Shi et al report a set of statistical associations in cross-national HBSC survey data between national cannabis policies and adolescent use in three time periods; ever, past-year and regular. Their summary of the results is that liberal policies are associated with increased adolescent use. They write in their conclusion: 'Our study showed that liberalization policy in general was associated with higher levels of cannabis use, and depenalization and partial-prohibition policies were particularly correlated with regular use.' This is a potentially important finding, especially as it contradicts the finding of a similar, multilevel study by Vuolo (2013), who found “lower” odds of drug use in countries which had ‘liberalized’ policies on cannabis possession.

We would like to note two potential issues with the analysis by Shi et al.

The first issue is that the authors’ conclusion seems based on the results for girls exclusively, and may not hold for boys, or for adolescents in general. As the authors note in reference to the boy-policy interaction term, 'the associations between cannabis use and cannabis control policies were in general smaller in boys.' They do not mention that the effects for boys were so much smaller that liberal policies in several cases were associated with apparent reductions in male adolescent use. For instance, while the 'liberalization' odds ratio (OR) for girls on the 'regular use' outcome was 1.23 in table 3, the analogous OR for boys would be 1.23 x 0.77 = 0.947 which is below 1 and so indicates lower use. Given that boys are far more likely to be regular users (the baseline OR for boys is 3.08), this means that the association between overall adolescent use and liberalization will be a mix of an increase for girls and a decrease for a larger number of boys.

To attempt an estimate of the association between policies and overall adolescent use levels, we may calculate the 'average' OR across the genders relative to the baseline odds of use by girls in ‘full prohibition’ countries. Continuing with the previous example and assuming a 50% gender split, this gives us 0.5 x (1 + 3.08) = 2.04 for the baseline policy (‘full prohibition’). A liberal policy is associated with an “increase” to 1.23 for girls and a “reduction” to 0.947 for boys, giving an average OR of 0.5 x (1.23 + 3.08 x 1.23 x 0.77) = 2.07. With the rounded numbers, this suggests that the liberalized policies are associated with a 2.07/2.04 = 1.01 change in the OR for regular use for adolescents in general, which is substantively equal to no association.

Repeating this exercise for the various policies reported, this pattern holds for many of the outcomes and policies analysed. Results for the ORs in the 'liberalization' analysis of table 3 are here: http://screencast.com/t/n... and results for the more narrowly defined 'liberalization' policies are here: http://screencast.com/t/v...

This analysis suggests that a more correct summary of the analysis would be that ‘liberalization’ is not associated with higher adolescent use (ORs for ever use 0.96, past year use 0.96, and regular use 1.02), and that the same holds for both depenalization and decriminalization (with 'net effects' in the range of 0.93 to 1.02 depending on policy and outcome measure). Only 'partial prohibition' remains associated with increased use (ever use 1.44, past year use 1.64 and regular use 2.07).

The second issue concerns the mapping of countries onto policy regimes. The authors write that they use a policy typology presented in Room et al (2010) and additional data from the EMCDDA, but provide few details as to how the EMCDDA data were used to map countries into the typology. Some countries that are not covered by either the EMCDDA or Room et al (e.g. Armenia, Israel) are included in the analysis. The UK is apparently included in the ‘depenalisation’ group in 2005/6 and in the ‘full prohibition’ group after 2009, despite there having been no change in the discretionary depenalization of first-time cannabis possession offences by adults in England and Wales through the issuing of on-street cannabis warnings. These were introduced in 2004 and are still used now.

A very important analytical decision is where to place the USA. Throughout the period covered by the analysis, the USA as a whole had relatively high rates of arrest and imprisonment for cannabis possession, despite formal decriminalization in some states. Other cited authors (e.g. Reinarman, Cohen & Kaal 2004) have used the USA as a direct opposite case in comparison with the more liberal policy of the Netherlands. Shi et al places these two countries together in the ‘partial prohibition’ group. Their supplementary analysis shows that the association between ‘partial prohibition’ and higher rates of adolescent cannabis use reduces if the USA is excluded from the analysis. Given the relative scale of the law enforcement approach to cannabis in the USA as a whole, and the high rates of use that US adolescents have reported to the HBSC survey, it would be interesting to see how the inclusion of the USA as an example of ‘full prohibition’ would affect the results of the analysis.

Authors:

Ole Rogeberg (Senior Research Fellow, Frisch Centre) and Alex Stevens (Professor in Criminal Justice, University of Kent)

“Reference”

Vuolo M (2013) National-level drug policy and young people’s illicit drug use: A multilevel analysis of the European Union. “Drug and Alcohol Dependence” 131(1-2): 149–156.