Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
Understanding how market structure affects health outcomes is difficult because of the nonrandom nature of where patients receive care. I develop a control function approach that is robust not just to selection on unobservable differences in patients’ condition severity, but also to selection on unobservable variation in how patients respond to different treatments. I apply the approach to data from 2004 to 2008 for all hemodialysis patients in Atlanta, Georgia. The results indicate that a one-facility acquisition by the average provider in the average market would lead the average patient to spend 28 percent more days in the ICU/CCU. Estimates that do not control for the role of idiosyncratic matching would predict smaller, though still economically and statistically significant, effects on the affected patients. I interpret the results as indicating the importance of idiosyncratic matching in this market, which multifacility firms address by differentiating their operations.