ArticlePDF Available

Unemployment, drugs and attitudes among European youth

Authors:
  • Fundação Instituto de Pesquisas Econômicas

Abstract and Figures

This paper studies changes in the patterns of drug consumption and attitudes towards drugs in relation to sky-high (youth) unemployment rates brought about by the Great Recession. Our analysis is based on data for 28 European countries that refer to young people. We find that the consumption of cannabis and 'new substances' is positively related to increasing unemployment rates. An increase of 1% in the regional unemployment rate is associated with an increase of 0.7 percentage points in the ratio of young people who state that they have consumed cannabis at some point in time. Our findings also indicate that higher unemployment may be associated with more young people perceiving that access to drugs has become more difficult, particularly access to ecstasy, cocaine and heroin. According to young Europeans, when the economy worsens, anti-drug policies should focus on the reduction of poverty and unemployment, and not on implementing tougher measures against users.
Content may be subject to copyright.
Journal
of
Health
Economics
57
(2018)
236–248
Contents
lists
available
at
ScienceDirect
Journal
of
Health
Economics
jo
u
r
n
al
homep
age:
www.elsevier.com/locate/econbase
Unemployment,
drugs
and
attitudes
among
European
youth
Sara
Ayllón,
Natalia
N.
Ferreira-Batista
Department
of
Economics
&
EQUALITAS,
University
of
Girona,
Spain
a
r
t
i
c
l
e
i
n
f
o
Article
history:
Received
3
October
2016
Received
in
revised
form
10
August
2017
Accepted
21
August
2017
Available
online
26
August
2017
JEL
classifications:
I12
K42
E32
D12
Keywords:
Unemployment
Drugs
Youth
Attitudes
Effective
policies
Great
recession
Europe
a
b
s
t
r
a
c
t
This
paper
studies
changes
in
the
patterns
of
drug
consumption
and
attitudes
towards
drugs
in
relation
to
sky-high
(youth)
unemployment
rates
brought
about
by
the
Great
Recession.
Our
analysis
is
based
on
data
for
28
European
countries
that
refer
to
young
people.
We
find
that
the
consumption
of
cannabis
and
‘new
substances’
is
positively
related
to
increasing
unemployment
rates.
An
increase
of
1%
in
the
regional
unemployment
rate
is
associated
with
an
increase
of
0.7
percentage
points
in
the
ratio
of
young
people
who
state
that
they
have
consumed
cannabis
at
some
point
in
time.
Our
findings
also
indicate
that
higher
unemployment
may
be
associated
with
more
young
people
perceiving
that
access
to
drugs
has
become
more
difficult,
particularly
access
to
ecstasy,
cocaine
and
heroin.
According
to
young
Europeans,
when
the
economy
worsens,
anti-drug
policies
should
focus
on
the
reduction
of
poverty
and
unemployment,
and
not
on
implementing
tougher
measures
against
users.
©
2017
The
Authors.
Published
by
Elsevier
B.V.
This
is
an
open
access
article
under
the
CC
BY-NC-ND
license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
1.
Introduction
If
there
is
an
age
group
that
has
been
particularly
hit
by
the
Great
Recession
in
Europe,
it
is
certainly
that
of
young
people.
Youth
unemployment
rates
have
reached
unprecedented
levels
in
many
countries,
labour
market
opportunities
have
clearly
worsened,
and
the
careers
of
many
young
people
have
been
abruptly
interrupted.
The
consequences
of
these
deteriorating
labour
market
conditions
in
several
different
spheres
of
life
currently
form
the
subject
of
This
paper
has
received
funding
from
the
European
Union’s
Horizon
2020
Research
and
Innovation
Programme
under
grant
agreement
no.
649395,
project
title:
NEGOTIATE
Overcoming
early
job-insecurity
in
Europe.
Sara
Ayllón
also
acknowledges
support
from
the
projects
ECO2013-46516-C4-1-R,
ECO2016-76506-
C4-4-R
and
2014-SGR-1279
and
is
very
grateful
for
the
warm
hospitality
received
in
the
Department
of
Social
Sciences
at
the
University
of
Eastern
Finland,
where
this
paper
was
revised.
Participants
at
the
Negotiate
meeting
in
Brighton
(March
2016),
at
the
Early
Job
Insecurity
Workshop
in
Poznan
(September
2016)
and
at
the
41st
Symposium
of
the
Spanish
Economic
Association
in
Bilbao
(December
2016)
are
thanked
for
their
useful
comments.
We
would
also
like
to
thank
Richard
Williams
at
the
University
of
Notre
Dame
for
answering
our
questions
regarding
the
Stata
command
gologit2.
Any
errors
or
misinterpretations
are
our
own.
Corresponding
author
at:
C/Universtat
de
Girona
10.
17003,
Girona,
Spain.
E-mail
address:
sara.ayllon@udg.edu
(S.
Ayllón).
analysis
of
much
research
i.e.
career
prospects,
the
possibility
for
young
people
to
leave
the
parental
home
and
set
up
their
own
fam-
ilies,
subjective
well-being,
etc.
This
paper
adds
to
this
literature
by
analysing
the
extent
to
which
changes
in
the
labour
market
have
also
translated
into
changes
in
the
patterns
of
drug
consumption
and
youth
attitudes
toward
drugs.
Drug
consumption
among
European
youth
is
not
a
minor
problem:
17.8
million
young
adults
(15–34)
used
drugs
in
2015
according
to
the
European
Drug
Report
2016,
published
by
the
European
Monitoring
Centre
for
Drugs
and
Drug
Addiction
(EMCDDA,
2016).
The
same
study
estimates
that
cannabis
was
used
in
2015
by
16.6
million
young
adults
that
is,
by
13.3%
of
the
age
group.
In
the
case
of
cocaine,
the
figure
was
2.4
million
(1.9%),
and
for
ecstasy
(MDMA)
and
amphetamines,
2.1
million
and
1.3
million
(1.7%
and
1.0%),
respectively.
Moreover,
it
is
estimated
that
8%
of
the
youngest
group
(15–24)
have
used
new
psychoactive
substances
at
some
time.
The
same
source
estimates
that
EUR
24.3
billion
were
spent
in
2013
on
illicit
drugs
in
Europe;
that
there
were
1.6
million
https://doi.org/10.1016/j.jhealeco.2017.08.005
0167-6296/©
2017
The
Authors.
Published
by
Elsevier
B.V.
This
is
an
open
access
article
under
the
CC
BY-NC-ND
license
(http://creativecommons.org/licenses/by-nc-nd/4.
0/).
S.
Ayllón,
N.N.
Ferreira-Batista
/
Journal
of
Health
Economics
57
(2018)
236–248
237
reported
drug
offences
in
2014;
and
that
at
least
6800
overdose
deaths
occurred
that
same
year
(EMCDDA,
2016).1
But
why
should
changes
in
the
labour
market
be
related
to
changes
in
drug
consumption
or
attitudes
toward
drugs?
The
theo-
retical
literature
analysing
the
link
between
the
business
cycle
and
substance
use
highlights
three
causal
pathways:
an
‘income
effect’,
‘economic
stress’
and
an
‘opportunity
cost’
(also
called
‘substitution
effect’).
The
‘income
effect’
asserts
that
if
drugs
are
normal
goods,
then
in
a
bad
economy
consumers
should
adapt
their
demand
to
a
tighter
budget,
and
therefore
reduce
their
consumption.
Evidence
for
this
pro-cyclical
result
can
be
found
in
Neumayer
(2004),
Tapia
Granados
(2005),
Gerdtham
and
Ruhm
(2006),
Charles
and
DeCicca
(2008),
Catalano
et
al.
(2011),
Freeman
(1999),
Ruhm
and
Black
(2002),
Ruhm
(2005),
Ásgeirsdóttir
et
al.
(2012)
and
Xu
(2013),
and
references
within.2
The
‘economic
stress’
mechanism
links
substance
consumption
to
psychological
reasons.
In
this
case,
to
deal
with
uncertainty
about
future
income,
the
increased
probability
of
being
unemployed
or
the
lack
of
opportunities
found
in
the
labour
market,
young
peo-
ple
may
resort
to
self-medication,
which
causes
an
increase
in
substance
use.
Furthermore,
economic
recessions
can
change
the
‘opportunity
cost’
of
substance
consumption:
given
that
fewer
jobs
are
available
and
wages
are
lower,
spending
time
using
drugs
has
a
smaller
opportunity
cost,
which
may
enhance
consumption.3
Böckerman
and
Ilmakunnas
(2009),
Arkes
(2011,
2012),
Dee
(2001),
Bradford
and
Lastrapes
(2014)
and
Currie
and
Tekin
(2015)
present
evidence
for
such
counter-cyclical
results.4All
in
all,
the
rela-
tionship
between
substance
use
and
the
business
cycle
will
be
pro-cyclical
if
the
income
effect
offsets
the
other
two
mechanisms,
and
counter-cyclical
if
it
is
the
other
way
round.
This
paper
is
the
first
to
provide
evidence
of
the
relationship
between
sky-high
(youth)
unemployment
rates
brought
about
by
the
Great
Recession
and
drug
use
among
young
people
across
Europe
(28
countries).
More
importantly,
our
study
goes
beyond
an
analysis
based
solely
on
consumption
to
analyse
certain
atti-
tudes
towards
drugs
that
have
been
identified
in
the
literature
as
good
predictors
of
future
drug
consumption.
In
particular,
we
analyse
how
changes
in
the
local
labour
market
may
have
had
an
effect
on
the
availability
of
drugs
and
on
young
people’s
perceived
access
to
them;
on
the
perceived
risk
of
drug
consumption;
and
on
youth
opinion
as
to
the
most
effective
ways
of
combating
the
problems
that
drugs
cause
in
society.
The
analysis
of
good
pre-
dictors
is
important
because,
particularly
when
interviewed
about
drugs,
individuals
may
be
reluctant
to
admit
actual
consumption.
Moreover,
and
as
explained
by
Bachman
et
al.
(1990),
any
com-
plete
explanation
of
drug
use
needs
to
account
for
drug-specific
factors
for
example,
perceived
risk
and
perceived
availability.
As
those
authors
point
out,
only
by
accounting
for
drug-specific
factors
are
we
able
to
understand
the
different
patterns
of
trends.5Indeed
drug-specific
factors
are
good
predictors
because
they
are
more
1The
literature
that
links
drug
use
and
violence
is
large
see
Boles
and
Miotto
(2003)
for
a
review.
2Results
are
called
pro-cyclical
because
drug
consumption
changes
have
the
same
sign
as
economic
growth
rates:
when
an
economy
grows
(GDP
increases
and
unem-
ployment
falls),
consumption
is
found
to
increase.
Instead
results
are
found
to
be
counter-cyclical
when
in
a
growing
economy
(with
lower
unemployment
rates),
consumption
diminishes.
3The
planning
horizon
of
young
people
during
an
economic
crisis
could
be
shorter
than
during
better
economic
years,
inducing
short-term
recreational
consumption.
4The
rise
of
unemployment
during
recessions
increases
people’s
free
time
to
engage
in
other
activities
outside
the
labour
market.
On
the
other
hand,
it
is
also
true
that
with
more
free
time,
adults
can
increase
their
control
over
activities
undertaken
by
youth
or
teenagers
(Arkes,
2007).
5As
defined
in
Bachman
et
al.
(1988:
95),
drug-specific
factors
are
‘those
factors
which
relate
primarily
to
the
use
of
a
particular
drug
rather
than
to
drug
use
in
general
(or
problem
behaviour)’.
Such
factors
include
awareness
of
the
drug,
per-
likely
to
change
over
time
than
are
general
factors
related
to
the
broad
range
of
problem
behaviour.
In
this
respect,
this
paper
adds
to
the
literature
a
new
analysis
of
how
the
economic
environment
may
have
shifted
attitudes
toward
drugs.6
Our
study
focuses
solely
on
young
people
(aged
15–24
years).
This
is
important
because,
as
the
specialized
literature
shows,
early
consumption
is
one
of
the
factors
that
can
lead
to
progressive
dependence
(Swift
et
al.,
2008;
Coffey
et
al.,
2003;
von
Sydow
et
al.,
2001).
In
the
case
of
cannabis,
for
example,
several
authors
point
to
the
possibility
that
early
consumption
can
work
as
a
gateway
to
harder
drugs
(Melberg
et
al.,
2010;
Beenstock
and
Rahav,
2002;
Pudney,
2003;
Van
Ours,
2003;
Fergusson
et
al.,
2006;
Bretteville-
Jensen
et
al.,
2008).
Our
main
findings
indicate
that
rising
levels
of
total
and
youth
unemployment
may
be
associated
with
increased
consumption
of
cannabis
and
‘new
substances’
by
European
youth.
According
to
our
results
for
the
period
between
2011
and
2014,
a
1%
increase
in
the
regional
unemployment
rate
at
the
NUTS-1
level
is
associated
with
an
increase
of
0.7
percentage
points
in
the
probability
that
young
people
respond
that
they
have
consumed
cannabis.
In
the
case
of
new
substances,
the
figure
is
0.5
percentage
points.
Our
results
also
indicate
that
a
link
can
be
established
between
changes
in
the
local
labour
markets
and
perceived
availability
of
drugs:
as
the
unem-
ployment
rate
rises,
so
access
to
drugs
becomes
more
difficult,
in
the
opinion
of
young
Europeans
(in
particular,
access
to
cocaine,
heroin
and
ecstasy).7On
the
other
hand,
no
link
could
be
estab-
lished
between
changes
in
the
local
labour
markets
and
changes
in
young
people’s
opinion
of
the
health
risk
of
using
drugs.
Finally,
and
when
asked
about
effective
policies
to
combat
the
problems
that
drugs
cause
in
society,
in
contexts
of
rising
unemployment
young
people
say
they
are
more
in
favour
of
measures
that
reduce
poverty
and
unemployment
than
they
are
of
tougher
measures
against
users.
Our
findings
are
important
because
they
provide
evidence
of
other
effects
of
the
Great
Recession
on
young
people
that
go
beyond
those
more
closely
associated
with
the
labour
market.
Moreover,
our
results
should
prove
informative
to
policy
makers:
drug
con-
sumption
is
linked
to
the
opportunities
afforded
to
young
people
in
the
local
labour
markets,
and
so
special
attention
needs
to
be
focused
where
career
prospects
have
worsened
the
most.
Further-
more,
our
paper
takes
into
account
young
people’s
own
views
on
which
policies
are
effective
in
combating
the
problems
caused
by
drugs.
Anti-drug
policy
should
not
ignore
these
views.
After
this
introduction,
the
paper
continues
as
follows.
The
next
section
presents
the
dataset
and
some
descriptive
statistics.
Sec-
tion
3
introduces
the
methodology
and
the
econometric
techniques
used.
Section
4
shows
our
results
on
consumption,
perceived
avail-
ability,
perceived
risk
of
drug
use
and
young
people’s
opinions
regarding
effective
anti-drug
policies.
Finally,
the
conclusions
sum-
marize
our
main
results
and
discuss
avenues
for
future
research.
2.
Data
The
data
used
come
from
four
Eurobarometer
surveys
on
‘Young
people
and
drugs’,
collected
across
Europe
in
four
different
years:
2004
(Flash
Eurobarometer
(EB)
158),
2008
(Flash
EB
233),
2011
(Flash
EB
330)
and
2014
(Flash
EB
401).8The
pooled
dataset
suits
ception
of
the
effects
of
using
it,
availability,
perceptions
by
friends
and
others
and
perceptions
that
friends
and
relatives
disapprove
of
using
a
certain
drug.
6Attitudes,
as
personality
traits,
are
not
set
in
stone
and
change
with
the
social
and
economic
environment
(Almlund
et
al.,
2011).
7Our
paper
does
not
make
any
claims
for
causality:
we
simply
wish
to
consider
different
predictors
to
try
to
understand
possible
future
trends
in
consumption.
8The
surveys
used
in
this
study
have
been
explicitly
designed
to
obtain
informa-
tion
on
drugs
among
young
people
between
the
ages
of
15
and
24,
and
so
our
results
238
S.
Ayllón,
N.N.
Ferreira-Batista
/
Journal
of
Health
Economics
57
(2018)
236–248
the
purposes
of
our
analysis,
because
it
covers
more
than
a
decade
and
deals
with
the
period
prior
to
the
Great
Recession
(2004),
the
time
when
countries
were
hardest
hit
(2008–2011)
and
the
post-crisis
period
(2014)
though
some
countries
are
still
hav-
ing
difficulty
in
escaping
from
the
economic
downturn.
Data
from
the
Eurobarometer
surveys
is
free
of
charge
and
can
be
down-
loaded
from
the
European
Union
Open
Data
Portal
(http://open-
data.europa.eu/data/dataset).
The
pooled
dataset
contains
detailed
information
relating
to
young
people
between
the
ages
of
15
and
24
on
drug
consump-
tion,
access
to
drugs,
youth
opinion
on
the
most
effective
ways
to
combat
the
problems
caused
by
drugs,
perceived
risk
of
drug
consumption,
etc.
The
surveys
are
nationally
representative
of
the
specified
age
group,
and
respondents
were
selected
randomly.9In
total,
28
European
countries
are
present
in
the
pooled
sample:
15
countries
participated
in
2004,
27
countries
in
2008
and
2011,
and
28
countries
in
2014.
Table
A1
in
the
online
Appendix
contains
detailed
information
on
the
countries
and
years
covered.
When
working
with
data
at
the
regional
level,
we
use
information
at
NUTS-1
(Nomenclature
of
Territorial
Units
for
Statistics)
and
we
count
a
total
of
96
regions.10 Our
pooled
sample
contains
45,412
observations.
One
important
issue
of
the
data
to
hand
is
that
the
four
Eurobarometer
surveys
do
not
contain
exactly
the
same
vari-
ables,
and
sometimes
answer
codes
differ.
Thus
a
long
process
of
harmonization
was
required
to
make
target
variables
comparable
across
time.
Also
not
all
the
variables
exist
in
all
the
years.
For
this
reason,
we
clearly
indicate
in
the
tables
and
in
the
text
which
period
is
covered
by
the
results.
Finally,
Table
1
summarizes
some
of
the
most
important
char-
acteristics
of
our
sample
(used
as
control
variables):
51%
are
males;
average
age
is
19.7
years;
31%
live
in
a
rural
area,
43%
in
a
medium-
sized
town
and
25%
in
a
large
city;
66%
are
students,
nearly
23%
are
employed,
2%
are
self-employed
and
9%
are
unemployed.
Finally,
almost
52%
of
the
sample
had
completed
secondary
school,
while
18%
held
a
university
degree.
The
sample
is
largest
for
2014,
as
that
was
the
year
when
most
countries
participated;
each
country
rep-
resents
less
than
5%
of
the
total
sample
(see
Table
A1
in
the
online
Appendix).11 Naturally,
we
would
have
liked
to
include
more
con-
trol
variables,
had
they
been
available,
but
it
should
be
noted
that
some
of
the
variables
used
in
previous
literature
are
not
so
rele-
vant
for
the
age
group
we
focus
on
(for
example,
marital
status
or
self-assessed
health).
Furthermore,
we
have
treated
as
‘missing’
the
answers
‘does
not
know’,
‘not
available’
or
‘refuses
to
answer’;
however,
this
is
a
very
minor
issue
in
the
dataset
to
hand.12
are
not
derived
from
general
health
surveys,
where
typically
the
reported
incidence
of
drug
use
is
very
low.
9For
example,
in
2014,
respondents
were
called
both
on
fixed
and
mobile
phones,
which
may
provide
sufficient
privacy
for
young
people
to
decide
to
participate
in
the
survey
and
to
give
truthful
answers.
The
basic
sample
design
applied
in
all
countries
was
multi-stage
random
probability.
In
each
household,
the
respondent
was
drawn
at
random
following
the
‘last
birthday
rule’.
10 Note
that
there
are
three
levels
officially
defined
by
the
European
Commis-
sion,
with
two
levels
of
local
administrative
units
(NUTS-1
and
NUTS-2)
see
an
interactive
map
at
http://ec.europa.eu/eurostat/web/nuts.
However,
some
coun-
tries,
because
of
their
small
size,
do
not
have
such
a
division.
For
example,
in
Luxembourg
the
three
levels
correspond
to
the
entire
country.
Moreover,
while
some
countries
provided
information
at
the
NUTS-2
and
NUTS-3
level,
other
coun-
tries
did
not,
which
prevented
us
from
working
with
more
disaggregated
data.
This
means
that,
in
some
countries,
results
at
the
regional
level
can
refer
to
large
geo-
graphical
areas
(as,
for
example,
in
Finland).
See
all
the
detail
in
Table
A1
in
the
online
Appendix
(www.saraayllon.eu/web-appendices.html).
11 The
descriptive
statistics
for
the
dependent
variables
are
presented
at
the
begin-
ning
of
each
section
and
the
unemployment
rates
used
across
time
are
shown
in
the
next
section.
12 The
number
of
observations
treated
as
missing
differs
for
each
dependent
vari-
able.
For
example,
in
the
case
of
cannabis
consumption,
it
is
0.6%
of
the
sample
used,
Table
1
Summary
statistics.
Variables
Mean
Std.
deviation
Min.
Max.
Male
0.510
0.50
0
1
Age
19.725
2.74
15
24
Age
squared
396.585
108.24
225
576
Rural
area
or
village
0.316
0.46
0
1
Small
or
medium-sized
town
0.436
0.50
0
1
Large
town
or
city
0.249
0.43
0
1
Student
0.664
0.47
0
1
Employee
0.227
0.42
0
1
Self-employed
0.018
0.13
0
1
Not
working/unemployed
0.091
0.29
0
1
Never
been
in
full-time
education
0.004
0.07
0
1
Primary
education
0.295
0.46
0
1
Secondary
education
0.521
0.50
0
1
Higher
education
0.179
0.38
0
1
2004
0.169
0.37
0
1
2008
0.271
0.44
0
1
2011
0.271
0.44
0
1
2014
0.289
0.45
0
1
Source:
Authors’
computation,
based
on
the
Eurobarometer
‘Young
people
and
drugs’
surveys
for
2004,
2008,
2011
and
2014.
Weighted
results.
3.
Methodology
In
order
to
capture
the
possible
relationship
between
the
busi-
ness
cycle
and
drug
consumption
or
changes
in
attitudes
toward
drugs,
we
matched
the
harmonized
dataset
of
the
four
Eurobarom-
eter
surveys
with
data
from
Eurostat
on
the
total
and
the
youth
unemployment
rates
at
both
country
and
regional
level.
The
dif-
ferent
impact
of
the
Great
Recession
in
Europe
and
the
important
variability
of
unemployment
rates
over
time
and
across
countries
and
regions
allow
identification
of
different
consumption
patterns
or
changes
in
opinions
and
attitudes
toward
drugs
that
are
related
to
changes
in
macroeconomic
conditions.
Fig.
1
shows
the
great
variability
in
the
incidence
of
unemploy-
ment
within
the
population
and
especially
among
young
people
(under
25)
over
time
and
across
countries
according
to
data
from
Eurostat.
As
can
be
seen
from
the
four
years
considered
and
the
28
European
countries
under
analysis,
the
youth
unemployment
rate
varies
from
as
low
as
7.7%
in
Germany
(2014)
to
as
high
as
53.2%
in
Spain
(2014).
The
variability
is
even
greater
if
we
con-
sider
all
the
regions
(not
shown)
with
the
figure
lowest
in
Bayern
(Germany)
with
4.4%
in
2014
and
highest
in
Andalucía
(Spain)
with
60.2%,
also
in
2014.
Furthermore,
the
within-country
deviation
in
the
annual
youth
unemployment
rate
for
the
period
under
analysis
ranges
from
0.8
to
13.4;
Greece
and
Spain
are
the
countries
with
the
most
variation,
while
Malta
and
Austria
have
the
least.
The
within-
country
deviation
rang
is
smaller
for
the
total
unemployment
rate
(from
0.2
to
7.2),
but
larger
for
both
measures
of
unemployment
at
the
regional
level.
Moreover,
additional
graphs
(available
from
the
authors
upon
request)
show
that
not
all
changes
across
time
in
the
unemployment
rates
move
in
the
same
direction
and
not
all
the
changes
are
of
the
same
size.
Our
results
are
based
on
logit
models
and
generalized
ordered
logit
models
(depending
on
the
nature
of
the
dependent
variable
under
study)
with
fixed
effects.
In
the
case
of
simple
logits,
and
using
the
subscript
c
for
country
(or
r,
in
the
case
of
regions)
and
t
for
time,
the
basic
regression
can
be
specified
as
follows:
Yict =
˛
+
Xict
+
Unemplct
+
Cc+
Tt+
sict (1)
where
Yict represents
the
outcome
of
interest
and
Xict is
the
vec-
tor
of
our
control
variables,
which
include
gender,
age,
age
squared,
while
in
the
case
of
effective
policies
it
goes
up
to
4.0%.
However,
the
percentage
of
‘refuse
to
answer’
or
‘not
available’
never
exceeds
5%
in
the
variables
used.
S.
Ayllón,
N.N.
Ferreira-Batista
/
Journal
of
Health
Economics
57
(2018)
236–248
239
Fig.
1.
Total
and
youth
unemployment
rate
(less
than
25
years
of
age)
at
country
level,
2004,
2008,
2011
and
2014
(in
%).
Note:
In
the
graph,
1
refers
to
the
year
2004,
2
to
2008,
3
to
2011
and
4
to
2014.
The
missing
bars
in
the
graphs
indicate
that
the
country
is
not
present
in
the
analysis
in
that
year
as
it
does
not
participate
in
the
Eurobarometer
surveys.
Countries
are
displayed
in
alphabetical
order,
according
to
their
full
names
in
English.
Source:
Eurostat.
educational
level,
status
in
the
labour
market
and
living
in
an
urban,
semi-urban
or
rural
area.13 The
parameter
of
main
interest
is
,
which
captures
the
relationship
between
the
unemployment
rate
of
the
country
(or
region)
in
which
each
young
person
lives
and
the
different
outcomes
examined.
In
other
words,
captures
the
association
of
within-country
(or
within-region)
deviations
in
eco-
nomic
conditions
over
time
on
the
outcomes
of
interest.
Finally,
sict
is
the
usual
error
term.
The
fixed
effects
Cc(or
Rrin
the
case
of
regions)
control
for
time-
invariant
country
(or
region)
characteristics,
while
Ttaccounts
for
time
effects.
In
other
words,
country
and
region
fixed
effects
control
for
a
given
pattern
specific
to
an
area,
and
time
fixed
effects
control
for
possible
shocks
that
could
change,
for
example,
attitudes
toward
drugs
in
a
given
year
in
all
Europe.14 This
methodology
allows
us
to
control,
for
instance,
for
differences
in
the
prices
of
drugs
in
the
different
areas
and
over
time.
All
the
regressions
are
weighted
by
the
population
weights
provided
in
the
different
datasets
and
clus-
tered
standard
errors
at
the
country
(or
regional)
level
are
used
throughout
the
paper.
13 The
information
on
educational
level
is
not
available
for
2004,
but
various
robustness
checks
have
been
carried
out
to
ensure
that
our
results
are
not
depen-
dent
on
this.
It
is
important
to
note
that
there
are
potential
confounding
factors
that
can
influence
the
link
between
our
outcomes
of
interest
and
the
level
of
edu-
cation
(or
young
people’s
status
in
the
labour
market).
Particularly
for
the
youngest
individuals
in
our
sample
who
have
not
completed
education,
there
could
be
sev-
eral
unobservable
factors
that
may
cause
school
dropout
or
delay
that
can
have
an
impact
on
drug
consumption
or
attitudes
towards
drugs.
The
same
is
true
of
self-
selection
into
a
labour
market
status.
The
methodology
applied
in
this
paper
does
not
deal
with
these
issues.
14 Phrased
differently,
year
fixed
effects
capture
trends
that
are
not
specific
to
a
country
(or
region),
for
example,
a
cultural
factor.
In
the
case
of
ordinal
dependent
variables
for
example,
ease
of
access
to
substances
or
young
people’s
opinion
on
how
harmful
it
is
to
consume
drugs
we
use
generalized
ordered
logit
models
instead
of
the
standard
ordered
logit
model,
because
the
ordered
logit
model
assumption
of
parallel
lines
or
proportional
odds
is
violated
in
our
data.15 We
use
the
Stata
command
gologit2,
made
available
to
researchers
by
Williams
(2006).
When
using
general-
ized
ordered
logits,
it
is
important
to
remember
how
to
interpret
the
results:
generalized
ordered
logits
are
equivalent
to
a
series
of
binary
logistic
regressions
where
the
dependent
variable
is
com-
bined
in
different
categories.
So
if,
for
example,
the
dependent
variable
has
four
categories
(from
1
to
4),
then
the
results
for
cate-
gory
1
will
need
to
be
read
in
contrast
to
categories
2,
3
and
4.
In
a
similar
vein,
the
results
for
category
2
will
contrast
categories
1
and
2
against
3
and
4,
while
the
results
for
category
3
will
consider
1,
2
and
3
against
4.
Note
that
the
way
to
interpret
the
coefficients
is
different
from
the
way
in
which
results
for
marginal
effects
should
be
read:
marginal
effects
indicate
the
probability
that
an
outcome
occurs,
given
certain
values
of
the
independent
variables.
Finally,
and
as
will
be
shown
in
the
next
section,
we
present
all
the
results
using
four
different
measures
of
the
unemployment
rate:
(1)
the
total
unemployment
rate
at
country
level,
(2)
the
total
unemployment
rate
at
regional
level,
(3)
the
youth
unemployment
rate
(15–24)
at
country
level
and
(4)
the
youth
unemployment
rate
at
regional
level.
Which
of
these
rates
is
the
best
for
the
purposes
of
our
analysis
is
still
a
matter
of
discussion
in
on-going
research.
On
the
one
hand,
for
example,
Arkes
(2007)
argues
in
favour
of
using
a
15 The
assumption
requires
that
all
the
coefficients
are
the
same
for
each
category
of
the
dependent
variable.
We
tested
the
assumption
with
the
Brant
test
and,
in
all
cases,
the
assumption
was
violated.
240
S.
Ayllón,
N.N.
Ferreira-Batista
/
Journal
of
Health
Economics
57
(2018)
236–248
total
rate
because
it
is
derived
from
a
larger
number
of
observations
than
is
the
rate
for
young
people
(or
teenagers
in
the
case
of
Arkes,
2007).
That
way,
potential
sampling
error
may
be
less
important.
Moreover,
Arkes
(2007)
points
out
that
use
of
a
youth
unemploy-
ment
rate
could
introduce
some
endogeneity
if
youth
labour
supply
is
affected
by
drug
use.
But
such
an
effect
is
likely
to
have
a
mini-
mal
impact
on
the
rate
for
the
population
as
a
whole.
On
the
other
hand,
though,
it
can
be
argued
that
young
people
are
more
likely
to
make
decisions
about
drugs
while
assessing
the
opportunities
for
their
age
group
in
the
labour
market,
rather
than
the
opportunities
for
the
adult
population
as
a
whole.
Given
such
different
argu-
ments,
we
use
both
measures;
this
has
the
advantage
of
assessing
the
robustness
of
the
relationships
found.
4.
Results
We
present
our
results
in
four
main
sub-sections.
First,
we
show
our
findings
on
the
consumption
of
cannabis
and
new
substances.
Second,
we
explore
the
relationship
between
changes
in
(local)
labour
markets
and
perceived
drug
availability.
Third,
we
anal-
yse
risk
behaviour
toward
drugs.
And
finally,
we
study
changes
in
young
people’s
opinions
concerning
the
most
effective
ways
in
which
the
public
authorities
can
combat
the
problems
that
drugs
cause
in
society.
4.1.
Drug
consumption
The
literature
regarding
the
link
between
the
business
cycle
and
substance
use
is
vast
(see,
among
many
others,
Neumayer,
2004;
Tapia
Granados,
2005;
Gerdtham
and
Ruhm,
2006;
Charles
and
DeCicca,
2008;
Catalano
et
al.,
2011;
Arkes,
2012;
Bradford
and
Lastrapes,
2014;
Currie
and
Tekin,
2015).
However,
most
of
the
lit-
erature
has
focused
on
the
consumption
of
tobacco
and
alcohol,
while
the
specific
relationship
between
illicit
drug
consumption
and
changes
in
the
labour
market
has
been
less
thoroughly
ana-
lysed.
Among
studies
that
focus
on
legal
substances,
the
general
results
point
to
a
pro-cyclical
relationship
(Xu,
2013;
Ásgeirsdóttir
et
al.,
2012;
Dávalos
et
al.,
2012;
Charles
and
DeCicca,
2008;
Johansson
et
al.,
2006;
Ruhm,
2005;
Ruhm
and
Black,
2002).16 Instead,
results
are
not
so
clear
among
studies
that
include
illicit
drugs,
as
they
often
differ
by
age
group
or
by
country
(or
region).
An
example
is
the
work
of
Chalmers
and
Ritter
(2011).
These
authors
anal-
yse
the
implications
of
the
business
cycle
on
cannabis
and
alcohol
consumption
(number
of
users
and
frequency)
in
Australia
for
the
period
1991–2007.
The
results
for
cannabis
show
that
among
young
adults
(under
24)
the
relation
is
counter-cyclical
for
participation
and
frequency.
But
among
people
over
24,
although
the
frequency
of
cannabis
use
is
pro-cyclical,
the
effect
of
the
business
cycle
on
participation
is
mixed:
it
decreases
with
a
rise
in
the
unemploy-
ment
rate
(pro-cyclical),
but
goes
in
the
opposite
direction
if
the
income
per
capita
falls
(counter-cyclical).
In
their
conclusions,
the
authors
point
out
that
for
Australians,
cannabis
is
not
a
normal
good.
The
work
by
Chalmers
and
Ritter
(2011)
is
just
one
example
of
how
differences
in
the
age
group
analysed
can
lead
to
mis-
matched
results
in
the
literature.
The
key
to
understanding
such
differences
among
various
age
groups
lies
in
the
direction
of
the
‘income
effect’.
Arkes
(2007,
2012),
who
focused
on
young
Amer-
16 There
are
some
studies
that
point
to
a
counter-cyclical
relationship,
but
they
deal
with
specific
population
groups
as
teenagers
or
young
adults
(Arkes,
2012;
Dee
and
Evans,
2003)
or
use
individual
employment
status
as
a
business
cycle
indicator
(Aguilar-Palacio
et
al.,
2015;
Golden
and
Perreira,
2015).
See
Catalano
et
al.
(2011)
for
a
complete
review.
icans
(16–24),
shows
that
the
counter-cyclical
link
between
drug
consumption
and
the
business
cycle
for
this
group
is
related
to
their
limited
response
to
the
income
effect.
His
studies
indicate
that
dur-
ing
economic
recessions,
the
young
are
more
likely
to
sell
drugs
and
are
thus
better
able
to
fund
their
own
consumption
(Arkes,
2011).
This
finding
is
supported
by
other
studies,
which
show
that
young
people
can
even
get
cannabis
for
free
(Caulkins
and
Pacula,
2006;
Harrison
et
al.,
2007).17
Concerning
the
impact
of
the
Great
Recession
on
substance
use
in
Europe,
the
literature
is
still
very
scarce.
Zuccato
et
al.
(2011)
find
that
during
the
recession
in
the
North
of
Italy,
drug
consumers
replaced
expensive
substances
with
cheaper
ones.18
The
works
of
Colell
et
al.
(2014)
and
Martin-Bassols
and
Vall-
Castello
(2016)
centred
on
the
consumption
of
illicit
drugs
in
Spain.
Both
groups
of
researchers
found
a
counter-cyclical
relationship
between
cannabis
use
and
the
business
cycle.19 Despite
the
rele-
vance
of
their
findings,
these
works
do
not
address
the
situation
of
young
people.
As
for
evidence
on
the
United
States,
Carpenter
et
al.
(2017)
con-
tains
a
first
analysis
of
the
relationship
between
macroeconomic
conditions
and
the
use
of
several
illicit
drugs
from
2002
to
2015,
while
drawing
on
restricted
data
from
the
National
Survey
on
Drug
Use
and
Health
for
individuals
aged
12
and
older.
Their
results
provide
mixed
evidence
on
the
relationship
between
state
unem-
ployment
rates
and
drug
consumption:
economic
downturns
are
associated
with
increases
in
the
use
of
ecstasy
and
heroin,
and
with
decreases
in
the
use
of
LSD
and
crack
(though
results
are
sensitive
to
the
time
window
used).
In
the
case
of
cocaine,
tranquilizers
and
inhalants,
the
coefficients
for
the
unemployment
rate
do
not
differ
statistically
from
zero.
A
positive
relationship
is
also
found
between
the
unemployment
rate
and
the
timeframe
within
which
individ-
uals
report
having
used
‘any
illicit
drug’
or
analgesics.
Their
results
also
show
that
certain
substance-use
disorders
(based
on
profes-
sionally
developed
diagnostic
criteria)
are
countercyclical
e.g.
mild
disorder
for
marijuana
or
severe
disorder
for
hallucinogens,
to
give
two
examples.
The
authors
also
find
that
the
relationship
between
economic
conditions
and
use
disorders
is
symmetrical
(similar
in
magnitude
whether
the
state
of
the
economy
is
improv-
ing
or
worsening).
In
turn,
Pabilonia
(2017)
analyses
the
effects
of
the
Great
Recession
on
teenagers’
risky
behaviours
in
the
US,
including
the
use
of
marijuana
in
the
past
30
days.
She
finds,
for
the
period
2003–2011,
that
unemployment
rates
are
associated
with
an
increase
in
consumption,
but
only
among
black
teenage
males.
The
information
on
drug
consumption
in
the
Eurobarometer
surveys
focuses
on
just
two
drugs:
cannabis
and
the
so-called
‘new
substances’.20 In
the
case
of
cannabis,
the
question
asked
is
exactly
the
same
in
2011
and
2014,
‘Have
you
used
cannabis
yourself?’
with
four
possible
answers
(apart
from
‘do
not
want
to
answer’
or
‘do
not
know’):
(1)
‘No,
I
have
never
tried’,
(2)
‘Yes,
in
the
last
30
days’,
(3)
‘Yes,
in
the
last
12
months’,
and
(4)
‘Yes,
but
more
than
12
months
17 The
first
authors
showed
that
58%
of
cannabis
users
in
the
2001
National
House-
hold
Survey
on
Drug
Use
and
Health
in
the
United
States
obtained
cannabis
for
free;
the
second
study
found
similar
percentages
among
14–17
year-old
cannabis-using
students
in
Philadelphia,
Toronto
and
Montreal.
18 The
authors
used
wastewater
analyses
to
estimate
loads
of
cocaine,
heroin,
methamphetamine
and
cannabis
consumed
daily
in
two
cities
(Milan
and
Como)
for
the
period
between
2005
and
2009.
19 Both
studies
use
the
four
editions
of
the
Spanish
Household
Survey
on
Alcohol
and
Drugs
for
2005,
2007,
2009
and
2011.
Colell
et
al.
(2014)
focus
on
cannabis
frequency
use
and
Martin-Bassols
and
Vall-Castello
(2016)
on
the
consumption
of
legal
(alcohol
and
tobacco)
and
illegal
substances
(cannabis
and
‘hard
drugs’).
20 ‘New
substances’
refer
to
powders,
tablets,
pills
or
herbs
that
imitate
the
effect
of
illicit
drugs
some
of
these
are
sold
as
legal
substances
in
several
countries
and
are
often
known
as
‘legal
highs’.