The subject I wish to address today is the basic underlying conception of the New Deal, though this conception, I would argue, extends well beyond the New Deal and constitutes the basic underlying conception of the modern liberal state both here in the United States and abroad. That conception is the role of the government as an insurer: namely, that an essential purpose of government is to provide aid to citizens after they have suffered some form of economic or personal loss. It is a conception accepted for some contexts by virtually all citizens, yet it is a conception that, to my mind regrettably, conservatives have never effectively addressed. Modern conservatism surely emphasizes that this conception of the role of government must have limits, limits based chiefly upon costs. In addition, a prominent criticism--led by Charles Murray of this great institution--shows some forms of social insurance--welfare, for example--to be dysfunctional, a point which I totally accept. But this criticism has not been generalized, and even as telling as this criticism is, it does not itself strike at the heart of this liberal conception of government. As a consequence, much of the conservative movement remains reactionary in the sense that, in the absence of an alternative, affirmative conception of government, conservatives are left in a position where they can only react--become obstacles--to efforts to extend and effectuate this liberal conception. What I am hopeful of accomplishing today is to show the need for a reanalysis of this conception of the role of government, and to briefly suggest an alternative to this liberal view.
I. The Conception of the Government as an Insurer: The New Deal to the Modern Welfare State
Although I have not seen historical accounts that put it in exactly this language, I would argue that the central unifying conception of New Deal legislation is the provision of government insurance against the economic dislocations of life. The Hundred Days, for example, was of course devoted to addressing in an immediate sense some citizens' suffering from the Depression--the Emergency Banking Act, the Federal Emergency Relief Act, and the like--but Hundred Days' legislation also sought to establish permanent forms of government insurance for such losses. The Emergency Farm Mortgage Act created government guarantees of farm mortgages as did the Home Owners' Loan Act, of home mortgages. Even more explicitly economic legislation such as the Agricultural Adjustment Act, through price supports, and the National Industrial Recovery Act, through its codes of fair competition and its efforts to standardize prices and wages within industries, were rationalized as means for reducing the variance of economic outcomes over time, a clear insurance rationale.
After the Hundred Days, of course, the insurance cast of New Deal legislation became more obvious and more extensive. The creation of Federal Deposit Insurance, the Emergency Relief Appropriations Act dealing with catastrophes, the Social Security Act providing relief from unemployment, disability, and old age are all forms of direct government insurance. And surely this was Franklin Roosevelt's view. In his message to Congress on Social Security, Roosevelt explained his desire to introduce legislation to further "the security of the citizen and his family through social insurance . . . to provide at once security against several of the great disturbing factors in life--especially . . . unemployment and old age." And who can forget Roosevelt's great dictum in his remark to Frances Perkins: "I see no reason why every child, from the day he is born, shouldn't be a member of the social security system . . . I don't see why not. Cradle to grave--from the cradle to the grave they ought to be in a social insurance system."
In essential respects, the subsequent development of the liberal state represents only an extension of this basic idea: the Great Society with its creation of now-enormous health insurance programs: Medicaid and Medicare; the creation of multiple specific government insurance plans such as Federal Crop Insurance and Flood Insurance; the expansion of loan guarantees for small business, for student education, for foreign trade through the Export-Import Bank; the more recent creation of the Pension Benefit Guarantee Corporation; all of these are examples of government-provided insurance. Indeed, even ignoring the massive health insurance programs and Social Security, our government today engages in vast insurance activities. According to the 1996 Budget, government-provided direct insurance and loan guarantee programs totalled a 1994 face value of $7.3 trillion.
But the idea of the government's role as an insurer extends beyond mere numbers or specific programs. The conception of the state as an insurer or risk spreader provides the basis of the most powerful statement of moral philosophy of our century: John Rawls' A Theory of Justice. Rawls' philosophical and social judgments proceed from the evocative metaphor of the "veil of ignorance" behind which none of us as citizens can know what our abilities or endowments will be or what misfortunes in life we might face. According to this analysis, we should define our public institutions as if all differences among individual citizens and among personal experiences are risks. The role of the government, according to this conception, is to minimize the losses from those risks for each member of the citizenry.
Of course, our citizenry has not read John Rawls to a person, but the underlying idea extends broadly. More colloquially (and to present yet another related metaphor), there are few citizens and certainly no politicians in this election year who would dispute in a serious way that one important role of the government is to provide a "safety net." One can debate how tightly the safety net should be woven or how high off the ground the net should be strung, but to suggest the total elimination of the government safety net--that is, to deny the role of the government as an insurer--is in today's political debate largely unthinkable.
II. How Insurance Reduces Risks
Why is the government regarded as particularly effective in spreading risks? The central basis of the view is quite simple: to take a straightforward statement of the point, a loss ruinous to a single individual is less ruinous if spread over a larger set of individuals and, if spread broadly enough, can become miniscule. The state, of course, is the largest social entity. Moreover, the state, unlike any specific group of citizens, maintains a continuous existence over time, and so can spread losses not only within one generation, but over multiple generations, reducing the incidence of any loss to a small fraction. Because of its scope and inclusiveness, the state comprises a wide range of disparate activities--indeed, the widest range of disparate activities--so that it can achieve maximum diversification by pooling disparate risks, such as the risks of unemployment with the risks of floods with the risks of industrial disability, and on.
My ambition in this talk is to take this proposition seriously. We know a great deal about how insurance operates in real markets; nothing that I am going to present here in that regard is controversial in the slightest. My objective, however, is to take that received knowledge of market insurance and apply it to the context of government-provided insurance.
Before beginning the analysis, however, I think that it is important to distinguish between spreading risk, in the sense of distributing its incidence over a different or larger set of the population, and reducing risk, which is reducing the magnitude of losses that occur. Risk spreading or shifting is different than risk reduction. As I shall explain, the most important function of insurance is to reduce risks, not merely to shift them.
There are three principal features of insurance that reduce risks:
The first is the aggregation of uncorrelated or statistically independent risks. This is a statistical point, but an important one. The aggregation of statistically independent risks reduces the risk level by causing error terms to cancel out, reducing the cumulative magnitude of loss and, therefore, increasing predictive accuracy. What this means is that, if the insurer carefully aggregates risks that are truly statistically independent, then it will need to keep a lower--indeed, far lower--level of reserves for expected losses than if each person were individually to reserve for the same risks.
The second way insurance reduces risk is by segregating high-risk from low-risk insureds into separate risk pools, charging them premiums appropriate for only the risk that they bring to the pool. Segregation has a similar statistical effect: it reduces the variance of loss. But, in addition, by charging premiums that most closely approximate the risk level brought to the pool, segregation affects the underlying activity of insureds as to whether and how much to engage in the risky activity. As a commonplace example, charging higher auto premiums to 16- to 25-year old males (or their parents) and then rapidly increasing those premiums if the driver gets a ticket or has an accident serves as a market rationing device for teen-age male driving in general and for high-risk driving in particular. Put more simply, it reduces the accident rate. Segregating the high-risk from the low-risk also increases insurance availability. For example, segregating pilots of general aircraft or, more prosaically, smokers into separate life insurance pools reduces the life insurance premiums that the rest of us have to pay, and so increases our ability to buy life insurance protection.
The third way that insurance reduces risk is by controlling moral hazard which is the tendency of persons who have insurance to subsequently increase the level of risk-causing activity. Market insurance controls moral hazard through the introduction of deductibles, coinsurance, and exclusions of coverage in insurance contracts. For example, all life insurance policies exclude coverage of death by suicide. Hospitalization coverage that requires the patient to pay 20 percent of the costs encourages patients to stay in bed only if they are really sick. Yale University provides accident coverage if I attend a conference, but not if I go mountain climbing or caving. In market insurance, deductibles, coinsurance, and exclusions of coverage are omnipresent, and they reduce risk either by leading the insured to avoid or to reduce the level of risky behavior or they expand insurance availability by reducing insurance premiums for the rest of us by requiring the high-risk to pay some proportionate amount of the loss.
III. The Government versus the Market as a Risk Reduction Mechanism
These three ways in which insurance reduces the losses from risk-related activities are quite familiar--I assure you that there is nothing whatsoever controversial in what I have presented. But my question is how effective is the government or is government-provided insurance in reducing risks in these ways? Let us go through them one-by-one:
1. The government versus the market with respect to risk aggregation.
Perhaps the strongest intuitive case for government-provided insurance derives from the aggregation function. It seems plausible that the government, as the broadest and most inclusive social entity, would be far superior to any private insurer in its capacity to aggregate risks and spread them. Indeed, it is this intuition, I believe, that provides the strongest support for the government as insurer idea.
There is some reason at the outset, however, to be suspicious of this claim from what we see in the structure of private insurance markets. That is, purported economies of scope in risk spreading, one would think, would also be operative in the private sector. In fact, private insurance markets are extremely fragmented. As a quick illustration, the market share of the largest carrier in the property/casualty industry is only 12.6 percent.
Indeed, if one gives a little thought to the subject, an aggregation advantage to government insurance is not obvious at all. As the most basic statistical awareness makes clear, within any distribution, risk terms can be cancelled out by the aggregation of relatively small numbers, as long as the risks are independent and uncorrelated. The most obvious example is modern polling. Polling organizations, such as the Gallup or Roper polls or the polls published in American Enterprise, can estimate the voting behavior of 80 million voters or the consumption behavior of 200 million consumers from a sample of very small numbers, typically 1,000 or 1,200 persons.
Similarly, the advantage possessed by the government in terms of the scope of activities it comprises is not clear. Error terms are cancelled out and predictive ability is enhanced--which is to say the risk level is reduced--by aggregating independent risks, not by aggregating highly variant risks. The idea that diversification can be achieved by lumping together highly variant risks--disability risks, flood risks, unemployment risks--is misleading. To reduce the risk level by achieving predictive accuracy requires very careful risk sorting and evaluation and ultimately specialization. This is why market insurance firms that are small can spread risks very effectively. There is no advantage to government insurance from the broad scope of activities that it can incorporate.
Indeed, the ability of the government to reduce risks through aggregation is problematic in its essence. The government, by definition, must be inclusive in the insurance coverage it offers. That is, where the government provides insurance, it must make it available to all citizens desiring coverage. The compulsion of inclusiveness, however, requires offering coverage to highly correlated risks, not statistically independent risks. The government's obligation as government, thus, diminishes its ability to reduce risks through aggregation.
2. The government versus the market with respect to risk segregation.
What about the control of adverse selection by segregating high-risk from low-risk insureds and charging risk-appropriate premiums in order to reduce the level of risky behavior and increase insurance availability by reducing the premium level? Government insurance typically engages in no efforts to control adverse selection through risk pool segregation. Discrimination in government insurance premiums is suppressed and, most often, eliminated. Federal Deposit Insurance, the Pension Benefit Guarantee Corporation, Crop Insurance are common but, in terms of risk reduction, notorious examples. Even worker's compensation at the state level--privately managed, but state-controlled--incorporates very incomplete experience rating for high-risk firms.
Where insurance is offered, without discrimination, to all parties at some average premium, adverse selection necessarily follows. First, where the government insurance plan is self-supporting, some number of low-risk parties necessarily will find the insurance not worth the premium. Few government plans, however, are fully self-supporting. Adverse selection still occurs, but through the effect of a budget constraint imposed on the insurance plan. Because there is little government effort to control adverse selection, government insurance plans typically face severe budgetary problems. The most typical government response to a budget constraint, however, is not to increase the level of discrimination in order to expand insurance availability, but rather to lower the average level of benefits, reducing the attractiveness of insurance and reducing the extent to which the insurance serves to protect insureds. The experience of virtually every Western European country today with the impending bankruptcy of health and social welfare programs is only the most recent prominent example. 
3. The government versus the market with respect to the control of moral hazard.
Finally, what about the control of moral hazard through deductibles, coinsurance, and exclusions of coverage? The government as an insurer seldom makes efforts of this nature to control moral hazard. Though some government insurance plans incorporate small deductibles and limited amounts of coinsurance, the extent to which these contractual methods are relied upon to reduce risk is vastly less than in the private sector. As a consequence, virtually every study of government insurance activities shows moral hazard problems to be severe. Again, the savings and loan experience is an obvious example. Many commentators have proposed greater governmental regulation of risk-causing behavior as a means of controlling moral hazard. Indeed, there is a wide economic literature--led by Peter Diamond of M.I.T. --that simply assumes that moral hazard can be controlled equally by government regulation or market insurance contracts. But the market disproves the assumption. Insurance companies can achieve equal forms of regulation through underwriting guidelines. Thus, the existence of deductible and coinsurance provisions in virtually all property/casualty contexts despite full underwriting authority shows that regulation--whether by the goverment or in the market--is an imperfect means to achieve this end.
IV. Why is the Government Ineffective at Risk Reduction?
This brief review suggests that, in terms of each of the three risk-reduction functions of insurance--aggregation, risk segregation, and the control of moral hazard--government-provided insurance is less effective, not more effective, than private insurance. Why is this so? This is not a point about privatization. I support privatization in many respects. But my point here is not about bad management, sloppiness, or a lack of financial incentives. It is a deeper point. The ethic and the principles of government are antagonistic to risk reduction.
Effective risk reduction is achieved by market discipline: by differential charges according to risk level; by constraints on benefits to control moral hazard; and by discrimination and narrow risk pool definition to control adverse selection. Private insurers are rewarded in the marketplace according to their ability to reduce societal risks in these ways.
In contrast, the state, for very good reasons, is unable to engage in any of these forms of market discipline. The size and necessary inclusiveness of government obstructs optimal risk aggregation. By definition, the state must provide benefits to all citizens, not exclusively to citizens chosen according to risk proclivities in order to create appropriate risk pools.
Similarly, the state's commitment to non-discrimination prevents control of adverse selection through risk segregation. Our government, committed as it must be to principles of equal treatment, cannot engage in discrimination among citizens according to risk proclivity--even risk proclivity by class of citizen--to set premiums to correspond closely to the risk that the citizen brings to the insurance pool.
Finally, the political responsiveness of the state to voter interests in benefits cripples efforts to control moral hazard. Given that the reason that the government engages in insurance is a humanitarian response to loss, our government cannot, once it has entered the insurance business, introduce deductibles, coinsurance or, surely, exclude coverage entirely to those activities or to those individuals most likely to suffer the loss. As a result, our government's insurance programs do little to reduce risk levels.
It is frequently claimed that the government has an insurance role because it can offer coverage in contexts in which there are incomplete insurance markets, for example the special government programs providing flood and other disaster insurance. The sources of the special government expertise here, however, are not clear. Private insurance markets are incomplete, not by chance or lack of interest, but because risks are essentially uninsurable where moral hazard and adverse selection cannot be controlled. In these contexts, no insurance market can survive. The government cannot influence the survivability of an insurance pool. Indeed, because the government is less able to control moral hazard and adverse selection, the government is less able than the private market to make risks such as these truly insurable.
This is not simply an analytical point. When one reviews the empirical literature on the effects of government-provided insurance, the findings are unanimous:
1. To take the most prominent recent example, government-provided insurance for savings and loan deposits increased the risk level of investments by lenders. The best current estimate of these losses is $147 billion.
2. Government-provided unemployment insurance increases the extent and duration of unemployment.
3. Social Security disability insurance, because it does little to control moral hazard and adverse selection, increases claims of disability, especially among older workers.
4. Workers' Compensation Insurance, by limiting the extent of experience rating in order to redistribute wealth to small firms, increases the number of worker injuries.
I could go on with these examples, but they are cumulative--and uniform in their conclusions.
Now, you might ask, surely there must be exceptions? Surely, there is some role for government with regard to insurance? However much I am critical of government, can I possibly be claiming that government increases the rate of, say, catastrophes such as hurricanes, floods or crop damage?
I will not go so far as to claim that government-provided insurance increases the frequency of natural disasters. On the other hand, I have no doubt whatsoever that the government provision of insurance increases the magnitude of losses from natural disasters. Government flood insurance increases building on flood plains and, thus, increases losses from flooding. The government mandate of earthquake insurance increases building on fault lines and, thus, increases losses from earthquakes. In a recent paper on government insurance for catastrophes, I reviewed studies of the effects of a government disaster insurance program for specific enumerated crops, such as turnip greens and the like, introduced in the late 1980s. The studies showed that, since the introduction of the insurance program, 45 percent of federal disaster aid for turnip green crop losses was distributed to farmers in counties, none of which had reported any acreage of turnip green planting prior to the enactment of the disaster program.
V. The Moral Foundations for the Criticism of Government Insurance
As I mentioned, I am not making a point simply about the benefits of private versus government management. Instead, I believe that there are strong moral grounds to criticize the government provision of insurance.
The first ground should be obvious, though it is often ignored or, perhaps better, washed over in the tide of humanitarian feeling in favor of the provision of governmental benefits to citizens who have suffered loss.
No loss is ever fully compensable. This point is obvious with regard to personal injury loss, but it is true of every loss since, given a limited life span and limited life opportunities, every loss is a loss of an alternative opportunity. This means, however, that, to the extent that government insurance increases losses, it inflicts incommensurable loss on the society and on the citizenry. Citizens can debate issues of redistribution. Indeed, often the support for government insurance proceeds as if the only issue is one of redistributing wealth to those who have suffered loss from those who have escaped it. Preferences for redistribution cannot be morally defended, however, where the redistributional institution that has been selected increases the frequency and magnitude of loss to the society. This subject of the government's role in serving as an insurer of societal losses has been dominated by an admirable humanitarian impluse--that is why it is a liberal program. But humanitarian actions that increase the frequency and magnitude of incommensurable losses cannot be morally defended. On this point, conservatives need not yield the moral ground.
The second moral or normative criticism of government insurance relates to a conception of how government can advance our society. Here, again, I want to return to a comparison of market insurance to government insurance and address the question of insurance finance.
How do private insurers finance their activities? They charge premiums according to the best estimate of the risk level that the insured brings to the pool, and then they invest these premiums to create the largest possible fund from which to pay off subsequent losses.
How does the finance of government insurance compare? First, in all government insurance contexts, as described earlier, there is a very loose relationship between premium levels and risk levels. Indeed, often, there is no relationship: losses are charged against general revenues. The provision of "coverage" of this nature does not reduce risk, but only shifts risk to taxpayers. Many commentators have treated the government as risk neutral or comparatively risk neutral. But the risks to taxpayers associated with tax increases are not zero.
Assumptions about risk preference, however, ignore the more important comparison. Government provision of insurance systematically redistributes wealth toward relatively risky activities. As a consequence, the contrasting financial effects of government versus private provision of insurance are profound. Progressive federal income taxation is designed to most heavily tax those most productive individuals, corporations, and enterprises in the society. Under the progressive income tax, the burden of taxation falls most heavily on the greatest income-producing activities of the society. In contrast, private insurance, through premiums, taxes the riskiest activities of the society--and so reduces the risk level--and then invests these premiums in the most productive activities to reduce the effective cost of these risks even further. Thus, government insurance organizes taxes and investment exactly backwards. Private insurance taxes the most risky activities and invests the amounts most productively while government insurance taxes the most productive activities to redistribute to the most risky.
And here is the great divide between the conception of government implicit in my criticism and the liberal conception of the government as insurer. It is the difference between government investments in productive activities or individuals versus government investments in risky or unproductive activities. The New Deal and its subsequent liberal extensions--however admirably humanitarian--were totally turned around on this issue. The New Deal-Great Society-liberal conception of the state not only increases the frequency and magnitude of harm, it redistributes the wealth of the country through subsidization from productive and wealth-enhancing activities to risky and unproductive activities.
This is not the occasion to elaborate that affirmative vision of the state. Regrettably, but necessarily, I am on a solely critical mission today. But I believe that there is an affirmative mission implicit in this criticism. According to my conception of government, government expenditures, financed by taxation of the income of productive Americans, can only be justified if the return on the investment of the government expenditure exceeds the return to the citizen from the expenditure of his or her own money. This seldom happens, and it is why, at a crude level, that the government that governs least, governs best. But the point I want to make today is not simply a point about poor choices by our government or unwise decisions by decent people. Rather, it is a point about the underlying conception of what our government is about. We must reject the conception of government as savior. Given that we have a government, we must view the government as an instrument with the potential of enhancing the lives and the welfare of our citizenry. This does not come from the government as a safety net--as an insurer. That government increases harms, and it is morally wrong for it to continue to do so. The affirmative conception of government, instead, is a government that takes income from our citizens only where it can be invested more productively to increase the welfare of our citizenry by increasing its security and wealth. Eliminating the insurance provided by our government is only the first step.
1. Cited in Arthur J. Altmeyer, The Formative Years of Social Security at 3 (1966).
2. Cited in Arthur M. Schlesinger, Jr., The Coming of the New Deal at 308 (1958).
3. Budget of the United States Government Fiscal Year 1996: Analytical Perspectives at Table 9-1, page 122 (1995).
4. See generally, Guido Calabresi, Costs of Accidents (1970).
5. For a more complete discussion of these points, see Priest, The Government as an Insurer in the Context of Catastrophic Loss (mimeo, October 1995), available from the author.
6. For a demonstration of a dramatic increase in the accident rate following the prohibition of insurance discrimination in Quebec by age, sex, violation record, and accident experience, see Rose Ann Devlin, Liability Versus No-Fault Automobile Insurance Regimes: An Analysis of Quebec's Experience, U. Toronto Law & Econ. Programme (1988) (9.6% increase in fatal accidents, 26.7% increase in bodily injury claims, and 5.3% increase in property damage claims following adoption of non-discriminating no-fault plan).
7. Best's Aggregates and Averages Property-Casualty, 1994 ed. (measured in terms of net premiums written).
8. E.g. Robert Keatley, "Cure for Sweden's Current Economic Ills may Lie in Dismantling Model System," Wall St.J. at A6, col. 3, Apr.1, 1994; Roger Cohen, "Europe's Rescission Prompts New Look at Welfare Costs," New York Times, at A1, col. 1, August 9, 1993.
9. E.g., Peter Diamond & Eytan Sheshinski, Economic aspects of optimal disability benefits, 57 J.Pub.Econ. 1 (1993).
10. See Robert C. Merton & Zvi Bodie, Deposit insurance reform: a functional approach, 38 Carnegie-Rochester Ser. Pub. Polc'y, 1 (1993).
11. Eric Engen & Jonathan Gruber, Unemployment Insurance and Precautionary Savings, (mimeo, M.I.T., 1995).
12. Jonathan Gruber & Jeffrey D. Kubik, Disability Insurance Rejection Rates and the Labor Supply of Older Workers, NBER Working Paper Series, No. 4941 (1994).
13. J.W. Ruser, Workers' Compensation Insurance, Experience-Rating, and Occupational Injuries, 16 Rand J. Econ. 487 (1985).
14. U.S. Senate, Comm. on Agriculture, Nutrition and Forestry, Staff Rep., "Questionable Disaster Payments: $92 Million to 8 Non-Program Crops in 9 States, 1988-93," August 1994 (indicated "DRAFT 9/9").
15. In the first instance, one must compare the beneficial effect on the risk level of the incidence of private insurance premium charges versus the, at best, neutral incidence of government taxation. At the minimum, to the extent of the excess burden of taxation--the deadweight loss from the effect of an income tax on the choice between labor and leisure or the effect of an excise tax on commodity consumption--the financing of government insurance from general revenues is inferior to the financing of private insurance from premiums charged to risky activities. The incidence of the federal income tax is not clearly correlated with the extent of risky activities. The incidence of insurance premiums clearly is.
George L. Priest is the John M. Olin Professor of Law and Economics at Yale Law School.