Title
The Relationship Between Cannabis Legalization, Suicide and Mental Health Remains
Unchanged After Recreational Access
Jeffrey A. Singer, MD
1
, Jacob James Rich, MA
2
, Michael Schemenaur, MA
3
, and Robert
Capodilupo, MPhil
4
1. Health Policy Studies, Cato Institute, Washington, DC
2. Drug Policy Project, Reason Foundation, Washington, DC
3. Department of Economics, Eller College of Management, University of Arizona, Tucson, AZ
4. Yale Law School, New Haven, CT
Abstract:
Background. We extend the model introduced by Anderson et al. [M.D. Anderson, D.I. Rees, J.J.
Sabia, American Journal of Public Health 104, 2369-2376] to evaluate the public mental health
implications of new developments in marijuana policy, such as recreational marijuana access and
additional years of data on suicide mortality.
Methods. We obtained state-level suicide data from the National Vital Statistics System's
Mortality Detail Files for 1999-2018. We used panel regression analysis to examine the
association between suicides per 100 000 population and both medical and recreational
marijuana access. Three specifications of the equation were employed as a robustness test.
Results. After adjusting for economic conditions, alcohol taxes and consumption, time effects,
state fixed effects, and state-specific linear time trends, the legalization of recreational marijuana
was associated with a 2.9% (95% confidence interval [CI] = -5.6, -.03) reduction in the suicide
rate for the most rigorous specification, but didn’t pass the robustness check at the P < .05 level
for the other two specifications. However, recreational marijuana legalization was associated
with a 5.4% reduction (95% CI = -9.4%, -1.4%) in suicide rates for males in the 40 to 49 age
group, passing the robustness check for all specifications.
Conclusion. Suicide rates among men in the 40 to 49 age group fell after recreational marijuana
legalization compared with those in states that did not legalize, which is consistent with the
results found by Anderson et al. for medical marijuana when that same population was younger.
There was no consistent relationship between medical marijuana legalization and suicide rates
for any population. Additionally, mental health indicators at the state level were unaffected by
either recreational or medical marijuana legalization. We confirm the findings of Anderson et al.
(2014), observing no consistent relationship between marijuana legalization and observable
mental health outcomes. However, suicide rates increase at lower rates on average relative to
other states following recreational marijuana legalization.
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint
The copyright holder for thisthis version posted September 28, 2020. ; https://doi.org/10.1101/2020.09.25.20201848doi: medRxiv preprint
NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.
1
The Relationship Between Cannabis Legalization, Suicide and Mental Health Remains 1
Unchanged After Recreational Access 2
3
As of September 1 2020, 11 states and the District of Columbia have voted to legalize marijuana 4
for recreational use, while medical marijuana is available in 33 states. A November 2019 survey 5
from the Pew Research Center found that 67 percent of adults support national legalization of 6
marijuana, including 63 percent of the Baby Boomer generation, 76 percent of Millennials, and 7
65 percent of Gen X. (Daniller, 2019) In October 2018 Canada became the first G7 country to 8
federally legalize marijuana for recreational use, and the same month Mexico’s Supreme Court 9
ruled marijuana prohibition unconstitutional, setting the stage for legalization in that country. 10
(Hudak, 2020) In the not-too-distant future marijuana use in the United States might be federally 11
decriminalized and the matter devolved to the states. Legislation to that effect was introduced in 12
the 116
th
Congress. (Nadler, 2020) These developments reflect the increasing acceptance of the 13
recreational use of marijuana, which trends suggest will further increase in days ahead. 14
Policy makers considering marijuana legalization should consider the public health 15
implications of increasing access to a psychoactive substance that could lead to substance use 16
disorder. In a 2009 Norwegian study, Pedersen concluded “exposure to marijuana by itself does 17
not lead to depression but that it may be associated with later suicidal thoughts and attempts.” 18
(Pederson, 2008) A 2014 systematic review and meta-analysis of longitudinal studies by 19
Canadian researchers found an increased risk for developing depressive disorders may be 20
associated with heavy marijuana use. (Lev-Ran, Roerecke, George, McKenzie, & Rehm, 2014) 21
Agrawal et al., in a 2017 retrospective study, found that the correlation between marijuana use, 22
suicidal thoughts, and major depressive disorders in discordant twins was not solely attributable 23
to predisposing factors. (Agrawal, et al., 2017) This observation confirmed similar findings 24
reported by researchers at the same institution ten years earlier. (Lynskey, et al., 2004) A 2015 25
study using community-based samples from the Australian Twin Registry by Delforterie et al. 26
found a “modest association” of marijuana use with suicidal thoughts and attempts, concluding 27
that the association of suicidal thoughts and behaviour with marijuana use disorder requires 28
further study. (Delforterie, et al., 2015) In a 2020 study in JAMA Network Open, Gorfinkel et al. 29
found that people reporting depression tend to use marijuana at a higher rate. (Gorfinkel, Stohl, 30
& Hasin, 2020) 31
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint
The copyright holder for thisthis version posted September 28, 2020. ; https://doi.org/10.1101/2020.09.25.20201848doi: medRxiv preprint
2
However, much of the literature finds no relationship between marijuana use and mental 32
health outcomes. In particular a 2014 regression analysis of state-level data from the National 33
Vital Statistics System (NVSS) Mortality Detail Files from 1990-2007 by Anderson et al. found 34
no general association between medical marijuana legalization and suicide rates, but found 35
10.8% and 9.4% reductions in the suicide rates in males age 20 to 29 and 30 to 39 in the U.S., 36
respectively. (Anderson, Rees, & Sabia, 2014) Additionally, a 2018 Canadian study found no 37
correlation between suicidal behaviour and heavy marijuana use in male or female patients with 38
psychiatric disorders. (Naji, et al., 2018) With the mental outcomes of marijuana use contested 39
by much of the literature, policymakers would benefit from research investigating the population 40
effects on mental health after marijuana legalization. 41
Marijuana use is tracked by the Substance Abuse and Mental Health Services 42
Administration (SAMHSA) with the National Survey on Drug Use and Health (NSDUH). As of 43
2018, about 43.5 million people (or 15.9 percent of the U.S. population above the age of 12) 44
consumed marijuana at least once that year—a historic high. Of those marijuana users, 3.1 45
million used marijuana for the first time, which has been a relatively consistent finding every 46
year since Colorado became the first state to legalize marijuana in 2014. That year 12.9 percent 47
of the U.S. population used marijuana in the past year and the recent trend of increase is 48
predicted to continue into the near future. (SAMHSA, 2015) Marijuana use reported in the past 49
month has also steadily increased in the U.S. since 2007, after a subtler increase starting after an 50
all-time low in 1992. (Golub, Johnson, & Dunlap, 2005) 51
Reports of major depressive episodes have increased since 2010, while both serious 52
thoughts of suicide and severe mental illness have increased each year since the inception of the 53
survey questions in 2008. (SAMHSA, 2019) Suicide rates in the U.S. have been steadily 54
increasing since an all-time low in 1999 and were at an all-time high in 2018. Respondents who 55
reported past month marijuana use in 2017 were much more likely to experience major 56
depression (2.29 times), suicidal thoughts (3.2 times), and slightly more prone to any/severe 57
mental illness (1.16/1.64 times) than nonusers. (SAMHSA, 2018) This observation has garnered 58
much interest in the relationship between marijuana use and possible mental health deterioration. 59
But with increasing evidence of marijuana’s medicinal benefits, and with its growing popularity 60
as a recreational drug, it is necessary to definitively answer the questions surrounding heavy 61
marijuana use and risk of depression and suicide. 62
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint
The copyright holder for thisthis version posted September 28, 2020. ; https://doi.org/10.1101/2020.09.25.20201848doi: medRxiv preprint
3
In the years since the Anderson et al. review of data from 1990-2007, the advent of 63
recreational marijuana laws has provided unprecedented access to marijuana. Significant gaps 64
remain in the knowledge of the potential harms and benefits of marijuana use, and the effects 65
these laws have on various facets of mental illness are a fertile topic for new research. (National 66
Academies of Sciences, Engineering, and Medicine, 2017) 67
68
Methods 69
70
Employing most of the same techniques used by Anderson et al. (2014),
we use population-71
weighted ordinary least squares (OLS) regressions for the observed groups to conduct a state-72
level panel analysis for various time periods of available data to estimate the effect recreational 73
and medical marijuana laws have had on various proxies for mental health: suicide rates per 100 74
000 people for the general population, males, females, and both sexes separately for the age 75
groups 15-19, 20-29, 30-39, 40-49, 50-59, and 60 and above (1999-2018); the percentages of the 76
population that reported “serious thoughts of suicide in the past year,” “any mental illness in the 77
past year,” and “serious mental illness in the past year” for the age groups 18-25, 18 and up, and 78
26 and up (2009-2018); and the percentages of the population that reported they “had at least one 79
major depressive episode in the past year” for the age groups 12-17, 18-25, 18 and up, and 26 80
and up (2006-2018). 81
Similar to Anderson et al. (2014), we control for observable state-level characteristics the 82
literature considers risk factors for suicide: unemployment rates, GDP per capita, beer taxes, and 83
ethanol consumption per capita for the population age 21 and older. Also following Anderson et 84
al. (2014), we attempt to control for unobserved factors with time effects, fixed effects, and state-85
specific linear time trends. Since previous studies suggest stricter alcohol policies can reduce 86
suicides, (Markowitz, Chatterji, & Kaestner, 2003; Birckmayer & Hemenway, 1999; Jones, 87
Pieper, & Robertson, 1992; Robertson, 2014) Anderson et al. (2014) controlled for whether a 88
0.08 blood-alcohol-content (BAC) law was in effect, whether a zero-tolerance drunk-driving law 89
was in effect, and for the beer excise tax. We removed the controls for BAC laws because every 90
state had implemented a 0.08 BAC limit for our entire period, except for Utah, which 91
implemented a .05 limit December 25 2018. Additionally, we removed indicators for zero-92
tolerance drunk-driving laws because every state had implemented such restrictions by 2001. 93
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint
The copyright holder for thisthis version posted September 28, 2020. ; https://doi.org/10.1101/2020.09.25.20201848doi: medRxiv preprint
4
(Carpenter, Heavy alcohol use and youth suicide: Evidence from tougher drunk driving laws, 94
2004; Carpenter, How do Zero Tolerance Drunk Driving Laws work?, 2004) We argue those 95
variables ultimately attempted to capture state government interventions that reduce alcohol use, 96
which is why we added a measurement for per capita ethanol consumption. 97
To test the robustness of our findings, we follow Anderson et al. (2014) and employ a 98
sensitivity analysis involving three separate specifications for each dependent variable: first only 99
adjusting for state and year effects; then adjusting for state effects, year effects, and covariates; 100
and finally adjusting for state effects, year effects, covariates, and state-specific linear time 101
trends in the most rigorous model. Robust standard errors were employed and clustered at the 102
state level for all specifications. Additionally, we took the natural log transform of all dependent 103
variables, and because the log of zero is undefined, we dropped all observations of zero suicides. 104
If the coefficients for the marijuana law indicators remain relatively stable and significant for all 105
regression specifications, there is evidence that a predictive relationship exists between 106
marijuana legalization and the mental health variables of interest. (Miron & Tetelbaum, 2009; 107
Thornton, 2011) 108
Suicide mortality (X60-X84, Y87.0, and U03 intentional self-harm) data and population 109
estimates are sourced from the Centers for Disease Control and Prevention (CDC) Wide-ranging 110
Online Data for Epidemiologic Research (WONDER) NVSS database, reports of mental illness 111
from the SAMHSA NSDUH, unemployment rates and GDP per capita from the Bureau of Labor 112
Statistics, ethanol consumption from the National Institute on Alcohol Abuse and Alcoholism, 113
and beer excise tax figures from the Tax Policy Center at the Urban Institute and Brookings 114
Institution. Marijuana policy indicators were calculated referencing reports by the Marijuana 115
Policy Project and the National Organization for the Reform of Marijuana Laws (NORML). 116
These indicators consider the dates laws were enacted and for the first year assume a value 117
between 0 and 1 equal to the proportion of days the policies were effective. The year recreational 118
marijuana laws became effective were used to graph unadjusted suicides per 100 000 population 119
over the year of law change to compare rates with states that did not legalize recreational 120
marijuana anytime during the period. 121
122
Results 123
124
. CC-BY 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.(which was not certified by peer review)preprint
The copyright holder for thisthis version posted September 28, 2020. ; https://doi.org/10.1101/2020.09.25.20201848doi: medRxiv preprint
