When the Lewin Group speaks, people listen.  In 2009 Lewin projected that a public plan option could reduce private insurance coverage by two-thirds, a finding that was used to strip the public option from the reform law.  Knowing this impact, it sent shivers down my spine when, on Halloween, I read their latest study, published as a Health Affairs Web First article.  It suggests that defenders of the Affordable Care Act’s (ACA) individual mandate are wrong to claim that eliminating the mandate would necessarily cause an adverse selection premium spiral in the individual market.   Instead, John F. Sheils and Randall Haught concluded “that although the mandate has important effects on premiums and coverage, it might not be essential to the act’s successful implementation.”

“Essential to . . . implementation” is the same legal standard the 11th Circuit ruled the government had failed to meet in defending the mandate’s constitutionality.  Therefore, we can expect mandate opponents to cite Lewin’s study prominently in upcoming Supreme Court briefs.  The study might also help argue for severability of the mandate, to preserve the rest of the ACA, in the event the Court strikes the mandate.  Accordingly, it merits close scrutiny.

The study concludes that the mandate may not be essential because other ACA features encourage insurance purchase and guard against adverse selection — principally, subsidies and a limited open enrollment period.  Thus it projects that dropping the mandate would reduce coverage by only 8 million people, about half the 16 million loss projected by the Congressional Budget Office, and a third the 24 million loss projected by MIT economics professor Jon Gruber.  Not cited by Sheils and Haught, but also contradicting their findings, are studies by the RAND Corp. and the Urban Institute, which, using similar methods, projected coverage losses from dropping the mandate of 13 million and 18 million, respectively.

What Makes The Lewin Study An Outlier?

As Sheils and Haught acknowledge, microsimulation has considerable uncertainty due to key behavioral assumptions required in the absence of convincing empirical evidence.  One key assumption in dispute is whether the individual mandate will induce more purchase of insurance than might be expected based solely on its weak financial penalties.  Other modelers have assumed that people attach some benefit to complying with the law beyond merely the coarse economic calculation of avoiding a fine.  Sheils and Haught refuse to acknowledge any general law-abidance or social norm factor, claiming lack of any supporting evidence and speculating that it is equally possible people would behave the opposite way – refusing to comply simply because they oppose government mandates or subsidies.

These assumptions are admittedly within the realm of theoretical possibility, but it is stubborn to refuse to test other, more realistic, possibilities, in a modeling exercise whose main point is to measure possible effects of the mandate’s legal requirements.  Multiple sources of evidence from Massachusetts, which others rely on but Sheils and Haught do not cite or consider, strongly suggest that Massachusetts’ somewhat lighter mandate incentives have produced substantially more compliance than predicted.  The impossibility of quantifying these legal compliance effects for precise extrapolation under the ACA is no reason to ignore them in a modeling exercise already riddled with other similar uncertainties.

For instance, Lewin’s analysis makes assumptions about the behavioral boost that comes from limiting open enrollment (see here at pp. 24-25).  They note that the inability to enroll for up to 11 months (or, more accurately, an average of 5.5 months, subject to many exceptions) will encourage enrollment even without a mandate.  True enough, to an extent, but when Sheils and Haught quantify this enrollment boost, they assume that, for people with a chronic condition, it is just as strong as their current incentive to enroll in the face of medical underwriting and exclusion of pre-existing conditions.  Stated otherwise, they assume that banning medical underwriting produces no increased tendency among people with a chronic condition to wait to purchase coverage until they feel a strong need to use insurance.  That assumption is certainly a stretch lacking any real evidence, and again assumes away the very thing they should be modeling.

The Evidence To Date Does Not Support Lewin’s Conclusions

The authors’ conclusion of no serious adverse selection from dropping the mandate is also rebutted by real-world reality checks.  Market failure has already happened under two other ACA provisions.  When the new law barred medical underwriting for children three years prior to the mandate taking effect, most major insurers simply withdrew from the market for child-only coverage.  Just last month, HHS abandoned plans to implement the ACA’s CLASS Act provisions for long-term care insurance, based on actuaries’ analyses that adverse selection would make the program unsustainable.

The likelihood of crippling market effects is also suggested by experiences in every state that has attempted to implement guaranteed issue and community rating in the individual market without a purchase mandate.  Last year, Mr. Sheils drew from this evidence to conclude that universal coverage “makes it possible to require guaranteed issue without creating a premium spiral,” but here he and colleague Haught dismiss compelling evidence from New York and other states.  They explain there is no proof these laws have reduced coverage population-wide (including Medicaid and group insurance), but that’s a bit like ignoring a raging forest fire until it changes the weather.  Even without measurable declines in all forms of coverage, there is no reasonable dispute that banning medical underwriting without a purchase mandate has consistently crippled markets for individual insurance.

The leading New Jersey study concluded that its individual market “appears to be heading for collapse . . . [due to] an enrollment crisis that threatens its market stability,  . . . consistent with a marketwide adverse-selection death spiral spurred by open enrollment and pure community rating.”   A recent New York study documented that its market for individual health insurance “has nearly disappeared, declining by 96 percent since 1994.”  The insurance industry’s Supreme Court brief documents several other well-known examples.

Eliminating The Individual Mandate Would Be A Foolish Risk

In the end, what matters most is not whose model is the best: Lewin’s vs. any one of the other contenders (CBO, Urban, RAND, Gruber).  That, we’ll never know, since we’ll never get to see events play out both with and without a mandate.  But this we do know: with a mandate, all models predict stable markets.  Without a mandate, there is at least great uncertainty about whether the reform structures would survive intact, and a real risk they would crumble.  Without a mandate, perhaps the individual market would stumble along somehow, but at least it would deteriorate: coverage would drop, and prices would increase, forcing government to subsidize higher premiums in order to cover fewer people.  That is all one needs to know to conclude that the mandate is an essential lynchpin for the ACA’s success.