SlideShare a Scribd company logo
1 of 29
Download to read offline
1
Department of Economics
University of Warwick
Research in Applied Economics EC331
2015 - 2016
1304164
Price or Preference? The economic relationship
between alcohol and marijuana
A study of American Universities from 2002-2014
Word Count: 4,833
Abstract
This article investigates the relationship between marijuana prices and the use of marijuana
and alcohol for American university students aged 18 - 29 in the National Survey on Drug
Use and Health data from 2002 to 2014. Previous studies yield mixed evidence regarding the
economic relationship and have preliminarily utilised proxies to account for pricing effects. I
implement an instrumental variable specification that identifies drug use with variables that
are empirically unrelated to consumption of marijuana and alcohol. Results indicate that
marijuana and alcohol are gross substitutes. Alterations to the model through legal drinking
age change this relationship. The analysis is enriched by observing changing desires to
consume marijuana and alcohol for different age groups and those in different years of
university. Exogeneity tests reveal that the standard Probit estimates for alcohol
consumption are severely biased towards one.
**Acknowledgements: I would like to thank Dr. Robert Akerlof for his immense support and
guidance as a supervisor, Dr. Rocco d’Este, Dr. Jeremy Smith and Dr. Piotr Jelonek for their
insight on empirical techniques, SAMHSA for the data and Ms. Anuria Singh, Mr. Hameem
Raees Chowdhury, Ms. Sophia Karanicholas and Ms. Harshini Singh for their comments and
encouragement. Any remaining errors and omission are my own.
2
Table of Contents
ABSTRACT...........................................................................................................................................1
1. INTRODUCTION ........................................................................................................................3
2. LITERATURE AND THEORETICAL FRAMEWORK..........................................................4
2.1 RELATED LITERATURE ........................................................................................................................4
2.2 THEORETICAL FRAMEWORK...............................................................................................................5
3. DATA AND SUMMARY STATISTICS.....................................................................................6
3.1 DATASET ................................................................................................................................................6
3.2 VARIABLES.............................................................................................................................................7
3.2.1 Consumption likelihood measures.......................................................................................... 7
3.2.2 Pricing measure............................................................................................................................... 7
3.2.3 Control Variables............................................................................................................................ 8
3.2.4 Correlations among variables .................................................................................................. 8
4. RESEARCH METHODOLOGY..................................................................................................9
4.1 MAIN MODEL AND ESTIMATION METHOD.........................................................................................9
4.2 ENDOGENEITY.......................................................................................................................................9
5. RESULTS ....................................................................................................................................10
5.1 ARE MARIJUANA AND ALCOHOL ECONOMIC COMPLEMENTS OR SUBSTITUTES?.......................... 10
5.2 DO AGE AND UNIVERSITY YEAR AFFECT DESIRE TO CONSUME SUBSTANCES?.............................. 11
5.3 ARE UNIVERSITY STUDENTS OF LEGAL DRINKING AGE MORE LIKELY TO CONSUME ALCOHOL? 11
6. ROBUSTNESS CHECKS...............................................................................................................15
7. DISCUSSION AND CONCLUDING REMARKS .......................................................................15
8. APPENDIX .....................................................................................................................................17
8.1 APPENDIX I: TABLES ............................................................................................................................. 17
8.2 APPENDIX II: FIGURES .......................................................................................................................... 24
9. REFERENCES ................................................................................................................................25
3
1. Introduction
Substance abuse involving young adults has been an evolving phenomenon characterised by
growing affluent nations, requiring frequent and structured reassessment. Marijuana is
being seen as less and less of a 'problem drug' and more like a stress relaxant, advocated by
a rising number of young adults who have a strong impact on university students.1
There is a
common perception that marijuana (whose accessibility is prominent despite its illegal
status) is healthier and a less damaging option than alcohol. Due to the very nature of
marijuana, combining it with alcohol is hazardous which makes it interesting to note why
past research has led to viewing the pair as economic complements.2
Luthar (2003) provided an interesting outlook when she identified that inner-city students
showed more frequent usage of marijuana and alcohol. Material wealth, in her opinion, had
long lasting implications on the culture and psychological costs imparted onto children from
affluent backgrounds. Yet prior research has yet to look at how affluence affects preferences
towards the pair. There may be an underlying factor that promotes a psychological view that
marijuana is deemed a higher social class means of intoxication with alcohol being viewed as
the layman's substance (Thies and Register, 1993).
This study investigates the economic relationship between two commonly used substances
in American universities - alcohol and marijuana. By assessing own and cross-price effects, it
is possible to identify whether the goods are economic complements or substitutes.
Theoretical views present two alternating hypothesis. Marijuana and alcohol can be viewed
as either complements or substitutes, and in both cases, support standard consumer utility
maximisation.3
Empirical results from variations of a standard Probit model used to model
desire to consume either marijuana or alcohol are applied to test the theoretical hypothesis.
There is a particular focus as to whether an individual is of legal drinking age (LDA) affects
this relationship. There is reason to believe that the pair of substances are interrelated
(Cameron and Williams, 2001). Cannabis and alcohol consumption provide the similar initial
euphoric effects, albeit with differing end results.
Results from the analysis suggest that marijuana and alcohol are gross substitutes. However
variations to the model - including testing the individual's legal status suggest that for higher
marijuana prices the desire to consume alcohol, while positive, declines relative to previous
prices. Factors such as gender and geographic location are insignificant in contrast to prior
work. From a policymaker's perspective, this paper provides key insight into consumption
patterns at university and is a stepping-stone towards drug policy development. For instance
if marijuana and alcohol are in fact economic complements, a nation wide per-unit tax on
marijuana will dampen binge-drinking at universities, a high-priority problem to counter in
recent years (Hingson et al., 2009).
1
See Amonini (2005) for how youth perception influences substance use
2
There may be a time lag between consumption of alcohol and marijuana that would be
driving this relationship.
3
See section 2.2 for Theoretical Framework on utility maximisation
4
2. Literature and Theoretical Framework
2.1 Related Literature
The premise that alcohol and marijuana may satisfy similar consumption desires and as a
result, the restriction of one, leads to the increase in consumption of the other is not new.
Post the Volstead Act of 19204
, the prohibition of alcohol led to the first signs of increased
recreational use of marijuana5
. DiNardo and Lemieux (1992) were the first to research
whether youth substitute alcohol for marijuana6
. They used utility maximisation constraints
to analyse the effects of an increase in minimum drinking age on alcohol and drug
consumption and whether decriminalising marijuana had any effect on the two. They
speculated how the presence of legal restraints at state level would deter marijuana or
alcohol consumption and amplify the substitution relationship. The results advocated that
an increase in minimum drinking age led to an increase in the use of marijuana among
youth, and decriminalising marijuana led to significant declines in alcohol consumption.7
From these two results, they concluded that alcohol and marijuana were in fact economic
substitutes.
Alternatively, Chaloupka and Laixuthai (1994) examined the effects of pricing on alcohol
substitutability with marijuana. They added information from the American Chamber of
Commerce Researchers Association (ACCRA) to examine full and part pricing effects of
alcohol in tandem with the legalisation of marijuana. They accounted for this relationship by
monitoring traffic related fatalities to proxy for substance abuse and induced policy changes8
to identify the substitution effect between the two goods. There was an inherent thought of
how the price effect would create a substitution-based relationship. However, the analysis
was lacking in areas with regards to money measurement problems, where it became
difficult to place an approximate value on marijuana. Later research done by Saffer &
Chaloupka (1999) negated this view.
Similarly, Pacula (1998), worked with pricing effects to examine whether rising beer taxes
would have an impact on marijuana. She used individual demand level equations and a static
(censoring) model to better test the economic relationship along with using data from the
National Longitudinal Survey of Youth. Her analysis revealed that an increase in the federal
tax on beer was found to have a larger unconditional decline in marijuana than alcohol,
making the pair complements.9
Her view was built upon how exogenous price effects would
lead to an alternative outcome to past results. The analysis extended to include a racial
4 First infringed in Chicago although implemented across the 50 states.
5
The evidence for this came from the sudden use of marijuana ‘tea bags’ in New York City.
6
Utilising large sample survey data across 43 US states from the Monitoring the Futures Survey
7
Although the latter had no effect on marijuana use
8
Moving from where marijuana is fully criminalised to where it is fully decriminalised
9
Taxes were used to proxy pricing effects as they were thought to better address policy questions
and show fewer measurement errors.
5
break up where Pacula identified the prominence of a racial bias.10
It was interesting to note
that the majority of individuals getting caught for possession were under the influence
themselves, which would possibly distort the results.
Supporting work by Williams et al (2001) looked at alcohol use amongst college students on
the basis of Congress's Drug-Free Schools and Communities Act 1986. They utilised the
Harvard School of Public Health College Alcohol Study (1993, 1997, 1999) along with the
Illegal Drug Price/Purity Report (IDPPR) published by U.S. Department of Justice as a way of
segmenting potential effects on the basis of marijuana quality. By examining own and cross
price effects across 30 day prevalence equations, the group identified that marijuana may
form a broader social trend of consumption that isn't inherently linked to alcohol but
validates it as being an economic complement. People’s perception of the ‘legal cost’ of
using marijuana and how the relative views could alter the cross-price elasticity was their
key concern. A gender break up suggested that females showed the most prevalent effects
on alcohol and marijuana consumption in regards to a full alcohol ban. The results, while
significant, only highlighted the effect under certain price specifications and still left room as
to whether unobservable factors (e.g. cigarette use) are driving relationship between the
two variables. Indeed, Crost and Guerrero (2012) found that, by IV estimation, common
influencers between alcohol and marijuana, the two goods return back to being economic
substitutes. Their RDD model11
worked well to monitor the causal effect of the minimum
legal drinking age on the substitutability between alcohol and marijuana with more
significant effects being identified for men, contradicting the previous study done by
Williams et al (2001).
To the author’s knowledge, this paper is the first of its kind to directly examine pricing
effects through a standardised form of marijuana, in contrast to prior results that utilised
taxation, prohibition or alternative proxies to determine the economic relationship. Williams
et al (2001) pioneered categorical pricing effects by observing how the economic
relationship changed for different qualities of marijuana. Building on, this study focuses on
how categorical pricing effects are altered by an individual's’ legal status and demographic
characteristics respectively.
2.2 Theoretical Framework
Consumer Utility Maximisation
Standard individual utility maximisation theory purports increasing utility in respect to
increased consumption. Rational consumers have monotonically increasing consumption
functions, don't like high prices, and that this effect is amplified for university students who
on average are on a restricted budget (usually on student loans).
10
African-Americans were more likely to be accosted and arrested for possession
11
Regression Discontinuity Design models were first introduced by Thistlethwaite and Campbell
(1960) as a way of identifying treatment effects on individuals in a non-experimental way. Treatment
was determined by whether the observable variable passed a 'cut-off' point.
6
The following effects can be identified:
∂F(Consumption Function) = "Positive"
∂goodi
∂F(Consumption Function) = "Negative"
∂pricei
This study restricts utility maximisation to maximising consumer surplus and discounts for
exogenous factors (such as altruism) that may affect individual utility. Purchasing behaviour
is driven by prices, with consumer surplus shifting outwards for an increase in price by a
substitute and inwards for a complement.
Wealth Effects and Changing Preferences
Individual responsiveness to price changes in relation to overall net wealth. The absolute
elasticity of prices for relatively low value goods is higher for individuals in the lower income
bracket. A simple example of this would be how a middle-class individual considers eating at
a restaurant an occasional luxury whereas for someone wealthier it is more common.
There exists an upper bound impact of price increases, after which the consideration of price
as an influence on consumption behaviour automatically goes close to 0. This would
primarily occur for Veblen goods12
where individuals wealthy enough to purchase it would
not factor in price but rather alternative influencers such as celebrity promotion etc.
Marijuana, being primarily an illegal substance with high standard prices, would possibly
qualify for this wealth effect. Literature suggests that young adults from wealthy
backgrounds tend to purchase marijuana with greater frequency, possibly due to the status
associated with possession. Given their background, price is unlikely to significantly affect
their purchasing behaviour. Furthermore, the illegality of the drug would push up its price,
allowing mostly those wealthy young adults to purchase them at a regular frequency.
3. Data and Summary Statistics
3.1 Dataset
Data is obtained on individual consumption patterns and pricing response to illicit
substances from the National Survey on Drug Use and Health (NSDUH).13
The initial sample
contains all individuals sampled for the period 2002-2014. The analysis is restricted to
individuals at university, and within the age of 18-29 years to reduce anomaly effects. The
final sample consists of 6,308 observations over the 12-year period.
First launched in 1999, the NSDUH (formerly the National Household Survey on Drug Abuse)
primarily measures prevalence and correlates of drug use within the United States. Using
12 A luxury good whose price does not follow the usual laws of demand and supply
13 Courtesy of the Substance Abuse and Mental Health Services Administration (SAMHSA).
7
the NSDUH, this study extracts a rich set of socio-demographic and consumption
characteristics at individual level. Notably this includes the price last paid for marijuana, the
main variable of interest.14
Responses are categorised into pricing brackets depending on
amount paid.15
Restrictions on personal data made it difficult to obtain state specific
geographic information that would have provided an interesting extension to the analysis.
Future studies could test pricing effects on States that have legalised marijuana.
3.2 Variables
3.2.1 Consumption likelihood measures
Our primary measures for consumption of alcohol or marijuana are MJOYR2 and ALCYR,
binary variables recording the likelihood of consuming marijuana or alcohol in the past year
respectively. Self reported values have a tendency to be either under or over-reported
(Fisher and Katz, 2000). However, the nature of the data accounts for this by providing
anonymity and financial motivation to individuals sampled to encourage truthful responses.
The model used is an analogous variation to the model outlined by Williams et al.
(2001), with the binary variables simplifying the existing model in literature. The
likelihood of consumption of the substance in the previous year was used as a proxy
for utility derived by consumption. An implicit assumption of this study is that consuming
marijuana or alcohol gives positive utility otherwise individuals would choose not to
consume it. Refer to Appendix16
for summary statistics of consumption likelihood measures.
The average likelihood an individual has consumed marijuana and alcohol in the past year
was 44.2% and 98.6%, respectively, over the whole sampling period. On average this means
that an individual selected at random is nearly always likely to have consumed alcohol in the
past year compared to marijuana, which has roughly 50-50 chances.
3.2.2 Pricing measure
The key independent variable in this study is marijuana prices. Prior studies have all
attempted to establish the economic relationship via proxy methods, either through indirect
pricing effects by taxation (Pacula, 1998) or via legal implications on consumption (DiNardo
and Lemieux, 1992). However due to the nature of the dataset, we're able to obtain
amounts last paid for marijuana. This provides a more robust pricing outlook and lets us
examine the direct effect of pricing on the economic relationship between the two. Panel B
of Table 1 presents summary statistics of the pricing measure. The average price paid last for
marijuana is between $11 to $20.99.
14
Derived from the question "How much did you pay for the marijuana you bought this last time?".
15
See Table 1 in Appendix for detailed variable descriptions.
16
Panel A of Table 2
8
3.2.3 Control Variables
To provide a more robust analysis, key control variables are included that fall in line with
recent consumer preference literature (Caulkins & Pacula, 2006). These are broken up into
individual characteristics (represented by age and sex), locality (represented by university
year and county type) and purchasing capability (represented by whether the individual is of
legal drinking age or not). Panel C of Table 1 provides summary statistics for the control
variables used in this study. For instance, 62.2% of the sample was of legal drinking age
during the time of survey. The variables are designed to test variation in the pricing
relationship.
3.2.4 Correlations among variables
As a fundamental check for multi-collinearity, Table 3 in the appendix reports the
correlations amongst all the independent variables in the analysis. A general rule of thumb,
correlation values over 0.8 in absolute terms suggest possible multi-collinearity within
variables (Gujarati, 1995). Table 2 shows that the highest correlation coefficient of 0.84
between AGE2 and legal. Since legal is a proxy for an age-induced shock to an individual's
capability to purchase substances, this relationship is expected to occur and unlikely to
affect the overall analysis to the variables of interest.
9
4. Research Methodology
4.1 Main model and estimation method
The following is the main model regression for how marijuana prices affect the likelihood of
consuming marijuana:
𝑝𝑟𝑜𝑏(𝑀𝐽𝑂𝑌𝑅2 = 1) = 𝛽0 + 𝛽1 𝐴𝐺𝐸2𝑖 + 𝛽2 𝐼𝑅𝑆𝐸𝑋𝑖 + 𝛽3 𝐼𝑅𝐸𝐷𝑈𝐶2𝑖 + 𝛽4 𝐶𝑂𝑈𝑁𝑇𝑌𝑃2𝑖 +
𝛽5 𝑀𝐽𝐶𝐴𝑇𝐸𝐺𝑖 + 𝛽6 𝑙𝑒𝑔𝑎𝑙𝑖 + 𝜀𝑖
The following is the main model regression for how marijuana prices affect the likelihood of
consuming alcohol:
𝑝𝑟𝑜𝑏(𝑀𝐽𝑂𝑌𝑅2 = 1) = 𝛽0 + 𝛽1 𝐴𝐺𝐸2𝑖 + 𝛽2 𝐼𝑅𝑆𝐸𝑋𝑖 + 𝛽3 𝐼𝑅𝐸𝐷𝑈𝐶2𝑖 + 𝛽4 𝐶𝑂𝑈𝑁𝑇𝑌𝑃2𝑖 +
𝛽5 𝑀𝐽𝐶𝐴𝑇𝐸𝐺𝑖 + 𝛽6 𝑙𝑒𝑔𝑎𝑙𝑖 + 𝜀𝑖
MJCATEG is a categorical variable showing last price paid for marijuana. The price paid on
average tends to be normally distributed with a slight skew to the right. AGE2, IRSEX,
IREDUC2, COUTYP2 and legal are five sets of control variables related to age, gender,
educational year (either freshman, sophomore or senior), county type and legal
characteristics of the individual respectively.
There are two estimation methods commonly used in addiction and pricing literature. One is
a Probit model, controlling for environment effects and the other is a base latent variable
regression adapted from the model developed by Williams et al (2001). Due to the nature of
the data, individual level observations over time are unavailable and would restrict sampling
size to the period at which the individual was at university. Therefore, the Probit regression
on pooled cross sectional data is favoured as the main regression method. The linear
probability model, analogous to the Probit model, is avoided due to inconsistency and bias
in generated estimates as they are not bound to the unit interval (Horrace and Oaxaca,
2006). The reported standard errors are adjusted for potential heteroscedasticity.
4.2 Endogeneity
There is concern between endogeneity of marijuana prices and the desire to consume either
alcohol or marijuana, since marijuana prices can affect the decision to consume the two
goods but also affect an individual's purchasing power (Galea et al., 2007). An alternative
model specification uses instrument-variables ("IVs") to estimate the main regression model
via the two-stage least squares ("2SLS") method. The IVs used need to meet the model
exogeneity and instrument relevance conditions. Considering the price paid for marijuana
can be affected by the price of loose marijuana (MMLSPCAT) and the quantity purchased in
grams (MMLSGMS), the two alternative variables are utilised as IVs for MJCATEG.
10
5. Results
5.1 Are marijuana and alcohol economic complements or substitutes?
Initial regressions look at the pricing effect of marijuana on the desire to consume either
marijuana or alcohol17
. Individual Probit regressions are used with results outlined in
columns (1) and (4). Results modelling through instrument-variable estimation and for a
Bivariate Probit analysis are outlined in columns (2) and (5), and (3) and (6) respectively.18
Constant sample size is used and robust standard errors are used throughout.
The pricing effects are statistically significant and suggest an inverse relationship as
expected (Pacula et al., 2014). For instance, a rise in price bracket from <$5 (default category
in the analysis) to $5 - $10 reduces desire to consume marijuana by 20% and increases it for
alcohol by 62%. This fits in with standard utility maximisation theory outlined in section 2.2.
At gross level, marijuana and alcohol are observed to be economic substitutes. Specific to
alcohol consumption, there is a maximum pricing effect before a reduction in impact of
increased marijuana price, on the desire to consume alcohol (Glaeser et al., 2008). Lower
priced marijuana categories showed a more significant effect in contrast to higher prices,
which declined in impact as further control variables, were introduced (Tsuang et al., 2001).
Concerning the IVs estimation method, the instruments are tested to meet the exogeneity
and relevance conditions. Hansen's J instrument test19
is employed to examine if the IVs
mention in section 4.2 meet the exogeneity requirement. The null hypothesis is that the IV's
are not correlated with the error terms in each model. The regressions measured by ALCYR
and MJOYR2 both produce insignificant J statistics, measured at the 10% significance level,
failing to reject the null hypothesis. The relevance of each instrument is then examined,
treating MJCATEG as a continuous variable and undertaking a standard linear regression.
The F-test statistic is significant at the 1% level indicating that the instruments are jointly
significant. However, MMLSGMS individual F-test statistic is insignificant, indicating a
potentially weak instrument problem. The results based on IVs estimation indicate that
marijuana prices continue to have a negative impact on desire to consume marijuana and a
positive impact on desire to consume alcohol. Although, it is important to interpret the
magnitude and significance level of MJCATEG's coefficient with caution since weak IVs may
cause estimators to perform poorly.
A bivariate Probit analysis is run to test whether the decision to consume alcohol or
marijuana could potentially be related through the error term. The correlation coefficient
between the bivariate outcomes is -0.28 and is significant at the 1% significance level. The
decisions are therefore interdependent and should be estimated via bivariate analysis rather
than independently. However, the coefficient effects are nearly identical and for simple
analysis, independent Probit models would not alter the validity of the model. The
robustness of this is further discussed in section 6.
Noteworthy observations are that gender effects are insignificant although women are more
likely to consume marijuana than their male counterparts (Booth & Nolen, 2012).20
Also
17 Outlined in Table 3
18
See table 4 in the Appendix for Instrument Relevance test
19
Also know as the over-identification test
20
Women are considered to be more experimental during university as compared to men.
11
individual’s legal status is highly important in determining whether they consume marijuana
or alcohol (Yörük & Yörük, 2011). This is further analysed in section 5.3.
5.2 Do age and university year affect desire to consume substances?
To enrich the analysis, preferences changing across age and university year are analysed.
Marginal effects for each standard model across AGE2 and IREDUC2 are plotted and can be
seen in Figures 2 and 3. The graphs are created by considering age and university year as
categorical variables and plotting marginal effects for each.
General trends of Figure 1(a) and 1(b) suggests that there is an inverse and direct cohort
effect induced by university year on desire to consume marijuana and alcohol respectively.
Purchasing preferences may be subject to social norms invoked by the individuals you
interact with on a regular basis (Perkins, 2002). Within university, this may be peer-led
decisions, usually drawn from what is deemed socially acceptable at present time. This
effect is amplified by the inherent social culture of drinking promoted through US
fraternities and sororities (Wechsler et al., 2009). These social organisations usually offer
selective post-degree networks that are often highly prized and sought after (Marmaros and
Sacerdote, 2002). Due to this, college individuals may inherently self-select alcohol over
marijuana as a required trait of these clubs (Phua, 2011).
Figures 2(a) and 2(b) suggest that there is a direct and inverse age induced effect on desire
to consume marijuana and alcohol respectively. The initial spike in the desire to consume
alcohol once an individual turns 21 is further analysed in section 5.3. Older students on
general are more likely to have consumed alcohol although their consumption trends move
towards greater moderation with age (Engs and Hanson, 1986).
5.3 Are university students of legal drinking age more likely to
consume alcohol?
Perception of legal implications clearly plays an important role when deciding which
substance to consume (Williams et al, 2001). The standard Probit model is broken up by
individuals who are not of legal drinking age and by those who are.21
Figure 3(a) and 3(b) suggest that there is a straightforward substitution effect for those not
of legal drinking age. Increased marijuana price decreases desire to consume marijuana and
increases desire to consume alcohol. This effect is amplified for marijuana, for those who are
not of legal drinking age. Figure 4(a) suggests that this effect is retained for marijuana for
those of LDA. Theoretically, once an individual is given the freedom to undertake an action,
there is a spike in desire to do so followed by a slow decline (Fromme et al., 2010). The
action in this case is the ability to purchase either alcohol or marijuana. Being of LDA can
have a significant impact on consumers switching preference away from marijuana to
alcohol. Prior to being of LDA, purchasing alcohol and marijuana had the same legal
implications if the individual were to get caught. However, the results suggest that alcohol is
now the preferred choice. There is a legal bias towards the good where the individual's
ability to purchase the substance legally, fuels his choice (Fromme et al., 2010). Figure 4(b)
suggest that for those of legal drinking age, there seems to be a positive relationship
between alcohol and marijuana initially, till a maximum is reached, after which a decline in
21
Statutory law in the US indicates that the general legal drinking age is 21.
12
likelihood of consumption (complements). A possible “Wealth Effect” is identified. Pricing
effects are dampened for wealthy individuals. (Glaeser et al, 2008). An alternative outlook is
legal bias. Those who are of legal drinking age are more likely to consume alcohol and less
likely to be concerned with marijuana price (Yörük & Yörük, 2011). However the effect of
legal bias is only temporary. Once the individual gets accustomed to their newfound
freedom, the desire to consistently purchase alcohol over marijuana solely due to legal
ability dies out. Legal comfort is experienced and the individual returns back to his original
consumption preferences.
Table 3
Pricing Effects of Marijuana on desire to consume Marijuana or Alcohol
(MJOYR2) Desire to Consume Marijuana (ALCYR) Desire to Consume Alcohol
Standard Model
(1)
IVs Model
(2)
Bivariate Model
(3)
Standard Model
(4)
IVs Model
(5)
Bivariate Model
(6)
MJCATEG
$5 - $10.99
$11 - $20.99
$21 - $50.99
$51 - $100.99
>$101
-0.20*
(0.12)
-0.57***
(0.11)
-0.70***
(0.11)
-0.82***
(0.12)
-0.96***
(0.12)
-0.17***
(0.02) -0.20*
(0.12)
-0.57***
(0.11)
-0.70***
(0.11)
-0.82***
(0.12)
-0.96***
(0.12)
0.62***
(0.19)
0.81***
(0.19)
0.76***
(0.18)
0.64***
(0.20)
0.62***
(0.21)
0.02
(0.04) 0.63***
(0.19)
0.81***
(0.19)
0.75***
(0.18)
0.63***
(0.20)
0.63***
(0.21)
AGE2 0.11***
(0.02)
0.12***
(0.02)
0.11***
(0.02)
-0.19***
(0.06)
-0.19***
(0.06)
-0.20***
(0.06)
IRSEX 0.03
(0.03)
0.04
(0.03)
0.03
(0.02)
-0.09
(0.09)
-0.08
(0.09)
-0.10
(0.09)
IREDUC2 -0.08***
(0.03)
-0.9***
(0.03)
-0.08***
(0.03)
0.32***
(0.07)
0.34***
(0.07)
0.33***
(0.07)
COUTYP2 -0.07***
(0.02)
-0.07***
(0.02)
-0.07***
(0.02)
0.10
(0.07)
0.10
(0.07)
0.10
(0.07)
legal -0.17***
(0.06)
-0.17***
(0.06)
-0.17***
(0.06)
0.48***
(0.17)
0.47***
(0.17)
0.48***
(0.17)
Rho -0.03
(0.02)
-0.28***
(0.05)
0.07
(0.05)
-0.28***
(0.05)
Observations 6308 6308 6308 6308 6308 6308
Pseudo R2
0.03 0.06
The left panel reports the pricing effects of marijuana on the desire to consume marijuana.
The right panel reports the pricing effects of marijuana on the desire to consume alcohol.
*p<0.1; ** p<0.05; *** p<0.01
13
Figure 1(a) University year impact on desire to consume marijuana Figure 2(a) Age impact on desire to consume marijuana
Figure 1(b) University year impact on desire to consume alcohol Figure 2(b) Age impact on desire to consume alcohol
14
Figure 3(a) Pricing effects on marijuana consumption for those not of LDA Figure 3(b) Pricing effects on alcohol consumption for those not of LDA
Figure 4(a) Pricing effects on marijuana consumption for those of LDA Figure 4(b) Pricing effects on alcohol consumption for those of LDA
15
6. Robustness Checks
This section covers checking the model's validity in contrast to using a naive estimator to
simply gauge consumption preferences. The 'Hit-and-run' results for this can be see in Figure
5 (a) and 5(b) in the Appendix. 98.6% all individuals at university in the study have consumed
alcohol at some point during the past year. This is not surprising (Johnston et al., 2008) and
would mean that, in general, regardless of how well specified your model is, simply
assuming all individuals consume alcohol would be the best way to predict consumption
patterns. However, since the empirical analysis is targeted at pricing effects on standard
consumption, the model described still has valid implications. What's interesting to note is
that in comparison to the naive estimator of marijuana consumption22
, the model acts as a
better predictor for individual behaviour.
Robust standard errors are used to account for heteroscedasticity. A bivariate Probit analysis
is undertaken to test whether the decision to consume marijuana and alcohol can be related
through the error term (Greene, 1984). This can be seen in table 1. The results do suggest
this and under normal scenarios it would make sense to use a bivariate analysis as the main
model. However, as mentioned earlier, the purpose of this study is not to create the best-fit
model but to explicitly look at how pricing affects purchasing decisions (Yamada et al.,
1993). As a simple direct tool, independent Probit equations can be used to undertake the
analysis (Bray et al., 2000).
7. Discussion and Concluding Remarks
The relationship between alcohol and marijuana is interesting. Empirical results suggest that
the goods could be, both, economic complements or substitutes under specific scenarios.
The legal drinking age is an important factor that affects the pricing effect of marijuana on
alcohol. Generally, an individual who is of legal drinking age is more likely to consume
alcohol rather than marijuana. Social factors such as university year, age and sex provide a
more robust outlook at the economic relationship identified.
Standard policy implications stand towards stricter enforcement on alcohol sale around
university areas. Empirically a large proportion of students were observed to be engaging in
underage drinking. Building upon what Pacula et al. (2014) suggested, it would be beneficial
to look towards legalising marijuana and then heavily taxing consumption to deter
substance abuse.
Limitations to the analysis include the lack of personal data, which distorts the model's
accuracy. Reverse causality is another concerning issue. It is still unclear as to whether
individual motivation to consume substances affects price or vice versa. Due to the nature of
the dataset this paper is unable to account for changing consumption patterns over time.23
Extensions to this paper would include testing individual specific motivation over time in
tandem with additional demographic variables such as family wealth. It would also be
22
Assuming all individuals have not consumed marijuana.
23
This could be ignored considering that alcohol consumption at university has remained at nearly
100% over the years.
16
worthwhile to check how consumption patterns differ over states where purchasing
marijuana is legal.
The paper finds significant evidence supporting the hypothesis that marijuana and alcohol
are at gross, substitutes. However legal implications clearly play an important role an induce
variation into this economic relationship. Future tax policy changes should allow for
flexibility, adapting to suit the situational economic relationship at hand.
17
8. Appendix
8.1 Appendix I: Tables
Table 1: Description of the Variables Used
Variable Description
Panel A: Consumption Likelihood Measures
MJOYR2 Binary Variable; equals to 1 if marijuana was
consumed in the last 12 months and 0 otherwise
ALCYR Binary Variable; equals to 1 if alcohol was
consumed in the last 12 months and 0 otherwise
Panel B: Pricing Measure
MJCATEG The average categorical price paid for last using
marijuana. This is broken up into 6 categories of
under $5, $5 - $10.99, $11 - $20.99, $21 - $50.99,
$51 - $100.99, >=$101 with numerical assignment
of 1, 2, 3, 4, 5 and 6 respectively. Under $5 is the
assumed default price for our analysis.
Panel C: Control Variables
AGE2 Categorical variable for individuals aged between
19 -29. This is broken down into 6 categories of
19, 20, 21, 22 or 23, 24 or 25 and 26 to 29 with
numerical assignment of 8, 9, 10, 11, 12, 13
respectively.
IRSEX Gender Variable; assumes a value of 1 for male
and 2 for female.
IREDUC2 Describes individual current education year at
university. Broken up into freshmen, sophomores
and seniors with numerical assignment as 9, 10
and 11 respectively
COUTYP2 Geographic indicator. Individual can be in either a
large metropolitan area, small metropolitan area,
or a non metropolitan area with numerical
assignment 1, 2 and 3 respectively.
legal Dummy Variable; equals to 1 if individual is of
legal drinking age and 0 otherwise
18
Table 2: Summary Statistics
Variable Obs Mean Std. Dev. Min Max
Panel A: Consumption Likelihood Measures
MJOYR2 6308 44.2% .5 0 1
ALCYR 6308 98.6% .12 0 1
Panel B: Pricing Measure
MJCATEG 6308 3.72 1.2 1 6
Panel C: Control Variables
AGE2 6308 10.1 1.43 8 13
IRSEX 6308 1.40 0.49 1 2
IREDUC2 6308 9.83 0.67 9 11
COUTYP2 6308 1.66 0.71 1 3
legal 6308 62.2% 0.49 0 1
Table 3: Correlation Matrix
1 2 3 4 5 6 7 8
1 MJOYR2 1.00
2 ALCYR -0.07 1.00
3 MJCATEG -0.17 0.03 1.00
4 AGE2 0.03 0.01 0.15 1.00
5 IRSEX 0.04 -0.01 -0.14 -0.01 1.00
6 IREDUC2 -0.02 0.06 0.12 0.52 0.00 1.00
7 COUTYP2 -0.05 0.02 0.07 -0.02 -0.04 -0.02 1.00
8 legal 0.01 0.03 0.11 0.84 -0.01 0.49 -0.013 1.00
This table presents the correlation matrix for all the independent variables employed in this study.
19
Table 4: Instrument Relevance Testing
MJCATEG (Categorical Pricing Variable) Standard Model
MMLSGMS -0.0001
(0.00)
MMLSPCAT 0.041***
(0.00)
Observations 6308
R2
0.8328
*p<0.1; ** p<0.05; *** p<0.01
Table 5: Age effects on desire to consume alcohol or marijuana
(MJOYR2) Desire to Consume Marijuana (ALCYR) Desire to Consume Alcohol
Standard Model Standard Model
MJCATEG
$5 - $10.99
$11 - $20.99
$21 - $50.99
$51 - $100.99
>$101
-0.19
(0.12)
-0.56***
(0.11)
-0.69***
(0.11)
-0.82***
(0.12)
-0.96***
(0.12)
0.62***
(0.19)
0.81***
(0.19)
0.76***
(0.18)
0.64***
(0.20)
0.61***
(0.21)
AGE2
20
21
22/23
24/25
26-29
0.01
(0.05)
0.01
(0.06)
0.07
(0.06)
0.20***
(0.07)
0.47***
(0.09)
-0.10
(0.13)
0.14
(0.16)
0.02
(0.16)
-0.17
(0.17)
-0.49***
(0.18)
IRSEX 0.03
(0.03)
-0.09
(0.09)
IREDUC2 -0.06**
(0.03)
0.31***
(0.08)
COUTYP2 -0.07***
(0.02)
0.10
(0.07)
Observations 6308 6308
Pseudo R2
0.03 0.06
*p<0.1; ** p<0.05; *** p<0.01
20
Table 6: Testing Age Effect at Margins
(MJOYR2) Desire to Consume Marijuana (ALCYR) Desire to Consume Alcohol
At Margins At Margins
AGE2
19
20
21
22/23
24/25
26-29
0.41***
(0.02)
0.42***
(0.02)
0.42***
(0.02)
0.44***
(0.01)
0.49***
(0.02)
0.60***
(0.03)
0.99***
(0.00)
0.99***
(0.00)
0.99***
(0.00)
0.99***
(0.00)
0.98***
(0.00)
0.97***
(0.01)
*p<0.1; ** p<0.05; *** p<0.01
Table 7: University Year Effect on desire to consume alcohol or marijuana
(MJOYR2) Desire to Consume Marijuana (ALCYR) Desire to Consume Alcohol
Standard Model Standard Model
MJCATEG
$5 - $10.99
$11 - $20.99
$21 - $50.99
$51 - $100.99
>$101
-0.20*
(0.12)
-0.57***
(0.11)
-0.70***
(0.11)
-0.82***
(0.12)
-0.96***
(0.12)
0.62**
(0.19)
0.81***
(0.19)
0.76***
(0.18)
0.64***
(0.20)
0.62***
(0.21)
AGE2 0.11***
(0.02)
-0.19***
(0.06)
IRSEX 0.03
(0.03)
-0.09
(0.09)
IREDUC2
Sophomore
Senior
-0.13***
(0.04)
-0.13**
(0.06)
0.31***
(0.09)
0.67***
(0.18)
COUTYP2 -0.07***
(0.02)
0.10
(0.07)
legal -0.15** 0.49***
21
(0.06) (0.17)
Observations 6308 6308
Pseudo R2
0.03 0.06
*p<0.1; ** p<0.05; *** p<0.01
Table 8: Testing University Year Effect at Margins
(MJOYR2) Desire to Consume Marijuana (ALCYR) Desire to Consume Alcohol
At Margins At Margins
IREDUC2
Freshman
Sophomore
Senior
0.47***
(0.01)
0.42***
(0.01)
0.42***
(0.02)
0.98***
(0.00)
0.99***
(0.00)
1.00***
(0.00)
*p<0.1; ** p<0.05; *** p<0.01
Table 9: Price Effect on desire to consume alcohol or marijuana for those not of LDA
(MJOYR2) Desire to Consume Marijuana (ALCYR) Desire to Consume Alcohol
Standard Model Standard Model
MJCATEG
$5 - $10.99
$11 - $20.99
$21 - $50.99
$51 - $100.99
>$101
-0.29
(0.18)
-0.72***
(0.18)
-0.89***
(0.18)
-1.10***
(0.19)
-1.1***
(0.20)
0.68***
(0.24)
1.00***
(0.25)
0.89***
(0.25)
1.20***
(0.34)
1.30***
(0.42)
AGE2 0.04
(0.06)
-0.17
(0.16)
IRSEX 0.01
(0.05)
-0.10
(0.13)
IREDUC2 -0.10*
(0.06)
0.43**
(0.18)
COUTYP2 -0.11***
(0.04)
0.26**
(0.12)
Observations 2387 2387
Pseudo R2
0.04 0.09
*p<0.1; ** p<0.05; *** p<0.01
22
Table 10: Testing Pricing Effect for those not of LDA at Margins
(MJOYR2) Desire to Consume Marijuana (ALCYR) Desire to Consume Alcohol
At Margins At Margins
MJCATEG
<$5
$5 - $10.99
$11 - $20.99
$21 - $50.99
$51 - $100.99
>$101
0.72***
(0.06)
0.62***
(0.02)
0.45***
(0.02)
0.39***
(0.02)
0.30***
(0.03)
0.29***
(0.03)
0.90***
(0.04)
0.98***
(0.01)
0.99***
(0.00)
0.99***
(0.00)
0.99***
(0.00)
1.00***
(0.01)
*p<0.1; ** p<0.05; *** p<0.01
Table 11: Price Effect on desire to consume alcohol or marijuana for those of LDA
(MJOYR2) Desire to Consume Marijuana (ALCYR) Desire to Consume Alcohol
Standard Model Standard Model
MJCATEG
$5 - $10.99
$11 - $20.99
$21 - $50.99
$51 - $100.99
>$101
--0.13
(0.16 )
--0.46***
(0.15 )
--0.57***
(0.15 )
--0.65***
(0.15 )
--0.83***
(0.16 )
0.68**
(0.34)
0.59**
(0.30)
0.57**
(0.29)
0.27
(0.30)
0.24
(0.31)
AGE2 0.13***
(0.02)
-0.19***
(0.06)
IRSEX 0.04
(0.04)
-0.09
(0.12)
IREDUC2 -0.06*
(0.03)
0.28***
(0.08)
COUTYP2 -0.04
(0.03)
-0.01
(0.09)
Observations 3921 3921
Pseudo R2
0.02 0.06
*p<0.1; ** p<0.05; *** p<0.01
23
Table 12: Testing Pricing Effect for those of LDA at Margins
(MJOYR2) Desire to Consume Marijuana (ALCYR) Desire to Consume Alcohol
At Margins At Margins
MJCATEG
<$5
$5 - $10.99
$11 - $20.99
$21 - $50.99
$51 - $100.99
>$101
0.65***
(0.05)
0.59***
(0.02)
0.47***
(0.02)
0.42***
(0.01)
0.39***
(0.02)
0.32***
(0.02)
0.97***
(0.02)
0.99***
(0.00)
0.99***
(0.00)
0.99***
(0.00)
0.98***
(0.00)
0.98***
(0.01)
*p<0.1; ** p<0.05; *** p<0.01
24
8.2 Appendix II: Figures
Figure 5(a): Testing model validity - Probit Model
Table 5(b): Testing model validity - Naive Estimator
Marijuana Use Frequency Percent
No 3523 55.85
Yes 2785 44.15
Total 6308 100
Alcohol Use Frequency Percent
No 88 1.4
Yes 6220 98.60
Total 6308 100
25
9. References
Amonini, C. "The Relationship between Youth's Moral and Legal Perceptions of Alcohol, Tobacco and
Marijuana and Use of These Substances." Health Education Research 21.2 (2005): 276-86. Print.
Bailey, Jennifer A., Karl G. Hill, Meredith C. Meacham, Susan E. Young, and J. David Hawkins.
"Strategies for Characterizing Complex Phenotypes and Environments: General and Specific Family
Environmental Predictors of Young Adult Tobacco Dependence, Alcohol Use Disorder, and Co-
occurring Problems." Drug and Alcohol Dependence 118.2-3 (2011): 444-51. Print.
Booth, Alison, and Patrick Nolen. "Choosing to Compete: How Different Are Girls and Boys?" Journal
of Economic Behavior & Organization 81.2 (2012): 542-55. Print.
Booth, Alison L., and Patrick Nolen. "Gender Differences in Risk Behaviour: Does Nurture Matter?*."
The Economic Journal 122.558 (2012). Print.
Bray, Jeremy W., Gary A. Zarkin, Chris Ringwalt, and Junfeng Qi. "The Relationship between
Marijuana Initiation and Dropping out of High School." Health Econ. Health Economics 9.1 (2000): 9-
18. Print.
Cameron, Lisa, and Jenny Williams. "Cannabis, Alcohol and Cigarettes: Substitutes or
Complements?" Economic Record 77.236 (2001): 19-34. Print.
Caulkins, J. P., and R. L. Pacula. "Marijuana Markets: Inferences from Reports by the Household
Population." Journal of Drug Issues 36.1 (2006): 173-200. Print.
Chaloupka, Frank, and Adit Laixuthai. "Do Youths Substitute Alcohol and Marijuana? Some
Econometric Evidence." (1994). Print.
Crost, Benjamin, and Santiago Guerrero. "The Effect of Alcohol Availability on Marijuana Use:
Evidence from the Minimum Legal Drinking Age." Journal of Health Economics 31.1 (2012): 112-21.
Print.
26
Desimone, Jeff. "Illegal Drug Use and Employment." Journal of Labor Economics 20.4 (2002): 952-77.
Print.
Dinardo, John, and Thomas Lemieux. "Alcohol, Marijuana, and American Youth: The Unintended
Effects of Government Regulation." (1992). Print.
Engs, Ruth C., and David J. Hanson. "Age-Specific Alcohol Prohibition And College Students' Drinking
Problems." Psychological Reports 59.2 (1986): 979-84. Print.
Fisher, Robert J., and James E. Katz. "Social-desirability Bias and the Validity of Self-reported Values."
Psychology and Marketing Psychol. Mark. 17.2 (2000): 105-20. Print.
Fromme, Kim, Reagan R. Wetherill, and Dan J. Neal. "Turning 21 and the Associated Changes in
Drinking and Driving After Drinking Among College Students." Journal of American College
Health 59.1 (2010): 21-27. Print.
Galea, Sandro, Jennifer Ahern, Melissa Tracy, and David Vlahov. "Neighborhood Income and Income
Distribution and the Use of Cigarettes, Alcohol, and Marijuana." American Journal of Preventive
Medicine 32.6 (2007). Print.
Glaeser, Edward, Joseph Gyourko, and Albert Saiz. "Housing Supply and Housing Bubbles." (2008).
Print.
Greene, William H. "Estimation of the Correlation Coefficient in a Bivariate Probit Model Using the
Method of Moments." Economics Letters 16.3-4 (1984): 285-91. Print.
Gujarati, Damodar N. Basic Econometrics. New York: McGraw-Hill, 1995. Print.
Hanson, David J., and Ruth C. Engs. "College Students' Drinking Problems: A National Study, 1982-
1991." Psychological Reports 71.1 (1992): 39. Print.
Hingson, Ralph W., Wenxing Zha, and Elissa R. Weitzman. "Magnitude of and Trends in Alcohol-
Related Mortality and Morbidity Among U.S. College Students Ages 18-24, 1998-2005."Journal of
Studies on Alcohol and Drugs, Supplement J. Stud. Alcohol Drugs Suppl. S16 (2009): 12-20. Print.
Horrace, William C., and Ronald L. Oaxaca. "Results on the Bias and Inconsistency of Ordinary Least
Squares for the Linear Probability Model." Economics Letters 90.3 (2006): 321-27. Print.
27
"Intoxicating Liquors. Eighteenth Amendment. Interpretation of the Volstead Act." Harvard Law
Review 34.4 (1921): 437. Print.
Johnston, Lloyd D., Patrick M. O'malley, Jerald G. Bachman, and John E. Schulenberg. "Monitoring
the Future: National Results on Adolescent Drug Use: Overview of Key Findings, 2008." PsycEXTRA
Dataset. Print.
Manning, Paul. Drugs and Popular Culture: Drugs, Media and Identity in Contemporary Society.
Cullompton, Devon, England: Willan Pub., 2007. Print.
"Marginal Effects in the Bivariate Probit Model." By William H. Greene. Web. 14 Apr. 2016.
Marmaros, David, and Bruce Sacerdote. "Peer and Social Networks in Job Search." European
Economic Review 46.4-5 (2002): 870-79. Print.
"National Survey on Drug Use and Health." Encyclopedia of Substance Abuse Prevention, Treatment,
& Recovery. Print.
Pacula, Rosalie Liccardo. "Adolescent Alcohol and Marijuana Consumption: Is There Really a
Gateway Effect?" (1998). Print.
Perkins, H. Wesley. "Social Norms and the Prevention of Alcohol Misuse in Collegiate Contexts."
Journal of Studies on Alcohol, Supplement J. Stud. Alcohol Suppl. S14 (2002): 164-72. Print.
Phua, Joe. "The Influence of Peer Norms and Popularity on Smoking and Drinking Behavior among
College Fraternity Members: A Social Network Analysis." Social Influence 6.3 (2011): 153-68. Print.
Saffer, Henry, and Frank Chaloupka. "Tobacco Advertising: Economic Theory and International
Evidence." (1999). Print.
Sevigny, Eric L., Rosalie Liccardo Pacula, and Paul Heaton. "The Effects of Medical Marijuana Laws on
Potency." International Journal of Drug Policy 25.2 (2014): 308-19. Print.
Solomon, Michael R., Rebekah Russell-Bennett, and Josephine Previte. Consumer Behaviour: Buying,
Having, Being. Frenchs Forest, N.S.W.: Pearson Australia, 2013. Print.
Tsang, Hector W.h., Ashley Chan, Alvin Wong, and Robert Paul Liberman. "Vocational Outcomes of
an Integrated Supported Employment Program for Individuals with Persistent and Severe Mental
28
Illness." Journal of Behavior Therapy and Experimental Psychiatry 40.2 (2009): 292-305. Print.
Tsuang, Ming T., Jessica L. Bar, Rebecca M. Harley, and Michael J. Lyons. "The Harvard Twin Study of
Substance Abuse: What We Have Learned." Harv Rev Psychiatry Harvard Review of Psychiatry 9.6
(2001): 267-79. Print.
Wechsler, Henry, George Kuh, and Andrea E. Davenport. "Fraternities, Sororities and Binge Drinking:
Results from a National Study of American Colleges." Journal of Student Affairs Research and
Practice 46.3 (2009): 763-84. Print.
"When Can You Safely Ignore Multicollinearity? | Statistical Horizons." Statistical Horizons. Web. 22
Apr. 2016.
Williams, Jenny, Rosalie Liccardo Pacula, Frank Chaloupka, and Henry Wechsler. "Alcohol and
Marijuana Use Among College Students: Economic Complements or Substitutes?" (2001). Print.
Yamada, Tetsuji, Michael Kendix, and Tadashi Yamada. "The Impact of Alcohol Consumption and
Marijuana Use on High School Graduation." (1993). Print.
Yörük, Barış K., and Ceren Ertan Yörük. "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 30.4 (2011): 740-52. Print.
Zarkin, Gary A., Thomas A. Mroz, Jeremy W. Bray, and Michael T. French. "The Relationship between
Drug Use and Labor Supply for Young Men." Labour Economics 5.4 (1998): 385-409. Print
RAE - Final Submission1

More Related Content

Similar to RAE - Final Submission1

Measuring Drug and Alcohol Use AmongCollege Student-Athletes
Measuring Drug and Alcohol Use AmongCollege Student-AthletesMeasuring Drug and Alcohol Use AmongCollege Student-Athletes
Measuring Drug and Alcohol Use AmongCollege Student-AthletesAbramMartino96
 
Is Marijuana a Gateway DrugAuthor(s) Jeffrey DeSimoneSou.docx
Is Marijuana a Gateway DrugAuthor(s) Jeffrey DeSimoneSou.docxIs Marijuana a Gateway DrugAuthor(s) Jeffrey DeSimoneSou.docx
Is Marijuana a Gateway DrugAuthor(s) Jeffrey DeSimoneSou.docxpriestmanmable
 
Running head CRITICAL APPRAISAL SANITATION AND HOSPITAL ACQUIRED.docx
Running head CRITICAL APPRAISAL SANITATION AND HOSPITAL ACQUIRED.docxRunning head CRITICAL APPRAISAL SANITATION AND HOSPITAL ACQUIRED.docx
Running head CRITICAL APPRAISAL SANITATION AND HOSPITAL ACQUIRED.docxhealdkathaleen
 
Action research on drug safety assestment
Action research on drug safety assestmentAction research on drug safety assestment
Action research on drug safety assestmentReymart Bargamento
 
MONITORING FUTURENATIONAL SURVEY RESULTS ON DRUG U.docx
MONITORING     FUTURENATIONAL SURVEY RESULTS ON DRUG U.docxMONITORING     FUTURENATIONAL SURVEY RESULTS ON DRUG U.docx
MONITORING FUTURENATIONAL SURVEY RESULTS ON DRUG U.docxroushhsiu
 
A Model For Pharmacological Research Treatment Of Cocaine Dependence
A Model For Pharmacological Research Treatment Of Cocaine DependenceA Model For Pharmacological Research Treatment Of Cocaine Dependence
A Model For Pharmacological Research Treatment Of Cocaine DependenceRichard Hogue
 
ANALYZING SOLUTIONS FOR UNSAFE PHARMACEUTICAL DISPOSAL final
ANALYZING SOLUTIONS FOR UNSAFE PHARMACEUTICAL DISPOSAL finalANALYZING SOLUTIONS FOR UNSAFE PHARMACEUTICAL DISPOSAL final
ANALYZING SOLUTIONS FOR UNSAFE PHARMACEUTICAL DISPOSAL finalOlivia Chambliss
 
Effective Strategies For Intervening With Drug Abusing Offenders
Effective Strategies For Intervening With Drug Abusing OffendersEffective Strategies For Intervening With Drug Abusing Offenders
Effective Strategies For Intervening With Drug Abusing Offenderslakatos
 
163409237 the-effect-of-cigarette-prices
163409237 the-effect-of-cigarette-prices163409237 the-effect-of-cigarette-prices
163409237 the-effect-of-cigarette-priceshomeworkping7
 
JOURNAL OF SEX RESEARCH. 145-158, 2014RRoutledgeCopyrig.docx
JOURNAL OF SEX RESEARCH. 145-158, 2014RRoutledgeCopyrig.docxJOURNAL OF SEX RESEARCH. 145-158, 2014RRoutledgeCopyrig.docx
JOURNAL OF SEX RESEARCH. 145-158, 2014RRoutledgeCopyrig.docxtawnyataylor528
 
The Importance of Identifying Characteristics Underlyingthe .docx
The Importance of Identifying Characteristics Underlyingthe .docxThe Importance of Identifying Characteristics Underlyingthe .docx
The Importance of Identifying Characteristics Underlyingthe .docxrtodd33
 
Masters thesis differential effectiveness of substance abuse treatment by j f...
Masters thesis differential effectiveness of substance abuse treatment by j f...Masters thesis differential effectiveness of substance abuse treatment by j f...
Masters thesis differential effectiveness of substance abuse treatment by j f...Joyce Fuller
 
Differential Effectiveness of Substance Abuse Treatment by Joyce Fuller
Differential Effectiveness of Substance Abuse Treatment by Joyce FullerDifferential Effectiveness of Substance Abuse Treatment by Joyce Fuller
Differential Effectiveness of Substance Abuse Treatment by Joyce FullerJoyce Fuller
 
Cannabis Smoke Does Not Harm Your Lungs Like Tobacco Smoke Says New Study
Cannabis Smoke Does Not Harm Your Lungs Like Tobacco Smoke Says New StudyCannabis Smoke Does Not Harm Your Lungs Like Tobacco Smoke Says New Study
Cannabis Smoke Does Not Harm Your Lungs Like Tobacco Smoke Says New StudyEvergreen Buzz
 
Title of ProjectPresenter NameUniversity nameIntroduction .docx
Title of ProjectPresenter NameUniversity nameIntroduction .docxTitle of ProjectPresenter NameUniversity nameIntroduction .docx
Title of ProjectPresenter NameUniversity nameIntroduction .docxjuliennehar
 
Seminar paper 5
Seminar paper 5Seminar paper 5
Seminar paper 5juilice
 
Palazzolo 2013 (electronic cigarette)
Palazzolo 2013 (electronic cigarette)Palazzolo 2013 (electronic cigarette)
Palazzolo 2013 (electronic cigarette)ElviraYunita2
 
EconomicBenefits_of_Drug_Trx_02.05_
EconomicBenefits_of_Drug_Trx_02.05_EconomicBenefits_of_Drug_Trx_02.05_
EconomicBenefits_of_Drug_Trx_02.05_Nicholas Patapis
 

Similar to RAE - Final Submission1 (20)

Measuring Drug and Alcohol Use AmongCollege Student-Athletes
Measuring Drug and Alcohol Use AmongCollege Student-AthletesMeasuring Drug and Alcohol Use AmongCollege Student-Athletes
Measuring Drug and Alcohol Use AmongCollege Student-Athletes
 
Is Marijuana a Gateway DrugAuthor(s) Jeffrey DeSimoneSou.docx
Is Marijuana a Gateway DrugAuthor(s) Jeffrey DeSimoneSou.docxIs Marijuana a Gateway DrugAuthor(s) Jeffrey DeSimoneSou.docx
Is Marijuana a Gateway DrugAuthor(s) Jeffrey DeSimoneSou.docx
 
Running head CRITICAL APPRAISAL SANITATION AND HOSPITAL ACQUIRED.docx
Running head CRITICAL APPRAISAL SANITATION AND HOSPITAL ACQUIRED.docxRunning head CRITICAL APPRAISAL SANITATION AND HOSPITAL ACQUIRED.docx
Running head CRITICAL APPRAISAL SANITATION AND HOSPITAL ACQUIRED.docx
 
Action research on drug safety assestment
Action research on drug safety assestmentAction research on drug safety assestment
Action research on drug safety assestment
 
MONITORING FUTURENATIONAL SURVEY RESULTS ON DRUG U.docx
MONITORING     FUTURENATIONAL SURVEY RESULTS ON DRUG U.docxMONITORING     FUTURENATIONAL SURVEY RESULTS ON DRUG U.docx
MONITORING FUTURENATIONAL SURVEY RESULTS ON DRUG U.docx
 
A Model For Pharmacological Research Treatment Of Cocaine Dependence
A Model For Pharmacological Research Treatment Of Cocaine DependenceA Model For Pharmacological Research Treatment Of Cocaine Dependence
A Model For Pharmacological Research Treatment Of Cocaine Dependence
 
ANALYZING SOLUTIONS FOR UNSAFE PHARMACEUTICAL DISPOSAL final
ANALYZING SOLUTIONS FOR UNSAFE PHARMACEUTICAL DISPOSAL finalANALYZING SOLUTIONS FOR UNSAFE PHARMACEUTICAL DISPOSAL final
ANALYZING SOLUTIONS FOR UNSAFE PHARMACEUTICAL DISPOSAL final
 
Effective Strategies For Intervening With Drug Abusing Offenders
Effective Strategies For Intervening With Drug Abusing OffendersEffective Strategies For Intervening With Drug Abusing Offenders
Effective Strategies For Intervening With Drug Abusing Offenders
 
163409237 the-effect-of-cigarette-prices
163409237 the-effect-of-cigarette-prices163409237 the-effect-of-cigarette-prices
163409237 the-effect-of-cigarette-prices
 
JOURNAL OF SEX RESEARCH. 145-158, 2014RRoutledgeCopyrig.docx
JOURNAL OF SEX RESEARCH. 145-158, 2014RRoutledgeCopyrig.docxJOURNAL OF SEX RESEARCH. 145-158, 2014RRoutledgeCopyrig.docx
JOURNAL OF SEX RESEARCH. 145-158, 2014RRoutledgeCopyrig.docx
 
nihms799718
nihms799718nihms799718
nihms799718
 
The Importance of Identifying Characteristics Underlyingthe .docx
The Importance of Identifying Characteristics Underlyingthe .docxThe Importance of Identifying Characteristics Underlyingthe .docx
The Importance of Identifying Characteristics Underlyingthe .docx
 
Masters thesis differential effectiveness of substance abuse treatment by j f...
Masters thesis differential effectiveness of substance abuse treatment by j f...Masters thesis differential effectiveness of substance abuse treatment by j f...
Masters thesis differential effectiveness of substance abuse treatment by j f...
 
Differential Effectiveness of Substance Abuse Treatment by Joyce Fuller
Differential Effectiveness of Substance Abuse Treatment by Joyce FullerDifferential Effectiveness of Substance Abuse Treatment by Joyce Fuller
Differential Effectiveness of Substance Abuse Treatment by Joyce Fuller
 
Cannabis Smoke Does Not Harm Your Lungs Like Tobacco Smoke Says New Study
Cannabis Smoke Does Not Harm Your Lungs Like Tobacco Smoke Says New StudyCannabis Smoke Does Not Harm Your Lungs Like Tobacco Smoke Says New Study
Cannabis Smoke Does Not Harm Your Lungs Like Tobacco Smoke Says New Study
 
Title of ProjectPresenter NameUniversity nameIntroduction .docx
Title of ProjectPresenter NameUniversity nameIntroduction .docxTitle of ProjectPresenter NameUniversity nameIntroduction .docx
Title of ProjectPresenter NameUniversity nameIntroduction .docx
 
Seminar paper 5
Seminar paper 5Seminar paper 5
Seminar paper 5
 
Palazzolo 2013 (electronic cigarette)
Palazzolo 2013 (electronic cigarette)Palazzolo 2013 (electronic cigarette)
Palazzolo 2013 (electronic cigarette)
 
Layfield-Christopher-BHR-Baltimore-2014
Layfield-Christopher-BHR-Baltimore-2014Layfield-Christopher-BHR-Baltimore-2014
Layfield-Christopher-BHR-Baltimore-2014
 
EconomicBenefits_of_Drug_Trx_02.05_
EconomicBenefits_of_Drug_Trx_02.05_EconomicBenefits_of_Drug_Trx_02.05_
EconomicBenefits_of_Drug_Trx_02.05_
 

RAE - Final Submission1

  • 1. 1 Department of Economics University of Warwick Research in Applied Economics EC331 2015 - 2016 1304164 Price or Preference? The economic relationship between alcohol and marijuana A study of American Universities from 2002-2014 Word Count: 4,833 Abstract This article investigates the relationship between marijuana prices and the use of marijuana and alcohol for American university students aged 18 - 29 in the National Survey on Drug Use and Health data from 2002 to 2014. Previous studies yield mixed evidence regarding the economic relationship and have preliminarily utilised proxies to account for pricing effects. I implement an instrumental variable specification that identifies drug use with variables that are empirically unrelated to consumption of marijuana and alcohol. Results indicate that marijuana and alcohol are gross substitutes. Alterations to the model through legal drinking age change this relationship. The analysis is enriched by observing changing desires to consume marijuana and alcohol for different age groups and those in different years of university. Exogeneity tests reveal that the standard Probit estimates for alcohol consumption are severely biased towards one. **Acknowledgements: I would like to thank Dr. Robert Akerlof for his immense support and guidance as a supervisor, Dr. Rocco d’Este, Dr. Jeremy Smith and Dr. Piotr Jelonek for their insight on empirical techniques, SAMHSA for the data and Ms. Anuria Singh, Mr. Hameem Raees Chowdhury, Ms. Sophia Karanicholas and Ms. Harshini Singh for their comments and encouragement. Any remaining errors and omission are my own.
  • 2. 2 Table of Contents ABSTRACT...........................................................................................................................................1 1. INTRODUCTION ........................................................................................................................3 2. LITERATURE AND THEORETICAL FRAMEWORK..........................................................4 2.1 RELATED LITERATURE ........................................................................................................................4 2.2 THEORETICAL FRAMEWORK...............................................................................................................5 3. DATA AND SUMMARY STATISTICS.....................................................................................6 3.1 DATASET ................................................................................................................................................6 3.2 VARIABLES.............................................................................................................................................7 3.2.1 Consumption likelihood measures.......................................................................................... 7 3.2.2 Pricing measure............................................................................................................................... 7 3.2.3 Control Variables............................................................................................................................ 8 3.2.4 Correlations among variables .................................................................................................. 8 4. RESEARCH METHODOLOGY..................................................................................................9 4.1 MAIN MODEL AND ESTIMATION METHOD.........................................................................................9 4.2 ENDOGENEITY.......................................................................................................................................9 5. RESULTS ....................................................................................................................................10 5.1 ARE MARIJUANA AND ALCOHOL ECONOMIC COMPLEMENTS OR SUBSTITUTES?.......................... 10 5.2 DO AGE AND UNIVERSITY YEAR AFFECT DESIRE TO CONSUME SUBSTANCES?.............................. 11 5.3 ARE UNIVERSITY STUDENTS OF LEGAL DRINKING AGE MORE LIKELY TO CONSUME ALCOHOL? 11 6. ROBUSTNESS CHECKS...............................................................................................................15 7. DISCUSSION AND CONCLUDING REMARKS .......................................................................15 8. APPENDIX .....................................................................................................................................17 8.1 APPENDIX I: TABLES ............................................................................................................................. 17 8.2 APPENDIX II: FIGURES .......................................................................................................................... 24 9. REFERENCES ................................................................................................................................25
  • 3. 3 1. Introduction Substance abuse involving young adults has been an evolving phenomenon characterised by growing affluent nations, requiring frequent and structured reassessment. Marijuana is being seen as less and less of a 'problem drug' and more like a stress relaxant, advocated by a rising number of young adults who have a strong impact on university students.1 There is a common perception that marijuana (whose accessibility is prominent despite its illegal status) is healthier and a less damaging option than alcohol. Due to the very nature of marijuana, combining it with alcohol is hazardous which makes it interesting to note why past research has led to viewing the pair as economic complements.2 Luthar (2003) provided an interesting outlook when she identified that inner-city students showed more frequent usage of marijuana and alcohol. Material wealth, in her opinion, had long lasting implications on the culture and psychological costs imparted onto children from affluent backgrounds. Yet prior research has yet to look at how affluence affects preferences towards the pair. There may be an underlying factor that promotes a psychological view that marijuana is deemed a higher social class means of intoxication with alcohol being viewed as the layman's substance (Thies and Register, 1993). This study investigates the economic relationship between two commonly used substances in American universities - alcohol and marijuana. By assessing own and cross-price effects, it is possible to identify whether the goods are economic complements or substitutes. Theoretical views present two alternating hypothesis. Marijuana and alcohol can be viewed as either complements or substitutes, and in both cases, support standard consumer utility maximisation.3 Empirical results from variations of a standard Probit model used to model desire to consume either marijuana or alcohol are applied to test the theoretical hypothesis. There is a particular focus as to whether an individual is of legal drinking age (LDA) affects this relationship. There is reason to believe that the pair of substances are interrelated (Cameron and Williams, 2001). Cannabis and alcohol consumption provide the similar initial euphoric effects, albeit with differing end results. Results from the analysis suggest that marijuana and alcohol are gross substitutes. However variations to the model - including testing the individual's legal status suggest that for higher marijuana prices the desire to consume alcohol, while positive, declines relative to previous prices. Factors such as gender and geographic location are insignificant in contrast to prior work. From a policymaker's perspective, this paper provides key insight into consumption patterns at university and is a stepping-stone towards drug policy development. For instance if marijuana and alcohol are in fact economic complements, a nation wide per-unit tax on marijuana will dampen binge-drinking at universities, a high-priority problem to counter in recent years (Hingson et al., 2009). 1 See Amonini (2005) for how youth perception influences substance use 2 There may be a time lag between consumption of alcohol and marijuana that would be driving this relationship. 3 See section 2.2 for Theoretical Framework on utility maximisation
  • 4. 4 2. Literature and Theoretical Framework 2.1 Related Literature The premise that alcohol and marijuana may satisfy similar consumption desires and as a result, the restriction of one, leads to the increase in consumption of the other is not new. Post the Volstead Act of 19204 , the prohibition of alcohol led to the first signs of increased recreational use of marijuana5 . DiNardo and Lemieux (1992) were the first to research whether youth substitute alcohol for marijuana6 . They used utility maximisation constraints to analyse the effects of an increase in minimum drinking age on alcohol and drug consumption and whether decriminalising marijuana had any effect on the two. They speculated how the presence of legal restraints at state level would deter marijuana or alcohol consumption and amplify the substitution relationship. The results advocated that an increase in minimum drinking age led to an increase in the use of marijuana among youth, and decriminalising marijuana led to significant declines in alcohol consumption.7 From these two results, they concluded that alcohol and marijuana were in fact economic substitutes. Alternatively, Chaloupka and Laixuthai (1994) examined the effects of pricing on alcohol substitutability with marijuana. They added information from the American Chamber of Commerce Researchers Association (ACCRA) to examine full and part pricing effects of alcohol in tandem with the legalisation of marijuana. They accounted for this relationship by monitoring traffic related fatalities to proxy for substance abuse and induced policy changes8 to identify the substitution effect between the two goods. There was an inherent thought of how the price effect would create a substitution-based relationship. However, the analysis was lacking in areas with regards to money measurement problems, where it became difficult to place an approximate value on marijuana. Later research done by Saffer & Chaloupka (1999) negated this view. Similarly, Pacula (1998), worked with pricing effects to examine whether rising beer taxes would have an impact on marijuana. She used individual demand level equations and a static (censoring) model to better test the economic relationship along with using data from the National Longitudinal Survey of Youth. Her analysis revealed that an increase in the federal tax on beer was found to have a larger unconditional decline in marijuana than alcohol, making the pair complements.9 Her view was built upon how exogenous price effects would lead to an alternative outcome to past results. The analysis extended to include a racial 4 First infringed in Chicago although implemented across the 50 states. 5 The evidence for this came from the sudden use of marijuana ‘tea bags’ in New York City. 6 Utilising large sample survey data across 43 US states from the Monitoring the Futures Survey 7 Although the latter had no effect on marijuana use 8 Moving from where marijuana is fully criminalised to where it is fully decriminalised 9 Taxes were used to proxy pricing effects as they were thought to better address policy questions and show fewer measurement errors.
  • 5. 5 break up where Pacula identified the prominence of a racial bias.10 It was interesting to note that the majority of individuals getting caught for possession were under the influence themselves, which would possibly distort the results. Supporting work by Williams et al (2001) looked at alcohol use amongst college students on the basis of Congress's Drug-Free Schools and Communities Act 1986. They utilised the Harvard School of Public Health College Alcohol Study (1993, 1997, 1999) along with the Illegal Drug Price/Purity Report (IDPPR) published by U.S. Department of Justice as a way of segmenting potential effects on the basis of marijuana quality. By examining own and cross price effects across 30 day prevalence equations, the group identified that marijuana may form a broader social trend of consumption that isn't inherently linked to alcohol but validates it as being an economic complement. People’s perception of the ‘legal cost’ of using marijuana and how the relative views could alter the cross-price elasticity was their key concern. A gender break up suggested that females showed the most prevalent effects on alcohol and marijuana consumption in regards to a full alcohol ban. The results, while significant, only highlighted the effect under certain price specifications and still left room as to whether unobservable factors (e.g. cigarette use) are driving relationship between the two variables. Indeed, Crost and Guerrero (2012) found that, by IV estimation, common influencers between alcohol and marijuana, the two goods return back to being economic substitutes. Their RDD model11 worked well to monitor the causal effect of the minimum legal drinking age on the substitutability between alcohol and marijuana with more significant effects being identified for men, contradicting the previous study done by Williams et al (2001). To the author’s knowledge, this paper is the first of its kind to directly examine pricing effects through a standardised form of marijuana, in contrast to prior results that utilised taxation, prohibition or alternative proxies to determine the economic relationship. Williams et al (2001) pioneered categorical pricing effects by observing how the economic relationship changed for different qualities of marijuana. Building on, this study focuses on how categorical pricing effects are altered by an individual's’ legal status and demographic characteristics respectively. 2.2 Theoretical Framework Consumer Utility Maximisation Standard individual utility maximisation theory purports increasing utility in respect to increased consumption. Rational consumers have monotonically increasing consumption functions, don't like high prices, and that this effect is amplified for university students who on average are on a restricted budget (usually on student loans). 10 African-Americans were more likely to be accosted and arrested for possession 11 Regression Discontinuity Design models were first introduced by Thistlethwaite and Campbell (1960) as a way of identifying treatment effects on individuals in a non-experimental way. Treatment was determined by whether the observable variable passed a 'cut-off' point.
  • 6. 6 The following effects can be identified: ∂F(Consumption Function) = "Positive" ∂goodi ∂F(Consumption Function) = "Negative" ∂pricei This study restricts utility maximisation to maximising consumer surplus and discounts for exogenous factors (such as altruism) that may affect individual utility. Purchasing behaviour is driven by prices, with consumer surplus shifting outwards for an increase in price by a substitute and inwards for a complement. Wealth Effects and Changing Preferences Individual responsiveness to price changes in relation to overall net wealth. The absolute elasticity of prices for relatively low value goods is higher for individuals in the lower income bracket. A simple example of this would be how a middle-class individual considers eating at a restaurant an occasional luxury whereas for someone wealthier it is more common. There exists an upper bound impact of price increases, after which the consideration of price as an influence on consumption behaviour automatically goes close to 0. This would primarily occur for Veblen goods12 where individuals wealthy enough to purchase it would not factor in price but rather alternative influencers such as celebrity promotion etc. Marijuana, being primarily an illegal substance with high standard prices, would possibly qualify for this wealth effect. Literature suggests that young adults from wealthy backgrounds tend to purchase marijuana with greater frequency, possibly due to the status associated with possession. Given their background, price is unlikely to significantly affect their purchasing behaviour. Furthermore, the illegality of the drug would push up its price, allowing mostly those wealthy young adults to purchase them at a regular frequency. 3. Data and Summary Statistics 3.1 Dataset Data is obtained on individual consumption patterns and pricing response to illicit substances from the National Survey on Drug Use and Health (NSDUH).13 The initial sample contains all individuals sampled for the period 2002-2014. The analysis is restricted to individuals at university, and within the age of 18-29 years to reduce anomaly effects. The final sample consists of 6,308 observations over the 12-year period. First launched in 1999, the NSDUH (formerly the National Household Survey on Drug Abuse) primarily measures prevalence and correlates of drug use within the United States. Using 12 A luxury good whose price does not follow the usual laws of demand and supply 13 Courtesy of the Substance Abuse and Mental Health Services Administration (SAMHSA).
  • 7. 7 the NSDUH, this study extracts a rich set of socio-demographic and consumption characteristics at individual level. Notably this includes the price last paid for marijuana, the main variable of interest.14 Responses are categorised into pricing brackets depending on amount paid.15 Restrictions on personal data made it difficult to obtain state specific geographic information that would have provided an interesting extension to the analysis. Future studies could test pricing effects on States that have legalised marijuana. 3.2 Variables 3.2.1 Consumption likelihood measures Our primary measures for consumption of alcohol or marijuana are MJOYR2 and ALCYR, binary variables recording the likelihood of consuming marijuana or alcohol in the past year respectively. Self reported values have a tendency to be either under or over-reported (Fisher and Katz, 2000). However, the nature of the data accounts for this by providing anonymity and financial motivation to individuals sampled to encourage truthful responses. The model used is an analogous variation to the model outlined by Williams et al. (2001), with the binary variables simplifying the existing model in literature. The likelihood of consumption of the substance in the previous year was used as a proxy for utility derived by consumption. An implicit assumption of this study is that consuming marijuana or alcohol gives positive utility otherwise individuals would choose not to consume it. Refer to Appendix16 for summary statistics of consumption likelihood measures. The average likelihood an individual has consumed marijuana and alcohol in the past year was 44.2% and 98.6%, respectively, over the whole sampling period. On average this means that an individual selected at random is nearly always likely to have consumed alcohol in the past year compared to marijuana, which has roughly 50-50 chances. 3.2.2 Pricing measure The key independent variable in this study is marijuana prices. Prior studies have all attempted to establish the economic relationship via proxy methods, either through indirect pricing effects by taxation (Pacula, 1998) or via legal implications on consumption (DiNardo and Lemieux, 1992). However due to the nature of the dataset, we're able to obtain amounts last paid for marijuana. This provides a more robust pricing outlook and lets us examine the direct effect of pricing on the economic relationship between the two. Panel B of Table 1 presents summary statistics of the pricing measure. The average price paid last for marijuana is between $11 to $20.99. 14 Derived from the question "How much did you pay for the marijuana you bought this last time?". 15 See Table 1 in Appendix for detailed variable descriptions. 16 Panel A of Table 2
  • 8. 8 3.2.3 Control Variables To provide a more robust analysis, key control variables are included that fall in line with recent consumer preference literature (Caulkins & Pacula, 2006). These are broken up into individual characteristics (represented by age and sex), locality (represented by university year and county type) and purchasing capability (represented by whether the individual is of legal drinking age or not). Panel C of Table 1 provides summary statistics for the control variables used in this study. For instance, 62.2% of the sample was of legal drinking age during the time of survey. The variables are designed to test variation in the pricing relationship. 3.2.4 Correlations among variables As a fundamental check for multi-collinearity, Table 3 in the appendix reports the correlations amongst all the independent variables in the analysis. A general rule of thumb, correlation values over 0.8 in absolute terms suggest possible multi-collinearity within variables (Gujarati, 1995). Table 2 shows that the highest correlation coefficient of 0.84 between AGE2 and legal. Since legal is a proxy for an age-induced shock to an individual's capability to purchase substances, this relationship is expected to occur and unlikely to affect the overall analysis to the variables of interest.
  • 9. 9 4. Research Methodology 4.1 Main model and estimation method The following is the main model regression for how marijuana prices affect the likelihood of consuming marijuana: 𝑝𝑟𝑜𝑏(𝑀𝐽𝑂𝑌𝑅2 = 1) = 𝛽0 + 𝛽1 𝐴𝐺𝐸2𝑖 + 𝛽2 𝐼𝑅𝑆𝐸𝑋𝑖 + 𝛽3 𝐼𝑅𝐸𝐷𝑈𝐶2𝑖 + 𝛽4 𝐶𝑂𝑈𝑁𝑇𝑌𝑃2𝑖 + 𝛽5 𝑀𝐽𝐶𝐴𝑇𝐸𝐺𝑖 + 𝛽6 𝑙𝑒𝑔𝑎𝑙𝑖 + 𝜀𝑖 The following is the main model regression for how marijuana prices affect the likelihood of consuming alcohol: 𝑝𝑟𝑜𝑏(𝑀𝐽𝑂𝑌𝑅2 = 1) = 𝛽0 + 𝛽1 𝐴𝐺𝐸2𝑖 + 𝛽2 𝐼𝑅𝑆𝐸𝑋𝑖 + 𝛽3 𝐼𝑅𝐸𝐷𝑈𝐶2𝑖 + 𝛽4 𝐶𝑂𝑈𝑁𝑇𝑌𝑃2𝑖 + 𝛽5 𝑀𝐽𝐶𝐴𝑇𝐸𝐺𝑖 + 𝛽6 𝑙𝑒𝑔𝑎𝑙𝑖 + 𝜀𝑖 MJCATEG is a categorical variable showing last price paid for marijuana. The price paid on average tends to be normally distributed with a slight skew to the right. AGE2, IRSEX, IREDUC2, COUTYP2 and legal are five sets of control variables related to age, gender, educational year (either freshman, sophomore or senior), county type and legal characteristics of the individual respectively. There are two estimation methods commonly used in addiction and pricing literature. One is a Probit model, controlling for environment effects and the other is a base latent variable regression adapted from the model developed by Williams et al (2001). Due to the nature of the data, individual level observations over time are unavailable and would restrict sampling size to the period at which the individual was at university. Therefore, the Probit regression on pooled cross sectional data is favoured as the main regression method. The linear probability model, analogous to the Probit model, is avoided due to inconsistency and bias in generated estimates as they are not bound to the unit interval (Horrace and Oaxaca, 2006). The reported standard errors are adjusted for potential heteroscedasticity. 4.2 Endogeneity There is concern between endogeneity of marijuana prices and the desire to consume either alcohol or marijuana, since marijuana prices can affect the decision to consume the two goods but also affect an individual's purchasing power (Galea et al., 2007). An alternative model specification uses instrument-variables ("IVs") to estimate the main regression model via the two-stage least squares ("2SLS") method. The IVs used need to meet the model exogeneity and instrument relevance conditions. Considering the price paid for marijuana can be affected by the price of loose marijuana (MMLSPCAT) and the quantity purchased in grams (MMLSGMS), the two alternative variables are utilised as IVs for MJCATEG.
  • 10. 10 5. Results 5.1 Are marijuana and alcohol economic complements or substitutes? Initial regressions look at the pricing effect of marijuana on the desire to consume either marijuana or alcohol17 . Individual Probit regressions are used with results outlined in columns (1) and (4). Results modelling through instrument-variable estimation and for a Bivariate Probit analysis are outlined in columns (2) and (5), and (3) and (6) respectively.18 Constant sample size is used and robust standard errors are used throughout. The pricing effects are statistically significant and suggest an inverse relationship as expected (Pacula et al., 2014). For instance, a rise in price bracket from <$5 (default category in the analysis) to $5 - $10 reduces desire to consume marijuana by 20% and increases it for alcohol by 62%. This fits in with standard utility maximisation theory outlined in section 2.2. At gross level, marijuana and alcohol are observed to be economic substitutes. Specific to alcohol consumption, there is a maximum pricing effect before a reduction in impact of increased marijuana price, on the desire to consume alcohol (Glaeser et al., 2008). Lower priced marijuana categories showed a more significant effect in contrast to higher prices, which declined in impact as further control variables, were introduced (Tsuang et al., 2001). Concerning the IVs estimation method, the instruments are tested to meet the exogeneity and relevance conditions. Hansen's J instrument test19 is employed to examine if the IVs mention in section 4.2 meet the exogeneity requirement. The null hypothesis is that the IV's are not correlated with the error terms in each model. The regressions measured by ALCYR and MJOYR2 both produce insignificant J statistics, measured at the 10% significance level, failing to reject the null hypothesis. The relevance of each instrument is then examined, treating MJCATEG as a continuous variable and undertaking a standard linear regression. The F-test statistic is significant at the 1% level indicating that the instruments are jointly significant. However, MMLSGMS individual F-test statistic is insignificant, indicating a potentially weak instrument problem. The results based on IVs estimation indicate that marijuana prices continue to have a negative impact on desire to consume marijuana and a positive impact on desire to consume alcohol. Although, it is important to interpret the magnitude and significance level of MJCATEG's coefficient with caution since weak IVs may cause estimators to perform poorly. A bivariate Probit analysis is run to test whether the decision to consume alcohol or marijuana could potentially be related through the error term. The correlation coefficient between the bivariate outcomes is -0.28 and is significant at the 1% significance level. The decisions are therefore interdependent and should be estimated via bivariate analysis rather than independently. However, the coefficient effects are nearly identical and for simple analysis, independent Probit models would not alter the validity of the model. The robustness of this is further discussed in section 6. Noteworthy observations are that gender effects are insignificant although women are more likely to consume marijuana than their male counterparts (Booth & Nolen, 2012).20 Also 17 Outlined in Table 3 18 See table 4 in the Appendix for Instrument Relevance test 19 Also know as the over-identification test 20 Women are considered to be more experimental during university as compared to men.
  • 11. 11 individual’s legal status is highly important in determining whether they consume marijuana or alcohol (Yörük & Yörük, 2011). This is further analysed in section 5.3. 5.2 Do age and university year affect desire to consume substances? To enrich the analysis, preferences changing across age and university year are analysed. Marginal effects for each standard model across AGE2 and IREDUC2 are plotted and can be seen in Figures 2 and 3. The graphs are created by considering age and university year as categorical variables and plotting marginal effects for each. General trends of Figure 1(a) and 1(b) suggests that there is an inverse and direct cohort effect induced by university year on desire to consume marijuana and alcohol respectively. Purchasing preferences may be subject to social norms invoked by the individuals you interact with on a regular basis (Perkins, 2002). Within university, this may be peer-led decisions, usually drawn from what is deemed socially acceptable at present time. This effect is amplified by the inherent social culture of drinking promoted through US fraternities and sororities (Wechsler et al., 2009). These social organisations usually offer selective post-degree networks that are often highly prized and sought after (Marmaros and Sacerdote, 2002). Due to this, college individuals may inherently self-select alcohol over marijuana as a required trait of these clubs (Phua, 2011). Figures 2(a) and 2(b) suggest that there is a direct and inverse age induced effect on desire to consume marijuana and alcohol respectively. The initial spike in the desire to consume alcohol once an individual turns 21 is further analysed in section 5.3. Older students on general are more likely to have consumed alcohol although their consumption trends move towards greater moderation with age (Engs and Hanson, 1986). 5.3 Are university students of legal drinking age more likely to consume alcohol? Perception of legal implications clearly plays an important role when deciding which substance to consume (Williams et al, 2001). The standard Probit model is broken up by individuals who are not of legal drinking age and by those who are.21 Figure 3(a) and 3(b) suggest that there is a straightforward substitution effect for those not of legal drinking age. Increased marijuana price decreases desire to consume marijuana and increases desire to consume alcohol. This effect is amplified for marijuana, for those who are not of legal drinking age. Figure 4(a) suggests that this effect is retained for marijuana for those of LDA. Theoretically, once an individual is given the freedom to undertake an action, there is a spike in desire to do so followed by a slow decline (Fromme et al., 2010). The action in this case is the ability to purchase either alcohol or marijuana. Being of LDA can have a significant impact on consumers switching preference away from marijuana to alcohol. Prior to being of LDA, purchasing alcohol and marijuana had the same legal implications if the individual were to get caught. However, the results suggest that alcohol is now the preferred choice. There is a legal bias towards the good where the individual's ability to purchase the substance legally, fuels his choice (Fromme et al., 2010). Figure 4(b) suggest that for those of legal drinking age, there seems to be a positive relationship between alcohol and marijuana initially, till a maximum is reached, after which a decline in 21 Statutory law in the US indicates that the general legal drinking age is 21.
  • 12. 12 likelihood of consumption (complements). A possible “Wealth Effect” is identified. Pricing effects are dampened for wealthy individuals. (Glaeser et al, 2008). An alternative outlook is legal bias. Those who are of legal drinking age are more likely to consume alcohol and less likely to be concerned with marijuana price (Yörük & Yörük, 2011). However the effect of legal bias is only temporary. Once the individual gets accustomed to their newfound freedom, the desire to consistently purchase alcohol over marijuana solely due to legal ability dies out. Legal comfort is experienced and the individual returns back to his original consumption preferences. Table 3 Pricing Effects of Marijuana on desire to consume Marijuana or Alcohol (MJOYR2) Desire to Consume Marijuana (ALCYR) Desire to Consume Alcohol Standard Model (1) IVs Model (2) Bivariate Model (3) Standard Model (4) IVs Model (5) Bivariate Model (6) MJCATEG $5 - $10.99 $11 - $20.99 $21 - $50.99 $51 - $100.99 >$101 -0.20* (0.12) -0.57*** (0.11) -0.70*** (0.11) -0.82*** (0.12) -0.96*** (0.12) -0.17*** (0.02) -0.20* (0.12) -0.57*** (0.11) -0.70*** (0.11) -0.82*** (0.12) -0.96*** (0.12) 0.62*** (0.19) 0.81*** (0.19) 0.76*** (0.18) 0.64*** (0.20) 0.62*** (0.21) 0.02 (0.04) 0.63*** (0.19) 0.81*** (0.19) 0.75*** (0.18) 0.63*** (0.20) 0.63*** (0.21) AGE2 0.11*** (0.02) 0.12*** (0.02) 0.11*** (0.02) -0.19*** (0.06) -0.19*** (0.06) -0.20*** (0.06) IRSEX 0.03 (0.03) 0.04 (0.03) 0.03 (0.02) -0.09 (0.09) -0.08 (0.09) -0.10 (0.09) IREDUC2 -0.08*** (0.03) -0.9*** (0.03) -0.08*** (0.03) 0.32*** (0.07) 0.34*** (0.07) 0.33*** (0.07) COUTYP2 -0.07*** (0.02) -0.07*** (0.02) -0.07*** (0.02) 0.10 (0.07) 0.10 (0.07) 0.10 (0.07) legal -0.17*** (0.06) -0.17*** (0.06) -0.17*** (0.06) 0.48*** (0.17) 0.47*** (0.17) 0.48*** (0.17) Rho -0.03 (0.02) -0.28*** (0.05) 0.07 (0.05) -0.28*** (0.05) Observations 6308 6308 6308 6308 6308 6308 Pseudo R2 0.03 0.06 The left panel reports the pricing effects of marijuana on the desire to consume marijuana. The right panel reports the pricing effects of marijuana on the desire to consume alcohol. *p<0.1; ** p<0.05; *** p<0.01
  • 13. 13 Figure 1(a) University year impact on desire to consume marijuana Figure 2(a) Age impact on desire to consume marijuana Figure 1(b) University year impact on desire to consume alcohol Figure 2(b) Age impact on desire to consume alcohol
  • 14. 14 Figure 3(a) Pricing effects on marijuana consumption for those not of LDA Figure 3(b) Pricing effects on alcohol consumption for those not of LDA Figure 4(a) Pricing effects on marijuana consumption for those of LDA Figure 4(b) Pricing effects on alcohol consumption for those of LDA
  • 15. 15 6. Robustness Checks This section covers checking the model's validity in contrast to using a naive estimator to simply gauge consumption preferences. The 'Hit-and-run' results for this can be see in Figure 5 (a) and 5(b) in the Appendix. 98.6% all individuals at university in the study have consumed alcohol at some point during the past year. This is not surprising (Johnston et al., 2008) and would mean that, in general, regardless of how well specified your model is, simply assuming all individuals consume alcohol would be the best way to predict consumption patterns. However, since the empirical analysis is targeted at pricing effects on standard consumption, the model described still has valid implications. What's interesting to note is that in comparison to the naive estimator of marijuana consumption22 , the model acts as a better predictor for individual behaviour. Robust standard errors are used to account for heteroscedasticity. A bivariate Probit analysis is undertaken to test whether the decision to consume marijuana and alcohol can be related through the error term (Greene, 1984). This can be seen in table 1. The results do suggest this and under normal scenarios it would make sense to use a bivariate analysis as the main model. However, as mentioned earlier, the purpose of this study is not to create the best-fit model but to explicitly look at how pricing affects purchasing decisions (Yamada et al., 1993). As a simple direct tool, independent Probit equations can be used to undertake the analysis (Bray et al., 2000). 7. Discussion and Concluding Remarks The relationship between alcohol and marijuana is interesting. Empirical results suggest that the goods could be, both, economic complements or substitutes under specific scenarios. The legal drinking age is an important factor that affects the pricing effect of marijuana on alcohol. Generally, an individual who is of legal drinking age is more likely to consume alcohol rather than marijuana. Social factors such as university year, age and sex provide a more robust outlook at the economic relationship identified. Standard policy implications stand towards stricter enforcement on alcohol sale around university areas. Empirically a large proportion of students were observed to be engaging in underage drinking. Building upon what Pacula et al. (2014) suggested, it would be beneficial to look towards legalising marijuana and then heavily taxing consumption to deter substance abuse. Limitations to the analysis include the lack of personal data, which distorts the model's accuracy. Reverse causality is another concerning issue. It is still unclear as to whether individual motivation to consume substances affects price or vice versa. Due to the nature of the dataset this paper is unable to account for changing consumption patterns over time.23 Extensions to this paper would include testing individual specific motivation over time in tandem with additional demographic variables such as family wealth. It would also be 22 Assuming all individuals have not consumed marijuana. 23 This could be ignored considering that alcohol consumption at university has remained at nearly 100% over the years.
  • 16. 16 worthwhile to check how consumption patterns differ over states where purchasing marijuana is legal. The paper finds significant evidence supporting the hypothesis that marijuana and alcohol are at gross, substitutes. However legal implications clearly play an important role an induce variation into this economic relationship. Future tax policy changes should allow for flexibility, adapting to suit the situational economic relationship at hand.
  • 17. 17 8. Appendix 8.1 Appendix I: Tables Table 1: Description of the Variables Used Variable Description Panel A: Consumption Likelihood Measures MJOYR2 Binary Variable; equals to 1 if marijuana was consumed in the last 12 months and 0 otherwise ALCYR Binary Variable; equals to 1 if alcohol was consumed in the last 12 months and 0 otherwise Panel B: Pricing Measure MJCATEG The average categorical price paid for last using marijuana. This is broken up into 6 categories of under $5, $5 - $10.99, $11 - $20.99, $21 - $50.99, $51 - $100.99, >=$101 with numerical assignment of 1, 2, 3, 4, 5 and 6 respectively. Under $5 is the assumed default price for our analysis. Panel C: Control Variables AGE2 Categorical variable for individuals aged between 19 -29. This is broken down into 6 categories of 19, 20, 21, 22 or 23, 24 or 25 and 26 to 29 with numerical assignment of 8, 9, 10, 11, 12, 13 respectively. IRSEX Gender Variable; assumes a value of 1 for male and 2 for female. IREDUC2 Describes individual current education year at university. Broken up into freshmen, sophomores and seniors with numerical assignment as 9, 10 and 11 respectively COUTYP2 Geographic indicator. Individual can be in either a large metropolitan area, small metropolitan area, or a non metropolitan area with numerical assignment 1, 2 and 3 respectively. legal Dummy Variable; equals to 1 if individual is of legal drinking age and 0 otherwise
  • 18. 18 Table 2: Summary Statistics Variable Obs Mean Std. Dev. Min Max Panel A: Consumption Likelihood Measures MJOYR2 6308 44.2% .5 0 1 ALCYR 6308 98.6% .12 0 1 Panel B: Pricing Measure MJCATEG 6308 3.72 1.2 1 6 Panel C: Control Variables AGE2 6308 10.1 1.43 8 13 IRSEX 6308 1.40 0.49 1 2 IREDUC2 6308 9.83 0.67 9 11 COUTYP2 6308 1.66 0.71 1 3 legal 6308 62.2% 0.49 0 1 Table 3: Correlation Matrix 1 2 3 4 5 6 7 8 1 MJOYR2 1.00 2 ALCYR -0.07 1.00 3 MJCATEG -0.17 0.03 1.00 4 AGE2 0.03 0.01 0.15 1.00 5 IRSEX 0.04 -0.01 -0.14 -0.01 1.00 6 IREDUC2 -0.02 0.06 0.12 0.52 0.00 1.00 7 COUTYP2 -0.05 0.02 0.07 -0.02 -0.04 -0.02 1.00 8 legal 0.01 0.03 0.11 0.84 -0.01 0.49 -0.013 1.00 This table presents the correlation matrix for all the independent variables employed in this study.
  • 19. 19 Table 4: Instrument Relevance Testing MJCATEG (Categorical Pricing Variable) Standard Model MMLSGMS -0.0001 (0.00) MMLSPCAT 0.041*** (0.00) Observations 6308 R2 0.8328 *p<0.1; ** p<0.05; *** p<0.01 Table 5: Age effects on desire to consume alcohol or marijuana (MJOYR2) Desire to Consume Marijuana (ALCYR) Desire to Consume Alcohol Standard Model Standard Model MJCATEG $5 - $10.99 $11 - $20.99 $21 - $50.99 $51 - $100.99 >$101 -0.19 (0.12) -0.56*** (0.11) -0.69*** (0.11) -0.82*** (0.12) -0.96*** (0.12) 0.62*** (0.19) 0.81*** (0.19) 0.76*** (0.18) 0.64*** (0.20) 0.61*** (0.21) AGE2 20 21 22/23 24/25 26-29 0.01 (0.05) 0.01 (0.06) 0.07 (0.06) 0.20*** (0.07) 0.47*** (0.09) -0.10 (0.13) 0.14 (0.16) 0.02 (0.16) -0.17 (0.17) -0.49*** (0.18) IRSEX 0.03 (0.03) -0.09 (0.09) IREDUC2 -0.06** (0.03) 0.31*** (0.08) COUTYP2 -0.07*** (0.02) 0.10 (0.07) Observations 6308 6308 Pseudo R2 0.03 0.06 *p<0.1; ** p<0.05; *** p<0.01
  • 20. 20 Table 6: Testing Age Effect at Margins (MJOYR2) Desire to Consume Marijuana (ALCYR) Desire to Consume Alcohol At Margins At Margins AGE2 19 20 21 22/23 24/25 26-29 0.41*** (0.02) 0.42*** (0.02) 0.42*** (0.02) 0.44*** (0.01) 0.49*** (0.02) 0.60*** (0.03) 0.99*** (0.00) 0.99*** (0.00) 0.99*** (0.00) 0.99*** (0.00) 0.98*** (0.00) 0.97*** (0.01) *p<0.1; ** p<0.05; *** p<0.01 Table 7: University Year Effect on desire to consume alcohol or marijuana (MJOYR2) Desire to Consume Marijuana (ALCYR) Desire to Consume Alcohol Standard Model Standard Model MJCATEG $5 - $10.99 $11 - $20.99 $21 - $50.99 $51 - $100.99 >$101 -0.20* (0.12) -0.57*** (0.11) -0.70*** (0.11) -0.82*** (0.12) -0.96*** (0.12) 0.62** (0.19) 0.81*** (0.19) 0.76*** (0.18) 0.64*** (0.20) 0.62*** (0.21) AGE2 0.11*** (0.02) -0.19*** (0.06) IRSEX 0.03 (0.03) -0.09 (0.09) IREDUC2 Sophomore Senior -0.13*** (0.04) -0.13** (0.06) 0.31*** (0.09) 0.67*** (0.18) COUTYP2 -0.07*** (0.02) 0.10 (0.07) legal -0.15** 0.49***
  • 21. 21 (0.06) (0.17) Observations 6308 6308 Pseudo R2 0.03 0.06 *p<0.1; ** p<0.05; *** p<0.01 Table 8: Testing University Year Effect at Margins (MJOYR2) Desire to Consume Marijuana (ALCYR) Desire to Consume Alcohol At Margins At Margins IREDUC2 Freshman Sophomore Senior 0.47*** (0.01) 0.42*** (0.01) 0.42*** (0.02) 0.98*** (0.00) 0.99*** (0.00) 1.00*** (0.00) *p<0.1; ** p<0.05; *** p<0.01 Table 9: Price Effect on desire to consume alcohol or marijuana for those not of LDA (MJOYR2) Desire to Consume Marijuana (ALCYR) Desire to Consume Alcohol Standard Model Standard Model MJCATEG $5 - $10.99 $11 - $20.99 $21 - $50.99 $51 - $100.99 >$101 -0.29 (0.18) -0.72*** (0.18) -0.89*** (0.18) -1.10*** (0.19) -1.1*** (0.20) 0.68*** (0.24) 1.00*** (0.25) 0.89*** (0.25) 1.20*** (0.34) 1.30*** (0.42) AGE2 0.04 (0.06) -0.17 (0.16) IRSEX 0.01 (0.05) -0.10 (0.13) IREDUC2 -0.10* (0.06) 0.43** (0.18) COUTYP2 -0.11*** (0.04) 0.26** (0.12) Observations 2387 2387 Pseudo R2 0.04 0.09 *p<0.1; ** p<0.05; *** p<0.01
  • 22. 22 Table 10: Testing Pricing Effect for those not of LDA at Margins (MJOYR2) Desire to Consume Marijuana (ALCYR) Desire to Consume Alcohol At Margins At Margins MJCATEG <$5 $5 - $10.99 $11 - $20.99 $21 - $50.99 $51 - $100.99 >$101 0.72*** (0.06) 0.62*** (0.02) 0.45*** (0.02) 0.39*** (0.02) 0.30*** (0.03) 0.29*** (0.03) 0.90*** (0.04) 0.98*** (0.01) 0.99*** (0.00) 0.99*** (0.00) 0.99*** (0.00) 1.00*** (0.01) *p<0.1; ** p<0.05; *** p<0.01 Table 11: Price Effect on desire to consume alcohol or marijuana for those of LDA (MJOYR2) Desire to Consume Marijuana (ALCYR) Desire to Consume Alcohol Standard Model Standard Model MJCATEG $5 - $10.99 $11 - $20.99 $21 - $50.99 $51 - $100.99 >$101 --0.13 (0.16 ) --0.46*** (0.15 ) --0.57*** (0.15 ) --0.65*** (0.15 ) --0.83*** (0.16 ) 0.68** (0.34) 0.59** (0.30) 0.57** (0.29) 0.27 (0.30) 0.24 (0.31) AGE2 0.13*** (0.02) -0.19*** (0.06) IRSEX 0.04 (0.04) -0.09 (0.12) IREDUC2 -0.06* (0.03) 0.28*** (0.08) COUTYP2 -0.04 (0.03) -0.01 (0.09) Observations 3921 3921 Pseudo R2 0.02 0.06 *p<0.1; ** p<0.05; *** p<0.01
  • 23. 23 Table 12: Testing Pricing Effect for those of LDA at Margins (MJOYR2) Desire to Consume Marijuana (ALCYR) Desire to Consume Alcohol At Margins At Margins MJCATEG <$5 $5 - $10.99 $11 - $20.99 $21 - $50.99 $51 - $100.99 >$101 0.65*** (0.05) 0.59*** (0.02) 0.47*** (0.02) 0.42*** (0.01) 0.39*** (0.02) 0.32*** (0.02) 0.97*** (0.02) 0.99*** (0.00) 0.99*** (0.00) 0.99*** (0.00) 0.98*** (0.00) 0.98*** (0.01) *p<0.1; ** p<0.05; *** p<0.01
  • 24. 24 8.2 Appendix II: Figures Figure 5(a): Testing model validity - Probit Model Table 5(b): Testing model validity - Naive Estimator Marijuana Use Frequency Percent No 3523 55.85 Yes 2785 44.15 Total 6308 100 Alcohol Use Frequency Percent No 88 1.4 Yes 6220 98.60 Total 6308 100
  • 25. 25 9. References Amonini, C. "The Relationship between Youth's Moral and Legal Perceptions of Alcohol, Tobacco and Marijuana and Use of These Substances." Health Education Research 21.2 (2005): 276-86. Print. Bailey, Jennifer A., Karl G. Hill, Meredith C. Meacham, Susan E. Young, and J. David Hawkins. "Strategies for Characterizing Complex Phenotypes and Environments: General and Specific Family Environmental Predictors of Young Adult Tobacco Dependence, Alcohol Use Disorder, and Co- occurring Problems." Drug and Alcohol Dependence 118.2-3 (2011): 444-51. Print. Booth, Alison, and Patrick Nolen. "Choosing to Compete: How Different Are Girls and Boys?" Journal of Economic Behavior & Organization 81.2 (2012): 542-55. Print. Booth, Alison L., and Patrick Nolen. "Gender Differences in Risk Behaviour: Does Nurture Matter?*." The Economic Journal 122.558 (2012). Print. Bray, Jeremy W., Gary A. Zarkin, Chris Ringwalt, and Junfeng Qi. "The Relationship between Marijuana Initiation and Dropping out of High School." Health Econ. Health Economics 9.1 (2000): 9- 18. Print. Cameron, Lisa, and Jenny Williams. "Cannabis, Alcohol and Cigarettes: Substitutes or Complements?" Economic Record 77.236 (2001): 19-34. Print. Caulkins, J. P., and R. L. Pacula. "Marijuana Markets: Inferences from Reports by the Household Population." Journal of Drug Issues 36.1 (2006): 173-200. Print. Chaloupka, Frank, and Adit Laixuthai. "Do Youths Substitute Alcohol and Marijuana? Some Econometric Evidence." (1994). Print. Crost, Benjamin, and Santiago Guerrero. "The Effect of Alcohol Availability on Marijuana Use: Evidence from the Minimum Legal Drinking Age." Journal of Health Economics 31.1 (2012): 112-21. Print.
  • 26. 26 Desimone, Jeff. "Illegal Drug Use and Employment." Journal of Labor Economics 20.4 (2002): 952-77. Print. Dinardo, John, and Thomas Lemieux. "Alcohol, Marijuana, and American Youth: The Unintended Effects of Government Regulation." (1992). Print. Engs, Ruth C., and David J. Hanson. "Age-Specific Alcohol Prohibition And College Students' Drinking Problems." Psychological Reports 59.2 (1986): 979-84. Print. Fisher, Robert J., and James E. Katz. "Social-desirability Bias and the Validity of Self-reported Values." Psychology and Marketing Psychol. Mark. 17.2 (2000): 105-20. Print. Fromme, Kim, Reagan R. Wetherill, and Dan J. Neal. "Turning 21 and the Associated Changes in Drinking and Driving After Drinking Among College Students." Journal of American College Health 59.1 (2010): 21-27. Print. Galea, Sandro, Jennifer Ahern, Melissa Tracy, and David Vlahov. "Neighborhood Income and Income Distribution and the Use of Cigarettes, Alcohol, and Marijuana." American Journal of Preventive Medicine 32.6 (2007). Print. Glaeser, Edward, Joseph Gyourko, and Albert Saiz. "Housing Supply and Housing Bubbles." (2008). Print. Greene, William H. "Estimation of the Correlation Coefficient in a Bivariate Probit Model Using the Method of Moments." Economics Letters 16.3-4 (1984): 285-91. Print. Gujarati, Damodar N. Basic Econometrics. New York: McGraw-Hill, 1995. Print. Hanson, David J., and Ruth C. Engs. "College Students' Drinking Problems: A National Study, 1982- 1991." Psychological Reports 71.1 (1992): 39. Print. Hingson, Ralph W., Wenxing Zha, and Elissa R. Weitzman. "Magnitude of and Trends in Alcohol- Related Mortality and Morbidity Among U.S. College Students Ages 18-24, 1998-2005."Journal of Studies on Alcohol and Drugs, Supplement J. Stud. Alcohol Drugs Suppl. S16 (2009): 12-20. Print. Horrace, William C., and Ronald L. Oaxaca. "Results on the Bias and Inconsistency of Ordinary Least Squares for the Linear Probability Model." Economics Letters 90.3 (2006): 321-27. Print.
  • 27. 27 "Intoxicating Liquors. Eighteenth Amendment. Interpretation of the Volstead Act." Harvard Law Review 34.4 (1921): 437. Print. Johnston, Lloyd D., Patrick M. O'malley, Jerald G. Bachman, and John E. Schulenberg. "Monitoring the Future: National Results on Adolescent Drug Use: Overview of Key Findings, 2008." PsycEXTRA Dataset. Print. Manning, Paul. Drugs and Popular Culture: Drugs, Media and Identity in Contemporary Society. Cullompton, Devon, England: Willan Pub., 2007. Print. "Marginal Effects in the Bivariate Probit Model." By William H. Greene. Web. 14 Apr. 2016. Marmaros, David, and Bruce Sacerdote. "Peer and Social Networks in Job Search." European Economic Review 46.4-5 (2002): 870-79. Print. "National Survey on Drug Use and Health." Encyclopedia of Substance Abuse Prevention, Treatment, & Recovery. Print. Pacula, Rosalie Liccardo. "Adolescent Alcohol and Marijuana Consumption: Is There Really a Gateway Effect?" (1998). Print. Perkins, H. Wesley. "Social Norms and the Prevention of Alcohol Misuse in Collegiate Contexts." Journal of Studies on Alcohol, Supplement J. Stud. Alcohol Suppl. S14 (2002): 164-72. Print. Phua, Joe. "The Influence of Peer Norms and Popularity on Smoking and Drinking Behavior among College Fraternity Members: A Social Network Analysis." Social Influence 6.3 (2011): 153-68. Print. Saffer, Henry, and Frank Chaloupka. "Tobacco Advertising: Economic Theory and International Evidence." (1999). Print. Sevigny, Eric L., Rosalie Liccardo Pacula, and Paul Heaton. "The Effects of Medical Marijuana Laws on Potency." International Journal of Drug Policy 25.2 (2014): 308-19. Print. Solomon, Michael R., Rebekah Russell-Bennett, and Josephine Previte. Consumer Behaviour: Buying, Having, Being. Frenchs Forest, N.S.W.: Pearson Australia, 2013. Print. Tsang, Hector W.h., Ashley Chan, Alvin Wong, and Robert Paul Liberman. "Vocational Outcomes of an Integrated Supported Employment Program for Individuals with Persistent and Severe Mental
  • 28. 28 Illness." Journal of Behavior Therapy and Experimental Psychiatry 40.2 (2009): 292-305. Print. Tsuang, Ming T., Jessica L. Bar, Rebecca M. Harley, and Michael J. Lyons. "The Harvard Twin Study of Substance Abuse: What We Have Learned." Harv Rev Psychiatry Harvard Review of Psychiatry 9.6 (2001): 267-79. Print. Wechsler, Henry, George Kuh, and Andrea E. Davenport. "Fraternities, Sororities and Binge Drinking: Results from a National Study of American Colleges." Journal of Student Affairs Research and Practice 46.3 (2009): 763-84. Print. "When Can You Safely Ignore Multicollinearity? | Statistical Horizons." Statistical Horizons. Web. 22 Apr. 2016. Williams, Jenny, Rosalie Liccardo Pacula, Frank Chaloupka, and Henry Wechsler. "Alcohol and Marijuana Use Among College Students: Economic Complements or Substitutes?" (2001). Print. Yamada, Tetsuji, Michael Kendix, and Tadashi Yamada. "The Impact of Alcohol Consumption and Marijuana Use on High School Graduation." (1993). Print. Yörük, Barış K., and Ceren Ertan Yörük. "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 30.4 (2011): 740-52. Print. Zarkin, Gary A., Thomas A. Mroz, Jeremy W. Bray, and Michael T. French. "The Relationship between Drug Use and Labor Supply for Young Men." Labour Economics 5.4 (1998): 385-409. Print