Working Paper: NBER ID: w29802
Authors: Neil K. Mathur; Christopher J. Ruhm
Abstract: Over the last two decades there has been considerable movement at the state-level to legalize marijuana, initially for medical purposes and more recently for recreational consumption. Despite prior research, it is unclear how, if at all, these policies are related to rates of opioid-involved overdose deaths, which have trended rapidly upwards over time and represent a major public health problem. We provide two types of new information on this question. First, we replicate and extend upon previous investigations and show that the empirical results of those studies are frequently fragile and that, in most cases, the inclusion of more comprehensive controls, longer analysis periods and more correctly defined dependent variables results in less favorable estimates, often including predicted increases in opioid deaths. Second, we present new estimates from generalized differences-in-differences and event study models that incorporate more recent data and improvements developed in our replication and extension of early research. These results indicate that legal medical marijuana, particularly when available through retail dispensaries, is associated with higher opioid mortality. The results for recreational marijuana, while less reliable, also suggest that retail sales through dispensaries are associated with greater death rates relative to the counterfactual of no legal cannabis.
Keywords: marijuana legalization; opioid deaths; public health; differences-in-differences; event study models
JEL Codes: I10; I12; I18
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| legal medical marijuana (K40) | opioid mortality (I12) |
| retail dispensaries (L81) | opioid mortality (I12) |
| recreational marijuana legalization (L83) | opioid mortality (I12) |
| recreational marijuana legalization through retail dispensaries (H27) | opioid mortality (I12) |
| demographic variables (J10) | opioid mortality (I12) |