Legislation to make all enclosed areas in South Australian hotels and licensed clubs smoke-free was introduced on 1 November 2007.1 While smoke-free policies reduce exposure to secondhand tobacco smoke, they are inevitably accompanied by claims from industry that such legislation leads to adverse business sales in the hospitality industry.2 Previous studies examining the impact of smoke-free restaurant laws have demonstrated no adverse effect on sales in restaurants and cafes, but the number of published studies examining the effect of these laws in hotels and licensed clubs in Australia is limited.
Our study was designed to examine the economic impact of the smoke-free law on business revenue for hotels and licensed clubs in South Australia. We obtained seasonally adjusted monthy turnover data for hotels and licensed clubs from July 2002 to June 2008 from the Australian Bureau of Statistics Retail Business Survey,11 and adjusted for inflation using the ABS Consumer Price Index (CPI).12 To account for underlying economic trends, unemployment and population changes, we followed the procedure suggested by Glantz and Smith4,5 whereby the ratio of monthly turnover for hotels and licensed clubs to total monthly retail turnover (minus hotels and licensed clubs turnover) in South Australia was computed. This ratio (hereafter called RATIO) would be expected to decrease if the implementation of the smoke-free law had an adverse effect on retail turnover.
Linear regression was used to examine the effect of the smoke-free legislation on RATIO. To control for secular trend, a difference transformation was applied. This transformed variable was then used as the outcome (dependent) variable and a dummy (intervention) variable representing the timing of the smoke-free legislation was set up as a predictor (independent) variable. Once the series was de-trended, we used Ljung Box Q statistics to test the hypothesis of no autocorrelation.13 The presence of significant autocorrelation in most time series usually means techniques such as ordinary least squares (OLS) regression are not recommended, with analysts preferring instead such techniques as autoregressive integrated moving average (ARIMA) models developed by Box and Jenkins.14 This technique was not required in the present analysis as the serial dependence (autocorrelation) of the data disappeared once the series had been made stationary.
Two hypotheses for the impact of the legislation were considered: an abrupt and permanent impact (Model 1), and an abrupt but temporary impact (Model 2). Table 1 presents the results of the regressions testing the hypotheses that the introduction of the smoke-free legislation resulted in either an abrupt and permanent change (Model 1), or an abrupt but temporary change (Model 2) in retail turnover. While only Q1, Q6 and Q12 are presented, all Ljung-Box Q statistics were found to be non-significant for each model, suggesting the residuals were white noise.
|Model 1a||Model 2b|
Because the intervention coefficient was not significant for either model, neither of the hypotheses could be supported, suggesting the introduction of the smoke-free law had no impact on monthly turnover for hotels and licensed clubs in South Australia. This finding is consistent with research from Tasmania and overseas8–10 and studies in the restaurant setting where smoking bans did not reduce overall business.3–7 The results contribute to the wider body of evidence demonstrating that smoke-free hospitality laws do not negatively affect business sales.