Brant Avondet and Dr. Arden C. Pope III, Economics
My analysis of available data was unable to determine the effect of Utah’s ban on smoking in public buildings on restaurant revenues.
In 1994, Utah became the second state to enact a complete ban on smoking in public buildings. According to Americans for Nonsmokers’ Rights, 144 localities have passed similar ordinances barring smoking in all workplaces, including restaurants. One study has been published estimating the effects of smoking bans on tourism. This study found that tourism increased in all but one of the municipalities. I proposed to study the effects of Utah’s ban on smoking in public buildings on restaurant revenues.
I gathered three data sets to do my analysis. The first was monthly tax data by county of restaurants from 1992-1997. The second was yearly data from 1988-1998 of restaurant revenues. The final data set contained monthly restaurant revenues from 1995-1998. I also tried to obtain revenue data from several local restaurants as well as the National Restaurant Association. They were unwilling to share revenue information with me.
Control Variables – After gathering the revenue data sets, I also located data for three control variables. First, I contacted the Utah State Tax Commission and obtained seasonality adjustment factors for restaurant revenues. I also found the number of restaurants (dining establishments excluding fast food and bars) in Utah over time. The final control variable I located was Utah’s economic growth over time.
Analysis – The data sets were regressed using SAS and Shazam. I controlled for seasonality, growth, and number of establishments. A dummy variable was included to determine the effect of the smoking ban.
Data set 1 – Restaurant Tax Data – The monthly tax data set revealed no significant correlation between enactment of the law and Utah’s restaurant revenues. In fact, the year for each data point proved to be more correlated (though not statistically significant) than the law. I believe that this correlation exists because revenues increased over time. For actual regression, the date was not included in the model.
Data set 2 – Yearly Restaurant Revenue Data – The yearly data yielded no significant correlation between enactment of the law and Utah’s restaurant revenues. The data set only contained 10 data points. The statistical power of so few observations proved too small to determine if the law had any effect on restaurant revenues.
Data set 3 – Monthly Restaurant Revenues From 1995-1998 – This data set contained no observations prior to the smoking ban. After much searching, emailing, and calling, I found that the state of Utah did not compile that information prior to 1995. Because the data set contained no observations prior to the enactment of the law, the monthly data set was useless in determining the effect of the law on restaurant revenues.
Data collection – Data collection appears to be the most difficult part of research. At least 90% of my time spent on this project was attempting to gather data. Running the regressions and analyzing the results were the fun part. I found three data sets used for the analysis. Unfortunately, none of them yielded statistically significant results. Admittedly, I feel a bit awkward submitting a paper where I was unable to find any hard and fast results. However, my lack of concrete results was not for lack of effort. First, I contacted the National Restaurant Association for the data. As they oppose any law that prohibits smoking in restaurants, they would not provide revenue information. Then, I found data that contained the taxes paid by restaurants in various Utah Counties over time. This data set contained too few observations to make any results statistically significant and contained wide variations, the least of which was caused by restaurants paying taxes at different times of the year. The second and third data sets were obtained from the state of Utah. The second data set contained yearly data from 1998 back to 1988. The third data set contained monthly data from 1995 to 1998. I ran regressions on the yearly data first, controlling first for growth in the economy and later for number of establishments. As the number of observations was only 10, the smoking ban coefficient was not significant. I then proceeded to analyze the monthly data in hopes of obtaining further years before the ban. I began by controlling for seasonality. I used the seasonality adjustments reported by the Utah state government. I also produced my own seasonality constants and controlled a duplicate data set with my adjustments. Next, I proceeded to control for growth in the economy; however, this correction did not account for number of establishments. I then proceeded to locate the number of establishments in each year and divided the revenues by the number of establishments before controlling for growth in the economy. This adjustment produced data that appeared to be fairly controlled. Several colleagues mentioned that controlling for both growth in the economy and growth in establishments could be redundant, so I retraced my steps with the other data sets and recalculated the coefficients. The results were virtually identical to the “double” adjusted data sets. Returning to the third data set, I continued to email, phone, and fax the state of Utah for data from years before the ban. They were very helpful, suggesting other possible sources for data as well as helping me find data through them. Unfortunately, reports on the monthly revenues of the restaurants were not compiled before 1995. Without data from before the change in the law, I was stuck. I looked for other ways to gather the data, including contacting a few local restaurants and other restaurant organizations. None of them were willing to share revenue data with me.