Content uploaded by Meenakshi Sabina Subbaraman
Author content
All content in this area was uploaded by Meenakshi Sabina Subbaraman on May 20, 2022
Content may be subject to copyright.
Substitution and complementarity of alcohol and cannabis: A
review of the literature
Dr Meenakshi Sabina Subbaraman
Alcohol Research Group, Emeryville, United States
Meenakshi Sabina Subbaraman: msubbaraman@arg.org
Abstract
Background—Whether alcohol and cannabis are used as substitutes or complements remains
debated, and findings across various disciplines have not been synthesized to date.
Objective—This paper is a first step towards organizing the interdisciplinary literature on alcohol
and cannabis substitution and complementarity.
Method—Electronic searches were performed using PubMed and ISI Web of Knowledge.
Behavioral studies of humans with ‘alcohol’ (or ‘ethanol’) and ‘cannabis’ (or ‘marijuana”) and
‘complement*’ (or ‘substitut*’) in the title or as a keyword were considered. Studies were
organized according to sample characteristics (youth, general population, clinical and community-
based). These groups were not set
a priori
, but were informed by the literature review process.
Results—Of the 39 studies reviewed, 16 support substitution, ten support complementarity, 12
support neither and one supports both. Results from studies of youth suggest that youth may
reduce alcohol in more liberal cannabis environments (substitute), but reduce cannabis in more
stringent alcohol environments (complement). Results from the general population suggest that
substitution of cannabis for alcohol may occur under more lenient cannabis policies, though
cannabis-related laws may affect alcohol use differently across genders and racial groups.
Conclusions—Alcohol and cannabis act as both substitutes and complements. Policies aimed at
one substance may inadvertently affect consumption of other substances. Future studies should
collect fine-grained longitudinal, prospective data from the general population and subgroups of
interest, especially in locations likely to legalize cannabis.
Keywords
Cannabis; alcohol; substitutes; complements
INTRODUCTION
Alcohol and cannabis are two of the most commonly used drugs in the world (Substance
Abuse and Mental Health Services Administration, 2012). However, the question of whether
the two substances are consumed as substitutes or complements remains debated.
Correspondence to: Meenakshi Sabina Subbaraman, msubbaraman@arg.org.
HHS Public Access
Author manuscript
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
Published in final edited form as:
Subst Use Misuse
. 2016 September 18; 51(11): 1399–1414. doi:10.3109/10826084.2016.1170145.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Furthermore, findings across various disciplines (e.g., economics, epidemiology) have never
been synthesized.
Generally, a “substitute” is something that takes the place of something else and a
“complement” is something that completes something else or makes it better. These
definitions can be applied to categorize how two drugs interact with one another such that
they are “substitutes” if one drug can pharmacologically replace the other; “complements” if
the effects of one drug are enhanced by the other; and “independent” if the effects of one
drug are unaltered by the other (Hursh et al., 2005).
There are myriad individual and/or societal factors that may influence one’s substance use
and thus one’s tendency to substitute or complement substances. For example, patterns of
substance use can vary across income levels and social classes, and can be connected to
social inequality and marginalization (Room, 2005). Similarly, the prevalence of substance
use varies across countries due to differences in drug policies or cultures (Degenhardt et al.,
2008). Social setting and networks can also influence use patterns, for example through
developing a risk/reward tradeoff for substance use (Hunt, Evans, & Kares, 2007). All of
these individual and societal factors could affect one’s propensity to substitute/complement.
The goal of this paper is to review behavioral studies that explicitly examined substitution/
complementarity of alcohol and cannabis with empirical data. The recent movements
towards cannabis legalization in the US call for a better understanding of whether cannabis
and alcohol act like substitutes or complements in the general population and among
important subgroups (e.g., youth), especially if cannabis use becomes more prevalent as a
result. Identifying subgroups prone to using the two as complements is particularly
important because combined use can lead to greater impairment than ingestion of either
substance alone (Ronen et al., 2010). Furthermore, understanding how a policy aimed at
reducing the consumption of one substance (e.g., cannabis criminalization laws) affects
consumption of another substance (e.g., alcohol) is crucial for developing optimal policies
and recognizing potential unintended consequences (e.g., cannabis laws may inadvertently
affect alcohol use).
METHOD
Electronic searches were performed using PubMed and ISI Web of Knowledge. Articles
written in English with ‘
alcohol
’ (or ‘
ethanol
’) and ‘
cannabis
’ (or ‘
marijuana”
) and
‘
complement*
’ (or ‘
substitut*
’) in the title or as a keyword were considered. The Web of
Knowledge is the world’s largest accessible citation database and allows for in-depth
exploration of specialized sub-fields within social science disciplines (Thompson Reuters,
2016). Although “co-use” is a relevant term that was considered for searches, co-use is a
much broader concept that does not necessarily reflect the mechanisms of how individuals
use alcohol and cannabis together (or separately); thus “co-use” was not used as a search
term. Inclusion criteria were: 1) empirical studies of humans, and 2) if the independent
variable was cannabis-related, the study needed to include an alcohol-related dependent
variable; if the independent variable was alcohol-related, the study needed to include a
cannabis-related dependent variable. Based on the title and abstract, articles were considered
Subbaraman Page 2
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
for inclusion, then read and organized according to sample characteristics (i.e., adolescents
and young adults, general population, clinical and community-based; please see Tables 1–3).
These three groups were not set
a priori
, but were informed by the literature review process.
This study was approved by the Institutional Review Board of the Public Health Institute.
RESULTS
Searches of ‘
complement*
’ yielded 69 results; 21 met inclusion criteria. Searches of
‘
substitut*
’ yielded 48 results; 27 met inclusion criteria. Nine studies overlapped between
searches for a total of 39 studies reviewed. Of the 39 studies, 17 were conducted among
youth/young adults, nine were conducted using the general population, and 13 were
conducted with clinical samples. Twenty-one studies relied on cross-sectional designs and
17 utilized longitudinal data. The majority of studies (29/39) used individual-level outcomes,
while 10 studies employed aggregate-level outcomes. In terms of independent/exposure
variables, almost half (16/39) used cannabis-related policies as the primary independent
variable. For details pertaining to each study, please see Tables 1–3.
DISCUSSION
Adolescents and young adults
Almost all studies of substitution/complementarity in youth involve cannabis- or alcohol-
related policy. The strongest results come from these “natural experiments” of policy
changes using longitudinal data (Angrist & Pischke, 2010), which account for both cross-
sectional variation (e.g., differences in laws across states) and variation across time (e.g.,
changes in laws within a state over time).
Longitudinal studies of adolescents and young adults
Two longitudinal studies have focused on youth. First, Pacula (1998) used annual data from
the National Longitudinal Survey of Youth (NLSY) 1979 cohort (N=8,008) to examine the
effects of state-level beer taxes on past 30-day number of drinks and cannabis frequency
(Pacula, 1998). The NLSY79 data showed complementary between cannabis and alcohol:
doubling the beer tax reduced the probability of cannabis use by 11.4%, while only reducing
the probability of drinking by 3.2%. Pacula (1998) also assessed effects of cannabis
decriminalization using a dichotomous indicator and controlling for aggregate factors such
as alcohol and cannabis prices and the ratio of crimes: officers in each state;
decriminalization appeared to significantly predict a higher prevalence of alcohol
consumption, also suggesting complementarity. The only other longitudinal study of youth
focused on the effects of Medical Marijuana Laws (MMLs) and showed that time-varying
state-level MML indicators were not significantly related to past 30-day alcohol use
(Anderson, Hansen, & Rees, 2012). Thus, results from studies with the strongest designs are
mixed, perhaps due to differences in samples and variables examined.
Cross-sectional studies of adolescents and young adults
The majority of studies on youth rely on cross-sectional surveys. In terms of aggregate-level
outcomes, the earliest published study of substitution occurred in response to Operation
Subbaraman Page 3
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Intercept, an anti-drug measure headed by President Nixon in May 1969. Operation
Intercept involved closing the Mexican/American border, which reduced the cannabis supply
in the US (McGlothlin, Jamison, & Rosenblatt, 1970). Among a sample of students and free
clinic patients who had used cannabis ≥ 10 times, 44–51% reported reduced cannabis
frequency as a result of cannabis unavailability between May and October 1969. Although
results rely on self-report, the vast majority of those reporting a cannabis shortage (76–84%)
reported increasing use of alcohol and other drugs because of the shortage (McGlothlin et
al., 1970).
Cross-national studies, on the other hand, appear to support complementarity. In a cross-
national comparison (USA, Canada, Netherlands) of 4,858 10th graders, alcohol and
cannabis laws were examined in relation to use. With USA as the referent (because it has the
strictest drug laws), relative risks of drinking were 1.30–2.0 for both Dutch and Canadian
boys and girls, though rates of cannabis use did not differ. Although it is impossible to tell if
laws preceded prevalence, the results show that alcohol use is higher among 10th graders in
countries with more liberal cannabis policies (Simons-Morton, Pickett, Boyce, ter Bogt, &
Vollebergh, 2010). Population-level analyses of adolescents from 35 European countries
(European School Project on Alcohol and Drugs/ESPAD) similarly indicated
complementarity: most cannabis users combined with alcohol at least once, and population-
level drinking and cannabis use correlated positively (Pape, Rossow, & Storvoll, 2009). Pape
and colleagues also analyzed individual-level data from a survey of 14–20 year-olds in
Norway (N=16,813), which showed a that 82% of cannabis users had used in combination
with alcohol, and 80% of cannabis use incidents involved alcohol (Pape et al., 2009). Thus,
cannabis appears to be used as a complement to alcohol in both cross-national studies.
Alcohol-related policies and substitution/complementarity among youth
Cross-sectional studies of beer taxes also show complementarity among youth. On the
aggregate level, findings from the 1993–1999 Harvard SPH College Alcohol Study surveys
(N=48,174) showed that higher beer taxes were related to lower alcohol and cannabis use,
and that the price of cannabis was negatively related to alcohol and cannabis use, though
decriminalization did not significantly affect either (Williams, Pacula, Chaloupka, &
Wechsler, 2004). On the individual-level, Farrelly and colleagues examined effects of state-
level cannabis fines/penalties among youth (age 12–20) and younger adults (age 21–30) in
the 1990–1996 National Household Survey of Drug Abuse (NHSDA); again, higher beer
prices led to decreased probability of cannabis use. Importantly, most effects disappeared
when including state effect indicators, implying possible endogeneity by unaccounted state
factors. Still, the authors concluded that increasing the price of alcohol would decrease
cannabis use among youths (Farrelly, Bray, Zarkin, Wendling, & Pacula, 1999). Importantly,
these studies depend on cross-sectional variation in beer taxes, which could be driven by
unobserved variables such as cultural attitudes or preferences within states. However, the
results are similar to what was found in Pacula’s longitudinal study (Pacula, 1998), which
suggests that the relationship between beer taxes and cannabis use among youth may be
robust. Furthermore, because youth generally have lower income levels than adults, the
reduced cannabis use associated with increased alcohol costs may reflect true economic
substitution, i.e., the limited financial capacity of younger respondents may influence their
Subbaraman Page 4
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
substance use patterns more than (or in addition to) the psychoactive associations between
alcohol and cannabis.
In terms of other alcohol-related policies, a number of analyses have examined the minimum
legal drinking age (MLDA) with mixed results. Using the 1982–1989 Monitoring the Future
(MTF) samples, DiNardo and Lemieux (2001) examined raising the MLDA from 18 to 21,
cannabis decriminalization and the price of alcohol on past 30-day cannabis and alcohol
consumption; increasing the MLDA from 18 to 21 decreased alcohol use by 4.5 percentage
points but increased the prevalence of cannabis use by 2.4 percentage points (DiNardo &
Lemieux, 2001). Based on these findings, the authors built an economic model of
consumption showing that the increase in cannabis use was attributable to standard
substitution. A separate analysis of the MLDA using the 2002–2007 National Survey on
Drug Use and Health (NSDUH) showed a sharp decrease in cannabis use at the age of 21:
the probability of past 30-day drinking increased from 60% to 70%, while the probability of
cannabis use decreased by two percentage points from a baseline of about 20% (Crost &
Guerrero, 2012). The opposing effects suggest substitution; instrumental variable models
showed that a 9.8 percentage point increase in the probability of alcohol consumption led to
a two percentage point decrease in cannabis use, with stronger effects for women. The
authors concluded that policies limiting alcohol access might increase cannabis consumption
in young adults, especially women. Importantly, the authors noted that the results’ external
validity best applies to alcohol policies aimed at individuals who are close to 21 years old
and likely to comply with regulations like the MLDA (Crost & Guerrero, 2012). These
findings contradicted an earlier study reported that turning 21 was associated with an
increase in cannabis use in most model specifications in the NLSY97, consistent with
complementarity (Yörük & Yörük, 2011). However, in a re-analyses of the NLSY97, Crost
and Rees applied the same regression discontinuity design and found no evidence of
complementarity (Crost & Rees, 2013), perhaps because Yörük (2011) restricted the sample
to respondents who had used cannabis at least once since last interviewed. Applying the
same design to all respondents, Crost and Rees found no significant changes in cannabis use
at age 21; thus current cannabis users may complement more than the sample as a whole,
which would be expected. Most recently, a study of fatal accidents among 16–25 year olds
(N=7,191), which also used a regression discontinuity approach, showed that the prevalence
of cannabis-related accidents did not change significantly at the MLDA of 21 (Keyes, Brady,
& Li, 2015). While alcohol use did increase, cannabis use did not change at the MLDA of
21, which corroborates what Yörük and Yörük (2011) and Crost and Guerrero (2012) had
found with similar methods. Still, taken as a set, results from studies of the MLDA are
mixed.
Cannabis-related policy and substitution/complementarity among youth
On the other hand, results from studies of youth focused on cannabis-related policy as the
independent variable support substitution. In the 1982–1989 MTF samples, simulating the
effects of uniform cannabis criminalization (e.g., cannabis use is a criminal offense in all US
states across time) versus uniform decriminalization (e.g., cannabis use is not a criminal
offense in any US states across time) suggested that cannabis could substitute for alcohol
(Chaloupka & Laixuthai, 1997): moving from total criminalization to total decriminalization
Subbaraman Page 5
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
would increase the number of alcohol abstainers in the past year by 12%, while reducing the
number of frequent drinkers by 11%. In addition, simulations using the pooled time-series of
state cross-sections (1975–1988) for the Fatality Analysis Reporting System, (FARS, an
annual census of fatal motor vehicle accidents in the 48 contiguous US states) demonstrated
that changing total criminalization to decriminalization would reduce the fatal accident rate
by 6% among 15–24 year olds. This is consistent with Model's (1993) findings that
substitution can result from reductions in the “full price” of cannabis due to
decriminalization because the full price includes monetary costs related to criminality, e.g.,
lost wages (Model, 1993). Importantly, the results suggest that substitution resulting from
decriminalization may yield overall reductions in drug and alcohol-related consequences,
such as accidents (Chaloupka & Laixuthai, 1997). Furthermore, the findings support those
found in MTF using aggregate-level outcomes (DiNardo & Lemieux, 2001).
Another study of hypothetical legalization (N=281) showed differential effects according to
beverage preference: consumption of spirits decreased most with hypothetical legalization.
The largest anticipated reduction in drinking was for the daily cannabis users, suggesting
that they would be most likely to substitute in legal environments (Clements & Daryal,
2005). Similarly, data from the 1979, 1984, and 1988 NLSY cohorts (N=12,686 14–21 year
olds) showed that the frequency of drinking 6+ drinks in an episode went down in states that
had decriminalized cannabis, again supporting substitution (Thies & Register, 1993).
Other studies of adolescents and young adults
Similarly, results potentially supporting substitution have been observed in a more recent
MTF subsample (N=11,542): students who reported no alcohol use were more likely to
report cannabis use (Alter, Lohrmann, & Greene, 2006). Although the study focused on
perceived access harms, the authors concluded that cannabis may substitute for alcohol
among individuals who choose to completely abstain from alcohol (Alter et al., 2006);
however, this conclusion may be over-reaching since substitution implies previous use of a
substance with subsequent changes in its usage patterns because of concurrent changes in
usage patterns of another substance
Finally, an 8-week trial of naltrexone for alcohol dependence in 18–25 year olds (N=122)
showed that cannabis use did not affect alcohol use, alcohol-related consequences, or
motivation to reduce drinking in bivariate tests. The authors did not examine alcohol
outcomes in multivariate regressions as the focus of the study was medication adherence
(Peters et al., 2012). Thus, the single clinical study is not particularly informative here.
Summary of studies of adolescents and young adults
Overall, more than half (9/17) of the studies among adolescents and young adults used
cannabis policy as the primary independent variable; six of these concluded that alcohol and
cannabis are substitutes, indicating that youth may use less alcohol in environments with
more liberal cannabis policies (Alter et al., 2006; Chaloupka & Laixuthai, 1997; Clements &
Daryal, 2005; Crost & Guerrero, 2012; DiNardo & Lemieux, 2001; McGlothlin et al., 1970;
Thies & Register, 1993). Of the six studies concluding that alcohol and cannabis are
complements, three were based on beer tax associations (Farrelly et al., 1999; Pacula, 1998;
Subbaraman Page 6
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Williams et al., 2004); this includes both longitudinal and cross-sectional studies that
examined both individual and aggregate-level outcomes, which suggests that higher beer
taxes are associated with less cannabis use among youth. Overall, alcohol and cannabis are
used as both substitutes and complements among youth and young adults, and policies
aimed at one substance appear to affect consumption of the other. Results are summarized in
Table 1.
General population
Similarly to the studies of youth, the majority of the general population studies examined
cannabis policy relative to substance use in order to assess substitution/complementarity.
Specifically, most studies used indicators of MMLs or cannabis decriminalization as the
independent variable; results have been mixed.
Longitudinal studies of the general population
Data from the NSDUH and Behavioral Risk Factor Surveillance System (BRFSS) showed
that MMLs were associated with a 1.51 reduction in mean number of drinks/month by males
and a 0.65 reduction in females, with the strongest effects among 20–29 year olds (who are
more likely to use medical cannabis than other age groups (Anderson & Rees, 2011)).
Controlling for state-level traffic and alcohol laws and using neighboring states as controls,
MMLs were associated with 8.7% decrease in the rate of fatal accidents, a 12% decrease in
any-blood alcohol content (BAC) crashes, a 14% decrease in high BAC crashes, and 19%
decrease in fatality among 20–29 year olds specifically in the 1990–2009 FARS data
(Anderson & Rees, 2011). Comparable analyses of all 19 states with MMLs in 2013
supported these conclusions: one year after MMLs were passed, traffic fatalities tended to
fall by 8–11% (Anderson, Hansen, & Rees, 2013).
Conversely, Salomonsen-Sautel and colleagues used 1994–2011 FARS data to examine
whether MMLs in Colorado affected the proportion of drivers in fatal crashes who were
alcohol-impaired (BAC ≥ 0.08%), and found no change in either Colorado or the 34 control
states (Salomonsen-Sautel, Min, Sakai, Thurstone, & Hopfer, 2014). The mixed results from
the FARS data may be due to differences in the subgroups examined, as well as potentially
inconsistent testing and reporting across states and times.
The most recent longitudinal general population study looked at the effects of the
depenalization of cannabis possession in Lambeth, London, England on drug-related
hospital admissions. Using > 1 million public hospital records and difference-in-difference
regressions, Kelly and Rasul (2014) found a significant reduction in alcohol-related
admissions post-depenalization for the youngest cohort (15–24 years old), suggesting that
cannabis could substitute for alcohol in this age group.
Cross-sectional studies of the general population
While 3/4 of the longitudinal studies support substitution, results from cross-sectional
general population surveys support complementarity. First, Saffer and colleagues found that
higher alcohol taxes decreased drug use in the 1988–1991 NHSDA; the relationship did not
differ across races (Saffer & Chaloupka, 1998). Saffer (1999) used the same dataset to
Subbaraman Page 7
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
examine effects of cannabis decriminalization on alcohol and other drug use;
decriminalization was associated with more alcohol use (complementarity) for the full
sample, white males and African-Americans (H. Saffer & F. J. Chaloupka, 1999). However,
for Native Americans and Hispanics, alcohol and cannabis appeared to be economic
substitutes. No effects were found for Asians, women or youth. The differences across
different ethnic, gender and age groups are especially important in light of the fact that they
are often ignored when general policies are being considered. Similarly, results from the
Australian National Drug Strategy Household Survey (NDSHS; N=9,744) showed that the
criminal status of cannabis did not appear to affect alcohol use (Cameron & Williams,
2001). An extension of the NDSHS study likewise concluded that cannabis
decriminalization did not affect alcohol use (Williams & Mahmoudi, 2004). The strongest
result from (Williams & Mahmoudi, 2004) was that cannabis use was inversely associated
with fines for exceeding the legal BAC, implying complementarity; notably, this finding
contradicts Cameron and Williams (2001), which concluded that higher alcohol prices were
positively associated with cannabis use, implying substitution. Data from the 2004–2011
NSDUH surveys support complementarity as well: a dichotomous MML indicator was
associated with a 6–9% increase in the frequency of binge-drinking among those 21 and
older (Wen, Hockenberry, & Cummings, 2014). MMLs did not affect drinking behavior
among those 12–20 years old; however, earlier MML changes (e.g., California) might have
been missed due to the time-period analyzed.
Summary of studies of the general population
Table 2 summarizes results from the general population. Eight out of nine general population
studies used indicators of cannabis policy as the independent variable; most notably, all five
cross-sectional studies using individual-level consumption as the dependent variable
concluded that cannabis and alcohol are complements while the longitudinal studies using
state-level dependent variables concluded that they are substitutes. Importantly, cross-
sectional data cannot adequately measure substitution/complementarity because substitution
and complementarity inherently require the passage of time. Thus the discrepancy between
individual- and aggregate-level results may be due to individuals who substitute over time
but report both cannabis and alcohol use within a single cross-sectional time period. In
addition, differential rates of substitution/complementarity may occur within subgroups;
some of these general population studies begin to identify subgroups that may be more likely
to substitute, such as Native Americans and Hispanics (H. Saffer & F. Chaloupka, 1999) as
well as subgroups more likely to complement, such as whites, African Americans, males,
and polysubstance users (H. Saffer & F. Chaloupka, 1999; Williams & Mahmoudi, 2004).
Clinical and community-based samples
General population studies of adults, which usually rely on cross-sectional or retrospective
reports, are well supplemented by clinical and community-based studies, which tend to use
prospective data (Table 3). In addition, clinical samples include the heaviest substance users
and have more detailed measures of substance use, which may improve study validity.
Subbaraman Page 8
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Cannabis treatment seekers
The first set of clinical studies focuses on cannabis treatment seekers. In a prospective cohort
(N=212) of individuals seeking treatment for cannabis, posttreatment increases in alcohol-
related problems were not associated with reduced cannabis use (Stephens, Roffman, &
Simpson, 1994). A similar study (N=291) showed significant increases in alcohol problems
at every follow-up, though there were no changes in alcohol use (Stephens, Roffman, &
Curtin, 2000). A study of 207 individuals wanting help quitting/reducing cannabis found that
73% increased drinking at the one-year follow-up. However, increased alcohol was not
related to decreased cannabis, and those with lower baseline drinking were more likely to
increase alcohol use than those with high baseline drinking (who were more likely to reduce
drinking). The authors concluded that substitution was not occurring and that any increases
in drinking could be attributed to regression to the mean (Kadden, Litt, Kabela- Cormier, &
Petry, 2009). Overall, within studies of cannabis treatment-seeking individuals who reduce
cannabis, neither alcohol substitution nor complementarity appears to occur.
Cannabis users not seeking treatment
In contrast, alcohol might substitute for cannabis among users who are not seeking
treatment. Among 104 non-treatment-seeking cannabis smokers who reported at least one
cannabis quit attempt without treatment, half reported increased tobacco, alcohol, and/or
sleeping aids to cope with cannabis withdrawal (Copersino et al., 2006), suggesting some
substitution, though very few initiated new substance use (N=6). The authors concluded that
quitting cannabis spontaneously may lead to increases in legal substance use; however, the
results rely on small sample and retrospective self-report, and analyses were deemed
exploratory. Still, a community-based study of non-treatment-seeking, DSM-IV dependent
cannabis users (N=45) similarly found that two-week cannabis abstinence was related to
increases in alcohol that decreased once cannabis was resumed, especially among those with
low baseline alcohol use; alcohol did not increase among those who remained abstinent
through the one-month follow-up (Allsop et al., 2014).
Natural history descriptions
Prospective "natural history descriptions" of attempts to abstain from cannabis have also
been used. Kouri and colleagues (2000) examined 28-day diary data from 30 users and 30
non/former cannabis users and found that cannabis abstinence was not related to alcohol,
tobacco, or caffeine use (Kouri & Pope, 2000). Among a smaller (N=12) sample of daily
users, alcohol use did not differ between using and abstinent study phases (Budney, Hughes,
Moore, & Novy, 2001). A longer 50-day natural history study (N=18 users abstaining, 12
ex-users) documented withdrawal symptoms, and alcohol, cigarette and caffeine
consumption: overall, alcohol use did not change (Budney, Moore, Vandrey, & Hughes,
2003). Although these results suggest that substitution is not occurring among cannabis
users who abstain, participants in these three studies were asked not to change alcohol use,
which substantially limits interpretation. In a natural history study of daily cannabis users in
Vermont (N=19), neither cannabis abstinence nor reduction were related to changes in
alcohol use (Hughes, Peters, Callas, Budney, & Livingston, 2008). Though participants did
Subbaraman Page 9
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
not receive instructions about alcohol consumption, the sample was small and consisted of
daily dependent users, which may also limit generalizability.
Similar one-month diaries completed by 28 daily cannabis users showed that those with past
alcohol use disorders (AUD) significantly increased alcohol use during cannabis abstinence
(52% increase), while those without past AUD did not (3% increase), suggesting that
individuals with AUD may be more likely to substitute (Peters, 2010). On the other hand,
among individuals with AUD in psychiatric treatment, cannabis use increased the hazard of
alcohol relapse and decreased the likelihood of stable abstinence post-discharge, suggesting
complementarity (Aharonovich et al., 2005). Thus, results from AUD samples are mixed.
Medical cannabis users
Substitution does appear common among medical cannabis users. In a cross-sectional survey
of 350 medical cannabis patients, 40% reported substituting cannabis for alcohol (Reiman,
2009); 65% reported “less adverse side effects” as the reason for substitution. A similar
Canadian study (N=404) stated that 75% of the sample claimed to have used cannabis as a
substitute for some other substance, while 41% used cannabis as substitute for alcohol
specifically (Lucas et al., 2013). Those who reported using cannabis as a substitute for
alcohol were significantly more likely (p < 0.05) to be male, to make between $40,000 and
$59,000 annually, to be current drinkers, and to report a history of alcohol and substance
abuse than those who did not report substituting, suggesting other potential modifiers.
Summary of studies of clinical and community-based samples
Table 3 summarizes results from the 13 studies using clinical or community samples.
Among cannabis users and treatment seekers, alcohol does not appear to substitute for
cannabis during times of cannabis abstinence. Results from studies of individuals with AUD
are inconclusive. Only cross-sectional studies of medical cannabis users support substitution,
though those studies are limited by possible selection and recall bias. Thus, there is no clear
pattern of substitution/complementarity among clinical and community samples.
Overall summary of findings
Of the 39 studies reviewed, 16 support substitution, ten support complementarity, 12 support
neither and one supports both. Findings from longitudinal studies of youth lean towards
complementarity while findings from general population studies support substitution. The
inconsistent conclusions may be explained by 1) the examination of heterogeneous
subgroups both across and within studies (e.g., youth vs. adults, heavy vs. light drinkers), 2)
variation in independent and dependent variables (e.g., any vs. binge drinking), and 3)
reliance on a binary model of cannabis and alcohol co-use (i.e., substitution vs.
complementarity), especially because co-use patterns are more complex. In reality,
“concurrently available reinforcers” (e.g., alcohol and cannabis) fall on a continuum such
that they can be substitutes, complements, or independent of one another to different people
at different times (Bickel, DeGrandpre, & Higgins, 1995) (Green & Freed, 1993). On one
end, “substitutable reinforcers” can be easily traded for one another due to similar
functionalities. At the opposite end “complementary reinforcers” are consumed together
proportionately, and therefore cannot be traded for one another (Green & Freed, 1993). The
Subbaraman Page 10
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
current results indicate that longitudinal studies of youth support complementarity while
general population studies support substitution; this further corroborates the notion that as
psychoactive reinforcers, alcohol and cannabis indeed fall on a substitution/complementarity
spectrum.
Alcohol and cannabis policies
Studies of MMLs and cannabis decriminalization suggest that adolescents and young adults
may reduce alcohol use in more liberal cannabis environments (substitute). Conversely,
studies of beer taxes showed that youth reduced cannabis use when taxes were raised,
suggesting complementarity in more stringent alcohol environments. Studies of the MLDA
were mixed; whether changing the MLDA affects cannabis use among youth remains an
open question.
In the general population, three of the four longitudinal studies concluded that alcohol and
cannabis are substitutes, while all five cross-sectional studies using individual-level
consumption as the dependent variable concluded that cannabis and alcohol are
complements. Notably, all of these studies used cannabis decriminalization or MML
indicators as the independent variable. As discussed above, cross-sectional data cannot
capture the passage of time and are therefore suboptimal for assessing substitution/
complementarity. The inconsistencies between the cross-sectional and longitudinal findings
may be explained by individuals who actually do substitute over time but report both
cannabis and alcohol use within a single cross-sectional time period. Thus, taken as a set,
findings from the general population suggest that substitution may occur in more liberal
cannabis environments.
Notable subgroups
Numerous individual and/or societal factors influence individual substance use and the
propensity to substitute/complement. The studies reviewed here suggest that cannabis-
related laws may affect alcohol use differently across genders and races (H. Saffer & F.
Chaloupka, 1999; Williams & Mahmoudi, 2004). For example, white males and African
Americans were found to complement in more liberal cannabis environments, while Native
Americans and Hispanics were found to substitute (H. Saffer & F. Chaloupka, 1999).
Polysubstance users and males were also found to complement under cannabis
decriminalization (Williams & Mahmoudi, 2004). Finally, substitution appeared to occur
among medical cannabis users.
Common study limitations
Many of the studies cited used dichotomous measures of cannabis laws as their primary
independent variable, which may overlook variations due to nuances, e.g., many non-
decriminalized states have conditional discharges for first offenders (Pacula, Powell, Heaton,
& Sevigny, 2013). Some studies may have under-sampled risky consumers as well. Within
the general population studies, early studies exclude the price of cannabis, which may lead
to omitted variable bias (Williams & Mahmoudi, 2004). Furthermore, the use of pooled
cross-sectional data in many of the studies reviewed here might mask heterogeneity of
effects over time and across subgroups. Finally, based on the search criteria, relevant studies
Subbaraman Page 11
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
that did not explicitly refer to “substitution” and/or “complementarity” may have been
missed; a broader review of the literature regarding alcohol/cannabis co-use would
supplement the current findings and possibly explain some of what was found here.
Implications for future research and practice
Event-level data are crucial to understanding how individuals substitute/complement alcohol
and cannabis; future studies should collect detailed data regarding alcohol and cannabis
quantity and frequency, as well as regarding the order in which the substances are consumed.
For example, individuals may substitute cannabis for alcohol if using cannabis first, but
complement the two if drinking first. Whether cannabis can substitute for alcohol among
individuals with AUD also remains an important open question (Subbaraman, 2014),
especially because AUD treatment programs often expel patients for testing positive for
other drugs; current and future studies are investigating whether cannabis use during AUD
treatment affects post-treatment alcohol outcomes. In terms of policy, future studies should
collect fine-grained longitudinal, prospective data from the general population and
subgroups of interest, especially in locations that are likely to legalize cannabis in the near
future. These data will help us understand how various groups along the co-use continuum
respond to policy changes, and allow us to continue identifying high-risk groups and
consequences associated with various co-use patterns. Understanding whether laws aimed at
a particular substance have spillover effects on other substance use will help us develop
optimal policies, while identifying groups associated with particular co-use patterns will
inform prevention and intervention strategies.
Conclusion
Alcohol and cannabis act as both substitutes and complements, and policies aimed at one
substance may inadvertently affect consumption of other substances. Results from studies of
youth suggest that youth may reduce alcohol in more liberal cannabis environments
(substitute), but reduce cannabis in more stringent alcohol environments (complement).
Results from the general population suggest that substitution of cannabis for alcohol may
occur under more lenient cannabis policies, though cannabis-related laws may affect alcohol
use differently across genders and racial groups. Policymakers should consider spillover
effects when crafting legislation.
REFERENCES
Aharonovich E, Liu X, Samet S, Nunes E, Waxman R, Hasin D. Postdischarge cannabis use and its
relationship to cocaine, alcohol, and heroin use: a prospective study. The American Journal of
Psychiarty. 2005; 162(8):1507–1514.
Allsop DJ, Dunlop AJ, Sadler C, Rivas GR, McGregor IS, Copeland J. Changes in cigarette and
alcohol use during cannabis abstinence. Drug and Alcohol Dependence. 2014; 138:54–60.
[PubMed: 24613633]
Alter RJ, Lohrmann DK, Greene R. Substitution of marijuana for alcohol: the role of perceived access
and harm. Journal of Drug Education. 2006; 36(4):335–355. [PubMed: 17533805]
Anderson DM, Hansen B, Rees DI. Medical marijuana laws, traffic fatalities, and alcohol
consumption. Journal of Law and Economics. 2013; 56(2):333–369.
Subbaraman Page 12
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Anderson, DM.; Rees, DI. Medical marijuana laws, traffic fatalities, and alcohol consumption. Bonn,
Germany: Forschungsinstitut zur Zukunft der Arbeit (Institute for the Study of Labor); 2011. [IZA
Discussion Paper No. 6112]
Angrist JD, Pischke J-S. The credibility revolution in empirical economics: how better research design
is taking the con out of econometrics. Journal of Economic Perspectives. 2010; 24(2):3–30.
Bickel WK, DeGrandpre RJ, Higgins ST. The behavioral economics of concurrent drug reinforcers: a
review and reanalysis of drug self-administration research. Psychopharmacology. 1995; 118(3):250–
259. [PubMed: 7617816]
Budney AJ, Hughes JR, Moore BA, Novy PL. Marijuana abstinence effects in marijuana smokers
maintained in their home environment. Archives of General Psychiatry. 2001; 58(10):917–924.
[PubMed: 11576029]
Budney AJ, Moore BA, Vandrey RG, Hughes JR. The time course and significance of cannabis
withdrawal. Journal of Abnormal Psychology. 2003; 112(3):393–402. [PubMed: 12943018]
Cameron L, Williams J. Cannabis, alcohol and cigarettes: substitutes or compliments? The Economic
Record. 2001; 77(236):19–34.
Chaloupka FJ, Laixuthai A. Do youths substitute alcohol and marijuana? Some econometric evidence.
Eastern Economic Journal. 1997; 23(3):253–276.
Clements KW, Daryal M. Exogenous shocks and related goods: drinking and legalisation of marijuana.
Economics Letters. 2005; 89:101–106.
Copersino ML, Boyd SJ, Tashkin DP, Huestis MA, Heishman SJ, Dermand JC, Gorelick DA. Cannabis
withdrawal among non-treatment-seeking adult cannabis users. The American Journal on
Addictions. 2006; 15(1):8–14. [PubMed: 16449088]
Crost B, Guerrero S. The effect of alcohol availability on marijuana use: evidence from the minimum
legal drinking age. Journal of Health Economics. 2012; 31(1):112–121. [PubMed: 22381404]
Crost B, Rees DI. The minimum legal drinking age and marijuana use: new estimates from the
NLSY97. Journal of Health Economics. 2013; 32(2):474–476. [PubMed: 23199608]
Degenhardt L, Chiu W-T, Sampson N, Kessler RC, Anthony JC, Angermeyer M, Wells JE. Toward a
global view of alcohol, tobacco, cannabis, and cocaine use: findings from the WHO World Mental
Health Surveys. PLoS Medicine. 2008; 5(7):e141. [PubMed: 18597549]
DiNardo J, Lemieux T. Alcohol, marijuana, and American youth: the unintended consequences of
government regulation. Journal of Health Economics. 2001; 20(6):991–1010. [PubMed:
11758056]
Farrelly, MC.; Bray, JW.; Zarkin, GA.; Wendling, BW.; Pacula, RL. Cambridge, MA: National Bureau
of Economic Research; 1999. The effects of prices and policies on the demand for marijuana:
evidence from the National Household Surveys in Drug Abuse; p. 22[NBER Working Paper No.
6940]. Archived by WebCite® at http://www.webcitation.org/6XLlbgezE [Accessed: 2015-03-27]
Green L, Freed DE. The substitutability of reinforcers. Journal of the Experimental Analysis of
Behavior. 1993; 60(1):141–158. [PubMed: 16812696]
Hughes JR, Peters EN, Callas PW, Budney AJ, Livingston AE. Attempts to stop or reduce marijuana
use in non-treatment seekers. Drug and Alcohol Dependence. 2008; 97(1–2):180–184. [PubMed:
18485619]
Hunt GP, Evans K, Kares F. Drug use and meanings of risk and pleasure. Journal of Youth Studies.
2007; 10(1):73–96.
Kadden RM, Litt MD, Kabela- Cormier E, Petry NM. Increased drinking in a trial of treatments for
marijuana dependence: substance substitution? Drug and Alcohol Dependence. 2009; 105:168–
171. [PubMed: 19608353]
Keyes KM, Brady JE, Li G. Effects of minimum legal drinking age on alcohol and marijuana use:
evidence from toxicological testing data for fatally injured drivers aged 16 to 25 years. Injury
Epidemiology. 2015; 2:1. [PubMed: 26301177]
Kouri EM, Pope HG Jr. Abstinence symptoms during withdrawal from chronic marijuana use.
Experimental and Clinical Psychopharmacology. 2000; 8(4):483–492. [PubMed: 11127420]
Lucas P, Reiman A, Earlywine M, McGowan SK, Oleson M, Coward MP, Thomas B. Cannabis as a
substitute for alcohol and other drugs: a dispensary-based survey of substitution effect in Canadian
medical cannabis patients. Addiction Research and Theory. 2013; 21(5):435–442.
Subbaraman Page 13
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
McGlothlin W, Jamison K, Rosenblatt S. Marijuana and the use of other drugs. Nature. 1970;
228:1227–1229. [PubMed: 5487261]
Model KE. The effect of marijuana decriminalization on hospital emergency room drug episodes:
1975–1978. Journal of the American Statistical Association. 1993; 88(423):737–747.
Pacula RL. Does increasing the beer tax reduce marijuana consumption? Journal of Health Economics.
1998; 17(5):557–585. [PubMed: 10185512]
Pacula, RL.; Powell, D.; Heaton, P.; Sevigny, EL. Cambridge, MA: National Bureau of Economic
Research; 2013. Assessing the effects of medical marijuana laws on marijuana and alcohol and
alcohol use: the devil is in the details. Working Paper No. 19302. Archived by WebCite® at http://
www.webcitation.org/6JjNOpULy [Accessed: 2013-09-18]
Pape H, Rossow I, Storvoll EE. Under double influence: Assessment of simultaneous alcohol and
cannabis use in general youth populations. Drug and Alcohol Dependence. 2009; 101(1–2):69–73.
[PubMed: 19095380]
Peters EN, Leeman RF, Fucito LM, Toll BA, Corbin WR, O'Malley SS. Co-occuring marijuana use is
associated with medication nonadherence and nonplanning impulsivity in young adult heavy
drinkers. Addictive Behaviors. 2012; 37(4):420–426. [PubMed: 22189052]
Reiman A. Cannabis as a substitute for alcohol and other drugs. Harm Reduction Journal. 2009; 6(35)
online journal.
Ronen A, Chassidim HS, Gershon P, Parmet Y, Rabinovich A, Bar-Hamburger R, Shinar D. The effect
of alcohol, THC and their combination on perceived effects, willingness to drive and performance
of driving and non-driving tasks. Accident Analysis and Prevention. 2010; 42(6):1855–1865.
[PubMed: 20728636]
Room R. Stigma, social inequality and alcohol and drug use. Drug and Alcohol Review. 2005; 24(2):
143–155. [PubMed: 16076584]
Saffer, H.; Chaloupka, F. Cambridge, MA: National Bureau of Economic Research; 1998.
Demographic differentials in the demand for alcohol and illicit drugs. [NBER Working Paper No.
6432]
Saffer H, Chaloupka F. The demand for illicit drugs. Economic Inquiry. 1999; 37(3):401–411.
Saffer, H.; Chaloupka, FJ. Demographic differentials in the demand for alcohol and illicit drugs. In:
Chaloupka, FJ.; Grossman, M.; Bickel, WK.; Saffer, H., editors. The Economic Analysis of
Substance Use and Abuse: An integration of econometric and behavioral economic research.
Chicago, IL: University of Chicago Press; 1999. p. 187-212.
Salomonsen-Sautel S, Min S-J, Sakai JT, Thurstone C, Hopfer C. Trends in fatal motor vehicle crashes
before and after marijuana commercialization in Colorado. Drug and Alcohol Dependence. 2014;
140:137–144. [PubMed: 24831752]
Simons-Morton B, Pickett W, Boyce W, ter Bogt TFM, Vollebergh W. Cross-national comparison of
adolescent drinking and cannabis use in the United States, Canada, and the Netherlands. The
International Journal of Drug Policy. 2010; 21(1):64–69. [PubMed: 19303761]
Stephens RS, Roffman RA, Curtin L. Comparison of extended versus brief treatments for marijuana
use. Journal of Consulting and Clinical Psychology. 2000; 68(5):898–908. [PubMed: 11068976]
Stephens RS, Roffman RA, Simpson EE. Treating adult marijuana dependence: a test of the relapse
prevention model. Journal of Consulting and Clinical Psychology. 1994; 62(1):92–99. [PubMed:
8034835]
Subbaraman MS. Can cannabis be considered a substitute medication for alcohol? Alcohol and
Alcoholism. 2014; 49(3):292–298. [PubMed: 24402247]
Substance Abuse and Mental Health Services Administration. Rockville, MD: 2012. Results from the
2011 National Survey on Drug Use and Health: Summary of National Findings NSDUH Series
H-44, HHS Publication No. (SMA) 12-4713; p. 161Archived by WebCite® at http://
www.webcitation.org/6EMWVKtrn [Accessed: 2013-02-11]
Thies CF, Register CA. Decriminalization of marijuana and the demand for alcohol, marijuana and
cocaine. The Social Science Journal. 1993; 30(4):385–399.
Thompson, Reuters. Philadelphia, PA: 2016. Web of Science: The complete citation connection.
Archived by WebCite® at http://www.webcitation.org/6eLUhGXAI [Accessed: 2016-01-06]
Subbaraman Page 14
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Wen, H.; Hockenberry, JM.; Cummings, JR. Cambridge, MA: National Bureau of Economic Research;
2014. The effect of medical marijuana laws on marijuana, alcohol, and hard drug use. [NBER
Working Paper No. 20085]
Williams J, Mahmoudi P. Economic relationship between alcohol and cannabis revisited. The
Economic Record. 2004; 80(248):36–48.
Williams J, Pacula RL, Chaloupka FJ, Wechsler H. Alcohol and marijuana use among college students:
economic complements or substitutes? Health Economics. 2004; 13(9):825–243. [PubMed:
15362176]
Yörük BK, Yörük CE. The impact of minimum legal drinking age laws on alcohol consumption,
smoking, and marijuana use: evidence from a regression discontinuity design using exact date of
birth. Journal of Health Economics. 2011; 30(4):740–752. [PubMed: 21719131]
Subbaraman Page 15
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Subbaraman Page 16
Table 1
Summary of studies examining cannabis/alcohol substitution and/or complementarity among adolescents and young adults
Study Sample
size Study
design Population Independent
variable(s) Dependent
variable(s) Method Control
variable(s) Conclusion Validity
concerns
Longitudinal Studies
Pacula, 1998 8,008 Longitudinal
panel National
Longitudinal
Survey of Youth
(1979)
State-level beer tax Individual-level past
30-day # drinks, #
times used cannabis
Probit & OLS
regression Personal & family
characteristics Complements Collinearity of
crime/ officer
ratio and state
decriminalisation
Anderson, 2012
Varies by
dataset
Pre/post Youth Risk
Behavior
Surveillance
System, National
Longitudinal
Survey of Youth
(1997) Treatment
Episode Data Set
State-level MML
indicator Individual-level past
30-day cannabis &
alcohol use
a
&
frequency; State-
level cannabis-
related treatment
admissions
Fixed effects
regression Individual & state-
level demo’s;
state-level taxes,
BAC
b
law
Neither Inconsistencies
across states in
YRBSS
Cross-sectional Studies (Aggregate-level Outcomes)
McGlothlin, 1970 594 Cross-
sectional
survey
College students,
free clinic patients Occurrence of
Operation Intercept
(1969)
Prevalence of self-
reported substitution
of alcohol for
cannabis
Crude
prevalence None Substitutes Subject-specific
definition of
“substitution,” no
comparison
group
DiNardo, 2001 >156,000 Cross-
sectional
survey
Monitoring the
Future MLDA
b
, state-level
cannabis
decriminalisation
indicator, alcohol
price
State-level
prevalence of past
30-day cannabis &
alcohol frequency
Random
effects
regression
State-year demo’s
& policies Substitutes No agreement on
controls, MTF
excludes non-HS
grads
Crost, 2012 71 (state-
level obs) Cross-
sectional
survey
National Survey on
Drug Use and
Health
Turning MLDA of
21 Population-level past
month cannabis &
alcohol frequency
Regression
discontinuity,
instrumental
variables (IV)
None Substitutes Age might not be
precise; IV
estimates apply
to 21 year olds
likely to comply
with laws like
MLDA
Simons-Morton, 2010 4,858 Cross-
sectional
survey
Health Behavior in
School Age
Children (15 year
olds from USA,
Canada,
Netherlands)
Restrictiveness of
country-level
alcohol & cannabis
policies
Country-level alcohol
frequency, cannabis
use
Crude
prevalence
comparison
None Complements Cross-sectional;
Potential
endogeneity
(laws could
precede
prevalence)
Pape, 2009 16,813 Cross-
sectional
survey
European School
Project on Alcohol None Proportion of
cannabis/alcohol use
occasions
Crude
prevalence
comparison
None Complements Cross-sectional
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Subbaraman Page 17
Study Sample
size Study
design Population Independent
variable(s) Dependent
variable(s) Method Control
variable(s) Conclusion Validity
concerns
& Drugs
Williams, 2004a 48,174 Cross-
sectional
survey
Harvard School of
Public Health
College Alcohol
Study
State-level
cannabis
decriminalisation
indicator; cannabis
price
State-level past
month & past year
alcohol & cannabis
frequency
Fixed effects
regression Individual-level
demo’s; state-level
prices & policies;
college-level fines,
penalties, miles
from DEA
Complements Measurement
error in cannabis
price (one of
main IVs)
Cross-sectional Studies (Individual-level outcomes)
Farrelly, 1999 49,311 Cross-
sectional
survey
National
Household Survey
on Drug Abuse
(12–20 year olds)
State-level
cannabis-related
penalties, county-
level # cannabis
arrests
Individual-level past
month cannabis use Probit
regression Individual-level
demo’s, state-level
cigarette & alcohol
prices
Complements Effects disappear
when including
state-level
indicators
Chaloupka, 1997 25,430 Cross-
sectional
survey
Monitoring the
Future State-level
cannabis
decriminalisation
indicator
Individual-level
alcohol frequency,
heavy drinking,
accidents
Probit
regression Individual-level
demo’s, alcohol
prices, MLDA,
cannabis price
Substitutes Measurement
error in prices,
accidents
Clements, 2005 281 Cross-
sectional
survey
First-year college
economics
students
Hypothetical
cannabis
legalization
Individual-level
beverage-specific
alcohol use
b
Regression Sex, intensity of
cannabis use Substitutes Hypothetical;
price elasiticities
unclear
Thies, 1993 12,686 Cross-
sectional
survey
National
Longitudinal
Survey of Youth
(1979)
State-level
cannabis
decriminalisation
indicator, MLDA,
arrests
Individual-level
alcohol & cannabis
use & frequency
Multiple
regression Individual-level
demo’s Weak
substitutes Possible
measurement in
error in drug use
frequency
Alter, 2006 11,542 Cross-
sectional
survey
Monitoring the
Future:
Midwest
subsample
Perceived access
& harms of alcohol
& cannabis
Individual-level past
month alcohol &
cannabis frequency
Multiple
regression Individual-level
demo’s,
academics
Substitutes Pooled cross-
sections;
variation in
surveys
Yörük, 2011 ~9,000 Cross-
sectional
panel survey
National
Longitudinal
Survey of Youth
(1997)
Turning MLDA of
21 Individual-level past
30-day alcohol &
cannabis use &
frequency
Regression
discontinuity Individual-level
demo’s Complements Only included
respondents who
had used
cannabis at least
once since last
interviewed
Keyes, 2015 7,191 Pooled cross-
sections 16–25 years olds
from the 1999–
2011 Fatality
Analysis Reporting
System
Turning MLDA of
21 Individual-level fatal
injury due to alcohol
use
Regression
discontinuity;
Joinpoint
permutation
Sex, race/ethnicity,
# vehicle
occupants, #
deaths in incident,
year, state,
whether state has
MML
Neither Missing data,
inconsistent
reporting/testing
across states,
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Subbaraman Page 18
Study Sample
size Study
design Population Independent
variable(s) Dependent
variable(s) Method Control
variable(s) Conclusion Validity
concerns
Crost, 2013 28,089 Cross-
sectional
panel survey
National
Longitudinal
Survey of Youth
(1997)
Turning MLDA of
21 Individual-level past
30-day cannabis use
& frequency
Regression
discontinuity Individual-level
demo’s Neither Age might not be
precise (exact
birthdays are
unknown)
Other Relevant Studies
Peters, 2012 122
(18–25
year olds)
RCT 8-week
randomized control
trial for naltrexone
for alcohol use
Cannabis user Individual-level
alcohol use & related
problems
MANCOVA None Neither Focus was not on
alcohol, did not
examine in
regression
a
Use is dichotomous yes/no
b
BAC = Blood Alcohol Content
c
MLDA = Minimum legal drinking age
d
MML = Medical marijuana law
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Subbaraman Page 19
Table 2
Summary of studies examining cannabis/alcohol substitution and/or complementarity among the general population
Study Sample
size Study
design Population Independent
variable(s) Dependent
variable(s) Method Control
variable(s) Conclusion Validity
concerns
Longitudinal Studies
Anderson, 2011
Varies by
dataset
Pre/post National Survey
on Drug Use and
Health, Fatality
Analysis
Reporting System,
BRFSS
State-level MML &
cannabis
decriminalisation
indicators
State-level
alcohol-related
traffic fatality
rate;
monthly # drinks,
prevalence of
bingeing
Difference-in-
difference;
fixed effects
regression
State-level BAC
laws, demo’s,
driving
characteristics
Substitutes Ignores
nuances of laws
(as do most
studies); Does
not look at traffic
fatalities
involving
cannabis
Anderson, 2013
Varies by
dataset;
Expands
(Anderson,
2011)
to
19 states
with MMLs
Pre/post National Survey
on Drug Use and
Health, Fatality
Analysis
Reporting System,
BRFSS
State-level MML &
cannabis
decriminalisation
indicators
State-level
alcohol-related
traffic fatality
rate;
monthly # drinks,
prevalence of
bingeing
Difference-in-
difference;
fixed effects
regression
State-level BAC
laws, demo’s,
driving
characteristics
Substitutes Ignores
nuances of laws
(as do most
studies)
Salomonsen-Sautel, 2014 36 six-
month
intervals of
motor
vehicle
fatalities
Pre/post Fatality Analysis
Reporting System Indicator of
commercial
availability of
medical cannabis in
Colorado (2009)
State-level
proportion of
drivers in fatal
crashes who were
alcohol-impaired
or cannabis-
positive
Generalized
least squares
with AR(1)
34 states without
MMLs through
2011; Proportion
of male drivers,
proportion of
drivers age 21–
24, proportion
tested for
drugs/alcohol
Neither Inconsistent
testing and
reporting across
states and time
Kelly, 2014 <1 million
public
hospital
admissions
Pre/post Quarterly hospital
admissions from
1997–2009 in
London, England
Depenalization of
cannabis
possession in
Lambeth, London,
England
Hospitalizations
related to alcohol
use
Difference-in-
difference
regressions
Admission age
and gender;
models also
include time and
neighborhood
fixed effects
Substitutes Changes in ICD
coding over
time; potential
selection bias
due to use of
hospitalizations
Cross-sectional Studies
Saffer, 1998 49,082 Cross-
sectional
survey
National
Household Survey
on Drug Abuse
State-level alcohol
& drug prices Individual-level
alcohol frequency,
cannabis use
b
Regression Individual-level
demo’s, cost of
living, state-level
cannabis
decriminalisation
indicator
Complements Measurement
error in prices
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Subbaraman Page 20
Study Sample
size Study
design Population Independent
variable(s) Dependent
variable(s) Method Control
variable(s) Conclusion Validity
concerns
Saffer, 1999 49,082 Cross-
sectional
survey
National
Household Survey
on Drug Abuse
State-level
cannabis
decriminalisation
indicator
Individual-level
alcohol frequency,
cannabis use
Regression Individual-level
demo’s, cost of
living, state-level
alcohol & drug
prices
Complements
for full sample,
white males &
blacks;
substitutes for
Native Am’s &
Hispanics
Under-sampled
risky consumers
(a potential
concern of
numerous
studies); pooled
cross-sections
Cameron, 2001 9,744 Cross-
sectional
survey
National Drug
Strategy
Household Survey
(Australia)
State-level
cannabis
decriminalisation
indicator
Individual-level
past 12-month
alcohol &
cannabis use
Probit
regression Individual-level
demo’s; State-
level income,
alcohol,
cannabis,
cigarette prices
Both:
decriminalisation
increases
alcohol use
suggesting
complements;
price effects
suggest
substitution
Only
decriminalized
state is South
Australia, so
decriminalisation
effect might
really
be SA effect
Williams, 2004b 15,479
Expands
(Cameron,
2001)
to
polysubstance
users & adds
control for
drunk driving
Cross-
sectional
survey
National Drug
Strategy
Household Survey
(Australia)
State-level
cannabis
decriminalisation
indicator, fines for
drunk driving, price
of cannabis &
alcohol
Individual-level
past year
cannabis use
Probit
regression Individual-level
demo’s Complements,
especially
among
polysubstance
users & males
Results depend
on inclusion of
year effects
Wen, 2014 Not stated Cross-
sectional
survey
National Survey
on Drug Use and
Health
State-level MML
b
indicator Individual-level
alcohol, cannabis,
& other drug use,
binge-drinking
Fixed effects
regression Individual &
state-level
demo’s
Weak
complements
among those
older than 21
Short time-frame
(2004–2011)
might miss
effects of earlier
MMLs (e.g., CA)
a
MML = Medical marijuana law
b
Use is dichotomous yes/no
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Subbaraman Page 21
Table 3
Summary of studies examining cannabis/alcohol substitution and/or complementarity among adult clinical and community samples
Study Sample size Study design Population Independent
variable(s) Dependent
variable(s) Method Control
variable(s) Conclusion Validity concerns
Longitudinal Studies—Cannabis Treatment Seekers
Stephens, 1994 212 Prospective
cohort Cannabis
treatment-
seekers
Weekly alcohol
use Past 90-days
cannabis frequency Bivariate
MANOVA,
correlations
None Neither Focus is cannabis
treatment
outcomes;
generalizability
Stephens, 2000 291 Prospective
cohort Cannabis
treatment-
seekers
Past 90-day
alcohol frequency Monthly cannabis
frequency Partial
correlations Corresponding
pre-treatment
use measures
Weak
complements Cross-sectional;
Generalizability
Kadden, 2009 207 Prospective
cohort Cannabis
treatment-
seekers
Past 90-day
cannabis
frequency
Indicator of increased
(≥10%) past 90-day
proportion days using
alcohol, drinks per
drinking day
Logistic
regression Individual-level
demo’s,
psychiatric
status,
substance use
severity
Neither Cross-sectional;
Generalizability
Longitudinal Studies—Cannabis Users/Non-Treatment Seekers
Copersino, 2006 104 Prospective
cohort Non-treatment-
seeking
cannabis users
None Self-reported use of
alcohol to relieve
cannabis withdrawal
during quit attempt
Crude
prevalence None Some evidence
of substitution to
relieve
withdrawal
No comparison
group
Allsop, 2014 45 Prospective
community-
based cohort
Non-treatment-
seeking
cannabis users
Cannabis
abstinence Alcohol use, quantity Mixed models
for repeated
measures with
AR(1)
Individual-level
demo’s, past
use
Substitutes,
especially
among baseline
light drinkers
Generalizability
Longitudinal “Natural History” Studies
Kouri, 2000 60 Prospective
cohort Current & former
cannabis users Cannabis
withdrawal
symptoms (e.g.,
cravings)
Changes in alcohol
use (not clear) Within-subject
ANOVA None Neither Participants asked
not to drink more
than 2 drinks/day
during study
Budney, 2001 12 Prospective
cohort Daily cannabis
users Cannabis
abstinence Alcohol use Within-subject
ANOVA None Neither Participants asked
not to change
drinking; Small
sample
Budney, 2003 30 Prospective
cohort 18 cannabis
users, 12 ex-
users in parallel
Cannabis
abstinence Alcohol use Within-subject
ANOVA None Neither Participants asked
not to change
drinking; Small
sample;
Generalizability
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Subbaraman Page 22
Study Sample size Study design Population Independent
variable(s) Dependent
variable(s) Method Control
variable(s) Conclusion Validity concerns
Hughes, 2008 19 Prospective
cohort Daily cannabis
users trying to
stop or reduce
cannabis on
their own
Cannabis
abstinence Alcohol use Bivariate tests,
Wilcoxon rank
test
None Neither Small sample;
Generalizability
Longitudinal Studies—Individuals with Alcohol Use Disorders
Peters, 2010 28 Prospective
cohort Daily cannabis
users not trying
to stop or reduce
cannabis
Cannabis
abstinence Drinks/day Within-subject
ANOVA Individual-level
demo’s, past
use
Substitution
among those
with past AUD
Small sample;
Generalizability
Aharonovich, 2005 250 Prospective
cohort Inpatient
psychiatric/
substance
abuse treatment
Postdischarge
cannabis use Alcohol use post-
treatment Survival
analysis Individual-level
demo’s,
psychiatric
status, severity
Complements Generalizability
Cross-sectional Studies—Medical Cannabis Users
Reiman, 2009 350 Cross-sectional
survey Medical
cannabis
patients
None Self-reported
substitution of
cannabis for alcohol
Descriptive None Substitutes Subject-specific
definition of
substitution,
medical cannabis
patients may not
generalize
Lucas, 2012 404 Cross-sectional
survey Medical
cannabis
patients
None Self-reported
substitution of
cannabis for alcohol
Descriptive None Substitutes Subject-specific
definition of
substitution,
medical cannabis
patients may not
generalize
a
MML = Medical marijuana law
b
Use is dichotomous yes/no
Subst Use Misuse
. Author manuscript; available in PMC 2017 September 18.
