How tax structures for retail cannabis shape cannabis use among youth and young adults: evidence from a volumetric choice experiment

Abstract

Background

Despite ongoing debates about cannabis regulation, little is known about how tax policy design influences cannabis use among U.S. adolescents and young adults (AYAs). With states adopting diverse taxation schemes based on weight, price, or product potency, evaluating how these approaches affect consumption is critical for evidence-based policymaking.

Aim

This study uses a split sample volumetric choice experiment (VCE) to examine how variations in pretax prices, tax bases, tax rates, and THC levels influence both the amount of cannabis consumed and the overall intake of THC among US AYAs aged 15-20. We also estimate the own-price elasticities for four cannabis forms (legal flower, illegal flower, edibles, and cartridges) and cross-price elasticities between these products.

Methods

We use a nationally representative sample of 1,100 AYA who reported current use or susceptibility to use and completed a set of hypothetical purchase tasks featuring choices among four product forms (legal flower, illegal flower, edibles, cartridges). Respondents were randomized to three tax bases: weight-, price-, or potency-based taxation, where pre-tax price levels, tax rates, and THC levels additionally vary. We estimated own- and cross-price elasticities and assessed how AYA current and susceptible users adjusted cannabis consumption in response to the varying attributes using zero-inflated negative binomial and fixed effects models.

Results

Higher pre-tax prices and elevated tax rates significantly reduce both purchase quantities and total THC intake among current and susceptible AYA users. Products with higher THC levels increased THC intake but did not affect quantity consumption, suggesting that these users continue to purchase similar quantities but opt for more potent products. Compared to weight-based taxation, potency-based taxes (i.e., THC) were associated with a 30%-32% increase in cannabis quantity consumption. The price elasticities of cannabis demand were −0.3, with edibles being complements to other forms and illegal flowers being substitutes for legal flowers and cartridges. Compared to weight-based taxes, THC-based taxes significantly reduce the impact of THCs in increasing THC consumption. Finally, the impact of taxes on reducing THC consumption from legal products may be completely offset by shifting from legal to illegal products.

Conclusions

For AYAs who are using or susceptible to using cannabis, tax increases can effectively reduce the quantity and THC consumption, albeit a tax rate beyond 60% of pretax prices or equivalent generating no additional reductions. Compared to weight-based taxes, potency-based and price-based taxes reduce the consumption of high THC products, but potency-based taxes may also increase the consumption of low THC products. Given that higher prices on cartridges and edibles do not drive AYAs’ substitution using flowers, tiered tax structures may have the advantage over potency or weight-based taxes in balancing the need for targeting THC while reducing the burden to administer potency taxes. Finally, given that illegal flowers are substitutes for both legal flowers and cartridges, policy impacts are offset by the availability of illegal products.

Introduction

Cannabis is the most widely used illicit substance among adolescents and young adults (AYAs) in the U.S. [4, 21, 29, 38]. While federally illegal, the legal landscape of cannabis at the state and local levels has undergone significant transformation. As of April 2025, cannabis for medical use is legal in 38 U.S. states, with 24 states and the District of Columbia (DC) having additionally legalized recreational cannabis, up from only two states a decade earlier [28]. Other than DC and Virginia, states that have legalized recreational cannabis use have further established retail cannabis markets or have retail markets pending, making cannabis easily accessible to a majority of the U.S. population [6, 18, 21, 27, 28]. All the states with retail cannabis further impose excise taxes on these products, with the expectation that these taxes will reduce harmful cannabis use while generating tax revenues [65].

Unlike adult use of cannabis, which could have potential benefits in pain management, cannabis use at a young age can harm the developing brain and lead to lifelong addiction and adverse consequences (e.g., psychosis) [37]. While there is mixed evidence on whether cannabis legalization is associated with increased cannabis use among youth, concerns persist over the potential long-term impacts of cannabis access and use on adolescent health [17, 35, 51]. Studies have demonstrated that while recreational cannabis sales are restricted for AYAs younger than 21, a majority of AYA users reported buying cannabis from dispensaries or friends and family who have access to dispensaries [15, 69]. Since excise taxes are the most adopted policy to regulate retail cannabis, whether these taxes prevent the initiation and escalations of cannabis use among AYAs is of particular interest for policymakers and is the primary public health justification for taxing cannabis [20, 26, 30, 36, 37, 41, 46, 50, 70].

Economic theory further predicts that younger consumers who have limited budgets are more sensitive to cannabis price and taxes than older adults, making cannabis taxation a potentially powerful tool for use prevention. Using Monitoring the Future surveys, two studies estimated that a 10% increase in cannabis prices is associated with about 3% less cannabis use among youth [23, 43]. Additional research using observational surveys and behavioral purchasing experiments similarly shows that higher cannabis prices are associated with reduced AYA cannabis use [14, 17, 43, 48, 66, 68], with participation elasticities in the range of −0.2 to −0.3 [48, 71, 72]. However, many of these studies have limitations such as focusing on associations instead of causation or reliance on historical price and behavioral data collected when cannabis products were largely illegal [14, 17, 40, 71, 72]. As legal and illegal cannabis products, and as a result, taxed and untaxed products, increasingly co-exist in the US, how AYA consumers respond to changes in cannabis prices and taxes merits additional investigation.

In addition to knowing price or tax elasticity for AYAs’ cannabis demand, the real world policymaking involves more decisions than simply raising taxes. Tax structures, defined as the combination of bases and rates, are important determinants of tax policy effectiveness. Lessons from tobacco research suggest that uniform specific taxes based on pack sizes and indexed to inflation are superior to alternative tax structures in reducing cigarette consumption because such tax structures leave less room for tax avoidance [11, 53, 54]. However, this conclusion may only apply to manufactured cigarettes that are very uniform in sizes and harms [56]. In fact, alcohol research suggests that taxes based on harm-causing chemicals such as ethanol or tiered taxes based on ethanol have advantages of taxing more harmful products at a higher rate and addressing heterogeneity in product characteristics (e.g., beer vs. wine; concentrates vs. flowers) [39, 55, 57, 63]. In addition, since producing alcohol products with higher ethanol contents involves higher costs, price-based taxes are also often used to tax alcoholic beverages, likely imposing higher taxes on more harmful product types [55]. This consideration could also apply to cannabis, which costs more for higher THC products [61].

Currently, different states are adopting different tax bases and rates without knowing which combination is the most effective in reducing AYA cannabis use. While some states impose price-based or ad valorem taxes (e.g., a percentage of the retail price, similar to how some states tax e-cigarettes or wine and spirits), others levy specific taxes based on weight (e.g., per ounce of cannabis flower, similar to how states tax beer and cigarettes) or potency (e.g., per milligram of Tetrahydrocannabinol or THC) [45]. The potency or THC-based taxes can directly target the chemical that is responsible for the psychoactive effects of cannabis. A recent experimental cannabis marketplace study demonstrates that for U.S. adults, potency-based taxes may be more effective at reducing demand for high-potency products compared with price-based taxes [73]. However, this study does not consider weight-based taxes and focuses on the adult population. Tobacco and alcohol literature has also shown that excise tax bases not only have direct impacts on consumption but also interact with tax rates in driving use behaviors [52, 54, 57]. However, evidence on how cannabis tax bases and rates interact is lacking, especially on which tax base and rate combination is the most effective in curbing AYA cannabis use.

As the retail cannabis market thrives, the characteristics of cannabis products have undergone significant changes. The prices of legal products have decreased significantly over time, while the potency of cannabis products continues to increase [61]. The varieties and forms of cannabis have also expanded and currently encompass products such as flowers, edibles, vaping cartridges, and more [23, 61]. There is growing evidence showing that AYAs, compared to older adults, are more likely to vape cannabis [13, 24, 58]. These trends pose both challenges and opportunities for reforming or designing retail cannabis taxes to benefit public health. Specifically, with decreasing prices, how much tax rates must be raised to curb AYA use is important to both public health and revenue generation. Moreover, the economic relationship between various forms of cannabis (that is, whether they are complements, substitutes, or do not have a relationship) is unknown. Therefore, the current cannabis tax design has not taken into account these inter-form relationships that could have impacted tax effectiveness, especially in terms of preventing AYA use. This is a critical evidence gap as some states, such as Illinois, impose tiered taxes that tax cartridges and edibles at a higher rate than flowers. The effectiveness of such tiered structures may significantly vary depending on the inter-form relationships.

In summary, there are research gaps in the existing literature on the impact of retail cannabis taxes on AYA cannabis use, including the most effective tax structure design and own- and cross-price elasticities by forms. To address these gaps, this study conducted a Volumetric Choice Experiment (VCE) embedded in a nationally representative online survey of U.S. AYAs aged 15-20 who were using or susceptible to using cannabis in 2024 (N= 1,010). In the VCE, we randomized participants into three different tax bases (price, potency, weight), where pre-tax prices, tax rates, and potency levels additionally vary for four product forms: legal flowers, illegal flowers, edibles, and cartridges. This design allows us to capture not only the decision to purchase but also the purchase quantities and THC amount under different pricing and taxation scenarios [10]. It also expands the existing experimental methods by providing multiple forms that allow us to assess own price elasticities and cross-price elasticities between legal and illegal products and between different forms. [1, 1, 14, 19, 34, 62, 73, 74]. To our knowledge, this is the first VCE that evaluates how detailed tax policy design can be utilized to curb cannabis use among U.S. AYAs.

Materials and methods

Recruitment & sampling

The volumetric choice experiment (VCE) was embedded in a survey that we conducted during June-July 2024. We recruited a nationally representative sample of AYAs aged 15–20 years old who are currently using cannabis (i.e., using in the past 30 days) or are susceptible to cannabis use through a parent who is participating in the Ipsos Knowledge Panel. Susceptibility was determined using ever use and validated measures that assess the following five areas: willingness to use, intention to use, curiosity, anticipated pleasurable effects of use, and anticipated aversive effects of use. Susceptibility was identified if a participant does not strongly disagree with the use of cannabis or the pleasures associated with the use in either of the five areas [5]. Respondents who did not meet these criteria or who provided incomplete or inconsistent responses during the experimental tasks were excluded. These criteria ensured that all participants retained in the final sample are currently using cannabis or susceptible to cannabis use. In total, 1,010 participants were recruited, including 247 who reported past 30-day use (current users), 251 who reported ever use but no past 30-day use (former users), and 512 participants who never used cannabis but were susceptible (never users).

Volumetric choice experiment (VCE) design

Overview of VCE design

The study employed a split-sample VCE design to evaluate how AYAs’ cannabis consumption responds to different cannabis tax structures (bases and rates) and tax-related product features, which are pre-tax prices and THC levels [32, 49]. First, participants were randomly assigned to one of the following three tax base groups (i.e. split-sample): weightbased, price-based, or potency-based taxation. Next, they were presented with nine hypothetical choice sets where pre-tax prices, excise tax rates, and THC levels additionally vary. Therefore, this design utilizes between-subject variation to identify the impact of tax base on consumption and within-subject variation to identify the impact of other tax attributes (price, tax rates, THC levels) on consumption.

Compared to observational data, this VCE design has advantages in isolating the direct effects of each attribute (e.g., pretax prices vs. tax rates) and their interactions (e.g., tax rates × tax bases) on AYAs’ cannabis consumption. This is important given that in the real-world marketplace, both weight and THC are associated with cannabis prices, which makes isolating the effects of taxes based on these factors challenging [23, 60]. VCE also addresses several additional limitations in observational data. First, the effects of tax rates and bases on consumption cannot be estimated using observational data due to the lack of tax policy variation over time for causal inference methods (None of the taxing states have changed bases so far). Second, the survey data often ask questions about cannabis use rather than forms, which don’t allow for an estimation of inter-form price elasticities. Third, observational data are limited in their ability to identify optimal tax rate and base combinations, which could be beyond the existing choices in the real world. Compared to alternative experimental methods such as experimental marketplaces and hypothetical purchase tasks, VCEs have the advantages of varying three or more relevant attributes, further allowing for a comprehensive analysis of own- and cross- price or tax elasticities simultaneously.

Another common procedure of designing a choice experiment is to choose which products are explicitly listed as labeled products, along with the “none of the above” opt-out option. This is because it is not feasible to list all possible products in a marketplace, which is not only computationally cumbersome but also potentially confusing or driving participants’ fatigue. Therefore, in our design, four cannabis products are presented as the labeled choices: legal flower, illegal flower, edibles, and cartridges. These products are selected based on market share data and AYA use patterns. According to our cannabis youth survey, flower accounts for 30.83% of reported use, cartridges 27.45%, and edibles 17.84%, highlighting their prominence in the retail cannabis market. Legal flower and cartridges represent the most commonly purchased cannabis forms in licensed markets, and cartridges are the major method of cannabis vaping (Smart et al., 2018) [16, 31]. Among illegal products, illegal flower is included because it remains the most accessible and widely used illicit cannabis product among AYAs [69]. The rest of cannabis products in the marketplace, such as non-flower forms from illegal sources, while not explicitly listed in the choice set, are captured by the "none of the above" opt-out option. Our selection ensures that the experimental design reflects realistic and meaningful choice patterns within the cannabis marketplace, further allowing us to quantify the own- and cross-price elasticities among the selected products.

A detailed description of the VCE design, including screenshots of the purchase interface and example stimuli, is provided in Appendix B, and the VCE design examples are presented in Appendix C.

Attributes and levels

Table 1 summarizes the product attributes and levels used for each labeled product in the volumetric choice experiment, which are legal flower (1/8 ounce), illegal flower (1/8 ounce), edibles (one pack or unit), and cartridges (one unit). These sizes are chosen based on common units in the marketplace. The table also outlines the within-subject manipulations across three tax-related attributes: pre-tax price levels, tax levels, and THC levels, as well as the between-subject attribute - tax base.

Table 1.

Attributes and Levels, by Product

Attributes	Legal flower	Illegal flower	Edibles	Cartridge
	(Size=1/8	(Size=1/8	(Size=one)	(Size=one)
	ounce pack)	ounce pack)
Within-subject variations
Pre-tax price levels (4)	$ 10, $20, $30, $40	$10, $20, $30, $40	$10, $20, $30, $40	$20, $40, $60, $80
Tax levels	20%, 40%, 60%, 80% of of pre-tax prices or equivalent	0	20%, 40%, 60%, 80% of pre-tax prices or equivalent	20%, 40%, 60%, 80% of pre-tax prices or equivalent
THC levels (4) (2 for vaping)	10%, 15%, 25%, 30%	10%, 15%, 25%, 30%	20mg, 50mg, 100mg, 500mg	0.5g, 1g
Between-subject variations
Tax bases	Weight, price, THC mg	NA	Weight, price, THC mg	Weight, price, THC mg

Notes: Pre-tax prices represent 100%, 200%, and 300% increases from low prices

The levels of tax rates and bases were chosen based on existing and potential tax policies. Specifically, weight, price, and THC (potency) are tax bases that have been adopted by states in the real world. Tax levels are anchored at 20% of base prices or equivalent and raised significantly to 80%. These levels are chosen because of a variety of policy considerations [45]. First, due to the decreases in prices over time, there are discussions about significantly raising cannabis tax rates. Second, the tax shares (excise taxes + sales taxes) in retail prices could be benchmarked using tobacco taxes, which have been recommended by the World Health Organization to be at least 75%. Considering 8% to 10% sales tax rates, a 75% tax share in prices implies an approximately 80% price-based excise tax or equivalent for cannabis.

The levels for pre-tax prices and THC levels were chosen based on the marketplace data from Leafly.com and Weedmap.com at the time of survey design. We noticed that the THC levels are often shown in different units in the market (e.g., % for flowers, mgs for edibles, and grams for cartridges). Therefore, we presented the THC levels using common units assigned to each product type in the market but also provided conversion from % into mgs to assist comparisons. Finally, the tax levels with each different base were designed to represent similar total tax dollar amounts across bases. In total, there are five labeled options including opt-out, and each within-subject attribute contains two to four levels.

Choice set design

The number of attributes and their levels generate 16 (4²) profiles for illegal flowers, 32 (4² × 2) for cartridges, and 64(4³) for legal flowers and edibles. To enhance the experiment’s efficiency and reduce respondent fatigue, a fractional design is needed. We then used a D-efficiency design implemented using the Ngene software and chose 36 choice sets to balance the number of sets and the coverage of attribute and level combinations. [33]. These choice sets were further divided into four blocks, with each block containing 9 choice sets. The layout and presentation were previously tested in other volumetric choice studies. Participants were further randomized to one of the four blocks and answered 9 choice questions.

Other choice set features

Following the best practices, we incorporated cheap-talk scripts at the beginning of the experiment to encourage truthful responses, with the full text of these instructions provided in the Appendix [8, 67]. After reading the general instructions on what choice tasks involve, participants were randomized into seeing a tax base condition, where how taxes increase according to bases was presented (e.g., taxes are higher when prices are higher). In the choice sets, visual aids such as product images and descriptive labels were used to clearly differentiate between cannabis forms and attribute levels, thereby reducing cognitive burden and improving the authenticity of the decision-making process.

Before conducting the experiment, we asked users to report their weekly or monthly spending on cannabis. Participants currently using cannabis were instructed to make their purchasing decisions for the next month within the constraints of a personalized monthly cannabis budget, adding a layer of realism to the task. For participants who were susceptible but not using cannabis at the time of the survey, we asked them to make purchases with a hypothetical budget of $60, which was the average spending of users of similar ages. They were also instructed that they did not have to buy any of the cannabis products and instead opt out by clicking the “I would not buy any of these” button. When answering the choice sets, as participants filled in quantities, the total balance was displayed accordingly in front of them on the screen, along with their regular or recommended budget constraint to aid decision-making. However, these budget numbers served only as reminders since we did allow the participants to go over or below the usual or recommended budget to mimic real-world flexibility. An example of choice experiment questions is presented in the Appendix.

Measures

Outcomes

The outcome variables were the quantity of cannabis and associated THC consumption across product types, selfreported by participants in different purchasing scenarios. The detailed distribution of the cannabis unit consumption and THC consumption levels is shown in the appendix (Figures 1 & 2). Quantity consumption was measured based on popular or common units (e.g., 1/8 ounce for flowers and single servings for edibles and cartridges). THC consumption or intake was computed by multiplying the number of units chosen by the associated THC level, enabling an analysis of both quantity and potency dimensions of consumption.

Explanatory and control variables

The explanatory variables are manipulated conditions presented in Table 1, including tax bases, pre-tax prices, tax rates, and THC levels. When estimating price elasticities, the final retail prices that are inclusive of taxes and pre-tax base prices are used as explanatory variables. The details of these variables are described in the empirical strategy. The control variables include monthly budget constraint, alternative-specific constants for each product, and participants’ demographic and socioeconomic characteristics, such as age, gender, race/ethnicity, parents’ highest education, household income, and employment status. The detailed summary statistics for the demographic characteristics are shown in the appendix (Table 5).

Empirical Strategy

Analytical sample

Since we have 1,010 AYA participants who made 9 choice sets with 4 products, the total observations in the analytical sample were 36,360 (1,010 × 9 × 4). We further dropped 37 missing values when participants skipped the choice questions. The final analytical sample size was therefore 36,323.

Model to assess the impact of tax-related attributes on consumption

To evaluate how tax-related product attributes (pretax prices, tax rates, tax bases, and THC levels) and their levels influence cannabis consumption (units and THC) among AYA cannabis users and susceptible users, we employed a zero-inflated negative binomial (ZINB) regression that accommodates the excess zeros in the consumption measures. ZINB also jointly estimates the probability of nonpurchase and the number of units purchased and is widely used to estimate VCEs (Do et al. 2025). Key predictors are dummy-coded product attributes that we manipulated in the VCE, which are pretax price levels ($10 or $20 as omitted category, 100% increase, 200% increase, 300% increase), THC levels (low as omitted category, medium low, medium high, and high; and only low vs. high level for cartridges), tax rates (20% of price or equivalent, 40%, 60%, and 80%), and tax bases (weight based as omitted category, price based, THC based). This model is motivated by policymakers’ interest in choosing the combination of tax bases and rates, which are explicitly modeled in this specification (Tax Policy Center, 2024; bib6; Hoffer, 2023). The specification also allows us to test non-linear responses to changes in attributes, such as prices and taxes [2].

In the regressions, we also controlled for product-level alternative specific constants (ASC) for different cannabis product types (legal flower, illegal flower, edibles, and cartridges), which adjust for any systematic differences in product preference and features (e.g., differences in common units). Demographic variables such as age, gender, race/ethnicity, expenditures on cannabis, household income, and parents’ education were controlled for in the regressions. Our preferred model specification includes state fixed effects to account for the state-level differences such as policy environments. Nonetheless, we conducted sensitivity checks by excluding state fixed effects. We also conducted subgroup analyses by whether a participant has ever used cannabis given that ever users may have more knowledge about cannabis and more experience with purchasing. Standard errors were clustered at the individual level to adjust for correlations among the nine purchase decisions. Equation 1 shows the model specifications and the nesting natures of measurements (state-individual-choice sets-products) in more detail.

$$\begin{gathered} Consumptio{n}_{ist}={\alpha }_0+{\alpha }_{1}AS{C}_{ist} \\ +{\alpha }_{2}PriceLeve{l}_{ist}+{\alpha }_{3}THCLeve{l}_{ist}+{\alpha }_{4}TaxRate \\ Leve{l}_{ist}+{\alpha }_{5}TaxBas{e}_i+{\alpha }_6{X}_i+{\lambda }_m+{\epsilon }_{ist} \end{gathered}$$ (1)

where:

$Consumptio{n}_{ist}$ is the cannabis quantity or THC intake consumed by individual $i$ for product $s$ in choice set $t$;

$AS{C}_{ist}$ are dummy variables for each cannabis product type;

$PriceLeve{l}_{ist}$ are dummy variables for the pretax price levels $s$;

$THCLeve{l}_{ist}$ are dummy variables for the potency levels $s$;

$TaxRateLeve{l}_{ist}$ are dummy variables for the tax rate levels $s$;

$TaxBas{e}_i$ are dummy variables for tax bases;

$X_i$ is a vector of individual-level demographic covariates;

${\lambda }_m$ denotes state fixed effects;

$\epsilon$ is the error term.

In equation (1), $i$ indexes individuals, $s$ denotes the product alternative (e.g., legal flower, illegal flower, edible, and concentrate), and $t$ represents the choice set. PriceLevel, THCLevel, TaxRateLevel, and TaxBase are experimental conditions or manipulated attributes. ProductConstant comprises dummy variables for each cannabis product form, $X_i$ includes individual demographic characteristics, and ${\lambda }_m$ represents state fixed effects to control for unobserved state-level factors.

Model to assess own- and cross- price elasticities across forms or products

The labeled design of the VCE allows us to estimate how one product’s attributes impact the consumption of another product, such as how prices of legal flowers impact the consumption of illegal flowers and other forms (i.e., cross-price elasticities). In this specification, we assume that taxes are fully passed to prices and calculate the final retail prices that are inclusive of pretax prices and excise and sales taxes. Next, we use the ZINB model to estimate own- and cross-price elasticities across the following cannabis products or forms: legal flowers, illegal flowers, edibles, and cartridges. The results will provide price sensitivity by products, as well as illustrating whether and which products are substitutes or complements. We also included regressors such as ASCs, THC levels, tax bases, and sociodemographic variables similar to previous models. Motivated by the literature on illegal cannabis and high potency forms that are potentially more harmful than alternatives, this specification will inform how taxes or raising relative prices on one product (e.g., Illinois imposes higher taxes on edibles and cartridges than flowers) drives the consumption of another (i.e., inter-form or inter-product relationships). This specification can be expressed using the following equation 2 with both the product’s own price ($OwnPric{e}_{ist}$) and the prices of other alternatives ($OtherPric{e}_{i(j\ne s)t}$) in the same choice set entered into the equation:

$$\begin{gathered} Consumptio{n}_{ist}={\beta }_0+{\beta }_{1}AS{C}_{ist} \\ +{\beta }_{2}OwnPric{e}_{ist}+{\beta }_{3}OtherPric{e}_{i(j!=s)t} \\ +{\beta }_{4}THCLeve{l}_{ist}+{\beta }_{3}TaxBas{e}_i+{\beta }_6{X}_i+{\lambda }_m+{\epsilon }_{ist} \end{gathered}$$ (2)

where:

$OwnPric{e}_{ist}$ is the tax-inclusive final price of the selected product alternative;

$OtherPric{e}_{i(j!=s)t}$ represents the tax-inclusive final prices of all other available product alternatives in the same choice set;

All other variables are as previously defined.

Model to adjust for individual fixed effects and assess the interaction between tax bases and other tax attributes

Another advantage of the VCE design is the ability to control for individual fixed effects and fully rely on within-subject variation in attributes to identify the causal impacts of attributes on consumption. Although the tax base as a between-subject experimental condition is not estimable in an individual-fixed effects model, their interactions with those within-subject attributes - pretax prices, tax rates, and THC levels can be estimated. These interaction terms can also allow us to evaluate how tax bases are designed to impact consumption through unit, price, or THCs. Therefore, in the third specification (equation 3), we use a linear fixed effect model and regress consumption measures on within-subject attributes(pretax prices, prices, taxes, and THCs) and their interaction terms with tax bases (weight as the omitted category, price, and potency bases) to examine how tax bases moderate the impact of other tax-related attributes.

$$\begin{gathered} Consumptio{n}_{ist}={\alpha }_0+{\alpha }_{1}AS{C}_{ist} \\ +{\alpha }_{2}PriceLeve{l}_{ist}+{\alpha }_{3}THCLeve{l}_{ist} \\ +{\alpha }_{4}TaxRateLeve{l}_{ist}+{\alpha }_{5}TaxBas{e}_i \\ +{\alpha }_6{X}_i+{\lambda }_m+{\epsilon }_{ist} \end{gathered}$$ (3)

where:

$Pric{e}_{ist}$, $Ta{x}_{ist}$, and $TH{C}_{ist}$ are ordinal variables for attribute levels, respectively;

$I_i$ denotes individual fixed effect;

All other variables are as previously defined.

Results

Table 2 presents the semi-elasticities (percent change in consumption in response to a change from baseline) from the ZINB regressions that examine how product attributes and their levels impact cannabis consumption, measured both in units and THC. Our preferred results or models are in columns (2) and (4) when state fixed effects are controlled for. Compared with a base or pretax low price ($10 for flowers and edibles and $20 for cartridges), a 100% increase, a 200% increase, and a 300% increase lead to a 62%, 90%, and 132% reduction in unit consumption, and a 65%, 86%, and 133% reduction in THC consumption, respectively. These estimates also imply a price elasticity of demand ranging from −0.4 to −0.6. Post-estimation tests confirm that the coefficients for the different pre-tax price levels are statistically different from one another (p < 0.01) in both column (2) and (4). For example, moving from a 200% to a 300% increase in pre-tax price levels yields a significantly larger reduction in cannabis/THC consumption.

Table 2.

The Impact of Attributes on Cannabis Consumption, Estimated using Zero-Inflated Negative Binomial Regressions, N=35,323

	(1)	(2)	(3)	(4)
VARIABLES	Outcome=Unit		Outcome=THC
Within-Subject Effects:
Pre-Tax Price Levels (Base: $10 for flowers and edibles; $20 for cartridges)
100% increase	−0.63*** (0.04)	−0.63*** (0.04)	−0.66*** (0.03)	−0.65*** (0.03)
200% increase	−0.89*** (0.06)	−0.92*** (0.05)	−0.79*** (0.07)	−0.86*** (0.06)
300% increase	−1.34*** (0.07)	−1.36*** (0.07)	−1.30*** (0.07)	−1.34*** (0.07)
THC Levels (Base: Low)
Medium low	0.09 (0.07)	0.10 (0.06)	0.69*** (0.06)	0.71*** (0.06)
Medium high	0.07 (0.07)	0.06 (0.07)	1.22*** (0.07)	1.22*** (0.07)
High	−0.03 (0.07)	−0.04 (0.07)	1.98*** (0.07)	1.97*** (0.07)
Tax Rate Levels (Base: 20% or equivalent)
40% or equivalent	−0.29*** (0.09)	−0.24*** (0.07)	−0.42*** (0.09)	−0.34*** (0.07)
60% or equivalent	−0.37*** (0.08)	−0.35*** (0.06)	−0.46*** (0.08)	−0.43*** (0.06)
80% or equivalent	−0.40*** (0.07)	−0.41*** (0.07)	−0.41*** (0.07)	−0.43*** (0.07)
Between-Subject Effects:
Bases of Tax Imposed (Base: Weight)
Price	−0.05 (0.10)	−0.00 (0.11)	−0.10 (0.10)	−0.10 (0.10)
THC	0.31** (0.12)	0.29** (0.11)	0.22* (0.11)	0.16 (0.11)
State Fixed Effects:	No	Yes	No	Yes

Notes: Robust standard errors are reported in parentheses. In column (2), tax rate categories do not differ significantly across levels (p > 0.1), but the THC levels differ significantly (15% vs. 30%, 25% vs. 30%). For price levels, all comparisons ($20 vs. $30; $20 vs. $40; $30 vs. $40) show significant differences. In column (4), THC and price levels differ significantly, while tax rate categories do not differ significantly across levels (p > 0.1). *** p<0.01, ** p<0.05, * p<0.1

While THC levels are not statistically associated with unit consumption, they are positively associated with THC consumption. Compared with a 10% THC level, THC levels at 15%, 25%, and 30% increase THC consumption by 71%, 122%, and 197%. Post-estimation tests show that THC levels are statistically different (p < 0.01) for both column (2) and (4). Results for tax rates and bases also show significant impacts on consumption. Compared to a tax at 20% of pretax prices or equivalent, higher tax levels at 40% (100% increase), 60% (200% increase), and 80% (300% increase) lead to 25%, 33%, and 40% reductions in units, and 34%, 43%, and 43% reductions in THC consumption, respectively. These estimates imply a tax elasticity of demand ranging from −0.1 to −0.3. Post-estimation tests indicate that the coefficients for different tax rate categories are not statistically distinguishable from each other in either column (2) or column (4) (p >0.1). For example, The estimated reduction in consumption at a 40% tax rate is statistically distinguishable from the reduction observed at a 60% tax rate. Comparing to weight-based and price-based taxes, THC-based taxes lead to 30% more unit consumption but have no significant impacts on THC consumption when state fixed effects are controlled for.

Table 3 presents the estimated own-and cross-price elasticities for different cannabis forms. Our preferred model is Column 4 with the outcome being THC consumption and state fixed effects controlled for. The own-price elasticities for legal flower, illegal flower, edibles, and cartridges were all about −0.3 (p<0.01). Illegal and legal flowers are substitutes for each other; a 10% higher price of illegal flowers increases legal flower THC consumption by 1.3%, and a 10% higher price of legal flowers increases illegal flower THC consumption by 2%. Illegal flowers are also substitutes for cartridges. When cartridge prices increase by 10%, THC consumption from illegal flowers increases by 1%. Edibles are complements to other cannabis forms or products, including legal and illegal flowers and cartridges. If prices of other forms or products increase by 10%, edible THC consumption reduces by 0.3% to 0.8%. Cartridges are substitutes for legal flowers. When legal flower prices increase by 10%, THC consumption from cartridges increases by 0.7%. A summary of between form relationships is in Appendix Table 10. Finally, the average THC consumption by forms was 244.7mg (25.6%) for legal flowers, 469.3mg (49%) for illegal flowers, 78.4 mg (8.2%) for edibles, and 164.7mg (17.2%) for cartridges. A 10% increase in legal product prices will reduce overall THC consumption from legal products by 152 mg but increase THC consumption from illegal products by exactly the same amounts.

Table 3.

Own- and Cross- Price Elasticities of Demand Among Different Cannabis Forms, Estimated using Zero-Inflated Negative Binomial Regressions, N=35,323

	(1)	(2)	(3)	(4)
Variables	Outcome=Unit		Outcome=THC
Within-Subject Effects:
Own-Price Elasticity
Legal Flower	−0.26*** (0.05)	−0.23*** (0.05)	−0.24*** (0.05)	−0.27*** (0.05)
Illegal Flower	−0.30*** (0.02)	−0.28*** (0.03)	−0.29*** (0.02)	−0.29*** (0.02)
Edible	−0.40*** (0.04)	−0.29*** (0.03)	−0.30*** (0.02)	−0.30*** (0.02)
Cartridge	−0.30* (0.10)	−0.24* (0.11)	−0.26** (0.08)	−0.33*** (0.06)
Cross-Price Elasticity
Legal Flower
Illegal Flower Price	0.11*** (0.03)	0.08** (0.03)	0.14*** (0.03)	0.13*** (0.02)
Edible Price	0.01 (0.03)	−0.01 (0.04)	−0.00 (0.03)	0.00 (0.02)
Cartridge Price	−0.05 (0.04)	−0.06* (0.03)	−0.05 (0.04)	−0.05 (0.04)
Illegal Flower
Legal Flower Price	0.12** (0.04)	0.08 (0.04)	0.21*** (0.04)	0.21*** (0.03)
Edible Price	0.02 (0.02)	−0.00 (0.02)	0.00 (0.02)	0.00 (0.02)
Cartridge Price	0.03 (0.04)	0.02 (0.04)	0.09** (0.03)	0.10*** (0.03)
Edible
Legal Flower Price	0.00 (0.02)	0.01 (0.02)	−0.05** (0.02)	−0.005** (0.02)
Illegal flower Price	0.03*** (0.01)	0.03* (0.01)	−0.03*** (0.01)	−0.03*** (0.01)
Cartridge Price	0.01 (0.02)	0.01 (0.03)	−0.09*** (0.02)	−0.07*** (0.02)
Cartridge
Legal Flower Price	0.08*** (0.03)	0.06 (0.03)	0.10*** (0.04)	0.07** (0.04)
Illegal Flower Price	0.03 (0.02)	0.02 (0.05)	0.04 (0.03)	0.04 (0.03)
Edible Price	0.04 (0.04)	0.04 (0.07)	−0.03 (0.05)	−0.03 (0.04)
State Fixed Effects:	No	Yes	No	Yes

Robust standard errors clustered at the individual level in parentheses

^***p<0.01

^**p<0.05

^*p<0.1

Table 4 further explores how tax bases impact consumption through other tax-related factors. Columns 1 and 3 show results from estimating ordinal attribute levels, which are consistent with results in Table 2 where nonlinear effects of attribute levels are modeled. Columns 2 and 4 show how tax bases moderate the impacts of THC levels, prices, and taxes on consumption. The results suggest that under a weight tax base, higher THC levels lead to higher unit and THC consumption. This impact is significantly reduced in a potency or THC-based tax base. In addition, compared to a weight tax base, a price tax base also reduces the impacts of higher THC levels on increasing THC consumption. There are no significant differences in terms of tax and price impacts on consumption by tax bases. Additional results from sensitivity analyses and subgroup models are presented in Appendix Table 3, 4, and 8.

Table 4.

Effects of Product Attributes, Pricing, and Taxation on Cannabis Consumption N=35,399

	(1)	(2)	(3)	(4)
Variables	Unit	Unit	THC	THC
THC	0.00 (0.00)	0.01** (0.00)	0.12*** (0.01)	0.17*** (0.02)
THC × Price-based Tax		−0.01* (0.01)		−0.08*** (0.03)
THC × Potency-based Tax		−0.02*** (0.01)		−0.07** (0.03)
Price	−0.05*** (0.00)	−0.05*** (0.00)	−0.25*** (0.01)	−0.25*** (0.02)
Price × Price-based Tax		−0.00 (0.01)		−0.02 (0.03)
Price × Potency-based Tax		0.00 (0.01)		0.02 (0.03)
Tax	−0.01*** (0.00)	−0.02*** (0.00)	−0.08*** (0.01)	−0.10*** (0.02)
Tax × Price-based Tax		0.01 (0.01)		0.02 (0.04)
Tax × Potency-based Tax		0.01* (0.01)		0.06 (0.04)
Constant	0.27*** (0.01)	0.27*** (0.01)	1.19*** (0.06)	1.19*** (0.06)

Notes: Robust standard errors clustered at the individual level in parentheses. Price-based tax and Potency-based tax are dummy variables indicating tax bases. Regressions also controlled for product constants. *** p<0.01, ** p<0.05, * p<0.1

Discussions

Excise taxes are among the most effective policies to regulate substances, such as tobacco, alcohol, and retail cannabis. While economic theories primarily use negative externalities (e.g., secondhand smoke) to justify excise taxes and other regulations on substances, public health and real-world policymaking in the United States consider a broader set of factors, including substance use prevalence and addiction among young people. Therefore, evaluating how excise taxes on retail cannabis prevent or reduce cannabis use among AYAs is of particular importance to policy justification. Both the literature and real-world policymaking further suggest that tax designs are more complex than simply raising taxes. That is, policymakers can manipulate tax structures (i.e., rates and bases) to maximize the effectiveness of tax policies in affecting behaviors, or to target a subgroup of products that are more harmful than others.

In this study, we estimated the own-price elasticities of cannabis demand by AYA current and susceptible users in the U.S., with consumption measured in both quantities and THC mgs. The price elasticities were found to be around −0.3 regardless of forms, which can be further decomposed to a price elasticity ranging from −0.4 to −0.6 and a tax elasticity ranging from −0.1 to −0.3. These results align with previous literature showing that cannabis demand by U.S. AYAs is inelastic, with price elasticities mostly centered around −0.3 [17, 23, 43, 48, 71, 72]. Our estimates are also nearly identical to a recent study that analyzed Monitoring the Future Surveys conducted between 2015 and 2022 using an instrumental variable approach and found the price elasticity to be from −0.2 to −0.3. The alliance of price elasticities estimated using VCE with observational data further demonstrates the external validity of choice experiments and the robustness of price elasticity estimates. The conclusions are that raising cannabis taxes and prices not only reduces cannabis consumption among AYA users and potential initiations among AYA susceptible users but also leads to higher tax revenues.

In addition to elasticities, our findings elucidate the reduced salience in consumption responsiveness as cannabis prices increase. That is, while a 100% increase from $10 leads to a 65% reduction in consumption, a 300% increase only leads to a 133% reduction, suggesting diminishing responsiveness. Existing literature in tobacco research also shows that the price elasticity of cigarette demand is heterogeneous across the price distribution [64]. However, opposite to our findings, price responsiveness in cigarette consumption increases monotonically instead of decreasing [64]. Considering that retail cannabis prices have been decreasing since legalization, the effects of raising cannabis taxes and prices on consumption are likely to be salient at this stage [45, 61]. However, if cannabis prices start to increase instead of decrease over time, policymakers may need to consider means to improve the effectiveness or salience of cannabis taxes and prices in reducing cannabis consumption among AYAs.

The decreasing cannabis price trends also raise concerns about whether the current excise tax rates for cannabis, which approximately range between 10% to 40% of prices or equivalent, are sufficient in reducing or preventing AYAs’ cannabis use [45]. Our findings suggest that an increase in excise taxes from 20% of prices or equivalent to 40% significantly reduces consumption by current and susceptible AYA users. However, further increases to 60% or 80% of prices or equivalent do not lead to any additional reductions in consumption. When limiting the analysis to susceptible never users only, the results show that an increase in taxes to 60% of prices or equivalent significantly reduces purchasing among potential users, but the gain in reduction is marginal when moving from 60% to 80% of prices. These results can be directly used by policymakers to decide on excise tax rates for retail cannabis.

Our results demonstrate which tax bases are more effective in reducing AYA cannabis consumption or initiation. Compared to weight-based taxes, potency-based taxes are associated with higher unit cannabis consumption among AYAs but not higher THC consumption. This finding suggests that under a potency tax base where higher THC products cost more, AYAs shift toward consuming greater quantities of lower-THC cannabis. This observation is consistent with the existing literature in the following two areas: 1) a potency or THC tax base leads to less consumption of high-THC products compared to other bases [22] and 2) a potency tax base is not as effective as weight- or price-based taxes in reducing unit consumption [57]. If the goal of cannabis taxes is to reduce THC consumption and their psychoactive impacts among AYAs, THC and price tax bases are more effective than a weight tax base. However, if the goal of cannabis taxes is to reduce a broader set of health consequences, which tax base is better at curbing AYA use depends on the trade-off between using fewer units with high potency levels and using more units with low potency levels [42]. This trade-off could be further complicated by available product types and administration methods in the current marketplace, where high THC products tend to be concentrates or edibles and low THC products tend to be flowers. If low THC products are associated with a higher likelihood of combustible cannabis smoking, the overall risks of consuming low THC products may not be lower than consuming high THC products via vaping or ingestion [7]. Therefore, a cost-benefit analysis is needed to understand the consequences of potential shifts in use patterns and product choices due to tax bases [21].

Consistent with prior research [2, 44, 62], we find that higher THC levels are associated with greater THC consumption. When considering the interactions between tax bases and other tax-related attributes, the results further illustrate that the positive association between THC levels and consumption is stronger under a weight tax base. The potency tax base significantly reduces this positive association, showing protective effects. The price-based taxes, compared to weight-based taxes, also reduce the positive associations between THC levels and consumption. We also found that price-based taxes did not moderate the impact of prices on consumption. The combined findings suggest that price-based taxes shift consumption preference through choices of THC levels instead of price levels, which are consistent with observations that higher THC levels are associated with higher-priced products such as cartridges [59, 61].

Regardless of tax bases, cannabis excise taxes usually are not explicitly presented to consumers at the point of sale in the U.S.. All three tax bases: weight-, THC-, and price-based taxes are most often applied at some stage in the supply chain prior to retail, such as cultivation, distribution, or wholesale levels [3]. Even when excise taxes are based on retail prices and collected at the business-to-consumer transactions, they are often compounded with sales taxes at the point of sale and not necessarily distinguishable to consumers [25, 45]. However, this does not imply that consumers are unaware of tax bases, especially when most transactions at dispensaries are facilitated or assisted by retail staff such as budtenders who have the responsibility to explain taxes to price-sensitive consumers [9, 42]. Moreover, unlike Canada where cannabis is federally legal, cannabis taxes in the U.S. are set at state and local levels, with the detailed tax rates and bases going through the ballot initiatives and legislative actions. With the attention given to retail cannabis legalization and policy-setting at the local level, it is likely that the awareness of cannabis taxation is high in the U.S., which could impact how AYAs perceive the affordability of cannabis [47].

The results also reveal significant substitution and complementary effects. Higher prices on legal flowers and cartridges will drive AYA users to illegal products, which could undermine tax revenues and the effectiveness of taxes in reducing AYA cannabis use. Over-taxation of legal products (e.g., tax rates beyond 60%) may lead to marginal gains in use reduction and unintentionally drive consumers to the illegal market, which poses greater health risks due to potential contaminants and inconsistent THC levels [12]. Interestingly, edibles are complements to other products. When the prices of flowers (both legal and illegal ones) and cartridges increase, edible consumption will decrease. This finding suggests that policies increasing the price of flower or cartridges may also indirectly reduce edible consumption, amplifying the overall effect on total cannabis consumption. In contrast, if prices for other products fall, edible use may rise alongside them, highlighting the need for comprehensive pricing strategies across product types. The overall THC consumption impacts based on these inter-relationships suggest that the reduction from legal products may be completely offset by shifting THC consumption to illegal products, suggesting the importance of restricting the illegal marketplace to harvest the benefits of taxing cannabis. However, since we assume that there are no differences in cost or quality by sources and no penalties associated with obtaining illegal flowers and that illegal flowers are always cheaper than legal counterparts, the conclusion needs to be interpreted as the upper bound estimate for the tax evasion.

We also observe that higher flower prices may drive AYAs to cartridges, which tend to contain higher THC levels than flowers and are consumed via vaping. In contrast, higher taxes or prices on cartridges or edibles do not have significant impacts on flower consumption. The combined evidence suggests that tiered tax rates, such as the one in Illinois that taxes cannabis by forms and imposes higher rates on cartridges and edibles than on flowers, are unlikely to be undermined by substitutions between forms. Policymakers could consider applying such tiered tax structures, which have advantages over weight taxes by directly targeting THC levels, and over THC- or potency-based taxes by lowering the costs or efforts needed to calculate and administer a tax rate for each different level of THCs. Similar tiered tax designs have been successfully implemented to regulate alcoholic beverages, which are also dominated by three major product categories that contain differential levels of harm-causing chemicals.

As an experiment, our approach may have some limitations. First, our design assumes taxes to be fully passed to prices, whereas in the real world, taxes may be under- or over-shifted to prices. The policy impact prediction using our results therefore needs to adjust for real-world tax pass-through rates to prices. Another potential limitation is the reliance on stated preference or experimental data, which may contain hypothetical biases. However, given how close our price elasticity estimates are aligned with those estimated using survey data, the likelihood of our data containing severe hypothetical biases is low. We also adopted best-practice recommendations in the choice experiment literature by recruiting only participants who are at least susceptible to using cannabis and use cheap-talk and detailed instructions to encourage participants to answer questions truthfully. Finally, the implementation and compliance of tax policies also determine the effectiveness, which are beyond the scope of this study. Despite these limitations, our data provide policymakers with valuable insights for selecting appropriate tax bases and rates, as well as estimating the potential effects of novel taxation strategies on cannabis use.

Supplementary Material

Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/s10198-025-01875-3.

Appendix A

Fig. A.1.

The THC Unit by Product Category ¹

¹ In Figure A.1. For all categories, the average THC unit is 668, with standard deviation at 3,935.65; For legal consumption, the average THC unit is 740 (SD: 4,229.39); For illegal consumption, the avergae THC unit is 1649.20 (SD: 6,306.49); For Edible, the average THC unit is 104.97 For Concentrate, the average THC unit is 177.37 (SD: 1,585.74).

Fig A.2.

The Unit by Product Category ¹

¹ In Figure A.2, for all categories, the average unit is 1.02, with standard deviation at 5.30; For legal consumption, the average THC unit is 0.96 (SD: 4.96); For illegal consumption, the average THC unit is 2.22 (SD: 8.67); For edible, the average THC unit is 0.64 (SD: 2.24); For Concentrate, the average THC unit is 0.24 (SD: 2.29).

Table A.1.

Summary Statistics (Observation = 1010)

Variables	Percent/Mean (SD)
Sociodemographic Factors:
Age	17.68 (1.69)
Gender
Female	54.46%
Male	45.25%
Do not Know	0.30%
Race / Ethnicity
White, Non-Hispanic	59.21%
Black, Non-Hispanic	11.39%
Other, Non-Hispanic	3.07%
Hispanic	18.61%
2+ Races, Non-Hispanic	7.72%
Education
No High School or GED	43.27%
High School Graduate	28.22%
Some College	25.45%
Bachelor’s Degree	1.78%
Master’s Degree or Higher	1.29%
Household Income
Less than $10,000	5.35%
$10,000 to $24,999	6.34%
$25,000 to $49,999	12.57%
$50,000 to $74,999	13.56%
$75,000 to $99,999	15.54%
$100,000 to $149,999	21.39%
$150,000 or more	25.25%
Cannabis Use:
Current Users	24.46%
Former Users	24.85%
Susceptible Never Users	50.69%
Monthly Expenditure	119.34 (86.76)

Table A.2.

Other factors contributing to Cannabis Consumption, Estimated using ZINB, N=35,323

	(1)	(2)	(3)	(4)
VARIABLES	Outcome=Unit		Outcome=THC
Within-Subject Effects:
Product Constant (Base: Legal Flower)
Illegal Flower	0.50*** (0.16)	0.60*** (0.13)	0.45*** (0.15)	0.52*** (0.13)
Edible	0.82*** (0.13)	0.90*** (0.11)	−1.14*** (0.11)	−1.08*** (0.10)
Cartridge	−0.43** (0.21)	−0.49*** (0.14)	−0.55*** (0.19)	−0.63*** (0.13)
Between-Subject Effects:
Expenditures	0.63*** (0.07)	0.63*** (0.07)	0.30*** (0.07)	*0.31*** (0.06)
Age	2.15*** (0.83)	1.76** (0.80)	2.06*** (0.66)	1.75*** (0.63)
Gender (Base: Male)
Female	−0.04 (0.09)	−0.05 (0.08)	−0.09 (0.07)	−0.10* (0.06)
Race and Ethnicity (Base: White, Non-Hispanic)
Black, Non Hispanic	0.04 (0.16)	−0.07 (0.16)	0.20 (0.16)	0.16 (0.16)
Other, Non-Hispanic	−0.21 (0.22)	−0.28 (0.26)	−0.05 (0.11)	−0.11 (0.10)
Hispanic	0.22 (0.17)	0.06 (0.15)	0.17 (0.13)	0.11 (0.11)
2+ Races, Non-Hispanic	0.03 (0.15)	−0.02 (0.15)	0.15 (0.11)	0.9 (0.10)
Household Income (Base: Less than $10000)
$10000 to $24999	−0.56 (0.45)	−0.43 (0.40)	−0.48 (0.47)	−0.42 (0.40)
$25000 to $49999	−0.67 (0.38)	−0.65* (0.32)	−0.69* (0.41)	−0.70** (0.33)
$50000 to $74999	−0.43 (0.42)	−0.42 (0.33)	−0.57 (0.44)	−0.55 (0.34)
$75000 to $99999	−0.94** (0.38)	−0.86*** (0.32)	−0.81** (0.41)	−0.75** (0.33)
$100000 to $149999	−0.89** (0.37)	−0.79** (0.31)	−0.98** (0.40)	−0.87*** (0.32)
$150000 or more	−0.73* (0.38)	−0.65** (0.31)	−0.86** (0.41)	−0.75** (0.32)
Parents’ Education (Base: No High School diploma or GED)
High School Graduate	−0.47*** (0.17)	−0.36** (0.15)	−0.43*** (0.14)	−0.36*** (0.12)
Some College	−0.46** (0.18)	−0.35** (0.16)	−0.45*** (0.15)	−0.35*** (0.13)
Bachelor	−0.54** (0.24)	−0.49** (0.22)	−0.51*** (0.15)	−0.42*** (0.13)
Master or Higher	−1.11** (0.54)	−1.10* (0.65)	−0.22** (0.10)	−0.10 (0.13)
State Fixed Effects:	No	Yes	No	Yes

Note: Robust standard errors are reported in parentheses

^***p<0.01

^**p<0.05

^*p<0.1

Table A.3.

Own- and Cross- Price Elasticities of Demand Among Different Cannabis Forms, Estimated using Zero-Inflated Negative Binomial Regressions, N=35,323

	(1)	(2)	(3)	(4)
Variables	Outcome=Unit		Outcome=THC
Within-Subject Effects:
Product Constant (Base: Legal Flower)
Illegal Flower	0.43 (0.47)	0.40 (0.56)	0.08 (0.46)	0.05 (0.43)
Edible	5.18*** (1.14)	0.72 (0.44)	0.19 (0.38)	0.16 (0.36)
Cartridge	−0.81 (0.50)	−1.10* (0.66)	−0.59 (0.48)	−0.43 (0.40)
THC Levels (Base: 10%)
15%	0.08 (0.07)	−0.10 (0.10)	0.64*** (0.06)	0.66*** (0.06)
25%	0.10 (0.06)	0.01 (0.07)	1.19*** (0.06)	1.20*** (0.06)
30%	0.12* (0.07)	−0.04 (0.10)	2.10*** (0.07)	2.10*** (0.07)
Between-Subject Effects:
Bases of Tax Imposed (Base: Weight)
Price	−0.06 (0.11)	−0.18 (0.15)	−0.12 (0.10)	−0.14 (0.10)
THC	0.21* (0.11)	0.07 (0.17)	0.14 (0.11)	0.08 (0.11)
Expenditure	0.62*** (0.08)	0.26*** (0.07)	0.31*** (0.07)	0.32*** (0.07)
Age	2.07** (0.81)	2.98** (1.20)	2.15*** (0.64)	1.83*** (0.61)
Gender (Base: Male)
Female	−0.02 (0.09)	0.04 (0.19)	−0.09 (0.07)	0.03 (0.12)
Race and Ethnicity (Base: White, Non-Hispanic)
Black, Non-Hispanic	0.09 (0.15)	0.44 (0.37)	0.22 (0.16)	0.18 (0.15)
Other, Non-Hispanic	−0.23 (0.22)	−0.05 (0.25)	−0.07 (0.10)	−0.11 (0.09)
Hispanic	0.23 (0.15)	−0.06 (0.21)	0.16 (0.12)	0.10 (0.11)
2+ Races, Non-Hispanic	0.06 (0.14)	0.25 (0.19)	0.18* (0.11)	0.11 (0.10)
Household Income (Base: Less than $10000)
$10000 to $24999	−0.61 (0.42)	−0.44 (0.47)	−0.52 (0.46)	−0.44 (0.39)
$25000 to $49999	−0.66* (0.37)	−0.73* (0.42)	−0.65 (0.41)	−0.66** (0.33)
$50000 to $74999	−0.49 (0.40)	−0.48 (0.45)	−0.57 (0.44)	−0.53 (0.34)
$75000 to $99999	−0.94** (0.37)	−0.89** (0.42)	−0.81* (0.41)	−0.72** (0.33)
$100000 to $149999	−0.87** (0.36)	−0.95*** (0.35)	−0.97** (0.40)	−0.85*** (0.32)
$150000 or more	−0.73** (0.37)	−0.91** (0.41)	−0.86** (0.41)	−0.73** (0.32)
Parents’ Education (Base: No High School diploma or GED)
High School Graduate	−0.42*** (0.16)	−0.73** (0.29)	−0.45*** (0.13)	−0.37*** (0.12)
Some College	−0.44** (0.17)	−0.74*** (0.24)	−0.48*** (0.14)	−0.40*** (0.12)
Bachelor	−0.46** (0.23)	−0.80** (0.31)	−0.47*** (0.15)	−0.39*** (0.13)
Master or Higher	−1.17** (0.54)	−1.52* (0.92)	−0.25** (0.11)	−0.13 (0.14)
State Fixed Effects:	No	Yes	No	Yes

Note: Robust standard errors are reported in parentheses

^***p<0.01

^**p<0.05

^*p<0.1

Table A.4.

The Impact of Attributes on Cannabis Consumption, Estimated using Zero-Inflated Negative Binomial Regressions (current users), N=8,675

	(1)	(2)	(3)	(4)
VARIABLES	Outcome=Unit		Outcome=THC
Within-Subject Effects:
Pre-Tax Price Levels (Base: $10)
$20	−0.70*** (0.06)	−0.74*** (0.06)	−0.68*** (0.06)	−0.69*** (0.06)
$30	−0.87*** (0.11)	−0.95*** (0.10)	−0.69*** (0.10)	−0.83*** (0.08)
$40	−1.26*** (0.13)	−1.32*** (0.12)	−1.12*** (0.11)	−1.19*** (0.10)
THC Levels (Base: 10%)
15%	0.05 (0.10)	0.07 (0.10)	0.66*** (0.10)	0.73*** (0.09)
25%	0.01 (0.13)	0.02 (0.12)	1.16*** (0.10)	1.22*** (0.09)
30%	−0.08 (0.13)	−0.09 (0.12)	1.66*** (0.10)	1.66*** (0.09)
Tax Rate Levels (Base: 20%)
40%	−0.52*** (0.15)	−0.38*** (0.13)	−0.63*** (0.14)	−0.46*** (0.11)
60%	−0.44*** (0.14)	−0.44*** (0.12)	−0.38** (0.14)	−0.30*** (0.10)
80%	−0.30*** (0.12)	−0.33*** (0.11)	−0.20* (0.11)	−0.26** (0.10)
Between-Subject Effects:
Bases of Tax Imposed (Base: Weight)
Price	0.13 (0.17)	0.27 (0.19)	0.03 (0.18)	−0.01 (0.17)
THC	0.73*** (0.18)	0.64*** (0.22)	0.56*** (0.18)	0.35* (0.19)
	17816	17816	17816	17816
State Fixed Effects:	No	Yes	No	Yes

Note: Robust standard errors are reported in parentheses

^***p<0.01

^**p<0.05

^*p<0.1

Table A.5.

The Impact of Attributes on Cannabis Consumption, Estimated using Zero-Inflated Negative Binomial Regressions (potential users), N=17,740

	(1)	(2)	(3)	(4)
VARIABLES	Outcome=Unit		Outcome=THC
Within-Subject Effects:
Pre-Tax Price Levels (Base: $10)
$20	−0.47*** (0.06)	−0.48*** (0.05)	−0.58*** (0.05)	−0.58*** (0.05)
$30	−0.79*** (0.09)	−0.79*** (0.08)	−0.87*** (0.10)	−0.89**** (0.09)
$40	−1.30*** (0.10)	−1.28*** (0.10)	−1.43*** (0.10)	−1.45*** (0.10)
THC Levels (Base: 10%)
15%	0.00 (0.08)	0.02 (0.08)	0.77*** (0.09)	−0.79*** (0.08)
25%	−0.02 (0.08)	−0.005 (0.07)	1.28*** (0.08)	1.29*** (0.08)
30%	−0.15 (0.09)	−0.14* (0.08)	2.21*** (0.09)	2.21*** (0.10)
Tax Rate Levels (Base: 20%)
40%	−0.09 (0.11)	−0.05 (0.10)	−0.15 (0.10)	−0.14 (0.10)
60%	−0.34*** (0.08)	−0.35*** (0.07)	−0.38*** (0.07)	−0.39*** (0.07)
80%	−0.30*** (0.08)	−0.27*** (0.08)	−0.46*** (0.08)	−0.46*** (0.08)
Between-Subject Effects:
Bases of Tax Imposed (Base: Weight)
Price	−0.18 (0.17)	−0.08 (0.16)	−0.18 (0.15)	−0.17 (0.15)
THC	0.10 (0.17)	0.15 (0.16)	0.03 (0.15)	−0.01 (0.15)
	17816	17816	17816	17816
State Fixed Effects:	No	Yes	No	Yes

Note: Robust standard errors are reported in parentheses

^***p<0.01

^**p<0.05

^*p<0.1

Table A.6.

Economic Relationship based on THC Consumption between Different Forms in The Matrix

Consumption ╲ Price	Legal flowers	Illegal flowers	Edibles	Cartridges
Legal flowers	Reduction	Substitutes
Illegal flowers	Substitutes	Reduction		Substitutes
Edibles	Complements	Complements	Reduction	Complements
Cartridges	Substitutes			Reduction

Notes: Matrix of relationships based on Table 4, Column 4

Table A.2 shows product preferences in consumption and the associations between controls and cannabis consumption for AYA cannabis users and susceptible users. Compared with legal flowers, illegal flowers and edibles are associated with higher unit and THC consumption, whereas cartridges are associated with less unit and THC consumption. A 10% higher cannabis expenditure is associated with a 4% to 5% higher consumption. Age is positively associated with consumption. Gender was not significantly associated with consumption. Race/ethnicity is not significantly associated with cannabis consumption. AYAs with higher household incomes and parents’ educational attainment, particularly those whose parent(s) has some college or a bachelor’s degree, exhibit lower consumption compared to the reference group (household income less than $10,000 and no high school diploma or GED).

Data Availability

The datasets generated and analyzed during the current study are not publicly available due to data-use agreements but are available from the corresponding author on reasonable request.