> the data from the studies included in the aforementioned literature reviews (i.e. General Accounting Office (1993); World Health Organization (2004)) rely on small sample sizes and self-reported data regarding individ- uals’ drug use, and do not typically consider attrition nor spillover effects on those not directly treated. Additionally, many studies use data from Canada, Sweden, or New Zealand to serve as a comparison group for drug rates in the US. Such methods are problematic for addressing causality, given that other developed countries have differing policies on the operations of SEPs and greater access to substance abuse treatment.1 I overcome these existing limitations by cre- ating a novel dataset that combines information on SEP opening dates with restricted administrative county-level data on HIV diagnoses. These data proxy for drug use without relying on self-reporting behavior, and are representative of counties across the US.

It’s using newly-opened SEPs for a discontinuity analysis, which does provide evidence of causation (and furthermore notes that previous studies were correlational).