Table 5.

Difference-in-Differences Estimates for Doctor Shopping Outcomes.

	Panel A: State and Time Fixed Effects
	(1)	(2)	(3)
	Overlapping Prescriptions (per 1,000)	Number of Prescribers (per beneficiary)	New Patients (per beneficiary)
PMCL	−3.676** (1.536)	−0.012* (0.007)	−0.017** (0.008)
Counterfactual Mean	22.510	0.328	0.151
Implied Percent Change	−16.333	−3.584	−11.232
	Panel B: + Demographic and Policy Controls
PMCL	−4.781*** (1.175)	−0.011*** (0.003)	−0.022*** (0.004)
Counterfactual Mean	23.614	0.327	0.156
Implied Percent Change	−20.247	−3.295	−14.254
	Panel C: + OxyContin Misuse
PMCL	−4.818*** (1.055)	−0.011*** (0.002)	−0.022*** (0.004)
Counterfactual Mean	23.651	0.327	0.156
Implied Percent Change	−20.370	−3.267	−14.131

Notes:

^*10%,

^**5%,

^***1% statistical significance. Standard errors (adjusted for state-level clustering) provided. We use two-stage difference-in-differences, weighted by the number of beneficiaries. Only states east of the Mississippi River are included in the analysis. In the first step, we regress the outcome on state fixed effects, time fixed effects, and covariates using only untreated observations. We use the estimates to impute the counterfactuals for the treated units. We regress the difference between the observed outcome and estimated counterfactual on whether the state had enacted a PMCL. “Demographic and Policy Controls” include share of the state population that is White, share of Medicare beneficiaries ages 65+, and policy variables. The policy variables are ACA Medicaid expansion, legal and operational medical marijuana dispensaries, recreational marijuana laws, must-access PDMPs, and opioid prescribing guidelines. Panel C includes controls for the interaction of the 2004–2009 non-medical OxyContin use rate with year indicators. The “Counterfactual Mean” is the mean of the outcome for all treated observations after subtracting off the estimated treatment effect. “Percent Change” is the implied percent change of the estimate given the counterfactual mean.