Recreational cannabis excise taxation in the USA: Constructing a comparable tax measure for empirical analysis

Abstract

Background:

As of August 2023, 20 states in the US have established recreational cannabis retail markets and impose excise taxes on these products. However, there is significant heterogeneity in the bases (i.e., characteristics that taxes are applied to, such as price, weight, and potency), rates, and collection points (e.g., cultivation vs. wholesale) of excise taxes on recreational cannabis across states.

Methods:

We constructed a novel cannabis excise tax measure in $ per flower oz, which is comparable across different tax bases. Specifically, ad valorem excise taxes based on wholesale and retail prices and THC-based taxes were converted to excise taxes ($) per oz using monthly state-level prices between 2014 and 2023. We also calculated tax incidence (i.e., taxes as a percentage of the retail prices) and analyzed its association with tax bases and converted taxes using ordinary least square (OLS) regressions.

Results:

The mean and median values of converted excise taxes on recreational cannabis flowers were $37.93 and $37.55 per oz, respectively. The tax incidence for recreational cannabis was 18%, lower than the incidence of e-cigarette and cigarette excise taxes. During 2014–2023, real cannabis taxes and prices have decreased significantly over time. In addition, tax bases and converted excise taxes were not associated with tax incidence.

Conclusion:

As the prices and taxes of recreational cannabis continue to decrease, tax incidence remains low and is not significantly associated with tax bases or rates, posing concerns about whether the current levels of excise taxes are large enough in reducing cannabis use. Future research shall investigate this matter using converted cannabis taxes empirically. In addition, the wide range of tax magnitude and incidence across states suggests that tax avoidance opportunities may exist for recreational cannabis users who live in higher-taxed states to purchase in neighboring states with lower taxes.

1. Introduction

There has been a significant trend of relaxing recreational cannabis through legalizing the cannabis retail market in the US. As of August 2023, 24 states and the District of Columbia have legalized recreational cannabis use (Hansen et al., 2023). Among these locations, 20 states have established a retail marketplace of cannabis and imposed excise taxes on recreational cannabis to curb its use as well as collect tax revenue (Insurance Institute for Highway Safety, 2023). However, due to variations in state tax laws and the absence of federal recognition for the legality of cannabis, there is no unified model for taxing recreational cannabis, and various bases (i.e., specific criteria upon which a tax is levied,) and rates have been adopted for recreational cannabis taxes. As a result, it is challenging to assess and compare the magnitude of taxes across states with different tax bases (e.g., tax dollars as 10% of prices is not directly comparable to tax dollars as $50 per flower ounce (oz)). Hence, it is necessary to construct a comparable excise tax measure to allow for the evaluation of the magnitude of cannabis taxes with different bases (i.e., price vs. weight vs. tetrahydrocannabinol (THC)), which will further allow for future evaluation of how taxes influence downstream outcomes, such as consumption and tax revenues. The primary objective of this study is to construct a novel comparable measure of cannabis taxes for all states that have opened retail sales and implemented excise taxes.

The importance of constructing a comparable cannabis measure is further stressed by the lack of empirical evidence on the tax effectiveness in impacting cannabis use. It is widely acknowledged that a comparable tax measure is the first necessary step to facilitate tax effectiveness analysis. As a similar effort, e-cigarette taxes based on prices have been converted to a comparable measure per volume to estimate how e-cigarette taxes impact demand (Cotti et al., 2023; Diaz et al., 2023). Since cannabis use increases the risks of various adverse health outcomes, including schizophrenia and other psychoses, (National Academies of Sciences, Engineering, and Medicine, 2017), policymakers call for evidence on the effectiveness of taxation policies on health outcomes. This need has increased in recent years, as cannabis use in 2021 reached a record high of around 42.6% among adults aged 19 to 30 and 27.8% among US high school students (Patrick et al., 2022). Our novel tax measure will allow for future research to empirically examine whether cannabis excise taxes reduce cannabis use in adolescents and at-risk populations.

Other than behavioral outcomes, tax incidence, defined as taxes as a percent of retail prices, has been widely used by the World Health Organization as both a policy target and a measure of alcohol and tobacco tax levels (World Health Organization, 2022). The second objective of this study is to calculate cannabis excise tax incidence for each taxing state using comparable taxes and further assess how cannabis tax incidence is associated with tax bases and rates. With increasingly liberalizing cannabis regulations in the US and the availability of high-potency cannabis, there is a debate over choosing the optimal tax base among prices, weight, and potency (i.e., THC) to curb cannabis use disorder and protect public health (Firth et al., 2020; Pacula et al., 2022). Our analysis aims to identify which tax base leads to the highest tax incidence and has the greatest potential for raising prices and reducing problematic use such as substance use disorder.

Finally, for implementing sustainable cannabis tax policies, their revenue generation role needs to be assessed. Another goal of this study is to examine the trend of comparable cannabis taxes, which illustrates how tax revenue from each ounce of cannabis sales changes over time. Moreover, as a growing number of states legalizes and taxes retail cannabis, tax avoidance opportunities may exist for neighboring states that implement different levels of taxes. This study will plot a US map of converted cannabis taxes in 2023 to demonstrate whether there are incentives for consumers to travel to neighboring states with lower cannabis taxes to purchase cheaper cannabis products (i.e., cross-border shopping).

In summary, converting recreational cannabis taxes will allow researchers and policymakers to compare the magnitudes of taxes across states, evaluate the impacts of taxes on cannabis use, identify optimal tax bases, and assess revenue consequences and tax avoidance opportunities. However, such conversion has not been performed in the existing literature. To address this critical gap, this study innovatively converted the cannabis excise taxes across tax bases into common tax levels per flower ounce using cannabis price and tax data from 2014 to 2023, further calculating the tax incidence and assessing its association with tax bases and rates. As the first study to convert cannabis excise taxes across states into a comparable measure, our findings will pave the road for future studies that assess behavioral and tax revenue consequences.

2. Methods

2.1. Cannabis Tax Policy

To convert heterogeneous cannabis taxes across US jurisdictions into one unified measure, we comprehensively searched for cannabis tax policies across states to date (Alaska Department of Revenue, 2023; Arizona Department of Revenue, 2023; California Department of Tax and Fee Administration, 2023; Cannabis Control Commission Massachusetts, 2023a; Colorado Department of Revenue, 2023a; Connecticut State Department of Revenue Services, 2023; Washington State Department of Revenue, 2023; Illinois Department of Revenue, 2023; Maine Department of Administrative and Financial Services, 2023; Maryland Cannabis Administration, 2023; Michigan Department of Treasury, 2023; Missouri Department of Revenue, 2023; Montana Department of Revenue, 2023; New Jersey Division of Taxation, 2023; New Mexico Taxation & Revenue, 2023; New York Department of Taxation and Finance, 2023; Oregon Department of Revenue, 2023; Rhode Island Department of Revenue, 2023; State of Nevada Department of Taxation, 2023; Vermont Department of Taxes, 2023; Washington State Department of Revenue, 2023). Focusing on jurisdictions that legalized recreational cannabis use, Appendix Table 1 summarizes recreational cannabis tax policies between January 2014 and August 2023 following the National Institute on Alcohol Abuse and Alcoholism (NIAAA)’s format to break down a substance tax into different types of taxes (National Institute on Alcohol Abuse and Alcoholism, 2023). During the data period, we identified 20 recreational cannabis states that opened recreational dispensaries (i.e., retailers selling recreational cannabis products) by August 2023. It is worth noting that actual recreational dispensary opening dates may differ from the dates of each state’s legalization of recreational cannabis use. Therefore, we used the dispensary opening date instead of legalization dates to construct a comparable tax measure (Insurance Institute for Highway Safety, 2023; ProCon.org, 2023).

Among the three types of bases, ad valorem excise taxes are the most commonly used in the US (17 out of 20 states), where taxes are calculated as a percentage of wholesale or retail prices and increase as price increases. In addition, 5 out of 20 states apply a weight-based specific tax, where taxes increase as the weights of flowers or flowers used for edibles or extraction increase (Auxier & Airi, 2023). Two states - Connecticut and New York adopt potency-based taxes and impose higher excise taxes on products with higher contents of THC. THC is the primary psychoactive cannabinoid extracted from cannabis plants and is used recreationally due to its psychoactive effects (National Institute on Alcohol Abuse and Alcoholism, 2023). Potency-based taxes directly target the harm-inducing chemical, THC, and may be best suited to reduce adverse health consequences of risky cannabis use. However, potency-based taxes require accurate labeling and testing that incur extra administration costs.

In addition to tax bases, states may impose differential (i.e., tiered) rates based on product characteristics (e.g., forms and THC levels) and collect taxes at different points in the supply chain, such as cultivation, wholesale, and retail levels (see Figure 1), leading to additional complexity in tax policies and magnitudes. As of 2023, 6 states levy taxes at more than one collection point in the supply chain (e.g., cultivation and wholesale levels), while 14 states levy taxes only at one collection point in the supply chain. Figure 1 presents a brief supply chain model of cannabis products transferred in the market. Cultivators produce cannabis plants in raw forms (i.e., unprocessed cannabis plant components such as harvested flower, leaves, and stems) and they are transferred to wholesalers with processing, and then to retailers. In some cases, cultivators may function as wholesalers as well. In each stage of the supply chain, cannabis prices are determined given each agent’s determined markup rates, and cannabis taxes are levied according to each tax jurisdiction.

Figure 1.

Supply chain model of cannabis products

Note: In each stage of the supply chain, cannabis prices may be set by the supplying agent given their own markup rates determined. According to the tax jurisdictions, cannabis transfers may be subject to cannabis excise taxes at each stage. In addition, we take into account direct cannabis transfers from cultivators to retailers in the supply chain, where cannabis cultivators are expanding their role to not just cultivate the product but also distribute it to retailers in the form of vertical integration (“Stand-alone cultivators” as noted by Hansen et al. (2022)).

As shown in Appendix Table 1, cannabis tax bases and rates are heterogeneous across states and subject to subsequent changes since the legalization and opening of the retail markets. Importantly, cannabis flowers have been the most common type of cannabis product in the US (Panchalingam et al., 2023), and most states do not have taxes specific to other types of cannabis products. Hence, we focus on constructing a recreational cannabis tax measure per cannabis flower ounce at the state level (i.e., ignoring locality-level taxes given complexity). Further, we delineate cannabis excise taxes by each stage of the cannabis supply chain (i.e., cultivation, wholesale, and retail levels) as well as by types of taxes and tax bases (i.e., weight-based or potency-based specific excise tax or ad valorem excise tax) before aggregating the excise taxes from different stages or bases into a single measure.

We did not incorporate sales taxes in constructing converted taxes because these taxes are universally applied to all goods and services and are not considered the target of health taxes. The rich literature on alcohol and tobacco taxes has been focused on excise taxes instead of sales taxes (Chaloupka et al., 2012; Chaloupka et al., 2019). Nonetheless, in states where sales taxes are exempted for cannabis when excise taxes are in place, the effective excise tax rates should be adjusted to the published excise tax rates minus sales tax rates. This method has been adopted by the NIAAA to build their databases of effective excise tax rates for alcoholic beverages, for which sales tax exemption is common. Therefore, we reported both sales taxes and sales taxes adjusted excise tax rates along with other tax information in Appendix Table 1.

2.2. Cannabis Prices

On top of the cannabis tax information, we incorporate cannabis price information into our dataset to compute effective tax amounts for each taxing state. Due to the difference in price data availability, a variety of sources are used (see Appendix Table 2). In principle, if a state publishes cannabis prices at any stage in supply chains, we utilize those prices in converting taxes. If the prices are not publicly available, we use the Cannabis Benchmarks (CB) prices – a leading provider of financial, business, and industry data – for conversion. All the reportable tables and figures have included data from 16 out of 20 states between January 2014 and June 2023 (Maryland, Missouri, Montana, and New York are dropped due to unavailable price data).

Wholesale level

Three states: Colorado, Nevada, and Oregon published wholesale-level average prices at least quarterly (Colorado Department of Revenue, 2023b; State of Nevada Department of Taxation, 2023; Oregon Liquor and Cannabis Commission, 2023). For the rest of the states that do not publish their wholesale prices, we rely on the CB’s weekly transaction volume weighted cannabis prices (both medical and recreational) for flower across jurisdictions starting from April 2015, which collects price information through their network of licensed supply chain agents (e.g., cultivators, distributors, and retailers) and includes Alaska, Arizona, California, Colorado, Connecticut, Illinois, Maine, Massachusetts, Michigan, Nevada, New Mexico, Oregon, Rhode Island, Vermont, and Washington (New Leaf Data Services, 2023). We confirm that the CB’s wholesale prices and publicly available wholesale prices from Colorado, Nevada, and Oregon have overlapping 95% confidence intervals, further validating the quality of CB price data. (Colorado Department of Revenue, 2023b; State of Nevada Department of Taxation, 2023; Oregon Liquor and Cannabis Commission, 2023).

Cultivation level

Only one state – Colorado publishes cultivation-level prices in the form of average market rates (AMRs) from January 2014, which are reported about every quarter (sometimes more frequently). AMRs reflect median market prices when cannabis products are sold or transferred from cultivation facilities to manufacturing facilities or retail stores (Colorado Department of Revenue, 2023c). Therefore, the AMR prices reflect cultivation prices, as well as wholesale prices, if the sales or transfers happen directly between cultivators and retail stores (i.e., this indicates vertical integration that cultivators function also as wholesalers). We then compared Colorado’s wholesale prices from CB with the AMR prices and found that AMR prices are within the 95% confidence interval of CB prices, suggesting that cultivators likely play the role of wholesalers and sell directly to retail stores. Based on this observation, we consider that cultivation prices are equivalent to wholesale prices for all states when converting cannabis taxes.

Retail level

At the retail level, Colorado (monthly averages), Massachusetts (weekly averages), Michigan (monthly averages), and Oregon (weekly median prices) have published cannabis retail prices of flower since the legalization of each state (Colorado Marijuana Enforcement Division, 2023; Cannabis Control Commission Massachusetts, 2023b; Cannabis Regulatory Agency, 2023; Oregon Liquor and Cannabis Commission, 2023) and we use these prices when calculating taxes imposed at the retail level. For the rest of the states that do not publish retail prices but have wholesale prices available, we use wholesale to retail price markup rates to calculate retail prices. The detailed method is described in the next section.

Calculating wholesale to retail price markup rates to fill in missing wholesale or retail prices.

We had both wholesale and retail prices for four states – Colorado, Massachusetts, Michigan, and Oregon – allowing us to calculate wholesale to retail price markup rates for these states. In addition, California published their wholesale to retail price markup rates periodically until 2022. Using these data, we estimated the average markup rate from 2014 to 2023 to be 108%, ranging from 30% to 265%. Moreover, two out of the five states, Colorado and Massachusetts, show an average wholesale to retail markup rate of 97%−98%. Therefore, we use this 97% markup rate when converting wholesale prices into retail prices (and vice versa) to fill out unavailable retail or wholesale prices in the dataset (see Appendix Table 2).

Specifically, we convert wholesale prices into retail prices (and vice versa) to fill in missing prices using the following formula:

$$\mathit{\text{Wholesale}}\mathit{\text{price}}\mathit{\text{per}}o{z}_{st}=\mathit{\text{Retail}}\mathit{\text{price}}\mathit{\text{per}}o{z}_{st}/(1+\mathit{\text{markup}}\mathit{\text{rate}})$$ (1)

where s indicates a state, t represents a quarter, which ranges from Q1/2014 to Q2/2023, and the markup rate is set at 97%. We have also calculated missing prices using alternative rates such as 30% (minimum markup) and 265% (maximum markup). These results can be obtained from the authors upon request.

2.3. Overview of Methods

To construct a comparable cannabis tax measure across states, we focus on recreational cannabis taxes per flower oz at the tax jurisdiction level (i.e., state). Cannabis flowers have been the most common type of cannabis products in the US (Panchalingam et al., 2023), and most states do not have taxes specific to other types of cannabis products. All statistical analyses were conducted using Stata 18/SE 18.0 (StataCorp LLC, College Station, TX). Ordinary Least Squares regressions were employed for the necessary empirical analysis.

2.4. Effective Excise Tax Rate and Tax Conversion

Cannabis excise taxes can be imposed on different bases and in three different stages of the cannabis supply chain: cultivation level, wholesale level, and retail level. The taxes are usually levied on one of the three levels but can also be imposed on multiple supply chain levels. To construct a comparable cannabis excise tax measure across the US states, we first compute each stage’s effective excise tax rate using the following equation where s represents state and t represents quarter:

$$\mathit{\text{Excise}}\mathit{\text{tax}}\mathit{\text{per}}o{z}_{st}=\mathit{\text{Specific}}\mathit{\text{tax}}\mathit{\text{per}}o{z}_{st}+\mathit{\text{Cannabis}}\mathit{\text{price}}\mathit{\text{per}}o{z}_{st}*Ad\mathit{\text{valorem}}\mathit{\text{rat}}{e}_{st}$$ (2)

Note that states with specific taxes are either based on weight or THC mg. For states with taxes based on THC mg, we calculated the average THC mg per flower oz to be 5669.9, using an average THC level of 20% for flower products (Smart et al., 2017; Mahamad et al., 2020). In addition, corresponding cannabis prices that match the supply chain stage (e.g., wholesale prices for wholesale level excise taxes) are used for computing excise taxes per oz, following tax policies documented in Appendix Table 1. In all cases with data, states have set only one of the two types of tax bases per supply chain stage: either specific excise tax (in $ per unit) or ad valorem excise tax (in percentage of prices), so whenever specific taxes have been imposed for a state within a supply chain stage (e.g., cultivation level), ad valorem excise tax within the same supply chain stage is 0. Once each stage’s excise taxes have been computed for states given time, we sum them up to determine the converted cannabis excise tax dollar per flower oz as the following:

$$\mathit{\text{Converted}}\mathit{\text{excise}}\mathit{\text{tax}}\mathit{\text{per}}o{z}_{st}=\mathit{\text{Excise}}\mathit{\text{tax}}\mathit{\text{per}}o{z}_{\mathit{\text{cultivation}},st}+\mathit{\text{Excise}}\mathit{\text{tax}}\mathit{\text{per}}o{z}_{\mathit{\text{wholesale}},st}+\mathit{\text{Excise}}\mathit{\text{tax}}\mathit{\text{per}}o{z}_{\mathit{\text{retail}},st}$$ (3)

Detailed demonstration of types of cannabis taxes in each supply chain stage one by one is provided in the Appendix section.

2.5. Empirical Strategy

After converting cannabis excise taxes per flower oz as described, we evaluate the tax and price distribution using summary statistics at the state-quarter level between Q1/2014 and Q2/2023, which reports wholesale and retail prices and converted excise tax per flower oz in nominal and inflation-adjusted real terms using constant 2022 US dollars (Bureau of Labor Statistics, 2024).

In order to assess the magnitudes of taxes, we derive the effective tax incidence, defined as the converted excise tax as a percentage of the retail price per flower oz, and plot its distribution. Furthermore, we plot the average tax magnitudes and the effective tax incidence across states, which discern states that levy comparatively higher taxes on recreational cannabis. We also examine if tax bases (specific base as omitted category, ad valorem base dummy, and mixed base dummy indicating a mix of specific and ad valorem bases) are associated with tax incidence using ordinary least squares regressions after controlling for converted excise taxes per ounce and year fixed effects.

$$\mathit{\text{Tax}}\mathit{\text{incidenc}}{e}_{st}=Bas{e}_{st}+\mathit{\text{Converted}}\mathit{\text{excise}}\mathit{\text{tax}}\mathit{\text{per}}o{z}_{st}+\mathit{\text{Yea}}{r}_t+{ϵ}_{st}$$ (4)

Next, to provide a comprehensive view of the evolution of the cannabis market since the legalization of recreational cannabis, we chart the trends in cannabis prices and taxes. We further conduct trend analysis by regressing prices or taxes on a linear year trend, testing whether prices or excise taxes increase or decrease over time in both nominal and real terms. Further, because different states legalized recreational cannabis in different years, we conduct a leave-one-out analysis to examine whether the trends are sensitive to including states with varying legalization timing. Specifically, we dropped one state from the sampling each time and conducted the trend analysis 16 times.

Finally, given that taxes computed using prices may be endogenous and subject to supply chain responses to tax policies, we also generate alternative tax measures per ounce for ad valorem excise taxes using time-invariant average prices, which include 1) average price over all periods and 2) average price over each state’s first available years, respectively. These prices may be more appropriate for evaluating tax impacts on cannabis use using causal inference methods such as a two-way fixed effects model with state and year fixed effects. However, given the recency of recreational cannabis taxes, converted tax measures may be collinear with state and year fixed effects. Therefore, we examine the Variance Inflation Factor (VIF) – a measure to evaluate multicollinearity for multiple forms of converted excise tax (i.e., time-variant and time-invariant measures) by regressing these taxes on state and year fixed effects to compute the VIF as 1/(1–R²). The rule of thumb is that if the VIF value is over 10, then multicollinearity is high (Kutner et al., 2004). A higher VIF implies more severe multicollinearity and decreasing statistical power to identify the tax impacts on outcomes in a two-way fixed effects model.

3. Results

Based on the variables defined and methods of tax conversion, Table 1 presents summary statistics of wholesale price, retail price, and converted excise tax per flower ounce constructed using time-variant prices between Q1/2014 and Q2/2023. By definition, converted excise tax per flower ounce reflects both specific excise tax and ad valorem excise tax, given each state’s corresponding tax policies. Throughout the data period, the mean and median price per flower ounce were $242.31 ($267.82 in 2022 dollars) and $217.09 ($235.71 in 2022 dollars), respectively, at the retail level, and $121.11 ($133.65 in 2022 dollars) and $112.06 ($124.37 in 2022 dollars) at the wholesale level. The average converted excise tax was $37.93 ($42.14 in 2022 dollars) per flower ounce, with a median of $37.55 ($40.90 in 2022 dollars) and a range of $1.10 to $100.96. Excise tax accounted for 18.34% of retail prices per flower ounce. Previous studies estimated an average monthly cannabis consumption of 0.55 to 1.24 oz among cannabis users in Washington and New York (Kerr & Ye, 2022; Caulkins et al., 2020). This could approximately translate into a monthly cannabis purchase of $147.30 to $332.09 ($23.17 to $52.25 for excise taxes) for regular cannabis users in the US. The interquartile range (IQR) – a measure to assess dispersion – is smaller than mean and median values for prices and taxes, suggesting limited dispersions. Appendix Table 3 (i.e., Appendix Tables 3.1–3.6) contains state-quarter level data on converted cannabis taxes based on time-variant prices between Q1/2014 and Q2/2023 in addition to the converted taxes based on time-invariant prices.

Table 1.

Summary statistics of wholesale and retail prices and converted taxes in nominal and real terms (in 2022 dollars)

Variables	N	Mean	SD	Median	Min	Max	IQR
Wholesale price per ounce (nominal)	242	121.11	62.90	112.06	35.39	313.18	89.69
Retail price per ounce (nominal)	242	242.31	126.23	217.09	69.72	616.97	186.10
Converted excise tax per ounce (nominal)	250	37.93	15.55	37.55	1.10	100.96	24.00
Wholesale price per ounce (constant)	242	133.65	71.04	124.37	35.71	367.54	91.55
Retail price per ounce (constant)	242	267.82	143.96	235.71	73.24	724.06	206.93
Converted excise tax per ounce (constant)	250	42.14	18.65	40.90	1.08	124.30	26.28
Converted excise tax as % of retail price per ounce	242	18.34	8.39	17.00	8.12	50.38	7.51

Note: Data are state-quarter level between Q1/2014 and Q2/2023. Unit is USD per cannabis flower ounce. Whenever necessary, 97% of the retail-wholesale markup rate was used to convert prices. For data, 16 out of 20 states have been included: Maryland, Missouri, Montana, and New York were not included due to unavailability of prices. US CPI-U for all cities (base year = 2022) was used to adjust nominal values into real terms. Converted excise tax per ounce is a converted cannabis tax level applied per ounce reflecting both specific excise tax and ad valorem excise tax given each state’s corresponding tax policies. SD: Standard deviation; IQR: Interquartile range.

Figure 2 shows trends in cannabis price and excise tax between 2014 and 2023 in the US. Figure 2a and Figure 2b, respectively, depict trends in nominal and real terms. Since 2014, cannabis retail prices in real terms (i.e., red line with square marker) demonstrated an overall decreasing trend (p < 0.05). While wholesale prices (i.e., blue line with circle marker) show slightly decreasing trends in nominal and real terms (Figures 2a and 2b), these trends were insignificant. In addition, as many states have adopted retail-level ad valorem excise taxes that decrease with prices, cannabis converted excise taxes (i.e., green line with triangle marker) have shown a decreasing trend overall in both nominal and real terms (p < 0.01). In Appendix Figure A1, we further examine trends of cannabis price and excise tax (in real terms) based on leave-one-out analysis. Except for a few cases, the results were robust to omitting a random state’s observations in the analysis.

Figure 2.

Trends of cannabis price and excise tax between 2014 and 2023 in the US (in both nominal and real terms)

Note: Data are state-quarter level between Q1/2014 and Q2/2023. Unit is USD per cannabis flower ounce. Whenever necessary, 97% of the retail-wholesale markup rate was used to convert prices. For data, 16 out of 20 states have been included: Maryland, Missouri, Montana, and New York were not included due to the unavailability of prices. US CPI-U for all cities (base year = 2022) was used to adjust nominal values into real terms. P-values for testing the significance of decreasing trends are marked for each line.

As noted in the Methods section, we estimated the VIF values of different tax measures to examine whether there is sufficient variation for causal inference modeling. The VIFs for time-variant measures were 6.42 and 7.24 (in nominal and real terms). For the time-invariant measure based on the average price over all periods, it was 25.97 and 21.64 (in nominal and real terms). For the time-invariant measure based on the first available years’ average price, it was 35.65 and 28.12 (in nominal and real terms). The high VIF values over 10 for converted taxes based on time-invariant prices suggest that these measures have severe multicollinearity with state and year fixed effects and lead to insufficient variation in tax variables in a two-way fixed effect model. In contrast, tax measures based on time-variant prices meet the rule of thumb and may be more appropriate for conducting causal inference models.

We present the average cannabis tax magnitude and incidence levels by states between 2014 and 2023 in Figure 3 and further portray the tax magnitude as of Q1/2023 in a US state map in Figure 4. The tax magnitude is a converted cannabis tax level applied per ounce, reflecting both specific excise tax and ad valorem excise tax, and tax incidence is a converted cannabis excise tax level expressed as a percentage relative to the retail price per ounce. The results show that Illinois had the highest average excise taxes on recreational cannabis ($58.99 and $62.82 in nominal and real terms, respectively), whereas New Jersey had the lowest average excise taxes ($1.27 and $1.24, respectively). Figure 3c displays tax incidence; Washington had the highest tax incidence (37.42%) while Rhode Island and Michigan had the lowest incidence, corresponding to the 10% of the ad valorem excise tax at the retail level. Figure 4 shows that as of Q1/2023, there is a wide range of tax magnitudes, from $1.48 per flower ounce in New Jersey to $48.55 in Alaska.

Figure 3.

Cannabis Tax Magnitudes across the US States in nominal and real terms (in 2022 dollars), 2014–2023

Note: Data are state-quarter level between Q1/2014 and Q2/2023 and each bar represents the average values of the states during the sample period. For Figure 3c, New Jersey was not included due to the unavailability of price information. US CPI-U for all cities (base year = 2022) was used to adjust nominal values into real terms. Cannabis Tax Magnitude is a converted cannabis tax level applied per ounce, reflecting both specific excise tax and ad valorem excise tax, given each state’s corresponding tax policies. Cannabis Tax Incidence is a converted cannabis excise tax level, expressed as a percentage relative to the retail price per flower ounce.

Figure 4.

Cannabis Tax Magnitudes across the US States in 2022 dollars, as of Q1/2023

Note: Average values of converted tax per cannabis flower ounce (in real terms) were used to depict cannabis tax magnitudes in the US as of Q1/2023. US CPI-U for all cities (base year = 2022) was used to adjust nominal values into real terms. Maryland, Missouri, Montana, and New York were legal states with recreational cannabis, but tax conversion was not available due to the unavailability of prices. Grey-colored states denote states without legal recreational cannabis, including the four states without conversion. Cannabis Tax Magnitude is a converted cannabis tax level applied per ounce, reflecting both specific excise tax and ad valorem excise tax, given each state’s corresponding tax policies.

Table 2 demonstrates regression results of how tax incidence is associated with tax bases and converted excise taxes per flower ounce. The findings, however, show that tax incidence is not associated with either tax bases or converted excise taxes per flower ounce at any conventional level of statistical significance. This is not surprising given the decreasing trends of price and tax over time.

Table 2.

Regression analysis of tax incidence on converted excise tax and tax bases

Variables	Tax incidence
Converted excise tax	0.082 (0.443) [−0.141, 0.305]
Specific base (omitted category)
Mixed base	1.152 (0.802) [−8.515, 10.820]
Ad valorem base	−1.425 (0.714) [−9.597, 6.748]
Mean of Dep. Var.	18.310
Number of states	15
Number of observations	719
R-squared	0.131

Note: Data are state-month level between January 2014 and June 2023. P-values and 95% confidence intervals are in parentheses and brackets, respectively. Year fixed effects are included, and standard errors are clustered at the state level. New Jersey was dropped from the analysis as no price information is available as of June 2023. For the excise tax variable, US CPI-U for all cities (base year = 2022) was used to adjust nominal values into real terms. Tax incidence was defined as cannabis taxes as a percentage of the retail price per ounce and bounded in [0, 100]. *** p<0.01, ** p<0.05, * p<0.1

4. Discussion

There has been considerable heterogeneity in how the US states that have legalized recreational cannabis tax cannabis products. We create a measure of cannabis excise taxes, enabling empirical comparisons across states despite the variations and evolving nature of state-specific cannabis tax policies. Our study focused on the excise tax per cannabis flower ounce for conversion and tracking changes in recreational cannabis taxes across 16 out of 20 recreational cannabis states in the US.

In our novel dataset of recreational cannabis excise taxes, states have adopted various tax approaches at different levels of the supply chain. Excise taxes at each level may take either a specific form or an ad valorem form, and the choice varies from state to state. Our findings show that after accounting for the differences in tax rates, bases, and collection points, the converted total excise taxes account for about 18.34% of retail prices. In contrast, the state-level excise tax incidences in the US were 28% for cigarettes (42% if federal taxes are added) and 26% to 32% for e-cigarettes (Shang et al., 2023). Therefore, the tax incidence on recreational cannabis is much lower than that on cigarettes and e-cigarettes in the US. The World Health Organization Framework Convention on Tobacco Control recommends that the excise and sales tax incidences together need to reach 75% in order to reduce smoking effectively (World Health Organization, 2024). If policymakers intend to use higher prices to curb the adverse consequences of recreational cannabis use, there is room to increase the tax incidence on recreational cannabis.

Furthermore, our research exhibited trends in cannabis prices and excise tax rates between 2014 and 2023 in the US, showing decreasing retail prices and cannabis excise taxes observed over time. Existing literature suggests that the price elasticity of cannabis demand is likely inelastic (a 10% decrease in prices will lead to a less than 10% increase in consumption) (Pacula & Lundberg, 2014; Davis et al., 2016). As a result, the tax revenue from recreational cannabis could decrease as the decreasing trends in cannabis prices prolong, especially for those states that entirely or heavily rely on ad valorem excise taxes based on prices. In addition, existing literature on tobacco and alcohol taxes suggests that ad valorem excise taxes, compared to specific excise taxes, leads to greater price variability and less effective pricing or taxation policies in reducing consumption (Shang et al., 2014, 2015, 2018, 2019). Future research is needed to elucidate whether ad valorem cannabis excise taxes have similar disadvantages.

We found significant heterogeneity in tax incidence and the magnitude of converted taxes by state for recreational cannabis. The average amount of taxes per flower ounce (in real terms) was as low as $1.24 in New Jersey – merely 2% of the $62.82 taxes imposed in Illinois. In addition, among the 15 states for which the tax incidence was calculated, nine states were in the range of 10%−14%, five were between 16%−21%, and one state (Washington) had a tax incidence of 37%. The wide range of both tax magnitude and incidence not only suggests that there has not been a consensus on how to design tax policies across states but also implies that tax avoidance opportunities may exist for consumers who live in higher-taxed states to purchase in neighboring states with lower taxes (see Figure 4). In addition, given that illegal market preceded the legal market, simply raising cannabis excise taxes significantly could unintentionally lead cannabis users to unauthorized sellers offering cheaper prices. Future research should focus on how to implement higher tax rates while curbing sales by illegal sources. This includes examining reasonable anti-illicit market policies to disincentivize purchases from the black market (Han et al., 2024).

Existing evidence on cigarette taxes suggests that tax bases are associated with price distributions, including levels and variabilities (Chaloupka et al., 2013; Shang et al., 2014, 2015). However, we found cannabis tax bases and converted tax rates are not significantly associated with tax incidence. This finding raises concerns about whether existing excise tax bases and rates, which heavily rely on ad valorem taxes and impose a rate mostly centering around 10%−15% of retail prices, are sufficient to raise prices and reduce cannabis use. It is also likely that as prices and taxes of recreational cannabis continue to trend down, the ability of existing tax systems to raise tax incidence may be dampened. Future research may look into whether tax bases and rates of recreational cannabis are associated with price variability, if not levels, and ultimately how they impact cannabis use behaviors.

Finally, the VIFs demonstrate sufficient variability in the converted tax measures based on time-variant prices, suggesting that these tax measures can be used in future research to conduct causal inference analyses, such as difference-in-differences models. While Cotti et al. (2023) suggested mitigating endogeneity concerns (e.g., price adjustment in response to taxes) by converting e-cigarette taxes using time-invariant prices, this approach may not be necessary for recreational cannabis for several reasons. First, e-cigarettes have been sold in a legal market before the implementation of taxes, whereas taxes and legal sales are implemented or initiated simultaneously for recreational cannabis. It is thus less likely for retailers or manufacturers of cannabis to manipulate the pricing in response to excise taxes. Second, the e-cigarette market becomes more concentrated over time, leading to growing market power for suppliers to set prices. In contrast, the cannabis market has always been competitive due to the competition from the illegal market. Studies further indicate that as states open the legal sales market of recreational cannabis, the market concentration and prices have decreased (Davenport, 2021; Carliner et al., 2017; Caulkins et al., 2018; Levy et al., 2019). Therefore, cannabis suppliers may have limited power to set prices and respond to taxes. For these reasons, converted cannabis taxes based on time-varying prices are likely to have minimal, if any, endogeneity issues.

Nonetheless, we also construct converted recreational cannabis tax measures using time-invariant prices following the approach described in Cotti et al. (2023). Yet, these tax measures tend to have higher VIF values, indicating multicollinearity with state and year fixed effects. Therefore, researchers should exercise caution when applying these tax measures in causal inference approaches. Future datasets with more data periods and variations in tax rates are needed to facilitate the tax conversion using time-invariant prices. As we discussed earlier, converted cannabis taxes based on time-varying prices have a clear advantage at this stage to carry out causal inference analyses and are less likely to be endogenous.

The cannabis industry is still relatively new and rapidly evolving, not only in the US but also worldwide (Canavan et al., 2022; Caulkins et al., 2018). In this regard, how to effectively tax cannabis products, in order to incentivize decreases in cannabis use and ensure proper tax collection, is largely unknown. Our studies show the heterogeneity and diversities in cannabis tax designs, including bases, rates, and collection points. Future research may consider expanding our data to assess tax designs empirically and identify the optimal tax forms for raising prices and reducing the risky use of cannabis.

Our study contributes to the existing literature in several ways. First, we comprehensively document heterogeneous recreational cannabis taxes in the US using up-to-date tax policies since the first legalization of recreational cannabis. To our knowledge, this is the first to do so. Based on this, we converted heterogeneous cannabis excise taxes in the US into specific tax dollars per cannabis flower ounce, which has not been attempted in the literature. Our converted cannabis tax measure will allow empirical researchers to directly compare different cannabis tax levels across states and estimate the cannabis tax elasticity of demand and many other relevant outcomes for causal inferences using quasi-experimental designs in future research. Finally, our analysis of cannabis excise taxes informs policymakers and stakeholders to consider important decisions such as introducing or amending cannabis tax policies (e.g., raising taxes or reforming tax bases).

Yet, our study has some caveats. First, we were only able to report converted cannabis taxes for 16 out of the 20 recreational states to date, with some missing data points due to the unavailability of price data. Second, our tax measures did not consider local optional taxes on recreational cannabis due to the lack of local prices. Therefore, our measure may underestimate tax magnitudes in states with significant local taxes. Future research will need to address this limitation. Third, we converted the cannabis excise taxes focused on cannabis flower products. As the cannabis industry in the US is evolving rapidly, the composition of the types of cannabis products in the market may change significantly. Future research may update the converted cannabis taxes using the latest state-published price data and tax information for other types of cannabis products as available. Nonetheless, cannabis flowers remain the most prevalent form of cannabis products in the US (Panchalingam et al., 2023), and most states tax cannabis based on flower products or do not differentiate how to levy taxes by product types. Fourth, due to the lack of longitudinal THC data, we did not convert all taxes into taxes per % THC, which could have been decreasing due to the increasing THC levels in recent years (Freeman et al., 2020). The availability of state-level THC trends may address this limitation. Fifth, state-level data on markup rates in the supply chain are not available for all states. Future research and effort are needed to collect such data to improve the accuracy of tax conversion. Finally, there might be measurement errors in the price data, as not all states publish cannabis prices at all stages of the supply chain through their track-and-trace systems. While the CB price data are among the best alternatives for replacing unavailable state-published prices in analyses – since they closely resemble actual state-published prices with consistency – future research can improve the quality of our converted tax measures by updating price data using state-published estimates as they become available.

6. Conclusions

As the prices and taxes of recreational cannabis continue to decrease, tax incidence, which measures the total cannabis tax level as a percentage of the retail price, remains low and is not significantly associated with tax bases or rates. This poses concerns about whether the current levels of excise taxes are large enough in reducing cannabis use. Future research shall investigate this matter using converted cannabis taxes. In addition, the wide range of tax magnitude and incidence across states suggests that tax avoidance opportunities may exist for recreational cannabis consumers who live in higher-taxed states to purchase in neighboring states with lower taxes.

Data statement

All relevant data are published with this study.