Spatial Clustering of Hookah Lounges, Vape Shops, and All Tobacco Retailers Near Colleges

Abstract

Introduction

US college students smoke hookah and vape nicotine at higher rates than other young adults. Density and/or proximity of hookah lounges and vape shops near colleges has been described, but this study is the first to test whether tobacco retailers spatially cluster near college campuses.

Aims and Methods

We created and linked spatial shapefiles for community colleges and 4-year colleges in California with lists of hookah lounges, vape shops, and licensed tobacco retailers. We simulated 100 datasets, placing hookah lounges, vape shops, and tobacco retailers randomly in census tracts in proportion to population density. A modified version of Ripley’s K-function was computed using the radius (r) from each retailer within retail category.

Results

In 2018-2019, 50.5% of hookah lounges (n = 479), 42.5% of vape shops (n = 2467), and 42.0% of all tobacco retailers (n = 31 100) were located within 3 miles of a community college. Spatial clustering was significant (p < .05) from at least 0.4 miles for hookah lounges, 0.1 mile for vape shops, and 0.3 miles for all tobacco retailers. For 4-year colleges, approximately 46.8% of hookah lounges, 31.3% of vape shops, and 31.6% of all tobacco retailers were located within 3 miles. Clustering was significant from 0.2 miles for hookah lounges and 1.3 miles for all tobacco retailers but was not significant for vape shops.

Conclusions

Evidence that some types of tobacco retailers cluster near community colleges and 4-year colleges implies greater accessibility and exposure to advertising for students. It is also concerning because a higher probability of underage tobacco sales presumably exists near colleges.

Implications

Prior studies infer that hookah lounges and vape shops cluster near colleges from the density and closer proximity to campuses. This study modified a traditional test of spatial clustering and considered community colleges separately from 4-year universities. Spatial clustering of hookah lounges and all licensed tobacco retailers was evident near both types of campuses, but vape shops clustered only near community colleges. Place-based strategies to limit tobacco retail density could expand state and local laws that prohibit tobacco sales near schools to include retailers near college campuses. In addition, college environments should be a target for reducing hookah smoking and nicotine vaping.

Introduction

One in three US young adults still use tobacco, and college students smoke hookah and vape nicotine at significantly higher rates than other young adults.^1,2 According to the Monitoring the Future Survey, last-year hookah use was reported by 10.1% of college students in 2017 and 10.6% in 2019, but declined from 16.8% to 8.7% among same-age peers who were not in college.^2,3 The prevalence of vaping nicotine in the past 30 days among college students increased from 6.0% in 2017 to 22.1% in 2019 and from 7.9% to 18.4% among same-age peers who were not in college.^2,3

The majority of US college students’ exposure to tobacco marketing occurs at the point of sale.^4,5 Indeed, the tobacco retail environment near colleges is believed to contribute to higher rates of hookah smoking and nicotine vaping by providing convenient access to these products, by displaying more advertising for flavored e-cigarettes, and by offering discounts that appeal to price-sensitive consumers.^5,6 Such marketing may contribute to misperceptions that hookah is less harmful than smoking cigarettes,^7–9 that hookah and vape products contain less nicotine than cigarettes, and to overestimations of the prevalence of peer use.^10–12

For these reasons, a growing body of research examines the proximity of hookah lounges and vape shops to colleges, both in the United States and in Canada. A 2014 study identified 1690 waterpipe smoking establishments (ie, hookah lounges) in the United States, and estimated that 38.1% of 554 residential colleges were located within 3 miles of a hookah lounge.¹³ A 2015 study estimated that there were 9945 vape shops in the United States, finding that 66.5% of 2755 colleges were located within 3 miles of a vape shop.¹⁴ In addition, vape shop density was greater near 4-year than 2-year colleges, but not in models that adjusted for other variables, such as type of college (public or private), presence of campus housing, and location (urban, suburban, rural). In North Carolina and Virginia, higher density of hookah lounges, vape shops, and licensed tobacco retailers was associated with a greater proportion of college-enrolled residents in census tracts.¹⁵ Among a college cohort in those states, the proximity of hookah lounges and vape shops was associated with a greater likelihood waterpipe tobacco use in the past 6 months.¹⁵ In the province of Quebec, researchers identified 299 vape shops in 2015, and observed a significantly shorter median distance to colleges (1231 m) than to secondary schools.¹⁶ Thus, prior studies infer that hookah lounges and vape shops cluster near colleges from the density and closer proximity to campuses.

The current study addresses three critical gaps in the current literature. First, it creates spatial data for college campus boundaries rather than estimating distances from the campus address point. Second, it considers community college campuses separately from 4-year colleges. The inclusion of community colleges is important because compared with students attending a 4-year colleges in the United States, community college students report higher odds of current and daily smoking.^17,18 Finally, no previous study examined whether hookah lounges and vape shops, in fact, cluster near college campuses. A traditional test for spatial clustering rejects the null hypothesis of spatial randomness. However, the current approach compares observed data with a null hypothesis that tobacco retailers locate near people, by simulating retail locations in census tracts in proportion to population density.

Methods

Study Setting

Data for the current study are from California, where approximately 4 in 10 of the state’s 3.9 million residents (ages 18–24) are enrolled in a community college and/or public university.¹⁹ California was the second US state to restrict the sales of flavored tobacco, with exemptions for hookah, large cigars, and pipe tobacco in age-restricted venues.^20,21 However, the tobacco industry delayed implementation by challenging the state law on a November 2022 ballot referendum.²²

The sources and methods used to create and link spatial data for college campuses, as well as hookah lounges, vape shops, and all tobacco retailers are described below.

Campus Boundary Shapefiles

To improve on prior estimates that used the street address for college campuses, we curated or created spatial shapefiles for main campus boundaries using assessor parcel data and aerial imagery. The resulting campus boundary shapefiles for 114 public community colleges and for 104 main campuses of 84 4-year colleges with housing for at least 250 students (70 private and 34 public institution campuses; 52 private and 32 public institutions) are online: http://www.californiaschoolcampusdatabase.org.

Hookah Lounges

This study identified hookah lounges as defined by keyword searches online. We searched Google Maps for the terms “hookah bar” and “hookah lounge” and obtained a list of establishments in the “Hookah Bars” category on Yelp between December 31, 2019 and January 2, 2020.²³ Each file was de-duplicated using street address and zip code to flag possible duplicates, and manual review of all available record information (street address, store name, and phone number) to confirm whether records were for the same retailer. An additional n = 8 records from the Yelp without a street address were removed, yielding a cleaned Google file with n = 354 likely hookah lounges, and a Yelp file with n = 393. These files were merged based on street address and zip code, and manually reviewed using a combination of business name and phone number. When street address records between sources matched but business name and phone number yielded ambiguous linking status, research staff performed online searches to determine whether records were for the same business. The merging process yielded a single file with n = 481 unique hookah lounges, which were geocoded to latitude and longitude using ArcMAP 10.6.1 (match rate = 100%).

Vape Shops

This study identified vape shops as defined by keyword searches online. Using customized keywords, we searched Google Maps for the terms “vape shop,” “vaporizer store,” and “vape store” and obtained a complete list of establishments in the “vape shops” category on Yelp in April 2019. Downloaded records in each source file were geocoded using ArcGIS v10.6.2 (Google file match rate = 98.7%, Yelp file match rate = 99.4%), and key variables were created with common formats and content to increase matching within and between files. Each file was de-duplicated using street address combined with business name and phone number, yielding a cleaned Google file with n = 1878 likely vape shops, and a Yelp file with n = 1616. These files were merged via statistical software using street address and zip code (IBM SPSS Statistics for Windows, v.25), and manually reviewed by research staff incorporating business name and phone number to identify additional duplications and matched stores that did not link due to variations in street addresses between the two sources. The final combined dataset contained n = 2467 unique vape shops, including some stores that also sold conventional tobacco products.

Tobacco Retailers

A list of all retailers with a state tobacco retail license was obtained from the California Department of Tax and Finance Administration in October 2018. Addresses for all licensed tobacco retailers (n = 31 100) were geocoded using ArcGIS 10.6.1 (match rate = 100%). The list did not provide information that classifies tobacco licenses by store type, which is why online searches were used to identify hookah lounges and vape shops.

Statistical Analyses

Descriptive Statistics

For comparison with previous studies,^13,14 we calculated the proportion of college campuses within 3 miles of a hookah lounge, vape shop, and any licensed tobacco retailer. These proportions were calculated using distances from 218 college campuses to hookah lounges, vape shops, and all tobacco retailers that were computed two ways: (1) straight line (Euclidean) distance from the campus polygons (boundary) and (2) straight line miles from the campus address points. Descriptive statistics for campuses with hookah lounges, vape shops, and licensed tobacco retailers within 3 miles were computed separately for type of campus, community college and 4-year college. We did not test for differences in the proportion of campuses near retailers because some campuses are located near one another and some of the same stores are located near both types of campuses. Therefore, tests for differences by type of campus would violate assumptions of independence. However, paired t tests examined whether the distance to the nearest retailer differed for campus polygons and address points, computed separately for vape shops, hookah lounges, and all tobacco retailers and by type of campus.

Clustering Analysis

Our novel approach differs from the traditional version of Ripley’s K-function, a statistic used for detecting spatial clustering.²⁴ Using the spatial data for campuses and retailers, we computed the straight line distance from each retailer to nearest community college and 4-year college campus boundary. To examine whether the retailers cluster around colleges, we developed a modified definition of the K-function, which counts the proportion of stores within r miles of a college, rather than the proportion of points within r miles of another point:

$${\displaystyle \begin{array}{l}{\displaystyle K(r)=\frac{1}{n}{\displaystyle \sum _{i=1}^n}I(\mathrm{s}\mathrm{t}\mathrm{o}\mathrm{r}\mathrm{e}i\mathrm{i}\mathrm{s}\mathrm{w}\mathrm{i}\mathrm{t}\mathrm{h}\mathrm{i}\mathrm{n}r\mathrm{m}\mathrm{i}\mathrm{l}\mathrm{e}\mathrm{s}\mathrm{o}\mathrm{f}\mathrm{a}\mathrm{c}\mathrm{o}\mathrm{l}\mathrm{l}\mathrm{e}\mathrm{g}\mathrm{e})}\end{array}}$$

Also, we modified the null hypothesis because the classical K-function analysis simulates points completely at random over the region, which is inappropriate for this application. Even under the assumption that stores do not cluster near colleges, we expect stores to be located where people live. For this reason, we used census tract data in order to simulate the expected store locations in proportion to population density. Therefore, our novel test of spatial clustering rejects a null hypothesis that assumes hookah lounges, vape shops, and tobacco retailers cluster where people live, rather than rejecting a null hypothesis that retailers are randomly distributed.

A custom Python script was used to compute the modified version of Ripley’s K-function described above. This K-function compares the distribution of observed events to a random generation of events assuming no clustering. To do this, 100 artificial datasets were created, randomly placing all vape shops across the state of California in proportion to the population in each census tract. The K-function was then computed by creating successive circles of radius r miles around each store and calculating the percentage of stores with a college within that circle. This function of r was plotted for the observed data and each of the 100 simulations. A p value was calculated as the proportion of simulated datasets that exhibited greater clustering than the realized dataset. These procedures were repeated for hookah lounges and all licensed tobacco retailers in California.

Results

California had 479 hookah lounges, 2467 vape shops, and 31 100 licensed tobacco retailers (including hookah lounges and vape shops) in 2018-2019.

Proximity of Colleges to Retailers

For comparison with previous studies,^13,14Table 1 summarizes two measures of distance from colleges to hookah lounges, vape shops, and all licensed tobacco retailers: (1) straight line from the campus boundary and (2) roadway miles from the campus address point, separately for community colleges and 4-year colleges. Using straight line distance from campus boundary, 57.0% of community colleges and 77.9% of 4-year colleges were within 3 miles of at least one hookah lounge; 93.9% of community colleges and 95.2% of 4-year colleges were within 3 miles of at least one vape shop. The percent of colleges whose address point was located within 3 roadway miles of at least one hookah lounge was 47.4% for community colleges and 76.0% for 4-year colleges; 89.5% for community colleges and 92.3% for 4-year colleges were within 3 roadway miles of at least one vape shop. By comparison, nearly all community colleges and 4-year colleges were within 3 miles at least one licensed tobacco retailer, regardless of how distance from campuses to retailers was measured.

Table 1.

Proximity of Hookah Lounges, Vape Shops, and Licensed Tobacco Retailers to Colleges, by Campus Boundary and Address Point: California, 2018-2019

	Community colleges (N = 114)		Four-year colleges (N = 104)
	Boundary	Address point	Boundary	Address point
Percent of campuses within 3 miles
Hookah lounges	57.0	47.4	77.9	76.0
Vape shops	93.9	89.5	95.2	92.3
Licensed tobacco retailers	97.4	95.6	100.0	98.1
	M (SD)	M (SD)	M (SD)	M (SD)
Distance from campus to nearest retailer
Hookah lounges	9.6 (18.8)	9.7 (18.3)	2.5 (3.7)	2.8 (3.8)
Vape shops	1.1 (1.3)	1.3 (1.3)	1.0 (1.6)	1.3 (1.7)
Licensed tobacco retailers	0.3 (0.7)	0.6 (0.7)	0.3 (0.5)	0.5 (0.6)

Cell entries are percent of campuses within 3 miles of at least one retailer and average distance (M) and SD from campuses to nearest retailer derived from two calculations: Euclidean (straight line) distance (miles) from campus boundary and address point.

Table 1 also summarizes the average distance to the nearest retailer, separately for community colleges and 4-year colleges and for each type of retailer. Results from paired t tests suggest significantly shorter distances to the nearest retailer from the campus boundary than the address point, with the exception of hookah lounges near community colleges (see Supplementary Table S1).

Clustering Analysis

Different from the descriptive statistics that summarized distance from college campuses to retailers, the clustering analysis focuses on the distance from retailers to campuses. Specifically, the clustering analysis focused on straight line distance from retailers to the nearest community college and 4-year college campus boundary. Approximately half (50.5%) of hookah lounges, 42.5% of vape shops, and 42.0% of licensed tobacco retailers were located within 3 miles of a community college campus boundary. Similarly, 46.8% of hookah lounges, 31.3% of vape shops, and 31.6% of all licensed tobacco retailers were located within 3 miles of a 4-year college campus boundary.

Figure 1 compares the modified K-function for the observed data (straight line distance from retailer to community college campus boundaries) with the modified K-functions from 100 simulations that assumed the locations of interest are predicted by population density but not by proximity to colleges. In each panel of Figure 1, the K-function for observed data is represented by the black line. The faint gray lines represent simulated data for community colleges, with separate figures for hookah lounges (Panel 1), vape shops (Panel 2), and all licensed tobacco retailers (Panel 3). Clustering is illustrated by a gap between the observed data (black line) and the simulated data (gray lines). When the observed data (black line) fall systematically above the K-functions for the simulated data (gray lines), the results indicate that the retailers cluster near college campuses, beyond what we would expect based on population density alone. Even if all of the functions are increasing with distance from campus, the observed function increases much faster than the simulated functions. For community colleges, spatial clustering was significant (p < .05) beginning at 0.4 miles for hookah lounges, 0.1 mile for vape shops, and 0.5 miles for all licensed tobacco retailers.

Figure 1.

Percent of hookah lounges, vape shops, all licensed tobacco retailers within r miles of community colleges: California 2018-2019. In each panel, the black line represents observed data and the faded lines represent simulations. Clustering is illustrated by a gap between the observed data (black line) and the simulated data (faded lines), beginning at 0.4 miles for hookah lounges, 0.1 mile for vape shops, and 0.5 miles for all licensed tobacco retailers.

For 4-year colleges, the K-function tests reveal a different pattern (see Supplementary Figure S1). Spatial clustering was significant (p < .05) beginning at 0.2 miles for hookah lounges and 1.3 miles for all licensed tobacco retailers (Supplementary Figure S1, Panels 1 and 3). However, vape shops did not cluster near 4-year colleges (Supplementary Figure S1, Panel 2).

Sensitivity analyses using the spatial data for distances from retailers to college address points reveal findings similar to the primary cluster analyses using campus boundaries (Supplementary Figure S2). Thus, different data sources (campus boundaries vs. address points) mattered more for determining the distance to the nearest hookah lounge, vape shop, and licensed tobacco retailer than for determining whether those stores were spatially clustered near colleges.

Discussion

This study is the first that we know of to test whether hookah lounges, vape shops, and all licensed tobacco retailers cluster near college campuses, rather than infer clustering from data for retailer proximity or density. Even under the assumption that stores do not cluster near colleges, we expect stores to be located where people live. Therefore, we used census tract data in order to simulate stores proportional to population density. For community colleges, spatial clustering was significant beginning at 0.4 miles for hookah lounges, at 0.1 mile for vape shops, and at 0.5 miles for all tobacco retailers. For 4-year colleges, clustering was significant beginning at 0.2 miles for hookah lounges and at 1.3 miles for all tobacco retailers, but was not significant for vape shops.

The tobacco industry claims that the proximity of tobacco retailers to secondary schools is not surprising, because retailers of all kinds locate near people.²⁵ Although this research focused on colleges and universities rather than secondary schools, evidence from simulations suggests that tobacco retailers do not solely locate near people. Instead, tobacco retailers are clustered near college campuses. Building on previous evidence about the density and proximity of hookah lounges and vape shops near colleges, the current study suggests that a pattern of spatial clustering also differs by store type. Spatial clustering near both community colleges and 4-year colleges was not unique to hookah lounges but evident for all licensed tobacco retailers, as well. For community colleges, spatial clustering of all three types of tobacco retailers was detected within 0.5 miles, but the distance at which clustering was detected near 4-year colleges was nearer to campus for hookah lounges than for all licensed tobacco retailers.

Notably, vape shops clustered near community colleges uniquely, and not near 4-year colleges. The reasons for this are not known. However, prior evidence suggests that campus smoke- and vape-free policies were associated with a lower density of vape shops near US 4-year colleges.^14,26,27 In California, the prevalence and strength of tobacco-free campus policies varies within and between community college and 4-year colleges, and was related to the presence of advertising for e-cigarettes near campuses.⁶ Future research should consider whether hookah lounges, vape shops, and other tobacco retailers are less likely to cluster or locate near colleges with comprehensive tobacco-free campus policies.

The main strength of the current study is the novel application of spatial statistics to study whether hookah lounges and vape shops cluster near college campuses. A traditional test for spatial clustering rejects the null hypothesis of spatial randomness. However, the current approach rejects the null hypothesis that tobacco retailers locate near people, by simulating their locations in census tracts in proportion to population density. The inclusion of community colleges is also noteworthy because this population is understudied.

Because address points could be a poor approximation of college campus centroids and distances from the campus boundaries, the creation and use of spatial shapefiles to compute distances between college campuses and retailers is another strength of this research. Shorter distances to the nearest retailer from the campus boundary than the campus address point were observed, with the exception of hookah lounges near community colleges. Notably, the results using spatial data for address points in the K-function tests showed roughly the same patterns of clustering as the data for campus boundaries.

A 2014 study that identified 1690 hookah lounges in the United States and 38.1% of 4-year colleges within 3 miles of at least one.¹³ By comparison, the current results from California, with 479 hookah lounges and 76.0% of 4-year colleges within 3 miles of a hookah lounge in 2018-2019, suggest that the number of hookah establishments and their proximity to colleges may have increased over time. However, further research is needed to determine how the two different sources of spatial data (address points and campus boundary shapefiles) affect the estimated distances to hookah lounges, vape shops, and other tobacco retailers.

A potential weakness of the current study is that the definition of vape shops is not limited to stores that sell vape shops exclusively, as in some previous research.^28,29 The inclusion of hybrid stores that sell other tobacco products affect measures of vape shop density and neighborhood correlates of vape shop locations.³⁰ However, the category of stores that we studied is consistent with how users would search online for vape shops nearby. Another limitation is that the online search strategies may overestimate or underestimate hookah lounges and vape shops, and the lists were not validated for sensitivity and specificity. In addition, we did not exclude hookah lounges and vape shops from all licensed tobacco retailers because the state licensing list does not categorize business types. However, it seems unlikely that hookah lounges and vape shops would entirely explain the clustering of all tobacco retailers near community colleges and 4-year colleges, particularly since vape shops did not cluster near the latter. In addition, the sum of hookah lounges and vape shops that were identified online represented a small proportion (17.4%) of the total number of tobacco retailers on the state licensing list.

Future research should consider whether California’s exemption for hookah from sales restrictions on flavored tobacco increase the number of hookah lounges and/or exacerbate their clustering near college campuses. The methods developed for this study could be used to test whether exemptions for other types of retailers (eg, liquor stores or adult-only smoking bars) elsewhere affects their clustering near colleges and/or schools. Of course, results about spatial clustering near California college campuses may not generalize to other US states. Future research should also consider other tobacco retail environments, particularly where licensing policies do not exist and other tobacco control measures are weaker.

Results from this study provide further evidence that college environments could be prioritized for interventions regarding hookah smoking and nicotine vaping.³¹ After data for the current study were collected, the prevalence of past-month nicotine vaping among college students decreased from 22.1% in 2019 to 18.6% in 2020 but increased from 18.4% in 2019 to 23.6% in 2020 among same-age peers who were not in college.³² Research is needed to assess whether the same reversal (with higher use among young adults not in college) characterizes hookah use, and how the retail environment for tobacco contributes to these trends.

Evidence from the current study also has implications for developing place-based strategies to reduce the number and density of tobacco retailers. States and localities could consider expanding restrictions on tobacco sales near K-12 schools to include college campuses. That tobacco retailers cluster near colleges is concerning because a higher probability of illegal sales to underage purchasers (ages 18–20) presumably exists in college neighborhoods. In California, the rate of sales violations to decoys ages 18–19 was 44.7% in a combination of vape shops and smoke shops in 2018.³³ Although lower violation rates were obtained from inspections in vape shops that were conducted by the US Food and Drug Administration (FDA), perhaps proximity to colleges should be considered in the sampling frame for underage sales inspections by FDA as well as state and local enforcement agencies.

Supplementary Material

A Contributorship Form detailing each author’s specific involvement with this content, as well as any supplementary data, are available online at https://academic.oup.com/ntr.

Funding

This work was supported by the National Institutes of Health/National Cancer Institute, grant #R01-CA217165 (PI: Henriksen).

Declaration of Interests

DLS was employed by Google, Inc. in a capacity unrelated to this research. NCS and LH receive research funds from the California Tobacco Control Branch of the California Department of Public Health.

Data Availability

The data underlying this article will be shared on reasonable request to the corresponding author.