When the Affordable Care Act or “Obamacare” passed five years ago, government prognosticators and private research organizations projected between 21 and 27 million exchange enrollees in 2016. But last month, the Obama administration announced that it reduced that target to just 10 million enrollees.

In 2010, the Urban Institute, a left-of-center think tank, projected that almost as many people with income above four times the poverty level ($47,080 for a single person and $97,000 for a family-of-four in 2015) would enroll in exchange plans as people with income less than half these amounts. Urban’s projections, which were fairly representative of other groups, are in hindsight wildly optimistic. It turns out that more than 25 times as many people with income below twice the poverty level purchased an exchange plan as people with incomes above four times the poverty level.

In a new research study published by the Mercatus Center at George Mason University, I explore why ACA enrollment projections were so wrong. Here are the most likely explanations:

First, for most uninsured people, the costs of enrolling in ACA plans exceed the benefits. For example, a recent economics study found that a typical single person making $40,000 is generally $1,500 to $3,500 better off by remaining uninsured as opposed to purchasing an exchange plan.

Second, exchange plan deductibles are very high. In 2015, deductibles averaged about $3,000 for single coverage and $6,000 for family coverage for the most common type of plan. Tellingly, the only people who enrolled in exchange plans in large numbers—those below twice the poverty level—are the only ones who qualify for large subsidies to reduce these deductibles.

Third, exchange plans contain fewer doctors and hospitals than expected.

Fourth, the individual mandate isn’t motivating as many people to enroll as anticipated. The economic modelers believed that the mandate would create a “new social norm to have health coverage [that] can lead to behavioral responses much stronger than the nominal amount of the penalty would suggest.” However, few people who paid the mandate penalty in 2014 actually signed up for exchange plans in 2015 during a six-week special enrollment period created just for them. Moreover, hardship exemptions that relieve people who had any difficulty paying premiums from having to pay the penalty have weakened the mandate’s effectiveness.

Fifth, the ACA requires insurers to offer coverage to all applicants at standard rates regardless of health condition. Large net attrition in exchange enrollment in both 2014 and 2015 suggests that many people may be dropping coverage after treatment.

In addition to lower-than-expected enrollment, insurers are losing more money than anticipated. Just last year, the Congressional Budget Office released projections that insurers would make profits on exchange plans.

Using data released by the administration, however, I estimate that insurers lost about 12 percent of ACA plan premiums in 2014. Insurers lost this much despite a huge government subsidy to cover most of the cost of their expensive enrollees. This subsidy delivered $8 billion to insurers in 2014, but it’s scheduled to end after the 2016 plan year. Starting in 2017, premiums will for the first time reflect the true cost of enrollees unless the administration devises a creative way to further subsidize insurers beyond what the law seemingly allows.

In sum, double-digit premium increases in 2016 are a product of disappointing enrollment and insurers’ realization that a larger proportion of their enrollees are sicker and older than they expected. As premiums increase, deductibles are rising and provider networks are actually shrinking. These changes make ACA plans even less desirable to people who aren’t already sick or don’t qualify for large subsidies.

The magnitude of the errors of initial predictions about the ACA’s effect is cause to re-examine our assumptions about health care markets. The failure of exchange plans to attract people who don’t receive giant subsidies should cause policymakers to revisit the law and allow people to purchase insurance products that they actually want.