<<

ABSTRACT

DOES A LEGAL “HIGH” LEAD TO HIGHER RENTS? AN ESTIMATION OF THE EFFECTS OF CULTIVATION ON COLORADO INDUSTRIAL PROPERTY VALUES

by Jack Edward Fetick

This paper analyzes the effect of a change in the legal status of cultivation on the value of industrial space. An often-cited positive impact of legalization is the increase in tax revenue to the state, yet an increase in demand for warehouse-like spaces may crowd out other industries. An initial estimation analyzes how different characteristics of Colorado counties affect ballot choices. The results indicate areas that supported Barack Obama in 2012 were 26-37 percent more likely to support legalized cannabis. I then estimate the effect of municipalities exercising a “local option” to ban cannabis production on the values of industrial spaces. The natural variation of the local option across the state is used to compare industrial space values across treated areas using a difference-in-differences design. Using data from real estate listing website LoopNet, this paper estimates a slightly higher difference in the value of industrial space in areas with legal cultivation status.

DOES A LEGAL “HIGH” LEAD TO HIGHER RENTS? AN ESTIMATION OF THE EFFECTS OF CULTIVATION LAWS ON COLORADO INDUSTRIAL PROPERTY VALUES

Thesis

Submitted to the

Faculty of Miami University

in partial fulfillment of

the requirements for the degree of

Master of Arts

by

Jack Edward Fetick

Miami University

Oxford, Ohio

2018

Advisor: Dr. Melissa Thomasson

Reader: Dr. Analisa Packham

Reader: Dr. Austin Smith

©2018 Jack Edward Fetick

This thesis titled

DOES A LEGAL “HIGH” LEAD TO HIGHER RENTS? AN ESTIMATION OF THE EFFECTS OF CULTIVATION LAWS ON COLORADO INDUSTRIAL PROPERTY VALUES

by

Jack Edward Fetick

has been approved for publication by

Farmer School of Business

and

Department of Economics

______

Dr. Melissa Thomasson

______

Dr. Analisa Packham

______

Dr. Austin Smith

Table of Contents I. Introduction 1 II. Literature 2 III. Market Background 3 IV. Data and Estimation 5 a. Limited Dependent Variable Model 5 b. Difference-in-Differences Model 6 V. Results 9 a. Limited Dependent Variable Model 9 b. Difference-in-Differences Model 11 VI. Conclusion 12 VII. Appendix 14 VIII. References 19

iii

List of Tables I. Table 1 5 II. Table 2 6 III. Table 3 9 IV. Table 4 10 V. Table 5 11 VI. Table 6 12 VII. Table 7 14 VIII. Table 8 17 IX. Table 9 18 X. Table 10 18

iv

List of Figures I. Figure 1 7 II. Figure 2 8 III. Figure 3 14 IV. Figure 4 15 V. Figure 5 15 VI. Figure 6 16

v

Dedication

To my friends, my family, and my parents whose encouragement supported me through my years at Miami.

vi

Acknowledgements

I wish to thank all of the faculty at Miami that made my experience great, in particular my advisor, Dr. Thomasson, and my music professor, Dr. Schillinger.

vi

Introduction The changing legality of cannabis in the is a national topic of discussion. Initiatives for legalization compare cannabis to and point towards the failure of alcohol prohibition to prevent negative societal effects. These groups also advocate for the increased tax revenue that a legal market would bring to cash-strapped states. Opponents of legalization argue that it would increase both usage and the negative effects associated with use. There has not been much research into the potential economic effects of creating a completely new industry on a local community. Areas that choose to end cannabis prohibition must be ready to deal with changes to their local economy associated with this change. Cannabis use, like any other , can have a great impact on health. Although prohibited for many years, there are many studies that measure the effect of cannabis use on personal health. There have been many areas across the world that have chosen to legalize, allowing for more research activities. Recently, there have been many studies utilizing the variation in laws from various countries as a natural experiment to study cannabis effects. Marie and Zölitz (2015) analyzed a change in cannabis accessibility for college students and found that restricting access positively affects academic performance. While some arguments against legalization cite the aforementioned research as evidence of the positive effects of restricting access on adolescents, other studies have shown that allowing a regulated medical industry has no effect on adolescent consumption. A common argument in recreational legalization efforts compares cannabis to alcohol. In an early twentieth century temperance wave that also led to the banning of cannabis, alcohol was prohibited for thirteen years in the U.S. While overall alcohol consumption fell, numerous negative effects such as and homicide skyrocketed. Prior research has shown that the decrease in direct effects of alcohol use was offset by the indirect effects of crime and the great increase in government expenditures, particularly on prohibition enforcement. With no legal alcohol, the government could not collect a sin tax to fund this increase in expenditures. These conditions led to the repeal of alcohol prohibition in 1933 and many cannabis activists argue the government is facing the same problem with prohibition today. Prohibition is ineffective at eliminating drug use and introduces worse negative crime externalities. Although a sin tax may increase the amount of drug consumption compared to outright banning, advocates note that there would be a decrease in related illegal activity. Policymakers following the “” approach tend to favor an approach of legalization, education, and taxation. The prohibition on consumption are similar to alcohol. While the societal impact of drug consumption often receives the most attention, economic outcomes can be harder to study. Drug trade (both legal and illegal) may negatively affect a local economy by discouraging other businesses from locating nearby. But this issue suffers from reverse causality; drug activity might flourish in neighborhoods with few other businesses. Additionally, individual consumption of illicit substances is quite difficult to observe, as are any associated transactions. This is a problem when studying the cannabis industry. Even though legalization may have repealed much of the illegal trade, a black market still persists.1 Still, it is evident that Colorado and other states have been economically impacted by legalization. Once

1 Light, Miles et al. (2016). The Economic Impact of Marijuana Legalization in Colorado. Marijuana Policy Group, Market Intelligence.

1 empty warehouse districts have become centers of cultivation and many local municipalities, in addition to the state itself, have reaped the rewards of taxation.2 Yet, there are negative implications as well. A preliminary study has found evidence that a retail cannabis dispensary can act as a negative externality and decrease residential property values (Thomas and Tian (2017)). Cannabis is still not as prevalent as alcohol, and many firms and households make an effort to avoid the newly legal industry.3 While there is some research on residential effects, there are no studies of legalization on surrounding businesses. Firms may attempt to distance themselves from new marijuana operations. Legalization has created a new industry with a significant economic impact, but unlike some societal effects, it can be difficult to measure. This paper analyzes the characteristics of Colorado counties that may influence support of the cannabis industry. A probit model applied to demographic and ballot data from 2012 found that counties that tend to support Barack Obama and that have a higher concentration of medical marijuana centers and liquor stores are more likely to support legalization than not. I also measure the effect of recreational cannabis legalization on the value of warehouse space in Colorado. Colorado’s cannabis allows for municipalities to ban different types of marijuana business, which creates both treatment and control counties. Using data from a real estate listing website, a difference-in-differences model is used to estimate effects across treatment groups. I find that areas that allow legal recreational cultivation tend to have lower commercial property values than those that have exercised the local option. The change in the law is also associated with an increase in industrial property values across the entire state. There may be some evidence that areas with legal cultivation have had a higher increase in value than other areas since legalization, but further study is needed. Literature There is not an extensive literature that studies the effects of recreational cannabis legalization. However, there has been some work that has studied the societal effects of medical marijuana and there has been research into the effects of alcohol prohibition, which have some parallels with the legal status of cannabis. One particular working paper uses a similar methodology to this paper to study the effect of retail dispensaries on nearby residential property values. It is important to note that while legal changes are of great interest to researchers, not much can actually be gleaned from these studies. The enormous uncertainty of legal protections and the heterogeneity of medical marijuana laws across jurisdictions means that we do not yet know much about the benefits and harms of marijuana liberalization policies (Pacula & Sevigny, 2014). Public health and epidemiology have driven much of the prior research into the effects of medical marijuana. A paper by Cerdá et al. (2011) found that areas with legal had higher rates of marijuana use. However, a later critique of earlier studies by Harper et al. (2012) controlled for unmeasured state characteristics and failed to find a causal relationship between legal medical cannabis and consumption. Choo et al. (2014) applied similar methods to Harper et al., and indicated that changes in medical cannabis laws have no effect on adolescent cannabis use. While there is not a wide breadth of literature studying the effects of recreational

2 https://www.usnews.com/news/best-states/colorado/articles/2017-07-20/colorado-pot-tax-revenue-surpasses-500- million 3 https://vtdigger.org/2016/12/26/lessons-colorado-one-city-bucks-pot-legalization/

2 legalization, the evidence from medical cannabis studies suggests that a legal recreational market may not affect consumption locally. The analysis of alcohol prohibition has found some interesting results comparable to cannabis laws. In 1991, Miron and Zwiebel looked at alcohol consumption before, during, and after prohibition and found that while consumption fell drastically at the onset, it steadily rose to about 60-70 percent of the pre-prohibition levels. A later study found that the government imposed prohibition led to substantial undesirable consequences, such as an increase in violent crime and alcoholism, and failed to reduce consumption over time (Miron (1998)). In another work, Miron and Zwiebel (1995) conclude from the available evidence that a free market for is superior to the current prohibition model. They discuss evidence from alcohol prohibition, and predict that consumption may increase once a drug is legalized, but that the legalization will eliminate some negative societal effects from the black market. Violent crime associated with private drug deals will decrease as drug users shift to purchasing from a regulated, legal business. A review of the prohibition literature may indicate that cannabis legalization may yield a positive effect on society. While the economics of alcohol prohibition have been widely studied, the economics of cannabis remains a relatively dry field. There have been some studies that analyze the spatial effects associated with dispensary location. Medical dispensaries in California were shown to be prone to open in areas with already existing high demand, low income, and numerous alcohol outlets (Morrison et al. (2013)). A study of the Washington recreational market found that a dispensary opening can serve as a negative externality to nearby home values (Thomas and Tian (2017)). This paper uses a difference–in-differences model to estimate the differential impact of dispensary opening on residential property values. My estimation will follow a similar design to the aforementioned paper. Significant research on legalization may seem scarce currently, but not for long. The new cannabis industry will have a significant impact on local economies and the natural experiment of legalization can be a fruitful area for research. Market Background The cannabis industry has changed dramatically in the last 20 years. In Colorado, medical cannabis was first legalized after a voter initiative in 2000, allowing patients to grow or receive cannabis with a doctor’s permission. However, guidance from the DEA barred “caregivers” from providing medical cannabis to more than five patients until overturned in a 2007 Colorado court decision.4 With medical cannabis now legally widely available, the state legislature passed the Colorado Medical Marijuana Code which created a state licensing agency and various business license types. The success of the medical industry helped drive the campaign to legalize recreational cannabis use and on November 6, 2012, voters in both Colorado and Washington approved ballot measures to legalize recreational cannabis. The Colorado legislature then created the Retail Marijuana Code, which enacted a series of regulations and six specific business license types for the recreational industry. The law also created the “local option,” a rule in which local municipalities can choose whether to allow certain license types within their jurisdiction. This option reflects the narrow margin by which Amendment 64 was passed and allows strongly opposed areas to maintain the status quo.5 Since the date of the vote, many areas have chosen to block parts of the industry from their area: As of April 2017, 176 of 272 municipalities have chosen to block at least one type of license in their area. Appendix B includes a map that displays the

4 https://www.denverpost.com/2007/06/23/medical-marijuana-user-sues-over-state-policy/ 5 See appendix A for vote summary

3 variation in laws across Colorado counties and a map that displays the variation of only cultivation status. In contrast with the recent movement in several states, federal drug laws have remained relatively unchanged since the 1970s. Cannabis remains in Schedule 1 of the Controlled Substances Act and it remains illegal to “manufacture, distribute or dispense, or possess…” cannabis in the United States. Both sides of this issue have attempted to exploit the disparity in legal status to overturn either state or federal laws, but the issue remains ambiguous. No state law has been overturned, but the federal law has been applied in court cases regarding employment and financing decisions. Since banks are federally regulated, no major lender has engaged with the industry for fear of federal prosecution; many cannabis operations are run on a completely cash basis.6 Although almost unheard of, these business operations are subject to federal prosecution, though this risk decreased on August 29th, 2013 when the Obama administration issued the Cole Memorandum, directing attorney generals to avoid prosecuting operations that are in harmony with state law. In this state of national uncertainty, legal recreational sales began in Colorado on January 1, 2014. The market has been subject to many changes in the past four years. High early profits led to many entrants and frequent law and regulation changes as the state learned how to police a new industry.7 Many medical cannabis businesses converted their operations into dual retail/medical establishments, though there were plenty of new entrants as well.8 Since the law passed in 2012, the impact has greatly varied across different areas in the state. Cultivation operations and dispensaries now dominate different neighborhoods. The state has reaped in record tax revenues from this large growth. This “” has found steadily increasing demand in the face of decreasing prices. Initial higher prices have begun to shrink in part due to increased competition and economies of scale. Although the market has mainly faced challenges from internal issues such as decreasing prices, worries of a property seizure threat posed by the Justice Department still loom overhead. Regulatory oversight is nested in a division of the Department of Revenue- the Marijuana Enforcement Division (MED). The division issues all business licenses, regulates operations, and provides information for the entire state cannabis industry. They provide useful data, including the address of every cannabis business location. Potential entrants can apply for a license to operate a retail marijuana store, cultivation site, manufactured product operation, testing facility, transportation operation, or a professional service operation. Some of these licenses allow licensees to test the quality cannabis, transport the drug to and from different establishments, or provide an operation service, such as security and protection to cultivation sites. Table 1 displays the current counts of each license in the state. Until October 2014, the MED required vertical integration-cultivation, processing, and sales had to be done by the same business. This has changed, but many firms remain integrated. Cultivation operations are required to be indoors and are very important for every business. Proper growing locations are in great demand currently. While some industrial spaces are more suited for growing than others, the plant itself is not that difficult to cultivate and the most important qualities of prospective sites are proper industrial

6 https://www.cnbc.com/2017/04/18/marijuana-companies-sending-a-huge-cash-roll-to-irs-on-tax-day.html 7 Light, Miles et al. (2016). The Economic Impact of Marijuana Legalization in Colorado. Marijuana Policy Group, Market Intelligence. 8 http://www.9news.com/article/entertainment/television/programs/next-with-kyle-clark/colorado-nears-3000-active- marijuana-businesses/368693213

4 zoning and a legal cultivation status, according to Rodman Schley, a commercial real estate appraiser in Colorado with experience in the industry. Table 1

# of Licensed Retail Marijuana Facilities

Stores 518

Cultivations 722

Operators 7

Product Manufacturers 280

Testing Facilities 12

Transporters 10

updated February 1, 2018 Notes: License counts retrieved from the Colorado Marijuana Enforcement Division website. Data and Estimation Limited Dependent Variable Model The probability analysis contains county-level demographic information from various sources. Income, race, and unemployment estimates for 2012 are all from the American Community Survey (ACS). The number of liquor licenses, net sales tax revenue, and medical marijuana centers come from the Colorado Department of Revenue’s 2012 report. Religious congregation information is available from the 2010 US Religion Census from the Association of Statisticians of American Religious Bodies. To calculate the number of congregations per 10,000, I use 2010 county population estimates from the US Census. Election results from both the presidential election and the state constitutional amendment are from the Colorado Board of Elections. Here, I estimate the likelihood of a county to vote yes on Amendment 64 and to allow recreational cultivation. This analysis uses a probit model to estimate the binary vote outcome. Various demographic details that may impact the cannabis stance of a county are included in a vector of characteristics on the right hand side. I use a probit model rather than a linear probability model because in the latter on average 58 per cent of the residuals are below zero or above one.9The probit estimation takes the following form

9 See Appendix D for LPM results.

5

Probit Model

푦퐶 = 훷(훽0 + 푋퐶훽퐶 + 휀퐶) where 푦퐶 is the election or policy outcome for county c and 푋퐶 is a vector of county demographic information. I estimate 훽퐶, the vector of marginal effects on the outcome variable. Difference-in-Differences Model This estimation uses data from the real estate listing website LoopNet. LoopNet, one of the internet’s largest commercial real estate listing services, allows owners to list properties and contact realtors and potential renters. LoopNet has been absorbed into parent company Costar since August 2017, when I obtained my data, and no longer offers free access.10 These historical data are from a listing of industrial properties across the state of Colorado. There are 2,199 sale or rent postings of 414 different industrial spaces. The dates on market range from 1997 to active in August 2017. Basic details of each property are included such as number of stories, space available, and age. Using the available listed city, each listing has been classified into two groups, based on the status of recreational in the municipality. A list of areas that allow recreational cannabis cultivation can be found on the MED website.11 I also cross-checked certain unclear listings from this list with the municipal code for the specific area. Table 2 Summary Statistics

All Listings Legal Cultivation Area Illegal Cultivation Area Mean (Standard Deviation)

Rental Rate 0.75 0.72 0.78 (.53) (.46) (.60) Price/SF 85 87 84 (54) (62) (46) Age 34 38 31 (17) (18) (16) Size (SF) 16103 16956 15274 (37734) (34731) (40439)

All Listings Legal Cultivation Area Illegal Cultivation Area Total # of Listings 2199 1086 1113 Rental # of Listings 1512 746 766 Sale # Listings 687 340 347 The value of industrial space in this hedonic pricing model can be represented by either the market rental rate or the price per square foot, depending on the transaction type. I partition my data into groups depending on the listing type. As mentioned earlier, the “local option” of

10 My access to LoopNet was limited to 500 property records, of which only 414 were viable. This small sample limits the strength of any inference. Future studies should find larger, more granular, data sources. 11https://www.colorado.gov/pacific/sites/default/files/Local%20Authority%20Status%20List%2001292018%20CUR RENT%20VERSION.pdf

6 municipalities to determine their own marijuana laws serves as a source of variation to study the local effect of cannabis legalization. Subsections of the state are divided into treated and control groups. Both the date of the passing of Amendment 64 and the first date of recreational sales are tested as the treatment date. The division of municipalities by legal status motivates a difference- in-differences (DD) design. There are some limitations with using this DD design. Although only certain areas receive the legal recreational cultivation treatment, industrial spaces in control counties are impacted by the change too. Entry into the cultivation market could crowd out industrial space in other industries, potentially driving up price. I only observe properties that were available for new tenants, and could not measure the change in value in contract renewals or by appraisal. These unobserved valuations are a source of bias for my estimation, although it is unclear in which direction the bias leans. There is also difficulty in determining an exact treatment date. Medical cannabis cultivation had been legal for 3 years at the passage of Amendment 64, and that certainly impacted the available industrial spaces. Local municipalities chose to ban cultivation at unspecified times after the passage of the law as well. I will test both November 6, 2012, the passing date of amendment 64, and January 1, 2014, the first day of recreational sales, as treatment dates. Different local areas varied when they responded to the law. These two dates allow for a comparison between the market response to the law passing and to the beginning of recreational sales. For a DD design, there must be an assumption of pre-treatment parallel trends across both groups. Figures 1 and 2 display the trends for both groups before and after the treatment date. There may be different values for industrial spaces in legal and illegal spaces, but for most graphs there is evidence of trends in the same direction before the treatment date. The years immediately before the 2012 election treatment date have parallel movement in both rental rate and price per square foot. Figure 1

7

Figure 2

My estimation equations take the following form

log(푅푒푛푡푎푙푅푎푡푒푖,푗,푡) = 휆푦 + 훼푗 + 훽푖푋푖 + 훽푗퐿퐶푗 + 훽푡푇푡 + 훾퐿퐶푗 ∗ 푇푡 + 휀푖,푗,푡

log(푃푟푖푐푒/푆퐹푖,푗,푡) = 휆푦 + 훼푗 + 훽푖푋푖 + 훽푗퐿퐶푗 + 훽푡푇푡 + 훾퐿퐶푗 ∗ 푇푡 + 휀푖,푗,푡

For the DD estimation, 푅푒푛푡푎푙푅푎푡푒푖,푡 is the rent per square foot per month of space i in city j at time t and 푃푟푖푐푒/푆퐹푖,푡 is the price per square foot of similarly specified spaces. 푋푖 is a vector of property characteristics of each space (Table 2) included to control for qualities of each warehouse that affect value. 휆푦 and 훼푗 are year and city fixed effects, respectively. 퐿퐶푗 is an indicator variable that is equal to one if it is legal to cultivate recreational cannabis in the municipality. Similarly, 푇푡 is an indicator variable that equals one for any date after the treatment. I estimate γ which measures the difference in percentage change of value between areas where it is legal to cultivate recreational cannabis after the treatment date.

8

Results Limited Dependent Variable Table 3

Pr(County vote yes on A 64) Marginal Effect

Median HH Income .00000429 (.00000231) Net Sales Tax PC -.0000238 (.0000349) Percentage Non White -.431 (.446) Unemployment Rate .0315 (.0265) Obama Won in 2012 .370*** (.074) Congregations per 10,000 -.00502*** (.00174) Liquor Licenses per 10,000 .00245*** (.000659) Medical MJ centers per 10,000 .0928*** (.0301)

Observations 64 Notes: The sample contains demographic and electoral information from 2012 from all Colorado counties. The dependent variable is the probability that a county would vote yes on Amendment 64. Standard errors are corrected for heteroskedasticity and are clustered at the county level.

9

Table 4

Pr(County allowing recreational cultivation) Marginal Effect

Median HH Income .00000479 (.00000308) Net Sales Tax PC -.000487* (.000254) Percentage Non White .599 (.592) Unemployment Rate .0607 (.0406) Obama Won in 2012 .264*** (.0659) Congregations per 10,000 -.00267 (.00243) Liquor Licenses per 10,000 .00346*** (.00111) Medical MJ centers per 10,000 .110*** (.0346)

Observations 64 Notes: The sample contains demographic and electoral information from 2012 from all Colorado counties. The dependent variable is the probability that a county would allow recreational cultivation. Standard errors are corrected for heteroskedasticity and are clustered at the county level. Tables 3 and 4 display the linear probability model results. For the vote on Amendment 64, there are four characteristics that have a statistically significant association with the vote outcome. Counties in which Barack Obama was reelected are about 37 percent more likely to choose to legalize. An increase of 1 per 10,000 residents increases the probability of a yes by .25 percent and an increase of one medical marijuana center per ten thousand residents increases the probability by 9.3 percent. Areas with higher concentrations of religious congregations were less likely to support the amendment. There are slightly different results for the model that estimates whether a county allows legal recreational cultivation. Here, counties that supported Obama are more likely to support legal cultivation, along with areas with many liquor licenses and medical marijuana centers per capita. However, there is no statistically significant relationship between religious congregations and cultivation status. There may be evidence that areas with higher net sales tax per capita are less likely to allow cannabis cultivation locally. Similar characteristics of these counties may lead to support for different forms of legalization, both statewide and local cultivation.12

12 LPM results in Appendix D are robust only to the significance and direction of the presidential electoral results.

10

Difference-in-Differences Table 5

Log(Rental Rate) (1) (2) (3) (4) (5)

PostLaw*LC 0.106 0.0586 0.0799 0.0618 0.0785 (0.0660) (0.0734) (0.0660) (0.0605) (0.0570) PostLaw 0.182*** 0.183*** 0.159*** (0.0400) (0.0463) (0.0459) LegalCultivation -0.167** -0.140 -0.141 (0.0725) (0.0845) (0.0922) City Fixed Effects N N N Y Y Year Fixed Effects N N Y N Y Property Characteristics N Y Y Y Y N 1,493 1,443 1,443 1,443 1,443 R-squared 0.068 0.128 0.174 0.284 0.323 Notes: The sample contains information from 1,443 industrial properties across Colorado. The dependent variable is the percentage change in rental rate per square foot per month. Standard errors are corrected for heteroskedasticity and are clustered at the city level. Results from my first regression are found in Table 5. Initially, without including property characteristics or any fixed effects, there is a 10.6 percent increase in areas that chose to allow recreational cultivation after the law passed compared to areas that did not. However this is not a statistically significant relationship. Yet, there are stronger relationships in this initial regression that indicate both an 18.2 percent increase in industrial space values across the entire sample after the law, although it is unclear if this is due specifically to legalization, and that areas that choose to allow recreational cultivation have before the law on average 16.7 percent lower property values than those that do not allow cultivation. Further regression add both year and city fixed effects in addition to property characteristics. The preferred regression is column (5) which controls for more sources of unobserved heterogeneity. This regression indicates that there may be a 7.85 percent increase in areas that chose to allow cultivation after the law passed compared to those that do not, although there is no significant level of confidence for this association. Across all columns that include this term, there is a significant positive relationship between the value of industrial space and the passing of the law. There is also a consistently negative relationship between areas that choose to legalize cultivation and the industrial space value before the law.

11

Table 6

Log(Price/SF) (1) (2) (3) (4) (5)

PostLaw*LC 0.219 0.245 0.258 0.161 0.198 (0.217) (0.222) (0.238) (0.208) (0.216) PostLaw 0.0535 0.0732 0.142 (0.164) (0.174) (0.147) LegalCultivation 0.0265 -0.00626 0.00753 (0.108) (0.0977) (0.105) City Fixed Effects N N N Y Y Year Fixed Effects N N Y N Y Property Characteristics N Y Y Y Y N 685 663 663 663 663 R-squared 0.015 0.042 0.106 0.240 0.300 Notes: The sample contains information from 663 industrial properties across Colorado. The dependent variable is the percentage change in price per square foot per month. Standard errors are corrected for heteroskedasticity and are clustered at the city level. Table 6 uses the log of price per square foot as the dependent variable. Perhaps due to the low sample size, there are no statistically significant relationships. The results for the first row are similar to Table 5 in direction and magnitude. There are also similar results for the PostLaw row for magnitude and direction, but these relationships are insignificant. The results for the legal cultivation status indicator variable are not robust to the change in dependent variable. As a robustness check, I also vary the treatment date. In addition to testing the date of the passage of Amendment 64, I have tested the date of the first recreational sales. These tables are included in appendix E. With rental rate as the dependent variable, the results of Table 5 are robust in significance and direction to a change in the treatment date. All of the coefficients on the treatment indicator variables remain positive and significant at the 99 percentage confidence level. These coefficients increase relative to the earlier treatment coefficients and increase by around 6 percentage points. There is still a negative relationship between areas that choose to legalize recreational cultivation and rental rates before the treatment goes into effect. The coefficient of interest maintains both the same direction and similar levels of magnitude, yet still does not have any statistical significance. When the treatment date for the price per square foot regression was changed, a positive, statistically significant relationship was found for the post treatment indicator variable, similar to the rental rate regressions. Unlike earlier estimations, there is a positive pretreatment relationship between areas with legal cultivation status and price per square foot. The coefficient of interest varies in direction as different controls are added, and there is no significant relationship. Conclusion This paper studies the characteristics of Colorado areas that may lead to support of cannabis legalization and the effects of those changes on industrial space property values. Initially, an estimation finds several significant relationships between different aspects of counties and the likelihood of supporting Amendment 64 and allowing local recreational cultivation. I then use the

12 variation in legal status of cultivation across municipalities to measure a change in industrial values. The analysis compares industrial space value between legal and illegal areas after the passage of the law using a difference-in-differences approach. The counties in Colorado that support recreational cannabis tend to be Democratic, have more liquor licenses, and more medical marijuana centers than other counties in the state. Those in the state that may be more tolerant to substance use already could be more accepting of recreational use of cannabis. There may also be a strong negative relationship between a large number or religious groups and voting for Amendment 64. However, this relationship loses significance when legal cultivation becomes the dependent variable. Perhaps previously strong religious opposition to legal cannabis faded away once the law had been effect for a while and county leaders may have seen some benefits to legalization. Further study could study whether there is varying support across urban and rural areas of the state. I find an increase in property values in legal cultivation areas greater than other areas in Colorado. This increase represents a change in space value, due to increased demand for cultivation space. Yet, there is more evidence that industrial property values have increased throughout the state. I estimate a 16-18 percent increase in values across the state after the law passed; this would mean a $1,932 increase in yearly rent of the average industrial space. Local policymakers should be aware of this increase in value when making the decision on whether or not to exercise the “local option.” Possible tax gains from higher rents might be unwelcome in the face of cannabis cultivation crowding out other industries. This may explain the lower rental rates in areas with legal cultivation. These may be poorer areas that choose to allow this industry to increase tax revenues. With the important social and economic effects regarding this issue, cannabis legalization remains an important area of public policy. Specific economic effects will continue to be a key area to study. Future research could continue to use the variation in municipal laws in Colorado to determine possible causal relationships between legalization choices and consumption, tax revenue, and the makeup of local industries.

13

Appendix A. Amendment 64 Vote by County Figure 3

Table 7

Results Votes % Yes 1,383,139 55.32% No 1,116,894 44.68% Total Votes 2,500,033 100.00% Registered voters/turnout 3,647,082 68.55%

14

B. Recreational Marijuana Status by County Figure 4

Figure 5

15

C. LoopNet listing locations and Recreational Cultivation Legality Figure 6

16

D. Linear Probability Model Estimation Table 8 Linear Probability Model A64Pass Legal Cultivation

Median HH Income .00000875 .00000339 (.00000596) (.0000048) Net Sales Tax PC -0.000102 .0000970** (.0000658) (.0000482) Percentage Non White 0.288 0.421 (0.511) (0.732) Unemployment Rate 0.0340 0.0477* (0.0280) (0.0275) Obama Won in 2012 0.614*** 0.445*** (0.120) (0.145) Congregations per 10,000 -0.00241 0.000720 (0.00278) (0.00253) Liquor License per 10,000 0.00226 0.00135 (0.00144) (0.00108) Medical MJ centers per 10,000 0.0168 0.0554** (0.0159) (0.0216) Constant -0.406 -0.369 (0.376) (0.321)

Observations 64 64 Notes: The sample contains demographic and electoral information from 2012 from all Colorado counties. The dependent variables are the probability that a county would vote yes on Amendment 64 and the probability that a county would allow recreational cultivation. Standard errors are corrected for heteroskedasticity and are clustered at the county level.

17

E. Varied Treatment Date for DD Estimation Table 9

Log(Rental Rate) (1) (2) (3) (4) (5)

PostEffect*LC 0.157** 0.102 0.104 0.0817 0.0817 (0.0647) (0.0668) (0.0736) (0.0699) (0.0779) PostEffect 0.236*** 0.243*** 0.229*** (0.0317) (0.0382) (0.0430) LegalCultivation -0.162** -0.135 -0.138 (0.0697) (0.0819) (0.0922) City Fixed Effects N N N Y Y Year Fixed Effects N N Y N Y Property Characteristics N Y Y Y Y N 1,493 1,443 1,443 1,443 1,443 R-squared 0.102 0.159 0.173 0.309 0.322 Notes: The sample contains information from 1,443 industrial properties across Colorado. The dependent variable is the percentage change in rental rate per square foot per month. Standard errors are corrected for heteroskedasticity and are clustered at the city level. Table 10

Log(Price/SF) (1) (2) (3) (4) (5)

PostEffect*LC -0.0119 0.0320 0.00785 0.0113 0.00617 (0.224) (0.219) (0.219) (0.231) (0.229) PostEffect 0.394*** 0.351** 0.386** (0.133) (0.135) (0.145) LegalCultivation 0.132 0.0887 0.107 (0.126) (0.109) (0.106) City Fixed Effects N N N Y Y Year Fixed Effects N N Y N Y Property Characteristics N Y Y Y Y N 685 663 663 663 663 R-squared 0.041 0.061 0.098 0.260 0.293 Notes: The sample contains information from 663 industrial properties across Colorado. The dependent variable is the percentage change in price per square foot per month. Standard errors are corrected for heteroskedasticity and are clustered at the city level.

18

References Cerdá, M. et al., (2011). Medical marijuana laws in 50 states: Investigating the relationship between state legalization of medical marijuana and marijuana use, abuse and dependence. Drug and Alcohol Dependence, 120 (2012) 22-27. Choo, Esther K., MD et al. (2014). The Impact of State Medical Marijuana Legislation on Adolescent Marijuana Use. Journal of Adolescent Health, 55 (2014) 160-166. Harper, Sam, PhD et al. (2012). Do Medical Marijuana Laws Increase Marijuana Use? Replication Study and Extension. Annals of Epidemiology, 22 (3) 207-212. Light, Miles et al. (2016). The Economic Impact of Marijuana Legalization in Colorado. Marijuana Policy Group, Market Intelligence. Marie, Olivier & Zölitz, Ulf. (2015). ‘High’ Achievers? Cannabis Access and Academic Performance. IZA Discussion Paper No. 8900. Miron, Jeffrey A. (1998). An Economic Analysis of Alcohol Prohibition. Journal of Drug Issues, 28(3) 741-762. Miron, Jeffrey A. & Zwiebel, Jeffrey (1991). Alcohol Consumption During Prohibition. Working Paper No. 3675, National Bureau of Economic Research. Miron, Jeffrey A. & Zwiebel, Jeffrey (1995). The Economic Case Against Drug Prohibition. Journal of Economic Perspectives, 9(4) 175-192. Morrison, Chris et al. (2013). The economic geography of medical cannabis dispensaries in California. International Journal of , 25(2014) 508-515. Pacula, Rosalie Liccardo & Sevigny, Eric L. (2014). Marijuana Liberalization Policies: Why We Can’t Learn Much from Policy Still in Motion. Journal of Policy Analysis and Management, 33(1) 212-221. Thomas, Danna, & Tian, Lin (2017). Hits from the Bong: The Impact of Recreational Marijuana Dispensaries on Property Values. Working Paper.

19