Where the U.S. tobacco epidemic still rages:Most remaining smokers have lower socioeconomic status
Objective. We estimated the proportion of U.S. smokers who have low socioeconomic status (SES). Methods. We used 2012 data from a national supplement to The Attitudes and Behaviors Survey on Health (TABS), a periodic population survey of Colorado adults. We estimated smoking prevalence and total smokers by education, poverty level, occupation, health insurance status, and combinations of these factors. Results. Smoking prevalence across low-SES categories ranged from 24.3% to 42.6%. Combining low-SES categories with the highest smoking prevalence accounted for 31.1% of U.S. adults but half (50.1%) of smokers. Combining all low-SES categories regardless of smoking prevalence accounted for roughly half (53.3%) of adults but nearly three-fourths (72.2%) of smokers. Conclusions. A majority of continuing U.S. smokers have low SES. Further progress against the U.S. cigarette epidemic depends on focusing tobacco research and program initiatives on reaching and engaging these smokers in cessation strategies that work for them.
Smoking, socioeconomic status, social justice, public health
During the past half-century, U.S. smoking prevalence has steadily declined while socioeconomic disparities in smoking have steadily widened. From 1965 to 1991, the proportion of current smokers went from 42.4% of adults to 25.7% of adults—a 39.4% reduction—but among those without a high school diploma, the reduction was less than half as great as among those with at least four years of college (24.7% vs. 60.6% relative decline).1
Smoking prevalence is now highly disparate across the social, economic, and occupational dimensions of socioeconomic status (SES).2,3,4 In 2012, current smokers numbered fewer than one in ten (9.1%) adults with a college degree, but one in four (24.7%) adults without a high school diploma,5 and one in six (17.0%) adults living at or above poverty, but more than one in four (27.9%) living in poverty.5 By occupation, pooled smoking prevalence during 1987-2004 was 39.7% among construction workers and miners, for example, compared with 11.3% among teachers, librarians, and counselors.6
Elevated smoking rates among lower SES (LSES) categories are not rooted in lack of motivation to quit: High school dropouts report levels of desire to quit and attempts to quit that are similar to those among college graduates.7-14 Rather, the disparities persist [End Page 100] and are widening because LSES smokers who try to quit are less likely than other smokers to attain and maintain abstinence.8,15-18 Suspected causes include underuse of evidence-based cessation treatment,17,19-21 and chronic exposure to psychosocial stressors without adequate social support and effective, health-supporting coping resources.22,23,24
A population with above average failure rates in efforts to attain a health objective deserves public health attention under the moral imperative of social justice to secure a sufficient level of health for all and to narrow unjust inequalities.25 However, public health ethics also require balancing commitment to social justice against “the injunction to maximize good aggregate or collective health outcomes.”26[p. 2] Indeed, public health impact depends less on serving unjustly burdened groups than on reaching the greatest number of people with effective interventions.27 Social justice and greatest-good mandates compete for resources, except in cases where a specific population simultaneously bears an unjustly elevated health burden and comprises a majority of individuals who bear the burden. The current study examines the relationship of elevated smoking burdens with the size of socioeconomic populations that currently smoke.
To our knowledge, little if any research has focused on the socioeconomic composition of the U.S. smoker population. We analyzed data from a one-time national survey to estimate the proportion of remaining smokers who have low socioeconomic status (LSES).
The Attitudes and Behaviors Survey (TABS) on Health is a periodic, population-level study among Colorado adults (aged 18+). In 2012, TABS supplemented the state sample (n = 14,998) with a large, representative national sample of adults (n = 3,230) in order to compare state and national health attitudes and behaviors. The Attitudes and Behaviors Survey methods are reported in detail elsewhere.17 Briefly, both landline and cell phone telephone exchange banks were stratified and randomly sampled, with oversampling in exchanges with higher concentrations of Latino and African American households. Sampled households were enumerated, and up to two residents aged 18+ were selected to complete a computer-assisted telephone interview (CATI) in the respondent’s choice of English or Spanish. The instrument, protocol, and participant consent were approved by the Colorado Multiple Institutional Review Board (COMIRB). Survey topics in 2012 included four chronic conditions—diabetes, hypertension, hyperlipidemia, and overweight/obesity—and tobacco-related attitudes and behaviors. Response rates28 were 55.2% nationally (58.4% in Colorado) at the landline household level and 51.7% nationally (55.8% in Colorado) at the landline respondent level, and 19.6% of sampled cell phones (24.7% in Colorado). Weights were constructed to account for individual selection probability and non-response, and to adjust the sample to match the U.S. population on sex, age, race/ethnicity, and education level.
For the current analyses, four measures were used to indicate SES: household income as a percentage of the federal poverty level (FPL; <100%, 100%-199%, 200%+); occupational status (employed for wages or salary, disabled or unable to work, unemployed, all other [homemaker, self-employed, student, retired]); education level (≤8 years, 9-12 [End Page 101] years without a high school diploma, GED, high school diploma, some college or post-high school, college degree, postgraduate degree), and health insurance (private, Medicaid, Medicare, none). Item-missing values in each SES variable were multiply imputed (10 imputations, 100 iterations) using chained equations (mi impute chained; StataCorp. 2013. Stata: Release 13. College Station, TX) conditioned on demographic factors (age, sex, ethnicity, primary language); self-reported health factors (general health status, mental health diagnosis or limitation, physical limitation, hypertension, hyperlipidemia, diabetes, overweight/obesity); health care access and utilization (past-year medical and dental visits; past-year financial inability to obtain needed medical, dental, pharmaceutical, or mental health treatment); smoking status, and Internet access. Fifteen variables were used in imputation, and 7.5% of item-values were imputed.
Low socioeconomic status (LSES) was defined as FPL <200%, disability, less education than a high school or general education development (GED) diploma, or Medic-aid or no health insurance. The proportion of smokers with LSES was estimated two ways. One strategy focused on the strongest claim for social justice treatment, i.e., smokers in LSES categories that have the highest smoking prevalence: partial high school education, FPL <100%, disability/inability to work, and Medicaid insurance; this estimate includes only a portion of LSES smokers but represents the groups where smoking is the most widespread and densely concentrated. The other estimation strategy included all LSES categories (LSES-inclusive). A third variable was constructed to represent low-income (<200% FPL) employed adults, i.e., the working poor and working near-poor.
Data were weighted to represent the U.S. adult population, and analyses estimated descriptive parameters (current smoking prevalence, total number of smokers, percentage of all smokers, and percentage of all adults) with 95% confidence intervals (CIs) for each SES variable and each LSES construct. All analyses used design-based methods to yield approximately unbiased parameter estimates and robust estimates of variance.
An estimated 19.7% (CI, 17.6-21.9%) or 48.4 million (CI, 42.5 million to 54.2 million) adults were current smokers in 2012 (Table 1). The highest smoking prevalence by income, education, insurance status and occupational status ranged from 29.7% to 42.6%, with Medicaid beneficiaries and disabled adults exceeding 40% prevalence. Combining low-SES categories with the highest smoking prevalence accounted for 31.1% of U.S. adults but half (50.1%) of smokers. Combining all low-SES categories regardless of smoking prevalence accounted for roughly half (53.3%) of adults but nearly three-fourths (72.2%) of smokers. Low-income workers accounted for nearly one-fourth (23.5%) of smokers but less than one-seventh (14.1%) of adults.
A majority of U.S. adult smokers lives in poverty, has less than high school education, is disabled, and/or is on Medicaid, and these socioeconomically defined populations have the highest rates of smoking. When the near-poor and the uninsured are also included, [End Page 102]
Click for larger view
View full resolution
[End Page 103]
nearly three-fourths of continuing U.S. smokers have low socioeconomic status. These new findings indicate that ethical principles that often conflict—social justice for the disadvantaged vs. greatest good for the greatest number—are perfectly aligned for U.S. tobacco control. A priority shift to LSES populations would be not only ethical but practical, since even complete eradication of smoking among other populations would leave nearly three-fourths of the continuing epidemic untouched. Efforts should include prevention of smoking initiation among LSES adolescents, but the emphasis should be on existing adult smokers, both because they outnumber new adult smokers roughly 100 to one, and because adult smoking models the behavior for adolescents.
Study limitations include the cross-sectional design, which supports only associative rather than directional analysis of the relationship between SES and smoking. The self-reported data may underrepresent smoking, especially among higher socioeconomic categories where antismoking norms predominate, in which case the findings may be overstated. At the same time, low SES respondents are underrepresented in the (unweighted) study sample; weights used in analyses adjusted for sample-to-population discrepancies in education but not other SES dimensions. If LSES smokers were also less likely than LSES nonsmokers to complete the survey, the findings would underestimate the extent of LSES among continuing smokers. These concerns are somewhat mitigated by the overlap of current smoking prevalence estimates with those from the National Health Interview Survey for the same year.29 Nevertheless, further research is needed to address the study limitations and validate the findings.
Further progress against the U.S. cigarette epidemic depends on focusing a majority of tobacco research and program initiatives to reach, engage, and support smokers with lower socioeconomic status in their cessation attempts, treatment uptake and adherence, and adjustment to life without cigarettes. Little smoking cessation intervention research has focused on challenges specific to lower SES, such as stressful environments and chronically stressful lives, which seem highly likely to undermine LSES smokers in quit-attempts. Many smokers use cigarettes as a moment-to-moment coping strategy, and LSES life brings more than its share of momentary needs for coping and resilience. Yet, little is known about the feasibility, efficacy, or sustainability of interventions designed to help low SES quit-attempters develop and strengthen coping skills and resources for managing nicotine withdrawal and sustaining abstinence after cigarettes are gone but living remains stressful. The challenges of LSES smoking cessation must become targets for concerted research and programmatic initiatives.
ARNOLD LEVINSON is associated with the University of Colorado Cancer Center and the Department of Community & Behavioral Health, Colorado School of Public Health. Arnold Levinson can be reached at 13001 East 17th Place, Mail Stop F542, Aurora, Colorado 80045, arnold.levinson@ucdenver. edu, Phone: 303-724-3541, Fax: 303-724-3544.
Data were collected under a grant (#13FLA45130) from the Colorado Department of Public Health and Environment. The author appreciates very helpful feedback on a previous version from Erik Augustson, and reviewer and editor suggestions that led to an improved manuscript.
The author has no competing interests. [End Page 104]