique, Journal of Personality and Social Psychology, 1966, 4, 195-202.

Kelley, H. H. Attribution theory in social psychology. In D. Levine (Ed.), Nebraska symposium on motivation. Lincoln: University of Nebraska Press, 1967.

Krugman, H. E. The impact of television: learning without involvement. Public Opinion Quarterly, 1965, 29, 349-356.

Lepper, M. R. Dissonance, self-perception, and honesty in children. Journal of Personality and Social Psychology, 1973, 28, 65-74.

McArthur, L., Kiesler, C. A. & Cook, B. Acting on an attitude as a function of self-percept and inequity. Journal of Personality and Social Psychology, 1969, 12, 295-302.

Nimmo, D. The political persuaders. Englewood Cliffs, New Jersey: Prentice-Hall, 1970.

Ross, M., Insko, C. A. & Ross, H. Self-attribution of attitude. Journal of Personality and Social Psychology, 1971, 17, 292-297.

Schachter, S. & Singer, J. E. Cognitive, social, and psychological determinants of emotional state. Psychological Review, 1962, 69, 379-399.

Stokes, D. E. & Miller, W. E. Party government and the saliency of Congress. In A. Campbell, P. E. Converse, W. E. Miller & D. E. Stokes (Eds.), Elections and the political order. New York: John Wiley & Sons, 1966.

Yalch, R. E. The prediction of voter turnout. In G. Zaltman & B. Sternthal (Eds.), Broadening the concept of consumer behavior, in press.

Zajonc, R. Attitudinal effects of mere exposure. Journal of Personality and Social Psychology Monograph Supplement, 1968, 9, 1-27.