Design and setting
Data were collected between February 2015 and February 2020 from the ongoing Smoking Toolkit Study (STS), a monthly repeated cross-sectional survey of a representative sample of adults in England [28]. The STS uses a hybrid of random location and quota sampling to select a new sample of approximately 1,700 adults (aged ≥16 years) each month. Locations are randomly selected from around 170,000 output areas in England, stratified by a geodemographic classification of the population. Interviews are performed with one household member until quotas based on factors influencing the probability of being at home (e.g. sex, age, working status) are fulfilled. Comparisons with sales data and other national surveys show that the STS recruits a representative sample of the population in England with regards to key demographic variables, smoking prevalence, and cigarette consumption [19].
Ethical approval and consent to participate
Ethical approval for the STS was granted by the UCL Ethics Committee (ID 0498/001). All participants are treated in accordance with the principles of the Declaration of Helsinki. Written informed consent is obtained by all participants. The data is not collected by UCL and is anonymised when received by the research team.
Measures
Outcome variable
In order to measure party voting intentions, respondents were asked “How would you vote if there were a General Election tomorrow?” (1) Conservative (reference category; traditionally a right leaning party), (2) Labour (traditionally a left leaning party), (3) Liberal Democrat (considered a centrist party), (4) Green Party (traditionally a left leaning party), (5) UK Independence Party (right leaning party), (6) Other, (7) intended not to vote, (8) Undecided, (9) Refused.
Explanatory variable
Smoking status was determined by asking, “Which of the following best applies to you” (1) I smoke cigarettes (including hand-rolled) every day, (2) I smoke cigarettes (including hand-rolled), but not every day, (3) I do not smoke cigarettes at all, but I do smoke tobacco of some kind (e.g. pipe, cigar or shisha), (4) I have stopped smoking completely in the last year, (5) I stopped smoking completely more than a year ago, (6) I have never been a smoker (i.e. smoked for a year or more). Those who reported currently smoking cigarettes or tobacco of another type were considered to be a smoker. All of those who reported having stopped smoking within the last year or before were considered former smokers. All others were considered never-smokers.
Covariates
Sex was categorised by female/other, and age by category (16-24, 25-34, 35-44, 45-54, 55-64, and ≥65 years).
Occupational social grade
As measured by the National Readership Survey [29], comprises AB (higher and intermediate managerial, administrative and professional), C1 (supervisory, clerical and junior managerial, administrative and professional), C2 (skilled manual workers), D (semi-skilled and unskilled manual workers) and E (state pensioners, casual and lowest-grade workers, unemployed with recourse to state benefits.
Region
Region is presented by four divisions of England, North, Central, South and London.
Alcohol
The AUDIT [30] score was used as a continuous measure for alcohol drinking and associated behaviour and is a known confounder of associations between smoking and health outcomes. The AUDIT is a widely used measure of alcohol use designed to indicate alcohol use which is potentially harmful to health. A score between 0 and 7 indicate the lower risk category, scores of 8-15 indicate increasing risk, 16-19 higher risk and those of 20+ indicate possible dependence.
Newspaper readership
Participants were asked which daily national newspaper they read regularly (e.g., Daily Mail, The Sun, The Telegraph, The Guardian, (list not exhaustive)).
Analyses
This study was preregistered on the Open Science Framework: https://osf.io/fq7rd/. Within the protocol current smoker was indicated as the reference category; however, because never smokers are the largest category this forms the reference category here instead. Analysis was conducted using SPSS v. 26. Data analysis was conducted on complete cases for all variables (<5% missing at random) and descriptive data were weighted to match the English population profile on age, social grade, region, tenure, ethnicity, and working status within sex. The dimensions are derived monthly from a combination of the English 2011 census, Office for National Statistics mid-year estimates, and an annual random probability survey conducted for the National Readership Survey.
Causal pathways
The covariate adjustment set was determined by constructing a directed acyclic graph (DAG) (Supplementary Fig. 1). Based on the published literature, the DAG illustrates the hypothesized causal and mediated pathways between smoking and observed health and social factors as well as unobserved latent factors, not able to be captured here, and their relationship to voting.
For research question 1, we used a multinomial regression model to estimate the unadjusted association (presenting the 95% confidence interval [CI]) between smoking status (never smoker as reference category) and voting intentions on unweighted data. The Conservative party was selected as the reference category because it was the government party at the time of data collection. For research question 2, we used a multinomial model to estimate the associations adjusting for AUDIT scores (as a continuous variable) and sociodemographic characteristics (categorical variables) on unweighted data. Across all models presented below, Goodness-of-fit tests indicated the full model statistically significantly predicted the dependent variable better than the intercept-only model alone (Likelihood ratio < 0.001). Independence of observations and multicollinearity were evaluated with simple correlations among the independent variables.
Sensitivity analysis: To assess the extent to which associations with voting intention reflect associations with political orientation, we planned to repeat the models with newspaper readership replacing party voting intention. However, from a visual inspection of the frequency with which people reported newspaper readership, the individual categories of newspapers did not correspond with voting intention. As there is a larger number and readership of right-leaning and mixed papers, we decided that because some papers have historically switched their leaning and to reduce skewness of right leaning, only two papers of each leaning, based on their clear political affiliation [31] would be selected for analysis. Guardian and Mirror were coded as left-leaning, Independent and Metro coded as centrist, and the Daily Express and Daily Mail as right-leaning, with mixed readership indicating readership across leanings.