Is Toke Cheap? Correspondence between Cannabis Demand and Purchase in the Laboratory

Abstract

Background and Aims:

The typical Marijuana Purchase Task (MPT) assesses hypothetical demand (i.e., relative reinforcing value) for cannabis across escalating prices. Cannabis demand has been related to use frequency, craving, cannabis use disorder symptoms, and cue-exposure response, among other outcomes. However, MPT performance for hypothetical consumption has yet to be examined in relation to in vivo behavior in the laboratory, a critical step in its validation.

Methods:

Individuals endorsing cannabis use at least twice weekly (n=92) participated in a laboratory cannabis self-administration study. Participants completed MPTs for Hypothetical and Actual cannabis. One trial (i.e., amount purchased at specified price) was randomly selected from the Actual MPT and participants with non-zero value trials (n=81) were permitted to smoke up to that amount during a 1-hour session in the laboratory.

Results:

Bivariate Pearson correlations demonstrated that cannabis consumption preferences were highly similar across the Hypothetical and Actual MPT at the price (rs = .45–.81) and index (rs = .46–.81) level. However, mean O_max (i.e., maximum expenditure), P_max (i.e., price at maximum expenditure), and breakpoint (i.e., price suppressing consumption to zero), were significantly higher (ds = .47–.51), and elasticity (i.e., consumption decline rate relative to price increase) was significantly lower (d = −.58), on the Actual MPT; this was also evident at certain price points. Self-reported anticipated consumption was significantly related to the amount of cannabis smoked during self-administration (R² = .66; p<.001) and was not moderated by price of the randomly selected trial.

Conclusions:

Correspondence was high between Hypothetical and Actual MPT performance, although small magnitude differences were present at both the index and price level. Demand for cannabis was often higher on the Actual MPT, suggesting that, rather than overestimating hypothetical cannabis demand, individuals underestimated preferences for actual cannabis outcomes. Despite this, the differences, are unlikely to be clinically meaningful. The robust relationship between anticipated consumption and actual cannabis quantity smoked in the laboratory suggests individual self-report accurately predicts subsequent self-administration, further supporting the construct validity of hypothetical MPTs. These findings support the validity of MPTs, suggesting that hypothetical state versions are broadly valid measures of cannabis demand.

Introduction

Behavioral economic theories integrate and synthesize principles from psychology and economics into a lens through which substance use can be understood (1). The hallmark of behavioral economics is its utility in understanding behavioral choice and may be applied to the decision to use or refrain from using substances (2). Within such a framework resides the relative value one places on a given reinforcer relative to other commodities, income, and time, among other factors (3). Behavioral economic demand quantifies a substance’s relative value by ascertaining how much one is willing to pay for a given substance under certain pre-specified conditions, and represents the association between substance consumption and its cost (4).

Historically, experimental paradigms have been used to investigate substance demand wherein participants must allocate resources (e.g., lever press, money, time) to receive a preferred reinforcer (5). Such paradigms are highly effective in quantifying substance-related motivation. However, practical and methodological limitations like operational costs associated with human laboratory drug administration research, high participant burden, and ethical and legal considerations around substance self-administration (e.g., inability to enroll adolescent populations or individuals seeking treatment) remain critical barriers to this line of research (6). To circumvent these aforementioned issues, hypothetical purchase tasks were developed to assess how much of a particular substance individuals would hypothetically purchase and use across escalating costs (e.g., money, time, effort) in various scenarios, thus no substance administration or consumption is required (7,8). The hypothetical nature of these tasks facilitates their administration to individuals endorsing substance use, including those belonging to vulnerable groups (9–12). Such hypothetical purchase tasks have been developed for a variety of substances, including alcohol (8), tobacco (13), and more recently, cannabis (14,15). Responses on hypothetical purchase tasks are typically graphed using demand curve analysis (16), translating the association between cost and purchase into different indices of substance demand: intensity (i.e., consumption absent of cost), O_max (i.e., maximum expenditure across price), P_max (i.e., price at maximum expenditure), breakpoint (i.e., price at which consumption is suppressed to zero), and elasticity (i.e., rate of consumption decline relative to price increase). Importantly, metanalytic work has demonstrated that substance purchase tasks are largely valid and reliable representations of consumer preference for and relative reinforcing value of a variety of substances (for reviews, see 17–22).

Behavioral economic cannabis demand research has expanded greatly since its inception (14). Studies in this growing area have demonstrated that cannabis demand is a robust indicator of risk for cannabis involvement (17), and may be used to differentiate those with cannabis use disorder (i.e., CUD; 23). Indices from the Marijuana Purchase Task (MPT) have been associated with a panoply of risk-related cannabis use behaviors, including preferences for elevated potency (24), cannabis craving (25), cannabis-related problems (26), CUD symptoms (27–30), unsafe behavior while under the influence (e.g., driving; 31,32), and use frequency (14,23), among others (for reviews, see 21,22,33,34). Experimental research in the laboratory utilizing the MPT suggests cannabis demand can be influenced by exposure to cannabis cues (25,35), and research in the field utilizing ecological momentary assessment has demonstrated that cannabis demand varies day-to-day in response to internal and external stimuli (36), underscoring the mutable nature of substance demand.

While demand for various substances is cross-sectionally and prospectively associated with substance use and related negative consequences (e.g., 21,37), determining the concordance between hypothetical and actual substance demand is a crucial endeavor. A small but growing literature has begun to examine the relation between hypothetical and actual purchase task performance for alcohol (6,38) and tobacco (39,40) using experimental laboratory paradigms. For example, in two separate studies, adults endorsing heavy episodic drinking completed two versions of an alcohol purchase task (APT) based on hypothetical or actual outcomes (6,38). Participants were allocated a $15 bar tab and were informed that a randomly selected price from their APT responses would determine the cost and number of mini-alcoholic beverages ($0.00-$15.00/drink) available during a subsequent alcohol administration session. Results indicated considerable overlap between choices for hypothetical and actual drinks, as well as alcohol demand indices. Importantly, there were large-effect relations between the number of available alcohol mini-drinks and the number of drinks actually consumed during the experimental session. These approaches have also been extended to hypothetical and actual tobacco purchasing and smoking behaviors. MacKillop and colleagues (39) used similar experimental methods involving a $10 monetary tab for tobacco cigarettes during a self-administration session based on participants’ prior responses on two cigarette purchase tasks linked to the behavioral outcome. In this case, participants smoked over 80% of available cigarettes and the number of available cigarettes was significantly associated with the number smoked, suggesting that participant estimates of their consumption levels are largely consistent with their actual behavior.

Despite the documented links between hypothetical and actualized alcohol and tobacco rewards, no research to date has examined the association between hypothetical (i.e., not linked to money and/or cannabis) and actual (i.e., linked to actualized money and/or cannabis) demand for cannabis on an MPT. Moreover, the MPT may be a useful tool to inform cannabis policy (41) considering that cannabis legislation is rapidly changing while public perceptions of cannabis are shifting towards greater acceptance (42,43). Equally important, it is essential to demonstrate that demand for cannabis on a hypothetical MPT is a strong representation of how behavior would manifest when such choices will result in real rewards and consequences. Accordingly, the present investigation aimed to assess the relation between hypothetical and actual cannabis demand, and subsequently, test whether cannabis demand on the MPT accurately represented decisions to purchase and use cannabis when such rewards were actualized.

Methods

Participants

Individuals endorsing regular cannabis use were recruited between 2018–2020 from the community in Rhode Island and Massachusetts via flyers and social media websites to participate in an experimental cannabis administration study. Potential participants were screened via phone for eligibility. Participants met the following inclusion criteria: 18 to 50 years of age, cannabis use ≥twice weekly in the past month on average and ≥monthly for the past 6 months, English-speaking, positive urine toxicology screen for cannabis and negative screen for drugs other than cannabis, non-treatment seeking, purchase of cannabis ≥twice in the past 6 months, negative pregnancy test and use of contraception for those who can become pregnant, not nursing, good physical health as determined by physical examination, zero breath alcohol concentration, absence of diagnosis of current depression, mania, psychosis, and panic disorder, no self-report of past serious adverse reaction to cannabis, no current psychotropic drug use, smoking 0–20 tobacco cigarettes per day, ability to abstain from cannabis for 15 hours, and body mass index in 18.5–30 kg/m2 range. At the time of data collection, cannabis was legal for medical use in Rhode Island and for medical and recreational use in Massachusetts. Participants were excluded if they reported having a current medical cannabis registration card due to the impact of card possession on cannabis access, cost, and legal ramifications. This research was approved by the Brown University Institutional Review Board and the study was registered on ClinicalTrials.gov (NCT03518567) prior to commencement, however, the full analytic plan was not preregistered.

Measures

Demographics.

Participants completed a baseline demographic measure assessing sex, gender, race, ethnicity, age, socioeconomic status, and other sociodemographic variables.

Substance use.

Participants completed the Daily Sessions, Frequency, Age of Onset, and Quantity of Cannabis Use Inventory (i.e., DFAQ-CU; 44) to assess cannabis use behaviors including age of onset of cannabis use, typical cannabis use quantity, typical mode of self-administration, and monthly monetary cannabis expenditures. Participants also underwent a Cannabis Timeline Follow-Back Calendar interview (45) for the past 60-days from baseline to assess cannabis use frequency and formulation.

Cannabis Use Disorder (CUD) symptoms.

The Structured Clinical Interview for DSM-5 (SCID-5) Research Version (46) was administered to assess CUD symptoms endorsed over the past 12-months.

Cannabis demand.

Participants completed a state Hypothetical MPT measure based on the trait MPT developed by Aston and colleagues (23) and the state MPT validated in a laboratory-based cannabis cue-reactivity paradigm by Metrik and colleagues (25). Participants were provided with an instructional vignette describing stipulations associated with hypothetical purchasing and consumption of cannabis. Participants were informed that this task version was hypothetical and were instructed to indicate the number of hits they would purchase across 22 escalating prices. Participants subsequently completed a state Actual MPT. The instructional vignette was identical to that of the Hypothetical MPT except that participants were informed that responses would be linked to actual cannabis and money in the laboratory (see supplemental materials). Participants were again instructed to indicate the number of hits they would purchase across the same 22 price points. Participants were allotted a monetary budget (i.e., $20) based on the maximum price per hit on the Actual MPT. Five metrics of cannabis demand were obtained from each MPT: intensity, O_max, P_max, breakpoint, and elasticity.

Procedure

Initially eligible participants attended the laboratory for a baseline session to confirm eligibility and provide informed consent prior to participation commencement. A positive urine THC screen was required to confirm cannabis use status, however, positive results for other substances resulted in ineligibility. Participants who could become pregnant completed a pregnancy test. Participants were administered a medical screening questionnaire and underwent a physical exam. Participants completed a structured clinical interview (SCID-5) to rule out depression, mania, psychosis, and panic disorder. Ineligible participants were paid $10. Eligible participants completed baseline measures and were scheduled for the experimental cannabis administration sessions (Session 1: cannabis purchase and use via modified version of a controlled paced puffing smoking procedure; Session 2: ad libitum smoking topography). Findings from Session 1 are presented herein.

Prior to experimental sessions, participants were asked to refrain from all cannabis and tobacco smoking for 15 hours, alcohol for 24 hours, and caffeine for 4 hours. Upon arrival to the experimental session, participants were required to provide a 0.00 g/dl alcohol breathanalysis reading via an Alco-Sensor IV breathalyzer (Intoximeters, Inc., St Louis, MO., USA) and an alveolar carbon monoxide (CO) reading ≤8ppm via a Bedfont Scientific Smokelyzer^® to confirm absence of recent smoking (47–49). Participants endorsing tobacco use were permitted to smoke a cigarette following the CO test to prevent nicotine withdrawal (49).

The experimental session occurred in a ventilated smoking room equipped with a one-way mirror. The cannabis administration methodology (i.e., cannabis purchase and use via a modified paced puffing smoking procedure) was based on established procedures that have been successfully used to assess demand for hypothetical versus actual alcohol (6). Participants first completed the Hypothetical (i.e., not associated with cannabis available to smoke in the laboratory) MPT, followed by the Actual MPT (i.e., identical to Hypothetical MPT, but a response choice was selected at random to determine amount of cannabis available to smoke in the laboratory), and self-report measures of craving and subjective effects.

Selection of cannabis amount available to smoke:

A response choice (i.e., number of cannabis hits) corresponding to an item price was randomly selected by the participant from the Actual MPT by selecting a poker chip from a bowl that corresponded to the price in one of the 22 MPT items (e.g., How many hits of marijuana would you smoke if they were $1 each?). Once the response choice was selected, the participant was informed of the corresponding number of cannabis hits available to smoke (i.e., the maximum number of hits available to purchase at the selected price per hit within the budget), that the administration session would last 1-hour, and that they were not required to smoke the available cannabis. Participants who randomly selected a response choice corresponding to zero hits (i.e., items wherein the participant indicated they would not purchase cannabis) were not given the opportunity to smoke but remained in the laboratory for the duration of the 1-hour smoking session plus 1 additional hour to ensure duration of time in the laboratory was unrelated to MPT responding.

Modified controlled paced-puffing procedure:

Blood pressure and heart rate were continuously assessed during the entire experimental session for safety monitoring. Participants were then permitted to purchase and smoke cannabis hits ad libitum during the session using a modified controlled paced-puffing procedure (48,50). They were instructed to raise their hand to indicate they wished to order a cannabis hit. The price of the hit was subsequently deducted from their budget. Research staff then initiated a recording instructing the participant to “light the cigarette,” “get ready” (5s), “inhale” (5s), “hold smoke in lungs” (10s), and “exhale.” This modified paced puffing procedure was implemented to best control the duration of each hit across participants. This procedure was repeated until the participant reached the number of cannabis hits available to smoke or chose to cease smoking. At the end of the 1-hour cannabis administration period, participants received any remaining money from the budget that was not used to purchase cannabis (e.g., budget = $20, item selected from Actual MPT = $5/hit, participant reported they would purchase 3 hits at this price, 3 units were available during the smoking procedure, 2 of the 3 available hits were smoked, participant subsequently received the remaining $10 from the budget).

Sobriety assessment:

Participants who consumed cannabis remained in the laboratory 3-hours following the beginning of the smoking procedure as psychotropic effects of smoked cannabis dissipate within 2–3 hours (51) and subsequently completed a field sobriety test (52) before being transported home. Participants were compensated up to $100 for completing the session.

Drug:

Cannabis cigarettes (5.9% THC) were provided by the National Institute on Drug Abuse Drug Supply Program, rolled at both ends, stored frozen, and humidified for 24-hours pre-administration.

Data Analysis Plan

Raw cannabis demand data were evaluated using criteria from Stein and colleagues (53). No participants failed for violations of bounce (i.e., frequent price-to-price increases in consumption) or reversal from zero (i.e., positive value after a reported zero) criteria. One participant failed for trend criteria (i.e., overall reduction in responding) due to reporting zero demand at all timepoints on both MPTs and was subsequently removed from demand analyses. Raw demand data were examined for missingness, outliers (Z > 3.29), and distributional properties (54). There were no extreme raw values or reversals in either MPT version. Elasticity from both MPT versions was positively skewed; logarithmic transformations reduced skew to acceptable levels. Observed demand indices (i.e., intensity, O_max, P_max, breakpoint) were calculated from raw MPT data using Excel spreadsheets with automated formulae. Elasticity was derived by fitting individual demand curves in GraphPad Prism using an exponentiated approach to Hursh & Silberberg’s (55) exponential demand equation (56):

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

where Q = quantity consumed, Q₀ = derived intensity, k = constant denoting range of dependent variable (i.e., cannabis hits), α = elasticity (i.e., rate constant determining the decline rate in log consumption based on increases in price), and C = commodity cost (USD). K was determined by subtracting the log10-transformed average consumption at the highest price ($10.00) from the log10-transformed average consumption at the lowest price used in curve fitting ($0.01) on both MPT versions and averaging the resulting values. The K parameter used in analyses was 1.22. An R² value was generated to reflect percentage of variance accounted for by the exponentiated equation (i.e., adequacy of model fit to the data). One participant reported zero demand on the Hypothetical MPT thus elasticity could not be generated for this individual. As such, they were removed from demand analyses (i.e., correlations, t-tests) but retained in regression analyses.

Pearson product–moment correlations assessed the correspondence between hypothetical and actual MPT performance at both the individual price- and index-level, as well as associations between the number of available cannabis hits and the actual number of hits participants smoked during the 1-hour ad libitum cannabis self-administration session. Paired samples t tests assessed whether sample mean cannabis consumption and expenditure differed between Hypothetical and Actual MPTs across the 22 individual prices and five demand indices.

Next, a hierarchical multiple regression was used to predict the number of cannabis hits consumed in four variable groupings using data from participants who had the opportunity to smoke cannabis during the session. Model step 1 covaried for annual individual income. Step 2 included the number of hits available to smoke for each participant based on their randomly selected response on the Actual MPT. Step 3 included the price corresponding to the participant’s randomly selected response. Step 4 included a response by price interaction term to examine whether the number of available hits and actual cannabis self-administration differed based on the price per hit. The variables included in the interaction term were grand mean centered.

To examine whether differences between Hypothetical and Actual MPT demand indices (i.e., O_max, P_max, breakpoint, intensity, elasticity) were smaller than a meaningful effect size, post hoc equivalence tests were conducted using two one-sided tests (i.e., TOST; 57). First, 90% confidence intervals (i.e., 100% - 2[α]) were built around the effect size difference. Next, lower and upper bounds were compared to the equivalence region that could provide 80% power to detect equivalence, similar to equivalences tests used in prior behavioral economic research (e.g., 27). The equivalence region was Cohen’s dz = 0.262, which constituted a small-to-medium effect size based on a sensitivity power analysis conducted in G*Power (58) with a type I error rate of α = .05.

Bonferroni corrections due to multiple testing were applied, thus statistical significance for each analysis type was as follows: p < .05 for bivariate analyses, p < .002 for paired samples t-tests with individual MPT prices (i.e., 0.05/22), p < .01 for paired samples t-tests with demand indices (i.e., 0.05/5), and p < .05 for regression analyses. Deidentified study data are available from the corresponding author upon reasonable request.

Results

Preliminary Results

Sociodemographic information and sample descriptives are presented in Table 1. As expected, participants with (versus without; n=11) the opportunity to smoke cannabis reported significantly higher demand for all indices, except intensity, on the Actual MPT (see supplementary materials). In general, cannabis demand was prototypical as evidenced by decreased consumption in response to escalating price on both MPTs (see Figure 1). Likewise, expenditure conformed to an inverted U-shaped function on the Hypothetical MPT, whereas prices increased initially but plateaued between $1–10 per hit on the Actual MPT (see Figure 2). The exponentiated demand curve model provided an excellent fit to the data (Hypothetical MPT: median R² = 0.92, interquartile range [IQR]: 0.88–0.93; Actual MPT: median R² = 0.93, IQR: 0.90–0.94).

Table 1.

Sample descriptive statistics

	Full sample (n=92)	Opportunity to smoke during ad libitum session (n=81)
Sociodemographics
Age (years)	23.08 (4.98), 18 – 37	23.27 (4.91), 18 – 37
Sex (% female)	39 (42.39)	36 (44.44)
Race
Asian	6 (6.52)	6 (7.41)
Black/African American	15 (16.30)	13 (16.05)
White	57 (61.97)	50 (61.73)
Other	4 (4.35)	3 (3.70)
>1 race	10 (10.87)	9 (11.11)
Ethnicity (% Latine)	18 (19.57)	15 (18.52)
Education level
High school diploma	5 (5.43)	5 (6.17)
Some college	63 (68.48)	54 (66.66)
Bachelor’s degree or higher	24 (26.09)	22 (27.16)
Employment (% employed)	41 (44.57)	36 (44.44)
Individual annual income (USD)
≤$19,999	64 (69.57)	54 (66.66)
$20,000 – 39,999	17 (18.48)	16 (19.75)
$40,000 – 59,999	8 (8.70)	8 (9.88)
≥$60,000	3 (3.26)	3 (3.70)
Cannabis use
Age of regular cannabis use (years)	18.79 (3.07), 13 – 29	18.91 (3.18), 13 – 29
DSM-5 CUD symptomsa	4.23 (2.29), 0 – 10	4.10 (2.12), 0 – 10
Past-month cannabis use days/week	5.05 (1.61), 2 – 7	5.09 (1.59), 2 – 7
Past-month weekly cannabis amount (grams)	4.44 (5.08), 0.13 – 28	4.51 (5.32), 0.13 – 28
Cannabis demanda
MPT – Hypothetical
Omax	14.81 (6.24), 1.20 – 20	15.56 (5.94), 1.20 – 20
Pmax	4.77 (4.00), 0.30 – 10	5.25 (4.03), 0.30 – 10
Breakpoint	6.61 (3.56), 0.75 – 10	7.12 (3.42), 1.25 – 10
Intensity	16.24 (5.14), 3 – 20	15.98 (5.12), 5 – 20
Elasticity	0.020 (0.022), 0.005 – 0.145	0.019 (0.021), 0.005 – 0.145
MPT – Actual
Omax	17.18 (4.72), 2.50 – 20	18.33 (3.60), 6.75 – 20
Pmax	6.58 (3.88), 0.50 – 10	7.28 (3.59), 0.75 – 10
Breakpoint	8.34 (2.90), 1.25 – 10	9.04 (2.20), 1.25 – 10
Intensity	16.93 (4.89), 3 – 20	16.73 (4.86), 4 – 20
Elasticity	0.013 (0.013), 0.005 – 0.112	0.011 (0.007), 0.005 – 0.050
Ad libitum self-administration session
Participants with opportunity to smoke	-	81 (88)
Total number of hits available	-	9.91 (5.55), 2 – 20
Total number of hits smoked	-	8.02 (5.09), 1 – 20
% available hits smoked	-	0.84 (0.25), 0.13 – 1.00

Notes: Values represent mean (sd), sample range or n (%).

^aBased on n=91 and n=80 for full sample and subsample with opportunity to smoke during ad libitum session, respectively.

Figure 1.

Sample mean demand across actual and hypothetical Marijuana Purchase Task assessments. Individual data points represent mean (± standard error of the mean). Solid lines denote maximum allowable hypothetical consumption at each price. Mean self-reported consumption significantly differed across MPTs at $2.00 and from $2.50–$10.00 (p<.002) based on paired samples t tests when applying Bonferroni corrections for multiple testing (see Table 2). One participant was removed due to non-systematic demand data resulting in n=80.

Figure 2.

Sample mean expenditure across Actual and Hypothetical Marijuana Purchase Task assessments. Individual data points represent mean (± standard error of the mean). Mean self-reported expenditure significantly differed across MPTs at $2.00 and from $2.50–$10.00 (p<.002) based on paired samples t tests when applying Bonferroni corrections for multiple testing. One participant was removed due to non-systematic demand data resulting in n=80.

Correspondence Between Hypothetical and Actual MPT Outcomes

Mean differences and bivariate correlations between corresponding demand indices and individual prices on the Actual and Hypothetical MPTs are presented in Table 2. Based on contemporary effect size interpretations (59), there were statistically significant, medium-to-large effect bivariate associations between all corresponding demand indices and individual prices across both versions of the MPT. Sample mean cannabis consumption and expenditure (other than at zero cost) at individual prices across both MPT versions did not significantly differ from $0.00-$0.40 and at $0.75 (ps ≥ .002; Cohen’s ds: 0.22–0.34). There were statistically significant, small-to-medium effect differences (ps ≤ .001; Cohen’s ds: 0.39–0.57) between actual and hypothetical consumption at $0.50 and from $1.00-$10.00 (see Table 2, Figure 1, and Figure 2). Additionally, all other demand indices were significantly higher on the Actual (versus Hypothetical) MPT with the exception of intensity (see Table 2). Similar to statistically significant mean differences across demand indices (see Table 2), post hoc equivalence tests comparing Actual and Hypothetical MPT indices further confirmed that cannabis demand indices did not demonstrate statistical equivalence (see Supplemental Figure 1). In the full sample, 89% (n = 81) and 78% (n = 71) reported purchasing the number of cannabis hits that corresponded to the budgetary limit at least once across the Actual and Hypothetical MPTs. Moreover, the percentage of responses that corresponded to the budgetary limit across all prices ranged from 15.4% ($1.25) to 59.3% ($0.00) on the Hypothetical MPT and from 25.3% ($1.25/hit) to 65.9% ($0.00) on the Actual MPT.

Table 2.

Hypothetical and Actual Marijuana Purchase Task performance

	Actual MPT			Hypothetical MPT			ra	Mean Difference [95% CI]b	dc
Demand Index
Intensity	16.93			16.24			.81	0.69 [0.04, 1.34]	0.22
Omax	17.18			14.81			.60	2.38 [1.31, 3.44]	0.47
Pmax	6.58			4.77			.52	1.81 [1.00, 2.61]	0.47
Breakpoint	8.34			6.61			.46	1.73 [1.02, 2.44]	0.51
Elasticityd	−1.98			−1.83			.67	-0.14 [−0.19, −0.09]	−0.58
	Expenditure	Hits	Sample Range	Expenditure	Hits	Sample Range
Individual Price
$0.00	$0.00	16.93	3 – 20	$0.00	16.24	3 – 20	.81	0.69 [0.04, 1.34]	0.22
$0.10	$1.66	16.60	3 – 20	$1.57	15.69	3 – 20	.74	0.91 [0.14, 1.68]	0.25
$0.20	$3.22	16.12	3 – 20	$3.07	15.34	3 – 20	.77	0.78 [0.04, 1.52]	0.22
$0.30	$4.77	15.89	3 – 20	$4.51	15.03	3 – 20	.75	0.86 [0.07, 1.65]	0.23
$0.40	$6.22	15.56	3 – 20	$5.71	14.26	2 – 20	.75	1.30 [0.50, 2.09]	0.34
$0.50	$7.57	15.13	2 – 20	$6.76	13.52	2 – 20	.76	1.62 [0.79, 2.45]	0.41
$0.75	$10.52	14.03	2 – 20	$9.30	12.41	0 – 20	.65	1.63 [0.59, 2.66]	0.33
$1.00	$13.02	13.02	2 – 20	$11.07	11.07	0 – 20	.66	1.96 [0.90, 3.01]	0.39
$1.25	$13.75	11.00	0 – 16	$11.18	8.95	0 – 16	.67	2.06 [1.21, 2.91]	0.50
$1.50	$14.01	9.34	0 – 13	$11.37	7.58	0 – 13	.66	1.76 [1.04, 2.48]	0.51
$1.75	$14.44	8.25	0 – 11	$11.48	6.56	0 – 11	.63	1.69 [1.04, 2.35]	0.54
$2.00	$15.21	7.60	0 – 10	$11.67	5.84	0 – 10	.57	1.77 [1.10, 2.44]	0.55
$2.25	$13.40	5.96	0 – 8	$10.46	4.65	0 – 8	.48	1.31 [0.69, 1.92]	0.44
$2.50	$14.56	5.82	0 – 8	$10.93	4.37	0 – 8	.48	1.45 [0.82, 2.08]	0.48
$2.75	$13.96	5.08	0 – 7	$10.18	3.7	0 – 7	.53	1.37 [0.84, 1.91]	0.54
$3.00	$13.35	4.45	0 – 6	$9.36	3.12	0 – 6	.50	1.33 [0.84, 1.82]	0.57
$3.25	$13.86	4.26	0 – 6	$9.86	3.03	0 – 6	.52	1.23 [0.75, 1.71]	0.53
$3.50	$12.81	3.66	0 – 5	$8.81	2.52	0 – 5	.50	1.14 [0.72, 1.57]	0.56
$4.00	$13.76	3.44	0 – 5	$9.01	2.25	0 – 5	.48	1.19 [0.74, 1.63]	0.56
$6.00	$12.07	2.01	0 – 3	$7.98	1.33	0 – 3	.45	0.68 [0.39, 0.97]	0.49
$8.00	$10.46	1.31	0 – 2	$6.77	0.85	0 – 2	.51	0.46 [0.27, 0.65]	0.50
$10.00	$12.20	1.22	0 – 2	$7.80	0.78	0 – 2	.52	0.44 [0.25, 0.63]	0.49

Notes: One participant was removed due to non-systematic demand data thus results are based on n=91; CI = Confidence Interval;

^aPaired samples correlations between demand indices and individual price points are all statistically significant at p<.001;

^bBolded values are statistically significant at p<.002 for paired samples t tests for Actual and Hypothetical Hits;

^cEffect size based on Cohen’s;

^dElasticity values for the Actual and Hypothetical MPT were logarithmic transformed. Demand indices and individual prices denote sample mean values. Sample ranges on Actual and Hypothetical MPTs were restricted to 0–20 based on maximum consumption limits and to correspond with the $20 tab available to participants during cannabis administration.

Correspondence Between Estimated and Actual Cannabis Smoked

Based on study procedures, 88% of the sample (81/92 participants) had the opportunity to smoke during the self-administration session; 12% of the sample (11/92 participants) randomly selected a response choice corresponding to zero cannabis hits. Of the participants who had the opportunity to smoke, all self-administered at least one hit (M hits available = 9.91, SD = 5.55, range = 2 – 20). On average, participants smoked 83.8% of the cannabis that was provided at the price associated with their choice based on the randomly selected response (M hits self-administered = 8.02, SD = 5.09, range = 1 – 20; see Table 1). There was a large magnitude bivariate relation between the number of hits available and the number actually smoked (r = 0.81, p < 0.001; see Figure 3).

Figure 3.

Relations between estimated cannabis consumption at a given price (hits available) and actual cannabis consumption during the 1-hour ad libitum self-administration session. The number of participants who were able to self-administer cannabis during the session was n=81. The number of participants represented by each data point is indicated by superscripts.

Hierarchical Regression Model Predicting Cannabis Self-Administration

Individual annual income and response price were not associated with ad libitum cannabis self-administration at first entry or in the final model (ps ≥ .07). Similar to bivariate correlations, the number of available hits significantly predicted the actual number of cannabis hits participants smoked (ps < .001). Likewise, the effect of available hits on actual cannabis self-administration was not moderated by price per hit (p = .43; see Table 3).

Table 3.

Hierarchical regression model using anticipated cannabis consumption to predict actual cannabis smoked

		First entry into model					Final model
Step	Predictor	B [95% CI]	SE	β	p	Adj. R2	B [95% CI]	SE	β	p	R2 Change
1	Individual incomea	.55 [−0.05, 1.15]	.30	.19	.07	0.03	.20 [−0.12, 0.52]	.16	.07	.22	0.04
2	Number of hits availableb	.75 [0.66, 0.85]	.05	.85	< .001	0.74	.81 [0.63, 0.99]	.09	.91	< .001	0.71
3	Pricec	.03 [−0.31, 0.36]	.17	.01	.88	0.73	.30 [−0.47, 1.08]	.39	.13	.44	0.00
4	Hits available*priced	-	-	-	-	0.73	.03 [−0.05, 0.12]	.04	.10	.43	0.00

Notes: Based on analytic sample of n=81. Adj. R² = Adjusted R² value corresponding to each step in the hierarchical regression. R² change = denotes the amount of variance explained in the outcome based on the inclusion of the predictor variables at each step.

^aTotal individual annual income in USD;

^bNumber of maximum cannabis hits available based on participants’ randomly selected response on the MPT-Actual Version;

^cPrice corresponding to participants’ randomly selected response on the MPT-Actual Version;

^dBoth variables grand mean centered to create interaction term.

Discussion

The present investigation evaluated the correspondence between performance on a Hypothetical and Actual MPT, as well as the correspondence between MPT performance and actual cannabis subsequently smoked in the laboratory. Performance on both MPTs was highly correlated, however, significant mean differences were found at the index and price level, albeit of modest magnitude. Additionally, estimated consumption on the Actual MPT predicted cannabis smoked in the laboratory when rewards were actualized. As such, this work makes a critical contribution to the literature demonstrating that hypothetical MPTs are valid representations of how cannabis use behaviors manifest when rewards are real.

Similar to previous work documenting the high correspondence across price and index on hypothetical and actual versions of the APT (6,38), the present study demonstrated high correspondence across MPT versions. Yet, significant mean differences were found at many price points and on nearly all indices, with demand on the Actual MPT always being significantly greater. Notably, these differences translated to nominal monetary (i.e., less than $1) and cannabis (i.e., less than one hit) amounts, and are therefore not considered to be substantively meaningful. Units available for purchase in previous work have been larger (i.e., mini-drinks, cigarettes; 6,39) in comparison to units utilized in this work (i.e., hits). Consequently, if smaller units were available in previous work (e.g., sips, puffs), significant mean differences may have been identified as well. In this regard, one previous study on tobacco cigarette demand did demonstrate significantly higher tobacco demand on an actual versus hypothetical task version (40), aligning with the current results. Moreover, while hits appeared to serve as an effective unit of purchase in the laboratory, previous qualitative work has suggested that grams are superior to hits when assessing cannabis demand (15). Importantly, grams were not used in the present study as such a unit is not amenable to self-administration in a single laboratory session and would likely have contributed to ceiling effects across the sample, as most individuals do not smoke more than one gram during a given session (60).

Results from the present study have implications that are germane to clinical, public health, and policy applications. First, given the brevity of purchase tasks and ability to manipulate instructional sets, MPTs can be a useful tool in clinical research and public health settings to distinguish individuals who may be at elevated risk for cannabis-related harms. Equally important, cannabis’ evolving legal status in the United States necessitates empirically-supported policies to mitigate potential harms. Indeed, results from the present study can inform future regulatory policies and lend further support to prior behavioral economic research that has identified how pricing legal cannabis can shift individuals away from utilizing illegal cannabis markets (28,41,61). Our results indicate that hypothetical MPTs are an efficient and generally valid indicator of how much cannabis a person would be expected to consume, increasing confidence in their use in various contexts.

Despite the strengths of the present investigation, findings should be considered in the context of limitations. First, due to the possibility of low demand on randomly selected prices, some participants did not have the opportunity to smoke in the laboratory and thus were omitted from regression analyses. Permitting responses reflecting low demand is a critical component in the validation of any purchase task as these reflect real possibilities that individuals face in the field. Second, this study aimed to determine if participants would smoke up to the amount of cannabis they indicated on the MPT, however, participants were not able to smoke more than the amount selected, nor were participants able to smoke at all if the amount of cannabis randomly selected from the MPT corresponded to zero. Thus, it is unclear whether participants would have elected to exceed the amount they anticipated smoking if more cannabis had been available. Third, similar to previous work validating the APT (6), cannabis prices at the high end restricted the number of cannabis hits participants could have purchased due to potential safety concerns associated with not placing appropriate limits on substance use in the laboratory. However, in the multivariate model, price did not moderate the relation between anticipated and actual cannabis use. Fourth, the Hypothetical MPT always preceded the Actual MPT, consistent with previous studies validating hypothetical APTs (6,38), thus potential order effects could not be examined. Moreover, as part of the procedure, participants were informed that a response option would be randomly selected from the Actual MPT and would subsequently be available for purchase and use in the laboratory, thus it is possible that this knowledge impacted MPT performance. Despite this, the elevations in demand on the Actual MPT were of modest effect size and represented nominal amounts of cannabis. Fifth, the present study enrolled a sample of individuals endorsing regular cannabis use (i.e., ≥twice weekly). As such, it is unclear how results might differ for samples reporting less frequent use patterns, or for individuals seeking treatment for severe CUD. Sixth, the time allotted for cannabis administration in the laboratory was one hour. Consequently, it is unclear how longer or shorter smoking durations would have impacted self-administration. Finally, the data analytic plan was not preregistered thus findings should be interpreted as exploratory.

Despite preferences being comparable on both MPT versions, which in turn predicted actual behavior in the laboratory, participants smoked cannabis in an isolated laboratory environment. Yet, substance use often takes place in social contexts (62) and prior research demonstrates cannabis use can increase social interactions in small group settings (63). In this regard, Acuff and colleagues (64) demonstrated that the presence of peers tends to increase cannabis demand using a social context MPT vignette manipulation. Subjective drug responses and substance use behaviors may be context-specific (e.g., 65). For example, in one 12-day residential study that implemented counterbalanced work and social-access periods, six males reporting frequent cannabis use received up to eight cannabis cigarettes to smoke throughout each day (66). While half of the sample smoked more cannabis during social periods, the others smoked more cannabis in the mornings irrespective of activity. Thus, future research would benefit by testing the concordance between hypothetical and actual demand when participants can self-administer cannabis alone and in small-group settings. Similarly, many real-world contexts involving cannabis use typically occur in proximity to other substances, such as alcohol (e.g., parties, bars). However, in the present study, participants did not have access to other reinforcers (e.g., alcohol, nicotine, other cannabis formulations) during the 1-hour smoking period. As such, another important extension of this research will be to determine the validity of cross-commodity demand when the price of one commodity is fixed while the other escalates in price. Given the rise in simultaneous alcohol and cannabis use (67,68) and increasingly potent novel cannabis products (69,70), a better understanding of how hypothetical decision-making involving multiple substances aligns with real-world behaviors is paramount for prevention and intervention efforts.

Collectively, the present findings demonstrate high correspondence between preferences for hypothetical and actual cannabis on a MPT, albeit with significant, modest magnitude differences indicating the two measures are not completely interchangeable, and that demand for cannabis accurately forecasts consumption behavior in contexts where real cannabis is available. As such, this study represents a critical step in validating the MPT and leveraging its application to understand cannabis use and misuse.