Decomposing the Welfare Impacts of Medicaid MCO Steering

Wednesday, June 15, 2016: 12:20 PM
G65 (Huntsman Hall)

Author(s): Ted C. Rosenbaum; Devesh Raval

Discussant: Chapin White

Many states have recently moved from Fee for Service (“FFS”) Medicaid programs to a model in which Managed Care Organizations (“MCOs”) compete with each other for enrollees. In contrast to FFS programs, these MCOs have financial incentives to steer their members to lower cost hospitals. However, the extent and welfare implications of this steering are open questions.

Using Florida discharge data on labor and delivery, we find that Florida MCOs steer their members away from teaching hospitals, which are typically higher cost. Preliminary OLS estimates suggest that this steering makes women in MCOs 4 percentage points less likely to go to teaching hospitals than women in FFS. However, this naïve estimate is likely biased upwards due to selection away from MCOs by women who particularly value teaching hospitals. Therefore, to obtain estimates of the causal impact, we use both fixed effects and instrumental variable (“IV”) approaches. We use the panel structure of the data to account for member specific fixed effects and an instrument for participation in MCOs that uses variation from two recent policy changes within the state. Both of these approaches suggest the steering makes women in MCOs 2 percentage points less likely to go to a teaching hospital than women in FFS.

We assess welfare impacts by estimating the impact of steering on members’ need to travel and hospital quality. To estimate these relative impacts, we disentangle consumers' valuation for proximity from unobserved tastes that are correlated with it. This correlation could arise from hospitals’ catering to people living in the proximal area or from local referral patterns, and is likely to be of concern in other applications as well.

In our application, we disentangle these valuations by using a fixed effects approach based on Chamberlain (1980) and Honorè (2000).[1] Our preliminary results show that a naïve approach which ignores the correlation between distance, MCO choice, and patient preferences, understates the relative impact of steering on both patient travel and quality by half. Our fixed effects estimator shows that MCO steering is equivalent to an additional 8 minutes of travel and 30% of the quality difference between the first and third quantile hospitals. We also find that in the fixed effects approach, steering’s impact on patients is equivalent to 70% of patients’ cost from switching hospitals on their own. The naïve approach understates this by 70%.

Finally, we present preliminary evidence suggesting that physicians are a primary mechanism for this steering. We find that women who switch into MCOs are less likely to have the same attending physician for their second child than other women. Further, when we focus on women that use the same attending physician for their first and second child, we find that there is no steering away from teaching hospitals.


[1] In future drafts we intend to intend to test the robustness of these results by using an inequality estimator based upon Pakes et.al. (2015) or Pakes and Porter (2015).