Article Text
PDF
Abstract
Objective To determine if the Ohio statewide smoke-free law is associated with economic activity in Ohio or Kentucky counties that lie on the border between the two states. In November 2006, Ohio implemented a comprehensive statewide smoke-free law for all indoor workplaces.
Design A feasible generalised least squares (FLGS) time series design to estimate the impact of the Ohio smoke-free law on Kentucky and Ohio border counties.
Setting Six Kentucky and six Ohio counties that lie on the border between the two states.
Subjects All reporting hospitality and accommodation establishments in all Kentucky and Ohio counties including but not limited to food and drinking establishments, hotels and casinos.
Main outcome measures Total number of employees, total wages paid and number of reported establishments in all hospitality and accommodation services, 6 years before Ohio's law and 1 year after.
Results There is no evidence of a disproportionate change in economic activity in Ohio or Kentucky border counties relative to their non-border counterparts. There was no evidence of a relation between Ohio's smoke-free law and economic activity in Kentucky border counties. The law generated a positive influence on wages and number of establishments in Ohio border counties. The null result cannot be explained by low test power, as minimum changes necessary in the dependent variables to detect a significant influence are very reasonable in size.
Conclusions Our data add to the large body of evidence that smoke-free laws are neutral with respect to the hospitality business across jurisdictions with and without laws.
- Smoke-free laws
- economic impact
- economics
Statistics from Altmetric.com
Over the last quarter century, numerous communities across the USA have enacted smoke-free legislation in indoor workplaces. One such example occurred in November 2006, when Ohio enacted a statewide smoke-free law prohibiting smoking in all workplaces and enclosed public places. The implementation of smoke-free laws has been the subject of much debate. It has been argued that smoking is a ‘right’ that is up to the discretion of the individual rather than legislative decisions. However, the scientific evidence is irrefutable that secondhand smoke is a significant public health risk and smoke-free legislation has clear public health benefits.1–10
Another issue that often spurs controversy is the potential effect of smoke-free laws on the local economy. Some policymakers fear that certain industries (namely hospitality and gaming industries) might suffer as a result of these legal restrictions and that businesses such as bars or casinos might need to be exempt from the legislation. The underlying argument is that people are more prone to smoke when they drink or gamble and that preventing the former activity might also prevent the latter (and, in addition, food consumption), with the net result being a significant decrease in revenues for the business.11 12 However, the overwhelming majority of published works disagree with this premise.13–15
Kentucky is traditionally heavily reliant on tobacco, both in production and use, while Ohio falls well behind in both areas. In spring 2004, Lexington-Fayette County implemented the first smoke-free ordinance in Kentucky. Pyles et al examined the association between this legislation and employment and business closures. They found no evidence of disproportionate change in economic indicators after the law.16 Since then a total of 27 cities and/or counties in Kentucky have enacted smoke-free ordinances or Board of Health regulations. Of those, the only Kentucky county bordering Ohio that has a smoke-free ordinance is Boyd County, in which the City of Ashland went smoke-free in October 2006.
Data and methods
We obtained employment data from the US Department of Labor's Bureau of Labor Statistics each quarter from January 2001 through December 2007. Specifically, we collected the total number of employees, total wages paid and total number of reporting establishments within the hospitality and accommodation services industry for each Kentucky and Ohio county.i Of these, we specifically focused on the counties that lie on the border between the two states, compared to their non-border counterparts. For Kentucky, the border counties include Boone, Campbell, Greenup, Kenton, Lewis and Mason.ii For Ohio, these are Hamilton, Clermont, Brown, Adams, Scioto and Lawrence. In addition, we collected economic data on the size of the labour force, population, income and unemployment rates for each county for each quarter from the Local Area Unemployment Statistics (LAUS), which is also from the US Department of Labor's Bureau of Labor Statistics.iii An examination of the descriptive statistics indicates that border counties are typically larger than non-border counties in each state, as evidenced by larger populations, larger labor forces and higher levels of each of the three indicator variables.
We examined all hospitality and accommodation services (ie, NAICS 72) in our reported analyses, including but not limited to food and drinking establishments, hotels and casinos.iv In unreported analyses we also examined several more restrictive samples, including restaurants only and drinking establishments only and find the results to be qualitatively unchanged from those reported. Since Ohio's smoke-free law went into effect during the fourth quarter of 2006, we classify the four quarters of 2007 as post-law and all preceding quarters as pre-law.v We began by examining the percentage change (before and after the Ohio law) in each of the three economic activity indicator variables for each specific county and then compared the changes for border versus non-border counties within each state. Post-law levels of each indicator were uniformly larger than pre-law levels (indicating increased economic activity) and the increase was typically larger for border counties in both states relative to their non-border counterparts. However, the differences in the increases between border and non-border counties were uniformly not significant.
To confirm, we implemented robust statistical techniques to control for county-specific anomalies. Specifically, we estimate the following feasible generalised least squares model:
where Dep is either LnTotEmp, calculated as the natural logarithm of the total number of employees, LnWages, calculated as the natural logarithm of total wages paid, or LnEstab, calculated as the natural logarithm of the number of reporting establishments. Each variable is for each county i during quarter t. βCounty and βQuarter are vectors of dummy variables represented each county and quarter, respectively. Ban is a dummy variable equal to 1 for county quarters following the Ohio smoke-free law and captures any post-law reactions in the dependent variables. BorderBan is a dummy variable equal to 1 if the observation is from a border county and following the implementation of the law, and captures an additional reaction for border (over non-bordering) counties following the law.
LnUnEmpRate is the natural logarithm of the unemployment rate for each county during each quarter. LnPopulation is the natural logarithm of the number of individuals in each county. LnOtherEmp is the natural logarithm of the number of employed individuals in all industries other than hospitality and accommodation services.vi
The analysis, like that in many studies utilising time series data, is complicated by serial autocorrelation in the dependent variables, as an examination of the Wooldridge test shows a need to control for this potentially biasing influence.17 18 Beck and Katz, building upon the Parks-Kmenta method, confirms that generalised least squares (GLS) has superior properties relative to traditional OLS methods when examining panel data.19–21 Beck and Katz conclude that if the variance-covariance matrix of the model standard errors are unknown, it is more appropriate to use feasible generalised least squares (FGLS). Thus, we examine the model using the FGLS and allow for group-wise heteroskedasticity and serial autocorrelation of order 1 with a common AR (1) coefficient for all panels.vii
Results
We first estimate the model on the total sample including all Kentucky counties (see table 1 panel A). We found no significant relation between Ban and LnTotEmp, and a positive and significant relation to LnWages and LnEstab.viii Kentucky counties experienced increased wages and number of establishments in the time period following Ohio's smoke-free law, while the number of employed individuals remained approximately the same.ix More importantly, the coefficient on BorderBan is not significant, indicating that the changes in activity in Kentucky border counties were not different from the change in non-border counties. Further, we examine only border counties (and excluded BorderBan from the model) and find no relation between Ban and any of the dependent variables at any reasonable significance level, indicating levels of economic activity were unchanged for border counties.
Panel data models using FLGS
We also examine Ohio counties in the same manner (table 1 panel B) and find similar results. We find a positive relation between Ban and LnWages, but again no significance on BorderBan, which indicates no evidence that Ohio border counties reacted differently from Ohio non-border counties to the smoke-free law. The positive coefficients on Ban when examining wages and establishments in only the border counties suggests that Ohio border counties benefited economically following the law. While the null finding could raise issues of model power, the largest necessary change in the dependent variables for a rejection of the null is less than 9%, with most models requiring a considerably smaller change.
Discussion
In November 2006, Ohio enacted a statewide smoke-free law prohibiting smoking in all workplaces and enclosed public places. Some predicted that the local economy in the northern Kentucky counties bordering Ohio would benefit from patrons crossing the state line to smoke following Ohio's smoke-free law.x As such, future implementation of similar smoke-free legislation in northern Kentucky might mitigate the presumed economic benefits. This theory, the ‘spillover smoker’ hypothesis, suggests that, following the law, smoke-free Ohio counties bordering Kentucky would experience a disproportionately lower level of economic activity than smoke-free Ohio counties that do not border Kentucky. Conversely, bordering Kentucky counties without smoke-free legislation would experience a disproportionate gain in activity relative to Kentucky counties that do not border smoke-free Ohio.
Unfortunately, we do not have the data available to definitively measure the ‘flow’ of patronage from one state to another. However, the lack of change in Kentucky border counties and the positive change in Ohio border counties (in wages and establishments) following the Ohio smoke-free law are not consistent with smoker migration from Ohio to Kentucky following the enactment of the law. Our data add to the large body of evidence that smoke-free laws are neutral with respect to the hospitality business across jurisdictions with and without laws.
What this paper adds
Previous studies have found that smoke-free legislation does not bring economic harm to local communities.
Our contribution is to focus on possible spillover economic effects when there is a smoke-free law in one state and smoking is not prohibited in the communities bordering that state. We find no evidence of the spillover effect either way. We find Kentucky and Ohio border counties did not disproportionately change economically from Ohio's state smoke-free law. Any concern over a change in business activity should similar legislation be enacted in bordering northern Kentucky counties appears unfounded. Our findings add to the evidence that suggests that both local and contiguous economies need not fear significant macroeconomic changes following smoke-free legislation.
Acknowledgments
This study was funded by the Kentucky Prevention Research Center through a grant from the National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention; Cooperative Agreement # U48/DP000039. The authors would also like to thank Mary Jones for valuable assistance with data collection.
References
Footnotes
Funding The paper was supported financially by the Kentucky Prevention Research Center through a grant from the National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention; Cooperative Agreement # U48/DP000039.University of Kentucky Medical Center, CC444 Roach Facility Markey Center, Lexington, KY 40506-0093, USA.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
↵i Aggregate variables such as those examined here may be slow to reflect changes in economic activity, particularly since the majority of consumers are non-smokers. Thus, it would be prudent to re-examine our findings over a longer time period once the data become available.
↵ii There are a total of nine Kentucky counties that border Ohio. Of these, Boyd is excluded since it had one city with a comprehensive smoke-free law. In unreported analysis, we include Boyd in the border counties and find the conclusions of the paper unchanged. In addition, data were unavailable for Bracken and Pendleton counties in Kentucky.
↵iii Per capita income for Ohio counties was graciously provided by the Ohio Department of Development.
↵iv The specific subcategories that make up NAICS 72 can be found at http://www.census.gov/eos/www/naics/.
↵v In unreported analyses, we included the fourth quarter of 2006 as post-law but the results were unchanged. Since the law was in effect for less than half the quarter, we chose to classify it as pre-law in the reported analyses.
↵vi We also examined models that include the counties of both states simultaneously and find the results to be consistent with those reported here.
↵vii An alternative approach, also suggested by Beck and Katz, is to produce panel-corrected SE estimates. The results provided were unchanged with this method.
↵viii For classification, we assume that significance is determined by a p value ≤0.05.
↵ix The positive association with total wages paid is likely due to increases in the minimum wage requirement. While many employees in the hospitality industry (ie, tipped employees) are not subject to the minimum level, a subset is and the increase likely skews the results. In January 2007, Ohio increased their minimum wage from $5.15 to $6.85, while Kentucky increased the minimum wage requirement from $5.15 to $5.85 in June 2007. Thus, any conclusions drawn from the positive relation between the post-law period and wages should be tempered.
↵x An alternative line of thought is that patrons from northern Kentucky would cross the state line to enjoy smoke-free venues in Ohio. This might suggest a positive reaction in Ohio border counties and a corresponding negative reaction in bordering Kentucky counties.