The Effects of Arkansas Master Settlement Spending on Disparities in Smoking

Abstract

Objectives. We assessed the effect of Master Settlement Agreement (MSA) spending on smoking disparities in Arkansas, which distinguished itself from other states by investing all of its MSA funds in health-related programs.

Methods. In 1996–2009 data from the Behavioral Risk Factor Surveillance System, we specified multivariate logistic models to examine gender and racial/ethnic disparities in smoking rates within Arkansas (a pre–post analysis) and between Arkansas and its 6 neighboring states.

Results. Before the MSA programs started in 2001, male Arkansans smoked more than did female Arkansans (P < .05). After the programs, smoking declined significantly among men (but not women), eliminating the gender disparity by 2009. Smoking among men in Arkansas also declined more than it did in neighboring states (P < .05). Hispanics showed a greater decline in smoking than did non-Hispanic Whites in Arkansas (but not in neighboring states). In 2001, Hispanic Arkansans smoked more than did non-Hispanic Whites (P < .05); by 2009, Hispanic Arkansans smoked significantly less than did non-Hispanic Whites (P < .05).

Conclusions. MSA-funded programs were more effective in some segments of the Arkansas population than in others. Policymakers should consider targeting future MSA tobacco control programs to populations most resistant to change.

The landmark 1998 Master Settlement Agreement (MSA) awarded 46 states $216 billion over 25 years from the 4 US tobacco companies (Brown & Williamson, Lorillard, Philip Morris, and RJ Reynolds). These funds were compensation for Medicaid expenses related to tobacco use and a penalty to the tobacco companies for deceptive practices.^1,2 Although the MSA funds could be used by each state at its own discretion, at the time of the settlement, many states declared their intent to use them to defray smoking-related Medicaid costs and to support tobacco control activities.^3,4

Since the MSA went into effect, however, states have used the settlement funds for an array of nontobacco purposes.^4–6 In fiscal year 2010, for example, states spent just 2.3% ($567.5 million) of MSA and tobacco tax funds on tobacco prevention and cessation programs,⁴ and in 2009, 31 states provided less than a quarter of the Centers for Disease Control and Prevention–recommended funding for tobacco control.⁴ At different times since the settlement, various states have not spent any MSA funds on tobacco control at all.^4,6,7 The impact of moving funds away from tobacco-related initiatives is reflected in a recent standstill in smoking rates, which had been declining. The nationwide adult smoking rate in 2008 (20.6%) was essentially unchanged from 2004 (20.9%).⁸

Arkansas is unique among the states in its commitment to invest all of its MSA funds in activities designed to improve the health of its residents. The Tobacco Settlement Proceeds Act, passed by Arkansans in a 2000 referendum, invested the state's MSA funds in health-related programs and created a commission to monitor and evaluate the performance of its MSA-funded programs.⁹ These funds support 7 broad programs⁹:

1. the University of Arkansas College of Public Health, which provides professional education, research, and services to Arkansans;

2. the Arkansas Bioscience Institute, which develops new tobacco-related medical and agricultural research initiatives;

3. the Delta Area Health Education Center, which provides clinical education to the Arkansas Delta, an underserved and disproportionately poor region of the state;

4. the Arkansas Aging Initiative, which provides health education for seniors and professionals;

5. the Minority Health Initiative, which identifies and addresses the special health needs of minority communities;

6. Medicaid Expansion Programs, which supplement standard Medicaid benefits through targeted packages; and

7. the Tobacco Prevention and Cessation Program, which aims to reduce the initiation of tobacco use and resulting negative health and economic effects.

Since the settlement proceeds law was enacted, Arkansas has substantially increased its tobacco control spending. In fiscal year 2010, Arkansas was ranked ninth in tobacco control spending among the 46 states, 5 US territories, and the District of Columbia,⁴ a remarkable improvement over its status in fiscal year 2000, when it was one of 9 states that spent nothing on tobacco control.⁴ Increases in Arkansas's tobacco control spending have been associated with significant reductions in the state's overall smoking rate.^4,9 This is consistent with evidence from other states indicating that MSA funds can reduce smoking when they are spent on tobacco control.^4,10,11

Research on the effects of MSA tobacco control spending, however, has largely focused on smoking rates overall, and studies of the effectiveness of MSA-funded activities are typically conducted at the state level. State-level reports may obscure differences in the effectiveness of policies or intervention between groups. Consequently, the effects of MSA spending on gender or race disparities in smoking are largely unknown.^11,12 Epidemiological research suggests that certain demographic groups—such as men,^13,14 some ethnic minorities,¹⁵ persons with low educational attainment,⁸ and persons with mental illness¹⁶—are more likely to smoke and that some subgroups, such as women^17,18 and African Americans,¹⁹ have more difficulty quitting smoking once they start. Ideally, MSA spending would reduce disparities in smoking rates by decreasing the rates for all groups and making greater reductions among subgroups with higher smoking rates. It is not clear from the literature how MSA spending affects smoking disparities.

We aimed to illustrate how Arkansas's MSA spending affects disparities in smoking among different demographic groups and to identify populations that should be targeted by future MSA tobacco control spending. We examined the effectiveness of Arkansas's MSA-supported programs on smoking rates in women (vs men) and ethnic minorities (vs non-Hispanic Whites). To our knowledge, ours was the first study to examine how MSA spending affects disparities in smoking outcomes.

We explored whether smoking rates changed among gender and racial/ethnic groups in Arkansas before and after implementation of MSA-funded smoking cessation programs. Our primary hypothesis was that these programs would reduce disparities in smoking. We expected that Arkansas's MSA programs would reduce disparities in smoking because other large-scale tobacco control efforts have reduced such disparities. For example, increased cigarette taxation has been shown to reduce disparities in smoking between high- and low-income adults.¹⁹

We also compared disparities in smoking rates between Arkansas and the 6 states with which it shares a border. Our hypothesis for this analysis was that after implementation of the MSA programs, Arkansas would show a greater reduction in smoking disparities than its neighboring states.

METHODS

We conducted a pre–post analysis to estimate changes in smoking rates separately for gender and racial/ethnic groups in Arkansas before and after implementation of the state's MSA programs. Arkansas's MSA-funded programs incorporate broad policies designed to reach all Arkansans as well as some specific interventions focusing on tobacco use in underserved populations.

Our geographic analysis identified disparities in smoking rates between Arkansas and its 6 neighboring states. We treated Arkansas as the intervention group and the other states as the comparison group. We chose Arkansas's 6 neighboring states (Louisiana, Mississippi, Missouri, Oklahoma, Tennessee, and Texas) as the comparison group because they are geographically close to Arkansas and culturally similar to it.²⁰ Unlike all 6 of its neighbors, Arkansas spends at least 50% of the amount recommended by the Centers for Disease Control and Prevention on tobacco prevention programs (including MSA funds).⁴

Our 2 analytic approaches involved different samples. The pre–post analysis focused on a retrospective cohort in Arkansas only, and the comparison of geographic groups examined a larger retrospective cohort in Arkansas and its 6 neighboring states.

Data and Sample

We used data collected by the Behavioral Risk Factor Surveillance System (BRFSS), an ongoing telephone health survey that tracks health conditions and risk behaviors in the United States. The BRFSS used a disproportionate stratified sample design in 2009 to draw a probability sample of all households with telephones in each state.²¹ More than 350 000 adults are interviewed each year, making the BRFSS the largest telephone health survey in the world. Because the BRFSS collects state-specific data on risk behaviors, previous studies have used the data to study smoking rates at the state level.^22,23

Our study period (1996–2009) covered 6 years before the Arkansas tobacco settlement programs (1996–2001) and 8 years after the programs (2002–2009). The study sample consisted of 475 075 adults aged 18 years and older (Arkansas, n = 53 638; 6 neighboring states, n = 421 437). More detail about the sample is shown in Table A (available as a supplement to the online version of this article at http://www.ajph.org).

Measures

Status and disparities.

The BRFSS included 2 questions for determining cigarette smoking status: “Have you smoked at least 100 cigarettes in your entire life?” and “Do you smoke cigarettes now?” Current smokers were individuals who answered yes to both questions.^22,23 We used the term “disparities” to refer to between-group differences in smoking status.^24,25

Demographics.

We defined gender as a categorical variable with 2 categories: men and women. The BRFSS collected self-reported information about race/ethnicity. It first categorized the self-identified groups as White, Black, Asian, Native Hawaiian or other Pacific Islander, American Indian and Alaska Native, and others. It also asked if the respondent considered him- or herself as Hispanic. We used this information to define race/ethnicity as a categorical variable with 4 categories: non-Hispanic White, non-Hispanic Black, non-Hispanic other, and Hispanic. Because the groups of non-Hispanic others were too small (4.0% and 4.3% of the sample in Arkansas and its neighboring states, respectively) to provide interpretable results, we did not include them in the analyses of racial/ethnic disparities in smoking.

Socioeconomic status.

We controlled for the following socioeconomic status variables in all analyses: marital status (married; divorced, widowed, or separated; never married; and member of an unmarried couple), educational attainment (less than high school, high school graduate, some college or trade school, and college graduate or higher), employment status (employed, self-employed, no job for < 12 months, no job for ≥ 12 months, homemaker, student, and retired), annual household income (< $15 000, $15 000–$24 999, $25 000–$34 999, $35 000–$49 999, $50 000–$74 999, and ≥ $75 000), number of adults living in the household (1, 2, 3, and ≥ 4), and number of children younger than 18 years living in the household (0, 1, 2, and ≥ 3).

Analyses

We used Stata version 11 (StataCorp LP, College Station, TX) to perform all statistical analyses. We accounted for the complex sampling design of the BRFSS by using the survey design variables of weight, stratus, and sampling unit. We estimated 2 logistic models; the first focused on gender disparities in smoking rates, and the second focused on racial/ethnic disparities. The 2 models controlled for the same socioeconomic status variables but used different trend variables. For each model, we conducted F tests at α = 0.05 to determine (1) whether smoking rates among demographic groups within Arkansas had changed significantly since the tobacco settlement programs began and (2) whether trends in smoking rates in Arkansas were significantly different from trends in its bordering states.

To measure trends in smoking prevalence over time, we first constructed 3 variables: a linear year variable, a linear spline with a knot at 2001, and a second spline with a knot at 2006. We used 2001 as the cutoff because MSA-funded programs started in Arkansas in 2001.²⁶ Because it usually takes some years for statewide tobacco control programs to be fully implemented,²⁷ we designated the first 5 years of the MSA-funded programs (2002–2006) as the initial stage. In choosing this period we considered Arkansas's program implementation processes and similar processes in other states,^26,27 and we used 2006 as the cutoff to construct a second spline to represent the period when we expected the programs to be fully functioning.

We then assessed the interaction between these trend variables and demographic variables associated with disparities in smoking rates. For example, for our model on the difference in smoking rates between men and women, we constructed 12 variables by crossing the 3 trend variables with 4 dummy variables (men, women, Arkansas, and neighboring states). We incorporated these 12 interaction variables and the Arkansas dummy variable into our multivariate model, allowing us to compare different cessation trends for men and women within Arkansas and to compare different trends between Arkansas and its neighboring states. We similarly constructed and incorporated interaction variables and dummy variables in our analysis of racial disparities in smoking rates.

RESULTS

The pre–post analysis revealed a significant reduction in the difference between smoking rates of men and women (Figure 1). Male Arkansans smoked at a significantly higher rate than did female Arkansans, both in 2001 when the tobacco settlement programs began (25.5% vs 23.4%; P < .05) and in 2006 when the programs were fully operational (24.2% vs 20.9%; P < .05). By the end of the study period (2009), however, male and female Arkansans smoked at almost the same rate because there was a steep decline in smoking among men (20.4% in 2009 vs 24.2% in 2006; P < .05) and little change in smoking among women (21.8% in 2009 vs 20.9% in 2006; P > .05). The downward trend in male Arkansans' smoking rates starting in 2006 was consistent with the assumption that MSA-funded programs were fully operational by that year.

FIGURE 1.

Adjusted smoking rates for women and men: Behavioral Risk Factor Surveillance System, Arkansas, 1996–2009.

Figure 2a illustrates smoking rates among men in Arkansas and its neighboring states. Our comparative analysis suggested that the decline in smoking among male Arkansans after the full implementation of MSA-funded programs in 2006 was unique to Arkansas. Before the start (2001) and full operation (2006) of MSA-funded programming, we found no significant differences in smoking prevalence among men in Arkansas and its neighboring states. However, by 2009 male Arkansans were significantly less likely to smoke than were their neighbors (20.4% vs 24.8%; P < .05), suggesting a positive impact of MSA-funded programs on smoking among men. By comparison, smoking prevalence among women was not significantly different in Arkansas and its neighboring states during the study period, suggesting that Arkansas's MSA-funded programs had no significant impact on smoking among women (Figure 2b).

FIGURE 2.

Adjusted smoking rates for (a) men and (b) women: Behavioral Risk Factor Surveillance System, Arkansas and its 6 neighboring states, 1996–2009.

Racial/Ethnic Disparities

Our comparison of the effects of the Arkansas tobacco settlement programs on smoking rates among racial/ethnic groups used non-Hispanic Whites as the reference group. As shown in Figure 3, non-Hispanic White Arkansans had a slight downward trend in smoking prevalence following the start of tobacco settlement programming in 2001 and a significantly lower smoking rate in 2009 than in 2001 (19.9% vs 23.8%; P < .05).

FIGURE 3.

Adjusted adult smoking rates among racial/ethnic groups: Behavioral Risk Factor Surveillance System, Arkansas, 1996–2009.

Note. The study sample contained too few other non-Hispanics (4.0% in Arkansas and 4.3% in its neighboring states) to provide interpretable results.

Comparisons of non-Hispanic Whites to other groups of Arkansans showed disparities in smoking prevalence over the study period. At the beginning of the MSA-funded programs, non-Hispanic Black Arkansans were less likely than were non-Hispanic Whites to smoke (15.1% vs 23.8%; P < .05). However, non-Hispanic Blacks had an upward trend in smoking, particularly once tobacco settlement programs were fully operational, and by 2009, rates of smoking between non-Hispanic Whites and non-Hispanic Blacks were no longer significantly different.

Finally, in 2001 Hispanic Arkansans had a significantly higher smoking rate than did non-Hispanic Whites (26.1% vs 23.8%; P < .05). During program implementation, however, Hispanics more dramatically reduced their smoking rate than did non-Hispanic Whites. Consequently, by 2009 Hispanic Arkansans were significantly less likely than were non-Hispanic Whites to smoke (14.7% vs 19.9%; P < .05).

Geographic Differences

We also investigated racial/ethnic differences in smoking rates between Arkansas and its neighboring states. We observed similar downward trends in non-Hispanic Whites in Arkansas and in the bordering states (Figure 4a) and no significant differences in smoking prevalence in 2001, 2006, or 2009.

FIGURE 4.

Adjusted adult smoking rates among (a) non-Hispanic Whites, (b) non-Hispanic Blacks, and (c) Hispanics: Behavioral Risk Factor Surveillance System, Arkansas and its 6 neighboring states, 1996–2009.

Although the data revealed different trends in smoking rates between non-Hispanic Blacks in Arkansas and its neighbors, we found no significant geographic differences in smoking prevalence throughout the study period, suggesting that rates of smoking among non-Hispanic Black Arkansans may not have been affected by the tobacco settlement programs (Figure 4b).

Finally, Hispanic Arkansans had a greater reduction in smoking rates during the study than did their counterparts in the bordering states (Figure 4c). Although Hispanic Arkansans smoked more than did Hispanics in neighboring states in 2001 (26.1% vs 20.0%; P < .05), we observed no significant geographic differences between these 2 groups after the MSA programs became fully operational, either in 2006 or in 2009. This suggests that MSA programs may have been associated with reduced smoking among Hispanic Arkansans.

DISCUSSION

We used a large, longitudinal, state-level representative sample to estimate trends in smoking prevalence before and after Arkansas initiated smoking cessation programs funded by the tobacco settlement. We also controlled for confounding variables through a comparison group design and by including time-varying covariates associated with smoking prevalence. Overall, we found that Arkansas's tobacco settlement spending was associated with different trends in smoking across sociodemographic groups and with changes in smoking disparities.

Male and female Arkansans had different trends in smoking after the initiation of tobacco settlement programming. During the study period, the prevalence of men smoking decreased in Arkansas, and the rate among women was unchanged. Consequently, the gender disparity in smoking rates was eliminated. Our comparison of Arkansas with its 6 bordering states showed that a significant reduction in men smoking occurred only in Arkansas, suggesting that tobacco settlement programs may have reduced smoking among men but not among women.

The greater reduction in smoking among Arkansas men than among men in neighboring states may be attributable to the fact that neighboring states had a significantly higher proportion of non-Hispanic Blacks (12% vs 8%), a group whose smoking increased during the study period. This trend could reflect the low levels of spending on tobacco control in Arkansas's bordering states. Indeed, Arkansas may continue to see greater declines in smoking rates among certain groups because of state investment in prevention and cessation activities. Whereas Arkansas spends at least 50% of the amount recommended by the Centers for Disease Control and Prevention for effective tobacco control, Mississippi and Oklahoma spent less than 49%, Louisiana spent 24%, and Missouri, Tennessee, and Texas spent only 10%.⁴

The finding that Arkansas's tobacco settlement spending reduced the gender disparity in smoking rates by reducing the prevalence of smoking among men is consistent with previous research. Studies of American adults,¹³ adults in other Western countries,¹⁴ and adolescents²⁸ show that overall, rates of smoking among men are declining and are beginning to resemble rates of smoking among women, thus eliminating the historical gender disparity in smoking rates.

Our analyses also suggested that racial/ethnic groups were differentially affected by Arkansas's tobacco settlement spending. Hispanic Arkansans had smoking rates comparable to those of non-Hispanic White Arkansans in 2001, but by the end of the study period, Hispanics were significantly less likely than Whites to smoke. This downward trend among Hispanic Arkansans was unique to Arkansas among the study states and could therefore be related to tobacco settlement programming. On the other hand, our results are consistent with population survey results comparing trends in smoking across ethnic groups, which show Hispanics experiencing a greater decline in smoking.¹² Further research is needed to identify factors associated with downward trends in smoking among Hispanic Americans.

Two disturbing trends we observed were increases in smoking prevalence among women and non-Hispanic Blacks. These trends could be attributable to iatrogenic effects of MSA-funded programs, such as inadvertent glamorization of smoking for these groups. It is more likely, however, that external factors that became influential around 2006 affected smoking rates in women and non-Hispanic Blacks in ways that were not mitigated by the MSA programs. For instance, tobacco companies may have implemented advertising campaigns that targeted these groups because of previous research showing lower smoking rates among these groups than among other groups (men, non-Hispanic Whites). Another possibility is that migration patterns and other stresses associated with Hurricane Katrina resulted in unprecedented changes to Arkansas's female and Hispanic populations around this time. Further research is needed to identify factors that influence smoking in groups that remain relatively unaffected by traditional tobacco control activities.

Limitations

Our measure of smoking status was derived from self-reported survey data, which many studies have found to be consistent with biochemically verified smoking rates,²⁹ although some have not.³⁰ We attempted to control for confounding variables, but our analyses were correlational, and we could not assume causality.

The BRFSS did not trace participation in tobacco settlement programs; thus, we could not determine whether changes in smoking were attributable to participation in the programs or to some other factor. For instance, we found that the groups with the highest smoking rates showed decreases in smoking and that groups with lower rates tended to experience little change or an increase in smoking rates. Although these changes might have been related to MSA-funded programming, they might also have been the result of smoking patterns regressing toward the mean. Finally, the number of non-Hispanic others in the Arkansas BRFSS data was small, making analyses of this group unreliable.

Conclusions

This is one of the first studies to examine disparities in smoking outcomes related to MSA spending. Although disparities in smoking are well established in epidemiological and treatment literatures, very few studies have investigated the effects of tobacco control spending on disparities. Our study begins to fill this gap. Because MSA funds constitute the vast majority of tobacco control spending in Arkansas, we are taking important steps toward understanding how tobacco control funds affect different groups of smokers and learning which groups need more targeted interventions to help them quit.

Our findings have important policy implications. Researchers and policymakers should monitor the effects of MSA programs on sociodemographic groups and then consider targeting programs to groups whose rates of smoking remain high. Our analyses indicated that populations with relatively high smoking rates before the programs started, such as men and Hispanics, may benefit the most from MSA-funded programs, and that groups with relatively low smoking rates, such as women and non-Hispanic Whites, are likely to experience little or no program-related change. Although these results are consistent with the economic theory of diminishing marginal return³¹ (i.e., groups with low rates might not be easy to influence), they also suggest that Arkansas MSA funds were appropriately spent on moving smoking rates among groups with relatively high rates down toward the rates among groups with relatively low rates.

Our analyses also revealed that, after 8 years of implementing the Arkansas tobacco settlement programs, smoking cessation outcomes were still unimpressive among women, non-Hispanic Blacks, and other non-Hispanic persons. Policymakers may therefore want to target these groups in future MSA spending.

Policymakers can also learn from trends among groups with the largest reductions in smoking rate. Determining the specific programs that led to these reductions will also be of interest to other states that spend MSA funds on tobacco control. Similarly, it is not clear why non-Hispanic Black Arkansans had an upward trend in smoking rates after implementation of the MSA programs, so research on this population could help inform future tobacco control spending.