Skip to main content
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Mar 23.
Published in final edited form as: Addict Res Theory. 2022 Mar 23;30(5):360–367. doi: 10.1080/16066359.2022.2049255

Estimating Usual Grams per Day of Marijuana Use from Purchases

William C Kerr 1, Yu Ye 1
PMCID: PMC9523691  NIHMSID: NIHMS1793720  PMID: 36189378

Abstract

Background:

Quantifying average amounts of marijuana used per day or per occasion of use helps inform understanding population-level patterns of use and use-related harm, but better estimates and estimation methods are needed. Users have difficulty in reporting use amounts, but purchase amounts may be more clearly recalled.

Methods:

Measures of individual’s use and purchasing frequency and details of purchases such as the amounts and products bought and the cost of these were collected in six cross-sectional representative surveys of the population 18 and older in the state of Washington from 2014 to 2016. Analyses utilize purchase information on both flower and other marijuana products to estimate mean amounts per use day and predict use amounts for non-purchasers.

Results:

Mean marijuana use per use day among purchasers was 1.35 grams and non-purchasers estimated mean use amount per day was 0.71 grams. Lower mean use per day was a found for women and the most frequent users. Based on these estimates, total past year marijuana use for purchasers had a mean of 184.8 grams and the mean for non-purchasers was 28 grams.

Conclusions:

Methods based on purchasing details can be used to estimate individual’s marijuana quantity per occasion and total use amount per year, providing additional outcome measures for analyses of predictors of individual marijuana consumption and facilitating more detailed analyses of risks for marijuana harms.

Keywords: marijuana, purchasing, amount per day, measurement, quantity

INTRODUCTION

Research on marijuana use prevalence, trends, risk factors for use and risks from use have largely relied on use frequency, rarely including amounts per use occasion, day, month or year. Quantifying cannabis use amounts is an important underdeveloped area of measurement relevant to all types of health, social, economic, educational and other outcomes (Freeman and Lorenzetti 2020; Volkow and Weiss 2020). The present study develops methods for quantifying use amounts per day and year based on respondents reported marijuana purchases and sharing with and from others as well as methods for estimating use amounts for non-purchasers.

Methods for measuring cannabis use amounts will necessarily vary by assessment mode and legal status. For example, an in-person interview could utilize a sample amount, a smartphone event-level assessment could take a picture of the amount and use method, and a web survey could show pictures of representative amounts. For a telephone survey assessment, the options are more limited and include asking about typical use amounts in terms of grams or joints (Zeisser et al. 2012). However, there are concerns regarding the accuracy of responses as users do not closely track use amounts, which can vary substantially. An alternative strategy for telephone surveys is to estimate a mean daily use amount from measures of cannabis purchased. An advantage of this method is that users may better recall the amounts that they purchase and how often purchases occur, likely increasing reliability compared to direct reports of use amounts. Complications with this method include the sharing of cannabis between users and the fact that some users never purchase cannabis. In a series of surveys in the US state of Washington, questions were included regarding purchase amounts and expenditures and sharing with and from others as well as cannabis use frequency and how high users typically were during use events. This study describes methods for estimating usual use amounts from the purchasing, sharing and use questions along with models of predicted use amounts for respondents who did not report any past year purchases.

Prior studies have utilized frequency of use data from the National Surveys on Drug Use and Health (NSDUH) series, along with data on typical use amounts from other sources, to estimate use amounts and expenditures for Washington cannabis users (Caulkins et al. 2019). Studies have also used data on purchasing in the NSDUH to estimate sales and other aspects of how marijuana is obtained in the US (Davenport and Caulkins 2016). However, our study is the first to utilize individual’s use and purchasing data to estimate average use amounts and to extrapolate these estimates to non-purchasing users. The validity of recent purchase questions is supported by a study that included multiple purchases and found that the most recent was representative of other purchases (Bond et al. 2014).

Adult use legalization in Washington occurred in 2012 but retail stores did not open until July of 2014, with 31 stores open by the end of August, then rising to 311 stores open as of September 2016. Prior to the store openings in 2014, Washington already had liberal medical sales policy and established illicit market with high prevalence of use (Kerr, Ye, et al. 2018). Cannabis market regulations included restrictions on store numbers, seed to sale tracing, bans on home growing and delivery and higher tax rates than other early legalizing states, although the tax structure was simplified and rates were reduced in 2015 (Cambron et al. 2017). Bans on stores and location restrictions were also in place at local levels during this period (Dilley et al. 2017).

Studies of the Washington market have found increased use over time after legalization. These include a raw wastewater analysis from 2014 to 2016 that found evidence of increasing use in 2016 specifically (Burgard et al. 2019) and a survey analysis that found increased any and frequent use after legalization to be associated with access to retail stores, rather than legalization or store openings generally (Everson et al. 2019). The details of cannabis purchases in the legal market from 2014 to 2016 have also been evaluated, with findings that the purchases mostly involved high-THC cannabis flower with a growing share of cannabis extracts and a declining price per gram, particularly from 2014 to 2015 (Smart et al. 2017). More recently, marijuana products, particularly concentrates, vape pens and edibles, have become popular, increasing the importance of more detailed assessment and tracking (Carlini et al. 2020; Schauer et al. 2020).

This study utilizes data from six cross-sectional waves of an alcohol- and marijuana-focused representative survey series of Washington adults conducted in 2014, 2015 and 2016. Studies using these surveys have included retrospective analyses of the privatization of the liquor monopoly and cannabis legalization in 2012 (Kerr, Williams, et al. 2018; Kerr, Ye, et al. 2018). Analyses focus on utilizing measures of purchasing and sharing to estimate average use amounts among past-year purchasers and utilizing predictors of these average use amounts to estimate use amounts for those who did not report purchasing cannabis. Results contribute to the understanding of variation in cannabis use amounts and provide estimates of individual yearly cannabis use amounts for analyses in future studies.

MATERIALS AND METHODS

Sample

The series of Privatization of Spirits in Washington (PSW) Surveys, conducted between January 2014 and December 2016 by ICF International, was designed to evaluate impacts over time of the privatization of spirits sales and the legalization of marijuana in Washington state. The data analyzed consist of six cross-sectional representative surveys of adults (aged 18 and over), with sample recruitment taking place separately in January-April 2014 (Wave 1, N=1202), September-October 2014 (Wave 2, N=804), March-May 2015 (Wave 3, N=823), August-October 2015 (Wave 4, N=662), March-April 2016 (Wave 5, N=610), and September-December 2016 (Wave 6, N=1391). At each wave, respondents were selected using a state random probability sample obtained via random digit dial (RDD). The sample was approximately evenly split between landline and cell interviews. The AAPOR2 cooperation rate (The American Association for Public Opinion Research 2011), complete and partial interviews as a percentage of identified eligible respondents, (landline, cell) were: Wave 1 (50.8%, 59.5%), Wave 2 (45.8%, 62.4%), Wave 3 (43.7%, 61.5%), Wave 4 (41.7%, 59.6%), Wave 5 (49.4%, 60.9%) and Wave 6 (45.3%, 63.0%). At survey completion, participants were issued $10 dollar gift cards. Surveys lasted about half an hour on average. Protocols were approved by the Public Health Institute Institutional Review Board (#I13–010).

Measures

To determine the frequency of marijuana use in the past-year (PY), all respondents were asked “How often have you used marijuana, hash or pot during the last twelve months”, with selection options including “Every day or nearly every day”, “About once a week”, “Once every 2 or 3 weeks”, “Once every month or two”, “less often than that” and “Never last year”. The number of days marijuana was used PY was derived coding the above categories into 300, 52, 21, 10, 5 and 0 days respectively. Respondents who reported any use were defined as PY marijuana users. For all respondents across six waves (n=5,492) there are altogether 1,177 PY marijuana users.

All PY marijuana users were asked “how often do you usually purchase marijuana”, with selection options “About once a week or more”, “About once every 2 weeks”, “About once every month”, “About once every two to three months”, “Less often than that” and “Never during the last year”. PY marijuana purchasing days were derived by coding the above categories into 78, 26, 12, 5, 2, and 0 days respectively. The usual quantity of marijuana flower purchased was based on the question “what amount of RAW marijuana do you usually buy”, with selection options ranging from “0.5 grams (Nickle Bag)” to “an ounce (28 grams)” and “Never”. The selection options also include “Other: Specify”, probing the respondents to provide the specific amount. Among the 1,177 PY marijuana users, 642 reported valid usual purchase quantity of flower PY. Further, for 52 purchasers with missing data on usual purchase quantity who reported usual expenditure, usual purchasing quantity was estimated from expenditure as described below. Finally, 16 purchasers missing both quantity and expenditure were assigned the sample median purchase amount (3.50 grams). Total grams of marijuana flower purchased PY was derived by multiplying the usual grams of flower purchased by the purchase frequency.

Other marijuana product purchase estimates were based on the question “how often do you usually purchase marijuana-related products such as hash, oil, edibles, teas or lotions” asked of all PY marijuana users. Options ranged from “About once a week or more” to “Never during the last year”. These were coded into PY other marijuana product purchasing days similar to flower above. Purchasers of other marijuana products were asked to provide the amount usually bought and how much that amount costs (open-ended questions) for up to three products. The usual amount information was not used, due to a large number of missing cases and the substantial potential errors in reporting and data entries. Instead, the usual expenditure (in $) on other marijuana products was used to estimate the equivalent flower amount (in grams). Since usual expenditure was asked up to three products, for those who reported expenditure on two or three products (n=64 and 38 respectively out of 391 PY other marijuana product purchasers), we estimated the usual expenditure assuming that half of the time the products were purchased separately and half of the time the products were purchased together. The RAND report on legal marijuana market size of Washington state (Kilmer et al. 2019) was used to convert their usual dollar expenditure on other marijuana products and flower (for those missing on usual flower quantity with valid expenditure) into usual flower-equivalent quantity (in grams). Specifically, the 2016 retail revenue (in $) and retail THC sold (in kg) by product types (Table 2.17 and 2.18), with 20.5% THC for flower (Table 2.5) were used for expenditure to quantity conversion. We estimated that $100 can purchase 10.9 grams of flower, 8.75 flower-equivalent grams of extract (hash or resin, oil, wax or dabs), 1.25 flower-equivalent grams of edibles (tea or other beverages, brownies, cookies, candy)/tincture/lotion, or 6.48 flower-equivalent grams of unspecified product. There were 342 PY other marijuana product purchasers who reported valid usual purchasing expenditures, which were converted to flower-equivalent usual purchase quantities; and 49 respondents who were PY product purchasers but missing on usual expenditure who were assigned the sample median for usual purchasing quantity (1.08 grams). Total PY flower-equivalent purchase quantity (in grams) of other marijuana products was derived by multiplying the number of purchasing days by the estimated usual purchase in flower-equivalent grams.

All PY marijuana purchasers were asked “How much of the marijuana or marijuana products that you bought over the past 30 days did you yourself consume?”, with selection options including “All or nearly all”, “About three-quarters”, “About half”, “About one-quarter”, and “None, all of it was consumed by others”. The other question asked purchasers “Thinking about the past month, how does the amount of marijuana that you have shared with others compare with the amounts others have shared with you?” with selection options including “No sharing with others or from others”, “About the same”, “Shared more with others”, and “Received more from others”. The answers from these two self-consumption and sharing questions were combined to calibrate marijuana purchase quantity and to derive the usual amount of use for the purchasers who reported a valid purchase quantity (described below).

Other marijuana use characteristics included Mode of marijuana use, asked as “how do you most commonly consume marijuana”. Four types of products were analyzed: smoked product (smoke marijuana only or with tobacco), inhalable (vaporize marijuana, smoke hashish or resin and oil, wax or dabs through inhalation), edible (eat food product or drink beverage infused with cannabis), and other product (cannabis tincture, lotion, salve, balm or spray, and other). Marijuana users were also asked, when they use marijuana or hashish, the level and length of getting high, based on questions “how high do you usually get?” and “how long do you usually stay high?”, respectively. Selection options for level of getting high ranged from “Not at all high” to “Very high”, and for length of staying high ranged from “Less than one hour” to “More than 24 hours”. Marijuana medical recommendation was based on the question “Do you have a medical recommendation from a health care professional for marijuana or cannabis?” Other marijuana use variables include the marijuana and alcohol co-use and marijuana onset of use. The first measure asked marijuana and alcohol users whether their use of two substances at the same time can be described as “Usually”, “Sometimes”, or “Never”, and the onset question asked the age at which marijuana was first used.

Socio-demographic characteristics included gender, age (18–29, 30–39, 40–49, 50–64, and 65+), race and ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic and all others), Education (high school graduate and less, some college, college graduate, and graduate school), marital status (married, separated/divorced, widowed, and never married), family income (≤$20,000, $20,001–50,000, $50,001–80,000, >$80,000 and missing).

Statistical Analyses

The aim of this study was to calibrate and predict the usual and total quantity of PY marijuana use for (1) purchasers of marijuana flower and other marijuana products based on their PY purchasing amounts and (2) PY marijuana users but non-purchasers.

For the combined six cross-sectional waves from 2014–2016, among 1,177 PY marijuana users, there were 777 purchasers of flower and/or other marijuana products whose PY purchasing amounts were reported, for most flower purchasers, or estimated from purchase expenditures, for purchasers of other marijuana products and a small number of flower purchasers. Among them, 324 were purchasers of both flower and other products, 386 were purchasers of flower only and 67 purchased products only.

As the first step, PY total purchase quantity (in grams) was derived for flower and other marijuana products for purchasers. Unadjusted usual quantity of marijuana use was then derived by dividing the total purchase quantity by PY frequency of use. The total purchase quantity is the summation of total purchases for flower and other products (multiplying usual purchased quantity for flower and products separately by their respective purchase frequencies). This unadjusted usual quantity of use measure was further adjusted by purchasers’ reported self-consumption and sharing to derive the adjusted usual quantity of use. See the detailed calibration algorithm in Appendix Table 1. A key parameter is the percentage used to quantify “received more from others” and “shared more with others”. In our main analysis, ±20% is used for the above two scenarios. For example, assume a purchaser reported “About three-quarters” for the self-consumption question and reported “received more from others” for the sharing question, the usual quantity of use would be 105% (=75%+25%*120%) of the unadjusted usual quantity. In addition, we also used ±10% and ±30% in sensitivity analysis. Last, the adjusted usual quantity of use for purchasers was capped at 14 grams (half ounce).

In order to predict usual quantity of use among non-purchasers, we first fit a negative binomial regression model predicting the adjusted usual quantity of use among purchasers from their marijuana use characteristics (usual frequency of use, level and length of getting high, mode of use, etc.) and socio-demographic factors. Negative binomial regression was utilized given the highly skewed distribution of the usual quantity measure. The fitted model was used to predict usual quantity of marijuana use for the 400 non-purchasers, utilizing their observed marijuana use characteristics and socio-demographic factors. Note that the above approach not only assumes that the associations between the predictive factors and usual quantity are constant between purchasers and non-purchasers, but also assumes that the predicted intercept (i.e. when all predictors have values of zero) is the same across the two groups. This latter assumption may be too strong, since, all things equal in terms of the predictive factors, a purchaser could be expected to use larger amount than a non-purchaser. A further step was taken to correct for this potential bias. Total PY marijuana market size for Washington was estimated: 1) by the purchase quantity of purchasers, 2) by the adjusted quantity of use among the purchasers. Theoretically, the difference between 1) and 2) should equal the total use quantity by the non-purchasing users. An adjustment factor was thus derived dividing the predicted total quantity (i.e. predicted usual quantity multiplied by frequency of use) based on the above regression model for non-purchasing users by the estimated total amount for that group. This adjustment factor was then applied to derive the final predicted quantity of use for non-purchasers. All analyses were performed in STATA version 15 survey command (StataCorp. 2017). The data were weighted to adjust for probability of selection and represent all adults (18 and older) residing in the State of Washington.

RESULTS

Table 1 shows the mean, 95% confidence intervals (CI) and standard deviation (SD) for the unadjusted (derived from purchasing) and adjusted (further accounting for purchasers’ reported self-consumption and sharing) usual quantity of marijuana use for the 777 marijuana purchasers, including both flower and other marijuana products. The adjusted usual quantity was further capped at 14 grams (half ounce), affecting six respondents. The mean unadjusted usual quantity was 1.56 grams per day of use, reduced to 1.35 grams for adjusted usual quantity. Results by gender, age and use frequency show that mean usual quantity was higher for men, highest among the 30–39 age group and lowest among those aged 65 and older, and also lowest for daily or nearly daily users.

Table 1.

Mean usual quantity of marijuana use (grams per day of use, combining flower and other marijuana products) for marijuana purchasers (n=777)

Unadjusted usual grams of use1 Adjusted usual grams of use2

Mean 95% CI SD Mean 95% CI SD

Total 1.56 (1.25,1.87) 3.65 1.35 (1.13,1.56) 2.25
Gender:
 Women 1.40 (0.94,1.85) 4.19 1.08 (0.87,1.29) 1.78
 Men 1.67 (1.26,2.08) 3.24 1.52 (1.19,1.85) 2.49
Age:
 18–30 1.68 (1.03,2.34) 4.78 1.35 (1.03,1.67) 2.28
 30–39 2.05 (1.17,2.92) 3.97 1.89 (1.10,2.69) 3.05
 40–49 1.26 (1.02,1.50) 1.36 1.18 (0.96,1.40) 1.24
 50–64 1.31 (0.97,1.65) 2.66 1.09 (0.83,1.35) 1.92
 65+ 1.03 (0.57,1.50) 1.59 0.97 (0.50,1.44) 1.60
Use frequency:
 (Nearly) every day 1.18 (0.81,1.55) 2.28 1.12 (0.78,1.47) 2.07
 About once a week 1.48 (1.00,1.97) 3.27 1.33 (0.97,1.68) 1.95
 Once every 2/3 weeks 2.25 (0.53,3.97) 5.58 1.69 (0.87,2.51) 2.67
 Once every month or 2 2.06 (1.44,2.69) 3.24 1.74 (1.24,2.25) 2.27
 Less often than that 2.27 (0.94,3.61) 6.34 1.67 (1.07,2.26) 2.87
1

Total grams of marijuana purchased last year divided by frequency of use last year

2

Unadjusted usual grams of use further adjusted by proportion of marijuana purchased that was self-consumed and shared with others, then further capped at 14 grams (half ounce), affecting six persons.

Table 2 presents results from a negative binomial regression model predicting adjusted usual quantity from marijuana use characteristics and socio-demographic factors for purchasers. Presented are incidence rate ratios (IRRs), their 95% CI and the p-values. The estimated IRR of 2.1 for the “very high” level indicates that being “very high” would increase the expected usual grams of use by a factor of 2.1 or 110% compared to “not at all high”. Usual quantity of marijuana use decreased with frequency of use, with daily or nearly daily users having only about 45% of the amount of the least frequent users. No substantial differences were observed across the main product use type categories, but there was a significant decrease associated with the “other type” group. Having a medical recommendation increased the expected usual grams by 24%, but the estimate was not significant. Several marijuana use characteristics (onset of use, length of time getting high, and marijuana and alcohol co-use) were not included based on high p-values (>0.20). For socio-demographic factors, being male significantly increased expected usual quantity, while graduate school education decreased usual quantity.

Table 2.

Negative binomial regression predicting adjusted usual quantity of marijuana use among purchasers (N=777)

IRR1 95% CI P

How high usually got:
 Not at all high 1.00
 A little high 0.76 (0.42, 1.35) 0.350
 Moderately high 1.25 (0.72, 2.15) 0.430
 Very high 2.09 (1.06, 4.11) 0.032
Frequency of use:
 (Nearly) every day 0.45 (0.31, 0.65) <0.001
 About once a week 0.67 (0.45, 0.99) 0.046
 Once every 2/3 weeks 0.87 (0.53, 1.46) 0.606
 Once every month or 2 1.06 (0.69, 1.61) 0.800
 Less often than that 1.00
Medical recommendation 1.24 (0.90, 1.70) 0.189
Type of use:
 Smokable 1.00
 Inhalable 1.16 (0.80, 1.68) 0.442
 Edible 0.78 (0.53, 1.17) 0.232
 Other 0.28 (0.13, 0.59) 0.001
Gender male 1.30 (1.05, 1.61) 0.014
Age :
 18–29 1.00
 30–39 1.34 (0.92, 1.94) 0.126
 40–49 1.05 (0.77, 1.45) 0.751
 50–64 0.95 (0.68, 1.32) 0.760
 65+ 0.84 (0.47, 1.53) 0.574
Race/ethnicity:
 White 1.00
 Black 1.23 (0.81, 1.88) 0.334
 Hispanic 1.01 (0.69, 1.49) 0.962
 Others 0.91 (0.64, 1.29) 0.600
Education:
 HS grad or less 1.00
 Some college 0.91 (0.69, 1.19) 0.481
 College grad 0.92 (0.66, 1.29) 0.630
 Graduate school 0.66 (0.44, 0.99) 0.043
Marital status:
 Married 1.00
 Separate/divorced 0.97 (0.69, 1.35) 0.834
 Widowed 1.22 (0.71, 2.10) 0.471
 Single 1.15 (0.86, 1.54) 0.355
Family income:
 ≤20,000 1.00
 $20,001–$50,000 0.84 (0.62, 1.16) 0.290
 $50,001–$80,000 1.03 (0.71, 1.50) 0.873
 >$80,000 1.05 (0.72, 1.53) 0.788
 Missing 1.26 (0.77, 2.08) 0.357
1

IRR = Incidence rate ratio

Table 3 compares marijuana purchasers and non-purchasing users on marijuana use characteristics and socio-demographic factors. Compared to the purchasers, the non-purchasing users were less likely to report “moderately high” for levels of getting high. The non-purchasers were much less likely to be daily or nearly daily users (14.0% versus 49.2%). They were also less likely to have a medical recommendation. Regarding types of use, they were more likely to have used edibles and less likely to have used inhalables. They were also more likely to be female and more likely to have higher household income.

Table 3.

Comparison of past-year marijuana purchasers and non-purchasers

Purchasers (N=777) Non-purchasing users (N=400) P

How high usually got: <0.001
 Not at all high 6.6% 13.9%
 A little high 28.1% 32.9%
 Moderately high 52.6% 35.7%
 Very high 12.7% 17.5%
Frequency of use: <0.001
 (Nearly) every day 49.2% 14.0%
 About once a week 18.8% 6.8%
 Once every 2/3 weeks 9.0% 5.2%
 Once every month or 2 11.4% 16.6%
 Less often than that 11.6% 57.4%
Medical recommendation 25.8% 12.5% <0.001
Type of use: <0.001
 Smokable 73.4% 77.2%
 Inhalable 15.1% 5.4%
 Edible 9.2% 15.9%
 Other 2.4% 1.5%
Gender male 60.2% 49.9% 0.011
Age: 0.397
 18–29 34.6% 30.5%
 30–39 20.8% 19.2%
 40–49 14.2% 14.2%
 50–64 23.2% 25.1%
 65+ 7.2% 11.0%
Race/ethnicity: 0.060
 White 71.6% 71.6%
 Black 5.4% 2.7%
 Hispanic 12.5% 8.3%
 Others 10.4% 17.3%
Education: 0.267
 HS grad or less 42.4% 35.8%
 Some college 34.5% 37.1%
 College grad 13.8% 17.3%
 Graduate school 9.3% 9.8%
Marital status: 0.389
 Married 47.6% 41.6%
 Separate/divorced 17.6% 20.5%
 Widowed 2.7% 2.2%
 Single 32.2% 35.7%
Family income: 0.039
 ≤20,000 29.7% 26.2%
 $20,001–$50,000 32.5% 26.9%
 $50,001–$80,000 15.3% 16.5%
 >$80,000 17.9% 20.0%
 Missing 4.6% 10.4%
Mean usual grams (95% CI)1 1.35 (1.13,1.56) 0.71 (0.65, 0.77)
Mean total grams (95% CI)1 184.8 (130.6, 239.1) 28.0 (14.5, 41.5)
1

Predicted values for the 400 marijuana non-purchasers

Regression predictions and a market size adjustment were used to estimate the marijuana use quantity for non-purchasing users. The total Washington marijuana market size in flower-equivalent quantity was estimated based on purchases reported by the purchaser group at 192 Metric Tons (MT) in 2014 (158 MT flowers and 33.3 MT flower-equivalent other marijuana products), 176 MT in 2015 (149 MT flower and 26.9 MT products) and 243 MT in 2016 (222 MT flower and 20.3 MT products). Total marijuana use quantity among purchasers, estimated from adjusted use quantity, was 168, 171 and 233 MT for 2014–2016, respectively. The difference between the two estimation methods from purchases or use quantity among purchasers defines the total use quantity of non-purchasing users. However, the estimated use quantity for non-purchasers from the regression model is much larger, requiring adjustment so that the total estimated use quantity by both purchasers and non-purchasers matches the total amount purchased for 2014–2016 combined. The final mean predicted usual quantity of the non-purchasing users was 0.71 grams per day of use, compared to the mean adjusted usual quantity of 1.35 grams for purchasers. The mean predicted PY total quantity of marijuana use for non-purchasers was 28.0 grams, compared to 184.8 grams for purchasers.

Figure 1 presents histograms of the distributions of adjusted or predicted usual quantity separately for purchasers and non-purchasers, as well as the respective PY total quantities. The distribution of use amounts per day among purchasers is skewed, with most cannabis users consuming less than a gram per using day. Estimated use amounts for non-purchasers are less skewed, with nearly all users having less than 2 grams per using day. Because non-purchasers tend to use cannabis less frequently, the estimated PY total quantities consumed are also substantially lower than for purchasers.

Figure 1.

Figure 1.

Distributions of adjusted usual and total quantity of marijuana use for marijuana purchasers (n=777) and predicted usual and total quantity for non-purchasing users (n=400)1

1 Capped at 7 grams for usual quantity (involving 26 purchasers with usual quantity larger than 7 grams) and at 500 grams for total past-year quantity (involving 45 purchasers with total PY quantity larger than 500 grams)

Sensitivity analyses adjusted several parameters in our estimation and prediction procedures. The first re-assigned the top purchasing category, i.e. at least once a week, to 104 days PY (instead of 78 days.). Total PY purchasing quantity market size for flower and products combined changed to 219, 196 and 293 MT for 2014–2016, compared to 192, 176 and 243 MT in the main analysis. The mean adjusted quantity per using day among purchasers increased to 1.42 grams from 1.35 grams. The second sensitivity analysis re-assigned top category of PY use frequency, i.e. everyday or nearly every day, to 234 days from 300 days in main analysis. The mean adjusted quantity per using day among purchasers increased to 1.47 grams. The negative association between usual quantity and frequency of use was still observed with an IRR for daily or nearly daily users of 0.57 (p=0.002), compared to IRR of 0.45 in main analysis. When estimating adjusted usual quantity, we also tested ±30% and ±10% for “receiving more from others” and “sharing more with others”, instead of ±20% in main analysis. The results barely changed, with mean adjusted usual quantity ranging from 1.34–1.35 grams. Capping at 7 rather than 14 grams when calculating adjusted usual quantity changed the mean quantity among purchasers from 1.35 grams to 1.21 grams. Finally, purchasers aged 18–20 had both higher usual and total quantity of use than those aged 21+, emphasizing the importance of including consumption among this group.

DISCUSSION

This study has demonstrated methods for the estimation of marijuana use quantity per day from questions on purchasing, sharing and use frequency in a telephone survey conducted in an environment with legal retail sales as well as methods for the prediction of use quantity for non-purchasers. Results for purchasing users indicate a typical daily amount of 1.35 grams with wide variation skewed toward lower amounts, with most less than 1 gram per day. Larger amounts were associated with male gender, the 30–39 age group and, unsurprisingly, reporting usually getting very high. Interestingly, greater frequency of use is associated with smaller amounts per day, in contrast to some prior studies conducted in non-legal adult use environments (Zeisser et al. 2012). Prior studies have highlighted the importance of the most frequent daily and near daily users for the cannabis market with a US study (Davenport and Caulkins 2016) finding that they used 77% of the cannabis in 2012–13. An Australian study also found daily users accounted for 85% of total cannabis use in 2016 (Chan and Hall 2020).

We also found that purchasers differed from those who did not purchase, with purchasers being more likely to be male and to have lower household incomes as well as to use inhalable products, use more frequently and to usually getting moderately high. Estimated use amounts per use day among users who did not purchase averaged 0.71, about half of the average amount used by purchasers. A key product of these analyses are estimates of users past year total grams of marijuana, usual grams per day multiplied by the frequency of use, which was 6.6 ounces (184.8 grams) for purchasers and 1 ounce (28 grams) for non-purchasers. Total past-year quantities are highly skewed with many low values as shown in Figure 1, emphasizing the importance of distinguishing the high quantity users.

Standard dose amounts for cannabis of 5 and 10 mg THC have been by states, with the 5 mg dose potentially preferred for safety and use in harm reduction strategies (Freeman and Lorenzetti 2020). Assuming a flower THC concentration of 20%, the mean daily use amount among purchasers would contain 270 mg THC, equal to 27 10 mg doses or 54 5 mg doses. These amounts seem large for typical quantities used and very different from the standard drink for alcohol, 14 grams of ethanol in the US, where the mean amount per use day for drinkers in these same Washington surveys was 1.6 drinks (Kerr, Williams, et al. 2018). Cannabis and alcohol differ considerably regarding ingestion methods, impairment, tolerance and other aspects complicating any direct comparisons (Macdonald 2018). However, such large numbers of standard units per day on average complicate the use of these standards for use in research, education and prevention.

Prior analyses from the same Washington survey series utilized here found that use prevalence did not substantially increase with legalization from 2012 to 2014 and 2015 (Kerr, Ye, et al. 2018), but, in 2016 use prevalence did increase significantly (Subbaraman and Kerr 2020). Other results from the same survey series found that prevalence rates of marijuana harms from others’ use were flat from 2014–2016 (Kerr et al. 2021) and that variation in individual’s alcohol and cannabis use in the linked longitudinal sample indicated that more frequent cannabis use was tied to risky drinking (Kerr et al. 2019). Further, voters support for legalization was shown to increase after implementation among those who voted both for and against I502, the initiative establishing legal adult use (Subbaraman and Kerr 2016), and population support for cannabis legalization in Washington continued to increase to 78% in 2016 (Subbaraman and Kerr 2017). The individual per occasion and per year use amounts generated in this study will be utilized in further analyses of the survey series focused on estimating the effects of location-based access to marijuana retail stores and other predictors of use amounts.

Limitations of this method for quantity of use per day estimation include a lack of direct reporting on amounts from non-purchasers and the need to having use frequency as part of the use per day quantity calculation, which may bias more frequent users toward smaller estimated amounts. The use of sharing with and from others in calculating use estimates requires strong assumptions that will vary in accuracy, although we show in sensitivity analyses that alternative assumptions do not have large effects on the estimates. Estimates do not consider variability in use amounts, which could be important particularly for frequent users. Use amounts by non-purchasers are indirectly estimated and may also be more variable as they will depend on others providing the cannabis as well as the users choices. Estimates do not include own-growing, which is illegal in Washington but likely occurred and would add substantial amounts for some users. There is potentially substantial under-reporting of use, although this may be lower in a legalizedstate (Kerr William C., Williams, et al. 2018). Other marijuana product use was estimated from expenditures rather than directly measured. Capturing these accurately has grown in importance and improved measures are needed for future studies (Carlini et al.; Schauer et al. 2020). Despite these significant limitations, the more accurate reporting of purchase amounts compared to use amounts makes this an important contribution to the literature on quantifying marijuana amounts per use day.

Measurement studies aimed at improving the accuracy of cannabis use quantity assessment remain a research priority and methods involving pictures and photographs, which could not be used in our telephone interview, may offer a better path to estimating use amounts for all users. Despite the noted limitations, this study has developed individual use amount estimates designed to sum to total market volume as estimated from self-reported purchase measures. These measures will facilitate planned analyses of relationships between marijuana patterns of use based on both frequency and quantity with relevant outcomes including marijuana problem measures, co-use with alcohol and alcohol-related problems. Past-year quantity will also be an important additional outcome for studies evaluating impacts of cannabis policies and policy-relevant measures such as access to retail stores.

Funding:

This work was supported by the U.S. National Institute on Drug Abuse (NIDA) (R01 DA048526) and the U.S. National Institute on Alcohol Abuse and Alcoholism (NIAAA) (R01 AA021742) at the National Institutes of Health (NIH). Content and opinions are those of authors and do not reflect official positions of NIDA, NIAAA or the National Institutes of Health.

Appendix Table 1: Algorithm how unadjusted usual quantity of use from purchasing is converted to adjusted usual use quantity

Proportion of marijuana bought self-consumed1 Sharing with others2 N (777) Unadjusted usual gram
Main analysis3 (20% inc/des)
Mean Min Max

All or nearly all No sharing with others or from other 100 0.96 0.004 10.92 100.0%
About the same 106 1.68 0.01 55.03 100.0%
Shared more with others 40 2.65 0.003 14.60 100.0%
Received more from others 32 1.05 0.06 7.00 120%
Don’t know or refused 0 100.0%
About three-quarters No sharing with others or from other 7 0.62 0.05 1.95 75.0%
About the same 51 0.98 0.01 13.88 100.0%
Shared more with others 30 1.35 0.06 11.86 95%
Received more from others 6 0.50 0.15 1.12 105%
Don’t know or refused 0 100.0%
About half No sharing with others or from other 15 1.78 0.14 7.32 50.0%
About the same 107 1.84 0.01 35.17 100.0%
Shared more with others 38 1.09 0.03 6.46 90%
Received more from others 5 0.19 0.10 0.67 110%
Don’t know or refused 1 0.45 0.45 0.45 100.0%
About one-quarter No sharing with others or from other 13 0.77 0.24 1.59 25.0%
About the same 44 2.51 0.04 43.08 100.0%
Shared more with others 19 2.87 0.01 17.44 85%
Received more from others 3 0.66 0.06 1.00 115%
Don’t know or refused 5 5.13 0.63 11.50 100.0%
None, all of it was consumed by others No sharing with others or from other 27 1.98 0.02 27.30 10.0%
About the same 34 1.17 0.02 11.63 100.0%
Shared more with others 8 2.68 0.20 14.00 82%
Received more from others 10 2.42 0.14 5.60 118%
Don’t know or refused 6 0.71 0.09 1.83 100.0%
Don’t know or refused No sharing with others or from other 11 0.96 0.18 6.47 100.0%
About the same 10 0.93 0.13 2.67 100.0%
Shared more with others 3 0.75 0.33 0.91 100.0%
Received more from others 9 1.30 0.002 5.60 100.0%
Don’t know or refused 37 1.01 0.002 8.42 100.0%
1

Survey question: How much of the marijuana or marijuana products that you bought over the past 30 days did you yourself consume?

2

Survey question: Thinking about the past month, how does the amount of marijuana that you have shared with others compare with the amounts others have shared with you? Did you share more with others than they shared with you, share less with others than they shared with you, or was it about the same amount?

3

In the main analysis, 20% increase or decrease is used to quantify “received more from others” or “shared more with others”, respectively. For example, assuming a person reporting 100 grams of marijuana usual purchase, and reporting “about half” of the purchase was self-consumed, and “shared more with others”. The adjusted quantity accounting for proportion of self-consumption and sharing with others would be 50%+(1−50%)*(1−20%)=90% i.e. 90 grams. In another scenario, assuming still reporting “about half”, but “received more from others”, the quantity would be 50%*(1−50%)*(1+20%)=110%, i.e. 110 grams.

4

Two types of sensitivity analysis was performed, using 10% or 30% increase/decrease to quantify “received more from others”/“shared more with others” separately.

Footnotes

Disclosure Statement: The authors report no conflict of interest.

REFERENCES

  1. Bond B, Caulkins JP, Kilmer B, Dietze P. 2014. Are users’ most recent drug purchases representative? Drug Alcohol Depend. 142:133–138. [DOI] [PubMed] [Google Scholar]
  2. Burgard DA, Williams J, Westerman D, Rushing R, Carpenter R, LaRock A, Sadetsky J, Clarke J, Fryhle H, Pellman M et al. 2019. Using wastewater-based analysis to monitor the effects of legalized retail sales on cannabis consumption in Washington State, USA. Addiction. 114(9):1582–1590. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Cambron C, Guttmannova K, Fleming CB. 2017. State and national contexts in evaluating cannabis laws: a case study of Washington State. J Drug Issues. 47(1):74–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Carlini B, Barbosa-Leiker C, Cuttler C, Dilley J, Firth C, Haggerty K, Kilmer J, McDonell M, Stella N, Walker D et al. 2020. Cannabis Concentration and Health Risks. Washington State Prevention Research Subcommittee (PRSC). [Google Scholar]
  5. Caulkins JP, Davenport S, Doanvo A, Furlong K, Siddique A, Turner M, Kilmer B. 2019. Triangulating web and general population surveys: do results match legal cannabis market sales? Int J Drug Policy. 73:293–300. [DOI] [PubMed] [Google Scholar]
  6. Chan GCK, Hall W. 2020. Estimation of the proportion of population cannabis consumption in Australia that is accounted for by daily users using Monte Carlo simulation. Addiction. 115(6):1182–1186. [DOI] [PubMed] [Google Scholar]
  7. Davenport SS, Caulkins JP. 2016. Evolution of the United States: marijuana market in the decade of liberalization before full legalization. J Drug Issues. 46(4):411–427. [Google Scholar]
  8. Dilley JA, Hitchcock L, McGroder N, Greto L, Richardson SM. 2017. Community-level policy responses to state marijuana legalization in Washington State. Int J Drug Policy. 42:102–108. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Everson EM, Dilley JA, Maher JE, Mack CE. 2019. Post-legalization opening of retail cannabis stores and adult cannabis use in Washington State, 2009–2016. Am J Public Health. 109(9):1294–1301. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Freeman TP, Lorenzetti V. 2020. ‘Standard THC units’: a proposal to standardize dose across all cannabis products and methods of administration. Addiction. 115(7):1207–1216. [DOI] [PubMed] [Google Scholar]
  11. Kerr WC, Williams E, Patterson D, Karriker-Jaffe KJ, Greenfield TK. 2021. Extending the harm to others paradigm: comparing marijuana- and alcohol-attributed harms in Washington State. J Psychoactive Drugs. 53(2):149–157. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Kerr WC, Williams E, Ye Y, Subbaraman MS, Greenfield TK. 2018. Survey estimates of changes in alcohol use patterns following the 2012 privatization of the Washington liquor monopoly. Alcohol Alcohol. 53(4):470–476. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Kerr WC, Ye Y, Greenfield TK. 2019. Spirits purchasing and marijuana use behaviors of risky drinkers in the state of Washington from 2014 to 2016. Drug Alcohol Depend. 198:7–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Kerr WC, Ye Y, Subbaraman MS, Williams E, Greenfield TK. 2018. Changes in marijuana use across the 2012 Washington state recreational legalization: Is retrospective assessment of use before legalization more accurate? J Stu Alcohol Drugs. 79(3):495–502. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Kilmer B, Davenport S, Smart R, Caulkins JP, Midgette G. 2019. After the Grand Opening: Assessing Cannabis Supply and Demand in Washington State. Santa Monica, CA: RAND Corporation. [Google Scholar]
  16. Macdonald S 2018. Cannabis Crashes: Myths and Truths. First Edition ed. Scott Macdonald. [Google Scholar]
  17. Schauer GL, Njai R, Grant-Lenzy AM. 2020. Modes of marijuana use--smoking, vaping, eating, and dabbing: results from the 2016 BRFSS in 12 states. Drug Alcohol Depend. 209. [DOI] [PubMed] [Google Scholar]
  18. Smart R, Caulkins JP, Kilmer B, Davenport S, Midgette G. 2017. Variation in cannabis potency and prices in a newly legal market: evidence from 30 million cannabis sales in Washington state. Addict Recovery. 112(12):2167–2177. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. StataCorp. 2017. Stata Statistical Software: Release 15 College Station, TX: StataCorp LLC. [Google Scholar]
  20. Subbaraman M, Kerr WC. 2020. Subgroup trends in alcohol and cannabis co-use and related harms during the rollout of recreational cannabis legalization in Washington state. Int J Drug Policy. 75:102508. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Subbaraman MS, Kerr WC. 2016. Marijuana policy opinions in Washington state since legalization: would voters vote the same way? Contemp Drug Probl. 43(4):369–380. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Subbaraman MS, Kerr WC. 2017. Support for marijuana legalization in the US state of Washington has continued to increase through 2016. Drug Alcohol Depend. 175:205–209. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. The American Association for Public Opinion Research. 2011. Standard Definitions: Final dispositions of case codes and outcome rates for surveys, Revised 2011, 7th Edition [Accessed: 2011–05-18. Archived by WebCite® at http://www.webcitation.org/5ymByeilL]. Deerfield, IL: The American Association for Public Opinion Research. [Google Scholar]
  24. Volkow ND, Weiss SRB. 2020. [Commentary] Importance of a standard unit dose for cannabis research. Addiction. 115(7):1219–1221. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Zeisser C, Thompson K, Stockwell T, Duff C, Chow C, Vallance K, Ivsins A, Michelow W, Marsh D, Lucas P. 2012. A ‘standard joint’? The role of quantity in predicting cannabis-related problems. Addict Res Theory. 20(1):82–92. [Google Scholar]

RESOURCES