Health Policy 119 (2015) 44–49
Contents lists available at ScienceDirect
Health Policy journal homepage: www.elsevier.com/locate/healthpol
The smoking ban next door: Do hospitality businesses in border areas have reduced sales after a statewide smoke-free policy? Elizabeth G. Klein a,∗ , Nancy E. Hood b a Ohio State University College of Public Health, Health Behavior & Health Promotion, 352 Cunz Hall, 1841 Neil Avenue, Columbus, OH 43210, United States b CPO Management, 910 East Broad St, Columbus, OH 43205, United States
a r t i c l e
i n f o
Article history: Received 6 January 2014 Received in revised form 18 September 2014 Accepted 22 September 2014 Keywords: Secondhand smoke Tobacco Smoke-free policy
a b s t r a c t Introduction: Despite numerous studies demonstrating no significant economic effects on hospitality businesses following a statewide smoke-free (SF) policy, regional concerns suggest that areas near states without SF policies may experience a loss of hospitality sales across the border. The present study evaluated the impact of Ohio’s statewide SF policy on taxable restaurant and bar sales in border and non-border areas. Methods: Spline regression analysis was used to assess changes in monthly taxable sales at the county level in full-service restaurants and bars in Ohio. Data were analyzed from four years prior to policy implementation to three years post-policy. Change in the differences in the slope of taxable sales for border (n = 21) and non-border (n = 67) counties were evaluated for changes following the statewide SF policy enforcement, adjusted for unemployment rates, general trends in the hospitality sector, and seasonality. Results: After adjusting for covariates, there was no statistically significant change in the difference in slope for taxable sales for either restaurants (ˇ = 0.9, p = 0.09) or bars (ˇ = 0.2, p = 0.07) following the SF policy for border areas compared to non-border areas of Ohio. Conclusions: Border regions in Ohio did not experience a significant change in bar and restaurant sales compared to non-border areas following a statewide SF policy. Results support that Ohio’s statewide SF policy did not impact these two areas differently, and provide additional evidence for the continued use of SF policies to provide protection from exposure to secondhand smoke for both workers and the general public. © 2014 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Compared to other occupational groups, bar and restaurant workers (also called hospitality workers) have historically experienced the least protection from secondhand smoke (SHS) exposure in the workplace [1] as
∗ Corresponding author. Tel.: +1 614 292 5424; fax: +1 614 688 3533. E-mail addresses: [email protected], [email protected] (E.G. Klein), [email protected] (N.E. Hood). http://dx.doi.org/10.1016/j.healthpol.2014.09.011 0168-8510/© 2014 Elsevier Ireland Ltd. All rights reserved.
hospitality businesses were often exempted from policies that restricted smoking in workplaces. These exemptions from SF policies have become less common; as of July 2014, 39 states had local 100% smoke-free (SF) policies that applied to workplaces including both bars and restaurants, representing protection for 81.6% of the population in the United States (U.S.) [2]. Despite well-documented health benefits of SF policies [3], opponents – including the tobacco industry – continue to argue that SF policies that apply to bars and restaurants may reduce the number of customers in these
E.G. Klein, N.E. Hood / Health Policy 119 (2015) 44–49
establishments, thereby significantly reducing their revenue, employment opportunities, and likelihood of remaining in business [4]. Findings from numerous reviews of economic impact studies of SF policies have refuted these concerns [4–6]. In fact, Scollo et al. [6] found all studies documenting a negative economic impact of SF policies on the hospitality sector were sponsored by the tobacco industry. Furthermore, these studies were significantly more likely to use a subjective outcome measure and less likely to be peer-reviewed than studies showing no impact or a positive impact. Several studies that evaluated bars and restaurants separately reported no significant economic effects on either type of establishment [7–9] despite the known correlation between smoking and drinking behaviors [10]. Opponents have also raised concerns that statewide evaluations of SF policies may mask differential effects in some geographic areas such as border communities and rural regions, which may be important given that approximately 39% of smokers in the U.S. live within 40 miles of another state [11]. Economic studies have shown 4–25% of smokers cross borders to purchase cigarettes in jurisdictions with lower cigarette taxes [11,12]. Based on 2003 Current Population Survey Tobacco Use Supplement data, smokers traveled approximately three miles to save one dollar on a pack of cigarettes [11]. If smokers are willing to travel across borders to purchase cigarettes, they may also be willing to travel across borders to patronize bars and restaurants that permit smoking if those in their own jurisdictions do not. Only one study to date has examined the effects of a statewide SF policy on the hospitality sector in border communities [13]. These researchers evaluated whether Ohio’s statewide SF policy implemented in December 2006 differentially affected several hospitality sector employment indicators (including the total number of employees, total wages paid, and number of establishments) in selection of Ohio counties bordering Kentucky compared to nonborder counties; Kentucky did not have a statewide SF policy during the study period. Additionally, changes over time were compared between Ohio counties and both border and non-border counties in Kentucky. From three years before Ohio’s policy began to one year after implementation, there were no disproportionate changes in hospitality sector employment indicators between border and nonborder counties in Ohio or Kentucky. Thus, opponents’ arguments that a statewide SF policy would drive business across the border to jurisdictions without SF policies were not supported. Despite these findings, the issue of economic effects in border areas deserves continued study. Ohio offers a unique opportunity to empirically examine whether statewide policies negatively impact restaurant and bar establishments because until recently, all Ohio border counties (n = 88) were adjacent to five individual states without SF policies. Further, Ohio has suffered substantial set-backs in tobacco control including the 2008 state legislative dissolution of the Ohio Tobacco Prevention Foundation and diversion of Ohio’s master tobacco settlement funds to nontobacco-related budget items [14]. In fact, in 2011, Ohio did not contribute any state funds to tobacco control and
45
federal tobacco control funding equaled only 1.5% of the level recommended by the Centers for Disease Control and Prevention [15]. As the prevalence of adult smoking in Ohio remains one of the highest in the nation at 23.3% of adults reporting current smoking [16], (a slight decline from a high of 25.1% in 2011), this slower pace of decline compared to other states may represent a negative effect of diminished state tobacco control funding. All told, Ohio presents a unique natural quasi-experiment to evaluate the economic impacts of a statewide SF policy in an area where smoking remains relatively prevalent. This study provides three important extensions to previous work: (1) The use of taxable sales data as an objective and direct assessment of economic effects, (2) separate evaluations of bars and restaurant businesses to improve sensitivity to changes to specific business types, and (3) evaluation of the robust number of Ohio counties that share the border with five states. If Ohio residents responded to the SF law by traveling across state lines to restaurants and bars in states without comprehensive SF laws, as is suggested by the tobacco industry, sales in restaurant and bar taxable sales might be expected to suffer more in border counties, where individuals can more feasibly make such choices, than in more centralized counties. As a result, we hypothesized that border counties would experience a significant reduction in the rate (slope) of taxable sales compared to non-border counties for both business types. 2. Methods The study sample includes all 88 counties in the state of Ohio, dichotomized into those counties which share a border with another state and non-border counties. Border counties were defined as those sharing a border with one of five neighboring states (Indiana, Michigan, Pennsylvania, West Virginia, and Kentucky) (n = 26). One county shared less than 0.83 square miles with a neighboring state, including a single village of 525 residents, and was treated as non-border. None of the bordering states had 100% statewide SF laws covering restaurants and bars during the study period with the exception of Michigan, which implemented a law in the last month of the study period. However, five border counties were adjacent to non-Ohio counties with complete county-level smoke-free policies during at least some of the study period. Because border effects would not be expected in these areas, for the entire study period these counties were reclassified as nonborder. The final sample included a total of 21 border and 67 non-border counties (n = 88). Linear mixed model regression analyses were selected to evaluate for change in taxable sales within bars and restaurants before and after a SF policy was enacted for the state. Data were collected and provided for this analysis by the Ohio Department of Taxation (referred to here as the Department). The primary outcome variable was monthly taxable sales (in dollars). Hospitality businesses report sales tax liability every month. Very small businesses (defined as less than $1200 in state sales tax liability over a six month period) report every six months; businesses with biannual reporting of taxable sales were excluded from analyses
46
E.G. Klein, N.E. Hood / Health Policy 119 (2015) 44–49
because monthly data were not available. Sales tax liability was transformed into taxable sales by dividing the reported sales tax liability by the local sales tax rate (multiplied by 100) for the appropriate month and county. The local sales tax rate included county and transit taxes, where applicable. Monthly taxable sales were further classified by industry. Businesses that were most likely to be licensed to sell alcohol were selected using the North American Industry Classification System (NAICS) industry codes: full-service restaurants (NAICS code 7221) and “drinking places” (NAICS code 7224) referred to here as bars. Data were also obtained for the total hospitality sector, including fastfood and other limited service restaurants, catering, and accommodation businesses (NAICS code 72). For all taxable sales values, the unit of analysis available through the Department was county-level data. The Department suppressed data when there were fewer than five reporting businesses for a given county and month. The number of each type business (referred to as entities) by NAICS code was summed by county and month. Unadjusted unemployment rates were obtained from U.S. Bureau of Labor Statistics [17] in order to derive county-level unemployment statistics by month. The study period was June 2003 through May 2010 (n = 84 months). The statewide comprehensive SF policy was enacted on December 7, 2006 but not enforced until May 2007, yielding 37 months following policy enforcement. 2.1. Statistical analysis The goal of this research was to determine whether trends in county taxable sales changed following the passage of the statewide SF law, and whether those changes differed in border and non-border counties. To evaluate this question, we modeled monthly taxable sales as a function of time, using a spline to allow the time trend to change following the statewide SF policy enactment. We interacted both time trends with the indicator of border status to determine whether taxable sales time trends, both before and after the policy, differed in border and nonborder counties. Specifically, the following linear mixed models were estimated for each restaurant sales and bar sales. A linear mixed effects model with random intercepts and random slopes was used to examine whether taxable sales changed within border counties (i) over time (j) after enforcement of the Ohio Smoke-free Workplace Act. The selected model allows for the slope to change after the designated time (t*) when the state SF policy was enforced. E(Yij |i ) = ˇ0 + ˇ1 border + ˇ2 timeij + ˇ3 (timeij − t∗) + ˇ4 timeij ∗ borderij + ˇ5 (timeij − t∗) ∗ borderij + ˇ6 unemploymenti + ˇ7 entitiesi + ˇ8 overall sales for hospitalityi + ˇ9 springi + ˇ10 summeri + ˇ11 falli +, where t* equals May 2007 for the date of policy enforcement, and (timei − t*)+ is equal to (timei − t*)+ when
timei > t* and is equal to zero when timei ≤ t* [18]. Of most interest for this analysis are ˇ4 and ˇ5 ; an insignificant value for ˇ4 with a significant value for ˇ5 would suggest that county taxable sales time trends differed in border and non-border counties after, but not before, the policy passage, lending support to the possibility that the passage impacted the two areas differently. Covariates were included to account for other relevant economic factors that might influence the primary outcome: county-level unemployment, number of each type of business (entities), taxable sales for hospitality businesses excluding bars or restaurants, and seasons of the year (indicator variables for spring, summer, and fall; winter was used as the reference group). An exponential covariance structure was specified for the within-subject errors that depended on time. Models were run separately for fullservice restaurants and bars. An alpha level of 0.05 was used to identify statistical significance. Analyses were completed using SAS 9.3 (SAS Institute, Inc. Cary, NC).
3. Results A description of state and county-levels characteristics for the study sample at the study baseline in 2003 is shown in Table 1. For the state of Ohio, counties had an average of 88 restaurants and 21 bars, with a wide range for both types of hospitality businesses. Both border and non-border counties reported over $5 million in monthly taxable sales for restaurants, with approximately a half a million dollars monthly taxable sales for bars. There were no significant differences between border and nonborder counties in the number of entities, taxable sales, or unemployment (p > 0.05) as of June 2003 at the start of the present study. Figs. 1 and 2 depict the unadjusted monthly sums of taxable sales for border and non-border counties over the period of study, respectively for restaurants (Fig. 1) and bars (Fig. 2). In Table 2, the parameter estimates for the regression model are presented for restaurants and bars, respectively. The primary outcome variable is shown per $10,000 in taxable sales revenue, and all covariates (ˇ6 through ˇ11 ) were mean centered for ease of interpretation. For restaurants during the period of June 2003 to May 2007 before the Ohio SF policy was enacted, border counties had a difference in slopes of nearly $14,000 compared to nonborder counties (ˇ4 ; p = 0.03). Following the Ohio SF policy enactment, there was no statistically significant change in the difference in slopes for taxable sales in restaurants in border counties (ˇ5 ; p = 0.09) after adjustment for employment, overall sales and seasons of the year; in other words, the difference in the taxable sales rate for restaurants in border counties remained after the statewide SF policy. For bars, there was no significant differences in the slopes for border compared to non-border counties before policy enactment (ˇ4 ; p = 0.91) after adjustment for employment, overall sales and seasons of the year. Following the statewide policy enactment, the change in the different in sales revenue rate was not statistically significant comparing border and non-border counties (ˇ5 ; p = 0.07).
E.G. Klein, N.E. Hood / Health Policy 119 (2015) 44–49
47
Table 1 State and county-level descriptive characteristics of hospitality businesses in Ohio, 2003. Description
State of Ohio (n = 88) Mean (range)
Border (n = 21) Mean (range)
Non-border (n = 67) Mean (range)
Restaurants businesses Bars businesses Monthly taxable restaurant sales
40 (5–1036) 10 (5–188) $5,403,563 ($85,669–$71,316,301) $514,715 ($32,413–$7,856,139) 7.1% (4.6%–11.5%)
102 (8–625) 22 (5–92) $5,697,434 ($191,511–$52,896,386) $431,658 ($39,548–$2,059,480) 7.4% (4.9%–9.9%)
84 (5–1036) 21 (5–188) $5,311,455 ($85,669–$71,316,301) $544,616 ($32,413–$7,856,139) 7.0% (4.6%–11.5%)
Monthly taxable bar sales Unemployment rate Indicates p < 0.05.
$500 $450 $400
Millions
$350 $300
Non-border Border
Ohio Smokefree Workplace Act
$250 $200 $150 $100 $50 2003 2004
2005
2006
2007
2008
2009
2010
Fig. 1. Unadjusted total monthly restaurant taxable sales for border and non-border counties of Ohio, 2003–2010.
4. Discussion This study found that a statewide SF policy in Ohio did not have significant economic impacts on taxable sales for restaurants and bars in border counties that differed from non-border counties, which is consistent with the findings for the state as a whole [19]. These results are consistent with one study for a narrow sample of Ohio counties bordering the state of Kentucky that were evaluated within a year of policy enactment [13]. Another recent study found Ohio’s statewide policy was not associated with significant changes in another objective economic measure—employment indicators—in rural or urban counties [20]. Further, the separate analysis of bars and
restaurants provides robust, empirical evidence that Ohio’s border county bars did not suffer significant economic effects from the statewide SF policy, a common argument from opponents of statewide clean indoor air policies. All told, these findings present important evidence against arguments perpetuated by the tobacco industry and concerns among lay people that SF policies adversely impact the hospitality sector in some parts of the state. Because almost 40% of smokers in the U.S. live within relatively short driving distance of another state (i.e., 40 miles), findings from this study are important considerations for the rest of the country. Evidence that statewide SF policies do not differentially impact hospitality sector business near state borders even when adjacent to areas with more
$50 $45 $40
Millions
$35 $30 $25
Non-border Border
Ohio Smokefree Workplace Act
$20 $15 $10 $5 2003 2004
2005
2006
2007
2008
2009
2010
Fig. 2. Unadjusted total monthly bar taxable sales for border and non-border counties of Ohio, 2003–2010.
48
E.G. Klein, N.E. Hood / Health Policy 119 (2015) 44–49
Table 2 Association of the passage of a statewide smoke-free policy with trends in restaurant and bar taxable sales, in both border and non-border counties of Ohio, 2003–2010. Variables
Intercept (ˇ0 ) Border county (ˇ1 ) Monthly time trend before smoke-free law (ˇ2 ) Monthly time trend after smoke-free law (ˇ3 ) Difference in taxable sales rate before policy enactment in border counties (ˇ4 ) Change in the difference in taxable sales rate after policy enactment in border counties (ˇ5 ) County-level unemployment rate (ˇ6 ) # of restaurants entities (ˇ7 ) Overall hospitality sales (ˇ8 ) Spring (ˇ9 ) Summer (ˇ10 ) Fall (ˇ11 )
Restaurants
Bars
Regression estimates
Standard error (df)
Regression estimates
Standard error (df)
128.2 126.3 1.8
44.2 (86) 50.3 (86) 0.56 (7293)
0.005 0.014 0.001
12.8 12.2 0.3
8.15 (81) 10.52 (81) 0.16 (5941)
0.166 0.249 0.085
−1.9
0.57 (7293)
0.001
−0.3
0.10 (5941)
0.002
−1.4
0.64 (7293)
0.031
−0.02
0.19 (5941)
0.910
0.9
0.54 (7293)
0.089
0.2
0.1 (5941)
0.069
0.6
1.25 (7293)
0.616
0.4
0.23 (5941)
0.104
5.0
0.12 (7293)
Contents lists available at ScienceDirect
Health Policy journal homepage: www.elsevier.com/locate/healthpol
The smoking ban next door: Do hospitality businesses in border areas have reduced sales after a statewide smoke-free policy? Elizabeth G. Klein a,∗ , Nancy E. Hood b a Ohio State University College of Public Health, Health Behavior & Health Promotion, 352 Cunz Hall, 1841 Neil Avenue, Columbus, OH 43210, United States b CPO Management, 910 East Broad St, Columbus, OH 43205, United States
a r t i c l e
i n f o
Article history: Received 6 January 2014 Received in revised form 18 September 2014 Accepted 22 September 2014 Keywords: Secondhand smoke Tobacco Smoke-free policy
a b s t r a c t Introduction: Despite numerous studies demonstrating no significant economic effects on hospitality businesses following a statewide smoke-free (SF) policy, regional concerns suggest that areas near states without SF policies may experience a loss of hospitality sales across the border. The present study evaluated the impact of Ohio’s statewide SF policy on taxable restaurant and bar sales in border and non-border areas. Methods: Spline regression analysis was used to assess changes in monthly taxable sales at the county level in full-service restaurants and bars in Ohio. Data were analyzed from four years prior to policy implementation to three years post-policy. Change in the differences in the slope of taxable sales for border (n = 21) and non-border (n = 67) counties were evaluated for changes following the statewide SF policy enforcement, adjusted for unemployment rates, general trends in the hospitality sector, and seasonality. Results: After adjusting for covariates, there was no statistically significant change in the difference in slope for taxable sales for either restaurants (ˇ = 0.9, p = 0.09) or bars (ˇ = 0.2, p = 0.07) following the SF policy for border areas compared to non-border areas of Ohio. Conclusions: Border regions in Ohio did not experience a significant change in bar and restaurant sales compared to non-border areas following a statewide SF policy. Results support that Ohio’s statewide SF policy did not impact these two areas differently, and provide additional evidence for the continued use of SF policies to provide protection from exposure to secondhand smoke for both workers and the general public. © 2014 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Compared to other occupational groups, bar and restaurant workers (also called hospitality workers) have historically experienced the least protection from secondhand smoke (SHS) exposure in the workplace [1] as
∗ Corresponding author. Tel.: +1 614 292 5424; fax: +1 614 688 3533. E-mail addresses: [email protected], [email protected] (E.G. Klein), [email protected] (N.E. Hood). http://dx.doi.org/10.1016/j.healthpol.2014.09.011 0168-8510/© 2014 Elsevier Ireland Ltd. All rights reserved.
hospitality businesses were often exempted from policies that restricted smoking in workplaces. These exemptions from SF policies have become less common; as of July 2014, 39 states had local 100% smoke-free (SF) policies that applied to workplaces including both bars and restaurants, representing protection for 81.6% of the population in the United States (U.S.) [2]. Despite well-documented health benefits of SF policies [3], opponents – including the tobacco industry – continue to argue that SF policies that apply to bars and restaurants may reduce the number of customers in these
E.G. Klein, N.E. Hood / Health Policy 119 (2015) 44–49
establishments, thereby significantly reducing their revenue, employment opportunities, and likelihood of remaining in business [4]. Findings from numerous reviews of economic impact studies of SF policies have refuted these concerns [4–6]. In fact, Scollo et al. [6] found all studies documenting a negative economic impact of SF policies on the hospitality sector were sponsored by the tobacco industry. Furthermore, these studies were significantly more likely to use a subjective outcome measure and less likely to be peer-reviewed than studies showing no impact or a positive impact. Several studies that evaluated bars and restaurants separately reported no significant economic effects on either type of establishment [7–9] despite the known correlation between smoking and drinking behaviors [10]. Opponents have also raised concerns that statewide evaluations of SF policies may mask differential effects in some geographic areas such as border communities and rural regions, which may be important given that approximately 39% of smokers in the U.S. live within 40 miles of another state [11]. Economic studies have shown 4–25% of smokers cross borders to purchase cigarettes in jurisdictions with lower cigarette taxes [11,12]. Based on 2003 Current Population Survey Tobacco Use Supplement data, smokers traveled approximately three miles to save one dollar on a pack of cigarettes [11]. If smokers are willing to travel across borders to purchase cigarettes, they may also be willing to travel across borders to patronize bars and restaurants that permit smoking if those in their own jurisdictions do not. Only one study to date has examined the effects of a statewide SF policy on the hospitality sector in border communities [13]. These researchers evaluated whether Ohio’s statewide SF policy implemented in December 2006 differentially affected several hospitality sector employment indicators (including the total number of employees, total wages paid, and number of establishments) in selection of Ohio counties bordering Kentucky compared to nonborder counties; Kentucky did not have a statewide SF policy during the study period. Additionally, changes over time were compared between Ohio counties and both border and non-border counties in Kentucky. From three years before Ohio’s policy began to one year after implementation, there were no disproportionate changes in hospitality sector employment indicators between border and nonborder counties in Ohio or Kentucky. Thus, opponents’ arguments that a statewide SF policy would drive business across the border to jurisdictions without SF policies were not supported. Despite these findings, the issue of economic effects in border areas deserves continued study. Ohio offers a unique opportunity to empirically examine whether statewide policies negatively impact restaurant and bar establishments because until recently, all Ohio border counties (n = 88) were adjacent to five individual states without SF policies. Further, Ohio has suffered substantial set-backs in tobacco control including the 2008 state legislative dissolution of the Ohio Tobacco Prevention Foundation and diversion of Ohio’s master tobacco settlement funds to nontobacco-related budget items [14]. In fact, in 2011, Ohio did not contribute any state funds to tobacco control and
45
federal tobacco control funding equaled only 1.5% of the level recommended by the Centers for Disease Control and Prevention [15]. As the prevalence of adult smoking in Ohio remains one of the highest in the nation at 23.3% of adults reporting current smoking [16], (a slight decline from a high of 25.1% in 2011), this slower pace of decline compared to other states may represent a negative effect of diminished state tobacco control funding. All told, Ohio presents a unique natural quasi-experiment to evaluate the economic impacts of a statewide SF policy in an area where smoking remains relatively prevalent. This study provides three important extensions to previous work: (1) The use of taxable sales data as an objective and direct assessment of economic effects, (2) separate evaluations of bars and restaurant businesses to improve sensitivity to changes to specific business types, and (3) evaluation of the robust number of Ohio counties that share the border with five states. If Ohio residents responded to the SF law by traveling across state lines to restaurants and bars in states without comprehensive SF laws, as is suggested by the tobacco industry, sales in restaurant and bar taxable sales might be expected to suffer more in border counties, where individuals can more feasibly make such choices, than in more centralized counties. As a result, we hypothesized that border counties would experience a significant reduction in the rate (slope) of taxable sales compared to non-border counties for both business types. 2. Methods The study sample includes all 88 counties in the state of Ohio, dichotomized into those counties which share a border with another state and non-border counties. Border counties were defined as those sharing a border with one of five neighboring states (Indiana, Michigan, Pennsylvania, West Virginia, and Kentucky) (n = 26). One county shared less than 0.83 square miles with a neighboring state, including a single village of 525 residents, and was treated as non-border. None of the bordering states had 100% statewide SF laws covering restaurants and bars during the study period with the exception of Michigan, which implemented a law in the last month of the study period. However, five border counties were adjacent to non-Ohio counties with complete county-level smoke-free policies during at least some of the study period. Because border effects would not be expected in these areas, for the entire study period these counties were reclassified as nonborder. The final sample included a total of 21 border and 67 non-border counties (n = 88). Linear mixed model regression analyses were selected to evaluate for change in taxable sales within bars and restaurants before and after a SF policy was enacted for the state. Data were collected and provided for this analysis by the Ohio Department of Taxation (referred to here as the Department). The primary outcome variable was monthly taxable sales (in dollars). Hospitality businesses report sales tax liability every month. Very small businesses (defined as less than $1200 in state sales tax liability over a six month period) report every six months; businesses with biannual reporting of taxable sales were excluded from analyses
46
E.G. Klein, N.E. Hood / Health Policy 119 (2015) 44–49
because monthly data were not available. Sales tax liability was transformed into taxable sales by dividing the reported sales tax liability by the local sales tax rate (multiplied by 100) for the appropriate month and county. The local sales tax rate included county and transit taxes, where applicable. Monthly taxable sales were further classified by industry. Businesses that were most likely to be licensed to sell alcohol were selected using the North American Industry Classification System (NAICS) industry codes: full-service restaurants (NAICS code 7221) and “drinking places” (NAICS code 7224) referred to here as bars. Data were also obtained for the total hospitality sector, including fastfood and other limited service restaurants, catering, and accommodation businesses (NAICS code 72). For all taxable sales values, the unit of analysis available through the Department was county-level data. The Department suppressed data when there were fewer than five reporting businesses for a given county and month. The number of each type business (referred to as entities) by NAICS code was summed by county and month. Unadjusted unemployment rates were obtained from U.S. Bureau of Labor Statistics [17] in order to derive county-level unemployment statistics by month. The study period was June 2003 through May 2010 (n = 84 months). The statewide comprehensive SF policy was enacted on December 7, 2006 but not enforced until May 2007, yielding 37 months following policy enforcement. 2.1. Statistical analysis The goal of this research was to determine whether trends in county taxable sales changed following the passage of the statewide SF law, and whether those changes differed in border and non-border counties. To evaluate this question, we modeled monthly taxable sales as a function of time, using a spline to allow the time trend to change following the statewide SF policy enactment. We interacted both time trends with the indicator of border status to determine whether taxable sales time trends, both before and after the policy, differed in border and nonborder counties. Specifically, the following linear mixed models were estimated for each restaurant sales and bar sales. A linear mixed effects model with random intercepts and random slopes was used to examine whether taxable sales changed within border counties (i) over time (j) after enforcement of the Ohio Smoke-free Workplace Act. The selected model allows for the slope to change after the designated time (t*) when the state SF policy was enforced. E(Yij |i ) = ˇ0 + ˇ1 border + ˇ2 timeij + ˇ3 (timeij − t∗) + ˇ4 timeij ∗ borderij + ˇ5 (timeij − t∗) ∗ borderij + ˇ6 unemploymenti + ˇ7 entitiesi + ˇ8 overall sales for hospitalityi + ˇ9 springi + ˇ10 summeri + ˇ11 falli +, where t* equals May 2007 for the date of policy enforcement, and (timei − t*)+ is equal to (timei − t*)+ when
timei > t* and is equal to zero when timei ≤ t* [18]. Of most interest for this analysis are ˇ4 and ˇ5 ; an insignificant value for ˇ4 with a significant value for ˇ5 would suggest that county taxable sales time trends differed in border and non-border counties after, but not before, the policy passage, lending support to the possibility that the passage impacted the two areas differently. Covariates were included to account for other relevant economic factors that might influence the primary outcome: county-level unemployment, number of each type of business (entities), taxable sales for hospitality businesses excluding bars or restaurants, and seasons of the year (indicator variables for spring, summer, and fall; winter was used as the reference group). An exponential covariance structure was specified for the within-subject errors that depended on time. Models were run separately for fullservice restaurants and bars. An alpha level of 0.05 was used to identify statistical significance. Analyses were completed using SAS 9.3 (SAS Institute, Inc. Cary, NC).
3. Results A description of state and county-levels characteristics for the study sample at the study baseline in 2003 is shown in Table 1. For the state of Ohio, counties had an average of 88 restaurants and 21 bars, with a wide range for both types of hospitality businesses. Both border and non-border counties reported over $5 million in monthly taxable sales for restaurants, with approximately a half a million dollars monthly taxable sales for bars. There were no significant differences between border and nonborder counties in the number of entities, taxable sales, or unemployment (p > 0.05) as of June 2003 at the start of the present study. Figs. 1 and 2 depict the unadjusted monthly sums of taxable sales for border and non-border counties over the period of study, respectively for restaurants (Fig. 1) and bars (Fig. 2). In Table 2, the parameter estimates for the regression model are presented for restaurants and bars, respectively. The primary outcome variable is shown per $10,000 in taxable sales revenue, and all covariates (ˇ6 through ˇ11 ) were mean centered for ease of interpretation. For restaurants during the period of June 2003 to May 2007 before the Ohio SF policy was enacted, border counties had a difference in slopes of nearly $14,000 compared to nonborder counties (ˇ4 ; p = 0.03). Following the Ohio SF policy enactment, there was no statistically significant change in the difference in slopes for taxable sales in restaurants in border counties (ˇ5 ; p = 0.09) after adjustment for employment, overall sales and seasons of the year; in other words, the difference in the taxable sales rate for restaurants in border counties remained after the statewide SF policy. For bars, there was no significant differences in the slopes for border compared to non-border counties before policy enactment (ˇ4 ; p = 0.91) after adjustment for employment, overall sales and seasons of the year. Following the statewide policy enactment, the change in the different in sales revenue rate was not statistically significant comparing border and non-border counties (ˇ5 ; p = 0.07).
E.G. Klein, N.E. Hood / Health Policy 119 (2015) 44–49
47
Table 1 State and county-level descriptive characteristics of hospitality businesses in Ohio, 2003. Description
State of Ohio (n = 88) Mean (range)
Border (n = 21) Mean (range)
Non-border (n = 67) Mean (range)
Restaurants businesses Bars businesses Monthly taxable restaurant sales
40 (5–1036) 10 (5–188) $5,403,563 ($85,669–$71,316,301) $514,715 ($32,413–$7,856,139) 7.1% (4.6%–11.5%)
102 (8–625) 22 (5–92) $5,697,434 ($191,511–$52,896,386) $431,658 ($39,548–$2,059,480) 7.4% (4.9%–9.9%)
84 (5–1036) 21 (5–188) $5,311,455 ($85,669–$71,316,301) $544,616 ($32,413–$7,856,139) 7.0% (4.6%–11.5%)
Monthly taxable bar sales Unemployment rate Indicates p < 0.05.
$500 $450 $400
Millions
$350 $300
Non-border Border
Ohio Smokefree Workplace Act
$250 $200 $150 $100 $50 2003 2004
2005
2006
2007
2008
2009
2010
Fig. 1. Unadjusted total monthly restaurant taxable sales for border and non-border counties of Ohio, 2003–2010.
4. Discussion This study found that a statewide SF policy in Ohio did not have significant economic impacts on taxable sales for restaurants and bars in border counties that differed from non-border counties, which is consistent with the findings for the state as a whole [19]. These results are consistent with one study for a narrow sample of Ohio counties bordering the state of Kentucky that were evaluated within a year of policy enactment [13]. Another recent study found Ohio’s statewide policy was not associated with significant changes in another objective economic measure—employment indicators—in rural or urban counties [20]. Further, the separate analysis of bars and
restaurants provides robust, empirical evidence that Ohio’s border county bars did not suffer significant economic effects from the statewide SF policy, a common argument from opponents of statewide clean indoor air policies. All told, these findings present important evidence against arguments perpetuated by the tobacco industry and concerns among lay people that SF policies adversely impact the hospitality sector in some parts of the state. Because almost 40% of smokers in the U.S. live within relatively short driving distance of another state (i.e., 40 miles), findings from this study are important considerations for the rest of the country. Evidence that statewide SF policies do not differentially impact hospitality sector business near state borders even when adjacent to areas with more
$50 $45 $40
Millions
$35 $30 $25
Non-border Border
Ohio Smokefree Workplace Act
$20 $15 $10 $5 2003 2004
2005
2006
2007
2008
2009
2010
Fig. 2. Unadjusted total monthly bar taxable sales for border and non-border counties of Ohio, 2003–2010.
48
E.G. Klein, N.E. Hood / Health Policy 119 (2015) 44–49
Table 2 Association of the passage of a statewide smoke-free policy with trends in restaurant and bar taxable sales, in both border and non-border counties of Ohio, 2003–2010. Variables
Intercept (ˇ0 ) Border county (ˇ1 ) Monthly time trend before smoke-free law (ˇ2 ) Monthly time trend after smoke-free law (ˇ3 ) Difference in taxable sales rate before policy enactment in border counties (ˇ4 ) Change in the difference in taxable sales rate after policy enactment in border counties (ˇ5 ) County-level unemployment rate (ˇ6 ) # of restaurants entities (ˇ7 ) Overall hospitality sales (ˇ8 ) Spring (ˇ9 ) Summer (ˇ10 ) Fall (ˇ11 )
Restaurants
Bars
Regression estimates
Standard error (df)
Regression estimates
Standard error (df)
128.2 126.3 1.8
44.2 (86) 50.3 (86) 0.56 (7293)
0.005 0.014 0.001
12.8 12.2 0.3
8.15 (81) 10.52 (81) 0.16 (5941)
0.166 0.249 0.085
−1.9
0.57 (7293)
0.001
−0.3
0.10 (5941)
0.002
−1.4
0.64 (7293)
0.031
−0.02
0.19 (5941)
0.910
0.9
0.54 (7293)
0.089
0.2
0.1 (5941)
0.069
0.6
1.25 (7293)
0.616
0.4
0.23 (5941)
0.104
5.0
0.12 (7293)
