A new NBER working paper by Hefei Wen and colleagues finds that public insurance expansions increases substance use disorder (SUD) treatment, and that treatment may reduce crime. There’s a long-documented connection between substance use and crime, with 52% of violent offenders reporting being under the influence of alcohol or other substances while committing the crime (39% for property offenders). What remains unclear—and what the authors aim to clarify in this paper—is how much of that correlation represents causality running from substance use to crime.
Enforcement and treatment are two interventions targeting substance use that might produce observable changes in the crime rate. However, the authors note that efforts targeting “substance control such as prohibition and the ‘war on drugs’ have not led to significant crime reduction.”
Treatment, on the other hand:
The study findings highlight that a relative 10 percent increase in the SUD treatment rate can reduce the robbery rate by 3 percent, reduce the aggravated assault rate by 4 to 9 percent, and reduce the larceny theft rate by 2 to 3 percent. […] A back-of-the- envelope calculation shows that a 10 percent relative increase in the SUD treatment rate at an average cost of $1.6 billion yields a crime reduction benefit of $2.5 billion to $4.8 billion.
The paper uses an instrumental variable approach, so standard caveats apply—our confidence in the findings is only as high as our confidence in the instrument. An OLS approach did not yield significant results, but the authors note that endogeneity and omitted-variable issues that might bias those estimates toward the null.
Once we address the endogeneity issue, we are able to demonstrate an economically meaningful reduction in the rates of robbery, aggravated assault and larceny theft attributable to an increased SUD treatment rate. […]
[W]e employ a set of instrumental variables that are strongly related to SUD treatment, but are otherwise unrelated to crime. The instruments are two state-level policy shocks that occurred during the 2000s, namely the Health Insurance Flexibility and Accountability (HIFA)-waiver expansions and SUD health insurance parity mandates.
The authors controlled for all of the usual suspects (demographic and economic characteristics of the study population), for law enforcement resources, and for other government expenditures that might indirectly influence the crime rate (including education, welfare, and alternate sources of health funding).
[T]he first stage of our IV estimation is of interest in its own right. It provides previously undocumented evidence of a significant increase in the SUD treatment rate arising from public insurance expansions. This has direct relevance to the current health care reform discussions surrounding insurance expansion and “mainstreaming” of SUD treatment. The Affordable Care Act (ACA) is expected to substantially expand insurance coverage. Much of this expansion will occur through Medicaid and in the health insurance exchanges, and will include coverage for those with SUDs who are also in the age groups more likely to commit these crimes.
The societal perspective is intuitive for these findings because the authors examine expansions in taxpayer-funded Medicaid coverage. Under the ACA, access to substance use treatment is also expected to expand in the private market, as substance use and mental health coverage are defined essential health benefits. These findings suggest that society, not insurers, might realize the higher return-on-investment for these services.
SUD treatment holds the potential not only to reduce individual substance use, but also to promote public safety by reducing crime. One contribution of our study is that we uncovered a heretofore unrecognized relationship between the implementation of HIFA-waiver expansions and the increase in the SUD treatment rate. While this finding is interesting in and of itself, it also provides a potential avenue for solving the issue of joint determination of SUD treatment and crime that may seriously bias the simple OLS estimates towards zero