OUP user menu

“…Take up thy bed, and vote” Measuring the relationship between voting behaviour and indicators of health

Kevin J. Denny, Orla M. Doyle
DOI: http://dx.doi.org/10.1093/eurpub/ckm002 400-401 First published online: 30 April 2007


In most countries health policy is an important part of the political agenda. Yet few studies have examined the relationship between the two. This study investigates the association between health and voter turnout in Britain using the National Child Development Study. Self-rated general health, the Malaise Inventory score and indicators of smoking and alcohol consumption, as measured at ages 23, 33 and 42, are regressed on voter turnout in the 1979, 1987 and 1997 general elections. The results indicate that individuals with poor general and mental health and smokers are less likely to vote at election time.

  • mental health
  • NCDS
  • self-rated general health
  • smoking and alcohol consumption
  • voter turnout

In most countries health policy is an important part of the political agenda. Yet few studies have quantified this link between health issues and voting behaviour. A study of mortality rates and voter turnout in Britain finds a negative correlation between the two.1 Similar results have been found in a study of life expectancy and turnout in Russia2, and general health and state level turnout in the US3. These studies were conducted at a constituency level and none to date have investigated whether this relationship exists at an individual level.

Individuals experiencing ill-health may be more likely to vote at election time since they are more likely to be users of the health system and favour public provision of health services. It is important to investigate if this is true given that poor health may also act as a barrier to voting. As voting requires both a physical and a mental effort, poor health may impair one's ability or willingness to vote. A study of voter turnout finds that those with disabilities are 20% less likely to vote compared to otherwise similar individuals.4 Adverse health may also lead to a reduction in psychological resources, such as political interest, political efficacy and civic values.

Evidence suggests that major life events, such as economic adversity5, marriage and having children6, can affect turnout decisions as time, energy and resources are drawn elsewhere. This paper examines whether the same applies for health adversity. It tests whether the relationship between health status and electoral participation holds at the individual level, while controlling for a host of social circumstances over the course of three British elections.


The data used in the analysis is the National Child Development Study, a longitudinal study of all persons living in Britain who were born in one week in March 1958. The dependent variable is whether the individual voted in the 1979, 1987 and 1997 General elections.

The first health variable is a self-assessed measure of general health indicating whether the respondent reported excellent/good health, or alternatively, poor/fair health. The second is the ‘malaise inventory score’ developed by Rutter et al.7 This self-completed scale is derived from summing 24 psychological and somatic items, such as anxiety and irritability. A score above 7 is associated with a high-risk of psychiatric morbidity. The third health variable indicates whether the respondent is a smoker or not. The final variable describes whether the respondent is a non-drinker, a moderate drinker or a heavy drinker and is based on alcohol consumption in the week prior to the survey. Moderate drinking is defined as consuming 14 or less units for a woman and 21 for a man. Consuming above these units corresponds to heavy drinking.

The control variables included are standard in the voting literature-region, sex, education and whether the respondent is married, has children, is unemployed or self-employed at the time of each election. To control for the possibility that the health indicators reflect social circumstances we also include the fathers/mothers social class at birth, overcrowding in the household at age 7, and three indicators of absence from school due to illness at age 11.

As turnout is a binary outcome a probit regression model is estimated for each election, with the four health indicators as covariates, in addition to the control variables. The models are estimated using Stata version 9. The table reports marginal effects: the effect of a unit change in the covariates on the probability of voting.


The results for each election are in Table 1. Controlling for a set of social circumstances measured in childhood and adulthood, poor general health, poor mental health, smoking and abstaining from drinking alcohol are associated with lower turnout. Individuals with poor general health are ∼4% less likely to vote in the 1979 and 1997 elections. Similarly, smokers are 4% less likely to vote in these elections and 3% less likely to vote in the 1987 election compared to non-smokers. There is a negative relationship between mental health and turnout in the 1987 election; however, this effect is small: a one standard deviation increase in the malaise score is associated with a 0.5% lower probability of voting. Moderate drinkers are more likely to vote in the last two elections than non-drinkers, while heavy drinkers have a greater probability of voting in the 1997 election.

View this table:
Table 1

Modelling voter turnout

Poor general health−0.043**−0.021−0.040***
Poor mental health−0.002−0.005**−0.002
Moderate drinker0.0160.028**0.041***
Heavy drinker0.0250.0070.024*
  • Note: All models estimated using probit. The following variables are included but not reported—region of residence, sex, fathers/mothers social class at birth, overcrowding in household at age 7 (defined as more than one person per room), three binary variables indicating how long respondent missed school due to illness at age 11, age left full-time education, whether stayed in school beyond 16 and whether the respondent is married, has children, is unemployed or self-employed at the time of each election. Marginal effects and standard errors (in parentheses) are reported. Significance levels: *:p < 0.1. **: p < 0.05. ***: p< 0.01

The control variables behave as per standard voter turnout models. Higher education, higher parental social class and being married increase the probability of voting across all three elections. Being self-employed and living in overcrowded housing as a child lowers the probability of voting by 5 and 3%, respectively. In the 1979 election, being male, having children and being unemployed are also associated with lower turnout. Health status in childhood has no impact on voter turnout in later life. The impact of the health variables are largely comparable in size with the impact of these additional control variables.


Our results are consistent with the hypothesis that poor health is a contributory factor to individuals not engaging in political participation and suggest that the perceived cost of voting for the unhealthy (the effort involved in voting), is greater than the perceived benefits of voting (based on the policy implications of the election outcome). If elections have no consequences for health services then those in ill-health may perceive that there are low benefits from voting. This suggests that the main political parties do not differ enough in their policy positions on health to induce the unhealthy to vote. While the probability that an individual's vote is decisive is negligible (the paradox of voting hypothesis), voters derive other, psychic, benefits from voting.

The results indicate that smoking is consistently associated with lower turnout. This may be because smoking is an indicator of bad health and should be interpreted in a similar manner to the poor physical/mental health hypothesis. Alternatively, smoking indicates a high-rate of time preference. Therefore, as voting concerns the future, the type of people who smoke are also the type of people who do not vote as they care less about the future. Alcohol consumption is an exception to the bad health/low turnout hypothesis, in that higher consumption predicts a higher probability of voting. One might speculate that non-drinkers are, on average, less sociable and hence less likely to engage in pro-social activities such as voting. While we have controlled for some aspects of childhood circumstances, it is impossible to control for all. Hence it is possible that smoking or drinking is correlated with some unobserved background effect or possibly a personality trait.

Electoral participation is one form of social capital and the level of voting is an important barometer of the health of civil society. A number of studies have noted the importance of social capital for generating both community and individual well-being.8,9 Understanding the relationship between public health and political participation is therefore important. This paper shows that poor health leads to lower electoral participation, which suggests that the interests of the unhealthy are less likely to be represented in government. Unhealthy non-voters may therefore represent an untapped source of electoral support.


Permission to use the NCDS given by the ESRC Data Archive at Essex is gratefully acknowledged. Thanks to Liam Delaney, David Madden and Pat Wall for comments and suggestions. There are no funding sources for this paper.

Key points

  • Areas with higher mortality in Britain have lower voter turnout in general elections, but is this result replicated at an individual level?

  • Individuals experiencing poor general and mental health and who smoke are less likely to turn out to vote at election time.

  • Given the low participation rates of the unhealthy, a political party which formulates an attractive policy package aimed at such potential voters could therefore mobilize a previously untapped segment of the electorate.


View Abstract