Validation of a Brief Behavioral Economic Assessment of Demand among Cigarette Smokers

Abstract

Basic and clinical addiction research use demand measures and analysis extensively to characterize drug use motivations. Hence, obtaining an accurate and brief measurement of demand that can be easily utilized in different settings is highly valued. In the current study, two versions of a breakpoint measure, designed to capture cigarette demand, were investigated in 119 smokers who were recruited from an online crowdsourcing platform. The first version determines the maximum price a smoker is willing to pay for 1 cigarette received right now when paid out of pocket, and the second determines the maximum price when paid using a hypothetical $100 gift card received for free. The breakpoint measures were administered along with the Cigarette purchase Task (CPT), Fagerström Test for Cigarette Dependence (FTCD), and The Questionnaire of Smoking Urges (QSU-brief). Both single-item breakpoint versions were significantly correlated with CPT-derived demand measures loaded on the persistence factor (i.e., elasticity of demand, breakpoint, Pmax, Omax), but not with those loaded on the amplitude factor (i.e., intensity of demand). In addition, both single-item measures were associated with metrics of tobacco dependence (e.g., FTCD, QSU) with effect sizes that are similar to the ones found between CPT-derived breakpoint and those same metrics. These findings suggest that the single-item breakpoint measure is a viable method for measuring demand that may provide a useful and efficient tool to capture crucial and distinct aspects of smoking. In addition, the breakpoint measures may help increase the utility of behavioral demand measures in novel research and clinical settings.

The field of behavioral economics integrates principles from psychology and economics to study human behaviors (Bickel & Vuchinich 2000). Behavioral economics can also be applied to understanding motivation and demand for drug use (e.g., tobacco and alcohol) by assessing an individual’s preference for a drug in terms of the response to costs associated with its use (Acker & MacKillop 2013; MacKillop, Brown et al. 2012). Behavioral economic demand, which refers to the quantitative relationship between consumption of a commodity and its cost, has demonstrated utility in preclinical (For a review, see Hursh & Silberberg 2008) and clinical (MacKillop, Brown et al. 2012) research that aims to assess drug abuse liability, craving, and withdrawal.

An empirical body of evidence has established the negative association between cigarette prices and consumption (Buczkowski, Marcinowicz, Czachowski, & Piszczek 2014; Douglas 1998; Forster & Jones 2001; Ross, Blecher, Yan, & Hyland 2011; Tauras & Chaloupka 1999). In addition, cigarette demand is widely used to assess smoking motivation (Heather & Vuchinich 2003). Hence, the development of methods for quantifying drug demand (Hursh & Silberberg 2008), or simply how much an individual values drug (e.g., nicotine), has been a major focus of the behavioral economic field.

One validated and widely used measure of demand is the purchase task, in which individuals make cost-benefit decisions about how much of a specific commodity to consume at different prices (Hursh 1980; Hursh & Silberberg 2008). For example, in the cigarette purchase task (CPT), individuals report the number of cigarettes they would buy at escalating levels of price (e.g., how many cigarettes would you buy if they were $0.13 each?, how many cigarettes would you buy if they were $0.25 each? etc.,), either hypothetically (HPT) (Jacobs & Bickel 1999; MacKillop, Murphy, Ray et al. 2008), potentially real (i.e., one randomly selected purchase scenario is actually received) (Koffamus, Wilson, & Bickel 2014; Wilson, Franck, Koffamus, & Bickel 2016), or real (i.e., participants are given the cigarettes purchased, and all excess money) (Wilson et al. 2016).

The responses across the different prices among the purchase task are translated into an overall demand curve with multiple conceptually related (but not redundant) indices (Bickel, Marsch, & Carroll 2000) including intensity of demand (i.e., consumption when cigarettes are free), elasticity of demand (i.e., the change in cigarettes consumption as cost increases), breakpoint (i.e., the first price at which consumption is zero), Omax (i.e., maximum expenditure allocated for cigarettes), and Pmax (i.e., the point at which consumption moves from being inelastic to elastic). Latent factor analysis examining the underlying structure of facets of incentive value generated from a demand curve indicated that elasticity, breakpoint, and Pmax loaded on a factor that represent persistence of demand (i.e., insensitivity to price change), while intensity loaded on a separate factor that represents amplitude of demand (i.e., the absolute levels of consumption and price), and Omax loaded on both factors (Aston, Farris, Mackillop, & Metrik 2017; Bidwell, Mackillop, Murphy, Tidey, & Colby 2012; MacKillop, Murphy, Tidey et al. 2008; O’Connor et al. 2016).

Previous research supported the validity of hypothetical purchase tasks as a time- and cost-efficient method for assessing cigarette demand (MacKillop, Murphy, Ray et al. 2008). However, the implementation of hypothetical cigarette purchase tasks and their use are precluded in some settings due to two major barriers: (1) the task is relatively long especially when assessment time is constrained or during multi-model assessments (ranges from approximately 13 prices presented (Stein et al. 2017) to more than 70 (MacKillop, Few et al. 2012), depending on the research question), and (2) the analysis is relatively complex (e.g., several indices require derivation, and nonlinear regression is needed to generate elasticity and model the demand curve) (Hursh & Silberberg 2008; Koffarnus, Frank, Stein, & Bickel 2015; Owens, Murphy, & MacKillop 2015).

Given smoking’s substantial negative health effects (Courtney 2015; WHO 2011), the increased interest in understanding smoking motivations, the important information provided by demand measures, and the barriers of implementation of existing purchase tasks in a variety of settings, the current study sought to investigate the reliability and usefulness of assessing cigarette valuation using brief breakpoint measures to capture cigarette demand while overcoming the purchase task barriers. The brief breakpoint measure for cigarette valuation was investigated in a group of smokers who were recruited from an online crowdsourcing platform. As the highest price one is willing to pay for a commodity might depend on income (e.g., those with higher income might be willing to pay higher price for a specific commodity compared to those with lower income but not necessarily value the commodity more), the current study used two versions of a single item breakpoint measure. The first version determines the maximum price a smoker is willing to pay for 1 cigarette received right now when paid out of pocket, and the second, to limit the effect of income, determines the maximum price a smoker is willing to pay for 1 cigarette received now when paid using a hypothetical $100 gift card received for free. Given prior latent factor analyses (Aston et al. 2017; Bidwell et al. 2012; Epstein et al. 2018, MacKillop, Murphy, Tidey et al. 2008; O’Connor et al. 2016) we hypothesized that cigarette demand from both versions of the breakpoint measure, as a persistence measure, would correlate with the demand measures shown previously to load on the persistence factor (i.e., elasticity of demand, breakpoint, Pmax. Omax) obtained using lengthier, but more standard tasks such as the CPT. In addition, based on the same factor analyses, we hypothesized that the brief breakpoint measures would not correlate with demand measures that load on the amplitude of demand factor (i.e., intensity of demand).

Methods

Participation in the study was voluntary. Completion and return of the survey was considered an implied consent to participate in the study. This study was approved by the Institutional Review Board (IRB) at Virginia Polytechnic Institute and State University.

Participants

The study was carried out using data collected online from mTurk, a crowdsourcing service in which employers post Human Intelligence Tasks (HITs) which may be completed online in exchange for a previously determined monetary payment. Participants received $1 compensation upon completion of this study and an additional bonus of $1 for providing consistent and systematic purchase data according to the standard diagnostic criteria for the PT (Stein et al. 2015). Eligibility was assessed using a brief screening questionnaire. To accept the posted HIT participants were required to (1) be located in the United States; (2) have a HIT approval rate greater than 90%; (3) smoke at least 5 cigarettes per day.

Design and procedure

Participants were asked to complete a brief demographics questionnaire that was followed by a randomly presented battery of assessments including the two versions of Breakpoint Measure, the Fagerstrom Test for Cigarette Dependence (FTCD), the Cigarette Purchase Task (CPT), and The Questionnaire of Smoking Urges (QSU-brief).

Study variables

We collected demographic and tobacco use data including age, monthly income, gender, race, ethnicity, education level, number of cigarettes per day, and time to last cigarette (Table 1).

Table 1.

Sample characteristics (N=119)

Characteristics	Frequency (%)/Mean (SD)
Male	79 (66.4)
Marital Status
Married	44 (37.0)
Single	67 (56.3)
Other	8 (6.6)
Education level
High school diploma or equivalency	22 (18.5)
Some college or vocational training	27 (22.7)
Completed a 2-year college degree	12 (10.1)
Completed a 4-year college degree or more	58 (48.8)
Income
Less than $9,999	11 (9.2)
$10,000 - $29,999	35 (29.4)
$30,000 - $49,999	20 (16.8)
$50,000 - $69,999	10 (11.9)
$70,000 +	12 (10.1)
Race
White	92 (77.3)
African American	11 (9.2)
Other	16 (13.5)
Time since last cigarette
Just now	15 (11.3)
Less than an hour ago	59 (44.4)
1-2 hours ago	40 (30.1)
More than 2 hours ago	5 (3.8)
Not-Hispanic	104 (87.4)
Age	33.7 (8.49)
Number of daily cigarettes	14.9 (7.43)
FTCD	4.97 (2.28)
Breakpoint Measure (Out of Pocket)	3.33 (7.11)
Breakpoint Measure ($100 gift card)	9.52 (20.38)

Breakpoint Measure.

The brief breakpoint measure consists of two versions of a single-item measure, designed to capture cigarette demand. The first version of the measure is modified from a brief assessment of alcohol demand (BAAD) that was previously tested and validated against a hypothetical alcohol purchase task (Amlung, McCarty, Morris, Tsai, & McCarthy 2015; Owens et al. 2015). BAAD is a 3-item measure, designed to capture three important indices of demand that are derived from demand curve modeling (intensity, Omax, and breakpoint) (Amlung et al. 2015). In this study, the brief intensity measure is explicitly asked using the CPT (described below).

Previous studies using hypothetical purchase tasks have indicated particularly high correlations between breakpoint and Omax measures (Amlung et al. 2015; Amlung & MacKillop 2014; MacKillop et al. 2010; Murphy & MacKillop 2006; Owens et al. 2015), suggesting that these two items are measuring a common underlying aspect of drug valuation and including one of them could be sufficient. However, when brief measures of Omax and breakpoint were compared with the corresponding measures obtained using the full-length purchase task, the effect size for the brief Omax was notably smaller and outside the confidence intervals while the effect size for the brief breakpoint measure in the same study was modest in magnitude and fell within 95% confidence intervals of the originally observed effect sizes (Owens et al. 2015). Hence, in this study, only the brief breakpoint item was investigated as a brief measure of cigarette valuation and demand compared to the indices of the full-length CPT.

We modified the BAAD breakpoint measure by asking about cigarettes instead of alcoholic drinks. To assess the valuation of cigarettes when paid out of pocket, we asked the following question: “What is the maximum price you would be willing to pay for 1 cigarette received now?” similar to the full CPT (Jacobs & Bickel 1999), participants were instructed to assume that they do not have an access to any cigarettes or nicotine products other than those offered in this question.

The second version was developed to limit the effect of income on the maximum price individuals would pay for one cigarette. In this version, we asked participants to imagine receiving a $100 gift card for free. Participants were asked to assume this gift card could be used only for buying cigarettes, the card does not expire, and they cannot sell, give away, or share the cigarettes purchased using the gift card. Then we asked: “Using only the $100 gift card, what is the maximum price you would be willing to pay for 1 cigarette received now?” followed with the exact same instructions about access to nicotine provided in the first version.

Fagerström Test for Cigarette Dependence (FTCD).

FTCD is a valid and reliable six-item measure that assesses the intensity of physical addiction to cigarettes (Fagerström 2012). The FTCD evaluates the frequency and temporal distribution of participants’ smoking behavior, as well as whether they are able to refrain from smoking when they are ill and in public places in which smoking is banned. Responses to each item are scored and the total FTCD score ranges from 0 to 10, with higher scores indicating more intense physical dependence on cigarettes (Fagerström 2012).

The Cigarette Purchase Task (CPT).

The CPT is a hypothetical purchase task that measures cigarette demand and is a validated measure for assessing the relative reinforcing value of nicotine in smokers (Jacobs & Bickel 1999; MacKillop, Murphy, Ray et al. 2008). In the current study, the CPT started with asking participants to report the number of cigarettes they would purchase and use over the next 24 hours at the following 13 prices in ascending order: $0.00, $0.03, $0.06, $0.12, $0.25, $0.50, $1, $2, $4, $8, $16, $32, and $64 per one cigarette. Participants were asked to assume that: (1) The available cigarettes were their usual brand, (2) they have no access to any cigarettes or nicotine products other than those offered at these prices, (3) they would smoke the purchased cigarettes over the next 24 hours without saving or stockpile them for a later date. (4) they may not give away any of the cigarettes they purchased.

The Questionnaire of Smoking Urges (QSU-brief).

The QSU used in this study is an abbreviated version of the original smoking urge questionnaire (QSU) (Tiffany & Drobes 1991). The questionnaire consists of ten ‘agree-disagree’ Likert items that assess cigarette craving using 2 factors. Factor 1 (5 items) represents a desire and intention to smoke, with smoking considered rewarding. Factor 2 represents an urgent desire to smoke to relieve negative affect and nicotine withdrawal (Cox, Tiffany, & Christen 2001). Individuals are instructed to respond to statements using a 7-point scale ranging from strongly disagree to strongly agree. The scores of both subscales were examined in this study, with higher scores indicating higher levels of craving to smoke.

Data Analysis

We generated an estimate of demand curves by fitting participant's purchases across the range of prices to Koffarnus et al. (2015) exponential demand curve equation:

$$Q=Q_0\times {10}^{k(e^{-\alpha Q_{0}C}-1)}$$

in which, Q is the quantity consumed at price C, Q₀ is the quantity consumed at price $0, k is the range of cigarette consumption in log₁₀ units, and α is the change in cigarettes consumption as cost increase. Larger α values reflect higher elasticity. The value of k is constant across all analyses. Here, we determined k by subtracting the log10- transformed average consumption at the highest price from log10-transformed average consumption at the lowest price (giving k=1.44).

We used the CPT to generate five demand indices: (1) intensity (i.e., Q₀, consumption at a cost of zero); (2) elasticity of demand (i.e., α, sensitivity of cigarette consumption as cost increases); (3) CPT-derived breakpoint (first price at which consumption of cigarettes is zero); (4) (Omax (maximum expenditure on cigarettes); and (5) Pmax (the price associated with the maximum expenditure). (Omax and Pmax were calculated using the essential value Pmax, and Omax automated calculator (Kaplan & Reed 2014) after supplying the demand curve parameters calculated previously (Q0, α, and k).

The demand metrics were examined for distribution of normality and non-parametric Spearman correlation was used to assess correlation between the breakpoint measures values and indices of demand (as measured by the CPT), participants’ daily consumption of cigarettes, nicotine dependence (as measured by the FTCD), and nicotine cravings (as measured by the QSU-brief). In addition, bivariate linear regression analyses of income with each of the two breakpoint measures versions were carried out to test the effect of income on participant’s maximum price they are willing to pay for 1 cigarette. The Wilcoxon signed-rank test analyses between the out of pocket and the $100 gift card items were run to determine the effect of offering a free $100 cigarette gift card on the maximum price one is willing to pay for 1 cigarette. In addition, a one-way ANOVA was conducted to compare means of the brief breakpoint measures by relevant demographic and tobacco use characteristics (e.g., gender, income, education, and time since last cigarette). All analyses were performed in Stata 13.1 (Stata Corp, 2013) at a significance level of 0.05.

Results

A total of 119 subjects completed the study and were included in the analysis. Table 1 provides information on socio-demographic characteristics as well as the mean number of daily cigarettes, nicotine dependence and breakpoint measure items for the cigarette smokers who participated in the study.

Correlational analyses revealed moderate relationships between the breakpoint measure items and the behavioral economic indices of demand and nicotine dependence (Table 2). Valuation of cigarettes when paid out of pocket and when paid using a $100 gift card was positively associated with the FTCD scores (p=.002 and p=.009, respectively), QSU factor 1 (p=.017 and p=.011, respectively), QSU factor 2 (p=.001 and p=.002, respectively), CPT-derived breakpoint (p<.001 and p<.001, respectively), Omax (p<.001 and p=.001, respectively), and Pmax (p=.002 and p=.021, respectively) and associated negatively with elasticity of demand (p<.001 and p=.001, respectively). However, no associations were found between valuation of cigarettes when paid out of pocket and when paid using a $100 gift card with intensity of demand (p=.974 and p=.766, respectively). These associations are similar to the ones reported of the CPT-derived breakpoint and those same metrics of nicotine dependence (e.g., FTCD, QSU, cigarettes per day) (Table 2). Bivariate linear regression analyses of income with both versions of breakpoint measure indicated no significant effect of income on the maximum price participants are willing to pay for one cigarette if paid out of pocket (p= .585) or using the gift card (p=.498). The two versions of the breakpoint measure were strongly correlated with each other (r= 0.792, p<.000). The Wilcoxon signed-rank test analyses indicated that the maximum price participants were willing to pay for one cigarette using the gift card was significantly higher than the maximum price paid out of pocket (M=$9.52 vs. $3.33, p<.001 respectively); thus, although these measures showed strong correspondence, the gift card measure produced substantially higher estimates of breakpoint.

Table 2.

Correlations between Breakpoint Measure items and demand indices and metrics of nicotine dependence.

	Breakpoint Measure (Out of Pocket)	Breakpoint Measure ($100 gift card)	Intensity of Demand	Elasticity of Demand	Breakpoint	Pmax	Omax
Breakpoint Measure ($100 gift card)	0.792***
Intensity of Demand	−0.003	0.028
Elasticity of Demand	−0.364***	−0.317***	−0.120
Breakpoint	0.534***	0.530***	−0.019	−0.521**
Pmax	0.286**	0.214*	−0.405	−0.738**	0.391**
Omax	0.325***	0.295**	0.136	−0.950**	0.473**	1.00***
N of daily cigarettes	−0.055	0.017	0.619**	−0.231*	0.048	−0.139	0.205*
FTCD	0.276**	0.237**	0.434**	−0.264**	0.257**	−0.045	0.235*
QSU factor 1	0.218*	0.233*	0.120	0.271**	0.216*	0.158	0.246**
QSU factor 2	0.297**	0.275**	0.098	−0.311**	0.229*	0.155	0.274**

^*p<.05

^**p<.01

^***p<.001

An one-way analysis of variance showed that the mean of the brief breakpoint measures when paid out of pocket and when paid using the $100 gift card were not significantly different by gender (p=.519 and p=.519, respectively), income (p=.627 and p=.089, respectively), education (p=.549 and p=.604, respectively), and time since last cigarette (p=.141 and p=.718, respectively).

Discussion

The purpose of this study was to pilot a brief behavioral economic assessment of cigarette demand that would capture cigarette valuation comparably to the full-length cigarette purchase task while overcoming its barriers. The brief measure assessed in this study used two versions of a breakpoint assessment to determine the maximum price a smoker would pay for 1 cigarette received now when paid out of pocket and when paid using a $100 gift card received for free. The gift card version was added to limit the effect of income on the maximum price one is willing to pay for one cigarette. However, no significant association between income and the maximum price reported was found in this study sample. The out of pocket breakpoint measure was strongly correlated with the $100 gift card but significantly lower in value. Hence, specifically for cigarettes (relatively cheap compared to other drugs), using the out of pocket breakpoint measure to assess cigarette valuation instead of both is viable. However, for more expensive drugs (e.g., cocaine, heroin) where income have an impact on demand (Gallet 2013; Saffer & Chaloupka 1998), valuation of drugs and its demand indices might be better determined using the gift card breakpoint measure. Future research that test both versions in expensive drugs might be necessary to better establish their utility and suitability for measuring demand for drugs.

Both versions of the brief breakpoint measures were strongly associated with the CPT-derived breakpoint and moderately associated with other indices of demand (except intensity of demand). In addition, the brief breakpoint measures were associated with metrics of tobacco dependence (e.g., FTCD, QSU, number of daily cigarettes) with effect sizes that are similar to the ones found between CPT-derived breakpoint and those same metrics. The associations reported in this study are consistent with previously conducted latent factor analysis (Aston et al. 2017; Bidwell et al. 2012; MacKillop, Murphy, Tidey et al. 2008; O’Connor et al. 2016) and provide preliminary support for the construct validity of both versions of the breakpoint measure as a brief measures of cigarette valuation. However, further testing of their construct validity is required, along with tests of their relationship with and predictive validity of response to treatment, success in cessation, and natural course of tobacco use.

Evidence suggests that level of nicotine dependence is associated with adverse consequences of smoking and is recognized as a predictor for success in cessation (Breslau & Johnson 2000). Hence, future research examining the psychometric properties of the brief breakpoint measures in a sample where higher levels of nicotine dependence is expected (e.g., smokers with higher number of daily cigarettes and/or longer smoking history), or using other measures of dependence that tap different aspects of nicotine dependence such as the clinical criteria for tobacco use disorder, or the Minnesota Nicotine Withdrawal Scale (MNWS) that assesses withdrawal symptom severity might be beneficial.

Our investigation has some potential limitations. In this study, the participants were mostly non-Hispanic (87.4%), with a high proportion of Caucasian population (77.3%). Generalizing the results to broader populations should be done with those characteristics in mind. Furthermore, we relied on participants’ self-reported data about smoking status, which might have some potential sources of bias such as selective memory and social desirability bias. However, prior studies have validated self-reporting about smoking status (Rebagliato 2002; Wong, Shields, Leatherdale, Malaison, & Hammond 2012). Moreover, collecting data using mTurk limited our sample to online data only. However, data obtained online and using mTurk studies on addiction (Kim & Hodgins 2017) including smoking (Athamneh, Stein, and Bickel 2017) have been validated in previous research. Finally, as an online study the sample size for this study is relatively small, repeating the study with larger sample size might be necessary to ensure a representative distribution of the population and increase confidence with the findings.

As the first study to test the brief breakpoint measures’ reliability and usefulness of assessing cigarette valuation, we believe the present findings contribute new knowledge that may has substantial implications for increasing the efficiency and utility of behavioral cigarette demand measures in novel research and clinical settings.

Conclusion

The study findings suggest that the single-item breakpoint measures are viable methods for measuring cigarette demand that may provide useful and efficient tool capture crucial and distinct aspects of smoking. In addition, the breakpoint measures may help increase the utility of behavioral demand measures in novel research and clinical settings. However, future research that wishes to assess demand using the brief breakpoint measure should consider including the brief intensity measure as well. Further testing to verify the utility and construct validity for the breakpoint measures is warranted.