Health Affairs, 24, no. 6 (2005): 1582-1591 doi: 10.1377/hlthaff.24.6.1582 © 2005 by Project HOPE	New Online   Senate Health Reform Bill   Rewarding Providers   Public Option Policy Brief   Health Reform & Abortion   Delivery System Reform

This Article Abstract Reprint (PDF) Submit a response to this article Alert me when this article is cited Alert me when Comments are posted Alert me if a correction is posted Services E-mail this article to a friend Similar articles in this journal Similar articles in Web of Science Similar articles in PubMed Alert me to new issues of the journal Add to My Personal Archive Download to Citation Manager Reprints & Permissions Citing Articles Citing Articles via HighWire Citing Articles via Web of Science (6) Citing Articles via Google Scholar Google Scholar Articles by Feldman, R. Articles by Taylor, R. Search for Related Content PubMed PubMed Citation Articles by Feldman, R. Articles by Taylor, R. Related Collections Access To Care Health Reform Insurance Coverage Managed Care Managed Care - Consumers Business Of Health Health Spending Consumer Issues Insurance Market

Health Tracking

MARKETWATCH

Health Savings Accounts: Early Estimates Of National Take-Up

Abstract

 
The 2003 Medicare Prescription Drug, Improvement, and Modernization Act (MMA) approved tax-advantaged health savings accounts (HSAs) for certain high-deductible health insurance plans. We predict that MMA could lead to approximately 3.2 million HSA contracts among Americans ages 19–64 who are not students, not enrolled in public health insurance plans, and not eligible for group coverage as a dependent. We simulate the effect of several additional tax subsidies for HSAs. We predict that the Bush administration’s refundable tax-credit proposal would double HSA take-up and reduce the number of uninsured people by 2.9 million, at an annual cost of $8.1 billion.

Until recently, these plans typically were offered to employees of large, self-insured employers, often as an option alongside traditional health insurance products.² Early indications are that they are a viable alternative to existing plan designs.³ They also were available to employees of small businesses and the self-employed through a 1997 federal demonstration project. That demonstration never caught on, because either the employer or employee, but not both, could contribute to the account and because the number of policies that could be sold under the demonstration was limited to 750,000, which dampened suppliers’ interest in selling such products.

Recently, however, consumer-directed plans received a boost from the Medicare Prescription Drug, Improvement, and Modernization Act (MMA) of 2003. Section 1201 of MMA (and subsequent guidance by the Treasury Department) approved a new form of plan known as a health savings account (HSA). Beginning 1 January 2004, anyone could purchase a health plan with an annual deductible of at least $1,000 for an individual and $2,000 for a family, coupled with a tax-advantaged account to which both the employer and enrollee may contribute. Total annual contributions can be as large as the plan’s deductible (up to $5,000 for an individual and $10,000 for a family). Unlike previous designs, the HSA is fully portable, so a person may use it without being dependent on the provisions of a particular employer.

Mainstream insurers such as Blue Cross Blue Shield plans and UnitedHealth Group are selling these tax-advantaged HSAs.⁴ To facilitate its entry into this market, in November 2004 UnitedHealth Group purchased Definity Health of Minneapolis, an early leader in marketing consumer-directed plans. United-Health had previously purchased Golden Rule Insurance Company of Indianapolis, which sold health plans with non-tax-preferred savings accounts to individuals and employers.

HSAs might receive another boost if Congress enacts a Bush administration proposal to create a refundable tax credit for people under age sixty-five to purchase health insurance plans with HSAs. In one form of this proposal, the credit would provide a subsidy of up to 90 percent of the insurance premium.⁵ The maximum credit for low-income taxpayers would be $1,000 per adult and $500 per child (up to two children). The subsidy would phase out as income increases.

Policy analysts have sharply conflicting opinions about the wisdom of this tax credit. Some critics are concerned that a further tax subsidy for individual coverage might lead to the "hollowing out" of the market for group coverage as low-risk enrollees leave group insurance pools.⁶ Others, however, believe that more favorable tax treatment would spur the development of a mass market for individual coverage that would make it more broadly attractive.⁷

The purpose of our research is to examine the potential of individual HSAs for increasing the number of insured Americans, especially those with low incomes. In particular, how much HSA take-up is expected from MMA? What is the additional impact of the administration’s proposed tax credit and other possible subsidies? By how much will these proposals reduce the number of uninsured people? And how much will the subsidies cost?

Data And Analytic Approach

Top Data And Analytic Approach Study Results Conclusions NOTES

 
We addressed these questions by estimating a health plan choice model for employees of three large companies that offer consumer-directed plans. Results were then used to simulate how many people in a nationally representative survey would choose HSAs. For each simulation, we predicted the number of people taking up HSAs and the cost of the proposal.

Data sources. We used three data sources for our analysis: the Medical Expenditure Panel Survey (MEPS); health plan choice data from three large employers; and data for individual HSA policies from the eHealth insurance.com Web site.

MEPS is an annual survey of the U.S. non-institutionalized, civilian population. We used two files from the 2001 MEPS (the latest data available). The first is the Household Component (MEPS-HC), which contains detailed demographic, employment, and insurance information on a nationally representative sample of individuals. We restricted our attention to people ages 19–64 who were not enrolled in public insurance programs and not full-time students. For reasons explained later, we also excluded adults who were not offered insurance by their employer but who could be covered by someone else’s group insurance plan. When weighted to produce population estimates, the sample used in this study corresponds to 121,535,688 nonelderly Americans.

The second file is the MEPS-HC-Insurance Component (IC), in which a random sample of people who reported being employed and offered health insurance in Round 1 of MEPS-HC were asked to provide contact information for their place of employment. Their employers were surveyed regarding the number and types of plans offered to eligible workers.

We had access to data on the 2002 health plan choices of employees, representing about 80,000 covered lives (including dependents), of three large employers.⁸ Each employer offered a consumer-directed plan that received a first-year take-up rate of 4–15 percent. The type of plan offered by these employers was a health reimbursement account (HRA) in which the employer places tax-free credits (not real dollars) in an account that the employee can use to pay for eligible medical expenses. The account rolls over from year to year, but, unlike for an HSA, the employee does not own it, and it is not portable.

Our data for individual policies came from two surveys conducted by the Health Insurance Association of America (HIAA) in 2002 and 2003, plus current information on individual HSA policies from eHealthinsurance.com, which provides an estimated monthly premium cost based on county, enrollee’s age, family size, and health history.⁹

These data sources were used for three major analysis steps: model estimation; choice set assignment/prediction; and policy simulation.

Model estimation. In the first step, using pooled data from the three employers that offer consumer-directed plans, we estimated a conditional logistic plan choice model, similar to our earlier work.¹⁰ This step produced coefficient estimates that represent the utility of each plan attribute to the employee.¹¹

Prediction of plan choices. In the second step we used the estimated choice-model coefficients to predict health plan choices for individuals in MEPS-HC. To complete this step, it was necessary to assign the number and types of health insurance choices that are available to each respondent in MEPS-HC. For this purpose, we turned to the smaller but more detailed MEPS-HC-IC file, which contained the needed information. We summarize the details of this process and direct interested readers to a longer version of the paper on our Web site, www.ehealthplan.org.

The process can be described as a "crosswalk" between the two MEPS files. To use a specific example, government employees in the linked MEPS file are offered more plans by their employers than are other workers, on average. Suppose the average government employee had three offers. We "walked" back to MEPS-HC and assigned three plans to each government employee in that data set. The actual crosswalk was done by multivariate regression models.

Next, in the linked MEPS file, we identified the types of plans that had the maximum probability among the options that were offered. For example, those who were offered three plans were most likely to have a choice of high- and low-coverage preferred provider organizations (PPOs) and one health maintenance organization (HMO). We assigned these choices to government employees in MEPS-HC.

To predict the premiums associated with these choices, we estimated "hedonic" premium equations from the linked MEPS file as a function of hospital and physician coinsurance (or copayment) rates and the plan’s annual deductible. We used the estimated coefficients and the characteristics of the two PPOs and one HMO to predict premiums for the plans that were offered to our typical government worker.

To obtain employees’ out-of-pocket premiums, we multiplied total premiums by the average proportion paid by employees for single or family coverage from the linked MEPS data. Out-of-pocket premiums were adjusted by 1 minus the employee’s estimated federal tax rate for employees who paid their share of the premium with pretax income.¹²

Consumer-directed plans were not available in 2001, and tax-advantaged HSAs were not available until 2004, so we had to use a different strategy to assign consumer-directed plans to some workers in MEPS-HC. Because large employers have shown the most interest in consumer-directed plans, we assumed that all workers in firms with more than 500 employees will be offered two consumer-directed plans.¹³ One of these was modeled on the HRAs offered by the three employers for which we had data in 2002. Another was a less-generous HRA with cost-sharing characteristics typical of an HSA. We did not assign a consumer-directed plan to employees in establishments with fewer than 500 workers.

Our approach has three potential shortcomings. First, because our employer data were collected before 2004, they did not include HSA offers. We assumed that employees’ preferences for an HSA could be simulated by their preferences for the less generous HRA. This assumption no doubt affects the accuracy of our simulations but seems the most reasonable strategy available.¹⁴ Second, the simulations could be affected if consumer-directed plans prove to be less popular among large employers than we assumed. Thus, our estimates may be an " upper bound" of consumer-directed plan enrollment in employer-sponsored health benefit programs. Third, the "crosswalk" we used can be applied only to people in MEPS-HC with an offer of employer coverage. Those who were not offered such coverage but who could be covered as dependents by someone else’s (usually a spouse’s) group insurance plan had to be excluded.

We used a different algorithm to assign plans to individuals in MEPS-HC who did not have an employer’s coverage offer and were not eligible for coverage as a dependent. Before 2004, we assumed that such people had four choices: high-, medium-, and low-coverage PPOs and no insurance. The plan characteristics used to define these options were taken from the 2003 HIAA survey of plans purchased in the individual market. We used the 25th, 50th, and 75th percentiles of the distribution of plan attributes to define the coverage levels.

Because health insurance premiums in the individual market vary by a person’s age, we created an index using information from the 2002 HIAA survey. The index was set equal to 1.0 for the age group corresponding to the median age of adults in MEPS-HC. Older people, who had higher premiums, had index values greater than 1.0. Younger people had index values less than 1.0. After developing these indices and applying them to 2002 data, we scaled all premiums to 2005 prices.

Starting in 2004, we assumed that all people in the nongroup market would have access to an HSA. We relied on eHealthinsurance .com for current information on two HSA policies offered in the largest two cities in every state. Next, we estimated a hedonic premium equation to predict the premium for different HSA designs. For all of the simulations except one (described below), we used an HSA with a $1,000 spending account and $3,500 deductible for single coverage and $2,000/$7,000 for families.¹⁵ The average monthly premium for our prototype HSA for a forty-year-old, non-smoking, single male was $102.78 per month; the same person with a spouse and two children under age ten would pay $226.97 per month. The total cost of the prototype HSA is equal to the premium plus the enrollee’s contribution to the tax-advantaged account.

Finally, our health insurance choice model did not include alternatives for turning down the employer’s offer or for being uninsured in the individual market. To account for these possibilities, we added a "turndown" choice to the model for workers with an offer and scaled the utility of that choice so that a nationally representative 15 percent of employees with an offer would refuse it.¹⁶ For people without an offer, we added an "uninsurance" option that would be chosen by a nationally representative 57 percent of this group. Because uninsurance and turndown rates vary greatly by income and we wanted to determine whether HSAs would appeal to the low-income uninsured, we performed these estimates separately by income quartile.

Simulations. The third step in our analysis was to specify and perform the simulations. All of the simulations were conducted in relation to the MMA "baseline." The first simulation is the administration’s proposal to provide a refundable tax credit of up to 90 percent of the insurance premium, with a maximum credit of $1,000 per adult for single taxpayers with no dependents and annual adjusted gross income (AGI) less than $15,000.¹⁷ The subsidy would be phased down to 50 percent at $20,000 and zero at $30,000. These parameters were used to develop a sliding scale of tax credits with kinks at $15,000 and $20,000. We also modeled the tax credit applying to dependents, starting at $500 per dependent (limit of two children) for families with annual incomes less than $25,000 and declining according to the higher income kinks ($40,000 and $60,000) associated with families.

Given that one of the objectives of the tax-credit proposal is to reduce the number of uninsured people, we simulated an even more generous policy that subsidizes part or all of the insurance premium for lower-income individuals and families. Specifically, we set the HSA premium at zero for taxpayers with incomes of $15,000 or less, 50 percent of the premium for those with incomes of $15,000–$40,000, and 75 percent of the premium for those with incomes of $40,000–$60,000.¹⁸

In the third simulation, we simply set the total price of an HSA (premium and savings account) at zero. In effect, this proposal is a complete subsidy for the prototype HSA. As a final policy targeted at people without jobs, we created a simulation in which anyone not employed received a full subsidy for the premium and HSA, regardless of income.

Study Results

Top Data And Analytic Approach Study Results Conclusions NOTES

 
Each simulation begins with a comparison to the 2005 baseline situation. In Exhibit 1, for example, we see a take-up of 9 percent for HSAs in the nonoffered market without any additional change in policy. We attribute this impact to the relatively lower premium of the HSA in our simulations, compared with a PPO, and the high price elasticity associated with coinsurance. However, the market for employer-sponsored HSAs remains small as long as the employee can select a PPO or HMO with an employer premium contribution.

Simulation 2: low-income buy-in subsidy. Under this proposal, a greater share of the previously uninsured population would take up coverage (Exhibit 2). However, the cost is much higher: $10.8 billion per year for the non-offered population and $1.4 billion for the offered population, in which slightly more than one million people turn down their employer’s offer in favor of an individual HSA.

When we changed the design of the HSA to a $2,500 deductible for single coverage and $5,000 for families (with doughnut holes of $1,500 and $3,000, respectively), the premiums are higher, but the take-up rate for a "free" HSA of this type is much greater, with only 3.8 million people remaining uninsured (Exhibit 3). In other words, a generous "free" HSA could nearly eliminate uninsurance among the population considered in our simulations. However, the annual cost of the subsidy would be approximately $211 billion, much of it incurred in a "buy-out" of the employer health insurance sector.¹⁹

This example points to a trade-off between HSA generosity and the cost of the subsidy. More generous designs (for example, with smaller doughnut holes) will increase the take-up rate but also increase the cost of the subsidy. Also, as some have feared, a full subsidy for a generous HSA would have a strong impact on the group market.

Simulation 4: full subsidy for non-workers. One possible approach to prevent the erosion of employer coverage would be to exclude the working population from the HSA subsidy. To explore the consequences of this approach, our final simulation was a full subsidy targeted at people without jobs, regardless of income. The result is more take-up than with the administration’s proposal, but the subsidy cost, at $11.2 billion, is higher (Exhibit 4).

Comparative attractiveness to the low-income uninsured. Another standard for comparing the subsidy proposals is to assess their attractiveness to the low-income uninsured, a group that might have the most difficulty finding affordable health insurance. Approximately 42 percent of those taking up insurance under the administration’s proposal would come from the lowest quartile of the income distribution, and 75 percent would come from the lower half (Exhibit 6). However, the take-up rate among the upper half of the income distribution would be somewhat higher. The free HSA is the most attractive policy for the low-income uninsured, with 95 percent of the take-up coming from the lower half of the income distribution. This subsidy works better for low-income people because of the strong association between low income and not working.

Top Data And Analytic Approach Study Results Conclusions NOTES

Impact of subsidies on coverage and costs. The Bush administration’s proposed tax-credit plan would double individual HSA take-up and reduce the number of uninsured people by 2.9 million, at an annual tax cost of $8.1 billion—an average of $2,761 per person. A low-income buy-in subsidy would reduce the number of uninsured people by 16.5 percent (about 4.5 million people) at a cost of $12.2 billion annually, or an average of $2,718 per person. Offering "free" individual HSAs could, in theory, almost eliminate uninsurance but at a much higher per capita cost.

In addition to higher costs, "free" individual HSAs could greatly erode the market for employer-sponsored health insurance, with reductions of almost 5.7 million covered employees for the prototype HSA and 31.6 million for the more generous design.

Offering a free HSA to the nonworking, non–publicly insured population would not erode the employer coverage market. This simulation reduces the number of uninsured people, but less efficiently than a combination of other subsidies.

At least 70 percent of the take-up for the subsidies considered here would come from the lower half of the U.S. income distribution. The "free" HSA for nonworkers is the most attractive policy for the low-income uninsured.

Further HSA design considerations. Our work shows that people are sensitive to the size of the doughnut hole, with much larger take-up of the more generous HSA compared with the prototype design. Further variation of the doughnut hole, as well as simulating the effect of various coinsurance rates for coverage above it, would be warranted. Risk-adjusted subsidies tied to health status might also be considered if health-related selection were found to be a problem.

Although our study does not consider possible behavioral responses, such as workers moving from jobs with insurance offers to those without offers as a consequence of the subsidy, it does indicate that a subsidy for HSAs can be simulated and that the effects of a subsidy might be important for expanding health insurance coverage in the United States.

Editor's Notes

 
The authors are all affiliated with the University of Minnesota in Minneapolis. Roger Feldman (feldm002{at}umn.edu) is the Blue Cross Professor of Health Insurance in the School of Public Health. Stephen Parente is an assistant professor in the Carlson School of Management. Jean Abraham is an assistant professor in the School of Public Health. Jon Christianson is the James A. Hamilton Chair in Health Policy and Management, School of Public Health. Ruth Taylor is associate director of the Medical Industry Leadership Institute, Carlson School.

This research was supported by Department of Health and Human Services (DHHS) Contract no. HHSP233200400573P: Analytic Support in Assessing the Impact of Health Savings Accounts on Insurance Coverage and Costs. The views expressed herein do not represent official DHHS policy.

NOTES

Top Data And Analytic Approach Study Results Conclusions NOTES

 

1. J.B. Christianson, S.T. Parente, and R. Taylor, "Defined-Contribution Health Insurance Products: Development and Prospects," Health Affairs 21, no. 1 (2002): 49–64.[Abstract/Free Full Text]
2. Mercer Human Resources Consulting, "U.S. Health Benefit Cost Rises 7.5 Percent in 2004, Lowest Increase in Five Years," Press Release, 22 November 2004.
3. S.T. Parente, R. Feldman, and J.B. Christianson, "Evaluation of the Effect of a Consumer-Driven Health Plan on Medical Care Expenditures and Utilization," Health Services Research 39, no. 4, Part 2 (2004): 1189–1210.[CrossRef][Web of Science][Medline]
4. R. Kazel, "Blues Pledge Nationwide Expansion of HSAs," American Medical News, 13 December 2004.
5. U.S. Department of the Treasury, General Explanations of the Administration’s Fiscal Year 2005 Revenue Proposals (Washington: U.S. Treasury, February 2004).
6. W.S. Custer, C.N. Kahn III, and T.F. Wildsmith IV, "Why We Should Keep the Employment-based Health Insurance System," Health Affairs 18, no. 6 (1999): 115–123.[Medline]
7. M. Pauly, A. Percy, and B. Herring, "Individual versus Job-based Health Insurance: Weighing the Pros and Cons," Health Affairs 18, no. 6 (1999): 28–44.[Abstract]
8. One employer is located primarily in Minneapolis; the other employers’ establishments are dispersed widely throughout the United States.
9. T.D. Musco and T.F. Wildsmith, "Individual Health Insurance: Access and Affordability," Brief Analysis (Washington: Health Insurance Association of America, October 2002); and T.F. Wildsmith, Individual Health Insurance: Wide Choice of Benefits Available (Washington: America’s Health Insurance Plans, 2004).
10. S.T. Parente, R. Feldman, and J.B. Christianson, "Employee Choice of Consumer-Driven Health Insurance in a Multiplan, Multiproduct Setting," Health Services Research 39, no. 4, Part 2 (2004): 1091–1112.[CrossRef][Web of Science][Medline]
11. We wanted to include employee (or family) health status in the choice model, but, unfortunately, we did not have these data. However, in our previous work (see Note 10), we found that health status was not a significant predictor of enrollment in a consumer-directed plan. Therefore, we do not believe that this is a serious omission.
12. Because state identification codes are not available in the MEPS-HC data, we could not adjust the out-of-pocket premium by the employee’s state income tax rate if applicable.
13. Mercer, "U.S. Health Benefit Cost."
14. The most likely bias from this approach is that we will underestimate the attractiveness of an HSA, which features employee "ownership" of the account.
15. Additional details about the HSA policies used in the simulations are available from the authors. Contact Roger Feldman, feldm002{at}umn.edu.
16. We do not attempt to predict the ultimate insurance choices of workers who turn down an employer offer but do not buy an individual HSA. Some will obtain insurance through a spouse’s policy, but others will remain uninsured.
17. Under the administration’s proposal, the tax credit would apply only to the insurance premium, not to the HSA account. However, MMA lets consumers fund their accounts with pretax dollars, and this subsidy would stay in place.
18. These subsidies applied both to single people with no dependents and to families. We did not attempt to "spline" the subsidies as in the administration’s proposal.
19. The $211 billion for the full subsidy of a generous HSA was derived by adding the subsidy in the individual market ($98.7 billion) to the subsidy for those who turn down employer coverage ($122.4 billion).
20. The subsidy claimed by those already insured could be viewed as a reward for "doing the right thing."

                      What's this?

This article has been cited by other articles:

				J. Barabas Not the Next IRA: How Health Savings Accounts Shape Public Opinion Journal of Health Politics Policy and Law, April 1, 2009; 34(2): 181 - 217. [Abstract] [PDF]

				K. Baicker, W. H. Dow, and J. Wolfson Lowering The Barriers To Consumer-Directed Health Care: Responding To Concerns Health Aff., September 1, 2007; 26(5): 1328 - 1332. [Abstract] [Full Text] [PDF]

				D. Blumenthal Employer-sponsored insurance--riding the health care tiger. N. Engl. J. Med., July 13, 2006; 355(2): 195 - 202. [Full Text] [PDF]