* Correspondence
Cameron M. Kaplan, PhD, Gehr Family Center for Health Systems Science and Innovation, Keck School of Medicine, University of Southern California, 2020 Zonal Avenue, IRD #327, Los Angeles, CA 90033, USA.
Email: ude.csu@cnalpak ,
The Affordable Care Act allows insurers to charge up to 50% higher premiums to tobacco users, making tobacco use the only behavioral factor that can be used to rate premiums in the nongroup insurance market. Some states have set more restrictive limits on rating for tobacco use, and several states have outlawed tobacco premium surcharges altogether. We examined the impact of state level tobacco surcharge policy on health insurance enrollment decisions among smokers.
We compared insurance enrollment in states that did and did not allow tobacco surcharges, using a difference‐in‐difference approach to compare the policy effects among smokers and nonsmokers. We also used geographic variation in tobacco surcharges to examine how the size of the surcharge affects insurance coverage, again comparing smokers to nonsmokers.
We linked data from two components of the Current Population Survey—the 2015 and 2019 Annual Social and Economic Supplement and the Tobacco Use Supplement, which we combined with data on marketplace plan premiums. We also collected qualitative data from a survey of smokers who did not have insurance through an employer or public program.
Allowing a tobacco surcharge reduced insurance enrollment among smokers by 4.0 percentage points (P = .01). Further, smokers without insurance through an employer or public program were 9.0 percentage points less likely (P < .01) to enroll in a nongroup plan if they were subject to a tobacco surcharge. In states with surcharges, enrollment among smokers was 3.4 percentage points lower (P < .01) for every 10 percentage point increase in the tobacco surcharge.
Tobacco use is the largest cause of preventable illness in the United States. State tobacco surcharge policy may have a substantial impact on whether tobacco users choose to remain insured and consequently their ability to receive care critical for preventing and treating tobacco‐related disease.
Keywords: affordable care act, enrollment, health insurance, marketplace, premiums, smoking, tobacco surcharge
In states that allow tobacco surcharges, tobacco users pay substantially higher premiums for nongroup health insurance than nonusers.
Smokers living in states with high tobacco surcharges are less likely to have health insurance.Allowing higher premiums for tobacco use in the nongroup market results in lower health insurance enrollment among smokers, which is driven by decreased enrollment in the nongroup market.
Conditional on allowing a surcharge, higher surcharges result in lower enrollment among smokers in both any insurance plan and in nongroup insurance plans in particular.
High premiums and tobacco surcharges are reported by smokers to be barriers to enrollment in marketplace coverage.
One of the most important aspects of the Affordable Care Act (ACA) was the introduction of modified community rating in the nongroup insurance market, which drastically changed how premiums could be set. Since 2014, all nongroup health insurance plans, including those sold on the individual marketplaces, can only rate premiums based on four characteristics: age, family size, geographic region, and tobacco use. Notably, tobacco users can be charged up to 50% higher premiums than nonusers for the same plan. These tobacco surcharges can be substantial, especially for older tobacco users, and they have increased over time. 1 , 2
Proponents of tobacco surcharges argue that higher premiums for tobacco users could lead to lower premiums for nonusers, increase insurer participation, and create an incentive for tobacco users to quit. 3 On the other hand, tobacco surcharges may discourage insurance enrollment among tobacco users. Furthermore, since tobacco use is self‐reported and not verified, tobacco users may lie about tobacco use to avoid the penalty. Access to insurance and health care may be particularly important for tobacco users because they are at higher risk of developing numerous chronic health conditions and because all health insurance plans are required to provide tobacco cessation as an “essential health benefit” with no out‐of‐pocket costs.
Although federal regulations require that plans charge tobacco users no more than 50% higher premiums, several states have more strict regulations. Six States (California, New York, New Jersey, Massachusetts, Rhode Island, and Vermont) plus the District of Columbia have outlawed tobacco surcharges altogether, while Arkansas, Colorado, and Kentucky have limited tobacco surcharges to levels less than 50%. In addition, there is large variation in tobacco surcharges even within states that allow surcharges since most insurers do not charge the maximum allowable surcharge. 2 , 4 Since the ACA allows states to set their own policy regarding tobacco surcharges, it is one aspect of the ACA that may be more amenable to modification, especially during times when partisan divide in congress makes it difficult to amend federal regulations.
Tobacco surcharges and the state laws that limit them theoretically have varying effects on different segments of the health insurance market. Premium rating rules for nongroup insurance and small group plans (generally plans provided by employers with fewer than 50 employees) are similar. Plans in both market segments are allowed to charge enrollees different rates based on the same four factors. One key difference is that tobacco users in small group plans can avoid paying the surcharge by enrolling in a tobacco cessation program. Additionally, recent research has cast doubt on whether small employers complied with either the rating rules or the exemption for tobacco cessation, particularly in the early years of the ACA. 5 , 6 Although state tobacco rating rules typically do not apply to large employers, it is possible that there are spillover effects.
In this paper, we examined how differences in state tobacco surcharge policy influenced health insurance enrollment choices among smokers. In order to assess the impact of state policy on enrollment, we used a difference‐in‐difference specification, comparing enrollment among smokers and nonsmokers in states that did and did not allow tobacco surcharges. We hypothesized that there would be lower enrollment among smokers, relative to nonsmokers, in states that allowed tobacco surcharges compared to states that outlawed rating by tobacco use. Among states that allowed a tobacco surcharge, we also examined the effect of the surcharge size on enrollment, again comparing smokers to nonsmokers.
This paper contributes to a small existing literature on the impact of tobacco surcharge laws on health insurance enrollment. Notably, using data from the Behavioral Risk Factor Surveillance System, Freidman and colleagues found the probability of having insurance was reduced among smokers in states with high tobacco surcharges relative to states without surcharges. 7 Pesko et al found that marketplace enrollment among tobacco users grew at a substantially slower rate than enrollment among all other enrollees in the first two years of the marketplaces. 8 Our work extends and builds upon this prior work in the following important ways. First, we used data that link detailed health insurance information including plan type with tobacco use data, which allowed us to examine the impact of tobacco surcharges on enrollment decisions for each market segment, rather than just the likelihood of being insured in any health insurance plan. This allows us to account for whether the effect is concentrated in the nongroup market or whether there are spillover effects. Second, we separately estimated the effect of both the existence and the size of the tobacco surcharge, that is, whether allowing a tobacco surcharge influences insurance enrollment, and how the size of the surcharge implemented by insurers effects enrollment. Although Freidman et al grouped states into small surcharge, large surcharge, and no surcharge categories, they did not explicitly model each policy component. Our contribution is particularly important to state policy makers because they may want to understand the comparative effectiveness of banning tobacco rating versus limiting the size of potential surcharges. Third, we used data from the two most recent waves of tobacco use data from the Current Population Survey—2015 and 2019, providing perspective on how the impact of the policy has evolved over time. Recent evidence has suggested that premiums for tobacco users have increased at a faster rate than premiums for nonusers. 2 Additionally, the elimination of the individual mandate has led to a decrease in enrollment, 9 but it is unknown whether there are differential effects for smokers or those living in high surcharge states. Finally, we supplemented our main findings with qualitative data from a survey that provides more contexts for how tobacco surcharges may influence enrollment in nongroup plans.
Our main source of data was the Current Population Survey (CPS). The CPS is a nationally representative, longitudinal survey that follows households over the same four consecutive months for two consecutive years (ie, four months on, eight months off, four months on). We used two important supplements to the CPS. First, each March, participants respond to the Annual Social and Economic Supplement (CPS‐ASEC), which includes detailed questions about the source and type of health insurance coverage as well as exhaustive income and employment questions. Approximately every four years, the CPS includes a Tobacco Use Supplement (CPS‐TUS), which asks questions about smoking and tobacco use. The two most recent waves of the CPS‐TUS in 2014‐15 and 2018‐19 were conducted after the ACA rating rules went into effect. In these years, adult members of households included in the basic monthly CPS were surveyed in June of the first year (2014 and 2018) and January and May of the second year (2015 and 2019). In order to examine health insurance enrollment by tobacco use, we linked data from both CPS‐ASEC and CPS‐TUS. Due to the sampling design of CPS, all households from CPS‐ASEC can theoretically be linked to the CPS‐TUS in either January or May of 2015 or 2019.
Within households, we defined health insurance units (HIUs) which include an individual, their spouse if they are married, and all dependent children of either the reference individual or their spouse. Within each HIU, we estimated the family income that would be used to determine eligibility for marketplace subsidies or Medicaid enrollment by summing the simulated adjusted gross income (AGI) variable from the CPS across all members of the HIU. Consistent with the family income calculation for health insurance qualification, we excluded income under $12 200 for dependents. Although health insurance eligibility is actually determined by Modified Adjusted Gross Income (MAGI) rather than AGI, previous studies have found that CPS‐simulated AGI is actually closer to MAGI than AGI. 10 Finally, we calculated income as a percentage of the federal poverty level (FPL) using family income, the number of people in the HIU, and the appropriate FPL based on year and state of residence.
We measured individual and household characteristics to be used as covariates in our regression models from CPS‐ASEC. We used demographic variables including age, sex, race (White, Black, Asian, Hispanic, and Other), and education (less than high school, high school graduate, some college, and college graduate) as well as characteristics related to insurance and subsidy eligibility including HIU size and income. We also calculated state policy variables to adjust for differences that may be correlated with tobacco surcharge policy and health insurance enrollment including whether the state was a Medicaid expansion state and the use of a state‐based marketplace. Further, we included state tobacco policy variables which we calculated from the CDC State Tobacco Activities Tracking and Evaluation System and included state cigarette taxes per pack and a measure of clean indoor air policies that were coded as 1 if the state banned smoking in all indoor bars, restaurants, and private workplaces and 0 otherwise. Finally, all models adjusted for year of survey (2015 or 2019).
In order to focus on health insurance coverage among individuals most likely to be affected by tobacco surcharges, we made several sample exclusions. After calculating AGI, we excluded adults over 65 since they would likely be covered by Medicare. We also excluded children under 19 who did not participate in the CPS‐TUS and who qualify for Medicaid/CHIP at higher income thresholds than adults. Finally, in several specifications we excluded families with incomes less than 138% of the federal poverty level for the following two reasons. First, there is known over‐reporting of nongroup health insurance plan participation among Medicaid eligible respondents in survey data including the CPS. 11 , 12 , 13 Second, due to Medicaid expansions, eligibility for marketplace subsidies and Medicaid overlaps between 100% and 138% FPL in some states, but not others. At incomes over 138% FPL, insurance eligibility is substantively similar across states. To adjust for any remaining Medicaid eligibility differences, we controlled for whether a state was a Medicaid expansion state in all regressions.
Individuals are able to report multiple sources of health insurance in CPS‐ASEC. We assigned primary coverage to each respondent based on the following hierarchy: Medicare, employer, nongroup, Medicaid, other public, and uninsured. In the CPS‐TUS, smoking status was defined based on how frequently individuals report smoking. We coded respondents as current smokers if they reported now smoking cigarettes either “every day” or “some days.” This definition is slightly different from the definition used to apply the tobacco surcharge, which is “the use of tobacco products four or more times, on average, per week within the past six months.” We used the narrower definition, considering only cigarette smokers because cigarette smokers make up most of the population of every day tobacco users 14 and nonsmokers who use tobacco may be less likely to consider themselves to be regular tobacco users. 15 Our results were not sensitive to the use of the alternative definition of tobacco use including the use of other tobacco products.
Finally, we gathered data on tobacco surcharges from the CMS Health Insurance Exchange Public Use Files for 2015 and 2019. These data were only available for the 38 states that participated in the federal exchange. In order to aggregate surcharge data to the state level, we first computed the median surcharge in each marketplace rating area by calculating the ratio between the plan premiums offered to tobacco and nontobacco users for a 45 years old. We then took the mean of the median surcharge across all rating areas in the state weighted by the population of the rating area age 20‐64. County populations were pulled from the 2014 and 2018 American Community Survey annual county level estimates and aggregated up to the rating area level.
We used linear probability models with a difference‐in‐difference specification. Specifically, our model compares the insurance status of smokers to nonsmokers across states with and without tobacco surcharges. To adjust for other potential differences across states, we controlled for Medicaid expansion status, the use of a state‐based marketplace, tobacco taxes, and clean indoor air policies. All models used CPS‐ASEC weights, and standard errors are clustered at the state level. Our approach differs from the triple‐difference approach used by Friedman et al, which used pre‐ and post‐ACA implementation as a third difference. 7 Unfortunately, the pre‐post approach may not yield accurate estimates because it does not account for the fact that tobacco rating was allowed in the nongroup and small group markets prior to 2014. 16
We first examined the impact of tobacco surcharges on the likelihood of having any type of health insurance. However, since theoretically tobacco surcharges should primarily affect the decision to purchase nongroup insurance, our preferred specification excluded individuals who had insurance either through an employer or a public program such as Medicare, Medicaid, or Military insurance. The 2019 ASEC included variables that specifically identify marketplace coverage, so we examined whether the effect is driven by those with marketplace plans using the 2019 data only. As additional robustness checks, we also examined subsamples including individuals with incomes between 138% FPL and 400% FPL, who are most likely to qualify for subsidies and those older than 25 who no longer qualify for dependent coverage. Finally, we ran a model using a definition of tobacco use closer to the definition used for enrollment in the ACA, which includes the use of other noncigarette tobacco products.
We also considered a comprehensive model of health insurance selection, by utilizing a multinomial logit regression to examine the likelihood of being covered under various types of insurance—employer, nongroup, public, or uninsured. This allowed us to examine the impact of surcharges on each type of insurance coverage without worrying about endogenous sample selection that could cause bias in the linear probability models where the sample is limited to individuals with nongroup plans and those who are uninsured. While surcharges unambiguously lead to higher premiums for tobacco users in nongroup markets and thus ought to lead to lower enrollment, theoretically, surcharges could lead to either higher or lower enrollment in employer coverage. For example, tobacco surcharges could lead to less employer coverage as a result of higher premiums for tobacco users in the small group market or if large employers were more likely to charge surcharges in states that allowed them in small group and nongroup markets. On the other hand, tobacco surcharges could lead to higher enrollment in employer coverage if surcharges make small employers more likely to offer insurance or if surcharges in nongroup marketplaces discourage smokers from becoming self‐employed (ie, there is more job lock).
Finally, we examined the impact of the size of tobacco surcharges on enrollment using a linear probability model with a difference‐in‐difference specification to predict the likelihood of any insurance coverage as well as the probability of nongroup coverage among those without insurance through an employer or public program. We limited the sample to individuals in states with tobacco surcharges, comparing the impact of the size of the average tobacco surcharge on differences in insurance coverage between smokers and nonsmokers. Because tobacco surcharges are changing over time, this specification also allowed us to include state fixed effects. The inclusion of state fixed effects holds constant any state specific factors that might affect differential insurance enrollment choices between smokers and nonsmokers.
In order to provide context to our results on the reasons tobacco users did not enroll in health insurance, we also included separate data from a June 2019 online survey conducted using Qualtrics Research Panel. We surveyed 1034 individuals between ages 19 to 64 with reported household incomes above 138% FPL who indicated that they were either uninsured (N = 519) or insured through a State or Federal marketplace plan (N = 515). Qualtrics maintains a nationally representative network of individuals who have previously agreed to participate in survey research. Over 150 000 emails were sent out to individuals who resided in the United States and were thought to be under 76 years old. 17 117 individuals opened the survey, and 1034 qualified for and completed the survey. For the purposes of this study, we limited the sample to uninsured individuals (N = 519). Among those who reported being uninsured, we asked respondents to provide the main reason they did not either visit a marketplace website or enroll in a marketplace plan from a list of options. We then asked respondents to select all reasons they did not enroll in a marketplace plan. The University of Southern California Institutional Review Board approved the survey data collection and deemed the analysis of data from the CPS as exempt from full review, and the relevant part of the survey questionnaire is included in Appendix S2.
Our final sample used pooled data from CPS‐ASEC in 2015 and 2019 and included 106 711 nonelderly adults. Table 1 compares sample characteristics across insurance type. Compared to those with insurance, uninsured individuals tended to be younger, have smaller family sizes, lower levels of family income and education, and they are more likely to be male and nonwhite. Notably, smoking rates were higher among those without health insurance than those with nongroup coverage. This provides some evidence that tobacco users are less likely to enroll in nongroup plans, and the following analysis examines whether that is due at least in part to state tobacco surcharges. Table TableA1: A1 : Appendix S1 presents summary statistics by insurance type separately for the 2015 and 2019 samples.
Comparison of demographic characteristics by primary insurance type for nonelderly adults
| Uninsured | Nongroup | Employer | Public | All | |
|---|---|---|---|---|---|
| N = 11 563 | N = 8572 | N = 74 690 | N = 11 886 | N = 106 711 | |
| Age (mean [SD]) | 39.4 [12.0] | 44.5 [12.8] | 43.0 [12.0] | 41.0 [13.1] | 42.5 [12.7] |
| Female (%) | 47.7 | 52.7 | 51.6 | 58.5 | 52.1 |
| White (%) | 45.7 | 68.4 | 70.0 | 51.3 | 65.1 |
| Black (%) | 10.9 | 7.7 | 8.9 | 13.7 | 9.5 |
| Hispanic (%) | 37.1 | 15.1 | 12.3 | 24.5 | 16.6 |
| Asian (%) | 4.1 | 7.0 | 2.6 | 4.5 | 5.8 |
| Other (%) | 2.3 | 1.8 | 2.5 | 6.1 | 2.9 |
| Less than HS (%) | 22.9 | 7.7 | 4.0 | 16.2 | 7.7 |
| High school (%) | 35.5 | 26.0 | 22.4 | 34.0 | 25.4 |
| Some college (%) | 26.5 | 29.7 | 28.9 | 32.4 | 29.1 |
| College degree (%) | 15.2 | 36.6 | 44.7 | 17.4 | 37.8 |
| Health status—“Excellent” or “Good” (%) | 58.1 | 68.3 | 71.0 | 47.5 | 66.8 |
| Current smoker (%) | 20.1 | 10.5 | 10.0 | 19.0 | 12.1 |
| HIU Size (mean [SD]) | 2.3 [1.5] | 2.4 [1.4] | 2.6 [1.4] | 2.7 [1.5] | 2.5 [1.4] |
| HIU AGI (mean [SD]) | $42 166 [$57 454] | $77 007 [$106 712] | $105 008 [$107 367] | $41 333 [$63 934] | $88 857 [$102 361] |
| HIU under 138% FPL (%) | 37.8 | 22.5 | 7.6 | 51.6 | 17.0 |
| HIU over 400% FPL (%) | 13.6 | 33.4 | 54.9 | 13.2 | 44.1 |
Descriptive statistics from linked CPS‐ASEC and CPS‐TUS data pooled across 2015 and 2019. Public insurance includes Medicare, Medicaid, VA, CHAMPUS, Tricare, and Indian Health Service.
Abbreviations: AGI, Adjusted Gross Income; FPL, Federal Poverty Level; HIU, Health Insurance Unit; SD, standard deviation.
The results from the linear probability models are shown in Table 2 . Column 1 shows the effect of tobacco surcharges on the differences in insurance rates between smokers and nonsmokers for the full sample of nonelderly adults. Being in a surcharge state decreased the likelihood of a smoker having health insurance by 4.0 percentage points (P = .02). Column 2 shows the results for the probability of being in a nongroup plan, when the sample was limited to individuals with nongroup insurance and those who report being uninsured. We found that among individuals without insurance through an employer or public program, living in a surcharge state decreased the probability that a smoker enrolled in a nongroup plan by 0.9.0 percentage points (P < .01). Table TableA2: A2 : Appendix S1 presents results from these regressions separately for the 2015 and 2019 samples. Table TableA3: A3 : Appendix S1 presents results excluding individuals under 26 and classifying tobacco use as the use of any tobacco product rather than cigarettes only. Our results are robust to these alternative specifications.
Impact of allowing a state surcharge on enrollment in a health insurance plan—results from difference‐in‐difference linear probability model
| Outcome | (1) | (2) | (3) | (4) | (5) |
|---|---|---|---|---|---|
| Any Insurance | Nongroup Insurance | Nongroup Insurance | Marketplace Insurance | Nonmarketplace Insurance | |
| Surcharge state | −0.0128 * (0.00652) | −0.0104 (0.0181) | −0.00647 (0.0114) | 0.00926 (0.0335) | −0.0477 (0.0341) |
| Current smoker | −0.00681 (0.0148) | −0.0527 *** (0.0176) | −0.0250 (0.0225) | 0.0599 * (0.0325) | −0.0930 ** (0.0407) |
| Surcharge state × Current smoker | −0.0403 ** (0.0170) | −0.0900 *** (0.0227) | −0.133 *** (0.0275) | −0.115 *** (0.0370) | 0.000755 (0.0423) |
| Observations | 106 711 | 20 135 | 13 840 | 6163 | 6163 |
| R‐squared | 0.116 | 0.181 | 0.179 | 0.042 | 0.166 |
| Insurance Groups | Everyone | Nongroup and Uninsured | Nongroup and Uninsured | Nongroup and Uninsured | Nongroup and Uninsured |
|---|---|---|---|---|---|
| Income | All | All | >138% FPL | >138% FPL | >138% FPL |
| Year | 2015, 2019 | 2015, 2019 | 2015, 2019 | 2019 | 2019 |
This table presents estimates from a difference‐in‐difference linear probability model, examining the likelihood of the outcome of interest for smokers in surcharge states. Each column is a separate regression, and the interaction term between surcharge state and current smoker shows the key difference‐in‐difference coefficient. In addition to the coefficients shown, all regressions control for sex, age, age squared, race, family size, education, log family income, state Medicaid expansion status, state federal marketplace use, state cigarette taxes, clean indoor air laws, and year (2015 vs 2019). Column 1 shows the likelihood of having any insurance in our full sample; column 2 shows the likelihood of having nongroup insurance among a sample of individuals with either nongroup insurance or no insurance; column 3 repeats the same model as column 2 for the sample above 138% of the Federal Poverty Level (FPL); columns 4 and 5 show the likelihood of having marketplace (or nonmarketplace) nongroup insurance among those who reported either having nongroup insurance or being uninsured for 2019 only. All regressions are weighted using the appropriate weights from the Current Population Survey. Standard errors, clustered at the state level are shown in parentheses.
Relative risk ratios from multinomial logit regression on type of health insurance plan
| Insurance type | (1) | (2) | (3) |
|---|---|---|---|
| Nongroup | Employer | Public | |
| Surcharge state | 0.876 (0.0843) | 0.915 (0.0664) | 0.710 *** (0.0732) |
| Current smoker | 0.777 *** (0.0756) | 0.791 * (0.103) | 0.951 (0.165) |
| Surcharge state × Current smoker | 0.681 *** (0.0834) | 0.783 * (0.109) | 1.019 (0.187) |
| Observations | 106 711 | 106 711 | 106 711 |
This table presents estimates from a multinomial logit regression model, comparing the likelihood of being in one of four insurance categories—uninsured (reference group), nongroup, employer, or public. All columns are derived from a single multinomial logit regression, and each column presents the relative risk ratios (RRR), comparing the risk of being in the insurance type of interest and being uninsured. The interaction terms between surcharge state and current smoker show the key difference‐in‐difference results. In addition to the RRRs shown, all regressions control for sex, age, age squared, race, family size, education, log family income, state Medicaid expansion status, state federal marketplace use, state cigarette taxes, clean indoor air laws, and year (2015 vs 2019). All regressions are weighted using the appropriate weights from the Current Population Survey. Standard errors, clustered at the state level are shown in parentheses.
Columns 3, 4, and 5 show results limiting the sample to individuals with incomes above 138% FPL, which includes those who have incomes too high to qualify for Medicaid under the expansions, but who are eligible for subsidies in the marketplaces and are thus more likely to be affected by tobacco surcharges. Column 3 repeats the specification from column 2 and confirms that the relationship is stronger among this population. We found that living in a surcharge state decreased the probability of enrolling in a nongroup plan by 13.3 percentage points (P < .01) among these individuals who are most likely to benefit from the marketplaces—those with incomes over 138% FPL who do not have insurance through their employer or a public program.
Columns 4 and 5 show results when nongroup plan enrollment was split by whether the plan was offered through the ACA marketplaces for the 2019 sample. All plans offered on the marketplace adhere to the ACA tobacco surcharge restrictions; however, more than half of nongroup plans offered outside of the marketplaces are not ACA compliant. 17 Column 4 shows that smokers were 11.5 percentage points (P < .01) less likely to be enrolled in a marketplace plan if they lived in a state with a tobacco surcharge than if they lived in a state without a tobacco surcharge. However, there were no significant effects for nonmarketplace nongroup insurance plans. Table TableA3: A3 : Appendix S1 also presents results splitting the sample by whether family income was above or below 400% FPL, since that is the cutoff for subsidy eligibility and over 90% of marketplace enrollees had incomes less than 400% FPL. 18
Table 3 presents results from a multinomial logit regression, allowing for a more comprehensive picture of the impact of surcharges on insurance selection. The outcome variable represents four categories of insurance type—uninsured (omitted), nongroup, employer, and public. The table presents relative risk ratios from a difference‐in‐difference specification similar to the one presented in Table 2 and including all income levels. The results indicate that smokers were at an decreased risk of being in a nongroup plan or an employer plan relative to being uninsured. However, the interaction term between nonsurcharge state and current smoker was only statistically significant at P < .05 for nongroup insurance, which gives us confidence in the specifications presented in Table 2 that excluded individuals with insurance through an employer or public program. It is worth noting that the relative risk ratio for the interaction term was marginally significant for employer sponsored health insurance. This suggests that smokers living in nonsurcharge states might be more likely to have employer sponsored health insurance than smokers living in states that allow tobacco surcharges.
Results showing the impact of the size of tobacco surcharges are presented in Table 4 . By limiting the sample respondents in states with tobacco surcharges, we were able to estimate the effect of the size of the surcharge on insurance enrollment. Among states that allowed a tobacco surcharge, the statewide average of the median rating area surcharge ranged from 0% to 29.6% in 2015 and 7.0% to 32.3% in 2019. State fixed effects were included in columns (2) and (4) include state fixed effects; however, they did not substantially affect the results. Our estimates indicate that the probability that a smoker had insurance decreases by 3.4 percentage points (P < .01) relative to nonsmokers for a 10 percentage point increase in the size of the tobacco surcharge. Additionally, after limiting the sample to those most likely to benefit from the marketplace (individuals who did not have insurance through an employer or public program and who had incomes above 138% FPL), we found that the probability that a smoker was enrolled in a nongroup plan decreased by 8.6 percentage points (P = .02) relative to nonsmokers for every 10 percentage point increase in the size of the tobacco surcharge.
Impact of state surcharge size enrollment in a health insurance plan—results from difference‐in‐difference linear probability model
| Outcome | (1) | (2) | (3) | (4) |
|---|---|---|---|---|
| Any insurance | Any insurance | Nongroup insurance | Nongroup insurance | |
| Percent surcharge | 0.0310 (0.0470) | 0.124 (0.0999) | 0.00150 (0.167) | 0.0873 (0.267) |
| Current smoker | 0.00901 (0.0216) | 0.00878 (0.0212) | −0.00831 (0.0600) | −0.0100 (0.0624) |
| Percent surcharge × Current smoker | −0.340 *** (0.122) | −0.341 *** (0.120) | −0.861 ** (0.365) | −0.853 ** (0.382) |
| Observations | 72 369 | 72 369 | 9755 | 9755 |
| R‐squared | 0.130 | 0.132 | 0.179 | 0.187 |
| Sample | Everyone | Everyone | Nongroup and Uninsured | Nongroup and Uninsured |
|---|---|---|---|---|
| Income | All | All | >138% FPL | >138% FPL |
| State fixed effects | No | Yes | No | Yes |
This table presents estimates from a difference‐in‐difference linear probability model, examining the likelihood of the outcome of interest for smokers in surcharge states. The sample is limited only to states that allowed a tobacco surcharge and participated in the federal exchange. Each column is a separate regression, and the interaction term between the percent surcharge and current smoker shows the key difference‐in‐difference coefficient. In addition to the coefficients shown, all regressions control for sex, age, age squared, race, family size, education, log family income, self‐reported health status, state Medicaid expansion status, state cigarette taxes, clean indoor air laws, and year (2015 vs 2019). Column 1 shows the likelihood of having any insurance in our full sample; column 2 shows the likelihood of having nongroup insurance among a sample of individuals with either nongroup insurance or no insurance; columns 3 and 4 show the likelihood of having marketplace insurance among those reporting having nongroup insurance or being uninsured. All regressions are weighted using the appropriate weights from the Current Population Survey. Standard errors, clustered at the state level are shown in parentheses. FPL = Federal Poverty Level.
Figure 1 shows the results from our survey of uninsured tobacco users with incomes above 138% FPL. Participants overwhelmingly reported that cost of health insurance was a major reason for either not visiting their state's health insurance marketplace website or not purchasing a plan. 41% of respondents said this was the main reason they did not enroll, and 54% said it played a factor in their decision. Respondents also listed tobacco surcharges as a major factor. 18% said that having to pay more for premiums due to tobacco use was a reason that they did not enroll, and 7% said it was the main reason they did not enroll. Other important reasons for not enrolling included not knowing about the marketplaces and/or not knowing how to enroll (19% main reason, 29% a factor), believing they were not eligible (8% main reason, 16% a factor), coverage not meeting needs including provider networks or benefits (3% main reason, 14% a factor), not needing health insurance (8% main reason, 11% a factor), or missing the enrollment period (4% main reason, 8% a factor).
Reasons for not enrolling in marketplace coverage. Data comes from a June 2019 online survey of tobacco users using Qualtrics Research Panel collected and analyzed by the authors. The survey included tobacco users age 18‐64 with incomes above 138% FPL who reported being uninsured or insured through a marketplace plan. Among the uninsured, respondents were asked to provide the main reason why they did not enroll in a marketplace plan from a list of options and then asked to check all reasons they did not enroll[Color figure can be viewed at wileyonlinelibrary.com]
Overall, this study found that tobacco surcharges led to significantly lower enrollment in nongroup health insurance among smokers. Allowing tobacco surcharges at all and having larger tobacco surcharges were each associated with lower take‐up of insurance. More specifically, our results show that tobacco surcharges have a large negative impact on enrollment in nongroup plans sold on the ACA marketplaces among those without insurance through an employer or public program. This finding was robust across a variety of specifications. Additionally, increased enrollment in nongroup plans in states that did not allow surcharges was not offset by a decrease in the likelihood of having other types of insurance like employer provided insurance—if anything, employer insurance take‐up was higher in states without surcharges.
Evidence that tobacco surcharges lead to lower take‐up of marketplace plans is further bolstered by results from a separate survey of tobacco users. Among those who would be able to purchase health insurance through the marketplaces but chose to remain uninsured, the tobacco surcharge was reported as a major factor in the decision. Additionally, tobacco surcharges function to increase the cost of health insurance for smokers, which was the most commonly cited reason for not choosing to enroll in a nongroup plan.
Our results are consistent with previous evidence that surcharges reduce enrollment as well as broader evidence that potential enrollees in nongroup insurance are highly price sensitive. 7 , 8 , 19 , 20 , 21 Notably, our point estimates are in line with those reported by Friedman et al 7 We found that the likelihood of having insurance among smokers was 4.0 percentage points (P = .02) lower in surcharge states, while Friedman and coauthors found that enrollment was 4.3 percentage points lower in states with medium sized surcharges, compared to states without surcharges, although their estimate was not statistically significant. Our results also show that the impact was largest in the exact insurance segments that were exposed to tobacco surcharges, increasing confidence that the policy has an effect on enrollment. According to our estimates, the presence of a tobacco surcharge decreased the likelihood of enrollment in a nongroup plan by 9.0 percentage points (P < .01) among smokers without insurance through an employer or public program. Additionally, among those facing a tobacco surcharge, a 10 percentage point increase in the size of the tobacco surcharge decreased the likelihood of enrollment in nongroup insurance by 8.6 percentage points (P = .02).
One important limitation of our study is that although we compare enrollment rates by smoking status, smoking status is not experimentally varied across groups. Additionally, our identification strategy relies on the assumption that states that do not have tobacco surcharges do not have other unobservable characteristics that may differentially impact enrollment for tobacco users and nonusers that are not accounted for in our model. Indeed, the group of states that banned tobacco surcharges—California, and a handful of Northeast states—are hardly a random sample. However, our models adjusted for other state policies that may affect insurance enrollment and tobacco use. Additionally, the fact that the inclusion of state fixed effects in our model of the impact of surcharge size produced nearly identical results to the same model without state fixed effects gives us confidence that this identification assumption holds. Another potential data limitation is that we only had complete, linked smoking and health insurance information for 2015 and 2019, which we pooled in our main analysis. Although we did not have data from other years, the fact that our results were similar across the two years gives us confidence that pooling the data is a valid approach.
Tobacco use is the largest cause of preventable disease and death in the United States, 22 and quitting smoking can significantly improve health outcomes. 23 The ACA requires that all health insurance plans provide tobacco cessation and lung cancer screening with no out‐of‐pocket costs. As such, several previous studies have shown that gaining access to health insurance can have large effect on both quitting smoking and cancer screening. 24 , 25 , 26 Theoretically, surcharges could also lead to reduction in smoking by imposing a financial penalty for the behavior. Although evidence suggests that tobacco taxes and other policies that increase the cost of smoking have a direct impact on smoking, 27 there is yet little evidence that insurance surcharges lead to changes in tobacco use. 7 , 28 Thus, perhaps somewhat counterintuitively, eliminating or limiting surcharges could potentially lead to an increase in quitting.
Another potential consideration is the effect of tobacco surcharges on nonusers. Theoretically, tobacco surcharges could lead to lower premiums for nonusers, which could potentially lead to higher enrollment among that group. More research is needed to test whether tobacco surcharges decrease premiums for nonusers. Our models test the effect of the surcharge on enrollment and do not provide evidence of higher enrollment among nonusers.
Since tobacco surcharges can be set at the state level, this policy may be more easily altered than other aspects of the ACA which are subjected to federal regulation. Banning tobacco surcharges may be more effective than having a very low tobacco surcharge, since even asking about tobacco use during enrollment could have an effect on enrollment independent of the size of the surcharge. However, if banning surcharges is not feasible, policy makers might consider limiting surcharges to below the 50% federal cap. Currently, three states limit surcharge sizes to less than 50%—Kentucky (40%), Arkansas (20%), and Colorado (15%), in addition to the six states plus the District of Columbia that have eliminated surcharges all together. These limits may not be binding in all cases, as many plans charge less than the maximum allowed, and there is substantial geographic variation in the amount of the sucharge. 2 , 4 Nonetheless, this study shows that lower surcharges were associated with higher enrollment in nongroup plans. Eliminating tobacco surcharges or substantially limiting them is an option that state policy makers may consider in order to increase health insurance coverage among this particularly vulnerable population.
