Voting Participation in a Public Consumption Perspective

Randall G. Chapman, University of Chicago
Kristian S. Palda, Queen's University
ABSTRACT - Data from ten Canadian provincial elections and census-derived socioeconomic variables are used to estimate a model of voting participation. The model is developed, and the empirical results are interpreted, within the context of rational voter and household investment-consumption theories.
[ to cite ]:
Randall G. Chapman and Kristian S. Palda (1981) ,"Voting Participation in a Public Consumption Perspective", in NA - Advances in Consumer Research Volume 08, eds. Kent B. Monroe, Ann Abor, MI : Association for Consumer Research, Pages: 530-533.

Advances in Consumer Research Volume 8, 1981      Pages 530-533


Randall G. Chapman, University of Chicago

Kristian S. Palda, Queen's University

[Funding for this study was provided by a grant from the Social Science and Humanities Research Council of Canada.]


Data from ten Canadian provincial elections and census-derived socioeconomic variables are used to estimate a model of voting participation. The model is developed, and the empirical results are interpreted, within the context of rational voter and household investment-consumption theories.


Voting decisions, an integral part of the household investment-consumption process, can be studied from a consumer behavior perspective. Theories of rational voting and consumer behavior are invoked in this paper to develop a voting participation model. This model is a component of a larger integrative electoral market model currently under development (Chapman and Palda 1980). The empirical results presented here relate the number of votes cast in a constituency to a range of political and socioeconomic variables using data from ten Canadian provincial elections and from the 1970 census. A goal of this paper is to test the extent to which generalizations can be made about the forces affecting voting participation. The results obtained show substantial consistency across tine and provinces and are of particular interest with respect to income and education effects on voting turnout.


Participation in the political process is generally viewed as indicating the vigor of an individual's commitment to the democratic process. The most obvious manifestation of such participation is the act of voting. It is not surprising, therefore, to find that studies of electoral turnout stand at the center of attention of students of politics. Not unexpectedly, there are competing theories of voting turnout behavior.

The dominant voting paradigm among political scientists and non-economists is based on a social-psychological perspective which sees social integration as leading to a psychological attachment to the political system and thence to participation (Irvine 1976, Popkin et. al. 1975). The key concept is socialization: if an individual is socialized into the prevailing norms of society or its subgroups, he is likely to vote due to feelings of obligation. In empirical studies, education, income, age, race, and membership in certain organizations have been employed as indicators of the degree of political socialization (Kim, Petrosik, and Enokson 1975).

Economists and some political scientists prefer a theory of voting behavior which rests on the assumption, proposed by Downs, that voters are "rational": they evaluate available alternatives in light of their benefits and costs and implicitly employ a decision calculus based on maximizing expected utility (Downs 1957). In deciding whether or not to vote, the Downsian rational individuals cast their votes only when the expected utility of as doing is positive. This decision calculus can be represented compactly by the following behavioral equation (Frolich et. al. 1978):

U = p1 |D| + p2L - C   (1)


U = the net value (expected utility) an individual associates with voting

p1 = the probability of an individual's vote affecting the outcome of the election

D = the expected differential between the winning and the second-place party

p2 = the probability that an individual's vote will make a difference in the survival of the democratic political system

L = the long-run value of participation

C = the cost of voting.

The Downsian view of the voting participation decision implies that the act has combined consumption and investment features (Stigler 1975). Not only would lower costs lead to a higher turnout (as might be expected on both consumption and investment grounds) but, in addition, we would expect a larger turnout the greater the prospective investment return to the voter.

The accent on the investment perspective is more than a pedantic quibble because the socio-psychological and economic predictive turnout models use intersecting sets of variables to operationalize their underlying constructs. The socio-psychological school holds that income, education, and age level represent suitable proxies for social integration. From the Downsian perspective, these same variables represent measures of opportunity and information processing costs of voting. The differentiating elements between the two theories are most pronounced on the side of the long-run investment costs and benefits yielded by voting, much as taxation to the haves and government transfers to the have-nots.

Why should the student of consumer behavior contemplate these two conflicting theories of voting participation with other than benign indifference? The answer lies in the enormous economic consequences of voting in our post-industrial society. The share of government expenditures in the GNP ranges from 25 percent in Japan to 33 percent in the United States to more than 50 percent in the UK, among the OECD countries, and the taxes collected to finance such expenditures represent the single largest bite out of the income of the consumer-voter-taxpayer (Palda 1980). To the extent that the consumer can influence the size of the tax bite or his social benefits by hise vote, his decision to vote (and for whom) can be viewed as a discretionary 'buyer-type' activity. In this sense, the sphere of voting is of legitimate concern to the marketing and consumer behavior student.

To these writers, at least, the Downsian approach appears to lead more naturally to a marketing perspective of the voting participation decision. Such a perspective would stipulate that voting represents the selection of a purchasing agent (elective representative) for a period of one or more years. This agent is endowed with powers to decide not only on the mix of public goods to be provided but also on the amount of the voter's income to be appropriated. The voting decision can thus be seen as an investment in a purchasing agent. It exhibits standard investment aspects such as a multiperiod time horizon, revenues (public goods) and costs (taxes) to be compared on a net present value basis, scrap value of the agent's incumbency, a rate of discounting which reflects personal preferences, and risks associated with the prediction of the capital equipment's (the agent's) expected performance. (The riskiness could be proxied with survey-obtained and regression-estimated standard deviations of linear multi-attribute model coefficients. The D term in equation (1) seems most amenable to being modeled in terns of ideal point distance models and estimation procedures (Ordeshook ( 1976 ). )

To bring in consumption aspects of voting behavior, an appropriate integrating framework is that of the household production function (Wittman 1975). In the Beckerian (1965) perspective, households purchase intermediate goods in the market and combine these with their labor inputs and domestic capital equipment to produce final outputs which yield utility. The 'capital equipment' within the voting context (elected purchasing agents) is not a property of the household "factory" but rather a part of the "industrial" infrastructure which, unlike in the orthodox industrial case, the household members can influence by their vote. By virtue of their personal labor skills, perhaps acquired by education, certain households will be more proficient at processing electoral information and so will be more likely to vote. Practice may lead to consumption preference and so to direct utility derived from voting. Other household characteristics, such as the presence of small children, may raise electoral consumption costs by allowing less time to acquire political information and to go to the polls.

To sum up, with some additional effort the existing theory of rational voting can be brought closer to the traditional preoccupation of consumer behavior scholars. The central economic importance of voting behavior is a sufficient justification for their attention to the voting participation phenomenon. The next section outlines a model employed to account for electoral participation in Canadian provincial elections and sketches its relationship to a Downsian-household production function approach.


A comprehensive theory of voting must necessarily apply to both the decision to opt for a particular candidate (an elective purchasing agent) and the decision to vote, this paper's main concern. The Downsian rational voter theory has been substantially clarified, tested, and elaborated upon over the last 20 years. Its integration with household production theory is in its infancy. We cannot do entire justice to a complete Downs-Becker specification because we do not have access to individual level survey data which are needed to assess the party differential variable D and the p2L term in equation (1). The model of turnout which we propose and test in this paper does, nevertheless, drew its inspiration from the rational voter theory, supplemented by recourse to the economics of information.

Our model of voting participation may be described in the following general terms:



CAST = the number of votes cast (in 000s)

REGVOT = the number of registered voters (in 000s)

CLOSE = the difference between the vote share of the winner and the runner-up (ex post)

INC = average per capita income (in $000s)

ED = percent of population over 15 years of age with at least complete high school education

TOTEXP = the total reported campaign expenditures of all major candidates ( in $000s)

UNEMP = the percent unemployed workers in the labor force

FEMPAR = proportion of employed females in the total labor force.

All of the variables in equation (2) refer to a constituency in a given Canadian provincial election. The constituency socioeconomic variables- INC, ED, UNEMP, and FEMPAR -- were obtained from special Statistics Canada runs in which the 1970 census data were geocoded along the boundaries of the provincial constituencies.

REGVOT serves as a constituency-size scaling variable. It should be noted that in Canada, unlike in the United States, voter registration is done by enumerators and does not, therefore, require voluntary effort on the part of the individual qualified citizen-voter.

CLOSE, an ex-post variable, measures the probability (p1 in equation (I)) of an individual's vote affecting the outcome of the election (Barrel and Silberberg 1973). Theory predicts a negative sign on its coefficient: the narrower the margin, the greater the impact of each vote, the higher the incentive to turnout.

INC, the average par capita income (as of the 1970 census), has at least a three-fold interpretation. Simple economic theory suggests that there is both a substitution effect, in terms of opportunity costs facing higher income voters, as well as an income effect, related to an explanation of voting as a consumption good. These effects cut in opposite directions and offer no unambiguous qualitative conclusion as to the expected sign of this variable (Tollison, Crain, and Pautler 1975, Grain and Deaton 1977). In the absence of information on property and income subject to provincial taxation (i.e., on variables which Mould throw light on the investment aspect of voting), INC might reflect the taxation stake of the voter and be expected to exhibit a positive coefficient. The composite of all of these forces will determine the sign of INC. If it turns out to be negative, opportunity cost aspects predominate.

The economic theory of information would predict, ceteris paribus, a positive association between turnout and the level of education possessed by voters-consumers: the higher the education level, the lower the costs incurred in acquiring and processing political information (Goldman and Johansson 1978). As for ED, so for TOTEXP: campaign expenditures would be expected to have a positive influence on turnout as they represent the lowering of costs of information acquisition (Abrams and Settle 1978).

An economics of information perspective would also predict that the unemployed, likely to be on average less educated, could encounter higher costs in acquiring and processing electoral information. (The socio-psychological approach to electoral participation would suggest that the unemployed are disaffected with the political process, and thus would have a tendency to turn out in lesser numbers.) They would have, however, lower income opportunity costs of voting and possibly a greater stake in voting for left-leaning parties to the extent that provincial governments, rather than the federal government, can be held accountable for employment opportunities. Thus, theory is ambiguous with regard to the sign of UNEMP.

Finally, FEMPAR also has different possible interpretations as to expected sign. On one hand, high levels of female participation would be associated with higher incomes and possibly a greater "tightness" in household time allocation thus increasing opportunity costs. On the other hand, on-the-job discussions of politics and world events might lead to lower information costs.


The model in equation (2) was estimated using data iron each of ten Canadian provincial elections. A linear-additive functional form was chosen on the grounds of parsimony. To avoid constituency size affecting absolute campaign spending, the expenditure variable Was formulated in dollars spent par registered voter (i.e., as TOTEXP/REGVOT). The results of estimating the model are reported in Table 1. It should be noted that our a priori theory led us to use one-tailed tests of significance for REGVOT, TOTEXP/REGVOT, ED, and CLOSE (where we expected three positive and one negative signs, respectively) and two-tailed tests of significance for UNEMP, INC, and FEMPAR (where theory was ambiguous with respect to signs).

Two comments are in order regarding the linear-additive functional form of the voting participation model. The model was also estimated in logarithmic form to assess the possible presence of significant interactions among the independent variables. The results obtained were consistent with those from the simpler linear-additive functional form, as the more parsimonious additive version was chosen. A second functional form issue concerns the dependent variable. The usual dependent variable in a voting participation model is TURNOUT (=CAST/REGVOT). A linear-additive model with TURNOUT as the dependent variable was estimated for comparison purposes. Its results were also consistent with our results using CAST as the dependent variable. Again, parsimony led us to prefer the simpler version.

In interpreting the empirical results, patterns of generalizability were looked for. An important feature of the research reported in this paper is the large amount of data assembled to estimate the voting participation model. Most politometric research is based on a single electoral event only. The ability to generalize from such results may be open to question. With ten elections as a data base, the results of this study lend themselves to attempts to document generalizable and consistent patterns of voting behavior.

REGVOT, a scaling variable which standardizes for the size of the constituency, has naturally the largest (positive) impact. The high explanatory power of the model derives from simple correlations between CAST and REGVOT which are 0.90 or more in magnitude.

In three of the ten elections, per registered voter campaign expenditures have a significantly positive impact on turnout. A significantly negative effect is found in one election; no obvious explanation is available for this anomalous result. It may be noted that these four significant results occur in the largest and most urbanized Canadian provinces (Ontario and Quebec). A public policy question relates to the possible impact of campaign expenditure ceilings on voting behavior. Legislated ceilings on individual candidate expenditures were in effect in all provinces except British Colombia. The avowed purpose of spending limits is to restrict the 'buying' of seats on the assumption that campaign expenditures influence voting outcomes. This assumption has been repeatedly empirically verified (Chapman and Palda 1980). Spending limitations might, therefore, have the unintended and socially undesirable side effect of constraining the flow of information from candidates to voters, thus effectively increasing the information acquisition costs to the voters and possibly leading to reduced voter participation. Our empirical results here do not support, on the whole, the hypothesis that overall constituency spending affects voting participation.

The closeness of the electoral contest has a significant impact on participation in six of the ten elections. (The negative sign of CLOSE means that as the difference between winner and runner-up narrows, turnout increases.) This result is in line with previous empirical tests of the Downsian theory and supports the view that a stronger probability of influencing the election outcome moves the voter to cast his ballot.



Perhaps the most interesting results are derived with regard to the income and education variables, taken together. All of the ten INC coefficients are negative, of which five are significant. Nine of the ED coefficients have a positive sign; the one negative coefficient is not statistically significant. As far as we can tell, this is the first time (certainly with aggregate-level data) that the separate effects of income and education on voting participation are assessed -- and this in face of the high positive correlations between them (from 0.60 to 0.94 in our sample). The strong consistency of our results militates for the following interpretation. The substitution or opportunity cost effect predominates with regard to income: as income increases, holding education constant, voters tend to participate less since they value their tine more highly. (This result comes through even though the average income measure used is not totally "pure" for it would have been preferable to use income derived from employment, unavailable to us, and to have had access to an investment-type taxation-stake variable (Denver and Hands 1974).) Higher education, on the other hand, leads to increased voter turnout as it reduces the costs of acquiring and processing political information.

Nine out of the ten coefficients of UNEMP are negative, seven of these significantly so. Six of these significant results are obtained in the industrially-oriented provinces of British Columbia, Ontario and Quebec. Since education and income determinants are held constant at their sample means, the only Downsian interpretation plausible is that the unemployed do not perceive provincial governments as capable of job creation. (Note, however, that the socialization perspective would consider the structurally unemployed as participating less in the life of the polity.)

Six of the ten FEMPAR negative coefficients are statistically significant. While there is a pattern of positive correlation between FEMPAR and INC and between FEMPAR and ED (in the 0. 35 to 0.65 range), the added influence of feminine participation in the labour force is suggestive of non-trivial household time allocation costs.

In summary, the model's empirical results bear out rather convincingly the economic view of rational men faced with the decision to vote in light of his costs, benefits and the uncertainty he faces and taking account of the household context. The generally consistent pattern of the results over the ten elections has been arrived at despite the aggregate, constituency-level nature of the variables used and some strong collinearity among the socioeconomic determinants.


The model of voting participation reported upon here constitutes the initial component of a larger recursive electoral market model which will attempt to account for party candidate success (with special emphasis on the role of campaign expenditures), candidate expenditures and the contribution generation mechanism (Chapman and Palda 1980). The elaboration and testing of this integrated model with data from the ten Canadian provincial elections is being undertaken within the framework of the rational voter theory. The pattern of consistently plausible results obtained with regard to voting participation augurs well for the forthcoming study.


Abrams, Burton A. and Settle, Russell T. (1978), "The Economic Theory of Regulation and Public Financing of Presidential Elections," Journal of Political Economy, 86, 245-257.

Barzel, Yoram and Silberberg, Eugene (1973), "Is the Act of Voting Rational?' Public Choice, 22, 51-58.

Becker, Gary S. (1965), "A Theory of the Allocation of Time," The Economic Journal, 75, 493-517.

Chapman, Randall C. and Palda, Kristian S. (1980), "Voting Patterns and Campaign Expenditures in Canadian Provincial Elections," Forthcoming in Proceedings of the 52nd Annual Meetings of the Canadian Political Science Association.

Crain, W. Mark and Deacon, Thomas H. (1977), "'A Note on Political Participation as Consumption Behavior," Public Choice, 26, 131-135.

Denver, D. T. and Hands, H. T. G. {1974), "Marginality and Turnout in British General Elections," British Journal of Political Science, 4, 17-35.

Downs, Anthony (1957), An Economic Theory of Democracy, New York: Harper and Row.

Frolich, Norman, et. al. (1978), "A Test of Downsian Voter Rationality: 1964 Presidential Voting," American Political Science Review, 72, 178-197.

Goldman, Arieh and Johansson, J. K. (1978), "Determinants of Search for Lower Prices: An Empirical Assessment of the Economics of Information Theory," Journal of Consumer Research, 5, 176-186.

Irvine, William A. (1976), "Testing Explanations of Voting Turnout in Canada," in Ian Budge, et. al. (eds.), Party Identification and Beyond, New York: John Wiley and Sons.

Kim, Jae-On, Petrosik, J. R., and Enokson, S. N. (1975), "Voter Turnout Among the American States: Systemic and Individual Comparisons," American Political Science Review, 69, 107-131.

Ordeshook, Peter (1976), "The Spatial Theory of Elections: A Review and a Critique," in Ian Budge, et. al. eds.), Party Identification and Beyond, New York: John Wiley and Sons.

Palda, Kristian S. (1980), 'The Public Sector and Marketing: Should the Twain Meet?" Forthcoming in Proceedings of the Second Annual Special AMA Conference on Marketing Theory.

Popkin, Samuel L., et. al. (1976), "Comment: What Have You Done For Me Lately? Toward an Investment Theory of Voting," American Political Science Review, 70, 779-805.

Stigler, George J. (1975), "Comment," Journal of Law and Economics, 18, 743-744.

Tollison, Robert, Crain, Mark, and Pautler, Paul (1975), "Information and Voting,' Public Choice, 24, 43-49.

Wittman, Donald (1975), "Comment," Journal of Law and Economics, 18, 735-741.