Expanding on cross-price elasticity: Understanding tobacco product demand and substitution from the cross-price purchase task

Abstract

We examine whether cigarettes serve as substitutes for electronic nicotine delivery systems (ENDS) among ENDS users and demonstrate methodological extensions of data from a cross-price purchase task to inform policies and interventions. During a clinical laboratory study, n = 19 exclusive ENDS users and n = 17 dual cigarette/ENDS users completed a cross-price purchase task with cigarettes available at a fixed price while prices of own-brand ENDS increased. We estimated cross-price elasticity using linear models to examine substitutability. We defined five additional outcomes: nonzero cross-price intensity (purchasing cigarettes if ENDS were free), constant null demand (not purchasing cigarettes at any ENDS price), cross-product crossover point (first price where participants purchased more cigarettes than ENDS), dual-demand score (percentage of prices where both products were purchased), and dual-use break point (minimum relative price to force complete substitution). The cross-price elasticity results indicated that cigarettes could serve as substitutes for ENDS among ENDS users on average, but this average effect masked substantial heterogeneity in profiles of demand (here, a measure of the drug’s reinforcement potential). Policies and regulations that increase ENDS prices appear unlikely to steer most exclusive ENDS users toward cigarette use, as most would not purchase cigarettes at any ENDS price, but they could prompt some dual users to substitute cigarettes completely while others remain dual users. This heterogeneity in consumer responses suggests additional indices of cross-product demand are useful to characterize the anticipated and unanticipated effects of tobacco price policies more fully.

INTRODUCTION

Perhaps no public health policy strategy has proven as effective in reducing smoking as levying taxes on combustible tobacco cigarettes to increase their price and suppress demand (Hoffman & Tan, 2015), a construct that has been operationalized as the amount of a given product that someone opts to consume at its current price and a marker of the product’s reinforcing potential (Bickel et al., 2000). Given these successes, policy makers in numerous contexts have extended this approach to electronic nicotine delivery systems (ENDS; CDC, 2023a). However, several factors complicate ENDS price policies. The potential for higher prices to dissuade tobacco-naïve youth from using ENDS must be weighed against the possibility that those prices could deter adult cigarette smokers from switching to ENDS, a potentially lower harm product. Another relatively understudied complication is whether increased ENDS prices might prompt current ENDS users to substitute cigarettes for ENDS. This economic substitution effect—smoking more cigarettes—poses concerns given data suggesting that dual use is associated with greater biomarkers of inflammation, respiratory symptoms, and risk for cardiovascular disease relative to exclusive use of cigarettes or ENDS (Cobb et al., 2021; Osei et al., 2019; Reddy et al., 2021; Stokes et al., 2021) and may be associated with greater dependence (Snell et al., 2020; Strong et al., 2017). Consequently, the public health improvement resulting from taxing ENDS hinges in part on the economic relationship between cigarettes and ENDS.

Questions remain regarding how the demand for cigarettes responds to increased ENDS prices. For exclusive ENDS users, the economic relationship between ENDS and cigarettes may take one of three forms (Figure 1): the alternative product (cigarettes) could function as a substitute good (i.e., as the price of ENDS increases, so does demand for cigarettes), a complement good (i.e., as the price of ENDS increases, demand for cigarettes falls), or as an independent good (i.e., changes in the price of ENDS do not affect cigarette demand). Recent analyses examining the potential substitutability between cigarettes and ENDS using retail sales data have disagreed, with some studies suggesting a substitution relationship (Cotti et al., 2020; Saffer et al., 2020), and others indicating that cigarettes and ENDS function as complements (Abouk & Adams, 2017; Cotti et al., 2018) or independent goods (Allcott & Rafkin, 2020; Huang et al., 2014; Pesko et al., 2019).

FIGURE 1.

As the price of ENDS increase, cigarettes could function as a substitute good (increasing demand), a complement good (decreasing demand), or an independent good (no relationship)

As an alternative to traditional economic approaches, behavioral-economic methods can assess the economic relationship between two or more goods and predict possible unintended consequences under controlled experimental conditions (Bickel et al., 2017). For example, cross-price purchase tasks can assess demand for two tobacco products simultaneously (Johnson et al., 2017; Snider et al., 2017; Stein et al., 2018). In these tasks, participants are presented with a series of hypothetical scenarios in which a preferred tobacco product is available for purchase at several different price points and an alternative tobacco product is concurrently available at a fixed price. Participants report how many of each product they would purchase in each price scenario. Demand for the fixed-price alternative product as a function of the preferred product’s escalating price—cross-price elasticity—is the primary outcome of interest. Cross-price elasticity, a measure of the economic relationship between products, is often indexed by the average slope of a line relating the log of consumption for the alternative product to the log of the preferred product’s price. The resulting parameterized value obtained through linear modeling (versus a differentiated value obtained through nonlinear modeling) assumes that cross-price elasticity is constant across prices, an assumption that may not be met in practice. (For example, in a linear model where both demand for the alternative cigarettes and price of the preferred ENDS are log transformed, a positive cross-price elasticity estimate of X would suggest that demand for cigarettes will consistently increase by X% for every 1% increase in the price of ENDS, whereas a nonlinear model could capture whether cigarettes function as a substitute for ENDS at low ENDS prices but a complement at high ENDS prices.) However, linear modeling of cross-price elasticity is a common approach in the economic and behavioral literature given that such models are flexible, can be expanded to account for additional variables, and provide an easily interpretable parameter estimate indicating the rate and direction of change in consumption of the fixed-price alternative product in response to changes in the price of the preferred product (Gilroy et al., 2020; Gilroy, Waits, et al., 2022).

Still, more complete information on the level of demand for either product is needed to fully characterize how tobacco use behaviors may change in response to policies that alter the price and/or accessibility of tobacco products. This is particularly true of two products potentially differing in harms, such as cigarettes and ENDS, where quantifying the absolute change in demand for each and the likelihood of dual use is critical to advancing the public health intent of policy efforts. In this article, we build on the existing operant-demand literature to propose additional measures of cross-product demand that can be derived from the cross-price purchase task. The proposed outcomes were defined a priori, then measured and compared using experimental data from dual and exclusive ENDS users to highlight how data from cross-price purchase tasks can be leveraged to contextualize and extend cross-price elasticity estimates. The proposed outcomes are divided into three domains: demand for the alternative product, the strength of preference for the preferred product, and dual-use liability.

Demand for the alternative product

Cross-price elasticity is an index of price sensitivity between products and provides limited information about the absolute or independent demand for either product. However, demand for the alternative product in cross-price purchase tasks can provide useful information. For example, cross-price elasticities might suggest that both menthol and nonmenthol cigarettes substitute similarly for ENDS, but higher reported demand for menthol versus nonmenthol cigarettes could indicate that menthol cigarettes pose more significant risks for substitution or dual use.

Stein et al. (2018) introduced the measure of cross-price intensity, the quantity of the alternative product demanded when the preferred product is free, or the point at which the alternative product’s demand curve intersects the y-axis. If the preferred product is free while the alternative is not, opting to purchase any amount of the alternative product indicates a strong liking for that product, potentially implying an imperfect substitution relationship (Hursh & Roma, 2016) or potential for dual use (Stein et al., 2018). We transform cross-price intensity into a binary variable, nonzero cross-price intensity, representing whether or not a participant reported any demand for the alternative product when their preferred product was free (see Figure 2, Panel A; Bono et al., 2022). We propose a binary variable for three main reasons. First, under normal conditions, reporting any consumption of the alternative product when the preferred product is free is expected to be rare. Studies employing the cross-price purchase task—particularly those conducted in clinical laboratory settings—are likely underpowered for the most appropriate types of statistical models for analyzing continuous distributions with frequent zero values (e.g., zero-inflated Poisson models). Second, a binary variable allows for greater comparability across studies, which vary in prices and price frames (e.g., 10 puffs vs. 1 ml e-liquid) used in the cross-price purchase task. Third, whether a participant would use any amount of the alternative product if the preferred product were free may be more relevant than the absolute level of demand in some circumstances (e.g., when the two products have a drug interaction effect). This outcome may be particularly relevant to policy makers wishing to incentivize the use of a lower harm product by making that product free at the point of purchase, as is already the case with certain nicotine replacement therapies covered by health insurance: nonzero cross-price intensity signals that additional incentives could be necessary to promote uptake of that product.

FIGURE 2.

The same estimates for cross-price elasticity can mask differences in levels of demand for the alternative product. The dashed gray lines represent demand for ENDS (the preferred product), and the solid black lines represent demand for cigarettes (an alternative product) as prices for ENDS increase. In Panel A, the two black lines would produce the same cross-price elasticity estimates (i.e., the slopes are identical) but the baseline level of demand for the alternative product (i.e., the y-intercepts) differs. In Panel B, the black lines each indicate an independent relationship between the preferred and alternative products (i.e., a slope of zero). However, the line labeled constant null demand suggests zero interest in the alternative product, whereas the line labeled constant positive demand suggests constant interest in the alternative product.

Alternately, participants may report zero consumption across all prices, generating a null demand function (Stein et al., 2015) for the alternative product. This could occur if the product is unfamiliar or aversive to the participant (Stein et al., 2015) or if the experimental price of the alternative product is set too high to generate any demand response—all of which provide valuable feedback about the validity of the experimental design or the likelihood of substitution. Such information could be lost if only cross-price elasticity is reported, as null demand would produce a cross-price elasticity function indistinguishable from constant positive demand (Figure 2, Panel B). Each case has different policy implications: constant null demand for cigarettes could signal that raising the price of ENDS is unlikely to lead ENDS users to switch to cigarettes. In contrast, constant positive demand could suggest an underlying demand for cigarettes that could manifest in dual use or complete substitution at prices higher than those in the task. Therefore, we define constant null demand for the alternative product as a binary variable indicating whether or not a participant would purchase any amount of the alternative product in any price scenario.

Strength of preference for the preferred product

Cross-price purchase tasks typically instruct participants to consider their current income and savings while making purchase decisions. Given their budget constraints, participants are expected to maintain a relatively constant utility over the preferred product prices by varying their demand for both products. However, cross-price elasticity is estimated using the quantity demanded for the alternative product only, meaning that an alternative product could be considered a substitute (i.e., positive cross-price elasticity) even if demand for the alternative product never approaches demand for the preferred product or if demand for the preferred product does not change. Failing to examine the demand for the preferred product could undermine public health policy efforts. For example, policies that increase cigarette prices to encourage switching to ENDS could result in greater overall toxicant exposure if ENDS use increases without a corresponding decrease in cigarette use. Understanding the strength of the preference for the preferred product over the alternative could help contextualize cross-price elasticity findings.

Previous work reported the price at which average consumption of an alternative product exceeded average consumption of the preferred product (O’Connor et al., 2014). We adapt that measure to define the first price at which participants purchase a greater quantity of the alternative product than their preferred product (Figure 3; see the Supporting Information for additional examples). Echoing another behavioral economic task (Breland et al., 2020), we refer to this measure as the cross-product crossover point. Greater cross-product crossover points signal stronger loyalty to the preferred product. If the choice between products was based solely on each product’s relative prices, the cross-product crossover point should be approximately the same value as the fixed price used for the alternative product. Therefore, cross-product crossover points exceeding the fixed price suggest willingness to defend purchasing the preferred product even when doing so is more expensive, indicating that factors other than price drive product preferences or that individuals receive incomplete reinforcement from the alternative product (e.g., incomplete switching from ENDS to cigarettes could be due to lower nicotine delivery).

FIGURE 3.

The cross-product crossover point measures the strength of the demand for ENDS by indicating the price at which a participant begins predominantly using cigarettes

Dual-use liability

Cross-price purchase tasks involve reporting demand for both the preferred and alternative products across a range of prices, providing an opportunity to understand conditions leading to hypothetical dual use of the products. Here we define dual-use liability based on reporting nonzero demand for both products at a given price point.

We first propose a dual-demand score, expressed as the percentage of price points in a cross-price purchase task at which the participant would purchase both the alternative and preferred products (Figure 4). We suggest that a larger percentage of price points where demand for both products is occurring, and thus a larger dual-demand score, indicates a greater propensity toward dual use. Although this outcome is likely sensitive to the number and range of the price points used in the task and therefore may be difficult to compare across studies, the dual-demand score could allow for within-persons comparisons of dual-use liability across experimental conditions.

FIGURE 4.

The dual-demand score is the percentage of price points in a cross-price purchase task at which participants would purchase both products simultaneously. The shaded gray region of each panel indicates the range of price points at which both products are purchased. Panel A depicts a higher dual-demand score than Panel B.

Second, among those with a dual-demand score greater than zero, we defined the dual-use break point as the last price at which participants reported any demand for both products and began to use one product exclusively or ceased use of both products simultaneously (Figure 5). Like the break-point measure in own-price drug purchase tasks (MacKillop et al., 2008), the dual-use break point is the minimum price required to suppress dual use to zero. When the price of the alternative product is fixed at $1 in cross-price purchase tasks, the dual-use break point represents the minimum relative price at which the preferred product would need to be set to discourage dual use by forcing a complete product substitution. Relative prices below these thresholds are likely to encourage dual use or use of the preferred product alone. For example, as cigarette prices increase, smokers might report demand for both cigarettes and ENDS over several prices before the price of cigarettes becomes too high (e.g., $10 per 10 puffs) relative to the price of ENDS ($1 per 10 puffs) to support further cigarette consumption. To encourage ENDS use over exclusive or dual cigarette use, policy makers could adjust price-related policies to ensure that cigarettes cost 10 times more than ENDS per standardized unit.

FIGURE 5.

The dual-use break point represents the minimum ENDS price required to suppress dual use to zero. Each arrow indicates the dual-use break point for one of four hypothetical patterns of preferred product purchasing.

Objectives

This work employs a behavioral-economic task in a clinical laboratory setting to assess the substitution potential of cigarettes among ENDS users. We seek to understand how cigarette demand changes with ENDS price—a potential consequence of current and pending ENDS policies and regulations—and to demonstrate methodological extensions of cross-price purchase tasks that can provide additional insight regarding the influence of polices affecting tobacco product prices on dual-use behavior.

METHODS

Participants

Healthy adult ENDS users ages 21–55 were recruited from the Richmond, Virginia, area in 2019–2020 as part of a larger clinical laboratory experiment (Hoetger et al., 2022). Exclusive ENDS users used ≥ 1 ml ENDS liquid or approximately one pod/cartomizer containing ≥ 3 mg/ml nicotine per day for at least 3 months, with no past-month cigarette smoking. Dual cigarette/ENDS users used one product daily and the other product ≥ 3 days per week for the past three months or longer: either ENDS containing ≥ 3 mg/ml nicotine daily and cigarettes ≥ 3 days/week or cigarettes daily and ENDS containing ≥ 3 mg/ml nicotine ≥ 3 days/week. Nicotine use was verified by a semiquantitative urine cotinine result of “positive” (≥200 ng/ml; NicAlert, JANT Pharmacal Corporation, Encino, California) at screening. Participants were ineligible if they were planning to quit using nicotine or tobacco products in the next month or reported using any tobacco products other than cigarettes or ENDS weekly or more frequently.

Procedures

The study required six laboratory visits. The first visit consisted of an initial in-person screening. For those deemed eligible and who consented to participate, the in-person screening session was followed by four experimental sessions assessing four ENDS varying in device power and nicotine content and a final session assessing own-brand products. The current study uses data from the own-brand session only; results for selected outcomes from experimental sessions are reported in Hoetger et al. (2022). We collected demographic information and baseline data at the in-person screening, including tobacco use and dependence. During the final (own-brand) laboratory visit, participants completed computerized cross-price purchase tasks (Bono et al., 2022; Stein et al., 2018). In these tasks, participants were presented with a series of hypothetical scenarios in which one “preferred” product, own-brand ENDS, was available for purchase at 15 different prices (ranging from US$0.00 to $10.24 per 10 puffs) while an “alternative” product, cigarettes, was concurrently available at a fixed price of $1.00 per 10 puffs (see Supporting Information). In each price scenario, participants reported how many of each product they would purchase for personal use on a typical day, assuming they had no access to other products and had the same income and savings as they currently had when making real-world tobacco purchases. Dual cigarette/ENDS users were informed that the cigarettes available were their own brand. For exclusive ENDS users, the task specified “cigarettes” with no further description, so participants were free to respond to the question using only their previous knowledge and perceptions of cigarettes. The task was computer administered, prices were presented one at a time in ascending order, and the total expenditure on each product at each price point was displayed to the participant before they finalized their purchasing decisions. Responses to the task were not reinforced. The study was reviewed and approved by the relevant Institutional Review Board.

Measures

Cross-price elasticity

The elasticity of demand for cigarettes as own-brand ENDS prices increased in the cross-price purchase tasks was analyzed via linear mixed-effects regressions. In these regressions, the log price of the preferred product was regressed on log consumption of the alternative product (Heckman et al., 2017; Johnson et al., 2017; O’Connor et al., 2014). The resulting coefficient represents the cross-price elasticity, an index of price responsiveness, or the rate of change in consumption for the alternative product as prices for the preferred product change. A statistically significant positive coefficient suggests a substitution effect, indicating that consumption of the alternative product increases with increases in price for the preferred product. A statistically significant negative coefficient suggests that the two products are complement goods. A nonsignificant coefficient suggests that the two products are independent goods.

Demand for the alternative product

In our task, we defined nonzero cross-price intensity as 0 if a participant reported purchasing zero cigarettes at the first ENDS price scenario ($0) and 1 otherwise. Participants with nonzero cross-price intensity were presumed to have a particularly strong interest in cigarettes given that in this hypothetical price scenario they could demand an unlimited supply of their own-brand ENDS for free.

Constant null demand for the alternative product was defined as 0 if the participant reported zero consumption at all price scenarios and 1 if a participant would purchase any amount of the alternative product in any price scenario. Participants who would not purchase the alternative product regardless of the price of their preferred product are presumed not to have any interest in it.

Strength of preference for the preferred product

The cross-product crossover point is defined by the first price point at which a participant purchases a greater quantity of the fixed-price alternative product than the price-varying preferred product. For participants who never purchase greater quantities of the alternative product than the preferred product, cross-product crossover points are imputed as the highest price across scenarios (here, $10.24).

Dual-use liability

The dual-demand score was defined by summing the number of price points at which participants demanded any quantity of both the alternative and preferred products and dividing this sum by the total number of price points in the task (here, 15). Higher scores reflect greater propensity toward dual use; a score of zero indicates those with null demand for one product across all price points or those who cease using the preferred product completely before reporting any demand for the alternative product.

Among those with a dual-demand score greater than zero, we defined the dual-use break point as the last price at which participants reported any demand for both products. Beyond this price, participants could continue to purchase the alternative product or cease use of both products. For participants who never achieve complete substitution, the dual-use break point is imputed as the highest price across scenarios (here, $10.24). For participants who never report any dual use (i.e., a dual-demand score of zero), the dual-use break point is missing.

Demographics and history of tobacco use

At enrollment, participants reported demographics and tobacco use history and provided exhaled carbon monoxide measurements (Vitalograph, Lenexa, Kansas). Tobacco use measures included the time since ENDS use initiation (months), number of ENDS pods or ml e-liquid used per day, and cigarettes smoked per day (dual users only). ENDS dependence was measured by the ENDS Dependence Scale (EDS; Morean et al., 2019) and the Penn State ENDS Dependence Index (PSECDI; Foulds et al., 2015). Dual users also completed the four-item Nicotine Dependence for Daily and Nondaily Smokers scale (PROMIS; Shadel et al., 2014), the Penn State Cigarette Dependence Index (PSCDI), and the Fagerström Test for Nicotine Dependence (FTND; Heatherton et al., 1991). For dual users, two measures of total dependence were created by summing the EDS and PROMIS and the PSECDI and PSCDI (Yingst et al., 2023).

Analysis

The primary cross-price elasticity analysis involved separate regression models for dual and exclusive ENDS users. Values of zero were replaced with 0.001 (O’Connor et al., 2014), including for participants with constant null demand. The main conclusions of subsequent analyses were insensitive to instead using constant values of 0.01 and 0.1. Linear mixed models were fit via maximum likelihood and included participant-specific random intercepts to control for unobserved heterogeneity between participants; because the random effects term contained a single variable, the identity covariance matrix was used. Baseline analyses used the observed information matrix to estimate the variance–covariance matrix; in sensitivity analyses, we examined models that included robust standard errors (using the Huber–White or sandwich estimator of variance) to further account for multiple observations within persons, as each participant provided 15 responses per task (one for each price point). As a secondary analysis, we estimated individual cross-price elasticities using separate models for each participant and task. The average of these individual cross-price elasticities is equivalent to the coefficients from the primary analysis. The individual cross-price elasticities were used for comparison with the additional measures described below. Finally, we plotted total consumption (i.e., total units of ENDS plus total units of cigarettes purchased) at each price to illustrate the overall effect of price on tobacco product purchasing.

Sample characteristics and abuse liability measures were summarized using univariate statistics. We examined differences between dual users and exclusive ENDS users in nonzero cross-price intensity, constant null demand, cross-product crossover point, dual-demand score, and dual-use break point using Fisher’s exact tests and two-sample t tests. Overall and separately for dual users and exclusive ENDS users, we assessed interrelationships among outcomes, tobacco use at baseline, and dependence with t tests or Pearson correlations, depending on the distribution of the measure. All analyses were conducted in Stata version 15 (StataCorp, College Park, Texas) with a significance threshold of p < .05. The parent trial was registered on clinicaltrials.gov (NCT03830892), but the current analyses, which were not a primary end point of the parent trial, were not preregistered. Data and syntax are available from the authors upon request.

RESULTS

Sample characteristics

The participants (n = 36) were 17 exclusive ENDS users and 19 dual cigarette/ENDS users. The majority was male (64%) and non-Hispanic White (64%), and the average age was 29.69 (SD 9.52). Participants reported using ENDS for 24.25 (SD 25.47) months prior to study enrollment (Supporting Information, Table A1). All but one exclusive ENDS user (95%) had previously tried cigarettes, but few (16%) used any nicotine/tobacco products other than ENDS in the month prior to enrollment (Supporting Information, Table A2).

Cross-price elasticity

In baseline analyses, cigarettes had significant, positive cross-price elasticities. For exclusive ENDS users, cigarettes appeared to function as a substitute for own-brand ENDS (β = 0.29, p < .001; Table 1), and similarly for dual users, own-brand cigarettes served as a substitute for own-brand ENDS (β = 0.29, p < .001). Thus, the findings indicate that as ENDS prices increased, demand for cigarettes also increased among both exclusive and dual users. In the sensitivity analysis that allowed for robust standard errors, the substitution effect remained for exclusive ENDS users (p < .010) but not for dual users (p < .10). Despite the substitution effects identified, total consumption of tobacco products generally decreases as prices increase—including among those dual users who reported demand for both products in a single-price scenario—although some participants maintained constant total consumption across prices (Figure 6).

TABLE 1.

Cross-price purchase task outcomes for exclusive ENDS users (n = 19) and dual cigarette/ENDS users (n = 17)

Group	Exclusive ENDS users	Dual users	p
Cross-price elasticity
Baseline: β (95% CI) p value	0.29 (0.21, 0.37) p < .001	0.29 (0.16, 0.41) p < .001	--
Sensitivity: β (95% CI) p value	0.29 (0.08, 0.49) p < .010	0.29 (−0.05, 0.62) p < .100	--
Additional measures of demand
Demand for the alternative product†
Nonzero cross-price intensity: N (%)	0 (0%)	10 (59%)	<.001
Constant null demand: N (%)	13 (68%)	1 (6%)	<.001
Strength of preference for ENDS‡
Cross-product crossover point: M (SD)	$8.42 (3.09)	$2.95 (2.70)	<.001
Dual-use liability‡
Dual-demand score: M (SD)	0.70% (0.02)	35.29% (33.65)	<.001
Dual-use break point: M (SD)	--	$3.61 (2.78)	--

Note. Abbreviations: CI = confidence interval; M = mean; SD = standard deviation. The double dashes (--) indicate differences between exclusive and dual users in dual-use break point were not assessed, as only two of 19 exclusive ENDS users reported demand for both products at the same price point.

^†p values come from Fisher’s exact tests.

^‡p values come from t tests.

FIGURE 6.

Observed total consumption of tobacco products generally decreases as ENDS prices increase. Total consumption represents the sum of demand for ENDS and demand for cigarettes at each price. Dual users (shown in blue) were disaggregated by whether the participant reported demand for both ENDS and cigarettes in a single price scenario (Panel A: dual-demand score > 0; Panel B: dual-demand score of 0). Exclusive ENDS users are shown in purple (Panel C). A value of 0.001 is used to represent zero on the x-axis given the log scale. Figure displays empirical data (vs. fitted values).

Demand for the alternative product

No exclusive ENDS users had nonzero cross-price intensity, with 0% purchasing cigarettes if own-brand ENDS were free (Table 1), whereas a significantly higher fraction of dual users (59%) would still purchase own-brand cigarettes if own-brand ENDS were free, p < .001. Most exclusive ENDS users displayed constant null demand, with 68% opting not to purchase cigarettes in any price scenario in the task versus 6% (n = 1) among dual users, p < .001. Together, these outcomes indicate low demand for cigarettes among exclusive ENDS users regardless of how comparatively expensive ENDS became, whereas dual users would continue to purchase cigarettes even if they had virtually unlimited access to ENDS at no cost.

Strength of preference for the preferred product

When cigarettes were available at a fixed price of $1 and ENDS varied in price, the mean ENDS price at which participants would begin purchasing more cigarettes than own-brand ENDS was $8.42 (SD $3.09) for exclusive ENDS users and $2.95 (SD 2.70) for dual users, t(34) = 5.64, p < .001. That is, for exclusive ENDS users, cigarette consumption did not outpace own-brand ENDS consumption until the price of ENDS was, on average, 8.42 times that of cigarettes.

Dual-use liability

Exclusive ENDS users generally did not purchase both fixed-price cigarettes and price-varying own-brand ENDS concurrently, with mean dual-demand scores of just 0.7% (SD 0.02). Dual users showed significantly higher dual-demand scores, opting to purchase both fixed-price own-brand cigarettes and price-varying own-brand ENDS at 35.3% of prices on average, t(34) = −4.48, p < .001. Compared with exclusive ENDS users, dual users would purchase both products across a wider range of price scenarios, indicating relatively greater preference for dual use as opposed to complete substitution (Figure 7).

FIGURE 7.

Observed range of ENDS prices at which participants reported demand for both ENDS and cigarettes simultaneously. Each lettered row represents data observed for an individual participant. Participants who did not report demand for both ENDS and cigarettes simultaneously are not shown. Two of the 19 exclusive ENDS users (shown in purple) reported demand for both products simultaneously, and each did so at a single price point ($1.28 and $2.56). Eleven of the 17 dual users (shown in blue) reported demand for both products in one or more price scenarios. Participant G reported demand for both products when ENDS cost $0–$1.28 per 10 puffs and again when ENDS cost $3.84 per 10 puffs. Participant M reported demand for both products at a single price point ($3.84).

Most (65%) dual users purchased both products simultaneously at one or more price points. Among this group, the average dual-use break point was $3.61 (SD 2.78). That is, for those most at risk of dual use, policies that increase the price of ENDS to 3.61 times the price of cigarettes are likely to encourage cessation of ENDS through either complete substitution of ENDS in favor of cigarettes or cessation of both products. Few exclusive ENDS users opted to purchase cigarettes at any price, so differences between dual and exclusive ENDS users were not assessed.

Associations between cross-price elasticity, other measures of abuse liability, and tobacco use characteristics

We first compared outcomes between dual users with and without nonzero cross-price intensity. Individually estimated cross-price elasticity differed significantly by nonzero cross-price intensity: on average, dual users with cross-price intensities of zero had positive cross-price elasticities and dual users with nonzero cross-price intensities had negative cross-price elasticities (Supporting Information, Table A3). That is, dual users who would not buy cigarettes if ENDS were free may be more likely to treat cigarettes as a substitute for ENDS, whereas dual users who would continue to buy cigarettes even if ENDS were free may be more likely to treat cigarettes as a complement to ENDS (Figure 8). Nonzero cross-price intensity was also related to dual-demand score. Those who would not purchase cigarettes when ENDS were free would purchase both cigarettes and ENDS at 4% of price points, versus 57% of price points among those who would purchase cigarettes when ENDS were free, t(15) = −5.29, p < .001. These findings suggest that nonzero cross-price intensity could identify divergent responses to tobacco regulatory policies among dual users: those who are more willing to switch products in response to changes in price and less inclined toward dual use and those more committed to dual use and for whom changes in price may lead to reductions in demand for both products.

FIGURE 8.

Observed demand for ENDS and cigarettes among dual users, by nonzero cross-price intensity. The gray dashed lines represent observed (vs. fitted) demand for ENDS, and the solid black lines represent observed demand for cigarettes. One representative participant in each panel is highlighted in blue. A value of 0.001 is used to represent zero on both axes given the log-log scale. Among participants with nonzero cross-price intensity (Panel A), n = 6 had negative cross-price elasticities (i.e., cigarettes were a complement for ENDS), n = 2 had a cross-price elasticity of zero (i.e., an independent relationship between ENDS price and cigarette demand), and n = 2 had positive cross-price elasticities (i.e., cigarettes were a substitute for ENDS). Among participants with zero demand for cigarettes if ENDS were free (Panel B), n = 5 had positive cross-price elasticities (i.e., cigarettes were a substitute for ENDS) and n = 1 had constant null demand for ENDS.

We then examined correlations among outcomes and individual characteristics. Across all participants, cross-product crossover point was negatively correlated with dual-demand score (r = −.53, p =.001; Supporting Information, Table A4), indicating that stronger preferences for own-brand ENDS over cigarettes were linked to lower interest in dual use; similar correlations were seen among exclusive ENDS users (Supporting Information, Table A5) but not dual users (Supporting Information, Table A6).

Among exclusive ENDS users, individually estimated cross-price elasticity was negatively associated with cross-product crossover point (r = −.98, p < .001) and with ENDS use history (r = −.50, p = .028) but positively associated with dual-demand score (r = .62, p < .01). That is, those with a stronger propensity toward substituting cigarettes for ENDS tended to have weaker preferences for their own-brand ENDS, to have higher dual-use liability, and to have used ENDS for fewer months. Additionally, we note that among exclusive ENDS users who did not have constant null demand, average cross-price elasticity was 0.91 (SD 0.28; Figure 9). Among dual users only, cross-price elasticity was moderately negatively correlated with dual-demand scores, suggesting that those with a stronger propensity toward substituting one product for another also switch products more quickly, with a narrower window of dual use.

FIGURE 9.

Observed demand for ENDS and cigarettes among exclusive ENDS users who would purchase cigarettes at any ENDS price (n = 6). Six of 19 exclusive ENDS users would purchase cigarettes at any ENDS price. The gray dashed lines represent observed (vs. fitted) demand for ENDS, and the solid black lines represent observed demand for cigarettes. One representative participant is highlighted in purple. A value of 0.001 is used to represent zero on both axes given the log-log scale.

Nicotine dependence was generally not associated with outcomes. Greater levels of exhaled carbon monoxide at baseline were correlated with lower cross-product crossover points (i.e., weaker preferences for own-brand ENDS) and higher dual-demand scores (i.e., greater propensity toward dual use) overall, but in stratified analyses only the association with cross-product crossover point remained significant.

DISCUSSION

Among both dual users and exclusive ENDS users, we found evidence supporting a similar substitution effect for cigarettes: if prices for own-brand ENDS increased by 1%, consumption of combustible tobacco cigarettes would increase by 0.29%. For example, levying a 12% tax on ENDS, like in California and Maryland (CDC, 2023b), may increase demand for cigarettes by 3.3%. However, average cross-price elasticity effects masked important differences in demand for cigarettes, ENDS, or both. Specifically, exclusive ENDS users showed weak preferences for cigarettes even when available for free or when ENDS cost several times the price of cigarettes. Nearly 70% of exclusive ENDS users reported no demand for cigarettes at any ENDS price, so the substitution effect identified in this group was driven by the minority of participants, who may nonetheless be an important subpopulation for which to consider potential unintended consequences of ENDS price policies. In contrast, dual users generally displayed much stronger demand for cigarettes and interest in dual use. Yet sensitivity analyses using a more conservative statistical model for cross-price elasticity found a statistically significant substitution effect for exclusive ENDS users but not for dual users. This latter finding may be due to heterogeneity among the dual-user group, some of whom were more reluctant to give up ENDS, defending consumption of both products despite increasing prices. Thus, although a substitution effect may generally exist, responses to changes in the tobacco marketplace can vary widely both between and within groups of tobacco product users. In this case, our findings indicate that policies and regulations that increase ENDS prices appear unlikely to steer most exclusive ENDS users toward cigarette use but could prompt some dual users to completely substitute cigarettes while others will remain dual users. Ultimately, a high enough price burden is likely to encourage cessation or reduction in use of all tobacco products cumulatively.

Our findings highlight the importance of examining heterogeneity in responses to hypothetical price policies for tobacco products and demonstrate the potential utility of our proposed extensions to the cross-product purchase task for doing so. All five measures of demand and dual-use liability distinguished between exclusive ENDS users and dual users, providing suggestive evidence for the measures’ divergent validity. Some outcomes were also associated with behavioral and physiological markers of tobacco use (months since ENDS initiation; exhaled carbon monoxide). However, outcomes were not associated with measures of dependence, perhaps because of the distribution of nicotine-dependence scores in our sample, which did not differ between dual and exclusive ENDS users (Hoetger et al., 2022). Further, given that the outcomes were defined a priori, some outcomes may still obscure heterogeneity in demand profiles. For example, a dual-demand score of 33% may reflect different propensity for dual use if the prices at which both products are reported occur at the high or low end of the price range or are not contiguous. Future research should evaluate the validity of the additional outcomes proposed here and consider using alternative methods such as latent class analyses or latent class mixed modeling to examine the extent to which these outcomes reflect meaningful distinctions in demand profiles. These methods have previously been applied to determine latent structures of demand in single-commodity purchase tasks (Bidwell et al., 2012; MacKillop et al. 2009) and to determine patterns of nonsystematic responding in delay-discounting research (Gilroy, Strickland, et al., 2022). In the context of multicommodity purchase tasks, latent class methods could be leveraged to identify subgroups of consumers who may drive the effects observed in the overall marketplace as well as patterns in the types of tobacco products purchased in response to price changes. Such information could be useful for policy makers wishing to understand the potential consequences of tobacco regulation.

In addition to contextualizing cross-price elasticity for policy applications, the additional outcomes from cross-price purchase tasks could inform study design, recruitment, and evaluation. An unexpected finding of constant null demand in pilot studies may indicate a problem with the alternative product being tested or the sample being recruited (e.g., participants are unfamiliar with the product or find it aversive) or that the prices used in the task need to be adjusted (e.g., the price for the alternative product is too high). Alternately, in a trial directed at transitioning smokers to nicotine replacement therapy, nonzero cross-price intensity or constant null demand could be valuable screening items to distinguish potential participants who are open to using the alternative product from those who are not, and dual-use liability measures could be useful in evaluating whether the trial resulted in complete substitution.

Our work may provide insight into the disagreement among studies leveraging sales data, which have at times suggested that cigarettes and ENDS are economic substitutes, complements, or independent goods (e.g., Cotti et al., 2018, 2020; Huang et al., 2014). Group-level trends in cross-price elasticity may not provide a complete and accurate picture of how tobacco product users will respond to changes in the tobacco marketplace. Our conclusions align closely with those of Snider et al. (2017), who demonstrated that changes to cigarette prices may cause some smokers to quit and some to switch to ENDS. Further, evidence from discrete-choice experiments has suggested that features other than price (e.g., perceived harm reduction) can influence the likelihood of a smoker substituting with ENDS (Marti et al., 2019; Shang et al., 2020). Similarly, dual users in our study were sensitive to ENDS price, but the fact that many would purchase cigarettes even if ENDS were free suggests that factors other than price affect demand.

Limitations

The primary objective of the parent study required restrictive eligibility criteria, so ENDS users in this study may not be representative of ENDS users generally. For example, heterogeneity in previous tobacco use experience among exclusive ENDS users may have affected familiarity with combustible tobacco products and therefore demand observed in the task. Further, data collection was interrupted due to the COVID-19 pandemic and thus the sample size is limited (Hoetger et al., 2022), so policy implication interpretations from this work should be cautious.

Some caveats regarding the outcome definitions should be considered. First, unexpected patterns of demand such as nonzero cross-price intensity could represent participants’ inattention or misunderstanding of the task (Bono et al., 2022). Efforts to ensure participant engagement and comprehension (e.g., by reinforcing purchasing choices), limit unrealistic consumption responses (e.g., by setting a limited budget for tobacco products within the task), and conduct a careful examination of the data are critical. To our knowledge, there are no standardized criteria for determining whether data from cross-price purchase tasks are systematic (Stein et al., 2015), and no participants were excluded due to data quality. We note that when applying Stein et al.’s nonsystematic data-detection criteria to ENDS (the preferred product) only, two dual users failed the “trend” criterion because they reported the same consumption at every price; no other indicators of nonsystematic purchasing for the preferred product were detected. Second, all outcomes depend on a well-designed cross-price purchase task with parameters that are appropriate to the products and populations studied. In particular, cross-product crossover point may be sensitive to the price frame (i.e., unit of the product available for purchase) used in the task if, for instance, there are substantial differences in nicotine delivery between cigarettes and ENDS. Third, more work is needed to determine how the additional cross-product demand measures described here can be extended to the Experimental Tobacco Marketplace, which can assess more than two competing products at once (Bickel et al., 2018). Fourth, future studies may find additional utility in further examining patterns of responses among those with nonzero cross-price intensity to determine whether, for example, low levels of initial demand predict a substitute relation and high levels of initial demand predict a complement relation. We were unable to assess this possibility in the current study given the small sample size of dual users who were eligible for this subgroup analysis and the lack of meaningful cutoff points to distinguish low versus high initial demand.

Conclusions

Cross-price elasticities estimated from the cross-price purchase task indicated that cigarettes would substitute for ENDS among both exclusive ENDS users and dual users. However, additional focused analysis of the underlying data tempered this conclusion. Most exclusive ENDS users were unwilling to substitute with cigarettes regardless of ENDS prices. Some dual users responded to increasing ENDS prices by switching to cigarettes alone, whereas others defended dual use of both products across prices. Decision makers should be aware that broad policy and regulatory actions governing tobacco product prices may have diverse effects on different groups, and researchers should consider examining the additional indices of demand outlined here when evaluating tasks designed to assess economic relationships between products. Importantly, this work could be applied to not only questions regarding substance use but also any domain for which a more complete understanding of the substitutability of two potential alternative goods is of interest to policy makers and practitioners, such as health insurance plans, diet, and energy usage.