Exploring Heterogeneous Effects of Health Insurance Using the Oregon Health Insurance Experiment: An Application of Instrumental Forests

This event has now finished.
  • Date and time: Wednesday 2 November 2022, 1pm to 2pm
  • Location: A/D271, Alcuin College, Campus West, University of York (Map)
  • Audience: Open to staff, students (postgraduate researchers only)
  • Admission: Free admission, booking not required

Event details

This event is hosted by Nigel Rice.

Expanding Medicaid has been shown to impact health, health care utilisation, and financial hardship outcomes in the US. However, heterogeneity of these impacts among different subpopulations has not been fully explored.  Data: This study used data from the Oregon Health Insurance Experiment (OHIE), conducted in 2008 in Oregon, USA.

In the OHIE, individuals on a waiting list for health insurance were given the chance to apply for Medicaid if they were selected randomly by lottery, while those not selected were not offered this opportunity. Lottery selection acts as an instrument for whether the individual has health insurance. We focus on a sample of 12,229 respondents to a survey carried out in Portland, Oregon, metropolitan area between September 2009 and December 2010, an average of 25 months after the lottery began (Baicker et al, 2013).

We apply causal forest and instrumental forest methods to estimate the heterogeneous treatment effects of lottery selection and health insurance respectively. We compare our overall estimates to those obtained using ordinary least squares (OLS) and two-stage least squares (2SLS). As is often the case, we find that confidence intervals using instrumental variable approaches are quite wide. We conduct a simulation study to explore whether Stein-like estimation (Hansen, 2017) can be helpful in this context. 

The results of this study mostly coincide with the findings in the literature regarding the overall effects: insurance (and lottery selection) reduces out of pocket spending, increases the number of physician visits and the number of drugs prescribed, but has little (short-term) effect on the number of emergency department visits or the number of hospital admissions. Although the impact of the lottery on the probability of uptake of health insurance is heterogeneous among some pre-specified subgroups,
we detect fairly weak evidence of heterogeneity in the effects of the lottery and of health insurance across the outcomes considered.

The short-term effects of health insurance overall are fairly modest. While there is some evidence of heterogeneity in effect sizes across pre-specified subgroups, there is considerable uncertainty in the estimates rendering the evidence of heterogeneous effects of health insurance fairly weak overall, motivating the exploration of a Stein-like approach.


  • Baicker, K., Taubman, S.L., Allen, H.L., Bernstein, M., Gruber, J.H., Newhouse, J.P., Schneider,
    E.C., Wright, B.J., Zaslavsky, A.M. and Finkelstein, A.N., 2013. The Oregon experiment—effects of
    Medicaid on clinical outcomes. New England Journal of Medicine, 368(18), pp.1713-1722.
  • Hansen, B.E., 2017. Stein-like 2SLS estimator. Econometric Reviews, 36(6-9), pp.840-852.

About the speaker

Stephen O'Neill (London School of Hygiene and Tropical Medicine)