Asymmetries in the Effect of Campaign Expenditures on Voting Patterns

George Siomkos, New York University
Avijit Ghosh, New York University
ABSTRACT - Campaign expenditures by political candidates have significant effects on election results. Results from a sample of 1982 Congressional elections are analyzed to study the nature of this effect and to determine whether the incumbency status and party affiliation of the candidates lead to asymmetries in this effect. Results show that both incumbency status and party affiliation cause significant differences in the effect of campaign expenditures on votes polled Asymmetries exist in the overall effect of campaign expenditures as well as in the marginal effect of changes in the level of expenditures.
[ to cite ]:
George Siomkos and Avijit Ghosh (1987) ,"Asymmetries in the Effect of Campaign Expenditures on Voting Patterns", in NA - Advances in Consumer Research Volume 14, eds. Melanie Wallendorf and Paul Anderson, Provo, UT : Association for Consumer Research, Pages: 531-535.

Advances in Consumer Research Volume 14, 1987      Pages 531-535

ASYMMETRIES IN THE EFFECT OF CAMPAIGN EXPENDITURES ON VOTING PATTERNS

George Siomkos, New York University

Avijit Ghosh, New York University

ABSTRACT -

Campaign expenditures by political candidates have significant effects on election results. Results from a sample of 1982 Congressional elections are analyzed to study the nature of this effect and to determine whether the incumbency status and party affiliation of the candidates lead to asymmetries in this effect. Results show that both incumbency status and party affiliation cause significant differences in the effect of campaign expenditures on votes polled Asymmetries exist in the overall effect of campaign expenditures as well as in the marginal effect of changes in the level of expenditures.

INTRODUCTION

Researchers from a number of disciplines have investigated the role of advertising and mass communication on voting behavior It is now well established that campaign expenditures by political candidates, especially political advertising, has a significant effect on voting patterns and on the outcomes of political elections Surveys of individual voter attitudes as well as econometric studies of the effect of political advertising and promotional expenditures on actual voting patterns have provided evidence of this impact.

In this study the results of a sample of the 1982 Congressional elections are analyzed to test the nature of the relationship between campaign expenditures and voting patterns. Specifically, we seek to determine whether asymmetries exist in this effect that are caused by the incumbency status and party affiliations of the candidates. That is, does the effect of campaign expenditures on voting patterns differ between incumbents and challengers and between candidates with different party affiliations. A differential impact of campaign expenditures across incumbency status and par AS affiliations point to asymmetries in the institutional structure of the political markets (Crain 1977).

Advertising and promotions in the arena of political elections is believed to convey information to voters regarding candidates' stands on issues and their programs. This view of the role of political campaigns assumes that campaign promotion leads to a dissemination of information to the electorate and results in more reasoned choice on the part of individual voters. If, however, there are significant asymmetries in the impact of campaign expenditures on voting patterns, incumbency and party affiliation may be associated with "market power" (Bain 1968) that gives incumbents significant advantage over challengers. According to Welch (1976), the advantage of incumbency may result from two sources. First, incumbents can raise campaign funds more easily (see also Jacobson 1980; Soley and Reid 1984). Second, the level of awareness that results from holding office is hard to overcome by challengers Party affiliation may provide added advantage to the incumbent depending on the political climate in the country as a whole and in the local area in particular Because of this, expenditures by some candidates may have a greater impact than expenditures by others. We use two alternative econometric models to test for the existence of such asymmetries in our sample.

REVIEW OF PAST STUDIES

There has been a number of studies that have investigated how mass communication and advertising affect the cognitive, attitudinal and behavioral intentions of voters Most early studies of the impact of political advertising in fact argued that advertising had little or no effect on voting behavior This "limited effects" argument was supported by a number of early studies (see, for example, Lazarsfeld, Berrelson and Gaudet 1944), as well as by the more recent study by DeVries and Torrance (1972) DeVries and Torrance found that voting intentions were affected more by news and interpersonal communications than by political advertising

Drawing upon the results of more recent studies, Rothschild (1978) challenged the limited effects hypothesis and argued that the effects of political communication are moderated by the individual's involvement with the political process. Rothschild posits that in low involvement situations advertising has-a strong or at least a moderate effect on awareness and intentions but little or no effect on voter attitudes In high involvement situations, on the other hand, advertising may have some effect on awareness but little effect on attitude and behavior. These generalizations are supported by the results of a number of studies. Patterson and McLure (1976) report that viewers of political advertising who had less interest in the election had a greater attitude change than viewers with high interest. A similar pattern was found by Swinyard and Coney (1978) who studied the effect of direct mail campaigns and political canvassing. They, too, found that voting intentions were positively affected in low involvement situations while promotional effort had little effect on intentions when voters were highly involved with the campaign Rothschild (1978) provides additional evidence supporting the involvement hypothesis.

An alternative approach to studying the impact of campaign expenditures is to investigate its effect on actual voting patterns. Palda (1975), for example, examined the relationship between actual voting behavior and several independent variables in a sample of Canadian elections In addition to the level of advertising by the candidates, the independent variables included in the study were: level of other campaign expenditures; incumbency status of the candidate; urban-rural nature of the district; past party history in the district, ant, success of the candidate at the top of the party ticket in the provincial election. Palda's analysis of actual election results from 108 electoral districts in Quebec for 1966 and 1970 and 53 electoral districts in Manitoba for 1973 confirmed the significant impact of campaign expenditures on actual voting behavior

As Rothschild (1978) notes, the importance of Palda's study is enhanced by the fact that he used actual voting behavior rather than voting intentions gathered from survey data as the dependent variable. Moreover, use of an econometric model allowed him to assess the relative impact of the different independent variables In the 1966 elections, the most important variable was the level of non advertising campaign expenses The level of advertising was the second most important variable. In 1970 the effect of advertising followed that of other expenses and incumbency status In the aggregate, advertising returned about one-third of a vote to the candidate (Rothschild 1978).

In a more recent study, Chapman and Palda (1984) monitored the influence of campaign expenditures on voting behavior from the view point of sales response function modeling. They used simultaneous equations to describe relationships among votes, candidates and campaign expenditures. The study employed eight election events in order to test for the presence of a generalizable pattern of results, and analyzed "aggregate" electoral district data which were free of survey errors as opposed to individual survey responses data. Preference of the electoral body is, however, revealed by the total number of votes received by each candidate. The study's main empirical result was that a candidate's "own" communications efforts (represented by campaign expenditures) positively affect votes received, while competitive communications efforts draw votes away from the candidate.

In the United States, Grush, McKeough and Ahlering (1978) studied the relationship between advertising expenditures and voting patterns for the 1972 Congressional primaries. The authors found advertising to have a significant impact, but that the level of the impact was moderated by the level of previous exposure of the candidate. Advertising expenditures were less significantly related to election outcomes when the candidate had previous exposure

Soley and Reid (1984) report on a more recent study of the 1974 and 1976 elections to the U S Congress. They tested and confirmed the hypothesis that the percentage of votes received by a congressional candidate is significantly related to total promotional expenditure by the candidate. That was the case even when the candidate's incumbency status and party affiliation were controlled Based on an analysis of covariance of the percentage of votes received by each candidate they concluded that promotional expenditures were as important a determinant of votes received by each candidate as incumbency status and party affiliation. Further, they report that the strength of the relationship between promotional expenditures and votes received was the same for incumbents and challengers and did not vary with Party affiliation

AIM OF PRESENT STUDY

Previous research has established the important impact of promotional expenditures on election outcomes A candidate's share of votes in an election is related to the level of promotional expenditures. This is similar to the results from studies conducted in "product" markets which have shown that the level of promotional expenditure is significantly related to a brand's market share (see, for example, Naert and Weverbergh 1981; Ghosh, Neslin and Shoemnker 1984: Brodie and de Kluyver 1984)

Studies of promotional effectiveness in product markets also posit that there may exist significant asymmetries in this effect across individual brands. Comanor and Wilson (1979), for example, argue that there is asymmetry between new and established brands due to differences in the degree of experience consumers have with the brands The result is a significant difference among brands in the marginal impact of promotional expenditures on sales and market share. Schmalensee (1972) also posits such asymmetries due to differences in familiarity with products and promotional skills. Theoretically, the asymmetries result from lover cross elasticities of demand between established and new brands that give established brands significant market power (Bain 1968; Comanor and Wilson 1979)

In the context of voting, this implies that the relationship between campaign expenditures and the percentage of votes received (market share) may be different for incumbents and challengers. In addition, the party affiliation of the candidate may also moderate the impact of campaign expenditures on voting patterns Data from a sample 77 Congressional elections in 1982 are analyzed in this study. The sample of districts were chosen to achieve a balance between geographical area, urban and rural districts and party affiliation of incumbents The data were obtained from FEC's Report on Election Campaign Finances and FEC Monitoring Activities 1983. Only districts in which there was a straight contest between a Republican and a Democratic candidate were included in the study

METHODOLOGY

An important step in testing an econometric model is the specification of the functional form of the relationship between the dependent and the independent variables (Naert and Leeflang 1978) Since there is no previous evidence on the most appropriate functional form for testing the relationship between campaign expenditures and voting patterns, two different functional forms -- a linear and a non-linear model -- are used in this study In each model the dependent variable of interest is the percentage of the votes received by candidates (V) However, since the voting procedure is viewed as a binary choice by individual voters and the percent votes polled by each candidate is constrained between zero and one hundred, a logit transformation is used in calibrating the models. If Vi is the percent votes received by a candidate i, we define:

TVi=ln[Vi/(1.0 - Vi)]

to obtain the transformed variable. The independent variables used are: (i) campaign expenditure by the candidate as a percentage of the total expenditure in the district (CE), (ii) the incumbency status of the candidate and (iii) the candidate's party affiliation. Details of the models used in the study are shown in Table 1. We now briefly discuss the rationale behind the development of the models

If the functional relationship between TV and CE is linear, a simple model of this relationship can be written as follows:

TV = a + b(( E) + e   (1)

Equation (1), however, does not account for any potential asymmetry in the impact of CE on TV due to incumbency status and party affiliation. In order to test for the possible impact of incumbency status and party affiliation on the overall relationship, we define three dummy variables (INCD, CHD and CNR) to categorize the candidates into different incumbency-party affiliation groups. The variable INCD is assigned the value of one if the candidate is an incumbent and is affiliated with the Democratic party It takes the value of zero in all other cases. The variable CHD takes the value of one if the candidate is a Democrat and a challenger, otherwise it is assigned the value of zero Similarly, the value of CHR is one when the candidate is a challenger affiliated with the Republican party Assigning values to INCD, CHD and CHR in this manner also defines the fourth group of candidates. Candidates who are affiliated with the Republican party and are incumbents have the value of zero on each of these three variables.

There are two potential sources of asymmetry in the relationship postulated in equation (1) First, the overall effectiveness of campaign expenditures may be different for the four groups of candidates This may result, for example, from their differential abilities in raising campaign funds or because of historical voting patterns The existence of such asymmetries may be tested by modifying the original relationship as follows:

V = a0 + a1INCD + a2CHR + a3CHD + b(CE) + e   (2)

The hypothesis of differences in overall effectiveness will be supported if the estimated values of the parameters a1, a2, a3 are significantly different from zero.

A second source of asymmetry is differences in the marginal impact of campaign expenditures across the four groups of candidates Thus, the parameter b may itself vary depending on the incumbency status and party affiliation of the candidate. We assume that differences in marginal impact of campaign expenditures can be modeled as follows:

b = b0 + b1INCD + b2CHR + b3CHD    (3)

Nov substituting equation (3) in equation (2) we get:

V = a0 + a1INCD + a2CHR + a3CHD + b0CE + b1(CE) INCD + b2(CE)CHR + b3(CE)CHD + e   (4)

Equation (4) is the linear model shown in Table 1 Note that the parameters a1, a2 and a3 measure the extent of "mean shift" while parameters b1, b2 and b3 measure changes in the slope of the regression equation As symmetries in the relationship between V and CE can now be checked by testing the statistical significance of the estimates of these parameters.

TABLE 1

The second model shown in Table 1 is developed in a manner analogous to the preceding discussion and the parameters of these models can be interpreted similarly. The difference between the two models is that they assume a different underlying functional relationship between CE and TV The non linear model starts with the following relationship instead of the one shown in equation (1):

TV - a + b(ln CE) + e   (5)

The linear model is commonly used in many marketing studies Its drawback, however, is the assumption of a constant return to scale implicit in the linear functional form. In this model the change in votes polled due to a marginal change in the value of CE is the same for all values of CE In the non linear model, on the other hand, the impact of a change in CE on votes depends on the value of CE. Consequently, the implied elasticity of the percentage of votes polled by a candidate with respect to change in campaign expenditures is different Consider, for example, a candidate who is a Democrat and also an incumbent (INCD = 1, CHD - CHR = O). For this candidate, the change in TV due to a marginal change in CE is (b + b ), for the linear model. In the non linear model it is (b0+ b1)V/CE. In this mode the marginal change depends on the value of CE and TV Given the lack of any previous evidence on the functional form underlying the relationship between expenditures and votes, we test both these alternative specifications rather than a priori selecting one of them.

RESULTS

The results of estimating the two models are presented in Table 2 In general, the results confirm the presence of significant asymmetries in the relationship between campaign expenditures and voting patterns. Irrespective of the functional form, one or more of the parameters in the model are statistically significant at the p = .05 level. The asymmetries are caused by differences across the groups in the overall level of effectiveness as well as by differences in the marginal impact of campaign expenditures on voting patterns. The models in each case fits the data very well. The adjusted R value for the linear model is .83 and for the non linear model it is .85. These values, however, indicate the fit between the transformed actual and predicted values. To measure the true degree of fit an inverse transformation of the predicted logs odd ratios we e performed. The correlation between these values and the actual percentage of votes polled by candidates was then determined. The Pearson correlations were .918 and .926 for the linear and the non linear model, respectively. To test the reliability of the results split-half tests were performed. The data were randomly split into two groups and the models estimated for the two groups separately. In each case the results for the separate groups were similar to each other and to the results for the overall sample.

TABLE 2

PARAMETER ESTIMATION OF THE MODELS

The impact of party affiliation and incumbency status on the effectiveness of campaign expenditures can be judged from the estimated values of the different parameters. As mentioned earlier, there are two ways in which such asymmetries can arise. The parameters a1, a2 and a3 measure the difference in the effectiveness of campaign expenditures by incumbent Democrats, Republican challengers and Democratic challengers respectively from that of Republican incumbents. The parameters b1, b2 and b3, on the other hand, measure the differences in the marginal impact of campaign expenditures.

Parameter a1 is positive and statistically significant at p=.05 level in both models. This indicates that Democratic incumbents have an advantage over their Republican counterparts. Campaign expenditures by Democratic incumbents have a greater effect on votes than campaign expenditures of incumbents affiliated with the Republican party. The parameters a2 and a3 measure the difference between the two challenger groups from Republican incumbents. These parameters are negative in both models. Only in the non linear model, however, they are statistically significant from zero. The non linear model results suggest that campaign expenditures by challengers, whether Democratic or Republican, are less effective than those of the base group (Republican incumbents). Thus there is significant advantage associated with incumbency and this advantage extends further for Democratic incumbents. The advantage of Democratic incumbents in the sample is also evident in the linear model. However, in this case, the Republican incumbents have no advantage over challenging candidates.

In addition to asymmetries in overall effectiveness, the marginal impact of changes in the level of campaign expenditures also varies among the groups. The magnitude and direction of these differences exhibit a number of interesting patterns. First, the parameter b is statistically insignificant in both models. Since this parameter is associated with the Democratic challenger group, this implies that there is no difference in the marginal impact of changes in campaign expenditures by Republican incumbents and Democratic challengers. Second, in both models b and b are significantly less than zero. Compared to Republican incumbents, the marginal impact of expenditures by Democratic incumbents and Republican challengers are significantly less. Moreover, the magnitudes of these two parameters are very similar in both the linear and non linear models. It is interesting that the marginal impact of campaign expenditures does not vary significantly for the groups of candidates who would oppose each other in an election -- Republican incumbents versus Democratic challengers and Democratic incumbents versus Republican challengers. The marginal impact for the first pair, however, is significantly greater than that for the second pair.

DISCUSSION AND CONCLUSIONS

Based on the sample of Congressional elections included in our study, there are significant asymmetries in the impact of campaign expenditures on election outcomes. These asymmetries are related to the incumbency status and party affiliation of the candidates. The study identifies two sources of these asymmetries. First, there are differences in the overall level of effectiveness of campaign expenditures. Second, the marginal impact of changes in expenditures on voting outcomes differ among the groups. That is, incumbency status and party affiliations affect the elasticity of votes with respect to campaign expenditures.

In general, campaign expenditures by incumbents are more effective than those by challengers. Democratic incumbents, especially, enjoy considerable "market power". Our result is consistent with that of Soley and Reid (1984) who found similar advantages for incumbents in the 1974 and 1976 elections to the U.S. Congress. The finding is also consistent with the general argument that early entrants in a market have considerable advantages over later ones because of familiarity and loyalty.

The effectiveness of expenditures by incumbents is also moderated by their party affiliations. Democratic incumbents, in our sample, have an advantage over their Republican counterparts. On an average these candidates received 65.3% of the votes compared to 61.2% obtained by their Republican counterparts. The added advantage of Democratic incumbents may be related to the length of their incumbency. Unfortunately, unavailability of information precluded a formal test of whether the effectiveness of campaign expenditures varied with length of incumbency.

Even after controlling for differences in the overall effectiveness of campaign expenditures, asymmetries appear to exist in the marginal impact of these expenditures. The statistical significance of the b parameters in our results indicates that the elasticity of votes with respect to campaign expenditures varies among the different groups of candidates. In this respect our results differ from those of Soley and Reid (1984). In their covariance analysis they found no significant interaction between campaign expenditures and incumbency status. Note that if incumbency status or party affiliation of candidates were analyzed separately, there may be no interaction between campaign expenditures and these variables taken one at a time. Analyzing party affiliation and incumbency status jointly reveals an interesting pattern. As noted earlier, the marginal impact is higher for Republican incumbents and Democratic challengers compared to the two other groups. However, there is no significant difference in impact between these two groups. Similarly, changes in campaign expenditures by Democratic incumbents and Republican challengers have the same marginal impact. Thus, in any given election, changes in expenditure by either candidate would have a similar impact.

Our study is exploratory in that only a subset of elections were chosen for analysis. Results from a limited sample must always be interpreted with caution. Our findings do, however, point to significant asymmetries in the political market place. The consistency of findings from the two alternative models also adds to the validity of our results. Future studies of the impact of campaign expenditures on voting should probe further the nature of these asymmetries and their sources.

There are a number of avenues for extending our analysis. One interesting approach is to look at the dynamic patterns of fund raising and voting outcomes within a single framework. Presently, our analysis is essentially static. It looks only at the impact of campaign expenditures on voting patterns. A candidate's campaign expenditures, however, depend on his or her ability to raise funds. a is aspect of electoral competition is rather neglected in many studies. An exception is the study by Chapman and Palda (1984) who propose a simultaneous equation model in which campaign expenditures depend, among other things, on the strength of the candidate's party and the closeness of the election. An alternative approach to linking fund raising abilities and voting outcomes is a time series model. In such a model, campaign expenditures are linked to past performance of the candidate's party and the closeness of the previous election. In applying this model, information for a series of elections from the same electoral districts is necessary. This will provide a better insight of the dynamics of the electoral process.

REFERENCES

Bain, J.S. (1986), Industrial Organization, New York: Wiley & Sons, p. 252.

Brodie Roderick, and Cornelis A. de Kluyver (1984), "Attraction Versus Linear and Multiplicative Market Share Models: An Empirical Evaluation," Journal of Marketing Research, Vol. 21 (May), 194-201.

Chapman, Randall G., and Kristian S. Palda (1984), "Assessing the Influence of Campaign Expenditures on Voting Behavior with a Comprehensive Electoral Market Model," Marketing Science, Vol. 3, No. 3 (Summer), 207-226.

Comanor, W.S. and T.A. Wilson (1979), "> e Effect of Advertising on Competition: A Survey," Journal of Economic Literature, Vol. 17 (June), 453-476.

Crain, Mark W. (1977), 'tn the Structure and Stability of Political Markets," Journal of Political Economy, Vol. 85. No. 4, 829-842.

DeVries, Walter and V.L. Torrance (1972), The Ticket- Splitter: A New Force in American Politics, Grand Rapids, Michigan: William B. Erdmans Publishing Co.

Ghosh Avijit, Neslin Scott, and Robert Shoemaker (1984), "A Comparison of Market Share Models and Estimation Procedures " Journal of Marketing Research, Vol. 21 (May), 202-210.

Grush, Joseph E., Kevin L.McKeough, and Robert F. Ahlering (1978), "Extrapolating Laboratory Exposure Research to Actual Political Elections," Journal of Personality and Social Psychology, Vol. 36, 257-270.

Jacobson, G.C. (1980) Money in Congressional Elections, New Haven: Yale University Press.

Lazarsfeld, Paul, Bernard Berelson, and Hazel Gaudet (1944), The Peoples Choice: How the Voter Makes Up His Mind in a Presidential Campaign, New York: Columbia

Naert, Philippe, and Peter Leeflang (1978), Building Implementable Marketing Models, Boston, Mass.: Leiden, Martinus Nijhoff Social Sciences Division.

Naert, Philippe, and M. Weverbergh (1981), "On the Prediction Power of Market Share Attraction Models," Journal of Marketing Research, Vol. 18 (May), 146-153.

Palda, Kristian R. (1975), "The Effect of Expenditure on Political Success," Journal of Law and Economics, Vol 18. (December). 745-771.

Patterson, T.W. and R.D. McClure (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 (May), 88-97.

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

Schmalensee, R. (1972), The Economics of Advertising. Amsterdam: North-Holland.

Soley, Lawrence C., and Leonard N. Reid (1984), "Promotional Expenditures in U.S. Congressional Elections," Journal of Marketing and Public Policy, Vol. 1, 147-155.

Swinyard, W.R., and K.A. Coney (1978), "Promotional Effects on a High - Versus Low-Involvement Electorate," Journal of Consumer Research, Vol. 5 (June), 41-48.

Welch, W.P. (1976), "The Effectiveness of Expenditures in State Legislative Races," American Politics Quarterly, Vol. 4 (July), 333-356.

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