Note from the Co-Director
The Biden Administration's plan to combat drug overdoses and the pending opioid settlement funds will give states funding to address the relentlessly evolving opioid crisis. But decisionmakers need to know whether a given policy will have the intended effect, for the target population, in their state.
An OPTIC team conducted an extensive simulation study to compare the statistical performance of commonly used methods for evaluating state-level opioid policies, including popular difference-in-differences (DID) methods. We found that DID models had low power to detect potentially actionable state-level policy effects and generated high false positive rates when examining state-level outcomes. As a result, policy evaluations using such approaches are likely to have findings that are inaccurate, inflated, and even in the wrong direction. Our simulations showed that autoregressive models outperformed DID models in terms of power, false positives, and directional bias, and we strongly encourage the field to consider such models in future policy evaluations. You can learn more about this work in our recent webinar series.
A recent useful addition to the OPTIC website is a STATA tool to help researchers determine whether they have a "good match" between a control group of states that has not implemented a policy with a treated group that has.
The entire OPTIC team wishes you a safe and productive new year. As always, we welcome your comments and suggestions at OPTIC@rand.org.
Stay safe. – Beth Ann Griffin, Co-Director OPTIC
|
