- Email: [email protected]
Adverse selection: Does it preclude a competitive health insurance market?
Journal
of Health
Economics
1I (1992) 3533356. North-Holland
Comment
Adverse selection: Does it preclude a competitive health insurance market? Frank Vanderbilt
A. Sloan* Unioersity,
Final version
received
Nashciile,
TN,
US.4
May 1992
The main argument of those who would preserve a pluralistic multipayer health insurance system, as it now exists in the US, is that it permits satisfaction of diverse preferences. Although an appreciable portion of the population currently lacks insurance, universal coverage could be achieved with a fixed dollar subsidy, leaving the choice of plan to the individual. The simulations presented by Marquis in the previous issue show adverse selection against high-option plans. To the extent that her results are valid, one should be pessimistic about proposals that rely on private insurance markets with individual choice. Data from the Rand Health Insurance Experiment (HIE) provide empirical evidence for the demand side of Marquis’ analysis. A unique feature of these data for analysis of adverse selection is that the HIE obtained predictions of future health care expenditures which could then be compared with actual expenditures for the same year. Not surprisingly, predictions of high future expenditures are in fact realized. The HIE obtained no information on insurer behavior. Thus, the supply of insurance side of her analysis is completely based on assumptions about insurer behavior. In her analysis, premiums may be based on past losses of policyholders. But except for knowing this and the policyholder’s gender, age, and a rough measure of family size, the insurer knows nothing about the policyholder. This sets the stage for adverse selection and its effects. Insurers need not be so passive in classifying risks, relying only on information elicited from contract terms and premiums, very rudimentary information on policyholder characteristics, and a policyholder’s prior experience with the insurer. For one, the insurer may obtain information about Correspondence fo; Professor Frank A. Sloan, Department sity, Box 1503. Station B, Nashville, TN 37235, USA. *The author wishes to thank Thomas Hoerger, Vanderbilt
Ol67-6296/92/$05.00
g-2 1992-Elsevier
Science Publishers
of Economics, University,
Vanderbilt
for his comments.
B.V. All rights reserved
Univer-
354
F.A. Sloan. Adoerse selection
the policyholder at enrollment or develop incentives to encourage the policyholder to reveal truthful information about his anticipated risk. An insurer could ask a few health questions of a policyholder as is requested of applicants for life insurance. Life insurance contracts commonly contain a clause permitting payments to the policyholder’s heirs to be denied for a specified period if it subsequently was revealed that the policyholder had withheld information from the insurer when the policy was purchased. Or an applicant for insurance could be asked to reveal the applicant’s own best estimate of his or her risk class by agreeing to pay a premium penalty later if actual loss experience is statistically inconsistent than the initial declaration [D’Arcy and Doherty (1990)]. Insurers might require an insurance physical. Insurers might require all applicants to furnish evidence of a renewal invitation by another insurer and only accept applicants with renewal notices, thus limiting the high risk person’s choice to either paying an increased premium to the existing insurer or forgoing insurance altogether. Information on individuals might be pooled as by the Medical Information Bureau for life insurance, state motor vehicle records for automobile insurance [Insurance Research Council (1991)], or the national data bank on medical malpractice payments as required by the Health Care Quality Improvement Act of 1986. Then an insurer would not be limited to information generated from the insurer’s own experience with a policyholder. If the insurer cannot elicit accurate information on expected loss any other way, a multiperiod pricing regime might evolve in which all new applicants pay the same premium initially. After good news is obtained on a policyholder, the premium falls but not to the actuarial value of the loss. Given that the individual with the clean health record cannot secure insurance for less than the initial period’s pooled premium elsewhere, the insurer reaps a quasi rent from its information advantage by charging a second period premium slightly below the pooled risk premium the individual would pay competing insurers [Kunreuther and Pauly (1985)]. Many other possibilities exist. Three questions follow from this laundry list of possibilities. First, why have not health insurers been more aggressive in classifying risks to date? Second, should one expect more of the same if the US were to implement a market-oriented universal health insurance scheme with fixed dollar subsidies of premiums? Third, if insurers became more aggressive, would this be a desirable outcome? An answer to the first question is that information is obtained at a cost, and, for most of the period since the 1930s when health insurance first evolved, the profit-maximizing amount of information gathering was not very large. But premiums have not often even reflected policyholder characteristics that are instantly observable, such as age, gender, and family size. With use of these variables, less weight is placed on past loss experience for purposes
F.A. Sloan, Adverse selection
355
of risk classification [Sloan and Hassan (1990)]. Total reliance on past claims to differentiate risks may lead to substantial increases in premiums following a particular adverse event. Yet use of demographic variables to classify risks may be ‘inadmissable’ [Abraham (1986)]. Another possibility is that most private health insurance has been provided through employment, and there may be costs to employers of subdividing their employees into relined risk categories. Insurance has been provided on this basis because of economies of scale of group provision, and savings from this source may partly or even completely have offset the cross subsidies of high risk employees within employment groups. From anecdotal accounts, it appears that group insurers and even employers are becoming more interested in classifying risks accurately than they once were [see, e.g., Freudenheim (1992) Kolata (1992)], probably because health insurance premiums are now much higher than previously. The complaints probably come from high risk persons required to pay actuarially fair rates than from healthy persons who now pay relatively lower premiums. In response to the second question, because the tax subsidy of insurance premiums would be eliminated under a fixed dollar subsidy plan, demand curves facing insurers would probably become more elastic. Thus, insurers would want to obtain more information on insureds. But would a situation in which individuals received a fixed dollar subsidy, paid the actuarial value of their losses in premiums, and insurers earned zero profits be socially acceptable? If anything, the plan would probably not be acceptable because insurers knew too much and acted on this information rather than because of asymmetric information of the type described by Marquis and others before her. Either it would be necessary to vary the amount of the subsidy according to individual expected losses, which would not be easy to do, or a risk pool would be established to guarantee insurance to high risk persons at less than actuarially fair rates. Risk pools are common in automobile insurance, albeit with sometimes mixed results [Grabowski et al. (1989)]. In sum, although fixed dollar subsidies have the great virtue of ferreting out cross subsidies, society may not be satisfied with the results. The scenario described by Marquis is only one of many. People seem to want lifetime insurance offering low premiums if things go bad rather than premiums that change annually as health outcomes are realized [see, e.g., Light (1992)]. But nondiversible risk may be too great for a market in life contracts to exist. References legal theory, and public policy (Yale Abraham, K.S., 1986, Distributing risk: Insurance. University Press, New Haven, CT). D’Arcy, S.P. and N.A. Doherty, 1990, Adverse selection, private information, and lowballing in insurance markets, Journal of Business 63, no. 2, 145-164.
356
F.A. Sloan, Adverse
selection
Freudenheim, M., 1992, Employers winning right to cut back medical insurance, New York Times, March 29, 1 and 14. Grabowski. H., W.K. Viscusi and W.N. Evans, 1989, Price and availability tradeoffs of automobile insurance regulation, Journal of Risk and Insurance 56, no. 2, 274299. Insurance Research Council, 1991, Adequacy of motor vehicle records in evaluating driver performance (IRC, Oak Brook, IL). Kolata, G., 1992, New insurance practice: Dividing sick from well, New York Times, March 4, Al and A7. Kunreuther, H. and M. Pauly, 1985, Market equilibrium with private knowledge, Journal of Public Economics 26, no. 3, 2699288. Light. D.W., 1992, The practice and ethics of risk-rated health insurance, Journal of the American Medical Association 267, no. 18, 2503-2508. Sloan, F.A. and M. Hassan, 1990, Equity and accuracy in medical malpractice insurance pricing, Journal of Health Economics 9, no. 3. 2899320.
of Health
Economics
1I (1992) 3533356. North-Holland
Comment
Adverse selection: Does it preclude a competitive health insurance market? Frank Vanderbilt
A. Sloan* Unioersity,
Final version
received
Nashciile,
TN,
US.4
May 1992
The main argument of those who would preserve a pluralistic multipayer health insurance system, as it now exists in the US, is that it permits satisfaction of diverse preferences. Although an appreciable portion of the population currently lacks insurance, universal coverage could be achieved with a fixed dollar subsidy, leaving the choice of plan to the individual. The simulations presented by Marquis in the previous issue show adverse selection against high-option plans. To the extent that her results are valid, one should be pessimistic about proposals that rely on private insurance markets with individual choice. Data from the Rand Health Insurance Experiment (HIE) provide empirical evidence for the demand side of Marquis’ analysis. A unique feature of these data for analysis of adverse selection is that the HIE obtained predictions of future health care expenditures which could then be compared with actual expenditures for the same year. Not surprisingly, predictions of high future expenditures are in fact realized. The HIE obtained no information on insurer behavior. Thus, the supply of insurance side of her analysis is completely based on assumptions about insurer behavior. In her analysis, premiums may be based on past losses of policyholders. But except for knowing this and the policyholder’s gender, age, and a rough measure of family size, the insurer knows nothing about the policyholder. This sets the stage for adverse selection and its effects. Insurers need not be so passive in classifying risks, relying only on information elicited from contract terms and premiums, very rudimentary information on policyholder characteristics, and a policyholder’s prior experience with the insurer. For one, the insurer may obtain information about Correspondence fo; Professor Frank A. Sloan, Department sity, Box 1503. Station B, Nashville, TN 37235, USA. *The author wishes to thank Thomas Hoerger, Vanderbilt
Ol67-6296/92/$05.00
g-2 1992-Elsevier
Science Publishers
of Economics, University,
Vanderbilt
for his comments.
B.V. All rights reserved
Univer-
354
F.A. Sloan. Adoerse selection
the policyholder at enrollment or develop incentives to encourage the policyholder to reveal truthful information about his anticipated risk. An insurer could ask a few health questions of a policyholder as is requested of applicants for life insurance. Life insurance contracts commonly contain a clause permitting payments to the policyholder’s heirs to be denied for a specified period if it subsequently was revealed that the policyholder had withheld information from the insurer when the policy was purchased. Or an applicant for insurance could be asked to reveal the applicant’s own best estimate of his or her risk class by agreeing to pay a premium penalty later if actual loss experience is statistically inconsistent than the initial declaration [D’Arcy and Doherty (1990)]. Insurers might require an insurance physical. Insurers might require all applicants to furnish evidence of a renewal invitation by another insurer and only accept applicants with renewal notices, thus limiting the high risk person’s choice to either paying an increased premium to the existing insurer or forgoing insurance altogether. Information on individuals might be pooled as by the Medical Information Bureau for life insurance, state motor vehicle records for automobile insurance [Insurance Research Council (1991)], or the national data bank on medical malpractice payments as required by the Health Care Quality Improvement Act of 1986. Then an insurer would not be limited to information generated from the insurer’s own experience with a policyholder. If the insurer cannot elicit accurate information on expected loss any other way, a multiperiod pricing regime might evolve in which all new applicants pay the same premium initially. After good news is obtained on a policyholder, the premium falls but not to the actuarial value of the loss. Given that the individual with the clean health record cannot secure insurance for less than the initial period’s pooled premium elsewhere, the insurer reaps a quasi rent from its information advantage by charging a second period premium slightly below the pooled risk premium the individual would pay competing insurers [Kunreuther and Pauly (1985)]. Many other possibilities exist. Three questions follow from this laundry list of possibilities. First, why have not health insurers been more aggressive in classifying risks to date? Second, should one expect more of the same if the US were to implement a market-oriented universal health insurance scheme with fixed dollar subsidies of premiums? Third, if insurers became more aggressive, would this be a desirable outcome? An answer to the first question is that information is obtained at a cost, and, for most of the period since the 1930s when health insurance first evolved, the profit-maximizing amount of information gathering was not very large. But premiums have not often even reflected policyholder characteristics that are instantly observable, such as age, gender, and family size. With use of these variables, less weight is placed on past loss experience for purposes
F.A. Sloan, Adverse selection
355
of risk classification [Sloan and Hassan (1990)]. Total reliance on past claims to differentiate risks may lead to substantial increases in premiums following a particular adverse event. Yet use of demographic variables to classify risks may be ‘inadmissable’ [Abraham (1986)]. Another possibility is that most private health insurance has been provided through employment, and there may be costs to employers of subdividing their employees into relined risk categories. Insurance has been provided on this basis because of economies of scale of group provision, and savings from this source may partly or even completely have offset the cross subsidies of high risk employees within employment groups. From anecdotal accounts, it appears that group insurers and even employers are becoming more interested in classifying risks accurately than they once were [see, e.g., Freudenheim (1992) Kolata (1992)], probably because health insurance premiums are now much higher than previously. The complaints probably come from high risk persons required to pay actuarially fair rates than from healthy persons who now pay relatively lower premiums. In response to the second question, because the tax subsidy of insurance premiums would be eliminated under a fixed dollar subsidy plan, demand curves facing insurers would probably become more elastic. Thus, insurers would want to obtain more information on insureds. But would a situation in which individuals received a fixed dollar subsidy, paid the actuarial value of their losses in premiums, and insurers earned zero profits be socially acceptable? If anything, the plan would probably not be acceptable because insurers knew too much and acted on this information rather than because of asymmetric information of the type described by Marquis and others before her. Either it would be necessary to vary the amount of the subsidy according to individual expected losses, which would not be easy to do, or a risk pool would be established to guarantee insurance to high risk persons at less than actuarially fair rates. Risk pools are common in automobile insurance, albeit with sometimes mixed results [Grabowski et al. (1989)]. In sum, although fixed dollar subsidies have the great virtue of ferreting out cross subsidies, society may not be satisfied with the results. The scenario described by Marquis is only one of many. People seem to want lifetime insurance offering low premiums if things go bad rather than premiums that change annually as health outcomes are realized [see, e.g., Light (1992)]. But nondiversible risk may be too great for a market in life contracts to exist. References legal theory, and public policy (Yale Abraham, K.S., 1986, Distributing risk: Insurance. University Press, New Haven, CT). D’Arcy, S.P. and N.A. Doherty, 1990, Adverse selection, private information, and lowballing in insurance markets, Journal of Business 63, no. 2, 145-164.
356
F.A. Sloan, Adverse
selection
Freudenheim, M., 1992, Employers winning right to cut back medical insurance, New York Times, March 29, 1 and 14. Grabowski. H., W.K. Viscusi and W.N. Evans, 1989, Price and availability tradeoffs of automobile insurance regulation, Journal of Risk and Insurance 56, no. 2, 274299. Insurance Research Council, 1991, Adequacy of motor vehicle records in evaluating driver performance (IRC, Oak Brook, IL). Kolata, G., 1992, New insurance practice: Dividing sick from well, New York Times, March 4, Al and A7. Kunreuther, H. and M. Pauly, 1985, Market equilibrium with private knowledge, Journal of Public Economics 26, no. 3, 2699288. Light. D.W., 1992, The practice and ethics of risk-rated health insurance, Journal of the American Medical Association 267, no. 18, 2503-2508. Sloan, F.A. and M. Hassan, 1990, Equity and accuracy in medical malpractice insurance pricing, Journal of Health Economics 9, no. 3. 2899320.