Cannabis Home Cultivation and Residential Water Use: Evidence from New Mexico

1. Introduction

In April 2021, adult-use, recreational cannabis became legal in New Mexico. The state’s House Bill 2 (The Cannabis Regulation Act) anticipated use by 20% of adults, a rate similar to that of other states with legal adult-use cannabis. Although approved by the legislature and signed by the governor in March and April 2021, dispensaries did not open until April 2022. Home cultivation, however, became legal on June 29, 2021. With legal plant counts of up to six plants per person or 12 plants per household, a boom in home cultivation was expected, beginning in 2021.

Home gardening rates in the broader population and medical cannabis personal production licensing rates in New Mexico support that many New Mexicans likely cultivate cannabis at home or may do so in the future. The National Gardening Association estimates that 35% of U.S. households grow food,¹ i.e., many New Mexicans possess the necessary gardening skills for basic home cultivation of cannabis and these skills are easy to acquire. Among medical cannabis patients in New Mexico, all certified as suffering from severe, debilitating diseases, rates of personal production licensing ranged from 0.37 licenses per medical patient in the fourth quarter of 2012, the first year for which data are available, to 0.064 licenses in April 2021.² Recreational cannabis consumers are likely to be physically healthier than medical cannabis patients and do not have to complete a licensing process, meaning home cultivation by recreational users may exceed the rates seen for personal production licenses through the medical program. The decline in rates of personal production licensing among medical cannabis patients correlates with the availability of medical cannabis dispensaries, so home cultivation rates seem likely to be highest prior to dispensary entry in April 2022 with one caveat −29 June is a late start to the growing season and may have been too late for some outdoor cultivators to grow in 2021 as cannabis plants take 3–6 months to mature. For comparison, California’s growing season ranges from as early as April (Dillis et al. 2019) to October (Madhusoodanan 2019).

Cultivating cannabis requires water and electricity, with water being the dominant resource in outdoor growing and electricity being the primary resource for indoor growing (Zheng et al. 2021). Little research exists on how much water is used by commercial or residential growers. As summarized by a 2017 report on water and energy usage associated with cannabis cultivation in Colorado’s Pueblo County, “Searching for credible information on water and energy use in refereed journal articles and from other sources resulted in a significant and frustrating waste of time.”³ The report goes on to report a widespread but largely unsubstantiated estimate of 22 liters or around 6 gallons per day per plant along with their own estimate of half a gallon per day, based on interviews with six commercial growers. A 2019 article in Nature reported commercial outdoor growing uses six gallons of water per day but only for about three months of the five-month California growing season from June through October and water conservation methods can reduce commercial per-plant water use to only about half a gallon per plant per day (Madhusoodanan 2019). This is the first study to assess water use following HCL or for home cultivation more generally rather than for commercial cultivation. Because the associations between home cultivation and water use are measured at the respondent and household levels, they offer particularly policy-relevant results by accounting for actual household water use rather than offering per-plant estimates.

To evaluate whether or not HCL was associated with an increase in residential water consumption, this study uses surveys on home cultivation preferences and data on residential water use in Santa Fe, New Mexico. Survey responses suggest home-cultivated cannabis prevalence is increasing, and growing is most commonly indoors and using public utility water. Dispensary access and pricing affect the attractiveness of home cultivation, suggesting home cultivation and commercial cultivation are inextricably linked. Comparing residential water consumption prior to and following HCL using residential water utility data from Santa Fe, New Mexico, this study found evidence of a modest increase in water use associated with HCL.

Although this study contributes to the literature by using primary data to analyze an unanswered and relevant research question, it cannot provide a definitive answer to the question of what the effect of HCL on residential water use is for several reasons. Most importantly, the study lacks a control group rendering standard difference-in-differences analyses impossible. Furthermore, HCL occurred in the summer of 2021, the second summer of COVID-19, which the data show was associated with a significant increase in water use in summer 2020. Due to overlap in timing, this study cannot fully disentangle the effects of COVID from those of HCL. In addition, our data from Santa Fe do not include socioeconomic and demographic information for households or location information more granular than Census Tract, and such factors were correlated with water use in the survey data. The data also do not distinguish between rented and owned residences, which could affect water use and the permissibility of home cultivating. Finally, results from Santa Fe may not generalize to other locations with legal home cultivation. The analyses do control for pre- versus post-March 2020, when COVID-19 and associated policy responses were first experienced in New Mexico; we use household-level fixed effects to account for time-invariant differences across households and assess the role of Census-Tract-level differences in population density and household income, but in the end, this study could not establish a causal relationship between HCL and water use due to data limitations and the timing of COVID-19.

This work helps inform efforts by policymakers to create sustainable adult-use cannabis markets, while protecting limited water resources. In addition to its water policy relevance, this work contributes to a nascent academic literature on cannabis home cultivation. In particular, the extent to which home-cultivated cannabis substitutes for dispensary-sourced cannabis has important regulatory, public health and tax implications for states. Home-cultivated cannabis is not subject to health and safety regulations and does not generate tax revenue as currently regulated in state-legal cannabis markets in the United States (U.S.).⁴

The paper proceeds as follows. Section 2 describes the home cultivation survey data collection, analyses, results and several associated hypotheses that directly relate to the water utility data. Section 3 describes the water utility data, analyses and results. Section 4 discusses the results in the context of the literature, and Sec. 5 concludes with policy implications and suggestions for future research.

2. Home Cultivation Survey

2.1. Home cultivation survey data and methods

Given the dearth of information regarding home cultivation, we fielded a survey of New Mexicans to attempt to better understand who home cultivates, when they home cultivate, how home cultivators use water, the extent to which home cultivation substitutes for dispensary-sourced cannabis and the likely future trends in home cultivation in New Mexico as the legal cannabis market matures.

The Home Cultivation Survey occurred in two phases. The first phase involved an Opinio survey fielded through the Kurple Magazine Facebook page (a news publication focused on medical cannabis and headquartered in Albuquerque⁵), while the second phase purchased survey data collected from a compensated, broader sample by Qualtrics. The University of X Institutional Review Board approved the survey designs.

The pilot home cultivation survey was fielded in Spring 2022 (Stith and Chermak 2022). The sample included 27 participants, who responded to 38 questions on home cultivation, ranging from experience to growing methods and cost- and quality-based preferences for dispensary-sourced versus home-cultivated cannabis. These data were not included in this study’s analyses.

The Qualtrics-based survey was fielded in January and February 2023 and provided the data used in this study. The survey sample included 532 participants from throughout the state of New Mexico, who responded to 36 questions on home cultivation. The survey questions were updated versions of the pilot survey questions. The survey was administered online through Qualtrics, which recruits from “website intercepts, member referrals, targeted email lists, gaming sites, customer loyalty web portals, permission-based networks and social media.” Thus, the survey participants were compensated but the amount they were compensated depended on the app through which they accessed the survey. The target population was based on U.S. Census 2020 population estimates and research study requirements, but due to voluntary opt-in procedures, ultimately constitutes a convenience sample of New Mexicans. Despite this limitation, the survey responses provide the first information in the literature on water use associated with home cultivation. While potentially not representative of the broader New Mexican population, the responses likely are representative of the water use decisions made by home cultivators and individuals interested in home cultivation, providing important information for policymakers with respect to water use by home cultivators. The only exclusion criterion was being under the age of 21 years. The target sample was restricted to include a maximum of 40% of responses from Bernalillo County, the most populous county in New Mexico,⁶ to ensure coverage outside this county. In order to better inform the residential water utility data analysis, we also oversampled Santa Fe. Age, gender and ethnicity quotas were based on the U.S. Census 2021 American Community Survey estimates for the NM population 21 and older and included the following targets: Ages 21–34 - 25%, ages 35–54 - 33%, 55+ - 43%; male - 49%, female - 51%; Hispanic or Latino - 46% and Not Hispanic or Latino - 54%.

Table 1 shows descriptive statistics for the sample in Panel A. The first column reports the proportion for the full sample, the second column is restricted to cannabis consumers and the third column is restricted to current home cultivators. A total of 46% of the sample was from Albuquerque, 16% from Santa Fe and 38% from other locations within the state with 15% reporting living in a rural location rather than a city, small town, or suburb. Respondents were asked their age with 38% reporting being between 21 and 39, 42% reporting being between 40 and 64 and 20% reporting being 65 or older. The majority of the respondents were female (64%), White (70%) and Non-Hispanic (61%). A total of 51% reported an income of less than 40,000 USD per year with 18% reporting incomes of 100,000 USD or more. Cannabis consumers were younger, less White and more Hispanic than the full sample. Home cultivators are younger, less likely to live in a rural area or be White and more likely to be male, Hispanic and middle-income than both the full sample and cannabis users more generally.

Panel B summarizes our outcome variables related to cannabis use, home cultivation (current, past and future), growing methods and tradeoffs between home cultivating and sourcing cannabis from a dispensary. At the time of our survey, 63% of our sample consumed cannabis, 19% home cultivated, 12% had home-cultivated previously and 40% intended to home-cultivate in the future. Among home cultivators, retention was high with 90% planning on home cultivating again in the future. Experience home cultivating does not appear to have been entirely constrained by legality. Home cultivation even among medical patients and caregivers has only been allowed since 2007, however, several respondents report cultivating cannabis prior to 2007. The majority of those reporting home cultivating grew indoors, cultivated during the summer months of June, July and/or August and used public water. A total of 79% of home cultivators reported increasing their water use and that water use increased for an average of 3.5 months. About the same number of home cultivators reported increasing home gardening of non-cannabis plants (22%) as reported decreasing home gardening of non-cannabis plants (21%). Rates of indoor growing and public water use were lower among future home cultivators, but these methods still constituted the majority of the responses. A total of 20% of the sample reported intending to apply for a commercial cultivation license in the future. Cannabis use and current home cultivation are clearly correlated with future home cultivation, and about 50% of home cultivators expect to apply for a commercial grow license in the future.

Panel C of Table 2 shows how respondents felt dispensary-sourced cannabis compared with home-cultivated cannabis. A total of 15% felt home-cultivated cannabis is superior in quality while 30% reported that dispensary-sourced cannabis is higher quality. A total of 20% reported access to dispensaries influencing their likelihood of home cultivation and 43% reported dispensary prices as affecting their desire to home cultivate. Not surprisingly, home cultivators had a more favorable view of home-cultivated versus dispensary-sourced cannabis, however, they are also the most likely (at 80%) to report that their propensity to home cultivate is affected by dispensary prices.

A simple linear probability model was used to analyze associations between respondent characteristics and our outcome variables.

In the equation above, $Y_i$ represents our {0,1} outcome variables: Cannabis consumption, home cultivation (current, past and future), growing methods (indoors, summer, public water use), commercial cultivation expectations and comparisons between dispensary-sourced and home-cultivated cannabis. Our independent variables, represented by $X_i$, include age categories, race, ethnicity, family income and rural. For our regressions regarding cannabis use and future home cultivation or commercial cultivation, we use the full sample. In our regressions analyzing current home cultivation and growing methods, we include only home cultivators. The regressions analyzing tradeoffs between dispensary-sourced and home-cultivated cannabis are restricted to cannabis consumers. We use standard errors robust to heteroskedasticity throughout. In terms of the power of our analyses, i.e., our ability to identify an effect, we are limited by our sample size to only identifying statistically significant coefficients larger than 0.0608 for the full sample, 0.0765 for the cannabis user sample and 0.14 for the sample of home cultivators (power = 80, alpha = 0.05).

2.2. Home cultivation survey results

Tables 2–4 explore the associations in our home cultivation survey sample. As shown in Table 2, cannabis use declines with age, is higher among Hispanics and is lower among those from households earning 100,000 USD or more per year. Current home cultivation is less common among the oldest cohort and females and more common among Hispanics. Past home cultivation shows a similar pattern, but higher-income individuals are marginally statistically significantly more likely to have previously home cultivated than lower-income cohorts. Future home cultivation is also less common among the oldest cohort and more common among Hispanics, but there are no gender distinctions and Whites are statistically significantly less likely to intend to home cultivate than other races. Rural has no effect on cannabis consumption or home cultivation. Higher rates of home cultivation among males are consistent with surveys of home cultivation practices prior to recreational legalization (Azofeifa et al. 2021) and in Canada prior to legalization of extracts, when only flower was available to recreational users (Cristiano et al. 2022). Perhaps reflecting differences between the U.S. and Canada, this study and Wadsworth et al. (2022b), which uses U.S. data, found differences by age, gender and ethnicity, while Wadsworth et al. (2022a), which uses Canadian data, did not find differences by age, gender, or ethnicity.

Table 3 explores the current home cultivation practices in the first three columns, restricting the sample to only current home cultivators. Partly due to our small sample size, we find few associations. However, it appears that the oldest cohort is least likely to grow indoors, middle-income home cultivators are least likely to grow in the summer and the middle age cohort is least likely to use public utility water. The under-40 age group increased their water use the most, while Hispanics were less likely to increase their water use than non-Hispanics and high-income and rural respondents increased their water use less than other income groups. Whites were more likely than other races to grow both fewer cannabis plants and more non-cannabis plants. The middle-aged group was least likely to grow more non-cannabis plants in response to home cultivation. The R-squared suggests that demographic variables explain between 5.2% and 14.5% of the variation in home cultivation practices among current home cultivators.

Table 4 explores future home cultivation expectations using the full sample and finds that older cohorts are less likely to intend to grow indoors, use public water, or grow commercially. The youngest cohort is the most likely to anticipate applying for a commercial license. Whites are less likely to expect to grow indoors or to apply for a commercial license. Hispanics are more likely to anticipate applying for a commercial license than non-Hispanics. Not surprisingly, current home cultivation greatly increases the likelihood of anticipating future home cultivation. The R-squared suggests that demographic variables combined with current home cultivation explain between 14.5% and 20.5% of the variation in the future home cultivation variables.

Table 5 shows the results comparing dispensary-sourced and home-cultivated cannabis. Older adults and Hispanics were less likely to report home-cultivated cannabis as superior to dispensary cannabis, while home cultivators generally were more likely to believe their cannabis was superior to that sold in dispensaries. None of the coefficients is statistically significant for the dispensary access effect on home cultivation. The R-squared in both the first two columns is quite small as well, suggesting that the independent variables do not explain much of the variation in the outcome variables. The model for the effect of dispensary prices on the likelihood of home cultivation performs better with an R-squared of 10.3 and clearer associations. Those ages 40–64, women and middle-income were less likely to have their home cultivation decision affected by dispensary prices. Whites and current home cultivators were more likely to report effects from dispensary prices than non-Whites and non-home cultivators.

The results from the survey data show that home cultivation definitely occurs in New Mexico, it is increasing now that home cultivation has been legalized and that trend is likely to continue in the future. Home cultivators increase their water use relative to when they are not home cultivating and do so for an average of 3.5 months. Most growing occurs outdoors, in the summer and using public utility water. All these factors predict an increase in residential public utility water use, particularly during the summer months. Continued growing by those previously growing illegally will diminish the impact of legalization on water use and could lead to underestimates of the extent of water use related to now legal home cultivation. An interest in shifting towards indoor growing and substitution between home-cultivated and dispensary-sourced cannabis also may diminish the extent of the increase in water use following legalization, as well as reduce its persistence. Extensive heterogeneity exists in homes that are cultivated as captured by the demographic variables including age, gender, ethnicity and household income.

Two sets of hypotheses emerge from the survey data, which can be explored using the water utility data. First, HCL will be associated with an increase in water use relative to prior periods, and increases will occur primarily in the summer months. Second, water use will vary substantially across individuals, meaning overall estimates may mask substantially stronger associations between HCL and water use among some subgroups.

3. Santa Fe Residential Water Use and HCL

3.1. Santa Fe residential water use data & methods

This analysis seeks to assess the association between the June 29, 2021, legalization of home cultivation and residential water consumption with the expectation, based on the survey data, that HCL will be associated with increased water consumption.

In order to measure changes in residential water consumption, we obtained data on monthly household-level water consumption from the Santa Fe Public Utilities Department for the sample period from January 2017 through April 2022. Santa Fe is the third largest metropolitan area in New Mexico and its capital city with a population of about 89,000.⁷ As a city in the desert Southwest, Santa Fe has been successfully implementing water conservation programs for decades with an associated multi-decade decline in per capita water consumption since then. In addition to subsidies for low flow toilets and water-use-reducing landscaping, the City of Santa Fe has adopted policies to restrict summer irrigation (May through October), including higher prices, restricted hours (between 6 pm and 10am) and recommended limits on the number of days per week (3) households can irrigate outdoors.⁸

A key strength of the water utility data is that, by using overall water consumption rather than attempting to estimate per-plant water use as in prior studies, we are able to measure aggregate changes in water use at the household level, the level at which decisions to home cultivate occur and at which policy changes are likely to be implemented. The University of X Institutional Review Board deemed these data exempt.

The original data set included 2,214,106 observations from 35,978 residential water meters. We dropped the 98,132 observations with zero or less water consumption recorded in that month. In line with the literature (e.g., Price et al. 2014), we trimmed the bottom and top 1% of our data to remove outliers that might bias our results, which reduced our sample by another 42,215 observations. After these adjustments to the data, the analysis sample included 2,073,751 observations from 35,351 residential water meters. As shown in Table 6, the average meter in our sample reported 4,573 gallons of water consumed per month. United States Geological Survey data indicate that the average New Mexican used 81 gallons of water per day or about 2,462 gallons per month in 2015, the most recent year for which data are available, making the average in the sample data approximately the average water use for a family of two.⁹ We include all meter sizes in our main sample, even though this includes both single and multi-unit residential water meters, because we are seeking to evaluate the overall association between HCL and residential water use, not just the association for single-household dwellings. Cannabis plants can be grown indoors in a relatively small space, so even high-rise apartments could potentially be affected. In addition, apartment dwellers typically are more financially constrained making home cultivation potentially attractive as a substitute for more expensive dispensary-sourced cannabis. Furthermore, multi-unit meters mean water is not tied to a single unit and water costs are often paid by the landlord, which could increase the propensity to home cultivation. In addition, while apartment buildings may have more stringent rules than typical for single-family housing, higher-income neighborhoods almost universally have restrictive Home Owners’ Association rules that could limit home cultivation. Nonetheless, we run robustness checks on our main analyses constraining our sample to meters of size 5/8 inch, the most common size for single-family homes in Santa Fe.

Another concern with our main sample is that it includes both tenants and owner-occupied dwellings. The data do not allow us to distinguish between rented versus owned homes and the data do not include sufficient identifying information to match with real estate sales data. However, multi-unit housing is more likely to be rented than single-family units, so the smaller meter size results serve as a quasi-robustness check on the effect of renting versus home ownership. Furthermore, suggesting rental housing is not a major deterrent to home cultivation, among cannabis consumers in the survey sample who do not home cultivate, just 7.2% report not home cultivating because they cannot grow at home, with only 5% reporting that their landlord will not allow it. (Other reasons for not being able to grow at home include family and friends not approving and insufficient space.)

Relative to other counties in the U.S., the County of Santa Fe (dominated by the City of Santa Fe) has 5 percentage points more single-family housing structures and a 16 percentage point lower rental rate than the average county in the U.S. (Reagan 2021). While these factors may reduce generalizability, they do support that the results from our analyses are likely to be driven by water use in owner-occupied, single-family dwellings.

We use three measures of water consumption as outcome variables. For our main specification, we use 100s of gallons per day to simplify the interpretation of the results. We run two robustness checks on these outcome variables. First, we calculate the natural log of water consumption to potentially better account for outliers. (Histograms suggest the natural log of water use is distributed more normally than the total consumption in gallons, which absent trimming, includes large outliers.) Second, we calculate whether or not the household crossed the threshold between lower and higher cost per gallon consumed. The base residential water consumption charge for September through April in Santa Fe is $6.06/1,000 for the first 7,000 gallons and $21.72/1,000 for gallons thereafter and from May to August is $6.06/1,000 for the first 10,000 gallons and $21.72/1,000 thereafter.¹⁰

Measuring exposure to HCL is more complicated. At a basic level, our exposure variable measures whether water consumption occurred after June 2021, the month in which HCL occurred in New Mexico. Because all households were affected by HCL at the same time, we cannot construct a control group necessary for a standard difference-in-differences analysis. Instead, we must rely on household fixed effects. Including household fixed effects enables us to estimate essentially a within-household effect by controlling for differences between households, which do not change during our sample period. For example, for most households, the income level and ethnic, age and gender composition are unlikely to have changed materially over the approximately five-year sample period. Unlike in our survey data, we do not have socioeconomic or demographic information on individual households, and therefore, also must rely on household fixed effects to control for these factors. The household fixed effects allow us to effectively compare changes in water use within an individual household, allowing each individual household to serve as its own counterfactual.

Further threatening identification, New Mexico’s HCL occurred while the COVID pandemic was still ongoing, and we cannot ignore the potential influence of COVID-19 and the associated policy responses, which dramatically affected all aspects of life. Pandemic infection rates were associated with increased interest in home gardening (Lin et al. 2021), the stay-at-home orders have been linked to increased residential water use (Irwin et al. 2021)¹¹ and the first year of the pandemic showed a large surge in summer and fall residential water consumption in Denver, Colorado (Eastman et al. 2022). Reports from Santa Fe’s water authority describe a major decrease in commercial water use with a corresponding increase in residential water use during the business and school lockdowns of 2020.¹²

To disentangle the effects of HCL versus COVID, we also measure the effect of COVID-19 on water consumption. We measure the influence of COVID in two ways. In our simple pre/post analyses, we use a {0,1} dummy variable to control for whether water consumption occurred before or after March 2020, which coincides with the onset of COVID and the first stay-at-home orders.¹³ For our more detailed month-level analyses, we track water use by month beginning in January 2020, in addition to any changes that occurred more proximate to HCL in June 2021. Our sample includes 36 months pre-January 2020, 28 after January 2020 and 9 after June 2021.

The analysis also requires controls for the extensive and successful water conservation efforts by the City of Santa Fe; water use has been trending downward for decades. For ease of interpretation, we include a yearly trend in our main specification. As robustness checks on the year trend, we also conduct a more conservative analysis allowing for nonlinearities in the time trend by including year-fixed effects, and a specification that avoids the shorter-run COVID effects by comparing the post-HCL period from July 2021 through April 2022 with data from the pre-COVID period (January 2017 through February 2020), omitting data from March 2020 through June 2021 and both the year trend and year fixed effects.

Because seasonal changes, weather and elevation affect water use, we include 12 month-level fixed effects and control for total monthly precipitation and average monthly high temperatures throughout. The latter two variables are averaged across all reporting weather stations at the city-level, so we further adjust them by “zone”, which roughly translates into elevation and allows the effects of these variables to vary with elevation. Santa Fe elevation averages 7,198 ft above sea level but is as low as around 6,348 ft at the airport. The city is split into 11 different pressure zones ranging from the northeast of the city to the southwest, with higher numbered zones generally corresponding to higher elevation areas. Zone is not available for a subset of households and including it in our regressions reduces our sample size by 2,758 observations (51 households) to 35,300 households and 2,070,993 observations. During our sample period, total monthly precipitation and monthly high temperatures averaged 1.07 inches and 61 degrees, respectively.

Because the survey data identified substantial heterogeneity in preferences for home cultivation, we run specifications assessing whether population density or per capita income affects our results. These variables also capture differences in available resources that could affect growing, such as differences in household socioeconomic characteristics. We obtained these data for each of Santa Fe’s 35 Census Tracts from the 2019 American Community Survey administered by the U.S. Census Bureau.¹⁴ Population density ranges from four persons per square mile to 6,541, while average annual per capita income ranges from $18,309 to $95,198. Table 6 shows descriptive statistics for the variables.

We use two estimation methods. Our first use Least Squares regression techniques and a simple pre/post-HCL measure to estimate the within-household association between HCL and water consumption, controlling for the effect of COVID-19; household-level, time-invariant differences; month-level variation in water use across all households; a trend in water consumption common to all households; and total precipitation and average high temperature, both adjusted for elevation. The estimating equation for our first specification is as follows:

Our outcome variables are measured for household h at time t, where t refers to the month m and year y in which the water use occurred$.$ Our HCL and COVID variables are measured at the month-year level. We further control for the weather variables, adjusted by elevation (Zone), obtaining effect estimates for total precipitation and average daily maximum temperature, as well as for the relative effects of the weather variables by elevation. As the Zone variable does not vary at the household level and we control for time-invariant household characteristics, the main effect of the Zone variable is perfectly collinear with the household fixed effects and drops out of the equation, leaving only the relative effects of the weather variables by zone, with Zone = 1 the omitted zone. The remaining variables capture a constant term $\alpha$, the year trend $\tau (y)$, month fixed effects ${\theta }_m$, household fixed effects ${\omega }_h$ and the error term ${\epsilon }_{ht}$. Since observations within households may be arbitrarily correlated, we cluster our standard errors at the household level to avoid underestimating our standard errors and overestimating the statistical significance of our results. Clustering our standard errors also adjusts for heteroskedasticity, i.e., that the precision of the estimates varies systematically with the values of the independent variables. For our regressions allowing for heterogeneity by per capita income and population density, we interact these variables with the HCL variable.

For our second estimation method, we use the following event study specification, in line with work on the COVID-19 pandemic by Bacher-Hicks et al. (2021) and Goda et al. (2022).

Our outcome variables do not change, but we substitute a series of month-year-level pre- and post-HCL variables for the HCL and COVID indicator variables, tracing out the entire period from January 2020 through April 2022 with June 2021 as the omitted or baseline period relative to which the other periods’ water use is estimated. In other words, in January 2020, $t=-17$; in June 2021, $t=0$; and in April 2020, $t=10$. Observations occurring prior to January 2020 are included in a dummy variable Pre2020, capturing that the water use occurred pre-2020. As in our difference-in-differences approach, we adjust the estimates for seasonal differences using month fixed effects, for annual differences in years 2017–2019 using year fixed effects, for precipitation and temperature differences (adjusted by elevation) not captured by the month fixed effects and for time-invariant household characteristics using household fixed effects. We only include year fixed effects for 2017 through 2019 and drop the year-fixed effects for 2020–2022 as the effects of being in all months between 2020–2022 are traced out using the period-level variables.

The ${\beta }_t$’s measure how much within-household water use differs from predicted water use in each period t relative to use in June 2021, controlling for pre-existing water use, normal monthly variation in water use, COVID-19 effects and elevation-adjusted measures of temperature and precipitation. With ideal data, but for HCL, predicted water use in each period pre- and post-HCL should be equal to use in June 2021 after adjustment, i.e., the coefficients $({\beta }_t$’s) measuring the difference should be statistically indistinguishable from zero. Differences in predicted values in the pre-period could arise from anticipatory effects or be driven by other events entirely, e.g., COVID-19. Abnormal water use in the post-period captures the immediate and lagged effects of HCL but may also be influenced by other events unrelated to the HCL.

3.2. Santa Fe residential water use data results

We begin our analysis of the effect of HCL by graphing the raw data by month for the pre-COVID (January 2017–January 2020), COVID (February 2020–April 2022) and HCL (July 2021–April 2022) periods, as shown in Figure 1.

Throughout, seasonality is evident with higher water consumption in the summer and fall. During the winter months, COVID water use is lower than pre-COVID, but during the summer months, COVID water use is distinctly higher. Water use after HCL tracks similar to prior periods during the winter months but generally lies between pre-COVID and post-COVID water use during the summer months.

Table 7 shows the results from the regressions for our three outcome variables.

The estimates in Columns 1 and 2 can be directly interpreted as changes in 100’s of gallons of water consumed. In Column 1, the association between HCL and water use is measured without controlling for COVID. A negative coefficient corresponding to a reduction of 95.7 gallons is evident. However, the magnitude of the coefficient is too large due to the positive impact of COVID on water use in 2020. After controlling for the large surge in water use during the summer 2020, we estimate an HCL coefficient of −0.849, which indicates a reduction of 85 gallons per month in association with HCL, while the COVID coefficient indicates an increase of 189 gallons. In other words, water use was declining prior to COVID, increased with COVID in 2020 and then partially reverted to pre-COVID levels with HCL in Summer 2021. The coefficients together suggest an increase of 105 gallons per month in the HCL period relative to the pre-COVID period, but we can reject that the joint effect equals zero ($p<0.001$) at standard levels of statistical significance ($p\le 0.05$). This analysis suggests an aggregate increase of approximately 105 gallons per month times 35,300 households or 3,706,500 gallons per month in association with HCL. In Column 3, the HCL coefficient of −0.015 can be interpreted as showing that legalization of home cultivation was associated with a reduction in average monthly water consumption of 1.5% or 115 gallons per month.¹⁵ This counteracts an increase of 4.6% or 215 gallons during COVID. The combined coefficients for HCL and COVID again are statistically significantly different from zero ($p<0.001$). Although the outcome variable may better account for outliers in the natural log specification in Column 3, the model’s explanatory power (R-squared) is lower than for the model using hundreds of gallons in Column 1. The fourth column’s results show that HCL is associated with a 0.8 percentage point reduction in the likelihood of crossing into higher price-per-gallon consumption, with approximately 12% of households crossing the price threshold each month. COVID was associated with a 1.7 percentage point increase in the likelihood of crossing into higher price-per-gallon consumption. Jointly, the probability of crossing into higher price-per-gallon consumption was lower following HCL relative to COVID, but remained statistically significantly elevated relative to pre-pandemic levels ($p<0.001$). Combining the information in Columns 2 and 4, one can derive a rough estimate of a $29,891 increase in monthly payments from households’ post-HCL relative to pre-COVID.¹⁶ Dividing by the number of households yields an average per household increase in monthly payments of $0.85. In Table 8, we interact the HCL variable with Population Density in thousand persons per square mile and with per capita income in US$10,000 in order to evaluate whether the small, marginally statistically significant aggregate estimate is masking underlying variation in the effects of population density and per capita income. We adjust the original population density and per capita income variables by 1,000 and 10,000, respectively, to improve the interpretation of coefficients given small effect sizes.

In Table 8, we explore the heterogeneity in the association between HCL and water use by population density and per capita income. The main coefficients for HCL and COVID can be interpreted as baseline levels to which the coefficients for the interactions should be added. Despite the prior that cannabis cultivation requires sufficient space, census tracts with denser populations experienced more water use post-HCL, even after controlling for weather differences. Per capita income was also associated with greater water use post-HCL. It may be that these variables are proxying for other factors affecting the decision to home cultivate, such as race and education, which are highly correlated with income. Adding the HCL and COVID coefficients without accounting for population density and income yields baseline water use that is 39.9 gallons lower post-HCL. Accounting for the point estimates for population density and income in Column 1 above, for HCL to offset the negative baseline result and increase water use, population density must be more than 10,627 persons per square mile, which is not true of any Census Tract in Santa Fe, or income must be more than $86,656, which is true for only one of the 35 Census Tracts in Santa Fe. A combination of high enough population density and income would also yield positive net effects. Figure 2 below shows estimates of the effect of HCL on water use by Census Tract, where we base our estimates on Census Tract-level population density and per capita income. Clearly, substantial heterogeneity exists across Census Tracts as shown in Figure 2. All Census Tracts experience an increase in water use post-HCL. The magnitudes, however, vary substantially across counties from an increase of only 9 gallons near the airport (Census Tract 1304, average income $19,338) to an increase of 167 gallons in the large rural area on the northwest side of the city, which has the highest average per capita income at $95,198 (Census Tract 10204).

We run three robustness checks on our main specification, including year-fixed effects rather than a trend, omitting the COVID period from March 2020 through June 2021, and restricting our sample to 5/$8^{\prime \prime }$ meters only. All three specifications show similar patterns in the results to the main specification. In Table A.1, the results including year-fixed effects rather than a year-level trend are consistent with our main results, but the magnitudes are smaller. The HCL and COVID coefficients together suggest an increase of 36 gallons per month of water use. Table A.2 shows results omitting the COVID period. The coefficients for the association between HCL and water use measured in 100 gallons and the natural log of water consumption are negative, indicating a decrease associated with HCL relative to periods prior to COVID. However, heterogeneity across households exists, as the coefficients suggest a statistically significant switch to higher-priced water use following HCL despite the lack of an increase in average water use. Table A.3 shows results restricting the sample to meters sized 5/8 inches to better capture the effects on single-family dwellings. Both the coefficients for COVID and HCL are larger than in the main sample, suggesting that single-family dwellings have more volatile water use, i.e., water use that is more responsive to policy changes. The results are slightly larger in magnitude with an overall increase of 139 gallons in association with HCL versus 105 gallons in the main analyses.

To further tease out the relationship between HCL, COVID and residential water consumption, we use our second specification to generate the event studies in Figures 3–5. The reported outcomes are the average association between being in that period relative to June 2021, the last month before HCL, adjusted for month-, year- (for 2017–2019) and household-level characteristics, precipitation and high temperature.

All three figures show much higher levels of water consumption in summer 2020 than any time before or thereafter. This coincides with the period in which COVID lockdowns were most prevalent. Higher water use occurs in April 2021, likely reflecting efforts to preemptively water in advance of higher pricing and restricted irrigation starting in May. In July 2021, the first month in which cannabis home cultivation for adult use was legal, water use was abnormally low. The periods between August–October 2021 show water use that exceeds water use in June 2021, but is lower than during the first summer of COVID. (Table A.4 reports the estimated coefficients underlying these regressions.) These results are in line with the overall results, that water use increased post-HCL relative to pre-COVID, but was lower than during COVID and show that the increase in water use was driven by water use in September–November 2021 as well as possibly April 2022. (The April effect is evident in all three years depicted, indicating a disproportionate impact from irrigation restrictions when water demand is abnormally high.) September generally marks the end of the harvest season, suggesting that the increase in October and November may be driven by indoor grows begun at the end of June 2021 approaching maturity approximately 3–5 months after planting. A 3–5 month growing season falls within estimates of 3–8 months from Leafly, a popular cannabis information aggregator.¹⁷

Despite extensive efforts to disentangle the relationship between HCL and water use, we interpret all our coefficients from the regression analyses as correlational rather than causal, due to the lack of a suitable control group necessary for standard difference-in-differences analyses and the overlap in timing between COVID-19 and HCL.

4. Discussion

Responses from the survey on home cultivation indicate that home cultivators increase water use, typically do so for at least a few months and most anticipate continuing to home cultivate in the future. The survey results also indicate increasing interest in home cultivation over time, with many future home cultivators intending to use public utility water, suggesting that the current water utility results may be a lower bound on the association between HCL and residential water use. However, future home cultivators also report being more likely to grow indoors than current home cultivators, potentially decreasing overall water use. The commercial cannabis market will be an important determinant of how residential water use for home cultivation evolves. Many cannabis consumers and even home cultivators report that home-cultivated and dispensary-sourced cannabis are substitutes and that their propensity to home-cultivate is determined in part by dispensary proximity and pricing. The relative use of water in home cultivation versus commercial cultivation is unknown and depends on whether a lack of expertise or a profit motive leads to over- or under-watering.

As predicted by the survey responses, the Santa Fe water utility data indicate an increase in residential water use in association with HCL relative to water use pre-COVID. Post-HCL, households in Santa Fe are spending about $2.92 more per month on average for an additional 105 gallons (3.71 million gallons total) and Santa Fe Water Utility can expect $29,891 more per month. If households grow twelve plants and each plant uses between 0.5 gallons and 6 gallons per day, then a change of 3.71 million gallons would coincide with a home cultivation prevalence of between 58% and 5% of the adult population of Santa Fe, overlapping estimates from the literature, which range from 1.6% among cannabis consumers aged 12+ (Azofeifa et al. 2021) to 8.8% among adults in states with legalized adult-use cannabis and home cultivation (Wadsworth et al. 2022b). If 8.8% of the Santa Fe households in our data are home-cultivating, the estimates in this paper would imply 3.3 gallons are used per plant per day.

The Home Cultivation Survey data are limited by the small number of home cultivators in the sample and the opt-in nature of the survey design, but offer suggestive evidence of potential trajectories in cannabis home cultivation in the future. In particular, the main reported barrier to home cultivation, lack of know-how, can easily be addressed; more respondents intend to grow cannabis in the future than grow now; and home cultivated cannabis is a cheaper, potentially higher quality product as compared to dispensary-sourced cannabis. Expected growing methods point towards lower water-use methods, i.e., indoor growing, but the majority of water used for home cultivation is likely to come from public water utilities. Home-cultivated cannabis and dispensary-sourced cannabis are likely to continue to be substitutes with the former acting as a quality and price check on the commercial market. A majority of all respondents believe dispensary-sourced cannabis is no better or even worse than home-cultivated cannabis and a majority of cannabis consumers report that dispensary prices affect their likelihood of home cultivating. Our survey results further indicate that areas with younger populations, more males and more Hispanics are likely to see the most home cultivation, with younger and Hispanic respondents most interested in transitioning into commercial cultivation in the future. These results suggest that home cultivation may serve as a sort of training program for future commercial growers. Given widespread views that dispensary-sourced cannabis is not particularly high quality, increased growing expertise and associated higher quality products in the industry might be a positive spillover from HCL.

Although a range of policy-relevant implications arise from our water utility data results as well, we cannot rule out the influence of other factors, especially COVID, on the precision and generalizability of the water utility data results. Significant variation in water use exists across households, with average estimates not particularly representative of stronger correlations between HCL and water use seen in, for example, wealthier Census Tracts. Complicating the identification of causal effects, COVID caused an enormous perturbation in water use patterns in 2020 and it is possible that a spillover or continued COVID effect could be influencing the results in 2021 rather than HCL alone. Reducing the risk of contamination from spillover COVID effects, the COVID context changed substantially between the summers of 2020 and 2021. Although case counts and hospitalizations did not differ significantly across summers in 2020 and 2021, during the first summer stay-at-home orders were prevalent, while by summer 2021, most of those who wanted to be vaccinated were vaccinated and businesses were open. In its 2020 report, Santa Fe’s water utility reported that COVID primarily increased water use due to business lockdowns in Summer 2020, which temporarily shifted water use during the workday from commercial buildings to residences. This effect was no longer evident in the Summer of 2021 as per the city’s 2021 water utility report.¹⁸ Labor market analyses (e.g., Goda et al. 2022) show a general trend towards a resumption of normal labor market outcomes, suggesting that after the initial stay-at-home orders expired, temporary variations in individual policies and COVID-related cases, hospitalizations and deaths may have done little to perturb a general trend back to normalcy over the course of the pandemic. Data from the New Mexico Department of Health show peak cases in November 2020 and January 2022, peak deaths in December 2020, December 2021 and February 2022, which, apart from November 2020, are not months with particularly high-water use.¹⁹ Together these factors suggest that it is reasonable to conclude that most COVID effects on watering had resolved by Summer 2021.

Another major limitation of the water use data is the lack of household-level information, especially given the variation by demographics seen in the survey data. While the analyses control for household fixed effects to capture time-invariant household characteristics, they do not include household-level information on factors that changed during our sample period such as employment status, education, income, or rental status. Although we were unable to include these variables at the household level, we were able to evaluate Census Tract-level differences in the association between HCL and water use, focusing on population density to approximate lot size and per capita income to capture financial resources and because income is correlated with employment status, education and health. The positive correlation between population density and HCL was unexpected, but most likely explained by the scarcity of fertile land in New Mexico, which means that less populated areas are more likely to be desert rather than farmland as in less arid parts of the U.S. Less space seems unlikely to drive increased water use, so factors such as more private wells in less densely populated areas or increased population density in more fertile growing areas might be driving the relationship. Similarly, the positive coefficient on per capita income could be directly due to greater availability of resources or might arise because per capita income is correlated with other factors, e.g., health and education. Ultimately, Census Tract-level heterogeneity can only capture the general context in which a household lives and not the precise circumstances of an individual household. Nonetheless, the variation in the results suggests that average effects mask substantially underlying heterogeneity in the association between HCL and water use.

Another issue with the lack of individual information is that it limits the generalizability of the results beyond Santa Fe, as underlying household factors specific to Santa Fe could be affecting the results. Santa Fe has a disproportionate number of retirees and is a popular location for vacation homes (Goodnight 2021), both of which could affect water use and the propensity to home cultivation. While most desert cities in the U.S. pursue significant conservation programs, like Santa Fe, this would not be true of most other cities in the U.S. Comparing the 14 home cultivators from Santa Fe with the full sample of New Mexican home cultivators in the data, they do appear to skew younger, less female, more Native American and higher income. The Santa Fe home cultivators had more prior growing experience and were more likely to grow indoors and during the summer, but were less likely to use public utility water. These home cultivators were the most likely to respond that home-cultivated cannabis is of higher quality than dispensary-sourced cannabis, while simultaneously showing the greatest price effects from dispensaries. If anything, it appears that the Santa Fe home cultivators may be fairly sophisticated growers and their behavior could be predictive of effects we might see going forward in the broader sample, e.g., a shift toward indoor growing with potential associated reductions in water use.

Despite these data limitations, this study offers a substantial contribution to the literature by expanding knowledge of the association between HCL on residential water use through new survey data on growing practices and the use of residential water consumption data at the household level. No other studies of home cultivation have used water utility data, although several other studies have used survey methods to attempt to understand home cultivation. However, these survey-based studies primarily explore demographic differences in preferences for home cultivation and do not address growing methods or water use. Work on commercial cultivation is unlikely to be especially informative because home cultivators face different restrictions with respect to production, e.g., they are only permitted to cultivate a small number of plants but have fewer restrictions on water sources. For example, while both residential and commercial cultivators in New Mexico can use public utility water, only residential growers can use domestic, non-commercial wells. Most commercial grows in Northern California use commercial wells (Dillis et al. 2019). Thus, by analyzing a relatively large sample of New Mexicans post-HC legalization and using residential water utility data, this study offers an important contribution to the academic literature in terms of understanding home cultivation, growing practices and associated water use.

5. Conclusion

The data analyzed suggest that home cultivation occurs in New Mexico and it is likely to increase in the future, even if dispensary access is widespread. Home cultivation could lead to an increase in water use, but overall changes in water use through April 2022 were relatively small at around 2.2% of average monthly use and may not be solely driven by HCL given the outsize impact of COVID in Summer 2020. With respect to the future of water use in New Mexico, dispensary entry after the water utility sample period may have reduced home cultivation-related water use, as many survey respondents report that home-cultivated and dispensary-sourced cannabis are substitutes, but expectations about increased future growing suggest that dispensaries will not fully crowd out home cultivation. The interconnection between markets for home and commercially cultivated cannabis means that the overall effect of cannabis legalization on water resources used by all cannabis cultivators in New Mexico remains unknown. Furthermore, as this study serves as the only estimate of the association between HCL and residential water use and even studies of commercial water use are few, this study offers insights but not conclusions as to the final effects of cannabis legalization on water. The popularity of indoor growing methods likely mitigates the potential impact of home cultivation, but the use of public utility water has important policy implications for public water use, with the needs of home cultivators only one of many demands on local water supplies. In fact, the popularity of indoor growing may be partly due to seasonal water restrictions, as well as the risk of additional restrictions on outdoor water use in response to drought conditions as can happen throughout New Mexico, especially during the summer months.

While policymakers might prefer that cannabis be sourced through regulated dispensaries for public health and tax revenue reasons, dispensaries seem unlikely to drive out home cultivation for the foreseeable future and should provide an important safety, price and quality check on the commercial market. Dovetailing with the findings from the survey data in this study, perceptions in the popular press in April 2022 were that prices in New Mexico continue to be relatively high as compared to Colorado and quality is comparatively lower (Hooper 2022; Porter 2022), i.e., dispensaries are not seen as offering a particularly high-value product, even compared to dispensaries in competing markets, despite regulations. More recent price data are not available from the New Mexico Cannabis Control Division, only aggregate sales by city,²⁰ but consumer-sourced online data from Price of Weed shows high-quality cannabis retails at $283 per ounce in New Mexico, lower than most states, including Arizona ($297) but higher the $241 charged per ounce in Colorado. In general, more mature markets appear to display lower prices. Both home cultivation and proximity to Colorado are likely to continue to encourage New Mexican dispensaries to increase the quality of their products and decrease their prices, ultimately improving the cannabis consumer experience and the economic efficiency of the market. These effects need not be negative for state revenues — enough new dispensary customers will outweigh any negative tax effects from lower prices.

Although this study’s results offer suggestive evidence of potential impacts on residential water use from HCL, the survey data were self-reported and the analyses of the residential water utility data faced too many data limitations to allow for causal interpretation of the results. A future study using a longer sample period and an unaffected comparison of city and data with household-level identifiers could substantially improve up this study. Such a study would be able to better assess the long-run effects of COVID-19 and HCL, use recent developments in causal inference to identify causal effects rather than associations and assess the impact of dispensary access on water use by matching household addresses with dispensary location data. A possibly ideal location would be Kansas City, which straddles the border between the state of Missouri, which legalized home cultivation in 2022 and the state of Kansas, which does not allow for even medical use.

The results of this study, although not conclusive, still suggest a variety of policy recommendations. Relevant to water policy, these recommendations include educating and incentivizing growers to use low-water growing methods, with indoor growing a clear option for conserving water, developing training programs to facilitate transitioning from home cultivation to licensed micro-production with instruction on optimal water use procedures and continuing to monitor the evolution of local cannabis markets with respect to the interplay between unregulated home-cultivated cannabis and dispensary-sourced cannabis and the relative water use of each approach to cannabis cultivation. Policy recommendations related to ensuring quality and the substitution between home-cultivated and dispensary-sourced cannabis include expanding testing protocols that incorporate a broader range of contaminants — while New Mexico tests for 16, California tests for 66 and Oregon for 59 (Atapattu and Johnson, 2020); educating consumers on cannabis quality dimensions beyond THC; noting that widespread medical use underscores the importance of ensuring consistently high-quality, safe cannabis is sold by state-legal dispensaries; and requiring that regulatory authorities collect detailed data on prices and quantities sold overall and by product type. Making such data available to researchers is crucial for understanding substitution patterns between home-cultivated and dispensary-sourced cannabis and the implications for water use across types of products sold.

The information in this study points toward an important area for future work — why are regulations, significant competition among dispensaries and pressure from home cultivation and nearby states not ensuring that dispensaries consistently sell high-quality products? In other words, why is it that expert growers hired by dispensaries are unable to grow consistently better cannabis than many home cultivators? Are current testing protocols inadequate to ensure that products sold are free from contamination, e.g., from pesticides, mold, mites, or even just seeds, or that the industry itself lacks expertise on dimensions of quality beyond tetrahydrocannabinol (THC) potency? The lack of optimal quality in the industry suggests that other dimensions, e.g., water use, also are not optimal.

Acknowledgments

The authors thank the State of New Mexico Legislature for providing funding for this research; the Santa Fe Water Division, Santa Fe Public Utilities Department for residential water use data; Janie Chermak for guidance and mentorship; Anthony Ortiz and Kurple Magazine for fielding a pilot survey in Spring 2022, and Yuting Yang, David Van der Goes, two anonymous reviewers and seminar participants at the University of New Mexico’s Annual Economics Research Day in 2022 and 2023 for helpful comments. All materials presented herein are the authors’ work and in no way reflect the opinions or views of the Santa Fe Water Division or the State of New Mexico Legislature.