Factors driving the diffusion of medical marijuana legalisation in the United States

Abstract

In the past 20 years, the drive to legalise medical marijuana (cannabis) has gained national attention with the public and policy makers. Beginning in 1996 states started implementing MMLs (medical marijuana laws), and by the end of 2015, 23 states and the District of Columbia had adopted laws legalizing marijuana use for medical purposes. We find that measures of policy diffusion and political culture are important drivers of adoption. However, the relative effects of positive pressure from neighbouring states and negative pressure from the ideological distance between adopting and non-adopting states is such that complete diffusion is unlikely.

Keywords:

• Cannabis
• policy
• medical marijuana

Notes

1 For more details on state laws with links to legislative language, see ProCon.org. (2015). “23 Legal Medical Marijuna States and DC.” Retrieved December 18, 2015, from http://medicalmarijuana.procon.org/view.resource.php?resourceID=000881.

2 We present the state of the MML adoption up to the current time for discussion purposes. Our regression analysis stops in 2012 because of limitations in the data availability for all of the variables used in our analysis other than the MML themselves.

3 Other disease groups, such as individuals with cancer, also seem to benefit from marijuana use as part of their disease management. Unfortunately, consistent estimates of state-level cancer, glaucoma, etc., diseases are not available with sufficient frequency or over sufficient time periods to be included in our empirical models. For this reason, we will use the number of AIDS patients in each state as our measure of direct interest group pressure.

4 This modelling approach is referred to as a discrete time duration model in economics, and a hazard model in biostatistics.

5 Note that one might naturally consider that the level of criminal activity associated with marijuana would affect MML adoption timing. While this point is conceptually valid, it is empirically problematic: the volume of drug seizures, arrests, and even drug use will themselves be determined by marijuana policy and the signals that are sent to law enforcement leading up to the policy change. Thus, including them in our regression would lead to potential endogeneity bias. Similarly, drug prices will also be endogenous (as well as being difficult to measure for all 50 states in all years). Consequently, we do not include measures of drug activity in our causal model, which should be interpreted as a reduced form regression.

6 We tested splines defined with three through seven knots. Model performance (estimated value of the logged likelihood and pseudo-R²) improved through six knots, and overall model goodness of fit was not meaningfully improved for seven knots compared to six. Consequently, we used the six-knot specification.

7 Our primary concern is in correctly estimating the standard errors, since the parameters are not algebraically affected by clustering.

8 States that require patients to grow their own marijuana as of 2014 are Alaska and Hawaii; Michigan and Montana permit home cultivation and do not explicitly permit dispensaries.

9 States that are dispensary only are: Connecticut, Delaware, D.C., Illinois, Maryland, Minnesota, New Hampshire (in 2015), New Jersey, and New York.

10 States that permit both types of sourcing include: Arizona, California, Colorado, Maine, Massachusetts, Nevada, New Mexico, Oregon, Rhode Island, Vermont, and Washington.