The Social Returns to Mental Health Care: Evidence from a Dutch Payment Reform

I estimate the social returns to mental health treatment, based on exogenous variation in mental health care use and using a novel data set on mental health care use in the Netherlands linked to registry data on a wide variety on socioeconomic and demographic characteristics. A natural experiment--the nationwide introduction of a mental health care-specific annual deductible of €200 and the subsequent implementation of municipal policies to exempt lower-income individuals--allows me to obtain unbiased estimators of the effect of both a reduction of mental health care use and shifts in utilization between primary care, specialist care, and medication on (i) the use of other medical services, (ii) economic productivity and welfare receipts, and (iii) homelessness, crime and victimization.

Preliminary results indicate that, after four years of steady growth, the number of outpatient treatment episodes in the Netherlands decreased by 15·4 percent when cost sharing was raised in 2012. The response was stronger for patients from poorer neighborhoods compared to those in richer neighborhoods: a 16·9 percent versus a 13·1 percent decline, respectively. The number of emergency treatment episodes gradually increased throughout 2012, up 18·2 percent compared to the year before, possibly reflecting substitution to high-cost emergency care. We find evidence in support of the hypothesis that, while substantial, the demand response for treatment of severe mental illness is lower than for other illnesses.  

Unprecedented access to rich registry data on all 17 million inhabitants of the Netherlands allows me to investigate the relationship between mental health, treatment, and a variety of `hard' indicators of social functioning, and to obtain precise estimates even for very rare outcomes, such as mortality. Causal inference will primarily be based on variation in out-of-pocket expenditures along three dimensions: (i) above and below a certain municipality-specific income threshold, (ii) between municipalities, and (iii) over time, since municipalities adopted reimbursement policies at different moments in time. To this end, identification of causal effects will rely on instrumental variables, regression discontinuity design, and differences-in-differences methods. The outcomes of this project should be highly relevant for policymakers looking to reduce health expenditures, improve risk pooling, and to lower the social burden of mental illness.