September 2007

Turbo-Charging Consumer-Driven Health Plans with Tax Reform: Cost and Coverage Effects

Recent research by University of Minnesota professors Stephen Parente and Roger Feldman finds that combining the most recent models of consumer-driven (or directed) health plans (CDHPs) with a reform of the tax treatment of health insurance would substantially increase overall health insurance coverage. They estimate that adopting the George W. Bush plan for a standard health deduction in 2009--which levels the playing field between employer-purchased insurance and individual insurance--could reduce the ranks of the uninsured by more than 20 million people. A different universal tax credit proposal by Senator Tom Coburn (R-Okla.) and several Senate colleagues would have similarly powerful effects in reducing the ranks of the uninsured. Parente and Feldman also find that better-designed consumer-directed health plans help reduce health care costs. At this event, Parente and Feldman will present findings from their most recent national simulation model of recent health insurance reform options. They have used five years of employer and health plan experience in their study, making it the longest-running economic analysis of consumer-directed health plan design.

Gary Claxton, director of the Kaiser Family Foundation’s Health Care Marketplace Project; Anthony Lo Sasso, associate professor at the University of Illinois at Chicago; and Phil Ellis of the Congressional Budget Office discussed this research and other early lessons of consumer-driven health experiments.

Stephen Parente
University of Minnesota

CDHPs are health care designs that were first developed around 2000. They are high-deductible health insurance plans with accompanying health spending accounts (HSAs) created to cover out of pocket health costs. They generally have preventive health care engineered into the plan design. The high deductible is supposed to curb health care expenditures, and we do find that pharmacy expenditures tend to decrease among CDHP policyholders.

When CDHPs were first introduced, there was concern that such plans would have adverse selection effects. It was thought that only young or healthy people would choose such high-deductible plans. The evidence is more mixed. CDHPs come in two varieties, those with HSAs and those with health reimbursement accounts (HRAs). In HSAs, the spending account is comprised of real dollars, while in HRAs, the spending accounts are nominal dollars that belong to the employer. HSAs in some instances exhibit favorable selection, having enrollees who are healthier on average than high option HMO/PPO plan choosers. Bear in mind that when HMO plans were first introduced, they also experienced favorable selection. HRAs however, exhibit unfavorable selection, having enrollees among the sickest in the population and in some areas even more so than those choosing a high option PPO. This is an unintended consequence--the employers we looked at had no idea this was going on.

The adjusted choice and outcomes legislative assessment (ARCOLA) model predicts that Bush’s State of the Union proposal would lead to a 65 percent reduction in the uninsured by 2009 at a cost of $250 billion--$100 billion going to the individual market and the remainder going to the group market. This cost will decrease over time as people choose cheaper health insurance, but there is no evidence of a break-even point being reached within the next ten years. The model predicts a 39 percent reduction in uninsured by 2009 at a cost of $160 billion for Senator Coburn’s proposal. This cost would also drop over time, but once again, a breakeven point would not be reached within the next 10 years.

We can draw three conclusions from this analysis. The first is that these proposals could represent the most comprehensive U.S. health insurance reform ever, as evidenced by the predicted effects from the ARCOLA model. Second, Bush’s proposal would have greater impact on reducing the number of uninsured, but at a substantially greater cost than Coburn’s proposal. Third, and perhaps most important, is that it is possible to substantially reduce the number of uninsured without an employer mandate.

Roger Feldman
University of Minnesota

In his 2007 State of the Union address, Bush outlined a proposal to provide a standard tax deduction ($7,500 for an individual and $15,000 for a family) for purchasers of health insurance as an incentive for the uninsured to cover themselves. Senator Coburn proposed a tax credit ($2,000 for an individual, an additional $2,000 for a spouse, and $500 for each dependent up to a total of $5,000 per household). The ARCOLA model estimates the impact of these proposals, but in a way better suited to them than traditional models. Most models, are built around elasticities, or conversion factors, that allow one to estimate from the percentage change in one variable, such as the price of a policy, the percentage change in another variable, or the take-up rate of health insurance. Most models also try to convert proposals into price effects--a subsidy or tax--for analysis. We have modeled the price effects and elasticities on the behavior of over 250,000 employees in four national corporations as they made decisions over features of their health insurance plans, such as amount of deductible or coinsurance payment, choice of CDHP, size of spending account, and out-of-pocket expenditure gap. In this way, we could identify a premium elasticity response to policy without cramming policy into a price effect. The Bush proposal is not a subsidy as much as it is an addition to income. As such, it deserves this different modeling approach.

When the Medicare Modernization Act was passed in 2003, we used the ARCOLA model to predict the take-up of the newly authorized tax-preferred HSAs, and we obtained results that were fairly accurate. The model looks at a subset of the population and excludes children, students, those eligible for Medicare, and those eligible to be insured through someone else’s policy. There are 33.7 million uninsured in this subset of the U.S. population. We also assumed that tax deductions and credits would be actionable at the point of purchase, that individual coverage would be universally available, premiums would be adjusted to the relative health risk of a population, and the medical consumer price index would remain constant at 4 percent above general inflation. Our results are pegged against a moving target of the number of uninsured, which we project would grow to 49 million in our sample population if no measures are taken.

Gary Claxton
Kaiser Family Foundation

The Parente/Feldman model is good because it adds real people to population statistics on CDHPs. Kaiser’s models only get a couple hundred responses to surveys, and thus it is hard to get feedback on the same scale. From our research, we find that HRAs and HSAs can be very different in terms of claims, deposit amounts, and employer contributions. It could also be said that there are two HSA markets. One is the large employers trying to incorporate the HSA plans into their policy options, and these tend to resemble HRAs. The other market is comprised of small employers offering HSAs as high-deductible plans and not contributing to the savings account. There are things about large employers that cannot be generalized to all employers, so analysis based on research from a few large employers should be approached with caution. Also, while HRAs do have prevention worked into the plan, HSAs seem not to, or at least, not to the same extent as the HRAs.

HSAs make having insurance more desirable because one can pay for health expenditures with tax preferred dollars. A $2,000 high-deductible plan without an HSA has got to cost less than a $2,000 high-deductible plan with an HSA because consumers can use pretax dollars to pay for the cost share. So while there will be an incentive to buy non-group insurance because of HSAs, there also has to be an offsetting price effect that could make insurance more expensive, and I have yet to see this price effect modeled.

Finally, there is a difference between the people who would be affected by the Coburn proposal and those that would be affected by Bush’s proposal. At the low-income end, there are people who would benefit substantially from a tax credit but not at all from a tax deduction.

Anthony Lo Sasso
University of Illinois at Chicago

About half the uninsured face a marginal tax rate of zero. These people would be unaffected by Bush’s proposal for the tax treatment of health insurance. Another quarter have a marginal tax rate of ten percent, and the remaining quarter might have a marginal tax rate of 25 percent. Even if we assigned an arguably high elasticity between policy pricing and insurance take-up, the total reduction of the uninsured would be far below the estimates of the ARCOLA model. While elasticity between policy choices among people already insured might be high, evidence suggests that the take-up elasticity for the uninsured is much lower. We should keep in mind that there are people who do not take up free insurance in the form of Medicaid.

The preliminary results of my analysis of a health insurance company that has been in the small group market since 2000 show that each dollar in a HSA actually leads to three dollars more of health care spending. But each dollar in the “doughnut hole”--that expenditure area between the amount in the HSA and the total deductible--leads to fifty-seven cents less health care spending, but nine cents more out of pocket expenditure. This confirms the importance and potential promise of the new cost-sharing plans, but this research is incomplete, and more analysis will need to be done.

Phil Ellis
Congressional Budget Office (CBO)

CDHPs were created to address a trend in health expenditures. Out-of-pocket private expenditure for health services has remained relatively flat over the last two decades at roughly two percent of GDP, whereas the health expenditure as a percent of GDP has grown steadily. A RAND experiment found that conventional insurance, with a deduction and coinsurance for additional costs showed a dramatic reduction in health expenditure as opposed to zero-deductible insurance. Yet a very-high-deductible plan ($4,000) showed only marginal improvement over a conventional plan, and a $2,000 deductible plan actually resulted in more health expenditure than a conventional plan.

In terms of selection, we find that there are fewer selection pressures with an HRA design than with an HSA design, but there is adverse selection between high-deductible plans and conventional ones. This makes intuitive sense: healthy people would pay less in a consumer-directed design, as premiums are lower. Relatively, however, the cost of extensive care is cheaper in a conventional design with a lower deductible.

The question at the heart of how preferential tax treatment will affect the rate of the uninsured depends on three things: the premium of a particular plan, the tax deduction an individual receives, and the take-up response as a result of reduction of costs. The effects will be different for those with different incomes and cost of insurance. For a young, healthy person with high personal income, Bush’s tax proposal would more than pay for the low cost of insurance. For an older, less healthy person with lower income, the effect of a deduction would result in less of a discount from an already more expensive plan. This person would have far fewer incentives to participate. The CBO estimates that the Bush proposal would lead to a net increase in the insured population of about 7 million by 2010.

AEI research assistant Walton Dumas prepared this summary.