Association for Public Policy Analysis & Management

Autism Spectrum Disorder (ASD) is one of the fastest growing developmental disabilities in the US. According to the Centers for Disease Control, while 0.67% of the 1992 birth cohort had been diagnosed with ASD by age eight, this rate more than doubled by 2004. Though it affects a relatively small number of students overall, ASD is a very expensive disability to treat, requiring many hours of therapy services. A recent study found that children with ASD cost schools $8,610 more than the average non-ASD student while some studies calculate total treatment costs as high as $68,000 per year (Jarbrink, 2007; Knapp, Romeo and Beecham, 2009; Lavelle et al., 2014).

Recently, a number of states have sought to address spiraling costs of ASD treatment by introducing insurance mandates for therapy such as applied behavioral analysis (ABA). These therapies have been shown to generate substantial improvement in children's behavior, though there remains little evidence of their effects on students' academic performance. Hence, these mandates have the potential to help schools by shifting cost burdens to insurers, increasing early intervention and outside therapy that could improve student performance, and helping schools target and coordinate services.

In this study, we assess the effects of Michigan’s coverage mandate on autism identification, test scores, grade retention, test accommodations, support services, and educational settings using individual student level data from the Michigan Department of Education covering 2009 through 2015. Michigan adopted its autism insurance mandate in 2012 requiring that state regulated insurance plans cover therapy for ASD up through age 18. Self-insured plans, which are regulated under Federal law, were given generous financial incentives to adopt the coverage. The law also required Medicaid and CHIP to provide coverage but until 2016, lack of funding stopped coverage upon reaching age 6 considerably weakening the benefit. Since we cannot directly observe insurance coverage, we proxy for whether a child was covered under Medcaid/CHIP or uninsured using economic disadvantage status. Hence, we treat this as a dosage intention-to-treat where non-disadvantaged students are more likely to be privately insured and Medicaid/CHIP provides an inferior benefit. Census data for Michigan suggests only 31% of children income eligible for reduced-price lunch are privately insured as opposed to 89% of ineligible children.

Using a difference-in-differences strategy we first estimate impacts on ASD incidence and find no effect. We then estimate triple difference models using non-ASD special education students as a comparison group and find evidence of service reductions and moving of non-disadvantaged ASD students to less restrictive environments. Finally, preliminary triple difference models suggest a negative impact on achievement and higher retention. We will buttress this analysis with another source of identification – the availability of Board Certified Behavior Analysists (BCBAs) within or close to a school district. BCBAs are the primary providers of ABA therapy and, for many insurance companies, therapy must be provided directly or under the supervision of a BCBA. Hence, we will exploit geographical and temporal variation in the availability of nearby BCBAs to provide additional variation in potential treatment.