TABLE II.

Robustness Tests: Drug Overdose Death Rate

Nontriplicate ×	Baseline results (1)	Select on population size (2)	Select on PDMP states in 1996 (3)	Control for policy variables (4)
1996–2000	1.267** [0.062, 2.274]	2.919** [0.452, 5.067]	2.163 [−0.978, 4.828]	1.348** [0.176, 2.453]
2001–2010	3.561*** [1.321, 5.687]	5.543*** [2.671, 8.591]	5.869* [−0.711, 11.369]	3.628*** [1.671, 5.322]
2011–2017	5.240*** [3.213, 7.274]	6.045** [0.535, 12.242]	9.299*** [3.528, 14.946]	5.808*** [3.528, 8.030]
Joint p-value	.001	.073	.005	.000
Mean 1991–1995	4.436	5.090	5.294	4.436
N	1,377	216	405	1,377

Notes:

^***Significance 1%

^**significance 5%

^*significance 10%.

Outcome is all drug overdose deaths per 100,000. The reported coefficients refer to the interaction of the given time period and an indicator for whether the state did not have a triplicate program in 1996. Estimates are relative to preperiod 1991–1995. 95% confidence intervals reported in brackets are estimated by clustered (by state) wild bootstrap. All models include state and year fixed effects and time-varying covariates (see Table I for details). Column (1) repeats the column (3) results from Table I. Column (2) selects on the four nontriplicate states with the largest populations in 1990 along with the four largest triplicate states. Column (3) selects on states with some form of PDMP (triplicate, duplicate, electronic) in 1996. Column (4) includes policy controls for PDMPs (any PDMP, electronic PDMP, “must access” PDMPs), pain clinic regulation, medical marijuana laws, and operational/legal medical marijuana dispensaries. “Joint p-value” refers to the p-value from a joint hypothesis test that all three nontriplicate post effects are equal to zero and is also estimated using a restricted wild bootstrap.