Nicotine & Tobacco Research Advance Access published May 8, 2015
Smoking and (Not) Voting: The Negative Relationship Between a Health-Risk Behavior and Political Participation in Colorado Author list (in order):
an
us
a. Department of Community & Behavioral Health, Colorado School of Public Health, University of Colorado Anschutz Medical Campus, Aurora, Colorado, 80045, USA b. Colorado Health Outcomes Program, University of Colorado School of Medicine, Aurora, Colorado, 80045, USA c. Community Properties of Ohio (CPO) Management Services, 910 East Broad Street, Columbus, Ohio 43205, USA d. University of Colorado Cancer Center, Mail Stop F542, Aurora, Colorado, 80045, USA
Corresponding author:
pt e
d
M
Dr. Karen Albright University of Colorado, Anschutz Medical Campus Department of Community and Behavioral Health Mail Stop F443, 13199 E. Montview Blvd., Suite 300, Aurora, CO 80045 Phone: 303-724-3535 Fax: 303-724-1839 [email protected]
ce
Abstract word count: 248
Ac
Manuscript word count: 2143
Keywords:
smoking; voting; distrust; political engagement
Published by Oxford University Press on behalf of the Society for Research on Nicotine and Tobacco 2015. This work is written by (a) US Government employee(s) and is in the public domain in the US.
1
Downloaded from http://ntr.oxfordjournals.org/ at Universite Laval on September 22, 2015
cr ip t
Karen Albright, PhDa,b, Nancy Hood, PhDc, Ming Ma, MD MPH,a Arnold H. Levinson, PhDa,d
Smoking and (Not) Voting: The Negative Relationship Between a Health-Risk Behavior and Political Participation in Colorado ABSTRACT
cr ip t
Introduction: Considerable evidence suggests that cigarette smokers are an increasingly marginalized population, involved in fewer organizations and activities and with less interpersonal trust than their
nonsmoking counterparts. However, only two previous studies, both among Swedish populations, have investigated smokers’ attitudes toward political systems and institutions. The current, cross-sectional study
us
least partly reflects trust in formal political institutions.
an
Methods: Secondary analyses were conducted of interview data from 11,626 respondents in the Colorado Tobacco Attitudes and Behaviors Survey. Data were collected via telephone between October 2005 and
M
mid-April 2006 and included respondents’ reported voting behavior in the 2004 national election; the participation rate was 89.7%. Balanced multiple logistic regression was used to examine associations
pt e
variables.
d
between smoking and voting while controlling for other covariates known to be associated with both
Results: In the final model, daily smokers were less than half as likely as nonsmokers to report having
ce
voted in the election.
Conclusions: The results suggest possible consonance with previous work linking smoking with political
Ac
mistrust. Possible causal mechanisms are discussed. This study is the first to link a health-risk behavior with electoral participation, and provides initial evidence that smoking is negatively associated with political participation. Future research should investigate how public health might enhance tobacco control efforts by taking nonvoting behavior into consideration, or creatively combining smoking cessation interventions with voter registration and other civic engagement work, particularly among socioeconomically disadvantaged populations. 2
Downloaded from http://ntr.oxfordjournals.org/ at Universite Laval on September 22, 2015
examines smoking in relation to voting, a direct behavioral measure of civic and political engagement that at
Smoking and (Not) Voting: The Negative Relationship Between a Health-Risk Behavior and Political Participation in Colorado INTRODUCTION
cr ip t
Cigarette smokers are increasingly understood to be a marginalized population.
Numerous studies across multiple countries have found that smokers are involved in fewer
organizations and activities and have lower interpersonal trust than do nonsmokers or former
us
smoking was predicted by low participation in study circles, organizations, churches, and
an
cultural activities such as attending the theater/cinema and art exhibitions; in contrast, participation in study circles and church attendance predicted smoking cessation. Similarly, in an
M
earlier Swedish study of socioeconomic status (SES) and smoking,2 social participation predicted maintenance of abstinence from smoking, suggesting that successful quitting may be related to
d
differing social network resources and social capital between socioeconomic groups. “[Social
pt e
participation’s] influence on the socioeconomic differences in smoking cessation could reflect the marginalization of lower socioeconomic groups.”2 In a two-year British study, Giordano and Lindström found that interpersonal trust and
ce
social participation within the community predicted smoking cessation.3 In Beirut, married women who smoked reported lower trust in neighbors and neighboring businesses.4 A similar
Ac
relationship was found in England5 as well as in Australia, where Siahpush and colleagues6
found smoking associated with living in a community exhibiting less perceived trust and safety, more perceived income inequality, and lower perceived relative material well-being. Most recently, two U.S. studies reported similar associations. Using data drawn from the 2005 U.S. Health Information Trends Survey, Finney Rutten and colleagues7 found that current smokers 3
Downloaded from http://ntr.oxfordjournals.org/ at Universite Laval on September 22, 2015
smokers. In a Swedish study of smoking cessation behavior across one year,1 continuing
were less likely than nonsmokers to belong to community organizations, attend religious services, or trust information resources. Among rural U.S. adolescents, higher levels of social capital (defined as community cohesion, social control, and youths’ relationships with adults in
cr ip t
the community) were associated with lower rates of smoking among youth from low-income families after controlling for a range of individual, household, and community characteristics.8
Beyond this literature, few studies have investigated smokers’ attitudes toward systems
us
Two intriguing studies recently examined the relationship between political mistrust and
an
smoking, finding a negative association between smoking and system distrust. Lindström found that low political trust in government, independent of generalized trust in other people, was significantly associated with daily smoking and lower rates of smoking cessation among
M
Swedes.9 In a similar study population, Lindström and Janzon found that trust in the Swedish healthcare system, though not in the mass media, was significantly associated with lower odds of
d
daily smoking and higher odds of ever-smokers having quit.10
pt e
The current study extends these findings by examining whether smokers' distrustful attitudes toward political institutions translate to voting behaviors. Voting is a direct measure of
ce
civic and political engagement,11-13 and not to vote “is to withdraw from the political community.”12 Voting partly reflects trust in formal political institutions14 and may be
Ac
considered a measure of social participation. Though an extensive political science literature15-22 has previously investigated relationships between voting behavior and numerous other factors, including race and ethnicity, religion, gender, marital status, age, ideological orientation, and party identification, the current study is, to our knowledge, the first to examine the relationship between voting behavior and smoking.
4
Downloaded from http://ntr.oxfordjournals.org/ at Universite Laval on September 22, 2015
and institutions that undergird the society from which smokers are increasingly marginalized.
METHODS Study Population The current study analyzed data from a statewide tobacco survey. Data collection
cr ip t
methods are described in detail elsewhere;23 in brief, the Colorado Tobacco Attitudes and Behaviors Survey (TABS) uses random digit dialing to conduct computer-assisted telephone
interviews of adults (aged 18+) in either English or Spanish. Smokers and African American
us
(household) sampling. Constructed weights support inference to the Colorado adult population in
an
the survey year. The current analysis used data from the second wave, collected from October 2005 through mid-April 2006, when 12,257 adults completed interviews and were asked about
M
their voting behavior in the 2004 national election. The participation rate in the survey was 89.7%. Measures
d
The primary outcome of interest was response to the question, “In the election for
pt e
President and Congress last November, did things come up that kept you from voting, or did you happen to vote?” Since Colorado held no general election in November 2005, respondents who
ce
answered after that month were assumed to be reporting their voting behavior in the November 2004 national election. The primary independent variable was smoking status (lifetime
Ac
consumption of ≥100 cigarettes and current daily or some-day smoking). The following potential covariates were chosen a priori based on previously reported association with smoking or with voting:24 sex, age, race/ethnicity, education, employment status, marital status, household income relative to the federal poverty level, and self-reported general health status. Two demographic groups were omitted from analysis because they were unlikely to have been eligible to vote in
5
Downloaded from http://ntr.oxfordjournals.org/ at Universite Laval on September 22, 2015
respondents are oversampled. Respondents are selected through two-stage stratified cluster
2004: Spanish-dominant respondents, defined as those who spoke only Spanish at home and/or completed the survey in Spanish, and respondents aged 18 years old when interviewed (combined n=416, 3.4%). Cases with missing voting or smoking status were dropped (combined
cr ip t
n=215, 1.8% of remaining cases), and item-missing values in remaining observations were imputed using multiple imputation (PROC MI, SAS v. 9.4). The sample for analysis included 11,626 respondents, of whom 2,565 were current smokers.
us
Smoking and voting rates were separately described by covariate factors using chi-square
an
tests adjusted for complex survey sampling. Bivariate associations with voting were tested using survey logistic regression; sex was not significantly associated with smoking and was omitted from further analysis. A multivariate model was built with voting as the outcome, smoking as the
M
independent variable, and significant covariates, using purposeful selection for covariate entry.25 To address potential bias from unbalanced observational data, a final model was built
d
using inverse probability treatment (IPT) weighting to adjust the distribution of each measured
pt e
baseline covariate independently of smoking status.26 We constructed IPT weights by fitting a multinomial logistic model and predicted each individual’s probability (propensity score, PS) of
ce
belonging to a smoking category, conditioned on the baseline covariates and survey weights. All covariates related to both voting and smoking or only voting were included in the PS model.
Ac
Survey weights were also included in the PS model, rather than implementing a survey-weighted PS model, to help satisfy the assumption of no unmeasured confounders.27 IPT weights were calculated as the inverse of corresponding PSs. We stabilized IPT weights by multiplying by a constant equal to the expected PS of being in one of three smoking groups (daily smoker, nondaily smoker, nonsmoker), and truncated stabilized IPT weights greater than 10 or less than
6
Downloaded from http://ntr.oxfordjournals.org/ at Universite Laval on September 22, 2015
Data Analysis
0.1.28 We used the standardized difference method to evaluate the balance of the distribution of each baseline covariate, comparing the frequency of multilevel categorical covariates among three smoking groups with
Smoking and (Not) Voting: The Negative Relationship Between a Health-Risk Behavior and Political Participation in Colorado Author list (in order):
an
us
a. Department of Community & Behavioral Health, Colorado School of Public Health, University of Colorado Anschutz Medical Campus, Aurora, Colorado, 80045, USA b. Colorado Health Outcomes Program, University of Colorado School of Medicine, Aurora, Colorado, 80045, USA c. Community Properties of Ohio (CPO) Management Services, 910 East Broad Street, Columbus, Ohio 43205, USA d. University of Colorado Cancer Center, Mail Stop F542, Aurora, Colorado, 80045, USA
Corresponding author:
pt e
d
M
Dr. Karen Albright University of Colorado, Anschutz Medical Campus Department of Community and Behavioral Health Mail Stop F443, 13199 E. Montview Blvd., Suite 300, Aurora, CO 80045 Phone: 303-724-3535 Fax: 303-724-1839 [email protected]
ce
Abstract word count: 248
Ac
Manuscript word count: 2143
Keywords:
smoking; voting; distrust; political engagement
Published by Oxford University Press on behalf of the Society for Research on Nicotine and Tobacco 2015. This work is written by (a) US Government employee(s) and is in the public domain in the US.
1
Downloaded from http://ntr.oxfordjournals.org/ at Universite Laval on September 22, 2015
cr ip t
Karen Albright, PhDa,b, Nancy Hood, PhDc, Ming Ma, MD MPH,a Arnold H. Levinson, PhDa,d
Smoking and (Not) Voting: The Negative Relationship Between a Health-Risk Behavior and Political Participation in Colorado ABSTRACT
cr ip t
Introduction: Considerable evidence suggests that cigarette smokers are an increasingly marginalized population, involved in fewer organizations and activities and with less interpersonal trust than their
nonsmoking counterparts. However, only two previous studies, both among Swedish populations, have investigated smokers’ attitudes toward political systems and institutions. The current, cross-sectional study
us
least partly reflects trust in formal political institutions.
an
Methods: Secondary analyses were conducted of interview data from 11,626 respondents in the Colorado Tobacco Attitudes and Behaviors Survey. Data were collected via telephone between October 2005 and
M
mid-April 2006 and included respondents’ reported voting behavior in the 2004 national election; the participation rate was 89.7%. Balanced multiple logistic regression was used to examine associations
pt e
variables.
d
between smoking and voting while controlling for other covariates known to be associated with both
Results: In the final model, daily smokers were less than half as likely as nonsmokers to report having
ce
voted in the election.
Conclusions: The results suggest possible consonance with previous work linking smoking with political
Ac
mistrust. Possible causal mechanisms are discussed. This study is the first to link a health-risk behavior with electoral participation, and provides initial evidence that smoking is negatively associated with political participation. Future research should investigate how public health might enhance tobacco control efforts by taking nonvoting behavior into consideration, or creatively combining smoking cessation interventions with voter registration and other civic engagement work, particularly among socioeconomically disadvantaged populations. 2
Downloaded from http://ntr.oxfordjournals.org/ at Universite Laval on September 22, 2015
examines smoking in relation to voting, a direct behavioral measure of civic and political engagement that at
Smoking and (Not) Voting: The Negative Relationship Between a Health-Risk Behavior and Political Participation in Colorado INTRODUCTION
cr ip t
Cigarette smokers are increasingly understood to be a marginalized population.
Numerous studies across multiple countries have found that smokers are involved in fewer
organizations and activities and have lower interpersonal trust than do nonsmokers or former
us
smoking was predicted by low participation in study circles, organizations, churches, and
an
cultural activities such as attending the theater/cinema and art exhibitions; in contrast, participation in study circles and church attendance predicted smoking cessation. Similarly, in an
M
earlier Swedish study of socioeconomic status (SES) and smoking,2 social participation predicted maintenance of abstinence from smoking, suggesting that successful quitting may be related to
d
differing social network resources and social capital between socioeconomic groups. “[Social
pt e
participation’s] influence on the socioeconomic differences in smoking cessation could reflect the marginalization of lower socioeconomic groups.”2 In a two-year British study, Giordano and Lindström found that interpersonal trust and
ce
social participation within the community predicted smoking cessation.3 In Beirut, married women who smoked reported lower trust in neighbors and neighboring businesses.4 A similar
Ac
relationship was found in England5 as well as in Australia, where Siahpush and colleagues6
found smoking associated with living in a community exhibiting less perceived trust and safety, more perceived income inequality, and lower perceived relative material well-being. Most recently, two U.S. studies reported similar associations. Using data drawn from the 2005 U.S. Health Information Trends Survey, Finney Rutten and colleagues7 found that current smokers 3
Downloaded from http://ntr.oxfordjournals.org/ at Universite Laval on September 22, 2015
smokers. In a Swedish study of smoking cessation behavior across one year,1 continuing
were less likely than nonsmokers to belong to community organizations, attend religious services, or trust information resources. Among rural U.S. adolescents, higher levels of social capital (defined as community cohesion, social control, and youths’ relationships with adults in
cr ip t
the community) were associated with lower rates of smoking among youth from low-income families after controlling for a range of individual, household, and community characteristics.8
Beyond this literature, few studies have investigated smokers’ attitudes toward systems
us
Two intriguing studies recently examined the relationship between political mistrust and
an
smoking, finding a negative association between smoking and system distrust. Lindström found that low political trust in government, independent of generalized trust in other people, was significantly associated with daily smoking and lower rates of smoking cessation among
M
Swedes.9 In a similar study population, Lindström and Janzon found that trust in the Swedish healthcare system, though not in the mass media, was significantly associated with lower odds of
d
daily smoking and higher odds of ever-smokers having quit.10
pt e
The current study extends these findings by examining whether smokers' distrustful attitudes toward political institutions translate to voting behaviors. Voting is a direct measure of
ce
civic and political engagement,11-13 and not to vote “is to withdraw from the political community.”12 Voting partly reflects trust in formal political institutions14 and may be
Ac
considered a measure of social participation. Though an extensive political science literature15-22 has previously investigated relationships between voting behavior and numerous other factors, including race and ethnicity, religion, gender, marital status, age, ideological orientation, and party identification, the current study is, to our knowledge, the first to examine the relationship between voting behavior and smoking.
4
Downloaded from http://ntr.oxfordjournals.org/ at Universite Laval on September 22, 2015
and institutions that undergird the society from which smokers are increasingly marginalized.
METHODS Study Population The current study analyzed data from a statewide tobacco survey. Data collection
cr ip t
methods are described in detail elsewhere;23 in brief, the Colorado Tobacco Attitudes and Behaviors Survey (TABS) uses random digit dialing to conduct computer-assisted telephone
interviews of adults (aged 18+) in either English or Spanish. Smokers and African American
us
(household) sampling. Constructed weights support inference to the Colorado adult population in
an
the survey year. The current analysis used data from the second wave, collected from October 2005 through mid-April 2006, when 12,257 adults completed interviews and were asked about
M
their voting behavior in the 2004 national election. The participation rate in the survey was 89.7%. Measures
d
The primary outcome of interest was response to the question, “In the election for
pt e
President and Congress last November, did things come up that kept you from voting, or did you happen to vote?” Since Colorado held no general election in November 2005, respondents who
ce
answered after that month were assumed to be reporting their voting behavior in the November 2004 national election. The primary independent variable was smoking status (lifetime
Ac
consumption of ≥100 cigarettes and current daily or some-day smoking). The following potential covariates were chosen a priori based on previously reported association with smoking or with voting:24 sex, age, race/ethnicity, education, employment status, marital status, household income relative to the federal poverty level, and self-reported general health status. Two demographic groups were omitted from analysis because they were unlikely to have been eligible to vote in
5
Downloaded from http://ntr.oxfordjournals.org/ at Universite Laval on September 22, 2015
respondents are oversampled. Respondents are selected through two-stage stratified cluster
2004: Spanish-dominant respondents, defined as those who spoke only Spanish at home and/or completed the survey in Spanish, and respondents aged 18 years old when interviewed (combined n=416, 3.4%). Cases with missing voting or smoking status were dropped (combined
cr ip t
n=215, 1.8% of remaining cases), and item-missing values in remaining observations were imputed using multiple imputation (PROC MI, SAS v. 9.4). The sample for analysis included 11,626 respondents, of whom 2,565 were current smokers.
us
Smoking and voting rates were separately described by covariate factors using chi-square
an
tests adjusted for complex survey sampling. Bivariate associations with voting were tested using survey logistic regression; sex was not significantly associated with smoking and was omitted from further analysis. A multivariate model was built with voting as the outcome, smoking as the
M
independent variable, and significant covariates, using purposeful selection for covariate entry.25 To address potential bias from unbalanced observational data, a final model was built
d
using inverse probability treatment (IPT) weighting to adjust the distribution of each measured
pt e
baseline covariate independently of smoking status.26 We constructed IPT weights by fitting a multinomial logistic model and predicted each individual’s probability (propensity score, PS) of
ce
belonging to a smoking category, conditioned on the baseline covariates and survey weights. All covariates related to both voting and smoking or only voting were included in the PS model.
Ac
Survey weights were also included in the PS model, rather than implementing a survey-weighted PS model, to help satisfy the assumption of no unmeasured confounders.27 IPT weights were calculated as the inverse of corresponding PSs. We stabilized IPT weights by multiplying by a constant equal to the expected PS of being in one of three smoking groups (daily smoker, nondaily smoker, nonsmoker), and truncated stabilized IPT weights greater than 10 or less than
6
Downloaded from http://ntr.oxfordjournals.org/ at Universite Laval on September 22, 2015
Data Analysis
0.1.28 We used the standardized difference method to evaluate the balance of the distribution of each baseline covariate, comparing the frequency of multilevel categorical covariates among three smoking groups with
