Next week, America goes to the polls. Based on recent presidential elections, more than one-third of eligible voters will not cast a ballot. The figure for general elections in the UK is similar, which is actually a recovery since 2001 when just over 40% of registered voters did not vote. This high rate of abstention has led to an active research agenda on declining political trust and growing disaffection (Hetherington 1999).
Why bother voting?
Yet rather than focus on non-voting, perhaps the bigger research question should be why anyone bothers to vote. At least since Riker and Ordeshook (1968), rational voter theory notes that a person will only vote if the costs are outweighed by the benefits that flow from the preferred candidate winning, discounted by the probability of casting the deciding vote, and allowing for the consumption benefit that results from the feeling of satisfaction in performing one’s civic duty.
The probability of casting the deciding ballot is infinitesimally small in national elections. Hence, the cost of voting normally exceeds any plausible value of expected benefits from the preferred candidate winning.
For example, Feddersen (2004) presents calculations for a two-candidate election, with five million voters and where Candidate X has an expected vote share of 50.1%. In order for expected benefits to outweigh costs of voting for a voter who prefers candidate Y, the benefit to the voter from Candidate Y winning must be more than eight billion times greater than the cost of voting.
It is unlikely that any politician’s platform can deliver benefit to a voter that is eight billion times greater than the cost of voting. So if voting is rational it comes down to a comparison of the costs with the consumption benefits. But if these costs and benefits are small, the decision to vote is sensitive to small variations in either term. If these small effects are hard to measure, individual voter turnout will seem largely random (Matsusaka and Palda 1999) and random voting presents a seemingly difficult challenge for the rational voter model.
New research on the opportunity cost of voting
Along with my collaborators, I have recently reported results in Public Choice (Gibson et al. 2012) with very precise measures of the opportunity cost of voting. To calculate these costs we cross-referenced individual voter turnout in a general election from New Zealand with GIS estimates of the road distance from residential areas to the nearest polling station. By combining travel time estimates from Google Maps with estimated wages for the survey respondents, we obtained a detailed measure of the opportunity cost of time spent travelling to and from the polling station.
Our results show that even very small costs may deter voter turnout. Each extra kilometre – or two minutes’ driving time – to the nearest polling booth reduces turnout by one percentage point, all else the same. These effects are robust to various sources of confounding, including endogenous sorting of residential location, measurement error, and non-linearities. These results support one implication of rational voter theory, first made by Niemi (1976) that “if the B (benefits) or PB (benefits weighted by the probability that a person’s vote matters) term is indeed quite small, then a small increase in the cost of voting – such as driving a mile instead of a half-mile to the polls – would significantly reduce turnout.”
Non-nested testing shows that using our new measure of the opportunity cost of time spent voting, formed by combining estimated travel time with imputed wages, outperforms simpler distance-based measures of costs. We find that small increases in the opportunity costs of time can have large effects in reducing voter turnout. For example, at an opportunity cost of NZ$10 (equivalent to US$8) the predicted national turnout would be just 75%, which is down seven percentage points from the mean. In urban areas, predicted turnout falls even more sharply with respect to opportunity costs.
While the external validity of findings from New Zealand would typically be limited, a number of features of this setting allow for especially clean estimates of the impacts of opportunity costs on turnout.
- First, the general election voting is on a Saturday and almost always in person, so it is reasonable to assume that people are travelling from their home (the locations of which we use in the GIS algorithm).
- Second, registering to vote is compulsory, while voting is not, so there is no two-stage decision to model of whether first to register and then to vote.
- Third, our measure of individual turnout comes from an electoral survey which is validated by checking against the electoral rolls, so there is no over-reporting of voting as often happens with other surveys.
- Finally, this is a setting with ample polling places per voter, little road congestion, little use of absentee ballots, and no state/provincial governments, upper house, or an elected executive or judiciary.
Hence, the triennial election for national parliament is the only politically important election in New Zealand, as well as the only one that involves in-person voting.
Showing that small opportunity costs of voting matter for voter turnout even in this setting extends and corroborates the findings of the previous, more spatially limited, case studies in the political science literature. An important implication follows from finding that opportunity costs are low, but that turnout is still sensitive to those low costs. If these low costs were not able to be accurately measured (as has typically been the case previously), then the decision to vote would (erroneously) appear largely to be random. While many may choose not to vote in the forthcoming elections, such decisions can still be considered as rational ones.
Feddersen, T (2004) “Rational choice theory and the paradox of not voting”, Journal of Economic Perspectives, 18(1):99-112.
Gibson, J, B Kim, S Stillman, and G Boe-Gibson (2012), “Time to vote”, Public Choice.
Hetherington, M (1999), “The effect of political trust on the presidential vote, 1968–1996”, American Political Science Review 93(2):311-326.
Matsusaka, J and F Palda (1999), “Voter turnout: How much can we explain?”, Public Choice 98(3): 431-446.
Niemi, R (1976), “Costs of voting and nonvoting”, Public Choice 27(1):115-119.
Riker, W and P Ordeshook (1968), “A theory of the calculus of voting”, American Political Science Review 62(1):25-42.