September 13, 2004


by Edwin Park and Robert Greenstein

PDF of analysis

View Related Reports

If you cannot access the files through the links, right-click on the underlined text, click "Save Link As," download to your directory, and open the document in Adobe Acrobat Reader.

Debate continues over Health Savings Accounts.  Many leading health care analysts and economists have warned that HSAs pose a high risk of causing “adverse selection,” under which healthy people and less-healthy people separate into different insurance arrangements and the cost of insurance for the less-healthy consequently rises, which can place them at risk of becoming uninsured or underinsured.  Past studies by the Urban Institute, the American Academy of Actuaries, and RAND concluded that accounts like HSAs would have these effects if use of the accounts became widespread.[1]  Analyses by health and tax policy analysts also have concluded that HSAs are likely to be used extensively as tax shelters by high-income individuals.

HSA proponents have long dismissed these warnings and criticisms, and they recently have begun citing what they say are new data that refute these critiques.[2]  This analysis considers the new data.  As it demonstrates, careful examination of the new data shows that they do not support the claims of HSA proponents and shed little light on the issues being debated.  The use of these data to claim that concerns and criticisms about HSAs are unfounded does not withstand scrutiny.

Background on Health Savings Accounts

Health Savings Accounts were established as part of last year's Medicare drug legislation and made available as of January 1, 2004.  Any individual who enrolls in a high-deductible health insurance plan with a deductible of at least $1,000 for individual coverage and $2,000 for family coverage may establish a tax-favored savings account known as a Health Savings Account.[3]  An individual with a HSA may take a tax deduction for contributions he or she makes to the account (up to the amount of the deductible contained in his or her high-deductible insurance policy), as long as the contributions do not exceed an annual limit, set at $2,600 for individuals and $5,150 for family coverage in 2004.[4]  Both employers and employees may make deductible contributions to an employee’s HSA in the same year; the combined contributions made on behalf of an individual may not exceed the plan deductible or the contribution limits, whichever is lower.

Funds held in these accounts may be placed in various investment vehicles such as stocks and bonds, with the earnings accruing on a tax-free basis.  Withdrawals from the account also are exempt from tax as long as they are used to pay for out-of-pocket medical costs such as deductibles, co-payments, and other uncovered medical expenses.  Withdrawals for non-medical purposes are subject to income tax and a financial penalty, but no penalty applies to withdrawals for non-medical purposes made after an individual reaches age 65.[5]

Due to the structure of HSAs, they hold particular attractions for healthier and more affluent people.  For healthy individuals who do not expect to incur significant health care costs, HSAs provide a way to build up a new stream of tax-favored savings.  To the extent that funds in HSAs are not needed for health care costs, account-holders can build up account balances that accumulate over time and enjoy tax advantages that regular savings accounts do not have.  These tax advantages are worth the most to people at higher-income levels; the higher your tax bracket, the greater the benefit that the HSA tax breaks provide you.

HSAs do not provide the same benefits for less healthy individuals.  For such people, who tend to consume more health care, the high-deductible insurance policies that must be used in conjunction with HSAs can mean significantly greater out-of-pocket costs, as compared to the out-of-pocket costs typically borne under comprehensive insurance, which usually carries significantly lower deductibles.  These added out-of-pocket costs are of greatest concern for those less-healthy individuals who are not in the higher tax brackets, since they have fewer resources to draw upon and also would derive much less benefit from the HSA tax breaks.


Health Savings Accounts Raise Substantial Concerns

HSAs are a controversial element of last year's Medicare prescription drug legislation.  They raise two principal concerns:  that they are likely to weaken the comprehensive employer-based health insurance system through which the vast majority of Americans now obtain their health insurance; and that they will be used primarily as tax shelters by healthy, affluent individuals. 

  • Effects on employer-based coverage.  Under employer-based coverage, healthier and sicker employees are combined into a single insurance pool.  This enables less healthy individuals to obtain insurance at an affordable price.  If each individual had to purchase insurance individually based on his or her own health status, older, sicker workers would in many cases be priced out of the market.

    A major concern about HSAs is that if employers begin offering HSAs and high-deductible insurance as an option alongside traditional comprehensive insurance, then healthy and less-healthy workers may separate into different insurance arrangements, with the healthier workers shifting to HSAs and high-deductible policies and workers in poorer health seeking to remain in comprehensive coverage.  Numerous health policy experts believe this development is likely under HSAs.  Such a development would be highly problematic.  The cost of insuring any group of workers — and hence the price of insurance coverage for those workers — depends on the health status of the people in the group.  If the healthier, less-costly-to-insure employees opt out of comprehensive coverage to take advantage of the HSA tax breaks, then the average cost of insuring the people remaining in comprehensive coverage must go up, since those left in comprehensive coverage will be a less-healthy group that tends to use more health care services.

    The withdrawal of healthier workers from comprehensive employer-based coverage to take advantage of HSAs also could occur even if an employer does not offer HSAs and high-deductible insurance as an option.  Some affluent, healthier workers may conclude they would do better purchasing a high-deductible policy in the individual insurance market and setting up a HSA than remaining in employer-based coverage (especially if the employer-based coverage requires employees to bear a significant share of the premiums).

    If HSAs lead significant numbers of healthier workers to opt out of comprehensive employer-based coverage, making those who remain in such coverage more expensive, on average, to insure, then the comprehensive coverage that employers typically offer will become less affordable over time — and a growing number of employers may ultimately cease to provide it.  That would pose a particular problem for vulnerable workers in poorer health who need such coverage and seek to remain in it.
  • Creation of a lucrative new tax shelter.  The second concern stems from the fact that under HSAs, not only are contributions to the accounts tax-deductible, but withdrawals from the accounts to pay for out-of-pocket medical costs are tax-free.  This tax treatment — under which both contributions to a savings account and withdrawals from that account are tax advantaged — is without precedent in the tax code.  Retirement accounts such as traditional Individual Retirement Accounts (IRAs) and 401(k) plans permit deductible contributions, but withdrawals upon retirement are treated as taxable income.  Other plans, such as Roth IRAs, permit tax-free withdrawals but the contributions are not tax-deductible.

    Furthermore, unlike under traditional IRAs, there are no income limits on participation in HSAs.  As a result, affluent healthy individuals who have reached the maximum annual contribution limits on their IRA or 401(k) plans — or who are ineligible to make tax-deductible contributions to IRAs because their incomes exceed the IRA income limits — could use HSAs to shelter a greater share of their income for retirement.  HSAs consequently are likely to become a major tax shelter for affluent individuals, causing substantial revenue losses to the Treasury and adding to budget deficits.

    Of added concern, the exceptionally generous tax treatment that HSAs enjoy creates a dangerous precedent.  If this type of tax treatment, under which contributions to an account are deductible and withdrawals are tax free, is extended in whole or in part to other savings accounts — such as retirement accounts, as some Congressional leaders already are proposing — the adverse long-term fiscal consequences for the nation may be severe.[7]  For example, a proposal to convert a portion of 401(k) and IRA accounts into HSA-like accounts, which has been designed by Fidelity Investments and endorsed by Senate Majority Leader Bill Frist, would be likely to cost the Treasury hundreds of billions of dollars over coming decades (and, depending on how large a portion of retirement accounts were allowed to be converted to HSA-like accounts, could cost $1 trillion or more).


Analysis of the Data Cited by HSA Proponents

HSA proponents have claimed these concerns and criticisms are incorrect or overblown.  Recently, they have begun to cite data released earlier this year by eHealthInsurance and Assurant Health as evidence that concerns about HSAs are unfounded.[8]  eHealthInsurance is an online individual market health insurance broker that sells HSAs.  Assurant Health is an insurer specializing in offering HSAs to both employers and individuals.  Since both firms stand to make profits from expansion of HSAs, evidence from such sources should be subject to careful scrutiny.  As it happens, examination of these data shows they do not support these claims.

What HSA Proponents Say

HSA proponents contend these data demonstrate that concerns about adverse selection are unfounded.  They say the data refute concerns that HSAs will be used primarily by healthier or more affluent individuals.  They cite data from Assurant Health indicating that the company made an offer of coverage to 93 percent of those who applied for a high-deductible policy in conjunction with a HSA.  They say this shows that this coverage is available to nearly all who seek it, not just to those who are healthier.  They also cite Assurant data that 29 percent of those purchasing HSA-related coverage from Assurant had income below $50,000 as evidence there is not a tilt in favor of higher-income people.  And they cite data that 43 percent of HSA applicants lacked health insurance coverage in the months before purchasing a high-deductible policy in conjunction with a HSA as proof that HSAs can play a large role in reducing the ranks of the uninsured.

Weaknesses in These Data

The eHealthInsurance and Assurant Health data cited in support of these claims relate only to use of HSAs in the individual health insurance market; they include no data regarding HSA use in conjunction with employer-based health insurance.  Yet it is in their potential effects on employer-based coverage that the greatest dangers of HSAs, and the greatest risks of adverse selection, lie.  As a result, these data are not especially relevant to such concerns about HSAs.

Moreover, the data are highly preliminary.  The eHealthInsurance data cover only the first two months of 2004.  The Assurant data were collected only for the first four months of 2004.

Furthermore, eHealthInsurance is an online provider; its data may be affected by the characteristics of individuals who are using the Internet to apply for and purchase HSAs in the individual market.  In addition, it is unclear how many HSAs were purchased through eHealthInsurance and Assurant and served as the data set for these findings.  No information as to the number of HSAs has been provided by eHealthInsurance.  Assurant notes there were 56,396 applicants to Assurant for individual market HSA coverage during the four-month period the data cover, but the number of individuals who actually obtained coverage from Assurant is not provided.

In short, the limited data from these two sources, which apply only to the individual health insurance market, do not provide a rational basis for dismissing concerns about HSAs that stem from years of analysis conducted by leading institutions with no financial interest in these matters, such as the Urban Institute, the American Academy of Actuaries, and RAND.  We now proceed to examine the specific claims that HSA proponents have made with these data.

Unsupported Conclusions Drawn by HSA Advocates

HSA proponents have drawn three questionable conclusions from these data.

1.  HSA supporters have claimed that the data show HSAs are not primarily attracting healthy individuals and thus do not risk adverse selection.[9]  Nothing in the data supports this conclusion.  The eHealthInsurance data on HSA purchasers include no information about the health status of the purchasers.  The data include information only about the age, family size, premiums, and plan benefits of an unknown number of online HSA purchasers. 

The Assurant Health data similarly lack information about the health status of HSA purchasers.  HSA proponents have noted that Assurant Health provided an offer of coverage to 93 percent of the individuals who applied for HSA coverage,[10] but this factoid sheds little light on the health status of HSA users.  Given that HSA accounts and high-deductible insurance policies are considerably more attractive to healthier individuals than to less healthy people and pose risks for those in poorer health, it is likely that those who applied for HSA coverage were a healthier-than-average group.

Furthermore, insurers in the individual market can and do deny coverage to applicants, based on an applicant's health or medical historyThey also can offer coverage to a less-healthy applicant that carries higher premiums charges and/or excludes coverage for certain important medical conditions or health benefits.  The eHealthInsurance data and Assurant data include no information about premium or benefit variation among the offers of coverage that were made.  This further invalidates the drawing of inferences about the health status of those using HSAs from the limited data that have been provided. 

Moreover, some applicants who were in poorer health may have received an offer of coverage but found the premium they were quoted to be unaffordable and decided not to purchase health insurance.  The data made available lack information on this matter, as well.

Nor do the eHealthInsurance or the Assurant Health data offer any comparison between the average health status of HSA purchasers in the individual market and the average worker in employer-based health insurance.  The individual market is accessible primarily to healthier individuals, due to the widespread use in that market of medical underwriting, under which insurers can decline to offer coverage, offer more limited coverage, or charge much higher premiums to less healthy people.  As a consequence, it is likely that most people who have actually purchased a high-deductible policy in the individual market in conjunction with a HSA are people who are in better health, on average, than people who have employer-based coverage. To help assess the degree to which HSAs may result in adverse selection, it would be useful to have data on the health status of workers with employer-based coverage who have chosen a HSA and how their health status compares to that of workers in the same firm who have chosen comprehensive coverage.  No such data have been provided by eHealthInsurance or Assurant Health.

2.  HSA supporters have claimed these enrollment data show that HSA participants are not primarily higher-income individuals taking advantage of the tax shelter benefits of HSAs.  Here, as well, the data do not back up the claim.  The Assurant Health data indicate that 29 percent of individuals who purchased individual market coverage from Assurant in conjunction with a HSA had incomes below $50,000 per year.[11]  This means that 71 percent of HSA purchasers had incomes of more than $50,000.  The Assurant data do not provide a more detailed income breakdown of the purchasers with incomes above $50,000 to determine the extent to which high-income individuals purchased high-deductible plans and HSAs through Assurant.

Moreover, the Assurant Health data indicate that nearly 57 percent of purchasers were from professional or managerial occupations.[12]   People in those occupations tend to have higher salaries.  As a result, the Assurant data themselves seem to suggest that higher-income individuals may be more likely to purchase HSAs than those with low- or moderate-incomes.[13]

The eHealthInsurance data that have been made publicly available do not include any data on income.  One HSA proponent has cited eHealthInsurance as finding that 46 percent of HSA participants had family incomes of less than $50,000.[14]  Even if these data (which are not publicly available) are accurate, they provide little basis for drawing conclusions about HSAs.  It is important to recognize that data about HSAs that come solely from the individual health insurance market are likely to skew the income of the purchasers downward.  Lower-income workers tend to use the individual market in great proportions than higher-income workers due to their greater lack of access to employer-based coverage.  Lower-income workers often work for smaller businesses; such firms, particularly those with large numbers of low-wage employees, are among the least likely to offer health insurance coverage to their workers.  As a result, in 2003 — before HSAs came into existence — insured households with incomes below $25,000 were nearly 80 percent more likely to obtain their coverage through the individual market than households with incomes of $75,000 or more.[15]  Since it makes sense for people who already were purchasing high-deductible coverage in the individual market to set up HSAs, a sample of HSA purchasers in the individual market would likely be biased downward in terms of income.  Such data cannot be used to make inferences about the income of HSA participants generally.

A much more useful examination would look at individuals participating in HSAs in the employer-based health insurance system and compare their incomes to the incomes of individuals enrolled in comprehensive health insurance plans offered by the same employers.  Data on the income of employees who choose a HSA plan as compared to workers within the same firm who opt for a comprehensive plan would be particularly relevant.  Such data are not currently available since the large majority of employers do not yet offer HSAs.  Without such idea, conclusions on the matter are premature.

3.   HSA proponents also have claimed that the data show a large share of HSA users are people who previously were uninsured and thus that HSAs can be an important tool for expanding coverage.  According to the Assurant data, 43 percent of HSA applicants did not have health coverage in the months before purchasing a high-deductible policy in conjunction with a HSA.  This statistic is cited to show that HSAs can play an important role in making coverage more affordable for the uninsured.

These Assurant data, however, require considerable qualification.  The individual market is often a market of last resort, particularly for adults who have lost their jobs and health insurance, cannot afford COBRA coverage, and do not qualify for Medicaid.  Even before the advent of HSAs, many individuals purchasing insurance in the individual market are likely to have been uninsured for a period immediately preceding the purchase.  The fact that a certain percentage of people who purchased individual-market coverage in conjunction with a HSA were uninsured for the months before the purchase is not especially meaningful in assessing the contribution of HSAs.  Moreover, no data have been provided on the percentage of actual HSA purchasers — as distinguished from HSA applicants — who previously were uninsured.  Some uninsured applicants may have declined to purchase coverage once they were provided an offer of coverage if the offer carried a premium cost they considered unaffordable.  (It is curious that data on the extent to which purchasers previously were uninsured have not been made available along with data on applicants.)

Most important, these data are for the individual market only.  HSA use will become widespread only if HSAs are adopted by large numbers of employers, which many analysts now expect to occur.  Since the vast majority of employers who adopt HSAs are likely to be employers that already offer coverage to their workers, most employer-based HSA enrollment will involve workers who already are insured and are shifting their health insurance arrangements (or having the arrangements shifted by their employers) from comprehensive coverage to high-deductible plans attached to HSAs.  Over time, the vast majority of HSA participants thus is likely to consist of people who previously were insured.[16]


Some HSA proponents have claimed that recent data on HSA use in the individual market refute concerns that HSAs may weaken employer-based coverage through adverse selection and that HSAs may be used extensively as tax shelters by higher-income individuals.  These proponents also argue that the data show HSAs are effective in covering the uninsured.  In reality, the data do not support any of these conclusions.

HSAs are likely to become increasingly common in the employer-based health insurance system.  The Kaiser Family Foundation and the Health Research and Educational Trust determined 27 percent of employers were somewhat likely or very likely to offer a high-deductible health insurance plan attached to a HSA or other personal savings account in the next two years.  Among very large firms with 5,000 or more workers, 50 percent were somewhat likely or very likely to offer such plans.[17]  Similarly, a survey of 991 primarily large employers recently conducted by Mercer Human Resource Consulting found that nearly three-quarters of employers (73 percent) are somewhat likely or likely to offer Health Savings Accounts by 2006.[18]  A smaller employer survey conducted by Hewitt Associates found that 61 percent of large employers are likely to offer HSAs in the near future.[19]  As more employers adopt HSAs over time, more data will become available to evaluate the risks that HSAs pose.  These data will provide much better evidence than the preliminary, fragmentary, conflicting, and incomplete data from eHealthInsurance and Assurant Health.

It is likely that when better data become available, they will confirm the risks that health and tax policy experts believe HSAs pose.  Indeed, the recent Mercer survey of employers heightens these concerns.  The survey found that employers believe HSAs will be most attractive to healthy, higher-income workers.  A plurality of employers surveyed (44 percent) reported they believed their healthiest employees would be most likely to participate in HSAs.  A substantial majority of employers (61 percent) said they believed their higher-paid employees would be most likely to use HSAs.[20]

The Center on Budget and Policy Priorities is grateful to the Nathan Cummings Foundation for its support of this report.

End Notes:

[1] See Emmett B. Keeler, et. al., "Can Medical Savings Accounts for the Nonelderly Reduce Health Care Costs?" Journal of the American Medical Association, June 5, 1996, p. 1666-71; Len M. Nichols, et. al., "Tax-Preferred Medical Savings Accounts and Catastrophic Health Insurance Plans: A Numerical Analysis of Winners and Losers," Urban Institute, April 1996; and American Academy of Actuaries, "Medical Savings Accounts: Cost Implications and Design Issues," May 1995.

[2] See, for example, Laura Trueman, "Health Savings Accounts: Myth vs. Fact," National Center for Policy Analysis, July 19, 2004; Grace Marie Turner, "Health Savings Accounts Gain Popularity," Galen Institute, July 26, 2004; Richard Nadler and Dan Perrin, "The Center on Budget and Policy Priorities' Study on HSA Premium Tax Deduction Misses the Point," The HSA Coalition, May 25, 2004; and Derek Hunter, "New Data on Health Insurance, the Working Poor, and the Benefits of Health Care Tax Changes," Heritage Foundation, April 28, 2004.

[3] The high-deductible health insurance plan must have an out-of-pocket limit of no more than $5,000 for individuals and $10,000 for family coverage.  The out-of-pocket limit may be higher for out-of-network services.  Certain preventive services such as annual physicals and routine screenings may be exempted from the deductible.

[4] Individuals age 55 or older may make additional contributions (in excess of the limit) of another $500 in tax year 2004, rising to $1,000 by tax year 2009.  Individuals age 65 or older who are eligible and participating in Medicare are not eligible to make deductible contributions to HSAs.

[5] The financial penalty for a non-medical withdrawal prior to retirement age of 65 is equal to 10 percent.  Unlike other retirement accounts, there are no mandatory withdrawals upon retirement.

[6] See Emmett B. Keeler, et. al., "Can Medical Savings Accounts for the Nonelderly Reduce Health Care Costs?" Journal of the American Medical Association, June 5, 1996, p. 1666-71; Len M. Nichols, et. al., "Tax-Preferred Medical Savings Accounts and Catastrophic Health Insurance Plans: A Numerical Analysis of Winners and Losers," Urban Institute, April 1996; American Academy of Actuaries, "Medical Savings Accounts: Cost Implications and Design Issues," May 1995; Daniel Zabinski et. al., "Medical Savings Accounts: Microsimulation Results from a Model with Adverse Selection," Journal of Health Economics, April 1999, p.195-218; Gail Shearer, "The Health Care Divide: Unfair Financial Burdens," Consumers Union, August 10, 2000 (relying on Lewin Group estimates).

[7] See, for example, Edwin Park and Robert Greenstein, "New Retirement Medical Account Proposal Would Create Lucrative Tax Shelter and Swell Deficits but Do Little to Help Low- and Moderate-Income Seniors," Center on Budget and Policy Priorities, revised July 22, 2004.

[8] eHealthInsurance, "More than 70 Percent of Consumers Obtaining Health Savings Accounts Paid $100 or Less for their HSA-Eligible Health Plans," April 21, 2004; Assurant Health, "U.S. House Speaker Hastert Joins Business Leaders to Review Latest Data on Success of Health Savings Accounts," June 21, 2004.  Some of the data attributed to both eHealthInsurance and Assurant Health by HSA proponents are not included in the publicly available eHealthInsurance report or in the Assurant Health press release.  That makes such data difficult to evaluate.  It is assumed here that data that have been attributed to eHealthInsurance and Assurant Health but are not publicly available have been cited accurately.

[9] Trueman, p.3; Nadler and Perrin, p.15; Turner.

[10] Trueman, p.2; Coalition for Affordable Health Coverage, "Myth vs. Fact about HSAs," available at

[11] Assurant Health.

[12] Assurant Health.

[13] One HSA proponent also cites Assurant Health data as showing that 38 percent of HSA purchasers had homes with a market value of less than $125,000 and that 27 percent had a net worth of less than $25,000.  Trueman, p.1. This implies that 62 percent of HSA purchasers had home with market values in excess of $125,000 and 73 percent had a net worth in excess of $25,000.  Without a further breakdown of these data, which have not been made publicly available, one cannot infer the extent to which higher-income people are purchasing HSAs.

[14] Trueman, p.1.  This statistic was not included in the publicly available eHealthInsurance report.

[15] CBPP analysis of 2003 CPS data.

[16] The publicly available data from eHealthInsurance include no information about previous insurance status.  Two HSA proponents claim that eHealthInsurance found 32.8 percent of applicants did not have health insurance coverage in the prior six months (Nadler and Perrin, p.2; Trueman, p.1).  Assuming these data attributed to eHealthInsurance are accurate, qualifications similar to those just cited for the Assurant data are needed.  First, this statistic implies that 67.2 percent of applicants already had coverage.  Most important, these data apply only to a limited number of individuals who secured coverage in the individual market; the data do not involve the employer-based market, in which most HSA participants will, as just noted, consist of already-insured individuals who shift insurance arrangements.  In addition, these data apply to applicants for insurance, not to those who actually purchased coverage.

[17] Kaiser Family Foundation and Health Research and Educational Trust, “Employer Health Benefits: 2004 Annual Survey,” September 2004.

[18] Mercer Human Resource Consulting, "US Employers See a Role for New Health Savings Accounts in their Benefit Programs," April 26, 2004.

[19] Hewitt Associates, "Addition of HSAs Will Require Substantial Health Plan Design Changes," March 31, 2004.

[20] Mercer Human Resource Consulting, op cit.