health reform without side effects - hoover institution...health reform without side effects making...

Health Reformwithout Side EffectsMaking Markets Work forIndividual Health Insurance

MARK V. PAULY

Private health insurance in the United States protectsmost of the population against the risk of high med-

ical care spending, but the extent of that protection hasbeen declining for at least a decade. The reason for thedecline appears to be no mystery: overall health insur-ance premiums for given coverage have been rising rap-idly while consumer incomes have been near stagnant;consumers with little more to spend are increasingly re-luctant to spend it on something with a rapidly growingprice. But this simple story of prices growing relativeto incomes is not quite right, and one piece of evidencearguing for a more nuanced approach is the well-known,widely varying susceptibility to this problem across thepopulation. Most Americans are still getting good insur-ance protection for good medical care (at least in terms

PAGE 1

1

................. 17631$ $CH1 12-17-09 12:21:47 PS

M A R K V. PAU LY

of how they perceive their care), but for a sizeable frac-tion things are getting worse; not everything is broken,but some things for some people definitely are.

One of the flash points for this stressed system is theavailability of insurance for consumers who do not findit easy to access the dominant form of private insurance.Most people are offered insurance related to their job,and most still take it. But increasingly large numbers ofworkers have jobs where coverage is not offered, orwhere the terms under which it is offered do not inducethem to take it. It is the erosion of job-based coveragethat is the main problem. There is an alternative to job-based insurance: consumers could buy health insurancein the same way they buy other kinds of insurance—voluntarily, as individual consumers, from private insur-ance firms. A small fraction (about 7% of the nonelderlypopulation or 10% of those who have private insur-ance) do buy in this way.

However, the current individual insurance market isgenerally thought to not work well, less well, in mostviews, than even the misfiring employment-based mar-ket. In the individual market consumers tend to pay alot for what they get, and, even with that, may worryabout not being able to get and keep these overpricedproducts at all. Precisely because many of the uninsuredappear to be permanent dropouts from the employ-ment-based system, attention has turned to the develop-ment of an improved individual insurance alternative.

The current administration and Congress propose

PAGE 2

2

................. 17631$ $CH1 12-17-09 12:21:47 PS

H E A LT H R E F O R M W I T H O U T S I D E E F F E C T S

changes and alternatives to this ‘‘nongroup’’ market,some quite dramatic and far-reaching. Are these newmodels the best, or are there other alternatives bettermatched to actual needs and with fewer adverse side ef-fects? Can the current individual market serve as a foun-dation for reform, or must it be swept away, root andbranch, and replaced with something entirely new? Thecurrent, most popular model of individual insurance re-form proposes to revolutionize the way that marketworks, replacing the current market with a government-run insurance exchange offering a dominant public orconsumer cooperative plan along with heavily regu-lated, privately managed alternatives. Consideration ofits advantages and disadvantages may help in the choicebetween full replacement and modification.

I will begin with a discussion of the way in which in-dividual insurance markets currently fit into the overallpattern of health insurance markets (considering boththose who have and those who do not have insurance). Iwill then try to identify what is distinctive about currentperformance in that market relative to the alternatives:small-group private insurance, large-group private in-surance, and public insurance. Finally, I will suggest im-provements in the individual market that build on whatcurrently works well, contrast these with alternative re-forms, and offer a realistic assessment of how much im-provement we can demand and expect.

My main conclusion is that, while there surely aresome serious deficiencies in today’s individual insurance

PAGE 3

3

................. 17631$ $CH1 12-17-09 12:21:47 PS

M A R K V. PAU LY

market, the current criticisms are overly harsh, oftenbased on anecdote and speculation, and ignore some im-portant advantages in this market that should be pre-served. I will also argue that the most frequentlydiscussed alternative reforms reflect an excessively sim-plistic view of insurance markets that leads to policieslikely to have serious adverse side effects, very likely todo more harm than good by driving out some of thebeneficial features of the present market. I will arguethat there are better approaches that take into accountcurrent strengths of the individual market and can yieldmuch higher net benefits with much greater confidence.

Problem definition

Medical care is expensive because it is a prodigious con-sumer of high-value real resources, principally labor butalso some capital and technology. In the long run, it canbe made a little cheaper if consumers are willing to sac-rifice some current features of care, but it cannot bemade low cost. Its price can be temporarily lowered be-cause some labor inputs (principally health profession-als) have few equally attractive short-run alternatives:doctors’ net incomes and nurses’ wages can be squeezedfor a while, although this only redistributes welfare anddoes not lower real resource cost and must be tempo-rary. Health insurance premiums are high, and growing,because the medical costs they cover are high and grow-

PAGE 4

4

................. 17631$ $CH1 12-17-09 12:21:48 PS


ing. Health insurance and health care just cannot bemade permanently ‘‘affordable’’ to low-income house-holds except through subsidies. Someone has to pay forothers, and someone has to decide how much to payand for what.

However, medical care also provides high and grow-ing benefits for our lives and those of our loved ones,and, on average and up to this time, the benefits appearto exceed the cost. So the nonpoor bulk of the popula-tion will still want care and would want it even if theyhave to pay the full high cost. To evaluate how well al-ternative ways of financing perform, we need thereforeto begin by estimating how much care a householdshould have based on its own resources and then addingto that the contributions other citizens are willing tomake if the household’s own efforts fall short and it isthought to be in need.

How much care should a person have, and howshould it best be paid for? We focus primarily on thenonpoor, because most of the population is not poor,but we deal as well with attempts to help the privateindividual market to work for low-income families. Webegin with the premise that consumers should be wellinformed and that they should at least get the care theywant and are willing to pay for on their own behalf.The pedantic economic planning model gives the sameanswer to this question as for all goods and services:the person should get care up to the point at which itsadditional benefit to that person equals its real resource

PAGE 5

5

................. 17631$ $CH1 12-17-09 12:21:48 PS

M A R K V. PAU LY

cost given the person’s health state and income. The per-son and the person’s household should then pay for thatcare in a way that protects household wealth from thethreat of high spending because of the uncertain occur-rence of serious illness that is expensive to treat; thereshould be insurance.

These concepts of the ideal are abstract at this point,but they are more useful and more meaningful thansome other suggested alternatives. People talk about theright to medical care, but they do not say how much isrightful or how it should be produced and paid for. Orthey talk about ‘‘the right care of the right quality at theright time and at the right price’’ but do not tell us what‘‘right’’ means. It surely does not mean what doctorswould most prefer, or even what insured patients wouldmost prefer (given in both cases that insurance is pay-ing). Determining the benefit of more care to a givenperson requires knowing what the doctor knows (howsick the person is and what care provides what benefit),but this information must be combined with what theconsumer-patient knows about the value placed onhealth outcomes relative to other uses of income and onside effects and costs (monetary and nonmonetary) oftreatment.

Traditional health insurance usually gets in the wayof making a good joint decision because it makes expen-sive services appear artificially cheap. There needs to bea trade-off between this inappropriate stimulus to highuse and high cost (called ‘‘moral hazard’’ in the insur-

PAGE 6

6

................. 17631$ $CH1 12-17-09 12:21:49 PS


ance literature), on the one hand, and financial risk pro-tection on the other (Pauly 1968). With sufficientinformation, consumers of insurance can make all ofthese trade-offs, but getting the information can be chal-lenging and costly. Still, it seems plausible that the bulkof people who are middle class and above chooseenough coverage and enough care that others would nothave great concern for (and be willing to pay muchmore for) their additional use of care or additional in-surance benefits. If anything, among the middle classthere is more overuse (benefit less than cost) than un-deruse (the reverse), and a solution to the problem ofunderuse for them is better and more persuasive infor-mation for those unaware of the benefits from care, notmore insurance for all. As will be argued in more detailbelow, markets for care and markets for insurance forthe nonpoor have no serious intrinsic impediments togood functioning, but they do require care and feedingand are easy to mess up.

For lower-income households in America (given mod-erate risk levels) and for high-risk households (givenmoderate income), this reasonable state of affairs doesnot apply. Generous subsidies for better care are neededfor them. Fortunately, subsidized insurance can bothprovide financial protection and reverse moral hazardfor community benefit, stimulating the use of care thatothers value more than the person can or does.

The majority of uninsured Americans are not poor,obtain their income from employment (including self-

PAGE 7

7

................. 17631$ $CH1 12-17-09 12:21:50 PS

M A R K V. PAU LY

employment), but do not work in firms that offer thememployment-based group insurance coverage. While re-form could try to put many of these people into Medi-caid or other public plan or (through an employermandate) into employment-based insurance coverage,such changes would be large, awkward, and difficult. Amore natural setting for covering the bulk of the unin-sured is the setting already used by about 30% of peoplelike them—the individual insurance market (Pauly,Percy, and Herring 1999). But this market is generallyregarded (correctly, as we shall see, in its current form)as a relatively poor performer, both in terms of effi-ciency of administration and in terms of the efficiency-equity issues surrounding risk variation. In individualinsurance, to put it bluntly, you pay a lot for what youget, and, if you are unfortunate enough to become ahigh risk, you pay even more or get even less. However,this bad reputation masks some real advantages to thismarket, even as it is, and, more importantly, inappropri-ately excludes this market from consideration for modi-fications that could reduce the bad features and enhancethe good ones.

The lay of the land:How many people have what problems?

Not all parts of the private U.S. insurance market areyet in disarray. The market works well—at least as well

PAGE 8

8

................. 17631$ $CH1 12-17-09 12:21:50 PS


as is possible, given the cost of health care—for verymany people, but very poorly for others. We will usetwo measures of ‘‘works well.’’ One is whether peopleend up obtaining insurance. The other is what they haveto sacrifice to get insurance. The first indicator is mea-sured by the proportion of those not receiving publiccoverage who have private coverage. The measure of thesecond indicator is more complex, since the alternativeto paying nothing for health insurance is not usually get-ting good medical care for free; rather, it is being at riskfor high out-of-pocket payments for constrained care ofquestionable quality or going without. The real cost ofprotecting yourself against the risk of uncertain medicalspending is the difference between what you pay orwould pay for health insurance and what you expect onaverage to get back as benefits. For a set of similar risksbuying the same policy, this cost is the difference be-tween total premiums paid and total benefits received;in insurance parlance this is called the ‘‘administrativeloading’’ (or just ‘‘loading’’) on insurance. This measureof insurance price or cost is not perfect since it does notcapture the offsetting benefit of getting more care thanif one were uninsured, the offsetting cost of that care,and the nonmonetary cost of being limited in what careyou can get and from whom by a managed care insur-ance plan. But it will do for a start. Economic theoryand definitive empirical research tell us that the higherthe loading, the less attractive is insurance, other thingsequal, and the more likely people are to choose to be

PAGE 9

9

................. 17631$ $CH1 12-17-09 12:21:51 PS

M A R K V. PAU LY

uninsured or not make strong efforts to become andstay insured. The punchline: the worst off a person orgroup can be is when they are often uninsured andwould pay a high loading for insurance if they got it.

There are two primary determinants of who gets andkeeps health insurance in the United States: decent in-come and employment at a large firm. The role of in-come is shown in Table 1. As indicated there, thelikelihood of being uninsured for a person under 65 isonly half as high if the person is in a household whoseincome is at or above 400% of the poverty line than ifthe person is poor or near poor, even with Medicaid asa public safety-net program for the poor.

Even so, uninsurance is not unknown among the mid-dle class (the median U.S. household income is at about325% of the poverty line), and a quarter of the unin-sured (11.6 million out of 45 million) are above the300% cutoff. (See also Yegian et al. 2000.) The tablealso shows that, at about 160% of poverty and above,most people somehow get private coverage. The impli-cation is that, while income matters, there is more tocoverage than just income or affordability.

Table 2 indicates the second major influence: firmsize. The measure of size in this data is not ideal. Wemust look at ‘‘establishments’’ or places of employment,rather than firm size, because a large employer such as afast-food chain may be composed of many small estab-lishments. But since a large establishment cannot be partof a small firm, the measure still serves to indicate theeffect of firm size, and the message is clear: even for

PAGE 10

10

................. 17631$ $CH1 12-17-09 12:21:51 PS


Poverty Level Uninsured Government Private Total

<100% n 11.4 17.0 5.7 34.1% 33.4% 49.8% 16.8% 100.0%

100–125% n 3.6 4.5 3.1 11.2% 32.2% 40.5% 27.3% 100.0%

125–150% n 3.7 3.8 4.0 11.5% 32.6% 33.1% 34.3% 100.0%

150–175% n 2.8 2.9 4.7 10.5% 26.8% 28.1% 45.0% 100.0%

175–200% n 3.0 2.4 5.5 10.9% 27.3% 21.9% 50.8% 100.0%

200–300% n 8.8 7.0 27.9 43.7% 20.1% 16.0% 63.9% 100.0%

300–400% n 4.6 3.8 28.3 36.7% 12.5% 10.4% 77.1% 100.0%

>400% n 7.0 7.1 89.5 103.6% 6.8% 6.8% 86.4% 100.0%

Total n 45.0 48.5 168.7 262.2

TABLE 1Insurance Status of Individuals under 65 by Ratio of Household Income to Federal Poverty Level (FPL)

Source: Data from U.S. Census Bureau, Current Population Survey, March 2008Supplement.Note: Weighted (n in millions); individuals reporting both government and privateinsurance are included in government insurance.

workers in relatively low-income households, insuranceis common if they work for a large employer, while well-off workers at small firms may be uninsured.

I will discuss what are the ‘‘active ingredients’’ thatmake insurance markets work well for large firms in

PAGE 11

11

................. 17631$ $CH1 12-17-09 12:21:52 PS

M A R K V. PAU LY

Establishment Size

Income as % of Poverty Line

< 50 workers ≥ 50 workers

Below 300% 25.8% 13%

Above 300% 7.9% 2.3%

TABLE 2Percentage of Uninsured Adults by Ratio of Household Income to FPL and Establishment Size (households with a full-time worker)

Source: Data from Centers for Disease Control and Prevention, National Center forHealth Statistics, “National Health Interview Survey,” 2008.Note: Excludes unknown or other income categories.

more detail below, but for the present it is sufficient tonote their relatively good performance in terms of bothcoverage and administrative cost. The few uninsuredpeople who work full time in large firms are usually noteligible for coverage, often because they have just beguna job or are in a near-minimum-wage job, or, muchmore rarely, turned down offered coverage because theyare not willing to pay the explicit employee premium.About 4% of the under-65 population offered employ-ment-based insurance will not take insurance if theyhave to pay anything for it, no matter what. (For two-worker families, one worker will often enroll as a de-pendent on the spouse’s plan.) The table further subdi-vides households with a large-firm worker by householdincome; there is a modest effect of income on the pro-portion with coverage but the proportion is above thenational average regardless of household income.

PAGE 12

12

................. 17631$ $CH1 12-17-09 12:21:53 PS


There are two other employment-related categories:those who are employees at small firms and those whoare not employees of any firm, either because they areself-employed or not working. Here the picture is quitedifferent. In these categories, the number of uninsuredis high, even among relatively high-income households.It is highest of all among people who do not at presenthave a group insurance option (although in principleanyone could try to get a job at a large firm); it is alsohigh for middle-class people who work at small firms.

Can individual insurance help?

Those in households where no group insurance is of-fered (whether or not the person works as an employee)are ‘‘at risk’’ for purchasing individual insurance. Per-formance of this form of insurance is not impressive;in this case, insurance reaches only about 25% of thepopulation. But many of those who work as employeescould have taken jobs at firms offering coverage. A sub-population with no alternative but individual insurancewould be those who are self-employed or not employed;here the proportion with coverage is still low, at abouta third (Marquis et al. 2006).

The proportion of the population with private insur-ance has been trending downward, to some extent offset(and perhaps caused) by expansion of public coverage,especially through Medicaid and the State Children’sHealth Insurance Program (SCHIP). The fall in private

PAGE 13

13

................. 17631$ $CH1 12-17-09 12:21:53 PS

M A R K V. PAU LY

coverage has not generally occurred among people inhouseholds with a worker at a large firm, in part be-cause such firms continue to offer coverage and in partbecause household incomes for workers at large firmsare higher and therefore less subject to crowd-out frompublic plans. Moreover, although the data is somewhatsoft, it appears that the erosion in coverage for workersat small firms does not largely result from firms drop-ping existing coverage, but rather from firms that do notoffer coverage replacing firms that did. There has alsobeen some tightening of eligibility so that more workersin firms that offer coverage do not qualify.

Table 3 provides some rough data on the componentsof administrative cost for insurance bought as groupcoverage or as individual coverage. The loading ishigher the smaller the group, and it is also more vari-

General ProfitGroup Size Commissions Admin. (net) Taxesa Claims Total

Individual N/A N/A N/A N/A N/A 40–100%

1–20 8% 11% 7% 3% 5% 34%

100–500 1.6% 4% 4% 2.3% 4% 16%

10,000 + 0.1% 0.7% 0% 2.1% 3% 4–6%

TABLE 3Components of Insurer Administrative Costs (as percentage of claims)

Source: Author’s calculations based on data from the Congressional Research Service (2000).aAvoided if firm self-insures

PAGE 14

14

................. 17631$ $CH1 12-17-09 12:21:54 PS


able. This pattern and its rationale will be discussed inmore detail below, but the main message is clear: theadministrative loading, and therefore the true price ofinsurance, is much higher for individual than large-group insurance. This table differs from the originalcalculations (created by the Hay Group in 1999) in re-ducing the estimated charge for claims processing to amore uniform percentage; it also makes an estimate forindividual coverage based on extrapolating the verysmall group numbers and using other evidence in theinformal literature.

The time trend in the loading percentage has beenmodestly upward across the board; the average loadingin private insurance has probably increased by 1 to1.5% of premiums over the last 20 years. This is largelydue to the spread of managed care, which is somewhatmore costly to administer (per dollar of benefits or pre-mium) than was old-style insurance, which just paidclaims but did not select panels of providers, preapprovecoverage, or attempt to manage care. This growth incost seems to be fairly uniform by group size on average,although anecdotes of substantial increases in premiumsare much more common in the small-group and individ-ual markets; there is more dispersion in premiums insuch markets.

The extent of competition among private insurersmeasured by firm market shares (in states or cities) hasnever been very high. In most areas, the Blue Cross andBlue Shield plans have dominated for many decades,

PAGE 15

15

................. 17631$ $CH1 12-17-09 12:21:54 PS

M A R K V. PAU LY

often with market shares in the 80–90% range. Theshare of the largest non-Blue for-profit plan has neverbeen greater than 25%. There have been some excep-tions and some changes to these generalizations. In Cali-fornia and New York the Blue plans were never asdominant, and both private for-profit insurers and inde-pendent managed care plans were important. There hasbeen consolidation and change during the past decade;the independent HMOs were bought up or disappeared,some of the Blue plans converted to being investorowned, and the separate Blue plans with different geo-graphic territories merged. Data are soft here as well be-cause some consumers have policies from more than onefirm and because self-insured employers really do notbuy insurance from any insurance firm, even if they usean insurance company to process claims or organizepanels to supply medical services.

The Blues still dominate, with about 100 million outof about 170 million total privately insured. Accountingfor the Blue plans that are non-investor-owned, the mar-ket splits about 50–50 between for-profit and nonprofitor mutual (consumer cooperative) firms. Historically,the Blue plans were typically even more important thanother insurers in the individual market, although UnitedHealth Care now has a substantial individual presencedue to its acquisition of Golden Rule Insurance, and theinvestor-owned (but Blue) Wellpoint firm is active in theindividual market as well (though individual insuredsare only about 7% of the total population for which it

PAGE 16

16

................. 17631$ $CH1 12-17-09 12:21:55 PS


administers insurance). Only concentration, not owner-ship form, appears to affect competition, and there issome evidence that high concentration in managed caredoes lead to monopoly behavior in the HMO market(though not necessarily in the overall health insurancemarket) (Pauly et al. 2002). Profit levels do not yet sig-nal generally high market power, even when the numberof insurers selling in a market is small; net income ofprivate insurers was negative or very low in the later1990s, and has recovered, but profits as a proportion ofpremiums rarely exceed 5%. The local markets in whichone insurer is dominant are always dominated by a Bluefirm, which is usually nonprofit and subject to greaterstate scrutiny than commercial insurers.

There is, however, more competition among insurersthan simple market-share measures would suggest. Forone thing, most large groups self-insure, purchasing ad-ministrative services only (or some reinsurance) fromthe private insurers. These large buyers would not toler-ate prices (for coverage or administrative services) thatyield high insurer profits because they can turn both toadministrative-services-only firms (such as benefits con-sultants) and plans in other states to provide those ser-vices. Smaller employers and individuals are morevulnerable. Even here, however, there is potential entryinto the pure insurance business should any existing in-surer try to exercise serious market power. The mainpossible source of power is not insurance per se (per-forming the function of pooling risk); when there is

PAGE 17

17

................. 17631$ $CH1 12-17-09 12:21:56 PS

M A R K V. PAU LY

market power it comes from a dominant insurer’sprovider network and/or management tools, whichnew entrants may find hard to duplicate. Still, whileconsolidation in the industry is troubling, such changesdo not appear to be major factors in increasing premi-ums or reducing the proportion of the population withcoverage.

There does not appear to be comprehensive data onthe profitability of individual insurance. For-profithealth insurers as a whole earned a pretax return on eq-uity in 2008 of about 8% to 11%, a moderate rate bythe standard of other industries. They earned a low rate(at 2% to 4%) as a proportion of revenues; cuttingprofits in half would produce a barely noticeable reduc-tion in premiums. The return on equity of Golden RuleInsurance, the largest company specializing in individ-ual insurance before it was acquired by United HealthCare, was about 7% in 2000. This line of insurancedoes not appear to be so profitable that it is attractingentry or growth; the recent moderate expansion by Well-point was offset by the exit in 2002 of Mutual ofOmaha, which had been a major player. Profits for thistype of insurance are not consistently out of line orabove average relative to competitive returns elsewherein the economy. Individual insurer profits are in the bil-lions of dollars but are a tiny fraction of the industry’stotal revenue and costs, which are in the hundreds ofbillions. More recent (but incomplete) data found theaverage margin as a percentage of premiums for individ-

PAGE 18

18

................. 17631$ $CH1 12-17-09 12:21:56 PS


ual insurance to be about 2% or less in 2008, thoughit showed a loss for small-group insurance (Austin andHungerford 2009).

Coverage for high risks

One of the main criticisms of individual insurance, andone of the main motivations for dramatic increases inregulation, has to do with the belief that the individualmarkets treat poorly people of substantially greater riskthan average. As we will show, this is true to a consider-able extent, but group insurance does an equally poor(or worse) job of protecting high risks, especially thosewho work for small firms. But we first want to make akey point about relative importance: compared to thenumber of people under 65 who are average or normalrisks, the fraction who are high risks is very small.

In some ways, the point should be so obvious as to beundeserving of italics: the country cannot be some kindof perverse Lake Wobegon in which everyone is anabove-average risk. In fact, as well shall see, the shapeof the distribution of risk levels is skewed to the right,meaning that most people are below-average risks andthe average is driven up by only relatively few high risks.This point is important because common sense wouldsuggest that if the problem of high risks is relativelyrare, it would not generally be necessary to make drasticchanges in markets for the great majority of the popula-

PAGE 19

19

................. 17631$ $CH1 12-17-09 12:21:56 PS

M A R K V. PAU LY

tion; instead, specialized interventions targeted at thesmall minority of high risks (with associated modest im-pacts on the rest of the population) would be more effi-cient, more transparent, and more feasible. (As we willsee, the small fraction of high risks is a different phe-nomenon from the rarity of a high-cost illness for a per-son of average risk; in the latter case, insurance isdesigned to deal with the rare event that has yet to hap-pen but could occur.)

There is no definitive measure of the proportion ofthe under-65 population not covered by Medicaid orMedicare (as disabled or with kidney failure) who arehigh risk. Estimates of the proportion who are ‘‘unin-surable’’ are as low as less than 1% (Frakt, Pizer, andVrobel 2004), but I believe a reasonable number is theproportion with high-cost chronic conditions, which isabout 4% of this population (Pauly and Herring 1999).(This is the average proportion with expected expensesdue to chronic conditions that are more than twice theaverage, controlling for demographic characteristics;the proportion with expenses more than three times theaverage is 2%.) An implication of this small proportionis that the problem of a person becoming a high risk justabout the time individual insurance is initiated shouldbe even more rare, so the problem of preexisting-condi-tion exclusions must affect only a tiny fraction of thepopulation (7% with individual insurance times 4%high risk times the probability of becoming high riskjust when the individual sets out to buy insurance has

PAGE 20

20

................. 17631$ $CH1 12-17-09 12:21:57 PS


to result in a tiny number). There will still be ‘‘manythousands’’ of people with such problems, so somethingshould be done, but they make up such a small share ofthe total that wholesale changes hardly seem needed.

Risk in health policy analysis has many meanings(Pauly 2007), but the one to be discussed here is therisk of medical care spending (that will potentially becovered wholly or in part by insurance). Risk in thissense means expected or predicted medical spending; aperson will be said to be a high risk if, based on currentcharacteristics (such as age, ethnicity, health condition),the person would be expected to have expenses muchabove average. This meaning is to be contrasted withthe use of the term risk in the phrase ‘‘risk pooling.’’In risk pooling properly defined, risk is discussed in thecontext of a population with the same expected medicalspending whose actual spending will vary substantiallydepending on the random pattern of events. What ispooled is the unknown that is yet to happen, not theconsequences of known events that have already oc-curred. (Risk pooling will be discussed in more detaillater.)

One common characterization of medical care spend-ing is that it follows the ‘‘80–20 rule’’: in any year, 80%of spending is associated with 20% of the populationthat gets really sick; the other 4 out of 5 people are rea-sonably healthy and so spend less on average and intotal. Sometimes it is said that this fact alone shows thatthere will be a potential problem with treatment of high

PAGE 21

21

................. 17631$ $CH1 12-17-09 12:21:57 PS

M A R K V. PAU LY

risks. For example, Charlie Baker of Partners Health(2009) observed, ‘‘20% of the population incurs 80%of the costs, and vice versa. That’s called risk pooling,and it’s why Massachusetts (and Hillary Clinton dur-ing her presidential campaign) called for mandatinghealth insurance coverage as part of the state’s reformefforts.’’ While Baker’s discussion of the ‘‘rule’’ is accu-rate, his conclusion about the need for mandating isnot necessarily correct: even if people were all of thesame risk, the uncertain incidence of illness would re-sult in some spending much more than others, but allmight well voluntarily purchase insurance to spreadthis risk. Indeed, the whole basis of insurance is thatpeople of the same risk level at the beginning of a pe-riod can expect to realize quite different loss levels overthe period; insurance converts this lottery on bad out-comes into a uniform premium that all may pay andtherefore end up no worse off financially regardless ofwhat actually happens.

What is crucial to the policy discussion instead is thedistribution of risk, and that in turn depends on the dis-tribution of characteristics that predict high spending, adistribution that, as it turns out, does not fit the 80–20rule. There are more truly low risks and fewer truly highrisks than would be indicated by the rule, especially atthe extremes of the distribution of losses. However, itremains true that risk is unevenly distributed and dis-tributed in the skewed fashion implied by the rule.

PAGE 22

22

................. 17631$ $CH1 12-17-09 12:21:57 PS


Community rating: the worst possibleway to do a good thing

There is a strong case (to be treated in more detail later)that can be made against simple annual risk-rating ofhealth insurance and its anticipated effects on high risks.The problem is this: annual risk rating would exposepeople to risk. The risk they confront is called ‘‘reclassi-fication risk’’ and represents the possibility that unex-pected onset of a chronic condition may cause aperson’s future health insurance premiums (under sim-ple annual risk rating) to take a big jump. Not only maythis make insurance unaffordable for lower-income highrisks, but it also makes those who might become highrisks at all income levels more miserable than they needto be.

In the recent policy proposals from the Obama ad-ministration and the Democratic Congress, the solutionto this problem that is most obvious but has the worstside effects is the one that has been virtually the soletopic of discussion: regulate premiums by requiring in-surers to charge buyers with different expected expensesunder a given policy the same premium—so-called‘‘community rating’’—and close any loopholes by for-bidding insurers from limiting coverage for preexistingconditions at this premium. There has been almost nodissent expressing the general undesirability of thisapproach.

PAGE 23

23

................. 17631$ $CH1 12-17-09 12:21:57 PS

M A R K V. PAU LY

There are strong arguments against community ratingin the insurance literature (Pauly 1970). It is thereforesomewhat surprising (if not disappointing) that this oldclunker has such a powerful grip on policymakers. Hereis the simple critical argument in the literature. Assumethat competitive insurers can perfectly distinguish riskand so set premiums based on risk (so there is no reasonto expect adverse selection or cream-skimming).Higher-than-average premiums for people who becomehigher-than-average risks have some undesirable effects(just mentioned), so it is good public policy to assumethat those higher premiums are cushioned in some way.There may also be a social equity motivation for cush-ioning the financial status of those who become highrisks, even those with moderate incomes. A good wayto do so is to allow insurers to charge risk-rated buyershigher premiums but then to subsidize those premiumsfor those high-risk buyers who are thought to be deserv-ing of help—either because they unexpectedly becamehigh risk (risk reclassification) or because there is soci-etal concern about their ability to buy health insurance(unaffordability). The money for this subsidy should beraised by the most efficient and equitable tax available,usually thought to be the general income tax, although avalue added tax would possibly be superior. Other thansome administrative complexities concerned with iden-tifying the targets for subsidy and making sure the sub-sidy itself does not distort the amount of insurancebought, this simple policy solves the problem.

PAGE 24

24

................. 17631$ $CH1 12-17-09 12:21:57 PS


In the ideal arrangement, no price regulation isneeded; insurers should be able to charge more for highrisks. Community rating is a much inferior solutionstrategy. The most serious unintended adverse conse-quence is that, if it is implemented in otherwise competi-tive insurance markets without targeted subsidies,insurers required to underprice (relative to expectedbenefits cost plus loading) for higher risks must over-price policies sold to lower risks in order to cover theirtotal cost. The consequence of this price distortion isthat lower risks will drop or fail to buy coverage. Thereis substantial evidence that this is what happens, and tosuch an extent that community rating actually causesmore people in total to be uninsured, even as it modestlyreduces the number of uninsured high risks. Herringand Pauly (2007) found that rate regulation slightly in-creased the relative likelihood that a high risk would getindividual coverage (from 0.96 of the low risk likeli-hood to equal likelihood) but that the increase in low-risk premiums would increase the overall uninsuredpopulation, as more low risks were driven out than highrisks were brought in. Case studies of the initiation andfunctioning of community rating in different states havefailed to find evidence that it helps with the problem ofthe uninsured, or even that it consistently helps signifi-cant numbers of high-risk uninsured, compared to norate regulation (Swartz and Garnick 2000; Chollett andKirk 1998). Often it results in a meltdown of the entireindividual market as companies withdraw from that

PAGE 25

25

................. 17631$ $CH1 12-17-09 12:21:58 PS

M A R K V. PAU LY

business. It does provide more employment for insur-ance regulators.

One might still prefer community rating to doingnothing, if one attaches high enough value to coveringthe high risks and low enough or no negative value toincreasing the number of low-risk uninsured, but thepoint of insurance theory is that there is a better alterna-tive to doing nothing that does not have this side effect.Formally, community rating can be thought of as finan-cing a subsidy to insurance for higher risks (there beingpotential social value to the subsidy) through an excisetax on insurance sold to lower risks. The economics justreviewed tells us to avoid specific excise taxes in favorof more general taxes. The alternative subsidy/financingmodel outlined above does just that; there is no reasonto think that higher income taxes would appreciably af-fect the insurance purchasing plans of anyone, regard-less of risk level. From the viewpoint of economicefficiency, subsidizing high risks is often desirable, butpaying for that subsidy by taxing low risks never is. Noris there an obvious equity reason why low risks shouldpay. The reason for the political appeal of this policy isthat the tax on low risks is not counted as a tax, doesnot show up on any budget, and is opaque for all, evento low risks who are likely to blame the insurers, not theregulators, for their high premiums.

The primary rationale for community rating, it is im-portant to note, is not adverse selection that arises fromintrinsically better knowledge of risk by buyers than

PAGE 26

26

................. 17631$ $CH1 12-17-09 12:21:58 PS


sellers. Rather, the targeted policy problem is risk ratingwhere the insurer actually has figured out who is higher-than-average risk and is proposing to charge that personmore than average and likewise to charge less than aver-age to low risks. Community rating will then causeadverse selection (and its supply-side mirror image,cream-skimming) when it did not have to exist. Thereare theoretical reasons to consider community rating asa solution to the problem of true adverse selection, butthe rarity of this case combined with the availability ofalternative ways to ensure that low risks remain in themarket (such as subsidies or mandates) diminish the im-portance of this argument.1

Certainly one of the most ironic recent studies is aCommonwealth Fund study that asserts that reform,with community rating as its centerpiece, will helpyoung people and lower risks to get coverage. It notesproposals for state regulations requiring insurers to per-

1. Even here, a pooled policy charging the same premiums andselling the same coverage to high and low risks may temporarilyemerge, either because of community-rating rules or because it ispreferred by low risks to very skimpy though cheap coverage. Butthen insurers and low risks will realize that there is another policy,at a premium based on the low-risk rate though somewhat lessgenerous than the pooled one, that the low risks will prefer butthe high risks will not—so the market cannot be stable. But for-bidding this policy will make the low risks worse off. Mandatorycommunity-rated basic coverage can sometimes improve effi-ciency if people at different risk levels are permitted to supplementin different ways.

PAGE 27

27

................. 17631$ $CH1 12-17-09 12:21:58 PS

M A R K V. PAU LY

mit parents to keep dependents on their policies afterage 23 (at higher premiums) without noting the absenceof evidence that such provisions make a difference. Themost striking omission is the failure to note that majorreform proposals would have the effect (through com-munity rating that limits appropriate risk adjustmentsfor age) of dramatically raising premiums for youngpeople. Some of the reform proposals allow somewhatlower premiums for young people, but the reduction ismuch less than what would be consistent with theirlower risk. Even the subsidies offered for some youngpeople are unlikely to change their (correct) perceptionthat overpriced insurance is a less attractive buy, onethat rational young people will reject. There is no betterway to drive more healthy young people out of the indi-vidual market than to implement proposed insurance re-forms.

In addition to the flaw of distorted prices, the regula-tory process associated with community rating gener-ates a host of other problems. It provides incentives toinsurers to avoid or underserve high risks (when for-merly at risk-rated premiums, selling insurance to highrisks would be profitable). Rules need to be put in placeto risk-adjust the revenues insurers will collect, adjust-ments that are bound to be imperfect. So then theremust be heavy regulations to close off all avenues bywhich lower risks and insurers conspire together toavoid the cross subsidy: policy design, marketing, inno-vation, prevention programs, and provider networks all

PAGE 28

28

................. 17631$ $CH1 12-17-09 12:21:59 PS


must be controlled. Finally, the doors are thrown openfor political and special interests to lobby: public healthtypes against patient cost-sharing, managed care oppo-nents against aggressive plans, and politicians forwhichever provider or insurer group wants to make surethat they are part of the preferred plan, all in the nameof keeping out plans that might appeal to low risks.Imagination will be stifled, the uninsured will increase,and political rent-seeking will be rampant.

So as the discussion of health reform shifts to that ofhealth insurance reform, it is important to note that thecenterpiece of congressional legislation—federal com-munity-rating regulation—will itself increase the gen-eral incentives for the group already most at risk ofbeing uninsured, the good risks, to become even morelikely to be uninsured. Far from being an inevitable con-sequence of health reform intended to assure coverageof high risks, this adverse effect of community ratingcan be avoided by choosing a different way to reformthe individual market.

What’s wrong with individual insurance?

Individual insurance, with or without community rat-ing, has its serious problems, too. The primary problemwith individual insurance concerns its pricing. Its highadministrative cost makes its average price high, andthere is substantial variation in prices for the same prod-

PAGE 29

29

................. 17631$ $CH1 12-17-09 12:21:59 PS

M A R K V. PAU LY

uct across sellers. That is, regardless of risk level, indi-vidual premiums are high and variable relative to thebenefits received. There appears to be substantial varia-tion in the ratio of premiums to expected benefits acrossdifferent insurers’ individual products, as well as varia-tion across risk groups. The result is that, no matterwhat your age and risk level, if you enter the individualmarket you will face the unlovely prospect of high andvarying prices for essentially the same product, alongwith a great variety of products. Getting a reasonablygood deal is not going to be easy and (depending onhow you define ‘‘good’’) may not be possible. As a re-sult, some transactions occur at high prices, and alltransactions occur at premiums that are considerablyhigher than benefits received. A potential partial solu-tion to this problem is to use an insurance broker tohelp find a good price for a good product, but brokersadd commissions of their own.

There is no definitive data on administrative costs (or,as already noted, profit margins) across the set of allindividual insurance options in a local market.2 Fromthe information we have, it appears that the medical lossratio—the proportion of premiums actually paid outeach year as benefits—is about 65% to 75% in the indi-

2. We can and will make some rough calculations about the ob-served aggregate average ratio of premiums to benefits, but, becauseinsurance markets are local, this is not necessarily a good measureof what a given person will face.

PAGE 30

30

................. 17631$ $CH1 12-17-09 12:21:59 PS


vidual market. What does not go for benefits goes topay administrative costs (the ‘‘loading’’) and profits.

Good comprehensive data on the components of indi-vidual insurance loading is even harder to find, but itdoes appear that the cost of processing claims in individ-ual markets is comparable to that in group insurancemarkets, at about 4% to 6% of premiums. This shouldbe no surprise, because insurers who sell individualproducts usually also sell group and network productsand use the same claims-payment mechanisms for alltheir business. And, as already noted, profits (for the mi-nority of for-profit individual insurers) or ‘‘surplus reve-nues’’ (for the nonprofits) appear to be a small fractionof premiums.

It is the other two main categories of expense—selling costs and general administration—that aremuch higher in the individual market than in otherpublic or private group markets. The agent or entitythat persuades someone to buy this high-priced insur-ance will have to be paid.

Although an agent who sells a given policy to a largegroup will get a higher commission than if the same pol-icy were sold to an individual, the commission risesmuch less than proportionately—so the commission pernew person insured is much higher for the single-personsale than for the multiple-person one. General adminis-tration costs are also higher per person. These costs gofor billing and collection of premiums and for answer-

PAGE 31

31

................. 17631$ $CH1 12-17-09 12:21:59 PS

M A R K V. PAU LY

ing policyholders’ questions. There are also economiesof scale in billing: compared to individuals, largergroups spread billing costs over more people, and theyalso typically pay on time with less (costly) naggingfrom the seller and fewer confused inquiries from thebuyer. Surprisingly, most states add insult to injury bytaxing health insurance premiums under their jurisdic-tion (as they tax premiums for other kinds of insurance),generally at about 2% of premiums, though the tax goesas high as 10% of premiums in New York. State taxeson insurances are a major source of revenue, sometimesthe third largest component (behind income and salestaxes). While 2% sounds relatively modest, it wouldrepresent a 13% increase in the net cost of coverage(premiums minus benefits) for a policy with a typical15% loading. State government revenue-raising adds al-most as much to individual insurance premiums as dothe profits that private and public insurers (averagedover for-profit and nonprofit) receive. It is ironic thatstates ask for federal funding for new programs to coveruninsured children and high risks when they are impos-ing taxes that cause more people to be uninsured.

What is perhaps surprising is that the direct cost ofunderwriting—the process by which an application forcoverage is reviewed, accepted, or rejected, and if ac-cepted, quoted a premium that will be related to someextent to risk—is actually a trivial fraction, surely lessthan 0.5%, of administrative costs. And, as we shall see,for the person who buys a policy and then renews that

PAGE 32

32

................. 17631$ $CH1 12-17-09 12:22:00 PS


coverage for subsequent time periods, underwriting isdone only once—so any cost gets spread over time.

The other potentially surprising omission is that thereneed not be any contribution to higher charges relatedto supposed lower ‘‘risk spreading’’ in the individualmarket. The combination of a one-year time frame forinsurance along with the bundling of thousands of indi-vidual applicants into a single risk portfolio means that,even in individual markets, the potentially wide varia-tion in what can happen to any one person cancels outacross thousands, and so should generate no additionalcharge for risk bearing per se. To be sure, actuaries intheir calculations of suggested premiums will sometimesadd a charge for risk in recognition of the risk-poolingbenefit provided to individual consumers, but since thatbenefit has no cost to the insurer, that charge will not besupported in a competitive market. And, as notedabove, profits on individual insurance appear to be closeto the competitive level.

There are other potential risks of the individual mar-ket—for example, less reliable data on the risk of anyone buyer and the attendant possibility of the ‘‘winner’scurse’’ in which buyers gravitate toward the most(overly) optimistic seller. But there is no quantificationof this risk—it is similar to the risk in any businesswhere prices are set too low compared to what coststurn out to be—and in any case it should be pooled inequity markets. There may be an actuarial fudge factorapplied to estimated claims to account for persistent un-

PAGE 33

33

................. 17631$ $CH1 12-17-09 12:22:00 PS

M A R K V. PAU LY

derestimation of risk for people who choose to buy, butthis does not add to cost; it just corrects the underesti-mate. In the more reliable group data, this ‘‘charge’’would have been included in an accurate estimate of an-nual benefits cost; it is no higher a cost in individualmarkets than in other insurance markets. More gener-ally, there should be no risk charge for an estimate thattakes the same value year in and year out.

It is sometimes argued that people with higher-than-average expected expenses will not have their ‘‘risk’’(which is not a true risk, of something yet to happen)pooled with lower-than-average risks in individual mar-kets, in line with the supposed risk-pooling function ofinsurance markets (Pauly 2007). But if higher risks fareless well in individual markets—something which, as weshall see, is not always necessarily the case—lower riskswill fare better. So, on balance, individual markets donot harm the average risk. People in individual marketscannot all be higher risks by the law of averages. Themedian risk, who is actually below the mean, will notpay more for individual insurance for this reason. Vol-untary insurance markets are not intended to pool therisk of things that have already happened. If group in-surance markets really do anything fundamentally dif-ferent about variation in individuals’ risk fromindividual insurance markets (something that is far fromcertain), they do not ‘‘risk pool’’; if they do anything,they ‘‘risk average’’ or ‘‘risk-based transfer.’’

A clear way to see the difference between risk averag-

PAGE 34

34

................. 17631$ $CH1 12-17-09 12:22:01 PS


ing and risk pooling is to consider situations in whichhigh risks and low risks buy comprehensive insurancecoverage and pay a community rate. Because coverageis complete, consumers bear no risk, and only pay a pre-mium they know with certainty. But now let communityrating be replaced with risk rating. If coverage remainsthe same, consumers still bear no risk of out-of-pocketpayments depending on whether or not they get sick.And they still can expect to pay premia they know withcertainty. The only distinction is that the certain premi-ums are different from before, higher for higher risksand lower for lower risks. Compared to risk rating,community rating does not lower aggregate risk; it onlymakes transfers from low to high risks.

The administrative loading in individual insurance isgreater than in employment-based group insurance (em-ployer resources handle insurance issues, so somegroup-insurance costs are transferred from insurer toemployer). However, the individual insurance loadingapproximates that in most other insurances that peoplebuy as individuals: homeowners’ insurance, automobilecollision coverage, and tenants’ insurance (Kunreuther,Pauly, and McMorrow, forthcoming). Term life insur-ance and auto liability insurance appear modestly lesscostly to administer, but, in the former case, purchasesonce made are usually repeated automatically (and atlittle administrative cost) year after year, and, in the lat-ter case, the resources used to determine the value ofclaims are generated by the legal system and incorpo-

PAGE 35

35

................. 17631$ $CH1 12-17-09 12:22:01 PS

M A R K V. PAU LY

rated into the benefit payouts, and so are not counted asinsurer costs. Individual health insurance is expensivebecause it costs a lot to administer.

There are potential ways to improve administrativeefficiency in the individual market, but high administra-tive costs do not show that there is intrinsic inefficiency(or, apparently, higher-than-competitive profits) in thismarket, given the product it offers, the production proc-ess for that product, and the channel used to offer it.

The wide pricing variation for similar products al-ready noted does, however, seem to be a defect. Widevariation in prices apparently does not translate intohigh insurer profits across the board; this is not a marketwhere all confused buyers are ripped off. As we willnote below, there is evidence that the bulk of buyerseventually do find their way to reasonably good deals—more so the more money there is at stake. The real inef-ficiency here is the direct and indirect cost of buyersearch, along with the riskiness inherent in the wholeprocess. The advent of online selling of individual insur-ance may have helped somewhat in recent years.

Still, there is a strong element of arbitrary and capri-cious disarray in the individual market that has given itan appropriately tainted reputation. Add the occasionaltale of a scandalous company pulling fast ones on itscustomers, hiding in fine print, and reneging on implicitpromises, and it is no surprise that this market does nothave defenders even among those who will want to useit. For example, Douglas Holtz-Eakin, then advisor to

PAGE 36

36

................. 17631$ $CH1 12-17-09 12:22:02 PS


Senator John McCain, said that ‘‘the current individualmarket is not a good place to live’’ (2007). He thenwent on to overstate things a bit: ‘‘There is nothing inthe individual market that everybody should embraceand love.’’ I disagree. I will argue that there are somethings in this market that could be embraced (thoughnot necessarily loved). The generally jaundiced viewhealth policy analysts of all stripes have of the individ-ual market is not undeserved, but there are some ad-vantages to that market which reform should cultivaterather than discard.

What is good about the individual market?

Some of the advantages of the individual market aremirror images of the disadvantages associated with itsadministrative cost and pricing variation. One advan-tage is that the individual person has a wide choice ofwhat policy to take rather than being limited to the pol-icy or set of policies offered in a group. In a sense, thehigher administrative cost of individual insurance is inpart the cost of variety. Variety will be most valued bypeople whose choice would have been different fromwhat the group would have chosen, so this is an advan-tage that will have varying values across people; the in-dividual market is more valuable to eccentrics than toconformists. An aspect which should be valuable to all

PAGE 37

37

................. 17631$ $CH1 12-17-09 12:22:03 PS

M A R K V. PAU LY

is that the individual controls the choice, rather than anemployer, a union, or a benefits department. In individ-ual markets the person does not have to worry thatsome other agent, such as the employer or team of bene-fits consultants, will decide to change the policy or evenchange the pricing of the policy.

In the group health insurance market, you get itcheap, but you get what someone else wants. Employersoften are confused by their health benefits and make ar-bitrary decisions to cut benefits even when their employ-ees would prefer more generous benefits and would beprepared to take lower raises to pay for them. In theindividual market, the only relevant confusion is that ofthe buyer—and the buyer has the option of obtainingreasonably good advice from brokers or financial coun-selors, though at a price. This does not, I must hasten toadd, mean that the person would not prefer the groupmarket. Even if consumers do not get exactly what theywant, they may still be happy if it is a lot lower in costand not too different from what they most like.

The variety of choices in the individual market is tosome extent limited by what is available in the area—inPhiladelphia I cannot choose a group- or staff-modelHMO, such as Kaiser—but the restriction is much lessthan in group settings. The number of choices offered topeople in employment-based settings tends to shrink asfirm size shrinks. Very many people who get group in-surance from small employers have no choice at all. Sothe people working for small firms—who are paying

PAGE 38

38

................. 17631$ $CH1 12-17-09 12:22:04 PS


loadings not very much lower than in individual insur-ance—actually can buy much more choice with rela-tively little money by moving to the individual market.They do not get much reward for going along with thegroup. So why not embrace the individual market, inprinciple, at least? As we will see, it will also offer im-portant risk protection that small-group coverage islacking.

There is an important policy divide here. If you thinkthat almost everyone (properly informed) would trulywant about the same kind of health insurance—similarin both extent of out-of-pocket payment and strictnessof managed care policies—then the case for the individ-ual market, or any market, is greatly weakened. Even animperfect government should be able to figure out whatalmost everyone wants if they all want the same thing,and provide it at low administrative cost. The case formarkets rests on the hypothesis of—and the evidencefor—variation in what people want from their healthinsurance. The current pattern in which workers, givena choice, do sort themselves across a range of policiesfrom generous to frugal, and from permissive to restric-tive, does strongly suggest that, at least in the UnitedStates, there is substantial preference variation. (Thereis also evidence of variation from some other countries,such as Switzerland; despite what politicians favoringsolidarity may say, people are different.) Similar varia-tion across the Medicare population—in the amountand type of Medigap chosen, or in the form of the

PAGE 39

39

................. 17631$ $CH1 12-17-09 12:22:04 PS

M A R K V. PAU LY

Part D plan, likewise suggests different preferences.Markets, not government, are best at providing differ-ent things to different people.

Another advantage of the individual market is porta-bility. As the person changes jobs, takes off from workfor a period of time, or moves to self-employment, he orshe can retain the same coverage at the same price. Withgroup insurance, even after the passage of the HealthInsurance Portability and Accountability Act (HIPAA),a person who changes jobs often must change bene-fits—and learn a whole new system. There is also thepossibility that the person will not take a job at whichhe could be more productive because he does not wantto change benefits. This is in part the notorious ‘‘joblock,’’ which the Clinton plan for using local health in-surance purchasing cooperatives was supposed to allevi-ate, but it would not be a problem in the first place ifworkers had their insurance as individuals.

These first two advantages of today’s individual in-surance are fairly obvious. The third advantage is lessobvious. One of the threats to a family’s economic aswell as physical health is the unexpected onset of a high-cost chronic condition. Even with good insurance tocushion the often high explicit costs associated with firstsymptoms and diagnosis, the person faces the futureprospect (until going on Medicare at age 65 or becom-ing so incapacitated as to be disabled) of being a highrisk, someone insurers reasonably expect to incurabove-average medical costs under any given policy. In

PAGE 40

40

................. 17631$ $CH1 12-17-09 12:22:05 PS


an ordinary insurance market, a high-cost insuranceproduct, without a subsidy, would have to sell at a highprice, and so consumers rightly fear the compoundingof their bad luck as poor future health is accompaniedby very high future insurance premiums, incompletecoverage, or no coverage at all.

This dire prospect is a real threat for all healthyAmericans, and anecdotes abound of the misery experi-enced even by middle-class people (uninsured or withgroup insurance they lose) who unexpectedly becomehigh risks. The great bulk of individual insurance, evenwithout further reforms, offers good protection againstthis fate for people who were able and willing to antici-pate its possibility and buy individual insurance beforethey got sick. In particular, except for bridging insur-ance explicitly labeled as temporary, all individual in-surance now must contain a provision for ‘‘guaranteedrenewability at class average rates.’’3 That is, the insurerpromises to renew coverage each year if the person paysthe new premiums, for everyone who becomes a highrisk since initially buying, and it promises that the newpremiums will not be increased selectively because ofthe onset of this condition (or for any other reason). Pre-miums can be increased, but they must be increased thesame for all in a given rating or risk class. This feature

3. President Obama’s comment that his health plan will preventinsurers from ‘‘raising your premiums or canceling your coverage’’thus does not differ from the individual market.

PAGE 41

41

................. 17631$ $CH1 12-17-09 12:22:05 PS

M A R K V. PAU LY

solves the problem of ‘‘reclassification risk’’ (Pauly, Ku-nreuther, and Hirth 1995; Cochrane 1995).

This provision was present in most individual insur-ance policies for many years (since it avoids the cost of‘‘re-underwriting’’) before any state or federal regula-tion and is now required as part of state regulation bythe federal HIPAA law (Patel and Pauly 2002). Whilethe data for the pre-HIPAA period are imprecise andcombine health and disability insurance, it is estimatedthat 80% of individual insurance contained guaranteedrenewability provisions even before the federal law andits effect on state laws (Pauly, Percy, and Herring 1999).Thus, arguments in the health reform debate that, afterreform, insurers will not be allowed to ‘‘drop you if youget sick’’ are largely beside the point for individual in-surance since federal regulation already forbids such be-havior.4

That guaranteed renewability is a real commitmentfrom an insurer’s point of view is supported by its treat-ment in the actuarial literature. That literature empha-sizes that the premiums for health insurance withguaranteed renewability have to be larger than thosejust needed to cover current-period expected expensesin order to accumulate funds to pay higher costs of in-sureds who become high risks in future periods. Gener-ally, the actuary will set up a contract reserve to cover

4. Another potential benefit from guaranteed renewability isthat it gives strong incentives to the insurer to keep their insuredshealthy and thus avoid costly chronic conditions.

PAGE 42

42

................. 17631$ $CH1 12-17-09 12:22:05 PS


this commitment. Actuaries note that arbitrary actionswhich would force out high risks would deprive them oftheir ‘‘premium equity’’ in the policy.

There is a debate in the actuarial literature about thepossibility that insurers should or could act in ways thatundermine the guarantee (Bluhm 2006, 2007). The con-troversy over durational rating, where the time since thecontract was first written defines a rating class (andoften leads to an increase in premiums for contracts oflonger duration), is about whether such rating is consis-tent with guaranteed renewability or undermines it. Thecontroversy is not settled, but actuaries do generally ad-vise that insurers limit themselves to cancellations orrate increases strictly on a class underwriting basis, soat least part of the guaranteed renewability promise notto single out high risks for high premiums is preserved.Actuaries apparently are divided on whether insurerscould or should use more strategic activities to shedhigher risks.

While there are some ways an insurer willing to tar-nish its reputation or abandon its individual businesscan avoid the limits imposed by guaranteed renewabilityprovisions, this behavior appears to be uncommon. Thebest evidence that this good behavior actually occurs ona large scale is from data on the relationship betweenpremiums actually paid for individual insurance and therisk level of the person buying it (Herring and Pauly2007). Analysis of this data indicates that there is agreat deal of risk pooling—up to 60% for demographic

PAGE 43

43

................. 17631$ $CH1 12-17-09 12:22:05 PS

M A R K V. PAU LY

variation and up to 85% for chronic conditions, given aperson’s age, gender, family situation, and location.That is, only about 15% of the higher cost of a chroniccondition actually shows up in higher premiums paid bypeople with such conditions in the individual market;much of the variation in risk is not reflected in premi-ums. So, not only is individual insurance portable, onthe same terms and at the same premiums, across jobs,it is also similarly portable across health conditions. Thenet effect of such a provision is that higher risks in theoverall individual market are almost as likely as lowerrisks to end up with individual coverage—even withoutsubsidies and even without rate regulation. Additionalwork by Bundorf, Herring, and Pauly (forthcoming)that looked only at employed people and their familiesfound that high risks were actually more likely to havecoverage.

There is also other convincing and consistent empiri-cal evidence. Individual insurance premiums are front-loaded, as the theory predicts; they are higher relative toexpected expenses for young buyers than for older ones(Pauly and Herring 1999). And the age profile of indi-vidual premiums actually paid is consistent with thepath theory would suggest (Pauly and Herring 2006).Despite anecdotes and worries about insurers reneging,aggregate data on the individual insurance market looksmuch as it should if guaranteed renewability is working.

A key point is that group insurance does not at pres-

PAGE 44

44

................. 17631$ $CH1 12-17-09 12:22:06 PS


ent have the same kind of protection against risk reclas-sification as does individual insurance. As long as theperson stays in the group (keeps the job), both regula-tion and customary practice do prevent explicit grouppremiums from varying with risk. However, there areno insurance regulations that prevent money wagesfrom reflecting risk to some extent, and there is someevidence that wages are lower for those with high-riskconditions, such as obesity, or who are older (Sheiner1999; Pauly and Herring 1999). Even more to the point,the high-risk person who leaves employment at the in-sured job loses protection against re-underwriting andrisk rating. Indeed, research shows that workers in poorhealth who were covered by small-group insurance inone year are nearly twice as likely to be uninsured in thenext year than were workers with similar health statuswho had individual coverage (Pauly and Lieberthal2008). Ironically, when a high-risk person loses groupcoverage and then seeks individual insurance only to betreated as a high risk, it is the individual insurer who isoften excoriated for proposing to charge a high-riskrated and underwritten premium, even though it wasgroup insurance, with no guarantee of renewability, thatlured the person to buy insurance that then plunged himinto this predicament. Exploring the possibility of add-ing guaranteed renewability to group insurance for theindividual worker, not for the employer or group, seemshighly desirable. However, employers who offer health

PAGE 45

45

................. 17631$ $CH1 12-17-09 12:22:06 PS

M A R K V. PAU LY

benefits for the purpose of attracting and retaining goodworkers may not be so eager to have such insurance eas-ily portable across jobs.

How not to critique the individualhealth insurance market

Analysts at the Commonwealth Fund recently publisheda report that concluded that ‘‘the individual market inits current form does not provide a viable alternativeto employer-based group coverage’’ (2009, 2). Eventhough the study does correctly point out some prob-lems with the individual market, the bulk of its analysisis irrelevant, and its conclusion is incorrect. Reform isneeded in the individual market, but this kind of workis not helpful.

The analysis is based on a 2007 survey funded by thefoundation that surveyed people who were not currentlyenrolled in group insurance. The survey simply askedpeople if they ever ‘‘tried to buy’’ individual insurance,without defining ‘‘try’’ or asking about the circum-stances. In the sample of about 1,600 respondents, athird said they ‘‘tried to buy’’ individual insurance and atotal of 130 did end up with individual coverage. Whathappened to those who ‘‘tried’’ at some point but didnot buy individual coverage is not shown. Some mayhave remained uninsured, but surely many others ob-tained group coverage or went on Medicaid.

PAGE 46

46

................. 17631$ $CH1 12-17-09 12:22:06 PS


The reasons of high price and incomplete coverage—the main reasons given by people who ‘‘tried’’ but didnot buy—make perfect sense. Because it has lower ad-ministrative cost and commands a tax subsidy, groupinsurance is cheaper than individual insurance forthe great bulk of the population. And because it ischeaper—artificially so, because of the tax subsidy—people buy more of it.5 Thus the survey’s main findingthat people with individual insurance pay more and getless than people with group insurance is no surprise.Criticizing individual coverage in this context is likecriticizing Cinderella for poor fashion sense: given themore adverse circumstances in the individual market,the people who cannot find a group alternative do thebest they can, but that outcome has to be inferior togroup insurance. It is surely possible to find an individ-ual product with low cost-sharing and highly permissivemanaged care, but it is so expensive that people gener-ally do not choose it.

So the poor relative performance of this market is un-derstandable. The study’s conclusion is not. The conclu-sion is that ‘‘the individual market in its current formdoes not provide a viable alternative to employer-basedcoverage.’’ That is true enough for people who have anemployer-based alternative, but for the great bulk ofpeople in the individual market, there is no such alterna-

5. Group insurance premiums can be excluded from workers’taxable incomes and thus receive a tax subsidy that is generally notavailable to those who buy individual coverage.

PAGE 47

47

................. 17631$ $CH1 12-17-09 12:22:06 PS

M A R K V. PAU LY

tive that they find attractive. (In principle, anyone couldtake a job at a company providing group coverage, suchas Starbucks, but that may not be the preferred employ-ment option for most.) For these people, their only prac-tical option is individual coverage or no coverage. Forthem, as long as their income is not at the poverty level,at least some type of individual policy (even if only cata-strophic coverage with a high deductible but a low pre-mium) is a viable option. I would agree with anotherconclusion of the study—that the individual market ‘‘isinadequate as a source of affordable health insurancecoverage for those without access to employer-basedcoverage,’’ if ‘‘inadequate’’ refers to the ‘‘high unsubsi-dized price’’ (Commonwealth Fund 2009, 9). But thenthe solution should be an effort to get the administrativecost down and subsidies equalized so this poor stepsisterin the insurance market can at least make a reasonableshowing at the ball.

One other conclusion of the study—that those withhealth problems were ‘‘least likely’’ to ‘‘find an indi-vidual plan that meets their needs’’ is misleading, sinceother data analyses using larger samples of better datashow that higher risks are often as likely or morelikely to get individual coverage than lower risks(though, of course, they might have preferred muchmore generous coverage than they chose given theprice they faced). Finally, the study clearly shows thatoverinsurance—a large number of people choosing

PAGE 48

48

................. 17631$ $CH1 12-17-09 12:22:07 PS


subsidized coverage with zero deductibles and mini-mal cost-sharing—is a serious problem with em-ployer-based coverage, and to a much greater extentthan with individual coverage. Nearly a third of peo-ple had coverage with zero deductibles.6

The high proportion of disappointed potential buyersin the Commonwealth study is not a consistent finding.A recent study by America’s Health Insurance Plans(AHIP) of nearly two million applicants for individualinsurance showed that only 10% were excluded fromcoverage by medical underwriting and that, of those of-fered coverage, more people were offered discountsfrom standard coverage risk rates that were uprated(2007). The 2007 Commonwealth study in contrastasked only about those charged more, not about thosecharged less. That study did not say what proportion ofthose who got as far as underwriting then accepted theoffered coverage, nor what proportion of the potentialcustomer base got as far as applying to be underwritten.An earlier case study by the National Association ofHealth Underwriters (NAHU) found that almost everypotential buyer even with a chronic condition was even-tually able to get an insurance offer if they persisted insearching among companies despite being rejected atfirst application (2002).

6. The Commonwealth study found that 36% of those who‘‘tried to buy’’ were either rejected for coverage or charged a higherpremium.

PAGE 49

49

................. 17631$ $CH1 12-17-09 12:22:07 PS

M A R K V. PAU LY

Lowering administrative costs

Individual insurance is expensive for a reason: its ad-ministrative costs per insured person are high. Only arevolutionary shift to a single-payer system or futuristicmethods for electronic claims processing could affectthis cost, and that is not likely to happen nationwideany time soon. Moreover, the often-quoted 2% to 3%figure for administrative costs for the conventionalMedicare coverage is necessarily lower than the cost ofclaim processing in a private market, because (amongother reasons) private insurers do not have Medicare’sability to convert questionable provider reimbursementbehavior into a federal offense, and because Medicare’sresearch, development, and lobbying costs are notcharged off as insurer administrative expenses. Costs forselling, underwriting, billing, and general administra-tion that vary substantially across insurance types arealso important. The loading percentage is also loweredby the substantially larger dollar volume of Medicareclaims per beneficiary.

So let us take it as given that the cost of individualinsurance will be relatively high but that the uninsuredpeople we want to reach will be making individual vol-untary choices whether or not to take insurance andthat at least initially they will be charged the full pricefor that insurance. If individual insurance is the onlyfeasible game in town, is there any way to lower the costof reaching them?

PAGE 50

50

................. 17631$ $CH1 12-17-09 12:22:07 PS


Buying groups and exchanges: Howmany Chihuahuas equal a Great Dane?

There have been many attempts to obtain lower admin-istrative costs for individuals and small employers bycreating larger buying groups or exchanges. It is crucialto note that participation in these groups, as well as thedecision to offer insurance, had (before the advent ofthe Massachusetts Connector) been voluntary. Theseentities contract with a subset of all possible insurers,communicate information about premiums, and oftenthen help to collect premiums from people who agree toget their coverage in this way.

The key elements of an exchange, according to Kings-dale and Bertko (2009), are that it is organized by gov-ernment and that it is separate from alternative ways (ifpermitted) in which people can get individual or small-group insurance. Within these broad limits, there aremodels of varying degrees of exchange intrusiveness:from functioning just to list premium and policy charac-teristics for all licensed insurers in a market, throughvoluntary buying groups in which firms participating inthe exchange have advantages in getting the buyinggroup’s business, to having the exchange itself bargainwith insurers, regulate coverage and payment methods,and set premiums.

Voluntary buying groups have the longest record. Ido not believe there is a single example of permanentsuccess with voluntary buying groups here. The reason

PAGE 51

51

................. 17631$ $CH1 12-17-09 12:22:07 PS

M A R K V. PAU LY

for failure is ominous and obvious: whatever the sizeand configuration of the voluntary entity of which thebuyer becomes a part, the buyer still must choose indi-vidually or as a small firm to belong to and operatethrough the large group in the exchange. The buyer (in-dividual or small firm) does not shed skepticism, iner-tia, and apprehension just by joining or participatingin this entity; he or she will be subject to the same kindof selling job an individual customer would be subjectto. Once snared, the member might be handled atlower cost in an exchange or buying group (though thetwo-stage process may have duplicative costs of itsown), but even here there may still need to be individ-ual care and feeding and, compared to real group insur-ance, each person must explicitly pay the full premiumif there is no subsidy.

There are then few a priori reasons to expect substan-tial economies of scale in administrative costs from ex-changes. So then what might make voluntary ‘‘groupsof individuals’’ work better in the future? One approachmight involve new technology. There may be ways tolower the cost of searching for coverage, informingone’s self about the options, actually applying, and thenpaying premiums on a continuous basis. The advent ofelectronic methods for transacting creates some hopehere. Research suggests that online insurance purchas-ing has lowered administrative cost and increased com-petition for some other kinds of insurance, such as lifeinsurance.

PAGE 52

52

................. 17631$ $CH1 12-17-09 12:22:08 PS


Web-based health insurance offerings have emerged,and have had some success in terms of their volume,though little apparent effect on the total number of un-insured. There is some research on the potential advan-tages here, based on comparing information on actualpremiums paid in the pre-e-insurance period with thepremiums quoted on websites. That information issomewhat mixed. Under reasonable assumptions aboutthe cost trend, it seems that the pre-Web premiums peo-ple actually paid were lower than the average premiumsquoted on the Web, especially for older and presumablyhigher risks (Pauly, Herring, and Song 2006). So justgoing to a website, selecting an insurer at random, andcontemplating buying is not going to yield a premiumappreciably below that yielded by the traditionalprocess.

Upon reflection, this may be not much of a surprise.Before the web, people used brokers to search and,though the brokers required commissions, buyers werebenefited by being channeled to insurers with loweroverall prices, based on the broker’s previous highervolume of searches. This alternative intensive searchprocess would have been most rational for higher riskswho had more to lose from overpaying for insurance,and indeed the research suggests that the average Webpremiums were a worse deal for older risks.

A somewhat more optimistic outcome is possible ifwe assume that people would choose not the averagepremium plan on the Web but the lowest (or at least

PAGE 53

53

................. 17631$ $CH1 12-17-09 12:22:08 PS

M A R K V. PAU LY

a relatively low) quote. Then there is some evidence ofsavings, but the savings at best are quite modest. Proba-bly the reason is that the underwriting, and the ex-change of information, still needs to be done largelyoffline on a personal basis; the selling is not yet totally‘‘routinized,’’ like booking an airline ticket or payingcredit card bills online, perhaps for reasons we will ex-plore further below. Online insurance is still a boutiqueproduct; no all-electronic product has as yet emerged.

If the Web or neighborhood insurance broker cannotbe that much of a better deal on individual health insur-ance, can we expect that some type of quasi-group ar-rangement will do better? Here there is (as always) hopefor the future, but little evidence of past success. Ex-changes simply pass through administrative costs (suchas broker commissions), but they may allow buyers tomore easily determine which sellers have lower adminis-trative costs and lower target profits. Here again, wehave hopes, but not evidence. By encouraging or requir-ing insurers to post premiums in a common and orga-nized setting, some of the troublesome price dispersionalready noted might be mitigated. It is less obvious thataverage administrative costs will be lowered.

Massachusetts set up the Connector as an exchange-type vehicle for making coverage available to peoplewho sought insurance as members of small employ-ment-based groups, or as individuals, as a setting forenforcing rules about how such coverage could be sold,and as an institutional arrangement for subsidizing cov-

PAGE 54

54

................. 17631$ $CH1 12-17-09 12:22:08 PS


erage for those judged unable to afford insurance premi-ums. Has this arrangement improved the individualinsurance market? The main finding here is that premi-ums in this market are judged to be significantly lowerthan the former average premium for individual insur-ance in Massachusetts. Of course, Massachusetts wasone of the most expensive states in the country for indi-vidual insurance before the creation of the exchange,due to a combination of high medical expenses and ad-verse selection resulting from community rating in itsindividual market.

The evidence is not yet complete that would allow usto divide up the causes, but part of the reason for thelower premiums in the exchange is that its benchmarkaffordable plan involved higher cost-sharing than hadpreviously prevailed; an amusing story in the Boston pa-pers explained how the Blue Cross plan reacted to a de-mand from the group planning health reform for alower premium by finding a way to cut the pre-mium—by cutting coverage! Somewhat more specula-tively, the combination of new subsidies and new penal-ties for noncoverage probably caused some low risks toseek coverage, thus lowering the average community-rated premium. It is possible that this lower premiumis an example of a kind of low-level equilibrium underadverse selection.

The exchange almost surely provided benefit in someother ways. It did make it easier for people to shopacross insurers for a better deal (although the effect of

PAGE 55

55

................. 17631$ $CH1 12-17-09 12:22:08 PS

M A R K V. PAU LY

this feature on average premiums paid is hard to sepa-rate out). It probably was associated with lower admin-istrative costs, but these cost savings almost surely werecaused by the subsidy/mandate combination promptingpeople to seek coverage, not by economies of adminis-tration, per se. In fact, broker commissions continued tobe permitted in the exchange. Another potentially usefulfeature of the Massachusetts plan was a requirementthat employers set up Section 125 (‘‘cafeteria’’) planseven if there was no employer contribution, thus allow-ing all workers to get a tax exclusion for worker pre-mium contributions, but apparently no one has usedthis path.

The best part of the idea of an exchange is to reducebuyer search costs by standardizing the most popularpolicy options and bringing together a number of insur-ers to propose premiums on the exchange. However,this approach has so far only been used in states that donot allow explicit underwriting of coverage; it would bemore challenging to offer better information on pricesfor standardized coverage when insurers are also adjust-ing the risk levels they accept to the premiums theycharge. And we do not really know whether on balancebuyers ended up with better deals—only that, whateverthey paid, at least it came from an apparently tidymarket.

The other potential benefit from an exchange is a ben-efit of individual insurance now: insurance obtained in

PAGE 56

56

................. 17631$ $CH1 12-17-09 12:22:08 PS


the exchange would be portable across jobs. The mainnovelty here is that very-small-group insurance essen-tially is transformed into individual insurance. Since Ihave argued that individual insurance does have somepositive features, I regard this change as one for the bet-ter. One question is whether the tax and possibly someof the other advantages of group insurance can still bereaped in a setting where the insurance itself is ownedand paid for by the individual, not the group or the em-ployer. We still do not know the answer to this question.

Exchanges now being seriously discussed (or recentlycreated) have not been mild attempts to improve marketfunction; instead, some versions, including that used byMassachusetts, have come with a heavy overburden ofpremium and product regulation. It is this propensityfor dysfunctional regulation that sours the concept ofan exchange. The primary regulatory goal is communityrating, and the consequences of attempts to enforcecommunity-rating rules can be adverse.

As already noted, the main downside is that, becauseof a rule requiring insurers to charge the same premiumfor the same policy to all, incentives for adverse selec-tion and cream-skimming are created where they didnot need to exist. Any opportunity for choice across in-surance policies then raises the opportunity for adverseselection, which undoes risk transfers and may destroythe entire market. To attempt to prevent this, regulatorsare inexorably led to rule out offering lower-coverage

PAGE 57

57

................. 17631$ $CH1 12-17-09 12:22:09 PS

M A R K V. PAU LY

policies because those policies might pick off the lowrisks. Never mind that they might also be good for aver-age risks who are only moderately risk averse but veryconcerned about moral hazard, and never mind that in-novation overall may be stifled. At least low risks arenot avoiding their obligation to make transfers to highrisks. As one member of the Massachusetts board said,‘‘allowing people to pick and choose . . . underminesbroad spreading of risk’’ (Turnbull 2009). At present,high-deductible plans are included in the exchange, butthe future for this concession to Republicans cannot beregarded as very bright.

Finally, regulators are human. They develop affectionfor some policy designs and abhor others. Try as theymight, they fall prey to a soft paternalism in which theywant to channel people toward particular policy de-signs. These sentiments creep into the rules. The argu-ment that people have too many choices, that optionsshould be standardized yet further to allow more effi-cient shopping, feeds into these urges.

Does this kind of heavy-handed regulation actuallyoccur? According to Charlie Baker, it has occurred inMassachusetts. The state had to define ‘‘minimum cred-itable coverage,’’ and in so doing it both retained everymandated benefit that had previously been lobbied intoexistence and added a requirement for outpatient drugcoverage. It also imposed maxima on deductibles($2,000 for individuals, no matter what assets they heldor what attitude they had toward risk) and out-of-

PAGE 58

58

................. 17631$ $CH1 12-17-09 12:22:09 PS


pocket payments ($5,000 for an individual). The state isalso proposing to mandate new bundled ways to payfor health care that are both revolutionary and untried,although that policy is probably not the consequence ofthe exchange mechanism but rather is borne of quietdesperation as medical spending, already much aboveaverage in Massachusetts, has continued to grow evenas the uninsured are covered and supposedly are nowgetting the needed preventive care that saves on healthcare costs.

The problem here is not that these rules themselvesraised costs directly, since (especially with the tax sub-sidy and other subsidies) most people want more gener-ous coverage than the minimum. But in economictheory even a relatively small minority of people withstrong incentives to be aggressive shoppers keeps downthe cost for all, and yet regulation may prevent themfrom playing this role.

On balance, does the Massachusetts exchange repre-sent a model for a national plan (even one that imaginesfederally required but state-run exchanges)? As alreadynoted, many questions still cannot be answered. Weknow that the overall Massachusetts plan reduced thefraction of uninsured in that state, and we know thatthe subsidies that were effective in doing this are in-creasingly straining the state’s budget, but we do notknow what would have happened without subsidies,specification of minimum qualified plans, penalties forno insurance, or requirements for a Section 125 offer-

PAGE 59

59

................. 17631$ $CH1 12-17-09 12:22:09 PS

M A R K V. PAU LY

ing, without an exchange. There is no evidence of anequally dramatic drop in individual insurance adminis-trative cost in Massachusetts beyond what subsidizationitself would have generated. To score savings for the ex-change per se the administrative cost percentage wouldhave to be pushed below 15% for individual insurance,and that has not yet happened. It is also important toremember that much of the regulation and organizationembodied in the exchange was already present in Mas-sachusetts, one of the distinct minority of states that hasfound rate regulation (and many other kinds of healthcare regulation) economically and politically feasible. Ina state with lower incomes, more minorities, and less ofa tradition of competent regulation, the exchange con-cept will be more severely challenged.

Regulation and pricingwithout exchanges

In the absence of a formal exchange or an exchange-likearrangement by a dominant employer, how do peoplebuy small-group or individual insurance? In the statusquo, there is some regulation. As already noted, individ-ual coverage is required by current law to be guaranteedrenewable at class average rates, there are some guaran-teed issue provisions, and in a small number of statesthe variation in the premium charged to a new (not re-newing) insured is regulated with respect to limits on

PAGE 60

60

................. 17631$ $CH1 12-17-09 12:22:09 PS


variation with risk. In the small-group market there areno guaranteed renewability provisions applying to indi-vidual workers or to individual workers in groups, butpremium variation across groups is limited to some ex-tent in the great majority of states, usually with rating‘‘bands’’ (that is, premiums for any one group cannot bemore than twice the average). HIPAA rules do generallyrequire that people insured at one firm who switch jobsbe covered at the new firm (if it offers coverage) regard-less of risk, but this does not apply to some small firmsand, as noted, offers no protection in the individual mar-ket. When risk variation is limited, usually variation re-lated to demographics and location is allowed, but notvariation based on the prevalence of chronic conditionsor health status of a workforce. However, even in heavilyregulated states, premiums can vary over time based onpast group experience (for example, New York allowsexperience-rated groups); the rules apply to the initial of-fering and also depend to some extent on the size of thegroup, with larger groups having more ‘‘credibility’’ ap-plied to their claims experience. In some states, very smallgroups are lumped together with individual insurance.

Buyers approach unregulated markets in several poten-tially different ways. One option is to search online; thereare websites for individual insurance and small-groupplans, indicating both premiums and coverage. However,anyone seeking to buy what is on the website must applyfor coverage and be approved and underwritten—so alisting on a site is not really an unconditional offer to sell

PAGE 61

61

................. 17631$ $CH1 12-17-09 12:22:09 PS

M A R K V. PAU LY

at that price. Some individuals and many small employ-ers use brokers, middlemen who gather informationabout the prices and features of different plans, offer ad-vice to the buyers, and help guide the application andunderwriting process. If a broker arranges a transaction,the insurer pays an initial commission (and often a smallcommission at renewal). The payment system is similarto the way travel agents used to function, who were paidby the airlines rather than, as at present, getting an add-on to the seller’s price. Finally, buyers can approach themarket in a less organized way, contacting specific com-panies based on word-of-mouth information or namerecognition, with at best an informal search.

The evidence on efficiency of price search under exist-ing arrangements, as noted above, is mixed. There iswide variation in premiums posted or quoted by differ-ent insurance companies for apparently similar policies(Pollitz and Sorian 2001), but that is less relevant thaninformation about the variation in prices people actu-ally pay (Pauly, Herring, and Song 2006). Neither theuse of the internet nor the use of brokers appears to besufficient to squeeze out all transactions’ price variation(Sood et al. 2004).

The other unique dimension of transaction price vari-ation in this market is whether, at a given price, a selleris willing to underwrite a buyer. It is very clear that dif-ferent insurers target different risk segments of the mar-kets. Some target to sell only to the best or ‘‘cleanest’’risks, charging low premiums but rejecting many appli-

PAGE 62

62

................. 17631$ $CH1 12-17-09 12:22:10 PS


cants for what seem minor blemishes in their health re-cords (much to the irritation of the applicants andpolicymakers). But the rational strategy for the rejectedapplicant is to continue to search until he or she matcheswith an insurer whose premium/underwriting target fitsthat person’s risk profile. The evidence is strong thateven a relatively tarnished or high risk can get coverageif he or she persists in searching (NAHU 2002). Andsome of the observed variation in transactions premi-ums probably reflects this subtle variation across insur-ers in what they think predicts risk.

So in the end the individual market may not work allthat poorly—but it is both bothersome and maddeningto buyers. Brokers help, but their inherent conflict ofinterest (they get paid by the insurer, not the buyer)means that prices do not converge. Of course, if moreefficient searching were that valuable, one might imag-ine that other agents—financial advisors—wouldemerge to help, and some have. Still, this looks like aterrible way to run a railroad—or a market.

What is the fundamental problem here? Since techno-logical and organizational fixes for high administrativecosts are problematic, is there any other way to lowerthe cost of persuading people to buy insurance? Here weneed to go deeper and try to understand why such costlypersuasion is necessary. In terms of abstract economictheory, insurance is desirable because it allows risk-averse people to be better off, to have higher ‘‘expectedutility,’’ as they reduce the risk of a very high and finan-

PAGE 63

63

................. 17631$ $CH1 12-17-09 12:22:10 PS

M A R K V. PAU LY

cially crippling medical bill. But if insurance is as gooda thing as we think, why should people have to be talkedinto it? For many other products, selling costs are mod-est. I volunteer to go to the supermarket to get food fortonight’s dinner, even bearing some costs to get there; Ido not have to be persuaded. To be sure, there are someproducts that famously have to be ‘‘sold’’: soft drinks orready-to-eat cereals are products with high selling costsas a proportion of sales. But neither of these are prod-ucts that fulfill a fundamental human need or desire. Ifwe know why (or whether) health insurance has to besold rather than bought, perhaps that can point the waytoward methods to reduce costs. The answer to thisquestion depends, in large part, on the premium chargedfor coverage relative to the benefits the insured expectsto collect, but it also depends on buyer information andinsurer transparency. We look at the price-availabilityissue first, and then move on to consider other factors.

Insurance company economics, sharppractices, premiums, and exclusions

Insurers vary in the premiums they charge and the poli-cies they follow in deciding to issue insurance to newpolicyholders. One often-criticized feature of insurance,especially individual insurance, is a preexisting-condi-tion exclusion. This feature applies only to uninsuredpeople seeking insurance or those who are newly in-

PAGE 64

64

................. 17631$ $CH1 12-17-09 12:22:10 PS


sured. As a condition of coverage, insurers may excludebenefit payment for some period of time for conditionsthe person could reasonably have known about beforecoverage or for care in a current episode of treatment.Thus, if I knew I had an allergy (usually because a doc-tor told me so and/or because I sought care for it), treat-ment might not be covered for a period generallyranging from nine months to two years. Or if I was inthe middle of treatment for some symptoms (even if Idid not yet have a diagnosis), future care in that courseof treatment would not be covered. The rationale forexclusion is that the purpose of insurance is to pay foruncertain events, not those that are already known toexist. In the extreme case, were there no such exclusionsallowed, people would not buy insurance until they gotsick, but then the premiums needed to cover their costswould be very high because they could not be averagedover people who did not get sick. Obviously, a policywith no exclusions at all would be very expensive, butthere is choice among firms with a variety of differentexclusion policies. Different companies currently applythese exclusions with different degrees of strictness (andfor different periods of time). Less strict exclusion rulesare matched by higher premiums.

Why do people seek or accept policies that have theseexclusions? The answer that insurers require them is notsatisfactory, since insurers will offer anything consum-ers are willing to pay for; consumer acceptance is alsorequired. An alternative strategy for insurers would be

PAGE 65

65

................. 17631$ $CH1 12-17-09 12:22:10 PS

M A R K V. PAU LY

to agree to cover everything, but to (upwardly) risk-ad-just premiums to deal with the presence of chronic con-ditions—and such adjustment does occur to a fairlysignificant extent for new insurance applicants. The ex-clusions are present as a secondary device to deal withrisk variation and attendant possible adverse selection;they reflect the practical idea that premiums cannot beadjusted for everything and insurers might not find itworthwhile to discover everything about risk in ad-vance. Despite their recent publicity, these exclusions af-fect very few people and often have modest effects whenthey do.

But they bother people a lot. There is a chance thatan exclusion exposes a person to a large bill. Perhapspart of the reason is that a person who initially chosea plan with stricter rules to save just a few dollars onpremiums is frustrated if those rules end up affectingthat person when the prior condition does flare upagain. It is the frustration of a gamble gone wrong.What also bothers people is having to argue with theinsurance company, should they get sick soon after get-ting new coverage. The point to be made here is that youcan reduce frustration and argument by paying more.Consumers may well have a hard time knowing whichcompanies follow which policies (especially if they donot use a knowledgeable independent agent or financialadvisor), and so there may be some case for defining astandard set of provisions. And, as those experts know,some companies, even some large insurers, have a repu-

PAGE 66

66

................. 17631$ $CH1 12-17-09 12:22:10 PS


tation for being more likely to engage in sharp practices.Everyone who applies or who files a claim after havingjust begun insurance coverage will be asked about thepossible preexisting nature of the illness, and everyonewho is asked will be irritated by the question. Finally,even companies with middle-of-the-road policies willfrom time to time be overly aggressive in individualcases. In effect, the exclusion causes people to suffer ad-ministrative costs and even lack coverage on a basis thatis doubly unpredictable—depending on whether theperson gets sick and what their company then does.

The problem is clear. We would rather avoid thisbothersome lottery, but doing so completely wouldmake insurance less attractive to most people most ofthe time (since most people at any point in time are notsick or suffering a chronic condition) and unaffordablewhen they did seek it. Trying to deal with what is pres-ently a very minor problem even in the individual mar-ket (though occasionally serious for the rare household)runs the real risk of creating a much more serious prob-lem of bad incentives for healthy people thinking of(not) buying insurance. A minor flaw could be turnedinto a catastrophe. And, as we shall see, there are betterways than prohibition to deal with the few high riskswho would be caught by exclusions.

So why are such forms of ‘‘insurance reform’’ appar-ently popular with policymakers? Anger at insurancecompanies appeals to the inner lawyer (or locker-roomliberal) in all of us, but not all exclusions should be pro-

PAGE 67

67

................. 17631$ $CH1 12-17-09 12:22:11 PS

M A R K V. PAU LY

hibited. There is a temptation to insurance fraud; pub-lic-interest radio spots in Philadelphia have beenwarning people about buying auto insurance and thenmaking claims for ‘‘pre-existing collisions.’’ Insurance iscomplex, and, having paid a premium, we all want tocollect as much as we can and are upset when insurersturn down even those claims we thought that policyprovisions made them unlikely to pay. Some compro-mise, and an adult attitude, is needed as long as somepeople are sometimes uninsured because insurance pur-chase is voluntary.

Here is how to think about the trade-off. If there wereno exclusions and no waiting periods, virtually all peo-ple who became high risks would probably end up withcoverage (as long as they could raise the money for thevery high insurance premiums), but virtually all lowrisks would decline that very expensive insurance. As weallow exclusions or make the time we have to wait toget insurance longer, some high-risk people and high-risk conditions will be dropped out but more low-riskpeople and their illnesses will be covered. What is theright balance?

The answer to that question depends in part on socialvalues: how much does it bother me for a low-risk per-son to be without coverage, and how much for a high-risk person? Since we drive away more low risks forevery high risk we bring in, even if we feel worse aboutuninsured high risks than low risks, we still have a judg-ment call to make. Another kind of judgment call con-

PAGE 68

68

................. 17631$ $CH1 12-17-09 12:22:11 PS


cerns exclusions versus waiting periods for coverage tobegin. The trade-off is that an exclusion exposes the per-son to risk for the designated condition but provides im-mediate coverage for all other services, while a waitingperiod denies coverage for everything for some periodof time. Usually the exclusion period is longer than thewaiting period—but the waiting period is worse regard-ing access to care.

Some combination of incentives to get insurance be-fore you get sick—exclusions not based on ‘‘gotcha’’clauses but properly targeted at things people shouldhave anticipated, plus relatively brief waiting periods—might provide the best balanced incentive for coverageand care. Insurers sometimes and insurance critics fre-quently fall prey to the notion that denying legitimateclaims, or slipping out of them in some way, is the pathto high long-run profits. But in reality insurers have noprinciples about what claims they will or will not pay;as long as the buyer is willing to pay a high enough pre-mium and the policy language can be clear, insurerswould be willing to pay for whatever buyers want—experimental treatment, ongoing illnesses, aromather-apy, and others. (Consider the Nationwide autoinsurance policy that promises not to raise your pre-mium if you have an accident; sounds great, but thatinsurance is therefore more expensive upfront thanstrictly experience-rated coverage.) The evidence is thatconsumers are not really willing to pay for soft-heartedbut expensive health insurance policies but then get

PAGE 69

69

................. 17631$ $CH1 12-17-09 12:22:11 PS

M A R K V. PAU LY

frustrated when the rules kick in. We definitely do notyet know enough to mandate any particular policy, soat least some flexibility—with prohibitions on the arbi-trary and unreasonable, probably limiting aggressive in-surer behavior more for lower-income people—mightbe the best way to go. Heavy-handed blanket prohibi-tions motivated more by political sound-bite appealthan by good sense would not be good.

The answer also depends on what else we do abouthigh risks. If there is an alternative for high risks to con-ventional coverage with exclusions (for example, some-what more expensive coverage in a high-risk pool), thenthere is less need to prohibit exclusions in conventionalcoverage. It is interesting that in unregulated markets,waiting periods are rare and brief (if coverage is de-layed, it is usually for only a month or so), but exclu-sions are common, though usually limited to two yearsor less. The market test would suggest that waiting peri-ods are inferior to exclusions, but the former probablybenefit higher risks somewhat more than the latter. Butin unregulated markets, high risks can often get com-plete coverage for all conditions right away if they arewilling to pay a high-risk premium, something pre-cluded under community rating. On the other hand, anuninsured person with a high-risk condition could besure of getting coverage under a waiting period (just bywaiting long enough), whereas the condition might bepermanently excluded under the exclusion model. Butthis almost never happens, except for rare high-cost

PAGE 70

70

................. 17631$ $CH1 12-17-09 12:22:11 PS


conditions (for example, hemophilia or cystic fibrosis)because exclusions have time limits as well. Still anothercompromise is to give buyers an option to take coverageat a given point in time but then charge them a higherpremium in the future if they decline their first chance.

Whether the deterrent to being an uninsured high riskis the possibility of an exclusion or a waiting period,the best arrangement would be for consumers to takecoverage beforehand with guaranteed renewability. Themore regulation requires insurers to be permissive in ac-cepting people with prior conditions, the less likely thisis to happen. There needs to be a trade-off here as well.Some combination of penalty for being uninsured andadditional penalty for being high-risk uninsured shouldbe part of the mix. The more effective these incentives,the less effort needs to be put into monitoring cream-skimming and limiting policy designs that appeal to lowrisks—precisely because coverage should appeal to lowrisks. It may well be that excluding conditions is not thebest way to offer an incentive to take coverage whenhealthy. What is important here, then, is not the princi-ple but the practice.

Subsidies

The simplest version of the insurance-decision processexplains why a subsidy, even a modest one, can behighly effective in stimulating coverage (an obvious re-

PAGE 71

71

................. 17631$ $CH1 12-17-09 12:22:11 PS

M A R K V. PAU LY

sult) and in lowering administrative costs (a less obviousresult) and in reducing risk rating, exclusions, or wait-ing periods (an even less obvious result). The fundamen-tal idea is simple: in deciding on insurance, the personcompares the insurance premium to the out-of-pocketexpenses expected to be incurred in the absence of insur-ance, estimating the latter (in the economics version ofthe story) by weighting each possible expenditure by thesubjective probability that it will happen. If this ex-pected value of the benefits from the insurance policy isless than (or even close to) the premium paid for thepolicy, the person will undoubtedly gain from buyingthe policy and will buy it. If the premium remains lowrelative to the expected value of benefits, the person willbe eager to pay the bill for the renewal premium as well.

The reason why many do not buy individual healthinsurance is that they believe (whether correctly or not)that the premium is high relative to their expected bene-fit. It is not that they do not think insurance is valuable,or even that they could not, if pressed, come up with thepremium; it is that they do not think it is a good dealfor what they have to pay for it.

Why might so many think this? For one thing, be-cause it is often true. The loading that typically prevailsin individual insurance is high, and there is no tax sub-sidy. So on average, people who buy individual healthinsurance to cover a given prospective set of possiblemedical expenses will spend a lot more than the average

PAGE 72

72

................. 17631$ $CH1 12-17-09 12:22:12 PS


expense of those who take a chance without insurance.Put slightly differently, compared to shelling out highpremiums for individual insurance, most people who re-main uninsured will be better off financially at the endof the year; the benefits they get back would be less thanthe premium they paid. Most people will not get verysick and will not make money off their insurance, andso they may regard it as a poor investment.

However, people who are risk-averse enough will bewilling to pay a premium that is high relative to benefitsrather than risk the chance of an expense much higherthan the premium. One does not buy insurance becauseit is a good investment; one buys it because it helps enor-mously in the unlikely event of being unlucky in health.(The level of loading that prevails in the individualhealth insurance market is similar to that in the individ-ual life insurance market, and the takeup rate in bothmarkets is similar, so we probably do not want to gotoo far in suggesting more novel theories for nonpur-chase. The loading is also similar to that in homeownersand automobile collision insurance, but in these casesthere are insurance requirements from lenders that, forpeople with a mortgage or who are paying off their carover time, assure a high-proportion purchasing.)

Still, there is some evidence that people underestimatetheir expected losses for health care; this is especiallytrue of younger people who think they will almost neverneed medical care. There is also some evidence that peo-

PAGE 73

73

................. 17631$ $CH1 12-17-09 12:22:12 PS

M A R K V. PAU LY

ple believe individual health insurance to be more ex-pensive than it really is and so never even bother tosearch.

Some aggressive strategies for convincing people thatthey are at risk, that health insurance is a reasonablygood value, and that they should be worried about risk,appear to be only moderately successful. Although someinsurers have pushed such plans (‘‘Tonik’’ in Califor-nia), most have not, and low risks remain uncovered.An alternative and more effective strategy to induce cov-erage is to lower the explicit price paid for insurance;that is, to offer a ‘‘subsidized’’ price and so make theloading be negative. Then the takeup of insurance is soattractive that few resources need to be expended toconvince people in the first place or to get them torenew. The subsidy could be paid by others (taxpayers,other insureds), or it might even be paid by recipientsin the form of a compulsory contribution to insurancepremiums. The key point is that if the explicit premiumthe buyer could avoid paying by declining coverage canbe made less than the full premium, buyers will not needcostly persuasion. The best way to persuade people tobuy insurance is to make it less expensive; even ‘‘youngimmortals’’ will buy at a low enough price.

Lower price is part of the secret of success in group(especially large group) insurance. The worker eligiblefor such coverage now is usually charged an explicit pre-mium that he could avoid by declining to take the groupinsurance. But this premium is only a small fraction

PAGE 74

74

................. 17631$ $CH1 12-17-09 12:22:12 PS


(20%–25%) of the total premium and therefore looksreasonable (even relative to the low individual premiumfor young workers); almost all workers therefore takeoffered coverage.

But workers pay for the other part of the premium byaccepting lower money wages (Pauly 1997). Why dothey agree to this kind of trick or precommitment,rather than waiting to decide on paying, or not paying,the full premium for coverage? Tax subsidies are onemajor part of the reason. But another part may be thatworkers trust the insurance through their job to be areasonably good buy most of the time. They anticipatethat they might sometimes make mistakes in evaluatinginsurance if they were to do so themselves. Precom-miting a major share of the premium so that it will berational to drop coverage only in the rare case when itsvalue drops below 25% of its premium may be a goodcompromise, one that avoids the need to revisit the in-surance-purchase decision every year.

What to hope for and what to expect

There is already voluntary private insurance with thepremium subsidized at the point of purchase thatmight tell us what would happen were individual in-surance markets subsidized. Specifically, large-groupinsurance achieves two of the three objectives posedabove. In particular, it has low loading costs, estimated

PAGE 75

75

................. 17631$ $CH1 12-17-09 12:22:12 PS

M A R K V. PAU LY

to be as low as 5% of premiums for very large groups.This means that little is wasted on pure insurance ad-ministration, and it compares favorably with adminis-trative cost percentages in Medicare and Medicaid. Itis also attractive enough because of the tax subsidy tohave a very high participation rate among eligibles, al-most always above 90%.

Here is another example of subsidized insurance andits loading cost: Chile compels workers to devote a por-tion of their payroll tax to health insurance but permitsthem to elect a private individual insurance option.These ‘‘ISAPRE’’ insurances have loadings of about halfthose of individual health insurance in the United States,at about 14% (in 2007). The reason is not that the Chil-eans are much more efficient at the insurance businessthan the Americans are; the reason, rather, is that inChile the insurance companies only have to guide thechoice of which kind of insurance to take, not make thehard sell that some kind of insurance is better than beingtotally uninsured.

In an historical example in the United States, regula-tion for a time forced the near-monopolistic Blue Crossplan of New Jersey to cross-subsidize the individualmarket; with this subsidy, the administrative cost of in-dividual health insurance was also only about 14% ofpremiums.

The point here is that a subsidy does more than in-crease the takeup at a given loading; it actually permitsa dramatic reduction in loading itself because it reduces

PAGE 76

76

................. 17631$ $CH1 12-17-09 12:22:12 PS


the cost of persuading people to take coverage. Subsi-dized insurance is more efficient with regard to adminis-trative cost because of a kind of self-fulfilling prophecy:by making premiums more attractive, selling costs canbe lowered, but at lower selling costs, attractive premi-ums can eventually become financially feasible.

This powerful point, which I believe is not well recog-nized either in the health policy or the insurance eco-nomics literature, has two important implications, onefairly obvious and the other less so. The obvious impli-cation is that with a subsidy, the specific administrativearrangement for insurance becomes both less importantand less of a factor in takeup. For example, the Massa-chusetts Connector may well have relatively low admin-istrative cost, but so would almost any other way ofadministering insurance given that state’s large subsidyto people with incomes below 300% of poverty and itspenalty for uninsurance to others. It is not the wholebundle of features of a high-performance (and high-reg-ulation) health system that lowers individual insuranceadministrative cost; the active ingredient is the subsidy.A system with only brokers or agents would also becheap, because they would not have as much work todo in persuading people to take coverage as they did inthe pre-subsidy market. The use of an online websitemight also work quite well.

The best policy (though not the one followed in Mas-sachusetts) might be to offer subsidies especially tar-geted at buyers with lower insurance demand, but then

PAGE 77

77

................. 17631$ $CH1 12-17-09 12:22:13 PS

M A R K V. PAU LY

let consumers choose in a neutral way how they want toobtain insurance. They might well want to use an insur-ance exchange, but they might actually prefer othermethods—for example, the direct use of brokers oragents who know them personally and can offer betteradvice to the inexperienced buyer than the neutral andimpersonal exchange. The Connector is analogized toa stock exchange, but many people who use the stockexchange use a broker or investment advisor, and thestock exchange itself provides no investment advice.The distinctions in reality between an insurance ex-change and the more typical individual market withbrokers and agents are easy to overstate; it is most likelythat most consumers will not obtain their insurancewithout assistance, just as many small investors do notuse low-service discount stockbrokers.

The less obvious point is that lowering price with asubsidy might even lower administrative costs so muchas to eliminate (or at least greatly reduce) the need for asubsidy. More formally, there may be a break-even equi-librium at a low insurance premium with high volume,but markets may currently be trapped in another equi-librium with high premiums and low volume. Table 4shows an example.

The first row roughly reproduces the current situationin the individual insurance market. The second rowshows a hypothetical increase in quantity demandedwhen explicit premiums are cut by a third.

If sellers of insurance are rational profit maximizers,

PAGE 78

78

................. 17631$ $CH1 12-17-09 12:22:13 PS


Total Loading Selling Cost Other Cost Subsidy % Buying

30% 22% 8% 0 25%

10% 7% 8% (–5%) 80%

TABLE 4How a Subsidy Might Lower Administrative Loading as a Percentage of Premiums

they have presumably chosen their current (high) levelof selling effort because, were they to reduce sellingcosts, commissions, and the like, the money they wouldsave by so doing is more than offset by the reduction innet revenues they would expect from selling insurance,as buyers would disappear; the current position of highcosts, low quantity, but high prices paid by those fewbuyers remaining in the market is a ‘‘local profit-max-imizing equilibrium.’’ Suppose selling costs were greatlyreduced, but price was also lowered considerably. Thatlower price might bring forth enough demand to coverthe remaining costs and to counteract the consequencesof reduced persuasion, thereby leading to enough reve-nue. We could imagine price being close to the actuari-ally fair premium plus the small (and constant-over-units) claims-processing costs. There might be manypeople who would seek out insurance at that low price,thus permitting insurers to cover their cost without evena subsidy, or without much of a subsidy.

The key parameters here are the response of quantity

PAGE 79

79

................. 17631$ $CH1 12-17-09 12:22:13 PS

M A R K V. PAU LY

demanded, given selling costs, to premium, and the re-sponse of quantity demanded, given premium, to sellingcosts—in elasticity terms, the price and selling-cost elas-ticities of demand. The well-known Dorfman-Steinertheorem states that if these two elasticities are given andconstant, the amount of selling expense will be greaterthe higher the selling cost elasticity relative to the priceelasticity (in absolute value). But it is possible that theselling cost elasticity is lower at low levels of selling costthan at high levels; if so, there may be more than onelocal equilibrium.

More concretely, jumping to a ‘‘mass marketed’’ in-surance plan, priced low and sold with low-cost sellingmethods, might be profitable (or require only a modestsubsidy)—even if a slightly lower-priced and a slightlyless intensively marketed plan than the current one isnot. This hope is buoyed by noting an important prece-dent: auto insurance. Both third-party and collision in-surance used to be sold primarily by the Americanagency system, using brokers and agents paid high com-missions and leading to high premiums and high load-ings. Then, first Allstate and, subsequently and moreaggressively, GEICO pioneered the ‘‘direct writer’’ sys-tem, in which agents were salaried or nonexistent; gen-eral advertising was substituted for individual housecalls; and toll-free phone numbers were used to commu-nicate with potential customers, quote premiums, andcomplete applications. No government had to create anexchange. The same phenomenon might apply to indi-

PAGE 80

80

................. 17631$ $CH1 12-17-09 12:22:13 PS


vidual health insurance, especially since a low premiummight draw into the market people less interested in in-tensive service from their insurance and more interestedin low price and value. Costco on the West Coast hasbeen experimenting with an in-store sales system similarto what Allstate used in its early, Sears-owned stage. Itoffers individual health insurance to ‘‘Costco ExecutiveMembers’’ through the Pacificare plan, payable by elec-tronic funds transfer from the insured’s bank account.

Economic theorists are concerned that adverse selec-tion might limit the size of the individual market, andpolicymakers are concerned about risk rating. In reality,individual insurance markets in states without premiumregulation do not appear to be in an adverse selectionequilibrium, selling insurance at high premiums to a fewhigh risks. Instead, the usual criticism is just the reverse:individual insurers without regulation sell only to veryhealthy low risks at what must be a relatively low pre-mium and moderately high loading, with some part ofloading representing the cost of underwriting to screenout higher risks. (The argument that community ratingwill reduce the net numbers of uninsured by making in-surance more affordable to high risks is, as alreadydiscussed above, just wrong.) The explicit cost of under-writing in individual insurance already appears to be asmall fraction of premiums; probably some of theagents’ commission is for informal screening and chan-neling as well. But if insurance were subsidized mod-estly, so that more lower risks were attracted into

PAGE 81

81

................. 17631$ $CH1 12-17-09 12:22:14 PS

M A R K V. PAU LY

insurance markets, it would not pay for insurers to un-derwrite as vigorously as they do now. The purpose ofcostly underwriting is to identify the high-risk appli-cants, but underwriting requires that every applicant bereviewed. It only makes sense to spend resources on acareful review if you think there is a large enough frac-tion of applicants you would want to decline. If subsid-ies enrich the pool of applicants with many more goodrisks, it becomes much less cost effective to screen, andso screening efforts will decline. It would also no longerpay as much for low risks to search aggressively for poli-cies with premiums tied to their risk levels, if high risksare not segmented out.

Regardless of the current situation, a modest subsidymight change things. It might permit firms to cover thecosts of the small number of high risks and still be at-tractive to the much larger number of moderate-to-lowrisks. And if insurers know that, at this modestly subsi-dized premium, their pool of applicants will contain avery small proportion of high risks and a very large pro-portion of eager moderate-to-low risks, it will not evenbe worthwhile to spend substantial resources to screenmany applicants just to identify a few high risks.

Guaranteed renewability also limits adverse selectionnow, and could be improved to virtually eliminate itover time. The reason why guaranteed renewability isadverse-selection-proof is that good risks are alwayscharged a premium they find attractive in a guaranteed-renewability contract. Hence, there is no way an outside

PAGE 82

82

................. 17631$ $CH1 12-17-09 12:22:14 PS


insurer can pick them off. Those who have become highrisks are obviously delighted to be in their guaranteed-renewability contract and so have little to gain fromslipping into other lower-risk pools. Of course, guaran-teed renewability is no solution for people who are al-ready high risk and uninsured. There would be abacklog of uninsured high risks to be cleared even ifguaranteed renewability is made universal. Guaranteedrenewability requires signing people up while they arestill healthy and so cannot work with people who havealready become high risks at a young age. But the cur-rent stock of uninsured truly high risks is actually quitesmall among young adults; the fraction of young peoplewith birth defects, chronic conditions, or permanent dis-abilities (as to physical health) is only about three per-cent of their age group. So, targeted programs (ratherthan market-wide regulation) are better for dealing withthis small minority.

For these initial high risks, some subsidized arrange-ment would be desirable. It should, as an interim step,take the form of a high-risk pool financed by generalrevenue taxation; it should not (for reasons discussedearlier) take the form of community rating, ratingbands, or prohibition of exclusions.

Let me share a brief comment on the role of high-riskpools in the past and in a reformed system. Such poolshave been properly criticized as a weak solution to theproblem of uninsured high risks, but they could be madeto work better and to function as a fallback in a system

PAGE 83

83

................. 17631$ $CH1 12-17-09 12:22:14 PS

M A R K V. PAU LY

that relies on universal guaranteed renewability. Cur-rent high-risk pools are not attractive because theycover only a small fraction of high risks, are often fi-nanced by taxes or assessments that increase insurancepremiums, and sometimes close to new enrollment. Butwith robust guaranteed renewability, there would bemany fewer high risks needing help from a pool; in ef-fect, their current small scale may fit the new environ-ment. High-risk pools are also criticized because they donot more generously subsidize low-income high risks,although most of those low-income households couldnot afford coverage even at low-risk premiums. Theproblem here is not the high-risk pool concept per se, oreven their management (which appears to be compe-tent), but rather their inadequate budgets. In ideal re-form, budgets would be adequate but, as noted, may notneed to be so large if guaranteed renewability reducesthe need for this arrangement since there would then befew uninsured high risks remaining.

Offering the right product

The third kind of potentially desirable change in the indi-vidual health insurance market is related to product de-sign. Given some level of administrative cost and somelevel of subsidy, the way to maximize the chance that aperson will choose to obtain insurance is to offer the per-son the insurance design that he or she likes best for the

PAGE 84

84

................. 17631$ $CH1 12-17-09 12:22:14 PS


money. This sturdy truism is often ignored; policymakersand policy analysts alike endlessly debate what is the bestkind of insurance product in their personal opinion, withpassionate and conflicting views on high-deductiblehealth plans, managed care, limited out-of-pocket pay-ment, and coverage of preventive care. Much of the rea-son for the argument is that policymakers have skipped astep: they are trying to specify what coverage consumersshould have to get them closer to the care they need. Themissing step, of course, is that ideal coverage is only use-ful if the person has the insurance to begin with, and isable and willing to buy it. The insurance plan that greatlyimproves one’s health, or maximizes one’s health for themoney, will not affect one’s actual health unless one iswilling to choose that insurance.

This plain fact suggests a less-than-obvious policy:given some actuarial value, the best way to maximizethe chance that a person will buy coverage, given a sub-sidy (if any), is the policy design that the person likesbest, and the best way to make this happen is to let peo-ple choose their own options with very light minimum-benefit requirements. The least constraining version ofthis plan would be to specify an average actuarial valueand a subsidy, and allow people to choose whateverinsurance they wish, within broad limits, that has an av-erage cost equal to or greater than that value. The low-est-coverage policy might be limited and less than idealin policymakers’ judgment, but at least it would be freeand therefore universal.

PAGE 85

85

................. 17631$ $CH1 12-17-09 12:22:15 PS

M A R K V. PAU LY

Even this simple process still leaves open the questionof specifying the minimum actuarial value and the sub-sidy. A program that permitted eligible policies to havepremium values very close to the subsidy (so that itwould be virtually free) should technically reduce thenumber of people with literally zero coverage to zero,for who would reject free insurance? But, some mightobject, such a program would do so only by paying forminimal (or ‘‘nonmeaningful’’) coverage. There needs tobe some social decision on the contours of basic cover-age, which will probably be different for people at dif-ferent income levels and possibly vary with othercharacteristics, such as risk, location, ethnicity, and gen-der. Efforts to identify coverage considered to satisfy thebasic minimum from a societal perspective have notbeen very successful but must be undertaken. The goalof getting at least some coverage, even if it is not perfect,to everyone is probably preferable to a goal of gettinggood coverage to only some while leaving others totallyunprotected.

There are two other useful policy roles, both gener-ated by the observation that consumers may not all beinformed well enough to accurately determine whatcoverage is indeed best for them. Prohibition of fraudu-lent or illusory coverage or, more generally, of policytypes that no informed person would rationally choose,is a logical extension of standard consumer-product-quality regulation. Less limiting but more sensible maybe making the choice of a reasonably good plan the de-

PAGE 86

86

................. 17631$ $CH1 12-17-09 12:22:15 PS


fault option or line of least resistance, to nudge peoplein a decent (even if not the absolute best) direction(Sunstein and Thaler 2008). Even here there can be am-biguities and trade-offs. Suppose an uninsured person isterrified of cancer and therefore foolishly attaches high-est value to a policy that paid only for cancer care. Itwould seem desirable to forbid or discourage this typeof coverage on the grounds that it is not rational, yetdoing so would involve overriding the citizen’s prefer-ences and potentially having the person decide not totake insurance at all. So the hand of regulation shouldbe light and graceful (if that is not too much to ask).

The other potential action is developing new andmore attractive insurance designs and making sure theyare available in the market. It would seem to be in thebusiness interest of insurance plans themselves to do so,and yet the strongest controversy in the policy spherehas been generated by opinions of third parties aboutwhich private plan is ‘‘best.’’ Policymakers, physicians,and politicians without expertise in insurance try to de-cide what insurance people should have—whetherHMO, HSA, or prevention-friendly—rather than lettingconsumers choose for themselves.

One awkward issue here is the dearth of reliable in-formation on whether insurance necessarily causes sig-nificantly improved health across the board. There aremany correlational studies—people who did not obtaininsurance have worse health than those who did—butthat correlation may only reflect risky behaviors, includ-

PAGE 87

87

................. 17631$ $CH1 12-17-09 12:22:15 PS

M A R K V. PAU LY

ing not buying insurance and but also not taking goodcare of one’s self, that distinguish those who did not getinsurance from other similar people who did. Studies ofthe withdrawal of Medicaid show that insurance im-proves the health of poor children and mothers. Wehave good evidence on the impact of different policieson medical care use and spending, showing that cover-age increases both—since sellers of insurance need toknow this in order to price their product. What is lack-ing is evidence (as opposed to opinion) on what policieshave large impacts on health for those not covered byMedicaid—the nonpoor and low-income males. Thereis a trade-off. Any given level of out-of-pocket spendingpotentially can (and should, if things are efficient) leadto somewhat lower health status than a policy withhigher cost and lower out-of-pocket spending; theslightly lower expected health status is chosen becauseit is offset by the large savings on total costs. But we donot have any good idea of exactly what degree of healthstatus is traded for how much savings. We cannot withconfidence recommend a policy design, for a given pre-mium and/or actuarial value of the total spending,which will lead to a given level of expected health, norcan we say what the target actuarial value should be.Controversy over adequacy and affordability is never-ending and inconclusive because the facts are lacking.

The controversy over the very best study we have—the Rand Health Insurance Experiment—illustrates thisproblem (Newhouse et al. 1993). In the experiment,

PAGE 88

88

................. 17631$ $CH1 12-17-09 12:22:16 PS


higher patient cost-sharing lowered both total spendingand the use of medical care that health professionalsjudged to be essential for health. But, to general sur-prise, there was no significant relationship between cost-sharing and realized health (except for vision correc-tion) for the great bulk of the study population, withhealth effects limited to the 6% or so of the populationat low income and high initial risk of high blood pres-sure. How could people who used less care judged to beessential for health end up with no lower levels ofhealth? Perhaps professional judgments are not valid, orperhaps there were offsetting consumer behaviors, butthe point is that the current evidence on the linkage be-tween coverage and health outcomes is tangential andimprecise. There are always ‘‘concerns’’ about effects ofcost-sharing on health, as in the recent evaluation ofconsumer-directed health plans (Marquis et al. 2006),but no evidence to judge the health-spending trade-off,and certainly no basis for prohibiting a policy designthat some citizens find attractive. I am concerned aboutmany decisions made by my fellow citizens, but thatdoes not mean that I should be able to affect what theydo with their own resources (barring evidence for self-harm or serious financial distress).

Data on the effectiveness of different types of care isstill quite limited and quite insufficient for coverage de-sign. One needs information on both the value of careand the responsiveness of consumers to coverage. Medi-cal evidence alone is not sufficient.

PAGE 89

89

................. 17631$ $CH1 12-17-09 12:22:16 PS

M A R K V. PAU LY

This absence of good information on the connectionbetween coverage and health suggests in the short runthat restrictions on coverage should be light and designshould be primarily targeted at attracting people tovoluntarily purchase any reasonable health insurancepolicy and less concerned with their choosing the (as-yet-unknown) precisely ‘‘right’’ level of coverage. Provi-sion of information on the health outcomes associatedwith different plans may eventually fill the knowledgegap and guide consumers to choose voluntarily the bestcoverage. We can more usefully worry about ‘‘underin-surance’’ when we get everyone to have some coverage.But in our current state of fairly complete ignorance, wedo not want to specify design limits prematurely.

The ideal and the feasible:compromises and mixes

One ideal version of health reform would rely almostentirely on subsidies for low-income average risks, pen-alties for high-income average risks, and guaranteed re-newability to ensure that the population is insured withthe right level of insurance. The vision here is one inwhich the 96% of the population which is good to aver-age risk when they reach the age at which they ceasebeing dependents would all be induced to take coveragewith guaranteed renewability at that point. With feasi-ble adjustments to guaranteed renewability to permit

PAGE 90

90

................. 17631$ $CH1 12-17-09 12:22:16 PS


people to move across policies when they have good rea-sons for doing so, all would start with insurance cover-age and would be perfectly protected against higherfuture costs should they become high risk through guar-anteed renewability—without side effects and withoutfurther compulsion. The tiny fraction of people withdisabilities or conditions at young ages could be pickedup by existing Medicaid or Medicare programs.

This approach is superior to the high-performancehealth system relentlessly advocated by some. However,while I believe that serious consideration should begiven to implementing this vision and have so arguedfor many years, some compromises may be necessary, atleast for a while until almost everyone has coverage withguaranteed renewability provisions. The most obviousproblem is that some people may not, despite our bestefforts and best subsidies and penalties, initially take theneeded coverage. Then they may become higher risk andneed help. Practicality suggests that there must be some-thing in place for this group, some way of limiting thehigh-risk premium for those people who, through badluck or bad decisions, somehow end up as sick and un-insured. There is a trade-off here: the more we cushionthe consequences of skipping the chance to buy ideal(guaranteed renewable) coverage, the more we distortincentives—both incentives to protect against becominga high risk and even incentives to becoming insured atall regardless of risk level.

After all, the motivation for consumers to be willing

PAGE 91

91

................. 17631$ $CH1 12-17-09 12:22:16 PS

M A R K V. PAU LY

to pay in advance for a market mechanism to protectagainst risk reclassification—guaranteed renewabil-ity—is the possibility of reclassification. If, at the ex-treme, health insurance were to be fully communityrated and issue is guaranteed without exclusion of cov-erage for services related to current health status, thereis no risk and therefore no motive to get insurance. Ifrating instead puts bounds on the premiums that canbe charged to high risks but allows somewhat higherpremiums for them, there will then be a motive for guar-anteed renewability, but that motive will be attenuated.Similarly, a high-risk pool subsidized by others alsoweakens the incentive to protect one’s self against be-coming a high risk. Depending on the parameters, thereis some combination of a safety net for people who fallthrough the cracks and a subsidy for low-income peopleat all risk levels that will maximize the proportion whochoose to be insured, but what is it?

A combination of guaranteed renewability require-ments and a subsidy to that coverage feature (as wellas the assumed subsidy to high risks) might lead to areasonable compromise and to a solution that wouldnot require strict regulation of low-coverage policiessince there should be lessened demand for such policiesin the presence of guaranteed renewability. Some peopleinevitably will show up as high risks without insurance,and telling them ‘‘you should have bought guaranteedrenewability coverage when you had the chance’’ mayseem too hard-hearted. The key point, however, is that

PAGE 92

92

................. 17631$ $CH1 12-17-09 12:22:17 PS


this feature can substantially protect against the fatalflaw of community rating. But what if we do not wantto bet everything on subsidized coverage with guaran-teed renewability?

There are two alternative strategies. One strategy pe-nalizes people for being uninsured and high risk buttempers the penalty from what it would be under fullrisk rating for new applicants. Both rating-band regula-tion that allows some limited risk rating and high-riskpools that charge high risks more in the pool (or offerthem less attractive coverage) are examples here. Forreasons discussed earlier, high-risk pools are superior toinsurance regulation as a way of providing coverage asa last resort. In these cases there is still value to guaran-teed renewability, but because incentives are dissipatedthere may need to be requirements for it. Note thatguaranteed renewability still has value, keeping peoplewho do buy coverage (before they become high risks)out of the high-risk pool or the upper band.

The other strategy is to have strict community rating—high risks are entitled to the same premium level as allothers—but penalize people for becoming high risk anduninsured by denying coverage, either through exclud-ing high-risk conditions or by putting waiting periods orannual one-time sign-up periods on access to coverage.A feature in some of the reform bills is to forbid exclu-sions for people who previously had qualified coveragebut permit them to some extent for people who chose tobe uninsured for a time. Despite the apparently greater

PAGE 93

93

................. 17631$ $CH1 12-17-09 12:22:17 PS

M A R K V. PAU LY

political appeal of these steps, especially the latter, theyactually are inferior in concept to risk rating, since theycompel people to remain uninsured, for some or all con-ditions, who might have been willing to pay enough toget coverage. That is, even though one might be willingto pay to sign up now as a high risk, under the annualenrollment model one must wait uninsured until thetime comes. In principle there could be exceptions thatwould permit people to sign up earlier or sign up forcoverage of all conditions by paying a penalty; then riskrating returns de facto and there is a value again to guar-anteed renewability. Even in this case it may be hard forsomeone to pay more to avoid being anti-cream-skimmed or to have the option to take a more limitedcoverage plan. The situation is bound to be confusingand uncomfortable for buyers and insurers alike.

There is a system that is better than community ratingand regulation prohibiting exclusions; there is a betterway to achieve the same goals. Here is a way to thinkabout what a reformed system might look like thatmakes some reasonable compromises but still relies onguaranteed renewability and insurer competition todrive costs and coverage. Think of the two parts of thepopulation: those with high enough incomes that sub-sidies beyond a minimal level are both unneeded andunjust, versus those with lower incomes who need helpto obtain insurance at any risk level.

The higher-income part of the population should beincentivized (through a mandate or penalty) to get cata-

PAGE 94

94

................. 17631$ $CH1 12-17-09 12:22:17 PS


strophic insurance with guaranteed renewability. Cover-age can be sought from any source (employment-basedgroup, exchange, individual market) the consumer findsattractive. All policies regardless of source would beguaranteed renewable at class average premiums for theindividual insured. Policy specifications should be clari-fied and made transparent, and unreasonable pre-existing-condition provisions—the kind no one wouldbuy if they were careful—should be excluded. Ex-changes might be made available as an option. Beyondthese minimal steps, the market could be left free tofunction for the bulk of the population.

For those for whom low income or large family sizeprompts a social desire for more generous subsidies,those subsidies should be offered in the form of vouch-ers, with qualified coverage specified as parsimoniouslyas possible (and also taking the form of catastrophiccoverage, but the out-of-pocket maximum described ascatastrophic linked to income). The same modest rulesabout guaranteed renewability and policy transparencywould also apply.

For the small fraction of people who somehow do nothave guaranteed renewability protection and who areabove-average risk, the value of the credit would be aug-mented somewhat but the generosity of the policywould be limited; the person who falls into this situationwould pay more and get less than if they had boughtand retained guaranteed renewable coverage. Insurerscould charge full risk-based premiums to this popula-

PAGE 95

95

................. 17631$ $CH1 12-17-09 12:22:18 PS

M A R K V. PAU LY

tion but the credit would be risk-adjusted to some ex-tent to cushion the impact on the portion of thepremium the consumer would have to pay.

If political pressure and/or administrative conve-nience dictated some limits on risk rating of premiums,those limits would be modest in amount and wouldapply to coverage for which only a small fraction of thepopulation—those who are high risks and withouthealth insurance—would be eligible. Similar modestlimits would be placed on any efforts to constrain exclu-sion provisions that consumers found attractive as away to hold down premiums and deal with adverseselection.

Conclusion

I have not considered all possible improvements in theindividual market here. The amount of even reasonablydefinitive information on the comparative performanceof these devices is currently quite low, and so there islittle benefit from going into great detail about what itis we do not know.

In my judgment, the policy that follows from thisfrank admission of imperfect information is one thatmight be sensible in any case: the optimal governmentintervention is one that provides appropriate subsidiesto those who need them, a basic but initially quite lim-ited regulatory framework, the possibility for a large

PAGE 96

96

................. 17631$ $CH1 12-17-09 12:22:18 PS


number of different qualified sources of insurance—for-profit and nonprofit, public and private—and then per-mits consumers to make their own choices in a neutralway. If higher-income people who remain uninsured arethought to be a policy problem (rather than as evidencethat not all rich people are smart), a penalty or mandatemight be considered. Should additional subsidies andregulations be required, they can more easily be addedwhen needed than withdrawn if not needed.

For good reasons, reform of the individual health in-surance market is on the front burner in today’s policykitchen. But the recipe for reform in currently proposedlegislation is far from ideal and stands a good chance ofmaking things worse rather than better. It would run areal risk of increasing the number of uninsured (givensubsidies offered), stifling innovation in health insur-ance at a time when it is most needed, and creating dys-functional incentives for both consumers and insurancecompanies. Reform which uses rather than abuses mar-ket forces would be preferable. Competitive markets dohave their strongest rationale in settings in which theamount of knowledge a planner has or could have isincomplete, and the individual health insurance marketis a prime candidate for such a setting. There are someserious trade-offs between different policy features, butthe model of open, neutral competitive markets seemsbest suited to having those trade-offs made as well asthey can be in the world in which we now live.

PAGE 97

97

................. 17631$ $CH1 12-17-09 12:22:18 PS

health reform without side effects - hoover institution...health reform without side effects making...

Documents