Sociodemographic Characteristics

Sociodemographic variables at the census tract level from the 2015–2019 5-Year American Community Survey were downloaded from Social Explorer.² Variables included total population; percent of non-Hispanic white (white), non-Hispanic Black or African American (Black), and Hispanic or Latino (Hispanic/Latino) individuals; percent of population 65+, and median household income.

Premium Cigar Association Retailer Members

The Premium Cigar Association (PCA) provided a retailer-level dataset that included current members in 2019 through 2021, with all physical brick-and-mortar retailers that are PCA members and where the majority of sales were for premium cigars or pipe tobacco;³ retailer types, such as convenience stores, hookah/head shops, and vape stores, were excluded.

The original dataset included 1,316 unique retailers. However, upon review of the address fields and states, further cleaning of several observations was performed in consultation with the National Academies of Sciences, Engineering, and Medicine, briefly described below and detailed in Supplement A:^4,5

1.

25 observations were removed:

a.

Not in the USA or territory (n = 21)

b.

Store closed upon review of address (n = 1)

c.

Could not determine address (n = 3)

2.

Addresses corrected from PO Box to retailer address using Google (n = 10)

3.

Listed address corrected using Google (n = 4)

After cleaning the dataset, 1,291 unique retailers remained. Using ArcMap 10.8.1, all retailers were geocoded and assigned a latitude and longitude (see Figure G-1).

FIGURE G-1

Locations of 2019–2021 U.S. Premium Cigar Association retailers (n = 1,291).

Calculating Premium Cigar Retailer Density

All premium cigar retailers were spatially joined to their respective census tract and corresponding sociodemographic data, and the number of retailers per 1,000 people in a tract was calculated (Figure G-2). Tracts with zero population (n = 611; two PCA members) were omitted from analysis (total number of PCA members n = 1,289).

FIGURE G-2

2019–2021 Premium Cigar Association member retailer density (retailers per 1,000 people), U.S. census tracts (n = 72,410). NOTE: Census tracts with zero population (white) were omitted from the analysis (n = 611)

Analysis

Using SAS 9.4, Spearman correlations between each tract-level sociodemographic variable and premium cigar retailer density were calculated. In additional exploratory analyses, a binary variable indicating the presence (=1) of at least one premium cigar retailer (versus none = 0) was created. Unadjusted associations between sociodemographic characteristics and this binary indicator were fit. Tract-level age, race, and ethnicity variables were scaled to tens (e.g., 10 percent is coded 1.0) so that estimates may be interpreted as the expected difference for a census tract that has a 10-percentage point greater value in the sociodemographic variable. Median household income was scaled to $1,000. As census tracts are nested within states (and Washington, DC), general estimating equation (GEE) methods with an exchangeable working correlation matrix were used, which adjust both estimates and standard errors to account for this geographic nesting.

PCA Members Retailer Density

Figure G-2 shows very little variation in premium cigar retailer density (mean = 0.00, SD = 0.06), largely due to minimal variation in the total retailer count (Tables G-1 and G-2). Supplements B–D provide state-level maps and a table of the total count of retailers and density per 1,000,000 residents.⁶

Spearman correlations (see Table G-3) indicated very small positive correlations, though caution should be taken interpreting these results due to the large sample size (statistical significance is likely) and because geographic nesting is not accounted for in analyses.

Further exploration indicated that only 16 census tracts had a premium cigar retailer density of 1.00 or more, and this may be due to lower population sizes in the tract (Supplement E). These observations had no apparent patterns in sociodemographic characteristics.

Presence of Absence of Retailer

With so little variation in the total retailer count, unadjusted associations between each sociodemographic characteristic and the presence (versus absence) of a retailer were investigated. As indicated in Table G-2, 71,164 of census tracts (98.3 percent) did not have a premium cigar retailer, while 1,246 (1.7 percent) had one or more. Descriptive analyses indicating the average value of each sociodemographic characteristic are presented in Table G-4.

Results from the GEE models (see Table G-5) indicate that tracts with a higher percentage of white residents had greater odds of having at least one (versus none) premium cigar retailer (odds ratio [OR] = 1.06; 95 percent confidence interval [CI]: 1.03–1.09). In contrast, tracts with a higher percentage of Black (OR = 0.92; 95 percent CI: 0.88–0.96) and Hispanic/Latino residents (OR = 0.90; 95 percent CI: 0.84–0.96) had lower odds of having at least one premium cigar retailer. Adjusted models including all sociodemographic characteristics were not fit due to multicollinearity.