Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2019 May 1.
Published in final edited form as: Health Place. 2018 Apr 3;51:151–157. doi: 10.1016/j.healthplace.2018.03.009

The role of travel distance and price promotions in tobacco product purchase quantity

Nathan J Doogan 1,*, Sarah Cooper 1, Amanda J Quisenberry 1, Theodore M Brasky 1,3, Christopher R Browning 2, Elizabeth G Klein 1, Alice Hinton 1, Haikady N Nagaraja 1, Wenna Xi 1, Mary Ellen Wewers 1
PMCID: PMC5964010  NIHMSID: NIHMS957921  PMID: 29625358

Abstract

Introduction

Rural Americans are particularly vulnerable to tobacco use. For example, cigarette use is higher in rural areas than urban areas and the rate of decline is lower. Tobacco purchasing decisions, such as unit quantity purchased, may vary by geography and purchase related factors and are likely to affect tobacco use behavior. Therefore, explanation of variation in tobacco purchase quantity by factors associated with rural vulnerability and factors that fall under the regulatory scope of the Tobacco Control Act (TCA) of 2009 could be of value to regulatory proposals intended to equitably benefit public health. We examined whether (1) distance from home to the nearest tobacco outlet, which is larger on average in rural areas, and (2) price promotion use, which could be regulated through the TCA, and (3) their interaction explain variation in purchase quantity for cigarette and snuff purchases.

Methods

Our sample included 54 combustible tobacco users (298 purchase events) and 27 smokeless tobacco users (112 purchase events), who were asked to report all tobacco purchases on a smartphone application. We used an ecological momentary assessment methodology to collect data about tobacco users’ purchasing patterns, including products, quantity purchased, and use of price promotions. A parent cohort study provided relevant data for home-outlet distance calculation and covariates. Our analysis examined associations between our outcome—purchase quantity per purchase event—and distance from participant’s home to the nearest outlet, whether a price reducing promotion was used, and the interaction of these two factors.

Results

Combustible users showed an increased cigarette pack purchase quantity if they lived further from an outlet and used a price promotion (i.e., an interaction effect; RR=1.70, 95% CI [1.11, 2.62]). Smokeless users purchased more units of snuff when they used price promotions (RR=1.81, 95% CI [1.02, 3.20]).

Conclusions

Regulatory action that imposes restrictions on the availability or use of price promotions could alter the purchasing behavior of rural Americans in such a way that makes it easier to reduce tobacco use or quit. Such action would also restrict flexibility in the price of tobacco products, which is known as a powerful tobacco control lever.

INTRODUCTION

Rural areas are characterized by poorer health behaviors and poorer health outcomes compared with non-rural areas of the US.1–4 Tobacco use, particularly cigarette and smokeless tobacco (SLT) use, are higher in rural areas.5 The geographic disparity in the use of the most harmful tobacco product—cigarettes—appears to be growing, even after controlling for numerous sociodemographic characteristics of individuals.6 Thus, other contextual factors may explain the growing disparity. For example, less support for tobacco control efforts and the difficulty of enforcing policies uniformly in sparsely populated rural areas with relatively low resources7,8 could increase disparity. In general, rural residents have lower incomes than their urban counterparts,6 which may mean they are more likely to be targeted with price promotions from the tobacco industry.9 The population sparsity in rural areas could have additional impacts. Diffuse geographic distribution of tobacco retail outlets, resulting in longer travel times for consumers to obtain tobacco products, could alter purchasing patterns, and subsequent tobacco use behavior. This possibility has not been studied.

Why do purchasing patterns matter? The literature provides evidence that tobacco consumers use their tobacco products at a lower rate if they have a smaller supply on hand. According to marketing researchers, consumers of “vice goods” with concern about the long-term consequences of their continued use exhibit self-regulatory purchasing behaviors to control consumption behaviors.10,11 This research has included goods ranging from cookies and potato chips to alcohol and cigarettes.10 Similar findings have emerged within a behavioral economics framework; Marti and Sindelar found that some consumers will forgo quantity discounts and instead purchase premium-priced packs of 10 cigarettes (instead of 20) to constrain their product use.12 This behavior has been referred to as pre-commitment,13 as in the consumer pre-commits to a certain limit on tobacco use until the next purchase event by purchasing in a limited quantity despite an increased cost. This literature attributes consumer purchase quantity to concern about continued consumption. However, it seems to imply that purchasing patterns can affect use behavior regardless of health concerns. Theoretically, the limited supply and the increased transaction cost of renewing the supply are the factors altering consumption. These factors force the addicted consumer, who by definition of addiction has a reduced level of self-control over his or her use of the product, to think more carefully about use. Thus, factors that reduce purchase quantity may have the side effect of encouraging more judicious use of the product, regardless of the reason for limited quantity purchases.

Conversely, individuals who purchase in larger quantities may exhibit relatively high rates of product use because they lack the same supply limit as those who purchase in smaller quantities. Such consumers may purchase in larger quantities because they lack the motivation to restrict their use. Even if the consumers do have a motivation to restrict use, they may have been enticed with quantity discounts, which they deemed valuable enough to overwhelm a health-related motivation. These consumers could be seen to be committing to the product, or at least they are not committing to controlling their product use through their purchasing decisions. Marketing research suggests that consumers who have a larger supply of a product on hand may use it more liberally. This could be because they perceive the unit cost to be lower either because it is lower due to quantity discounts, or because of a misperception of cost due to an increased supply.14,15 Indeed, economic tobacco control literature clearly shows an inverse relationship between price and consumption.16 In sum, the reviewed literature provides ample evidence that purchase quantity can affect use behavior, which raises the importance of identifying factors that are related to tobacco product purchase quantity. However, this area of tobacco control is not well studied.

One factor that could be related to purchase quantity is geographic accessibility of products. The density of tobacco retail outlets is relatively low in rural areas. As such, rural tobacco users are more likely to incur a higher average transaction cost than urban consumers. These consumers may therefore have a larger incentive to minimize their transaction costs than consumers who live close to an outlet. Given previous research linking purchase quantity to tobacco use behavior,12,13 larger purchase quantities by individuals residing relatively far from a tobacco retail outlet could imply that these consumers have a decreased incentive to limit their tobacco use. Such a finding could complement ongoing research aimed at identifying the factors associated with increased tobacco use in rural areas of the United States relative to non-rural areas.5,6,17

Tobacco price promotions (e.g., discount coupons and reduced price multi-pack buys) may also be an important factor in the context of purchasing behavior. Cornelius and colleagues found that between 2002 and 2011, multi-pack purchases increased in popularity, likely because they balance the high total entry cost of carton-buying and the premium per-pack cost of single-pack purchases.18 The US Food and Drug Administration (FDA) has authority, through the Family Smoking Prevention and Tobacco Control Act (TCA) of 2009, to regulate the sale and promotion of tobacco products (FSPTCA, 906(d)), which could include limitations on price promotions that reduce price,16 increase purchase quantity, or both. Therefore, research on the associations between price promotions and purchase quantity could be valuable to those generating tobacco control policies within the FDA’s regulatory scope. With regard to relatively high rural tobacco product use, price promotions may be a particularly attractive way to offset travel costs involved in purchasing tobacco.

This study examined variation in tobacco purchase quantity of a subcohort of tobacco consumers living in urban and Appalachian rural areas of Ohio. Using repeated observations of purchase events collected from tobacco users enrolled in an ecological momentary assessment (EMA) study, the purpose of the reported analysis was to test the associations of contextual and purchase related factors with the quantity of tobacco products purchased. Specifically, one explanatory focus was the distance between a participant’s home and the nearest tobacco retail outlet. While this work is broadly premised on a disparity in rural tobacco use, our conceptual framework implies that travel distance is an underlying factor that determines the transaction cost, and therefore the extent to which bulk purchasing could offset that cost. We therefore do not focus on rural residential status, but more directly on distance. Our second explanatory focus was the use of price reducing promotions, which may affect purchase quantity directly by requiring a multi-pack purchase or indirectly by reducing the price of single packs. Additionally, the combination of a large travel distance and a price promotion could provide a uniquely strong incentive to purchase in larger quantities. We hypothesized that distance, the use of a price promotion during purchase, and the interaction between the two would be positively associated with purchase quantity.

METHOD

Sampling & Design

The sample was drawn from a cohort of rural and urban adult tobacco users established as part of a parent study entitled “Tobacco User Adult Cohort” (TUAC; P50CA180908). The study was designed for surveillance and to examine long-term dynamics of tobacco-related variables in participants classified as one of four types of tobacco users—exclusive combustible, exclusive SLT, or exclusive electronic cigarette, and dual use (described in more detail below)—living in urban and rural areas of Ohio. The TUAC study is a 36-month longitudinal prospective cohort design, which uses face-to-face interviews about tobacco use, consumption patterns, cognitive and affective factors, and purchasing factors. The investigators began enrollment for the cohort in October of 2014 using an address-based random sampling design. Details of the study and a description of the cohort can be found elsewhere.19

The EMA study reported here was designed to collect data about tobacco purchasing factors among a subset of the TUAC cohort. The study used a special application installed on a smartphone for data collection in the context of participants’ normal daily routines. Enrolled participants agreed to participate for 7–10 days, once per year, for a total of three years. At present, two of the three years of data collection are complete, and this study utilizes the data from both.

Participant sampling for the EMA study involved a quota sampling scheme and was done on a rolling basis. If a participant was eligible, he or she was approached about enrollment in the EMA study just prior to his or her upcoming six- or twelve-month TUAC study follow-up meeting. Eligibility was determined by whether the study group to which the TUAC participant would belong (described in the next paragraph) had not yet reached its quota.a If the participant was eligible, the interviewer provided a brief description of the EMA study during a scheduling phone call for the TUAC follow-up. The EMA study description included the requirement to: 1) carry a provided smart phone for 7–10 days; 2) report all tobacco purchases; and 3) receive several randomly timed daily requests for responses to tobacco related questions.b An incentive of $50 for enrolling was also described on the phone call along with ways to possibly receive additional compliance incentives. For participants expressing interest in enrolling in the EMA study, the interviewer scheduled an individual face-to-face meeting 7–10 days prior to the TUAC follow-up interview to drop off the smartphone and train the participant in the use of the phone and the installed application. A participant’s EMA study period therefore began 7–10 days prior to the TUAC follow-up interview, and ended at the follow-up interview.

In total, 233 TUAC participants were approached about the EMA study during scheduling calls, 190 agreed during the call to participate, and 184 (75.4% of those approached) enrolled. Those enrolled were classified into one of four groups based on product type use at their baseline TUAC interview. The study participants reported exclusive use of (1) combustible, (2) SLT, or (3) e-cigarette products currently every day or some days of the week, or some combination of the three previous classes, (4) dual use. However, participants could have changed their product use by the time they were enrolled in the EMA study (6 or 12 months after baseline). For analyses, we therefore reclassified participants according to their user type status at the follow-up interview that occurred immediately after the EMA study period. We briefly describe similarities and differences between the TUAC cohort and the EMA sample below in the results section.

For the analysis reported here, we focused on cohort members who were exclusive combustible (n=67) or exclusive SLT (n=38) users during at least one of the observation points (year 1 or year 2 of the EMA study) and with complete data for at least one purchase during that period. We excluded dual users to avoid complication with the purchase of multiple types of products, and E-cigarette users due to limited purchase data from this group (further described in following paragraph). If a participant was not an exclusive combustible or SLT user for one of the EMA years, the participant’s data for that year only was excluded. We excluded purchase data for purchases that occurred outside of the state of Ohio (by self-report) to avoid issues of varying price exposures due to significant excise tax differences between states. We excluded one SLT user that reported a PO-Box address, precluding the outlet distance calculations. These exclusions reduced the groups to 54 combustible users who reported 298 purchase events, and 27 SLT users who reported 112 purchase events.

Our analysis did not include exclusive E-cig users for several reasons. First, the purchase questionnaire requested e-liquid units in milliliters, but preliminary examination of the data suggests that some participants may have reported in ounce units without giving the units, and we have observed products in the field that do not report volume in milliliters; we therefore doubt the comparability of the quantity data for e-cig users across purchases. Second, the number of purchases by E-cigarette users was relatively low, possibly because e-liquid is frequently purchased in bulk. Finally, Ohio does not require licenses for stores selling only e-cigarettes and related supplies, which made it much more difficult to calculate distance to the nearest outlet. Most dual users in the TUAC used E-cigarettes as one of their products, further justifying the exclusion of dual users from this analysis.

Procedure

Study personnel provided all enrolled participants with a smart phone. A trained interviewer instructed the participant in critical aspects of the phone’s use, such as powering on and off, daily charging, opening and using the simple specially designed EMA application, and responding to phone questionnaires.

The interviewer asked participants to report all tobacco related purchases as soon as possible after making one during the EMA study period. To limit participant burden, we did not ask participants to report all retail store visits regardless of tobacco purchase; therefore our data do not include store visits in which zero tobacco products were purchased. Once a purchase was made, the participant was to open the phone application and press the icon labeled “Report a purchase.” The application then presented the participant with a set of purchase-related questions that could be answered immediately and directly on the phone application.

Measures

Dependent variable: Purchase quantity

When reporting a purchase event, participants indicated the type and unit quantity of the purchased products. The dependent variable in our analyses was the purchase quantity of either cigarette packs or snuff cans (the predominant product of use by SLT users in this sample) that was purchased during a single purchase event. By design, each purchase event involved the purchase of at least one tobacco product; rarely among exclusive combustible and SLT users, the purchased product was neither cigarettes nor snuff, in which case the target product purchase quantity was zero. We excluded cases involving zero units of the target product due to the design which did not require participants to report store visits unless they made a tobacco purchase. Purchases of cigarette cartons were counted as 10 packs of cigarettes.

Independent variables

We computed home-outlet distance as the distance in miles to the nearest outlet by comparing the coordinates of the participant’s home, available from the TUAC study, and the coordinates of all licensed tobacco retail outlets in Ohio, available from Ohio county auditors’ offices. The distance in miles was computed using the haversine formula. Participants self-reported price promotion use when reporting a tobacco purchase. The EMA purchase questionnaire included an item asking participants, “Did you use any price promotions when purchasing [product]” for each product type that the participant reported purchasing; price promotion use was coded either one (a promotion was used) or zero.

Covariates

Our models also included several covariates to control for participant characteristics that may vary according to home-outlet distance or price promotion use, and which may also be related to purchase quantity. We used participant addresses from the TUAC study to geo-locate them and assign them a status as either urban or rural according to the Census Bureau’s definition. Participants living in an urbanized area or urban cluster were coded as urban, and the remaining participants were coded as rural. Because rural status is correlated with distance to the nearest outlet, it is included as a covariate to separate the effect of rural, which may be representative of cultural or policy factors, from the specific effect of distance from home to a retail outlet. We calculated alternative tobacco product purchase, a variable that identifies whether a purchase involved multiple tobacco product types to control for a difference in the purchase quantity of the target product—either cigarettes or snuff—when a different type of tobacco product is also purchased. Finally, we included the indicator variable year-2 to control for unobserved factors affecting purchase quantity from the first study-year to the second.

Price is frequently cited as having a significant influence on consumption. However, we have little reason to believe that the prices our participants were exposed to varied among participants who all lived in the state of Ohio at the time of the study, and therefore experienced the same state and federal cigarette taxes. Some participants lived near a border to a state with significantly lower cigarette excise taxes. We excluded purchases that occurred across a state border, as reported by participants.

At the follow-up interview for the larger TUAC study, data was collected as usual, from which we derived additional person-level variables that could be confounders in our analyses. These variables included the following: age, the participant’s age in years; female, the self-reported gender; 30-day quit intention, a self-report of an intention to quit using tobacco in the next 30 days; household income < $15k, an indicator of a household income below $15,000 per year; education < HS, an indicator of less than a high school education; and use within 30 min of waking, an indicator of high nicotine dependence.20 All of these, except age, were scored such that a value of one meant ‘yes,’ and zero ‘no.’

Data Analysis

We first examined descriptive statistics of all variables according to user type—exclusive combustible and exclusive SLT. We further examined key independent variables broken down by rural status. This analysis is summarized in Table 1.

Table 1.

Descriptive statistics for exclusive combustible and exclusive smokeless users and their respective purchases of cigarettes and snuff.

Participant Characteristics Combustible (n=54) Smokeless (n=27)
 Dichotomous variables Percent Percent
  African American 26% 0%
  Female 57% 0%
  Income < $15K 31% 7%
  Education < high school 6% 4%
  30-day quit intention 9% 4%
  Use within 30 min of waking 65% 33%
 Continuous variables Mean (SD) Mean (SD)
  Age (years) 39.35 (12.82) 37.89 (11.48)
  Home-outlet distance (miles) 0.44 (0.75) 0.54 (0.71)
   Urban home-outlet distance 0.25 (0.21) 0.25 (0.27)
   Rural home-outlet distance 0.71 (1.10) 0.92 (0.91)
Purchase Characteristics Cigarettes (n=298) Snuff (n=112)
 Dichotomous variables Percent Percent
  Convenience store 79% 88%
  Grocery store 9% 6%
  Tobacco shop 5% 4%
  Alt. tobacco product purchase* 4% 3%
  Price promotion use 18% 21%
   Urban price promotion use 18% 21%
   Rural price promotion use 20% 21%
 Continuous variables Mean (SD) Mean (SD)
  Purchase quantity (packs or cans) 1.45 (0.95) 2.18 (2.08)
   Urban purchase quantity 1.31 (0.54) 1.73 (1.31)
   Rural purchase quantity 1.66 (1.33) 2.79 (2.71)
Open in a new tab
*

Purchasing another product in addition to the target product (cigarettes or snuff).

For multivariable analyses, we modeled purchase quantities during each purchase event using a truncated Poisson regression model—truncated below onec. Because some participants made multiple purchases during the EMA study, we included a random intercept to account for within-person correlation in purchase quantity. To isolate the within-person effect of using a price promotion, we used a “within-between” specification of the random effects model described similarly in multiple papers21–23 and a multilevel data analysis text by Snijders and Bosker.24 In this specification, the independent variable is decomposed into two components. The first component is a person-specific mean (i.e., the proportion of the individual’s purchases that included a price promotion) which allows for the estimation of the association between an individual’s overall tendency to use price promotions and purchase quantity (price promotion—between); this variable is included in the model as a covariate. Of focus to this study is the second component, an indicator of whether or not a price promotion was used during a specific purchase event (coded as 1 if yes, and 0 if no) with the person-specific mean (price promotion—between) subtracted from it. This second component allows for the estimation of the within-person change in purchase quantity when an individual uses a price promotion compared with when he or she does not, averaged over the sample. We therefore refer to it as price promotion—within.

We built two separate models, one of cigarette purchase quantities among exclusive combustible users only and another of snuff purchase quantities among exclusive SLT users only. We chose not to pool purchase events across both product types because a pack of cigarettes and a can of snuff are sized differently and have different costs, which may bear on the structure of multi-pack offers and on how consumers think about purchasing.

Both models included covariates and the primary independent variables, home-outlet distance and price promotion—within. We first fit models that also included the outlet distance by price promotion—within interaction; if the interaction was not significant, we removed it and re-fit the model. We centered the minimum outlet distance variable at the minimum observed value for each model (cigarettes: 0.04 miles; snuff: 0.06 miles), which allowed the coefficient of the price promotion—within variable to be interpretable as the effect of promotion use at the shortest observed distance from an outlet (as opposed to a less meaningful distance of zero miles). Due to the small sample size and lack of variability in some covariates, the snuff purchase model excluded covariates with little or no variability in the SLT sample—gender, race, education, quit intention, and alternative product purchase. Model coefficients were exponentiated, which resulted in a rate-ratio interpretation (i.e., a multiplicative change in purchase quantity associated with a one unit change in the independent variable), and 95% confidence intervals were also reported on the rate-ratio scale.

Finally, when a significant interaction effect was retained, interpretation of model coefficients is slightly more complicated; moreover, rate-ratio interpretations can be somewhat misleading in truncated modelsd. Therefore, instead of interpreting coefficients directly, we generated predictions from the fitted models on the scale of the outcome variable (purchase quantity), which incorporates the zero-truncated implications of the model. We used the predictions to calculate the ratio of mean purchase quantity between relevant conditions and report these in the text. We also visualized the predictions in Figures 1 and 2. Predictions were generated for the two levels of promotion use—within and for various values of home-outlet distance; all other variables were held constant at mean or modal values for continuous and dichotomous covariates, respectively. Despite the complications involved in interpretation of the coefficients, their 95% confidence intervals are still useful for statistical significance testing; when an interval does not contain zero, we consider the coefficient to be statistically significant. Models were fit and predictions were generated using the AD Model Builder (R package name: “glmmADMB”)25,26 in R.27

Figure 1.

Figure 1

Open in a new tab

Purchase quantities predicted by the cigarette purchase model for a range of outlet distances. It compares the within-person difference in purchase quantity given promotion use/non-use.

Figure 2.

Figure 2

Open in a new tab

Purchase quantities predicted by the snuff purchase model for a range of outlet distances. It compares the within-person difference in purchase quantity given promotion use/non-use.

RESULTS

Table 1 presents descriptive statistics of the sample of participants and the purchases made during their respective 7–10 day data collection periods. Among the 54 exclusive combustible users and the 27 exclusive smokeless users in the study, most were of European American descent, had a household income of greater than $15,000 per year, were at least high-school educated, and reported no intention to quit using tobacco in the next 30 days. The majority of combustible users were female (57%), and none of the SLT users were female. Combustible users were twice as likely as SLT users to report tobacco use within 30 minutes of waking. The mean age of combustible users was 39.35 years, and the mean age of SLT users was 37.89 years. The mean home-outlet distance was 0.44 miles (SD=0.75, min=0.04, max=4.27) among combustible users, and 0.54 miles (SD=0.71, min=0.06, max=4.24) among SLT users. As expected, the mean home-outlet distance in miles was larger for participants in rural areas (combustible: 0.71, SLT: 0.92) than those in urban areas (combustible: 0.25, SLT: 0.25).

Regarding the above characteristics (except home-outlet distance and purchase quantity, which were not calculated for the entire TUAC), the EMA study subcohort was quite similar to its parent, the TUAC.19 Two notable exceptions were that the EMA subcohort contained a smaller proportion of participants with an education level of greater than high-school in both user type groups (EMA: [combustible: 6%, SLT: 4%], TUAC: [combustible: 15%, SLT: 7%]19), and EMA SLT users were somewhat less likely to report product use within 30 minutes of waking (EMA: 33%, TUAC: 40%19).

Among the 298 purchase events by combustible users and 112 by SLT users, the majority were made in convenience stores, simultaneous purchase of multiple product types was very low, and purchases that involved the use of a price promotion were similar between the two groups (combustible: 18%; SLT: 21%). The mean number of units purchased per purchase event was higher among SLT users (M=2.17, SD=2.07) than combustible users (M=1.45, SD=0.95), and was higher for both products among rural (combustible: 1.66, SLT: 2.79) compared with urban (combustible: 1.31, SLT: 1.73) participants. The purchase quantity variable was positively skewed for both groups, as expected for a count variable with a small mean.

Table 2 presents the results of the fitted multivariable model of units purchased per purchase event for cigarettes and for snuff, separately. In the cigarette purchase model, the coefficient for the interaction of home-outlet distance and price promotion—within was positive and statistically significant (RR=1.70, 95% CI [1.11, 2.62]). The coefficients for price promotion—within (RR=0.98, 95% CI [0.56,1.72]) and home-outlet distance (RR=1.10, 95% CI [0.81,1.49]) themselves were not statistically significant. Taken together with the significant interaction effect, the centering of home-outlet distance at the minimum observed value, and that the referent group did not use price promotions, this implies that use of a price promotion has no relationship with purchase quantity at the minimum observed home-outlet distance, but the relationship becomes increasingly positive as home-outlet distance increases. The effects of price promotion—within and outlet distance are synergistic for cigarette purchasing. Figure 1 presents plotted model predictions clearly showing little difference between purchases with and without a price promotion at short home-outlet distances. However, as home-outlet distance increases, the purchase quantities of the two groups diverge rapidly. At the minimum observed home-outlet distance, the average purchase quantity when a price promotion is used versus not used is approximately equal. At home-outlet distances of 1, 2, and 3 miles, the purchase quantity associated with price promotion use is 25%, 75%, and 173% higher than that for non-use of promotions, respectively.

Table 2.

Multivariable truncated Poisson regression model results for cigarette (exclusive combustible users) and snuff (SLT users) purchase quantity, separately

Cigarettes Snuff
RR RR 95% CI RR RR 95% CI
lower upper lower upper
Intercept 0.31* 0.14 0.68 0.80 0.26 2.47
Home-outlet distance 1.10 0.81 1.49 0.84 0.44 1.62
Price promotion--within 0.98 0.56 1.72 1.81* 1.02 3.20
Outlet distance x Price promo--within interaction 1.70* 1.11 2.62
Price promotion--between 3.62* 1.40 9.40 0.37 0.04 3.23
Rural 1.15 0.65 2.03 2.71 0.87 8.38
Wave 2 fixed effect 0.60* 0.39 0.92 0.93 0.61 1.44
Alternative product purchase 1.54 0.57 4.17
Age 1.01 0.99 1.03 1.07* 1.01 1.12
Income < $15K 0.74 0.41 1.33 1.41 0.45 4.41
30-day quit intention 0.74 0.27 2.03
Use within 30m of waking 1.46 0.85 2.53 0.89 0.31 2.55
Education < HS 1.52 0.64 3.60
Female 1.94* 1.12 3.37
African American 0.54 0.26 1.10
Random Effects^ Est. Est.
Intercept Standard Deviation 0.45 1.12
Open in a new tab

RR: rate ratio

*

95% CI does not include one

^

Random effect standard deviation estimates are on the log rate ratio scale

In the snuff purchase model, the home-outlet distance by price promotion—within interaction was not significant and was therefore removed from the final model. In the final model, home-outlet distance was not statistically significant (RR=0.84, 95% CI [0.44, 1.62]). However, the use of a price promotion (price promotion—within) was associated with a greater purchase quantity (RR=1.81, 95% CI [1.02, 3.20]). Figure 2 contains plotted model predictions. Neither line shows a substantial change across home-outlet distance, consistent with the lack of statistical significance of the home-outlet distance coefficient. However, price promotion use is associated with an approximately 20% higher average purchase quantity.

DISCUSSION

This study was based on the premise that an individual’s personal supply of tobacco is associated with his or her use of the product. The purpose of the study was to examine the effect of two purchase factors, travel distance and the use of tobacco price promotions, on the number of tobacco product units purchased during a tobacco purchase event, and therefore their bearing on the user’s personal supply. We hypothesized that both factors and their interaction would be positively associated with purchase quantity. Overall, our hypotheses were partially supported, and the support varied by product type. Specifically, the cigarette purchase model results implied that the presence of both price promotion use and a relatively large distance between the participant’s home and the nearest tobacco outlet was associated with significantly and substantively larger purchases. Among smokeless users, we found no evidence that distance was an important factor, but price promotion use during a purchase significantly increased snuff purchase quantity overall. Thus, both factors play a role in tobacco purchase quantity, and therefore may also indirectly affect the consumer’s rate of product use.

Our findings are an important contribution to the literature concerning higher prevalence of tobacco use among US residents living in rural areas of the country.4–6 Travel distance for most goods in rural areas, including tobacco, is larger on average than that for individuals living in urban and suburban areas,28 a condition that our analyses suggest increases cigarette purchase quantity when price promotions are also used. If larger personal tobacco supplies discourage judicious use, then larger travel distance could be part of the explanation for higher cigarette use among rural residents. The reason our analyses did not show any travel distance effect for snuff purchases is unclear. Speculatively, it could be that the lower baseline unit cost of cans of snuff already encourage consumers to purchase in relatively large quantities, and that larger bulk purchases raise other purchase and use concerns, such as freshness of the product, for example. Alternatively, the travel distance and transaction cost framework that drove the development of this study could be incorrect, and should continue to be evaluated for its utility in this research. Regardless of whether the underlying mechanisms are understood, the data show an association between travel distance and cigarette purchase quantity in the context of promotion use, which is a cause for concern and should motivate additional research to replicate and nuance these findings. Moreover, snuff use is substantially more prevalent in rural areas, and our findings do show that snuff purchase quantity is positively associated with price promotion use regardless of the distance between a consumer’s home and a tobacco outlet. Thus, price promotions could be playing a role in maintaining the disparity of combustible and smokeless use between urban and rural areas.

Our findings are important from a tobacco regulatory perspective, particularly because of the apparent influence price promotions can have on purchase quantity of cigarettes and snuff. Because (1) we found a positive association between price promotions and purchase quantity in both cigarette and snuff purchases, (2) prior research implies an effect of personal supply on tobacco use behavior, and (3) the FDA can regulate the promotion and sale of tobacco products, our findings highlight a potential regulatory lever that may benefit public health generally, and reduce tobacco use disparities in rural areas. Additionally, our findings complement the large literature on the price elasticity of demand for cigarettes, which clearly shows that raising price reduces consumption.16,29,30 The FDA should seriously consider restricting the distribution and/or redemption of price reducing promotional material. We believe this will benefit the public health in several ways. First, it will limit price reducing strategies used by tobacco consumers. Second, our findings suggest it could have the effect of reducing an individual’s supply of tobacco, therefore increasing the saliency of managing the rate of consumption. Third, the synergy that we identified between travel distance to a tobacco retail outlet and price promotion use on cigarette purchasing implies that such a restriction could be an equity producing policy in regards to disparate rural tobacco use. And fourth, because price promotions are disproportionately targeted to9 and used by low income consumers,31,32 who also disproportionately use tobacco products,33 such a regulatory action could also reduce this serious and ubiquitous tobacco use disparity.

Our study has several notable limitations. The decision to report a tobacco purchase was the respondents’ and therefore our methods may have omitted purchases if the respondent forgot or decided for any reason to not input the data. Future work should consider mechanisms for proof-of-purchase that can facilitate rewarding purchase reports with incentives. All measures depended on self-report, and are therefore subject to inaccuracies associated with self-report data. Due to fairly small sample sizes, the statistical power of our analyses may have been insufficient to detect more modest associations that may deepen our understanding of what explains purchase quantity. Distance to the nearest tobacco outlet was defined as a function of the reported location of a participant’s home, and did not account for tobacco outlets that may have been present on travel pathways typically traversed by participants, such as a drive to work. In such a case, the importance of the distance from home to the nearest outlet might be diminished in the decision about purchase quantity. However, even when the purchase of tobacco is secondary to the purpose of an outing (e.g., travel to a worksite or grocery shopping), consumers may still be considerate of how far they would have to travel in the case that they needed to make a special trip for tobacco from home later; if that distance is large, they may choose to use the current opportunity to purchase in bulk. To further tease apart these possibilities, future studies can utilize geo-tracking data from cell phones to identify more precisely where and when tobacco is purchased and possibly other details of the individual or context that may moderate purchase quantity decisions, such as current craving for tobacco or the level of marketing exposure present at a particular retail outlet. Regardless of why promotion-using participants who live distant from tobacco outlets purchased in higher quantities, our data suggest that they do, and previous research suggests that this purchase behavior can affect use behavior.

In conclusion, this is the first study to ecologically examine tobacco purchase events. The study identified geographic context and tobacco price promotion as two purchase-related factors that appear to influence purchasing behavior and can be linked with previous research on how purchase behavior can affect tobacco use. Future research should replicate and expand this work in numerous ways to include more nuance in the hypothesized mechanisms by adding variations in or manipulating important demographic, social, and psychological dimensions. The generalizability of our findings to e-cigarette products also needs to be examined due to the growing use of these products. This area of research is understudied, but may carry the potential to significantly alter tobacco use and the associated health outcomes and disparities. Because this research can directly inform regulatory actions by the FDA, the policy impact of continued research in this area is potentially very high.

Highlights.

  • Tobacco purchase quantity, and therefore tobacco supply, affects tobacco use

  • The literature does not currently address factors explaining purchase quantity

  • Cigarette purchase quantity is explained by price promotion use and travel distance

  • Snuff purchase quantity is explained by price promotion use alone

  • FDA Regulation of price promotion distribution or use is recommended

Acknowledgments

FUNDING

This work was supported by the National Cancer Institute and the U.S. Food and Drug Administration (P50CA180908) and by the National Institutes of Health (UL1TR001070).

Footnotes

a

The quota was determined using a power analysis based on the research questions the full 3-year EMA study data are designed to answer, which are beyond the scope of this paper.

b

Responses to these items were not used in the reported study.

c

We also relaxed the equal mean and variance assumption of the Poisson model by fitting a truncated negative binomial model. The dispersion parameter indicated there was no over-dispersion after accounting for model predictors, and our inferences did not change.

d

Specifically, the rate ratio interpretation in the truncated Poisson model that we used refers to a change in the rate parameter for the underlying non-truncated distribution, and cannot be interpreted directly as a ratio of the mean number of purchases between the sample groups.

DECLARATION OF INTERESTS

All authors declare no conflicts of interest.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  • 1.Mansfield CJ, Wilson JL, Kobrinski EJ, Mitchell J. Premature mortality in the United States: the roles of geographic area, socioeconomic status, household type, and availability of medical care. Am J Public Health. 1999;89(6):893–898. doi: 10.2105/ajph.89.6.893. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.National Center for Health Statistics. Health, United States, 2011: With Special Feature on Socioeconomic Status and Health. Hyattsville, MD: 2012. [PubMed] [Google Scholar]
  • 3.Pettit ML, Nienhaus AR. The Current Scope of Health Disparities in the U.S.: A Review of Literature. Health Educ. 2010;42(2):47–55. [Google Scholar]
  • 4.Rural Health Reform Policy Research Center. The 2014 Update of the Rural-Urban Chartbook. Rural Health Reform Policy Research Center; 2014. https://ruralhealth.und.edu/projects/health-reform-policy-research-center/pdf/2014-rural-urban-chartbook-update.pdf. [Google Scholar]
  • 5.Roberts ME, Doogan NJ, Kurti AN, et al. Rural tobacco use across the United States: How rural and urban areas differ, broken down by census regions and divisions. Health Place. 2016;39:153–159. doi: 10.1016/j.healthplace.2016.04.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Doogan NJ, Roberts ME, Wewers ME, et al. A growing geographic disparity: Rural and urban cigarette smoking trends in the United States. Prev Med. 2017 doi: 10.1016/j.ypmed.2017.03.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.National Cancer Institute. Evaluating ASSIST: A Blueprint for Understanding State-Level Tobacco Control. Tobacco Control Monograph No. 17. NIH Pub. No. 06-6058. Bethesda, MD: U.S. Department of Health and Human Services, National Institutes of Health, National Cancer Institute; 2006. [Google Scholar]
  • 8.Stillman FA, Hartman AM, Graubard BI, Gilpin EA, Murray DM, Gibson JT. Evaluation of the American Stop Smoking Intervention Study (ASSIST): a report of outcomes. J Natl Cancer Inst. 2003;95(22):1681–1691. doi: 10.1093/jnci/djg098. [DOI] [PubMed] [Google Scholar]
  • 9.Caraballo RS, Wang X, Xu X. Can you refuse these discounts? An evaluation of the use and price discount impact of price-related promotions among US adult smokers by cigarette manufacturers. BMJ Open. 2014;4(6):e004685. doi: 10.1136/bmjopen-2013-004685. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Wertenbroch K. Consumption Self-Control by Rationing Purchase Quantities of Virtue and Vice. Mark Sci. 1998;17(4):317–337. doi: 10.1287/mksc.17.4.317. [DOI] [Google Scholar]
  • 11.Jain S. Marketing of Vice Goods: A Strategic Analysis of the Package Size Decision. Mark Sci. 2011;31(1):36–51. doi: 10.1287/mksc.1110.0657. [DOI] [Google Scholar]
  • 12.Marti J, Sindelar J. Smaller Cigarette Pack as a Commitment to Smoke Less? Insights from Behavioral Economics. Plos One. 2015;10(9):e0137520. doi: 10.1371/journal.pone.0137520. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Ladouceur R, Blaszczynski A, Lalande DR. Pre-commitment in gambling: a review of the empirical evidence. Int Gambl Stud. 2012;12(2):215–230. doi: 10.1080/14459795.2012.658078. [DOI] [Google Scholar]
  • 14.Wansink B. Can package size accelerate usage volume? J Mark. 1996;60(3):1. [Google Scholar]
  • 15.Wansink B, Cheney MM. Super Bowls: Serving Bowl Size and Food Consumption. JAMA. 2005;293(14):1723–1728. doi: 10.1001/jama.293.14.1727. [DOI] [PubMed] [Google Scholar]
  • 16.Chaloupka F, Warner KE. The economics of smoking. In: Jha P, Chaloupka F, editors. Tobacco Control in Developing Countries. 1. Oxford; New York: Oxford University Press; 2000. [Google Scholar]
  • 17.Roberts ME, Doogan NJ, Stanton CA, et al. Rural Versus Urban Use of Traditional and Emerging Tobacco Products in the United States, 2013–2014. Am J Public Health. 2017;107(10):1554–1559. doi: 10.2105/AJPH.2017.303967. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Cornelius ME, Driezen P, Hyland A, Fong GT, Chaloupka FJ, Cummings KM. Trends in cigarette pricing and purchasing patterns in a sample of US smokers: findings from the ITC US Surveys (2002–2011) Tob Control. 2015;24(Suppl 3):iii4–iii10. doi: 10.1136/tobaccocontrol-2013-051376. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Brasky TM, Hinton A, Doogan NJ, et al. Characteristics of the Tobacco User Adult Cohort in Urban and Rural Ohio. Tob Regul Sci. 2018;4(1):614–630. doi: 10.18001/TRS.4.1.8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Heatherton TF, Kozlowski LT, Frecker RC, Fagerström KO. The Fagerström Test for Nicotine Dependence: a revision of the Fagerström Tolerance Questionnaire. Br J Addict. 1991;86(9):1119–1127. doi: 10.1111/j.1360-0443.1991.tb01879.x. [DOI] [PubMed] [Google Scholar]
  • 21.Dieleman JL, Templin T. Random-Effects, Fixed-Effects and the within-between Specification for Clustered Data in Observational Health Studies: A Simulation Study. PLOS ONE. 2014;9(10):e110257. doi: 10.1371/journal.pone.0110257. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Bell A, Jones K. Explaining Fixed Effects: Random Effects Modeling of Time-Series Cross-Sectional and Panel Data. Polit Sci Res Methods. 2015;3(1):133–153. doi: 10.1017/psrm.2014.7. [DOI] [Google Scholar]
  • 23.Mundlak Y. On the Pooling of Time Series and Cross Section Data. Econometrica. 1978;46(1):69–85. doi: 10.2307/1913646. [DOI] [Google Scholar]
  • 24.Snijders T, Bosker R. Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modeling. 2. SAGE Publications Ltd; 2011. [Google Scholar]
  • 25.Fournier DA, Skaug HJ, Ancheta J, et al. AD Model Builder: using automatic differentiation for statistical inference of highly parameterized complex nonlinear models. Optim Methods Softw. 2012;27:233–249. [Google Scholar]
  • 26.Skaug H, Fournier D, Bolker B, Magnusson A, Nielsen A. Generalized Linear Mixed Models Using “AD Model Builder”. 2016. [Google Scholar]
  • 27.R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2017. http://www.R-project.org/ [Google Scholar]
  • 28.Luke DA, Ribisl KM, Smith C, Sorg AA. Family Smoking Prevention and Tobacco Control Act: Banning Outdoor Tobacco Advertising Near Schools and Playgrounds. Am J Prev Med. 2011;40(3):295–302. doi: 10.1016/j.amepre.2010.11.018. [DOI] [PubMed] [Google Scholar]
  • 29.Contreary KA, Chattopadhyay SK, Hopkins DP, et al. Economic Impact of Tobacco Price Increases Through Taxation: A Community Guide Systematic Review. Am J Prev Med. 2015;49(5):800–808. doi: 10.1016/j.amepre.2015.04.026. [DOI] [PubMed] [Google Scholar]
  • 30.Gallet CA, List JA. Cigarette demand: a meta-analysis of elasticities. Health Econ. 2003;12(10):821–835. doi: 10.1002/hec.765. [DOI] [PubMed] [Google Scholar]
  • 31.Licht AS, Hyland AJ, O’Connor RJ, et al. Socio-Economic Variation in Price Minimizing Behaviors: Findings from the International Tobacco Control (ITC) Four Country Survey. Int J Environ Res Public Health. 2011;8(1):234–252. doi: 10.3390/ijerph8010234. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Xu X, Pesko MF, Tynan MA, Gerzoff RB, Malarcher AM, Pechacek TF. Cigarette Price-Minimization Strategies by U.S. Smokers. Am J Prev Med. 2013;44(5):472–476. doi: 10.1016/j.amepre.2013.01.019. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Centers for Disease Control and Prevention (CDC) [Accessed November 9, 2016];Current Cigarette Smoking Among Adults — United States, 2005–2014. 2015 :1233–1240. http://www.cdc.gov/mmwr/preview/mmwrhtml/mm6444a2.htm?s_cid=mm6444a2_w.

RESOURCES