Health Savings Accounts and High-Deductible Health Plans: A Data Primer

Health Savings Accounts and
High-Deductible Health Plans: A Data Primer
Carol Rapaport
Analyst in Health Care Financing
Domestic Social Policy Division
Summary
Individuals began establishing health savings accounts (HSAs) in 2004. These
savings accounts are generally used to pay for unreimbursed medical expenses on a tax-
advantaged basis. Any unspent money accrues to the individual. To open an HSA, the
individual must enroll in a qualifying high-deductible health plan (HDHP). HSAs are
tax-advantaged and provide some incentives for people to monitor, and perhaps reduce,
their expenditures on health care.
Data covering enrollment and/or cost sharing during the first few years of HDHPs
and their associated HSAs are now available from at least six separate sources. Two
sources provide data on HSAs, two sources provide data on HSAs and Health
Reimbursement Accounts (HRAs) combined, and two sources provide data on HSA-
eligible HDHPs. Before analysts can evaluate the effects of HSAs, they must decide
which data source(s) to use. This primer provides basic guidance in that direction. The
primer also provides the most recent data available from each source on enrollment,
premiums and deductibes for HSAs, HSAs and HRAs combined, and HDHPs.
Individuals were first able to establish health savings accounts (HSAs) in 2004.
These accounts allow people to pay for out-of-pocket medical expenses on a tax-
advantaged basis. Individuals must have a qualifying high-deductible health plan (HDHP)
to establish an HSA. After establishing an HSA, individuals (or employers) can¹
contribute money to the account up to an annual maximum.
Although commonly discussed in combination, HSAs should not be confused with
Health Reimbursement Accounts (HRAs). Although HRAs are also used to pay for
unreimbursed medical expenses on a tax-advantaged basis, only employers may establish

¹ For self-only coverage, the annual deductible in 2008 for an HDHP must be at least $1,100
(with the plan’s annual out-of-pocket limit not exceeding $5,600). The annual HSA contribution
limit in 2008 for individuals with self-only coverage is $2,900. An explanation of the rules
governing HSAs can be found in CRS Report RL33257, Health Savings Accounts: Overview of
Rules for 2008, by Bob Lyke.

and contribute to an HRA. In addition, employees usually forfeit any remaining HRA
funds at the termination of employment.²
Data covering enrollment and/or cost sharing during the first few years of HDHPs
and their associated HSAs are now available from at least six separate sources. Two
sources provide data on HSAs, two sources provide data on HSAs and HRAs combined,
and two sources provide data on HSA-eligible HDHPs. Before analysts can evaluate the
effects of HSAs, they must decide which data source(s) to use. This primer provides basic
guidance in that direction. The primer also provides the most recent data available from
each source on enrollment, premiums and deductibles for HSAs, HSAs and HRAs
combined, and HDHPs.
Data Sources
Table 1 identifies the six data sources with data on HSAs, HSAs and HRAs
combined, and HSA-qualified HDHPs. The various data sources include a survey of large
firms, a survey of individuals, data on all policies reported to an association, data from
those who purchase individual policies online, and a sample of IRS tax returns. The data
sources are listed in alphabetical order.
Which data source to use depends primarily on the question being asked. If the
policy question truly requires information on HSAs — that is, the actual accounts rather
than the associated HDHPs — then only the IRS and Kaiser/HRET sources are suitable.³
The IRS data, which are broken down by tax reporting units, provide the total number of
tax deductions taken and the aggregate value of the deductions. The Kaiser/HRET data
include the number of working adults with HSAs, premiums and cost-sharing features of
the insurance plans, and various characteristics of the employer. Two disadvantages of
the IRS data are a total lack of information on the associated HDHPs and that the data are
released well after the other data sources.
Two sources combine data on HSAs and HRAs. These data can be used if separate
analyses of HSAs or HRAs are not necessary. EBRI provides enrollment estimates for
privately insured individuals aged 21 to 64, while Mercer provides enrollment estimates
for adults working in firms with at least 10 employees. The EBRI data are based on a
survey of individuals and contain information on the workers’ ages, incomes, health
status, and opinions of their health plan options. The Mercer survey is of firms and
contains information on firm size and whether the firm predicts it will offer an HSA or
HRA in the coming year. Choosing between these two data sources comes down to a
choice between an individual-level analysis (EBRI) or a firm-level analysis (Mercer).
Finally, two additional data sources provide information on HSA-qualified HDHPs.
The AHIP data are obtained from insurance plans and measure all covered lives in the

² For additional information on the differences between HRAs and HSAs, see CRS Report
RS21573, Tax-Advantaged Accounts for Health Care Expenses: Side-by-Side Comparison, by
Bob Lyke and Chris L. Peterson.
³ Approximately one-third of the insurers that provided HDHP information to AHIP also reported
information on enrollees’ HSA accounts. However, because this information was provided on
a voluntary basis, the sample is non-random and not necessarily representative.

plans. Both individual and group plans are analyzed. The data form virtually a census
of such policies among AHIP member companies. Thus, the AHIP data are based on a
large number of enrollees in high-deductible health plans. Along with the average
premiums and deductibles, information on enrollees’ age and state of residence is also
available. Compared with the AHIP data, the eHealthInsurance data are less
comprehensive because only individual policies purchased through the company’s website
are included. However, like the AHIP data, information on premiums and deductibles is
available. Of course, individuals who purchase insurance online through
eHealthInsurance may differ greatly from individuals who purchase insurance from
agents.
Enrollment
Table 2 presents the most recent available data on enrollment, premiums, and
deductibles for the six sources. Four of the sources contain data on enrollment. As
shown in Table 2, the enrollment estimates differ greatly. These differences occur
because each source measures a unique concept. Kaiser/HRET estimates that 1,900,000
working adults (in firms with at least three employees) were enrolled in an HSA in 2007.
The IRS data do not measure enrollment but report that 215,781 returns claimed an HSA
deduction in 2005. These populations, enrollment definitions, and years are too dissimilar
to provide meaningful comparisons. EBRI reports that 2,300,000 individuals between 21
and 64 were enrolled in either an HSA or HRA in 2006. Mercer reports that 5% of all
covered employees (in firms with at least 10 employees) have either an HSA or HRA, also
in 2006.
Although none of these numbers is directly comparable, it is reassuring that the
number of HSAs from Kaiser/HRET is smaller than the number of HSAs plus HRAs from
EBRI. The number of individuals who claim deductions for HSA contributions in 2005
is the smallest of all, as would be expected for two reasons: (1) the number of HSAs has
been growing over the 2005 to 2007 period, and (2) not all individuals contribute money
to the HSA — and of those who do, not all claim an HSA deduction. The AHIP
enrollment numbers are the largest because they show enrollment in an HSA-qualified
HDHP (regardless of whether an HSA account was actually established) from group
coverage as well as individual coverage.
Premiums and Deductibles
AHIP provides the most complete information on premiums and deductibles; the
average values are available for the small group and large group markets, and for three
age groups in the individual market. No other data source provides breakdowns for more
than one of these markets. In all cases, values for individual (and not family) insurance
plans are reported.
In general, individuals in small group markets are more costly to insure because the
risk of major illness is spread across fewer individuals and because there are fewer
economies of scale. Small group market deductibles should therefore be higher than large
group market deductibles, assuming benefits and other policy characteristics are
comparable across group size. The AHIP data display the expected pattern for HSA-
eligible HDHPs: The average deductible for small group policies is $2,244, and the

average deducible for a large group policy is $2,046. On the other hand, the premium
values are virtually identical between groups.
The data from eHealthInsurance provides information only from individual policies
sold through its website. The eHealthInsurance results show higher average deductibles
and lower premiums than the comparable AHIP data. However, this may reflect the
characteristics of individuals enrolling in health insurance through a website like
eHealthInsurance. In addition, the AHIP premium and deductible information were based
on reports of each insurer’s best-selling HDHP products, which may not equate to the
average of all HDHP premiums and deductibles.
Conclusion
HSAs have been available since 2004, and at least six data sources can be used to
uncover some basic facts about the recent experience. Nevertheless, the data sources
differ in the insurance markets analyzed; whether the information covers HSAs, HSAs
and HRAs combined, or HSA-eligible HDHPs; and whether the information is provided
by employers, insurance companies, or individuals. Information from different sources
therefore should be combined with extreme care. A more fruitful strategy would be to
decide on a specific question and use only the source which best answers that question.

CRS-5
Table 1. Characteristics of Data Covering HSAs, HSAs and HRAs Combined,
and HSA-Eligible HDHPs
^K^a^{iser Fa}^m^ily
^Am^erica’s^Health^E^m^{ployee B}^e^{nefit Res}^e^arch^F^o^unda^tⁱ^oⁿ^/H^ea^lt^{h Resea}^rch
^Iⁿ^s^u^{rance Plans}^Inst^it^ut^e⁽^E^B^R^I)^/^Iⁿ^tern^al^R^e^ven^u^e^a^{nd Educa}^t^io^{n Trust}
⁽^A^H^I^P⁾^eH^ea^lt^hInsuraⁿ^ce^Co^m^m^o^nw^ea^lt^{h F}^und^Service^(Kais^er/HRE^T⁾^M^ercer
ⁱ^p^tioⁿ^o^f^o^r^g^aⁿⁱ^zatioⁿ^a^sso^ciatioⁿ^o^f^h^ealth^s^e^l^l^e^{r of}ⁱⁿ^di^vⁱ^du^alⁿ^oⁿ^p^rof^{it res}^earch^fe^d^e^r^a^l^a^g^eⁿ^c^y^no^np^r^o^fi^t^fo^und^a^tⁱ^o^n/^no^np^r^o^fi^t^hum^aⁿ^-
ⁱⁿ^s^u^ran^{ce f}ⁱ^rm^sⁱⁿ^su^r^aⁿ^{ce p}^o^licies^o^r^gaⁿⁱ^z^a^tⁱ^o^n/^p^rⁱ^v^a^t^e^o^r^gaⁿⁱ^z^a^tⁱ^oⁿ^res^o^u^r^ce
^oⁿ^lin^e^fo^und^a^tⁱ^oⁿ^coⁿ^s^u^ltin^g
^fi^r^m
^r^{ce o}^f^d^a^taⁱⁿ^f^o^r^m^atioⁿ^r^e^p^o^r^ted^s^a^m^p^l^e^of^12,000^oⁿ^lⁱⁿ^e^a^nnua^l^sur^v^e^y^o^f^sam^p^{le o}^fⁿ^ear^ly^a^nnua^l^sur^v^e^y^o^f^ne^a^r^l^y²^,⁰⁰⁰^a^nnua^l^sur^v^e^y
^by^{97 A}^H^{IP m}^e^m^b^er^p^o^{licies so}^ld^th^r^o^u^g^h^{4,217 pri}^v^at^el^yⁱⁿ^s^u^red^{300,000 i}ⁿ^di^vⁱ^du^al^em^pl^oy^ers^ofⁿ^early
ⁱⁿ^s^u^ran^{ce com}^p^an^ies^com^p^an^y^w^e^bs^iteⁱⁿ^di^vⁱ^du^al^s^f^e^d^e^r^a^{l in}^co^m^e^tax^3,000
^ret^u^rn^s^em^pl^oy^ers
ⁱ^{ki/CRS-RS22877}^e^{l o}^f^d^a^taⁱⁿ^su^ran^{ce f}ⁱ^rm^sⁱⁿ^d^iv^id^u^a^{l HDHP}^p^o^{licies so}^ld^pri^v^at^el^yⁱⁿ^s^u^{red i}ⁿ^di^vⁱ^du^al^s^t^ax^reportⁱⁿ^g^uⁿⁱ^t^s^em^pl^oy^ers^(fⁱ^r^m^sⁱ^{ze of}^{3 or}^mo^r^e⁾^em^pl^oy^ers^(f^irm^sⁱ^{ze of}
^g/w^{10 or m}^o^re)
^s^.or^a^nc^e^m^a^r^k^e^t^{s c}^o^ve^r^e^dⁱ^ndⁱ^vⁱ^d^ua^l^aⁿ^d^gr^o^u^pⁱ^ndⁱ^vⁱ^d^ua^lⁿ^o^t^dⁱ^stⁱ^ngui^she^d^no^t^dⁱ^stⁱ^ngui^she^d^gr^o^u^p^g^r^o^up
^le^ak
^t^{detailed plan}^/accouⁿ^t^HDHP^(HSA^HDHP^(HSA^elig^ib^le)^H^SA^an^d^HR^A^co^m^bⁱⁿ^ed^HSA^H^SA^HSA^an^d^HR^A
^://wiki^r^m^atioⁿ^av^ailab^l^e^elig^ib^le)^com^bⁱⁿ^ed
^h^ttp^ailab^l^e
^T^o^{tal en}^r^o^llm^en^t^c^o^v^er^ed^liv^esⁿ^o^pri^v^at^el^yⁱⁿ^s^u^{red i}ⁿ^di^vⁱ^du^al^sⁿ^o^em^ploy^eesⁱⁿ^fⁱ^rm^s^w^ith^at^em^ploy^eesⁱⁿ
^report^e^{d by}^A^H^IP^ag^es^{21 t}^o⁶⁴^leas^{t 3 w}^o^rk^ers^fⁱ^r^m^{s w}^ith^at
^me^mb^e^r^p^l^aⁿ^s^leas^{t 10}
^wo^r^k^e^r^s
^A^v^er^ag^{e p}^r^em^iu^m^y^es^y^e^s^f^o^r^f^a^m^iliesⁿ^o^y^es^y^e^s
^A^v^erag^{e dedu}^ctible^y^e^s^y^es^y^e^sⁿ^o^y^es^y^e^s
^T^a^{x d}^e^d^u^c^tⁱ^o^{ns t}^a^keⁿⁿ^oⁿ^oⁿ^o^y^e^s^no^no
^A^v^erag^{e v}^a^lu^{e of}^no^no^no^y^e^sⁿ^oⁿ^o
^dedu^ctⁱ^oⁿ
^:^[^h^t^t^p^:^/^/^www.^a^hⁱ^p^r^e^s^e^a^r^c^h^.^o^r^g^/^p^d^f^s^/^2008_HSA^_C^en^s^u^s^.^pdf^{], [h}^ttp://^www.^e^h^e^a^l^t^hⁱⁿ^s^u^r^aⁿ^c^e^.^c^o^m/^c^oⁿ^t^eⁿ^t^/^R^e^p^o^r^t^N^e^w^/^2005H^S^A^F^u^l^l^Y^earR^e^port^-^05-^10-^06F^.pdf^],
^:^/^/^www.^e^b^rⁱ^.^o^r^g^/^p^d^f^/^b^rⁱ^e^f^s^p^d^f^/^E^B^R^I^_^I^B^_03-^2008.pdf^{], [h}^ttp://^www.ⁱ^r^s^.^g^o^v^/^p^u^b^/ⁱ^r^s^-^s^oⁱ^/⁰⁵ⁱⁿ^a^l^c^r^.^p^d^f^]^,^[^h^t^t^p^:^/^/^www.^k^f^f^.^o^r^g^/ⁱⁿ^s^u^r^aⁿ^c^e^/^7672/^u^p^l^o^ad/^76723.pdf^{], an}^d
^:^/^/^www.^me^r^c^e^r^.^c^o^m^/^r^e^f^e^r^eⁿ^cecon^t^en^t.j^h^tm^l?^idC^oⁿ^t^en^t=1287790].

CRS-6
Table 2. Comparisons of Enrollment, Premiums, and Deductibles Across HSA/HDHP Data Sources
^Em^plo^y^{ee B}^e^nefⁱ^t^K^a^{iser Fa}^m^ily
^Re^se^a^r^c^h^Inst^it^ut^e^F^o^unda^tⁱ^oⁿ^/H^ea^lht
^Am^erica’s^Health^(E^BRI⁾^/^Res^e^{arch and}
^Iⁿ^s^u^{rance Plans}^C^o^mmoⁿ^w^e^a^l^t^h^Iⁿ^tern^al^R^e^ven^u^e^Educa^t^io^{n Trust}
⁽^A^H^I^P⁾^eH^ea^lt^hInsuraⁿ^ce^F^und^Service^(Kais^er/HRE^T⁾^M^ercer
^e^o^f^p^l^an^HDHP^(HSA^elig^ib^le)^HDHP^(HSA^HSA^an^d^HR^A^HSA^H^SA^HSA^an^d^HR^A
^elig^ib^le)^com^bⁱⁿ^ed^com^bⁱⁿ^ed
^t^a^l^en^rol^l^m^en^t⁶^,118,000^a^2,300,000^1,900,000^5%^of^al^l^cov^e^red
^em^ploy^ees
^en^{t m}^eas^u^r^e^c^ov^{ered liv}^es^reported^priv^atelyⁱⁿ^s^u^red^em^ploy^eesⁱⁿ^fⁱ^rm^s^em^ploy^eesⁱⁿ^fⁱ^rm^s
^b^y^A^H^I^P^me^mb^e^r^p^l^aⁿ^sⁱⁿ^divⁱ^du^als^ag^es²¹^w^ith^{at least 3}^w^ith^{at least 1}⁰
^to⁶⁴^wo^r^k^e^r^s^wo^r^k^e^r^s
^m^o^s^t^recen^{t data}^Jan^u^a^ry²⁰⁰⁸²⁰⁰⁵²⁰⁰⁷^{2005 tax}^retu^rn^s²⁰⁰⁷²⁰⁰⁷
ⁱ^{ki/CRS-RS22877}^mⁱ^u^m^s
^g/w^A^v^erag^{e i}ⁿ^di^vⁱ^du^al^m^a^rk^et^{$1,319 t}^o^3,724^depen^dⁱⁿ^g^on^ag^e^$1,368
^s^.or
^le^ak^A^v^erag^{e s}^m^al^l^g^r^ou^p^$3,189
^A^v^erag^{e larg}^{e g}^r^ou^p^$3,185
^://wiki^A^v^erag^{e larg}^{e an}^{d s}^m^{all g}^r^ou^ps^$3,894^b
^h^ttp^u^ctib^les
^A^v^erag^{e in}^divⁱ^du^{al m}^a^rk^et^$2,600^c^$3,190
^A^v^erag^{e s}^m^al^l^g^r^ou^p^$2,244^c
^A^v^erag^{e larg}^{e g}^r^ou^p^$2,046^c^{$1,769 i}ⁿⁿ^e^t^w^ork
^A^v^erag^{e l}^a^rg^{e an}^{d s}^m^al^l^g^r^ou^ps^46%^uⁿ^d^{er $2,000}^$1,556
^tax^d^e^d^u^ctioⁿ^s
^N^u^m^b^{er of}^H^S^A^t^a^x^dedu^ctⁱ^oⁿ^s^215,781
^tak^eⁿ
^A^v^erag^{e v}^a^lu^{e of}^dedu^ction^$2,367
^:^[^h^t^t^p^:^/^/^www.^a^hⁱ^p^r^e^s^e^a^r^c^h^.^o^r^g^/^p^d^f^s^/^2008_HSA^_C^en^s^u^s^.^pdf^{], [h}^ttp://^www.^e^h^e^a^l^t^hⁱⁿ^s^u^r^aⁿ^c^e^.^c^o^m/^c^oⁿ^t^eⁿ^t^/^R^e^p^o^r^t^N^e^w^/^2005H^S^A^F^u^l^l^Y^earR^e^port^-^05-^10-^06F^.pdf^],
^:^/^/^www.^e^b^rⁱ^.^o^r^g^/^p^d^f^/^b^rⁱ^e^f^s^p^d^f^/^E^B^R^I^_^I^B^_03-^2008.pdf^{], [h}^ttp://^www.ⁱ^r^s^.^g^o^v^/^p^u^b^/ⁱ^r^s^-^s^oⁱ^/⁰⁵ⁱⁿ^a^l^c^r^.^p^d^f^]^,^[^h^t^t^p^:^/^/^www.^k^f^f^.^o^r^g^/ⁱⁿ^s^u^r^aⁿ^c^e^/^7672/^u^p^l^o^ad/^76723.pdf^{], an}^{d F}ⁱ^g^u^r^{e 8 at}
^:^/^/^www.^me^r^c^e^r^.^c^o^m^/^r^e^f^e^r^eⁿ^cecon^t^en^t.j^h^tm^l?^idC^oⁿ^t^en^t=1287790].
ⁿ^s^{ists o}^f¹^.⁵^m^illioⁿ^f^r^o^m^th^{e in}^dⁱ^vⁱ^d^u^{al m}^a^rk^{et, 1}^.⁸^m^illioⁿ^f^r^o^m^th^{e sm}^{all g}^r^o^u^p^m^a^rk^{et (as d}^e^fⁱⁿ^e^d^b^y^eachⁱⁿ^su^{rer), a}ⁿ^d²^.⁸^m^illioⁿ^f^r^o^m^th^{e lar}^g^{e g}^r^o^u^p^m^a^r^k^et.
ⁱ^d^es^t^h^{e “}^a^v^e^rag^e^cos^t^{per em}^pl^oy^{ee,” f}^o^{r i}ⁿ^di^vⁱ^du^al^as^w^e^l^l^as^f^a^mⁱ^l^y^cov^e^rag^e^{, of}^{$5,479, w}^hⁱ^c^hⁱ^sⁿ^o^t^com^p^ara^b^l^{e to}^th^{e p}^r^em^iu^m^f^o^rⁱⁿ^dⁱ^vⁱ^d^u^als.
^a^sed^oⁿ^eachⁱⁿ^su^rer’^{s b}^e^st-^s^ellin^g^p^r^o^d^u^ct.