Should There be Vertical Choice in Health Insurance Markets?

with Adrienne Sabety

Choice over coverage levels—"vertical choice"—is widely available in U.S. health insurance markets, but there is limited evidence of its effect on welfare. For a given consumer, the socially efficient level of coverage trades off the value of risk protection and the social cost from moral hazard. Providing choice does not necessarily lead consumers to select their efficient coverage level. We show that in regulated health insurance markets, vertical choice should be offered only if consumers with higher willingness to pay for insurance have a higher efficient level of coverage. We test for this condition empirically using administrative data from a large employer. Our estimates imply substantial heterogeneity in efficient coverage level, but we do not find that households with higher efficient coverage levels have higher willingness to pay. It is therefore optimal to offer only a single level of coverage. Relative to a status quo with vertical choice, mandating the optimal single level of coverage increases welfare by $330 per household per year.


Regulating Markups in U.S. Health Insurance

with Steve Cicala and Ethan Lieber

American Economic Journal: Applied Economics, October 2019

A health insurer’s Medical Loss Ratio (MLR) is the share of premiums spent on medical claims, or the inverse markup over average claims cost. The Affordable Care Act introduced minimum MLR provisions for all health insurance sold in fully-insured commercial markets, thereby capping insurer profit margins, but not levels. While intended to reduce premiums, we show this rule creates incentives to increase costs. Using variation created by the rule’s introduction as a natural experiment, we find medical claims rose nearly one-for-one with distance below the regulatory threshold: 7 percent in the individual market, and 2 percent in the group market. Premiums were unaffected.

Narrow Networks on the Health Insurance Marketplaces: Prevalence, Pricing, And The Cost Of Network Breadth

with Leemore Dafny, Igal Hendel, and Christopher Ody

Health Affairs, September 2017

Anecdotal reports and systematic research highlight the prevalence of narrow-network plans on the Affordable Care Act’s health insurance Marketplaces. At the same time, Marketplace premiums in the period 2014–16 were much lower than projected by the Congressional Budget Office in 2009. Using detailed data on the breadth of both hospital and physician networks, we studied the prevalence of narrow networks and quantified the association between network breadth and premiums. Controlling for many potentially confounding factors, we found that a plan with narrow physician and hospital networks was 16 percent cheaper than a plan with broad networks for both, and that narrowing the breadth of just one type of network was associated with a 6–9 percent decrease in premiums. Narrow-network plans also have a sizable impact on federal outlays, as they depress the premium of the second-lowest-price silver plan, to which subsidy amounts are linked. Holding all else constant, we estimate that federal subsidies would have been 10.8 percent higher in 2014 had Marketplaces required all plans to offer broad provider networks. Narrow networks are a promising source of potential savings for other segments of the commercial insurance market.


Oscar Health Insurance: What Lies Ahead for a Unicorn Insurance Entrant?

with Leemore Dafny

Harvard Business School Case 319-025, August 2018


All Medicaid Expansions Are Not Created Equal: The Geography and Targeting of the Affordable Care Act

with Craig Garthwaite, John Graves, Tal Gross, Zeynal Karaca, and Matthew Notowidigdo

Brookings Institute, September 2019; NBER Working Paper No. 26289

We use comprehensive patient-level discharge data to study the effect of Medicaid on the use of hospital services. Our analysis relies on cross-state variation in the Affordable Care Act’s Medicaid expansion, along with within-state variation across ZIP Codes in exposure to the expansion. We find that the Medicaid expansion increased Medicaid visits and decreased uninsured visits. The net effect is positive for all visits, suggesting that Medicaid leads the uninsured to consume more hospital services overall. The increase in emergency department visits is largely accounted for by “deferrable” medical conditions. Lastly, we find significant heterogeneity across Medicaid-expansion states in the effects of the expansion, with some states experiencing a large increase in total utilization and other states experiencing little change. We investigate the sources of this heterogeneity with an eye towards predicting the effects of future expansions or disenrollments.

Winners and Losers Under Counterfactual Health Risk Pooling

with Benjamin Vatter

We study public policy proposals that would decouple health risk pools from employment pools in the US. These policies, such as ‘Medicare for All,’ would pool health risk at the state or national level, overturning the status quo of pooling risks at the firm level. Generally speaking, firms with on average healthy employees would be worse off, and firms with on average sick employees would be better off, but little is known about the extent of existing variation along this dimension. We analyze a large, national data set of individuals with employer-sponsored health insurance and present novel evidence on the variation in average health spending across firms. We find that {55} percent of firms (representing {38} percent of individuals) have average health spending that is below the national average. We also find that {9} ({2}) percent of firms have average health spending that is two times lower (higher) than the national average, suggesting that big winners and big losers would exist. We then construct the joint distribution of firms’ average health spending and average income and find that pooling health risk across firms would on net be regressive, distributing {4} percent of the national health spending bill from the lowest to the highest firm income quartile.