International Journal of Drug Policy 43 (2017) 122–129

Contents lists available at ScienceDirect

International Journal of Drug Policy

journal homepage: www.elsevier.com/locate/drugpo

Research paper

Cannabis decriminalization and the age of onset of cannabis use§

Jakub Červený a, Pavla Chomynová b,c, Viktor Mravčı́k b,d,c, Jan C. van Ours e,f,g,h,*
a
Department of Economics, CentER, Tilburg University, The Netherlands
b
National Monitoring Center for Drugs and Addiction, The Office of the Government of the Czech Republic, Czech Republic
c
National Institute of Mental Health, Czech Republic
d
Department of Addictology, First Faculty of Medicine, Charles University and General University Hospital, Czech Republic
e
Erasmus School of Economics, Erasmus University Rotterdam, The Netherlands
f
Department of Economics, University of Melbourne, Australia
g
Tinbergen Institute, Amsterdam/Rotterdam, The Netherlands
h
CEPR, London, United Kingdom

A R T I C L E I N F O A B S T R A C T

Article history: Background: In the Czech Republic in 2010 a law was introduced decriminalizing personal possession of
Received 17 February 2016 small quantities of several illicit drugs, including cannabis.
Received in revised form 27 January 2017 Methods: We use 2012 survey data to examine the effect of a change in cannabis policy on the age of
Accepted 24 February 2017
onset of cannabis use. We estimate the effect of the policy change using a mixed proportional hazards
framework that models the transition to first cannabis use.
Keywords: Results: The change in cannabis policy did not affect the transition to first cannabis use.
Cannabis
Conclusion: We find no evidence of cannabis decriminalization affecting the age of onset of cannabis use.
Decriminalization
Age of onset
ß 2017 Elsevier B.V. All rights reserved.

Introduction Despite the prohibition policy, cannabis use has increased over
the past decades and there is a debate on whether this policy is
Cannabis has been and still is an illegal drug in almost all sensible (see for example Caulkins, Hawken, Kilmer, & Kleiman,
countries across the world. The main argument for prohibition of 2012; Cawley & Ruhm, 2011; Pudney, 2010). The cannabis policy
cannabis is the potential health risk associated with cannabis use. debate is often emotional, with strong views of both proponents
However, negative health effects of cannabis use are no robust and opponents (van Ours, 2012). Those who are in favor of
finding. For example, van Ours and Williams (2015) conclude from legalization tend to ignore the negative health effects of cannabis
an overview of the literature that there do not appear to be serious use. Those who are against legalization ignore the fact that legal
harmful health effects of moderate cannabis use. Only heavy use by substances such as alcohol and tobacco also have bad health effects
individuals who are susceptible to mental health problems may (see for example Hall & Lynskey, 2009; Nutt, King, & Phillips, 2010;
have negative effects on the mental well-being of these individua- Taylor et al., 2012). The debate on legalizing cannabis has gained
ls. This does not imply that cannabis use is harmless (see also Hall, momentum in recent years. Uruguay and several U.S. states have
2015). The age of onset of cannabis use is important as there is legalized cannabis use, allowing consumption and regulating
robust evidence that early cannabis use for example reduces supply. Other U.S. states and other countries have decriminalized
educational attainment. the possession of small quantities of cannabis or made assess to
cannabis for medical reasons easier.
Whether easier access leads to an increase in cannabis use is not
clear. Research on the relationship between cannabis policy and
§
The authors thank seminar participants at Tilburg University and an cannabis use varies from general comparative cross-country
anonymous referee for excellent comments on previous versions of this paper. studies to in-depth analysis of differences in cannabis use of
Chomynová and Mravčı́k gratefully acknowledge a grant from the Internal Grant individuals who are subject to different policy regimes. Among the
Agency of the Czech Ministry of Health No. NT/14064-3 and the project Nr. LO1611 general comparative studies is Reinarman, Cohen, and Kaal (2004)
with a financial support from the MEYS under the NPU 1 program.
who compare representative samples of experienced cannabis
* Corresponding author at: Erasmus School of Economics, Erasmus University
Rotterdam, The Netherlands. users in decriminalized-cannabis-use Amsterdam and criminal-
E-mail address: [email protected] (J.C. van Ours). ized-cannabis-use San Francisco finding no evidence to support

http://dx.doi.org/10.1016/j.drugpo.2017.02.014
0955-3959/ß 2017 Elsevier B.V. All rights reserved.
 J. Červený et al. / International Journal of Drug Policy 43 (2017) 122–129 123

claims that criminalization reduces use or evidence that decrimi- particularly interesting and important as many of the negative
nalization increases use. Reuter and Trautmann (2009) find that effects of cannabis use are related to an early age of onset. For
drug policies have limited effects on drug problems. Drug policies example, Lynskey et al. (2003) conclude that individuals who used
seem to be influenced by the drug situation rather than the other cannabis by age 17 had higher odds of other drug use and alcohol
way around. A study by the European Monitoring Centre for Drugs dependence than their co-twins, who did not use cannabis before
and Drug Addiction (2011) explores whether a significant change age 17. In our analysis, we exploit information on the age of onset
in the prevalence of cannabis use among the population aged to model transitions to first cannabis use. For this, we use data from
15–34 can be observed after a legislative change regarding a 2012 survey. We find that the policy change did not affect the age
cannabis use. Analyzing data from eight countries that changed of onset of cannabis use. To investigate the robustness of our
their cannabis policy in the past ten years (Italy, UK, Slovakia, findings we also use data from a 2008 survey as a counterfactual
Denmark, Finland, Portugal, Bulgaria and Greece) it is concluded analysis finding that indeed the ‘‘cannabis policy change that did
that cannabis legislation did not affect cannabis use. Pacula and not happen’’ did not affect the age of onset of cannabis use either.
Sevigny (2014a) argue that perceived harms or risks, changes in
social norms, changes in ability to access cannabis and changes in Background
the organization of supply may all influence the effect of cannabis
decriminalization on cannabis consumption. Pacula and Sevigny Change in cannabis policy
(2014a, 2014b) argue that a more liberal cannabis use policy does
not necessarily translate into a higher incidence of cannabis use. Shortly after the fall of communist regime in 1989, the Czech
The effects depend on whether or not cannabis supply is also penal code was revised to remove repressive practices of the
affected. If not, cannabis decriminalization leading to increased previous regime. Illicit drug possession was not a crime from
demand will increase cannabis price thus dampening the effects of 1990 to 1998. With the development of drug problems during the
cannabis consumption. Empirically, there is also an issue of timing 1990s, social and political concerns originated for a more
of events. Once cannabis is formally decriminalized it make take repressive approach in the Czech drug policy. As a result, the
some time before this decriminalization is implemented. penal code was amended defining the possession of drugs for
Another strand of studies on the relationship between cannabis personal use as a criminal offense and introducing the term
policy and cannabis use focuses on the U.S. in which some states ‘‘greater than small’’ quantity as a threshold distinguishing
have medical marijuana dispensaries which make access to between a criminal offense and an administrative offense. The
cannabis easy (In the U.S. cannabis is usually referred to as interpretation of the term ‘‘greater than small’’ was left to judicial
marijuana). The findings in these studies are not uniform. Some practice. The ‘‘greater than small’’ quantity became a focus of
studies conclude that easier access to cannabis through the debate on illicit drug regulation and prosecution in the Czech
dispensaries has a positive effect on cannabis use while other Republic (Radimecky, 2007; Zabransky, 2004; Zeman, 2007).
studies find no effect whatsoever. Pacula, Kilmery, Grossman, and In 2001, the government decided to differentiate drugs in the
Chaloupka (2010) conclude that in states where medical marijuana penal code according to their health and social risks – initially in
laws were introduced cannabis use increased. Wall et al. (2011) three, and finally in two groups – cannabis and other drugs
find that states with medical marijuana laws have higher rates of (Zabransky (2004)). Due to a complex re-codification of the old
cannabis use. Chu (2015) concludes that cannabis arrest rates penal code, this decriminalization of cannabis possession was
significantly increased after medical marijuana laws were passed. implemented as late as January 2010. The penalties for possession
However, Cerdá, Wall, Keyes, Galea, and Hasin (2012) conclude of cannabis and for growing cannabis for personal use were
that cannabis abuse and cannabis dependence rates among substantially reduced. Table 1 provides a summary overview. For
cannabis users are very similar in states with and without medical example, possession of ‘‘greater than small’’ quantities of cannabis
marijuana laws. Harper, Strumpf, and Kaufman (2012) find could result in a jail sentence of up to one year from 2010 onward
medical marijuana laws not to have increased cannabis use. while before 2010 this could have been a jail sentence of up to two
Anderson, Hansen, and Rees (2015) and Anderson and Rees (2014) years. Similarly, growing of cannabis for personal use in ‘‘greater
also find no evidence that medical marijuana dispensaries than small’’ quantities before 2010 could have been punished with
increased cannabis use. Wagenaar, Lynne-Landsman, and Living- 1–5 years in prison while from 2010 onward this was up to
ston (2013) find that neither the prevalence rate nor the frequency 6 months. For small quantities, the maximum penalty for
of cannabis use seem to have been affected by the dispensaries. possession of cannabis of s 550 did not change. However, for
Finally, there are studies on the effect of decriminalization on the growing of small quantities of cannabis for personal use the
the uptake of cannabis use. Williams and Bretteville-Jensen (2014) penalty changed from 1 to 5 years in prison to a maximum fine of s
analyze Australian data exploiting variation of the timing of 550. Under the new law, possession of less than 15 g (or five plants)
decriminalization over Australian States finding that cannabis of herbal cannabis and 5 g of hashish was not considered a criminal
decriminalization affected the uptake of cannabis among young- offense. Substantial changes were also introduced with regard to
sters in the first five years following the policy change. In the cannabis cultivation. Before the policy change, growing for
Netherlands, consumption of cannabis is quasi-legalized since the cannabis even for personal use could have been treated as drug
mid 1970s. Small quantities of cannabis can be bought in cannabis production and punishable with penalties from 1 to 5 years in
shops, retail outlets which are referred to as coffeeshops. These prison. The new law introduced in 2010 no longer considered
coffeeshops are subject to strict rules. Some of the fundamental growing of small quantities as drug production and amended the
rules are: no sale of hard drugs, no advertising, no sale to maximum penalty as a fine up to CZK 15,000 (s 550).
youngsters below 18 years of age, no sale above 5 g per transaction To summarize, cannabis possession was legal between
and no more than 500 g of cannabis on the premises. Palali and van 1990 and 1998, illegal between 1998 and 2010, and decriminalized
Ours (2015) find that individuals who grew up within 20 km of a for personal possession since January 2010. The focus of our
cannabis shop have a lower age of onset of cannabis use. analysis is on the effect of the decriminalization law passed in
In our paper, we focus on the Czech Republic where in 2010 a 2010. The 2010 intervention began in 2001 but was not fully
legislative change was introduced decriminalizing cannabis implemented until 2010. In our analysis, we investigate whether
possession. The question we address in our paper is how this the formal change in cannabis law had an effect on the uptake of
policy change affected the uptake of cannabis use. This is cannabis use.
124 J. Červený et al. / International Journal of Drug Policy 43 (2017) 122–129

Table 1
Penalties for possession of cannabis and growing of cannabis for personal use.

Before 2010 From 1.1.2010 onward

Panel A. Possession of cannabis

Small quantities Maximum fine of CZK 15,000 (s 550) Maximum fine of CZK 15,000 (s 550)
Quantities greater than small Up to 2 years in prison Up to 1 year in prison
Significant scale 1–5 years 6 months to 5 years
Substantial scale Was not defined 2–8 years

Panel B. Growing of cannabis for personal use

Small quantities 1–5 yearsa Maximum fine of CZK 15,000 (s 550)
Quantities greater than small 1–5 yearsa Up to 6 months
Significant scale 2–10 yearsa Up to 3 years
Substantial scale 8–12 yearsa 6 months to 5 years
Large scale 10–15 yearsa Not defined

Source: Ministry of Justice.

Note: Growing of cannabis for personal use could be regarded as a ‘‘possession crime’’, especially if the scale was small and intention for
personal use was clear (see Panel A.).
a
Applied if treated as drug production.

As Csete (2012) remarks, the new cannabis policy aligned the last 30 days. These numbers do not show large fluctuations. The
Czech Republic with a growing number of EU countries that prevalence of use in last 12 months is decreasing, falling from
effectively decriminalized some cannabis offenses. In July 2013, 15.3 percent in 2008 to 9.2 percent in 2012. Apart from 2008,
the Constitutional Court annulled substantial parts of the cannabis use in last 30 days is relatively stable at a level of around
aforementioned regulation as it was found contradictory to the four percent.
Constitution of the Czech Republic and the Charter of Fundamental
Rights and Freedoms, according to which any criminal offense (and
Methods
thus also the definition of greater-than-small quantity of a narcotic
or psychotropic substances) may only be defined by a law (Mravcik
Data
et al. (2013)). As a consequence, the Supreme Court decreased the
threshold limit for herbal cannabis from 15 to 10 g; see also
In our analysis, we use data from two surveys carried out in the
Mravcik (2015).
Czech Republic from October to December 2008 and from
Fig. 1 shows lifetime prevalence of cannabis use from 2008 to
September to November 2012 by the Czech NMC, the National
2012. There is a substantial variation from close to 35 percent in
Monitoring Center for Drugs and Addiction. The main goals were to
2008 to about 25 percent in 2011. Lifetime prevalence of cannabis
provide information on the extent of substance use and attitudes
use is relatively high in Czech Republic. It was 27.4 percent in
towards psychotropic substances and to determine the extent of
2012 compared to the European average of 17.6 percent. Once an
selected health risk behaviors associated with illicit drugs in the
individual has used cannabis, by definition he or she will always be
Czech population. The questionnaires are based on the European
an ever user of cannabis. A change in lifetime prevalence only
Model Questionnaire, a set of standard questions recommended
occurs at the margin, by older individuals leaving the sample
for general population surveys by the European Monitoring Center
because of death or emigration or younger individuals starting to
for Drugs and Drug Addiction (EMCDDA). For our paper, we use
use cannabis. Therefore, lifetime prevalence can only change
information on the age of onset of cannabis use which is available
slowly over time. The fluctuations in Fig. 1 are ‘‘noise’’ rather than
only in the 2008 and 2012 surveys. To ensure the anonymity of
‘‘signal’’, most likely introduced by changes in the sampling frame.
the respondents, the names and addresses were not recorded. The
Fig. 1 also shows changes in cannabis use in the last 12 months and
data were collected by means of face-to-face interviews with
[(Fig._1)TD$IG] respondents in randomly selected households using a paper
questionnaire (PAPI). Both samples were obtained by multi-stage
stratified sampling procedures and are nationally representative
for the Czech population aged 15–64 years with regard to gender,
age, region and level of education. All individuals participating in
the study had to be of Czech nationality. Further information about
the sampling design, weighting and stratification is provided in the
Appendix.
When studying the effect of a policy change on the uptake of
cannabis use, it makes sense to focus on younger generations.
Furthermore, for the 2012 sample we need to exclude individuals
who might have started using cannabis during the 1990–1998
period when cannabis was legal. Therefore, in the remainder of our
paper we focus in individuals of age 25 or younger at the time of
the survey. The appendix provides the definitions of the relevant
variables in our analysis and presents descriptive statistics.

Cannabis use dynamics

Fig. 1. Cannabis use trends in the Czech Republic individuals age 15–64; 2008–
Starting to use cannabis is a phenomenon that is highly age-
2012. related. Individuals most often decide on the use of cannabis when
Source: National Monitoring Center for Drugs and Addiction. they are in the age range 15–25. Individuals who have never used
 J. Červený et al. / International Journal of Drug Policy 43 (2017) 122–129 125

cannabis by age 25 are very unlikely to start using cannabis later on Set-up of the analysis
in life (van Ours (2005)). In many countries cannabis use among
younger generations is substantially higher than among older The focus of our analysis is on the effect of cannabis
generations simply because cannabis was a rare commodity when decriminalization on the uptake of cannabis. Since individuals
older generations grew up (see for example European Monitoring were asked about the age of their first use of cannabis and also
Centre for Drugs & Drug Addiction, 2011). their age at the time of survey, we are able to determine the time-
In the following empirical analysis, we focus on the starting frame in which they might have been affected by the new
rate of cannabis use. We use the age of first cannabis use to policy. Using retrospective information to establish a calendar
calculate the starting rate by age, from age 10 onward, taking into year effect is less sensitive to sampling procedures as the
account that some individuals have never used cannabis. For these information comes from one survey. Of course, retrospective
individuals we assume that the duration until cannabis use is information is subject to recollection errors but this is unlikely to
right-censored at their survey age. Panel A of Fig. 2 plots the be important since we focus on a sample of young individuals for
evolution of the unconditional starting rate over the age of the whom events concerning cannabis use have happened only
individuals. Cannabis use starts at age 11 but only a few percent of recently.
the individuals do this at such a young age. There are differences To estimate the effect of the policy change, we use a mixed
between the starting rates in the samples of 2008 and proportional hazard framework which allows the rate of transition
2012. The starting rates in 2008 are higher than in 2012. The from one state to the other to be affected by individual observed
peak in the starting rates of 2008 is at age 16, while this is age 18 for and unobserved characteristics as well as the duration of stay in a
the sample of 2012. From age 21 inwards the starting rates are very state (Heckman and Singer (1984)). In our case, the state of origin is
low. Panel B of Fig. 2 shows the related cumulative starting never having used cannabis while the destination state is cannabis
probabilities of cannabis use. Clearly, the cumulative starting use. The transition rate is equivalent to the age-specific starting
probabilities level off after age 20. For the 2008 sample the rate of cannabis use. The duration of stay in the first state is
cumulative starting probability by age 20 is 62 percent, for the equivalent to the age of the individual from age 10 inwards. Thus,
2012 sample this is 48 percent. the starting rate for cannabis use at age t conditional on observed
characteristics x, the age at which the new cannabis policy was
[(Fig._2)TD$IG] introduced tp and unobserved characteristics u, is specified as
follows:

uðtjx; t p ; uÞ ¼ lðtÞ expðx0 b þ dIðt > tp Þ þ uÞ (1)

where l(t) represents individual age dependence and b is a vector

of parameters. The observed characteristics we include in the
analysis are gender, education, birth-cohort and region of
residence. The birth-cohort variable accounts for potential trends
in the uptake of cannabis across birth-cohorts but does not account
for secular trends over time. The parameter d describes how the
hazard rate shifts at the age when the new law was introduced in
the year 2010 and thus measures the effect of the policy change on
the uptake of cannabis. Age dependence is flexibly modeled using a
step function:
!
X
lðtÞ ¼ exp lk Ik ðtÞ (2)
k

where k(= 1, . . ., K) is a subscript for age-intervals and Ik(t) are time-

varying dummy variables for subsequent age-intervals. We
assume that individuals are being exposed to cannabis from age
10 inwards. The first age interval is 10–14, subsequent age
intervals are annually specified from age 15 to age 20, and the last
interval refers to ages over 21. We estimate a constant and
normalize l0 = 0. Note that we are able to make a distinction
between age dependence and policy effect because the 2010
policy affected individuals at a different age. Nevertheless, we are
aware of the fact that the effects of the policy change may have
been contaminated by other policy changes that occurred around
2010.
The conditional density function for the completed durations of
non-use can be written as:
 Z t 
f ðtjx; t p ; uÞ ¼ uðtjx; t p ; uÞ exp  uðsjx; tp ; uÞ ds (3)
0

We assume that the random effects u come from a discrete

distribution G with two points of support (u1, u2), representing two
types of individuals who differ in unobserved characteristics. We
also investigated whether we could identify a third type of
Fig. 2. Cannabis use starting rates and cumulative starting probabilities by age. individuals but were unable to do so. The associated probabilities
126 J. Červený et al. / International Journal of Drug Policy 43 (2017) 122–129

are denoted as follows: Pr(u = u1) = p1, Pr(u = u2) = p2, where pj where K denotes dataset consisting of i = 1, . . ., N individuals, di
(j = 1, 2) is assumed to have a logistic distribution: denotes an indicator whether an individual started using cannabis
and F is the distribution function related to f.
expðan Þ To check the robustness of our findings we also estimate the
pn ¼ P ; n ¼ 1; 2 (4)
n expðan Þ
same model on 2008 data. By way of counterfactual analysis we
introduced in 2006 a ‘‘policy change that did not occur’’. This is
with a1 normalized to zero. Because we include a constant we also about three years before the survey, similar to the 2010 policy
normalize u1 = 0. We remove the unobserved heterogeneity change that occurred about three years before the 2012 survey.
distribution through integration:
Z Results
f ðtjx; t p Þ ¼ f ðtjx; t p ; uÞ dGðuÞ (5)
u Baseline estimates

In the estimation we take into account that we do not know the Table 2 reports the parameter estimates obtained by using the
birthday of the individual nor the exact day at which an individual method of Maximum Likelihood. Column (1) show the estimates
started using cannabis. So, if an individual indicated having used based on the 2012 data, column (2) present the results of the
cannabis for the first time at age 17, this could be at the 17th counterfactual analysis based on the 2008 data and column (3)
birthday or the day before the 18th birthday. The resulting log- presents estimates based on the joint 2008–2012 dataset. The top
likelihood function equals: row of the table shows the effect of the decriminalization law on
the starting rate of cannabis use. In the 2012 estimate, the effect is
X
N positive but statistically not different from zero at conventional
L¼ logfdi ðFðt þ 1ÞFðtÞÞ þ ð1di ÞFðt þ 1Þg (6) levels of significance. For 2008 and the joint estimate for
i¼1
2008 and 2012 the estimated effect of the decriminalization law

Table 2
Parameter estimates starting rate of cannabis use.

2012 2008 2008–2012

(1) (2) (3)

Panel A. Personal characteristics

Effect decriminalization (d) 0.51 (1.2) 0.01 (0.1) 0.29 (1.1)
Men 0.45 (1.2) 0.46** (2.2) 0.36** (2.3)
Survey 2012 0.80*** (3.8)

Education
Secondary 0.26 (0.5) 0.36 (1.5) 0.31 (1.5)
Secondary w. grad. 0.90** (2.0) 0.74** (3.2) 0.76*** (3.6)
Vocational 2.11*** (4.2) 0.80 (1.4) 0.63 (1.5)
University 1.27 (0.9) 0.20 (0.6) 0.40 (1.2)

Panel B. NUTS 2 Region

Central Bohemia 0.19 (0.1) 0.52* (1.9) 0.26 (1.1)
Southwest 0.13 (0.2) 0.57* (1.8) 0.46* (1.9)
Northwest 0.21 (0.2) 0.17 (0.7) 0.34 (1.4)
Northeast 0.82 (1.2) 0.16 (0.6) 0.29 (1.3)
Southeast 0.51 (0.9) 0.26 (0.9) 0.49** (2.2)
Central Moravia 0.30 (0.5) 0.28 (0.9) 0.28 (1.1)
Moravia-Silesia 1.00 (1.4) 0.20 (0.7) 0.45** (2.0)

Panel C. Age effects

Age 15 2.22** (2.4) 2.28*** (10.2) 2.27*** (8.1)
Age 16 3.23*** (3.5) 3.07*** (13.4) 3.09*** (10.7)
Age 17 3.84*** (3.7) 3.95*** (13.1) 3.82*** (11.3)
Age 18 4.95*** (3.8) 4.36*** (10.0) 4.33*** (10.4)
Age 19 5.58*** (4.4) 4.44*** (8.2) 4.36*** (9.3)
Age 20 4.91*** (3.9) 4.28*** (6.1) 3.97*** (7.1)
Age 21 5.33*** (4.8) 3.35*** (3.8) 3.54*** (5.3)
Birth-Cohort 0.10 (0.9) 0.13*** (3.5) 0.08** (2.4)

Panel D. Unobserved heterogeneity

u2 4.60*** (7.2) 2.87*** (4.7) 2.85*** (5.9)
a2 0.46** (2.2) 0.22 (0.6) 0.09 (0.3)
p2 (%) 61.4 44.5 47.8

Observations
438 1086 1524
Log likelihood
709.4 1385.8 2116.3

Constants not reported; absolute t statistics in parentheses.

*
p < 0.10.
**
p < 0.05.
***
p < 0.001.
 J. Červený et al. / International Journal of Drug Policy 43 (2017) 122–129 127

Table 3
Parameter estimates effects of the decriminalization policy on the starting rate of cannabis use; sensitivity analysis.

2012 Log L N 2008 Log L N 2008–2012 Log L N

Panel A. Baseline
0.51 (1.2) 709.4 438 0.01 (0.1) 1385.8 1086 0.29 (1.1) 2116.3 1524

Panel B. Men age  30

0.13 (0.4) 656.5 334 0.09 (0.2) 1307.6 857 0.30 (0.8) 1979.8 1191

Panel C. Women age  30

0.48 (1.6) 568.6 371 0.32 (1.2) 1060.6 815 0.09 (0.2) 1653.5 1186

Absolute t statistics in parentheses; N = number of observations.

is negative and insignificant. The remainder of Table 2 shows Table 3 shows additional parameter estimates of the effects of
how personal characteristics affect the uptake of cannabis use. the decriminalization policy when distinguishing men and
Men have a higher starting rate than women but only for women. For reasons of comparison Panel A repeats the main
2008 analysis and the joint analysis this effect is significantly parameter estimates of Table 2. As the sample is relatively small,
different from zero. In 2012 those with a vocational education have we increased the sample size up to age 30 or less. Panel B shows the
the lowest starting rate of cannabis use by approximately parameter estimates for men. For the 2008 and 2012 sample, as
(exp(2.11)  1)  100  88 percent compared to those with only well as for the joint sample, the effect of the policy remains
primary education, while in 2008 it is 55 percent. Birth-cohort has statistically insignificant. Panel C presents the parameter estimates
no effect in the 2012 analysis and a positive effect in the 2008 and for women. Again we find that neither for the actual policy change
joint 2008–2012 analysis. There are also some differences between in the 2012 sample nor for the joint estimate on the pooled
regions. Finally, there is clear age dependence in the starting rate 2008 and 2012 dataset there is a significant effect on the age of
while unobserved heterogeneity is present. In all samples, we find onset of cannabis use.
that the distribution of unobserved heterogeneity in the starting
rates can be described by a discrete distribution with two points of How to explain our findings?
support. There is one type of individuals that has a substantial
lower starting rate than the other type. This implies that some Despite the fact that cannabis was de-iure illegal between
individuals with a very low starting rate are unlikely to ever start 1999 and 2010 and decriminalized since 2010, the period of
using cannabis. The distribution of the types is different for the two prohibition left de-facto enforcement of the law up to judicial
surveys. In 2012, 61 percent have low starting rate, compared to practice, mainly due to the unclear definition of the term ‘‘greater
45 percent in 2008 and 48 in the joint estimate. We attribute these than small’’ (Belackova, Maalsté, Zabransky, and Grund (2015)). To
differences to the difference in sampling design in 2008 and 2012. illustrate how the law was enacted in practice, we provide
information about cannabis-related drug offenses and arrests
Sensitivity analysis summarized in Table 4. The first column shows that the number of
persons arrested for production, trafficking and selling increased
The main conclusion from Table 2 is that there is no significant from 608 in 2008 to 885 in 2011. Similarly, the second column
effect of the decriminalization policy on the starting rate of shows that the number of arrests for possession for personal use
cannabis use. To investigate the robustness of this finding we increased between 2008 and 2012, and almost doubled between
excluded birth-cohort as an explanatory variable. This did not 2011 and 2012. The 2010 decriminalization law somehow created
affect our main findings. We also investigated the importance of confusion and was sometimes mistakenly presented as legaliza-
regional within-country mobility by estimating the model tion by media. As a reaction, the Czech police prioritized the fight
excluding the regions as explanatory variables. This does not against drug-related crime. This is also reflected by the increasing
change our main findings either. number of persons charged for production, trafficking and selling,
where only 64% of individuals arrested in 2008 was in fact charged,
compared to 85–86% in 2011 and 2012.
However, despite the increasing number of arrests for posses-
sion for personal use, the actual number of persons charged
Table 4 remains relatively stable over 2008–2012 period. The fact that
Cannabis related arrests and charges. charges remain stable between 2010 and 2012 can possibly be
Year Production, trafficking Possession for Total explained by the introduction of the 2010 decriminalization law.
and selling personal use Individuals arrested after 2010 perhaps possessed amounts
Panel A. Number of persons arrested
smaller than the newly introduced thresholds and therefore had
2008 608 138 746 to be cleared.
2009 661 125 786 In further search for an explanation of our findings in the
2010 744a[1_TD$IF] 152 896 previous sections of the paper, we exploit two opinion questions
2011 885a 178 1063
related to cannabis use. One potential explanation of why the age
2012 870a 372 1242
of onset was not affected by the decriminalization policy is that
Panel B. Number of persons charged potential consumers may have found the access to cannabis as easy
2008 392 121 513
as before. In both 2008 and 2012 surveys respondents were asked
2009 520 116 636
2010 573a 97 670 the question ‘‘How difficult do you think it would be to obtain
2011 769a 111 880 cannabis within next 24 h?’’1 In 2008, 96.4 percent of the
2012 742a 127 869

Source: The Czech drug situation annual reports, Ministry of Justice. 1

Note that these results are based on responses from individuals aged 15–64, as
a
Includes unauthorized cultivation of plants containing a narcotic or psycho- the response to this part of questionnaire was not mandatory and therefore many
tropic substance. participants ignored it.
128 J. Červený et al. / International Journal of Drug Policy 43 (2017) 122–129

respondents indicated that it was easy or very easy to obtain Appendix: Definition & descriptives of variables
cannabis within 24 h. In 2012, after the policy change, this was
95.7 percent. Apparently the policy did not change the availability Definition of variables
of cannabis. This is not surprising as almost every respondent
indicated that cannabis was easily obtained. Cannabis: Have you ever tried cannabis (marijuana or
We also examined the perception of the risk of being caught by hashish)? If so, indicate at what age did you try cannabis for first
the police when obtaining cannabis. In both surveys respondents time.
were asked a question ‘‘Have you felt threatened the last time you Did you use cannabis within the last year? If so, did you use
obtained cannabis?’’. There is a difference in response, 21 percent cannabis within the last 30 days?
of cannabis users felt threatened by the police before the Cannabis opinions: How difficult do you think it would be to
decriminalization law was introduced. After the introduction of obtain cannabis within next 24 h? Impossible, very difficult,
the law this number decreased to 9 percent. This finding suggests relatively difficult, relatively easy, very easy, don’t know (only
that the introduction of decriminalization in 2010 had an effect on 2012 survey) – When you obtained cannabis last time, have you
perception of law enforcement but the importance of this was felt threatened while doing it? By a police, by a seller/dealer, by a
probably too small to influence the cannabis uptake behavior of side that did not take part in transaction, other threats, did not feel
youngsters. Alternatively, it is possible that the effect of the change any threat.
in cannabis policy is contaminated by other policy changes around Education: Dummy variables: Secondary: Special schools
2010 such as the National Drug Policy Strategy approved in May including technical schools, specialized in construction, chemistry,
2010 aiming for a reduction of drug use among youngsters. engineering etc. Without graduation. Secondary with graduation:
Grammar schools; higher vocational, higher specialized education,
without university diploma/degree. Higher vocational: Higher
Conclusions vocational education. University: University degree. Reference
group: Primary – compulsory up to age 15.
In 2010, as part of a new drugs policy, the Czech Republic Birth-Cohort: (Birth year – 1900)/10.
decriminalized the possession of small quantities of cannabis. We Regions: Dummy variables: Central Bohemia, Southwest,
study the effect of this decriminalization policy on the age of Northwest, Northeast, Southeast, Central Moravia, Moravia-
onset of cannabis use among youngsters. We estimate the policy Silesia; Reference group: Capital Prague.
effect using a mixed proportional hazards framework that Table A1 provides an overview of the means of the variables we
models the transition to cannabis use and allows us to distinguish use in our analysis. As shown, the 2008 sample has 1086 individu-
between the effect of observed and unobserved personal als, while the 2012 sample consists of 438 individuals. As shown,
characteristics as well as the effect of the change in cannabis the composition of the samples according to region and education
policy. We find that the policy change did not affect the age of onset are about the same, but there are differences in terms of prevalence
of cannabis use. of cannabis use. For example, in the 2008 survey 57 percent of the
We suggest several explanations of our main findings but in the respondents of age 25 and younger indicated to have ever used
end we can only speculate about the reasons why the decriminal- cannabis while in the 2012 survey this was only 45 percent. In the
ization did not affect the uptake of cannabis. It could be that 2008 survey, 36 percent of individuals used cannabis within last
initially the interpretation of the law by youngsters was wrong, year, while in 2012 it was 22 percent. The share of individuals
causing decriminalization to be confused with legalization. As a reporting cannabis use in last 30 days is also higher in 2008,
response, the police increased monitoring cannabis related reaching 21 percent compared to 11 percent in 2012.
activities which resulted in increasing numbers of cannabis-
related arrests. Another potential explanation is that after
Table A1
decriminalization cannabis lost its status as a ‘‘forbidden fruit’’ Means of variables.
and hence was no longer appealing for new users to try.
Alternatively, the uptake of cannabis did not change because 2008 2012

supply of cannabis was not affected. Cannabis consumers found it Panel A. Personal characteristics
as easy to access cannabis after the decriminalization as they did Age 20.56 20.26
Men 0.47 0.50
before the policy change. Furthermore, it might have been the case
Birth cohort 9.14 8.77
that the decriminalization changed the law but not the perceived
risks related to cannabis use. Finally, it is worth mentioning that Cannabis use
Lifetime 0.45 0.57
the lack of statistically significant results might also be caused by a
Last year 0.22 0.36
lack of power due to relatively small sample sizes. Last 30 days 0.11 0.21
All in all, the lack of evidence of an effect of cannabis
Education
decriminalization on the age of onset of cannabis is a fortunate
Primary 0.31 0.43
finding for those who worry about negative health effects related Secondary 0.17 0.26
to early onset of cannabis use. Secondary w. grad. 0.40 0.27
Whether the findings of our study carry external validity is Higher vocational 0.04 0.01
unclear. As mentioned in the introduction, similar research on University 0.08 0.03

Australian data finds contradictory results to those presented in Panel B. NUTS 2 region
this paper. The lifetime prevalence of cannabis use in Czech Prague 0.12 0.12
Central Bohemia 0.10 0.11
Republic prior to decriminalization was substantially higher than
Southwest 0.11 0.11
the average in other countries. It is conceivable that cannabis use Northwest 0.09 0.10
was already at its peak and therefore did not change significantly Northeast 0.14 0.14
even after a less stringent policy was introduced. Southeast 0.20 0.17
Central Moravia 0.11 0.12
Moravia-Silesia 0.13 0.13
Conflict of interest Observations 438 1086
The authors report no conflict of interest.
 J. Červený et al. / International Journal of Drug Policy 43 (2017) 122–129 129

Survey design Cerdá, M., Wall, M., Keyes, K. M., Galea, S., & Hasin, D. (2012). Medical marijuana laws in
50 states: Investigating the relationship between state legalization of medical
marijuana and marijuana use, abuse and dependence. Drug and Alcohol Depen-
General Population Survey on Drug Use and Attitudes dence, 120, 22–27.
Chu, Y. L. (2015). Do medical marijuana laws increase hard drug use? Journal of Law and
towards Drug Use in the Czech Republic in 2008 – Quota
Economics, 58, 481–517.
sampling: Csete, J. (2012). A balancing act: Policymaking on illicit drugs in the Czech Republic.
Technical report. New York: Open Society Foundations.
1. Voting district selection (around 14 000 for the whole Czech European Monitoring Centre for Drugs and Drug Addiction (2011). The state of the drugs
problem in Europe 2011. Annual report. Luxembourg: Publications Office of the
Republic). Quotas were assigned to reach targeted number of European Union.
respondents. Hall, W. (2015). What has research over the past two decades revealed about the
2. Household selection. Using random walk with given guidelines adverse effects of recreational cannabis use? Addiction, 110, 19–35.
Hall, W., & Lynskey, M. (2009). The challenges in developing a rational cannabis policy.
which household to choose. Current Opinion in Psychiatry, 22, 258–262.
5613 individuals contacted, 4506 agreed to be interviewed. Most Harper, S., Strumpf, E. C., & Kaufman, J. S. (2012). Do medical marijuana
people who declined to be interviewed indicated lack of time or laws increase marijuana use? Replication study and extension. Annals of Epidemi-
ology, 22, 207–212.
disinterest in the survey as a reason. After restricting the sample to Heckman, J., & Singer, B. (1984). A method for minimizing the impact of
age 25 years or younger 1086 individuals remain. distributional assumptions in econometric models for duration data. Econometrica,
Czech Republic National Survey on Substance Use 2012 – (2), 271–320.
Lynskey, M., Heath, A., Bucholz, K., Slutske, W., Madden, P., Nelson, E., et al. (2003).
Stratified random sampling: Escalation of drug use in early-onset cannabis users vs co-twin controls. JAMA,
289(4), 427–433.
1. City and town selection (23 strata). Regional stratification was Mravcik, V. (2015). (De)criminilisation of possession of drugs for personal use – A view
from the Czech Republic. International Journal of Drug Policy, 26, 705–707.
based upon NUTS2 classification and size of a residence,
Mravcik, V., Chomynova, P., Grohmannova, K., Necas, V., Grolmusova, L., Kissova, L.,
together with a registry of towns and cities collected in et al. (2013). The Czech Republic drug situation in 2012. Annual report. Urad vlady CR.
2011 national census. Nutt, D., King, L., & Phillips, L. (2010). Drug harms in the UK: A multicriteria decision
2. Street selection within a city or town. In each city/town a street analysis. Lancet, 376, 1558–1565.
Pacula, R. L., Kilmery, B., Grossman, M., & Chaloupka, F. J. (2010). Risks and prices: The
was randomly selected as initial point for questioning. In total, role of user sanctions in marijuana markets. The B.E.. Journal of Economic Analysis
177 initial points was selected, together with extra 50 for and Policy, 10(1), 1–38.
additional purposes. Pacula, R. L., & Sevigny, E. L. (2014a). Marijuana liberalization policies: Why we can’t
learn much from policy still in motion. Journal of Policy Analysis and Management,
3. Household selection. Every third household of every third 33(1), 212–221.
inhabited residence was included. Pacula, R. L., & Sevigny, E. L. (2014b). Natural experiments in a complex and dynamic
4. Respondent selection. Respondent was selected according to the environment: The need for a measured assessment of the evidence. Journal of Policy
Analysis and Management, 33(1), 232–235.
Kish selection grid. Palali, A., & van Ours, J. C. (2015). Distance to cannabis-shops and age of onset of
6210 households contacted, 2383 had unknown eligibility status cannabis use. Health Economics, 24, 1483–1501.
(none present or opened the door). 3827 had known eligibility status. Pudney, S. (2010). Drugs policy – What should we do about cannabis? Economic Policy,
61, 165–211.
Based on Kish tables, a respondent was selected (always 1 per Radimecky, J. (2007). Rhetoric versus practice in Czech drug policy. Journal of Drug
household) - 1693 were not eligible for the survey (no permanently Issues, 37, 11–44.
living respondent, different age group, language barrier, incompetent Reinarman, C., Cohen, P., & Kaal, H. (2004). The limited relevance of drug policy:
Cannabis in Amsterdam and in San Francisco. American Journal of Public Health, 94,
to answer, refused to participate when selected). There are 2134 com-
836–842.
pleted questionnaires (621023831693 = 2134). After restricting the Reuter, P., & Trautmann, F. (2009). A report on Global Illicit Drugs Markets 1998–2007.
sample to age 25 years or younger 438 individuals remain. Brussels: European Commission.
Both datasets include sampling weights to account for non- Taylor, M., Mackay, K., Murphy, J., Mcintosh, A., Mcintosh, C., Anderson, S., et al. (2012).
Quantifying the RR of harm to self and others from substance misuse: Results from
response. Weights are based on the following variables: age, a survey of clinical experts across Scotland. BMJ Open, 2(4), 1–7.
gender, education, NUTS2 region, size of residence and economic van Ours, J. C. (2005). Dynamics in the use of drugs. Health Economics, 15(12),
status. 1283–1294.
van Ours, J. C. (2012). The long and winding road to cannabis legalization. Addiction,
107, 872–873.
References van Ours, J. C., & Williams, J. (2015). Cannabis use and its effects on
health, education and labor market success. Journal of Economic Surveys, 29(5),
Anderson, D. M., Hansen, B., & Rees, D. I. (2015). Medical marijuana laws and teen 993–1010.
marijuana use. American Law and Economics Review, 17, 495–528. Wagenaar, A. C., Lynne-Landsman, S. D., & Livingston, M. D. (2013). Effects of state
Anderson, D. M., & Rees, D. I. (2014). The legalization of recreational marijuana: How medical marijuana laws on adolescent marijuana use. American Journal of Public
likely is the worst-case scenario? Journal of Policy Analysis and Management, 33, Health, 103, 1500–1506.
221–232. Wall, M. M., Poh, E., Cerda, M., Keyes, K. M., Galea, S., & Hasin, D. S. (2011). Adolescent
Belackova, V., Maalsté, N., Zabransky, T., & Grund, J. (2015). Should I buy or should I marijuana use from 2002 to 2008: Higher in States with medical marijuana laws,
grow? How drug policy institutions and drug market transaction costs shape the cause still unclear. Annals of Epidemiology, 21, 714–716.
decision to self-supply with cannabis in the Netherlands and the Czech Republic. Williams, J., & Bretteville-Jensen, A. L. (2014). Does liberalizing cannabis laws increase
International Journal of Drug Policy, 26, 296–310. cannabis use? Journal of Health Economics, 36, 20–32.
Caulkins, J., Hawken, A., Kilmer, B., & Kleiman, M. (2012). Marijuana legalization: What Zabransky, T. (2004). Czech drug laws as an arena of drug policy battle. Journal of Drug
everyone needs to know. Oxford, UK: Oxford University Press. Issues, 34, 661–686.
Cawley, J., & Ruhm, C. J. (2011). The economics of risky health behaviors. In M. V. Pauly, Zeman, P. (2007). Legislation and practice concerning prosecution of drug offenses in
T. G. Mcguire, & P. P. Barros (Eds.), Handbook of health economics, volume 2 of the Czech Republic. Journal of Drug Issues, 37, 45–72.
handbook of health economics (pp. 95–199). Elsevier.