A longitudinal examination of the influence of a smoke-free law on the bar–crime relationship

Abstract

Researchers frequently identify relationships between categories of facilities and crime, particularly between alcohol-serving businesses and violent crimes. One question facing this research is whether categories of facilities, including bars, function as crime generators or whether analyses may be identifying spurious relationships that are the result of facilities locating in spaces encompassed by higher criminal opportunities. One technique for addressing this question is to examine whether policy changes that alter social functioning of facilities are associated with changes in criminal offending around the effected facilities. Previous research utilized pre- and post-policy change examinations. This paper offers a different method. Using time series analysis spanning multiple years, this paper examines the relationship between the implementation of a law prohibiting smoking in most public facilities, including bars and taverns, and reported crimes across aggregates in Minneapolis. Results indicate no change in crimes associated with the implementation of a smoke-free law, regardless of the number of bars within an aggregate. Implications for policy and examinations of categories of facilities as crime generators are discussed.

Keywords:

• Bars
• violent crime
• public policy

Acknowledgments

This research was conducted without support from a funding agency.

Notes

1. Facilities is a term used to describe homogeneous categories of spatially disaggregated land uses such as categories of businesses, government uses including schools and public housing units, and nature of residential housing units such as apartment complexes (Eck, Clarke, and Guerette 2007).

2. Minneapolis has 87 neighborhoods; however, there were geographical neighborhood changes during the time frame of the study. In 2001, the Phillips neighborhood was split into three neighborhoods: East Phillips, Midtown Phillips, and West Phillips. To keep all time series the same length, these three neighborhoods were merged into the original Phillips neighborhood. Furthermore, the North River Industrial Area merged with the Near North Neighborhood in January of 2006. These two series were merged into one series to be consistent.

3. While the data were provided for each neighborhood, these data were not available geocoded by address.

4. The counts of bars per neighborhood in the following analyses are static. Over the duration of the study period, the count of on-sale liquor licenses was highest in 2014, the year selected for neighborhood counts. Given that the locations of businesses is very stable over time as well as the general clustering of similar businesses together, year over year counts and reclassification of neighborhoods is likely to result in few to no neighborhoods being reassigned from high bar to low bar or vice versa.

5. The outlying neighborhood with 129 taverns was removed to calculate the mean and standard deviation, and the neighborhood was removed from the main analysis as a result of its outlier status. However, analyses were conducted including this neighborhood, and the results did not substantially change. Moreover, subsequent analyses are performed on this neighborhood alone and the results are reported subsequently.

6. While we agree with others (see McCleary and Hay, Jr. 1980; Myer and Chamlin 2011) that theory should guide the selection of a transfer function to reduce the probability of finding a significant impact of an intervention on a series by chance alone, it is plausible that a smoking ban produced a gradual effect on crime (first order) or a short-term change in crime levels that returned to normal (pulse). Both first-order and pulse transfer functions were estimated to assess whether any of these changes occurred. The results of these analyses did not produce any substantive difference from the results presented in this paper.

7. Bouffard, Bergeron, and Bouffard (2007) ARIMA analysis identifies a significant increase in police stops for DUI following the passing of a law extending bar closing times. Since this law took place during the time frame of our study (1 July 2003), we ran analyses that included a transfer function for the change in both the extended closing time and the indoor clear air law (e.g., see Chamlin and Scott 2014). Results did not change the results presented in this paper. In other words, modeling and estimating the impact of extending bar hours does not significantly impact the (lack of) relationship between a smoking ban and crime.

8. One alternative is to examine sales tax collections from the effected businesses. We obtained sales tax data from bars and taverns in Minneapolis from the Minnesota Bureau of Revenue to try and assess whether the policy change could have impacted criminal opportunities by altering the number of motivated offenders and suitable targets. Unfortunately, this agency was unable to provide sufficient data before the implementation of the smoke-free law to provide accurate estimates using the interrupted time series techniques we employed with the crime data. After adjusting the figures for inflation, a visual inspection of the data suggests no significant changes in sales at bars. While this indirect indicator can be interpreted as suggesting that the policy change did not alter facility functioning, we believe restraint is warranted with such an interpretation. The implementation of the smoke-free law may also change alcohol consumption patterns among patrons. Moreover, previous analyses of the impact of smoke-free laws focus on patrons gathering outside impacted businesses to continue smoking tobacco products. Without additional data information on the distribution of consumption among patrons or the concentration of patrons exiting facilities to consume tobacco products, pre–post-intervention comparisons may be inappropriate.