The Role of Private Sector Annuities Markets in an Individual Accounts Reform of a Public Pension Plan

Contributor Notes

Author’s E-Mail Address: gmackenzie@imf.org

Pension reforms that establish individual accounts will diminish the relative importance of the traditional state pension while creating a significant role for individual accounts in providing income for retirement. This paper surveys the policy issues this new role entails. It offers general advice to countries considering such issues as the restrictions to be placed on the timing, extent, and form of withdrawals from individual accounts and the need for mandatory annuitization (conversion into annuities) of accumulated account balances. The paper also considers the role that private annuity markets should play and related regulatory, social safety net, tax, and administrative questions.

Abstract

Pension reforms that establish individual accounts will diminish the relative importance of the traditional state pension while creating a significant role for individual accounts in providing income for retirement. This paper surveys the policy issues this new role entails. It offers general advice to countries considering such issues as the restrictions to be placed on the timing, extent, and form of withdrawals from individual accounts and the need for mandatory annuitization (conversion into annuities) of accumulated account balances. The paper also considers the role that private annuity markets should play and related regulatory, social safety net, tax, and administrative questions.

At present we want to encourage prudence in the sense of distributing income through a man’s life. When that time comes all sorts of fancy devices possibly with a counter-insurance element in it, e.g., annuities on joint lives… from the notes of Lord Keynes for the National Debt Enquiry, 1945.2

“Father, I have often thought that life is very short.”–This was so distinctly one of his subjects that he interposed.

“It is short, no doubt, my dear. Still, the average duration of human life is proved to have increased of late years. The calculations of various life assurance and annuity offices, among other figures which cannot go wrong, have established the fact.”

“I speak of my own life, father.”

“O indeed? Still,” said Mr. Gradgrind, “I need not point out to you, Louisa, that it is governed by the laws which govern lives in the aggregate.” from Hard Times, by Charles Dickens.

I. Introduction and Summary of Conclusions

A. Introduction: The Need for Private Sector Annuities

“Social security privatization” has come to describe a reform that involves the partial or complete replacement of the public defined-benefits system by a privately managed but publicly regulated defined-contributions system. Specifically, the introduction of a system of individual accounts (hereinafter referred to for convenience as an IA reform) supplants the payroll tax, the traditional source of finance for public pension systems. On the benefits side, an IA reform results in the funds that accumulate in the IA of a contributor becoming a source of income in retirement that replaces, wholly or in part, the pension, typically an indexed life annuity that the state provides.

This general pattern of reform has many variants. However, all of them—apart from a reform that effectively maintained the existing system by requiring that the funds that accumulate in an IA be used to purchase an annuity from the state—would lessen the role of the state pension.3 Although the merits of an IA reform have been endlessly debated, this aspect of an IA reform has not yet received all the attention that it deserves.4 The state pension plays a major role in providing income to the elderly in most advanced economies, and in not a few emerging market economies, although its role has declined in countries like the United Kingdom and Chile that have moved away from the standard model.

The partial or complete elimination of the indexed life annuity that a conventional public pension system provides raises some important questions. This paper aims to provide a general guide to countries that are coming to grips with them. A separate paper (Mackenzie (2002) deals with the same issues in the specific case of the United States, in the light of recent proposals to introduce individual accounts. Indexed life annuities can play a vital role in protecting the elderly from the risk of outliving their resources or underestimating the effects of inflation in real income, even though it is possible that the provision of an annuity by the state means that too large a share of a retiring person’s wealth is annuitized (takes the form of an annuity).5 Typically, there is no close substitute for the state annuity, apart from occupational (employer-provided) pensions, and the coverage of this second tier is usually far from universal.

The full consequences of a diminished role for the state pension are incalculable. Private sector life annuities have typically not been regarded as a good investment, in part because of the adverse selection that affects annuities markets.6 There are good reasons for thinking that demand for private annuities will not spontaneously substitute for the state’s pension. Many retired persons, if left to their own devices may well have difficulty formulating and sticking to a spending plan that will spread their resources over their remaining lifetime. The state pension ensures these individuals enjoy some amount of regular income. Although the across-the-board reduction or elimination of the state pension could benefit some individuals, it could cause hardship to many others. It will be important, in consequence, to examine carefully the implications of mandatory annuitization or other restrictions on the use of the funds accumulated in IAs.

A government contemplating the reform of a public pension plan, and in particular, the distribution phase must make two basic decisions: (i) should there be restrictions on the distribution of the funds that accumulate in an IA, such as mandatory annuitization; and (ii) if annuitization is mandatory, should the provision of annuities be privatized. Although these choices are obviously fundamental, in that they define the basic model of reform the government will follow, an informed choice between them requires that a government investigate at some length all the ramifications they entail. The government must address a series of design issues whatever its final choice will be. Some of these are relevant whether or not the provision of annuities remains in the hands of the state; others arise only if annuitization becomes mandatory but the government cedes responsibility for the provision of annuities to the private sector.

The basic design issues that arise with mandatory annuitization with or without private sector provision are as follows:

How should the degree or extent of annuitization be determined? Four broad possibilities are:

  • Minimum premium value: the premium paid for the annuity must equal or exceed some minimum value.

  • Minimum annuity payment: the income the annuity generates must equal or exceed some minimum value (e.g., US$700 or €500 per month).

  • Minimum replacement ratio: the income the annuity generates must equal or exceed some specified ratio of a measure of the annuitant’s earned income.

  • Mixed minimum annuity payment and replacement ratio: the income the annuity generates must equal or exceed the greater of some minimum value or some specified ratio of a measure of the annuitant’s earned income.

Should the value of the funds accumulated in the account or the income that the annuity generates be guaranteed? If so, should the full amount of the mandatory annuity as determined by one of the rules set out above be guaranteed or only part of it?

What form should the mandatory annuity assume? Annuities can be classified according to how long or under what conditions the annuitant receives income under the terms of the annuity contract; the timing of the payment of the premium; the form of the income the annuity provides; and the survivorship entitlement.

The following list, although it is not exhaustive, provides a general idea of the possible forms an annuity can take:

  • duration: simple life, term (or period) certain or term certain and life;

  • timing of payment: (i) immediately prior to annuitization, or some time before (a deferred annuity), and (ii) in one payment or a series of payments;

  • form of income: fixed nominal, fixed real, variable with guaranteed minimum, variable but no guaranteed minimum; and

  • survivorship entitlement (various possibilities, including joint and survivor).

What exceptions should be allowed? Should the terminally ill be allowed to opt out of mandatory annuitization?

How will the annuity fit into the existing social safety net? Should IA holders who have failed to save enough to finance the minimum annuity required of them have their account topped up, or should they be handled by the existing safety net? The state pension normally includes safety net features for family members that an annuity cannot provide—e.g., survivors’ benefits that do not lower the value of the pension. How are these benefits to be treated? How should the redistributive features of the state pension be accommodated?

The issues below take on particular relevance in the case of private sector provision:

How much liberty should annuity providers have in pricing their products? Private sector annuities differentiate between the sexes. A female annuitant pays more for a given income than her male coeval because the average woman lives longer than the average man.7 A similar issue with regard to premium differentiation along racial lines may conceivably arise in the United States and possibly in other countries. Is such differentiation to be allowed?

Will the framework of financial regulation and supervision be adequate in the brave new world of an enlarged annuities market? In the United States and the United Kingdom, the incidence of life insurance companies not honoring their commitments to annuitants has been very low, but insurance companies do get into difficulty. Some companies default, and some find themselves unable to honor their commitments to annuitants (and other creditors). Whether annuitization is mandatory or not, an IA reform will increase the demand for annuities, and with it the importance of sound regulation and supervision.

Should the state provide a guarantee of some kind to protect annuitants from the failure of their annuity provider, if it requires the purchase of annuities? This question cannot be answered satisfactorily without considering the potential for moral hazard that a guarantee will create, and is closely related to the previous question.

Are there differences in the tax treatment of the state pension and private annuities, and if so how should they be resolved?

What new procedures or institutions are required to ensure that the market for annuities plays its proper role in the provision of income for the retired, and what sequencing issues arise? The shape these procedures will assume will depend on the administrative and legal role that the state will play and on the extent of the choice IA holders have when they withdraw funds from their accounts. Revamping the current regulatory framework could take time. Even if the government is no longer providing the annuity, it has a vested interest in ensuring that annuitants contract with sound financial institutions and receive good value for their money. If annuity purchase is mandated, the government may have to devise exceptions to the general policy, and institutions or procedures to minimize abuse or gaming.

These words give the general flavor of the implementation issues that would need to be addressed.

B. Conclusions

Since the paper is intended to be a general guide to any country engaged in or contemplating an IA reform, rather than a blueprint for a particular country, its conclusions are necessarily general and somewhat tentative and conditional. The generality is necessary both because the paper cannot survey all the relevant institutions in each of the countries that might implement an IA reform, and because so much depends on the exact design of the reform. It makes a great deal of difference, for example, whether the IA reform is an “add-on” reform, which might leave the state pension basically intact.8 In addition, the answer to one question must depend on the answers to all the others. A decision to require the purchase of only a minimum annuity (rather than one intended to achieve some target replacement ratio) has different implications for financial regulation and supervision than a decision to require the annuitization of all the funds accumulated in an IA, for example.

To start with the basic question, some degree of mandatory annuitization is highly desirable for most countries implementing an IA reform, particularly a radical version that completely eliminated the state pension. The same arguments that justify compulsory participation in a social security system—the need to discourage improvident behavior and free-riding—can justify some degree of mandatory annuitization. Compulsory annuitization would also reduce adverse selection.

An informed decision as to the optimal degree of annuitization is ideally based on comprehensive surveys of the share of wealth at retirement (apart from the funds in the IA) that is already annuitized. In most countries, such information is not available. In the United States, however, the poorest fifth of OASI recipients are estimated to have virtually no occupational pension or other annuitized income. It is likely that the state pension dominates other forms of annuitized wealth in the same income class in other countries as well. Without mandatory annuitization, a radical IA reform would mean that the elderly poor would be bereft of the valuable longevity and inflation insurance that the state pension provides unless they turned to the private annuities market. That many would do so is unlikely for a number of reasons.

As to the form a mandatory annuity might take, an indexed life annuity with survivor benefits would best replicate the most desirable features of the typical state pension. A minimum annuity payment defined in terms of a replacement ratio (over a certain income range) with a minimum value would replicate a key design feature of many state pension systems. The greater the share of IA funds that are annuitized, however, the greater the risk of over-annuitization and also of moral hazard if the value of the guaranteed annuity is set equal to the minimum mandatory annuity. In practice, the political environment in some countries might require that an informal guarantee be extended to cover that part of the mandatory annuity that was not covered by the official guarantee.

Exceptions to mandatory annuitization should be kept to a minimum, being extended at most to IA holders suffering from terminal illness. Annuity pricing poses problems. Differentiation is fair by one standard but not by another. Enforcing uniform pricing is administratively messy. Group annuitization, if organized by the government would reduce annuity premiums and simplify administration. These advantages come at the cost of reduced flexibility.

The major issue for financial regulation and supervision that arises with mandatory annuitization is the threat of moral hazard. In fact, there are two points in the accumulation-distribution cycle where moral hazard can arise. The first arises during the accumulation phase if a guarantee applies to the minimum value of the funds in an IA upon retirement, since this can create an incentive for an IA holder to invest his holdings recklessly, knowing he will be bailed out. This risk can be contained by limiting the value of the guarantee on each IA. The second arises when a guarantee on annuity payments creates an incentive for an annuity provider to offer excessively generous terms. Addressing the second problem effectively requires both a limit on the guarantee (as is typically the case with deposit insurance) and effective overall regulation.

The successful introduction of mandatory private sector annuities would require a substantial amount of preparation. This is particularly true of countries where annuities markets are limited and financial regulation needs strengthening. Even in countries where financial markets are sophisticated and financial regulation is strong, it is easy to underestimate the preparatory requirements for a successful reform. From the publicity campaign that informs IA holders of the changes to their retirement regime, to the implications of increased choice regarding the form that retirement income will take, to the regulations that will govern exceptions to the annuitization rules, to the necessary changes to the framework of financial regulation, the introduction of mandatory annuitization would require careful planning.

Finally, and depending on the eligibility of contributors to the state pension system for participation in the new IA system and on the treatment of their accrued pension rights under the old system, the first annuities under the new system may not be contracted until some time after the reform is enacted. While this delay in the start of annuitization may seem to make the distribution phase of the reform less urgent; it is not a good reason for avoiding detailed and early planning.

The corpus of the paper begins in Section II with a description of the basic features of the market for life annuities that draws on the more technical discussion of the economics of annuities in Appendix I. Section II also discusses some relevant aspects of the regulation of life insurance companies. Sections III, IV, and V then turn to the basic issues this introduction has highlighted. Section III discusses the problems that can arise with private sector provision of annuities, including the risks of over and under-annuitization, and the consequences of unpredictable financial markets. Section IV deals with regulatory issues, and Section V with tax and administrative issues.

II. Basic Features of Annuities Markets in the United States and Elsewhere9

A. The Market for Annuities

Despite the very long history of annuities—their lineage extends back to ancient Rome—the market for private sector annuities in our own times is small in most countries, if it exists at all.10 The exceptions to this general pattern include Australia, Canada, Chile, Singapore, Switzerland, United States, and the United Kingdom, although in none of these countries is the annuities market really large in comparison with the market for other financial instruments. Markets for annuities are also developing in Argentina and other countries where IA reforms have been implemented.

The annuities market in the United States offers many different kinds of annuities, although its overall size is modest in comparison with that of other financial assets. The variable annuity, which combines features of a savings plan with an equity mutual fund, is by far the most important part of the market. Although a minimum payment is normally guaranteed once the holder opts for annuitization, payments vary with the return on the investments that fund the annuity. Sales of variable annuities in 1999 are estimated to have amounted to US$122 billion. Sales of fixed annuities amounted to US$37 billion, or roughly US$19,000 per person reaching the age of 65 in 1999. The single payment immediate fixed life annuity— which in return for a lump sum payment provides a stream of payments fixed in dollar terms that commence upon purchase and are paid for as long as the purchaser (annuitant) lives—or in the case of an annuity with a joint survivor, as long as either he or his spouse lives—is not popular, with only US$2 billion in sales in 1998.11

In the United Kingdom, the life annuity is more important in relative terms than it is in the United States. Individuals with a defined-contribution scheme who have opted out of the state earnings-related pension scheme (SERPS) must annuitize 100 percent of the funds that would otherwise have gone to finance a state pension. The annuity is to be purchased between the ages of 60 and 75. Funds attributable to additional contributions should be 75 percent annuitized. This annuity is to be purchased between the ages of 50 and 75. The market for annuities is now comparatively small, but can be expected to grow as more and more individuals with defined-contributions plans reach retirement age (Banks and Emerson (1999).

Life annuities have become fairly common in Chile with the growth in the number of Chileans retiring under the system of individual accounts that was introduced in 1981. This development reflects the restrictions the law imposes on the form that withdrawals from individual accounts may take, the restrictions being intended to ensure that account holders have sufficient funds to acquire an annuity that will allow them a reasonable standard of living in retirement. The accumulated funds in individual accounts typically finance either an indexed life annuity or a series of programmed withdrawals (essentially an indexed term-certain annuity), although the funds in the account that exceed the equivalent of what is necessary to finance a pension that equals or exceeds both a specified replacement rate and value may be withdrawn in a lump sum. As of December 2001, the share of distributions taking the form of life annuities accounted for about 53 percent of the total chosen by retiring plan participants.12

Singapore has also witnessed a rapid growth in recent years in the market for life annuities. Doyle, Mitchell, and Piggott (2001) note that about one-sixth of the workforce retiring in 1999 purchased an annuity. This growth results from recent changes to the regulations governing withdrawals from the Central Provident Fund, a publicly-run defined-contributions scheme—that encourage an account holder to devote a specified minimum of the funds accumulated in his or her account to the purchase of a life annuity when the funds become accessible for withdrawal at age 62. Australia’s pension system has a privately managed individual accounts component, known as the Superannuation Guarantee, that applies to most workers. A contribution of 9 percent of salary finances either a defined-contribution pension or fully funded defined-benefit pension. At age 55, the “preservation age,” the account holder may elect to receive either a lump sum withdrawal or an income stream. The standard age (i.e., old age) pension is subject a minimum retirement age of 65 for men (61½ for women) and to a means test that applies to both income and wealth. These conditions create an incentive for some account holders to elect the lump sum withdrawal option and spend the money quickly in order to increase the size of their age pension. Concern over this incentive effect has led to a change in the law that will eventually increase the preservation age from 55 to 60.13 Information on the annuities markets of countries that have implemented an IA reform in the 1990s is scanty.

Annuities markets in France, Germany, Italy, and Japan are apparently small, mainly because of the relative generosity of the state pension system and the important complementary role played by occupational pensions. In Italy’s case its inflationary past may also play a role.

Indexed life annuities depend on the existence of indexed government securities. Markets for these instruments have been functioning for some time in the United Kingdom, Chile, and Israel. They are popular in Chile and Israel, two countries with moderately low inflation rates today but with an inflationary past. (Nominal annuities are not available in either country.) A market for indexed annuities exists in the United Kingdom as well, but is only just starting up in the United States, although the Treasury has been issuing indexed debt (TIPS) since 1997. Markets for indexed annuities, albeit very small ones, are also found in Colombia and Peru, where the IA reforms implemented in the 1990s include provisions requiring IA holders to buy either an indexed life annuity (which may be deferred) or engage in a series of programmed withdrawals when they retire.

This brief review of the international experience lends support for the view that annuities markets are relatively larger, or are growing at a faster pace, in those countries that have public pension systems with an important defined-contributions component and that encourage or require the annuitization of the funds accumulated in an IA upon retirement. A corollary of this point is that the annuities market remains to be developed in those countries with generous public systems that have not adopted any variant of the Chilean model.

The life annuity sold by life insurance companies and other financial institutions in the countries just named and a few others is similar to the retirement benefit provided by the state. There are important differences between them, however. Most state pensions are indexed, usually to consumer prices, unlike the typical private sector annuity, although indexation in practice has been less than full in some countries. Legislatures can vote to change the value of the state pension, although the political influence of retirees and the restraining effect of legal and constitutional provisions would make wholesale or abrupt changes unlikely in many countries including the United States. In contrast, life insurance companies are bound by their contract, although fraud or managerial incompetence (or possibly very bad luck with investments) can disrupt payments to annuitants.

B. Influences on the Demand for Life Annuities

Economists have proposed five basic explanations for the small size of markets for private sector annuities, and life annuities in particular. Specifically, demand for life annuities may be inhibited by: (i) the substitutes provided by the state pension and occupational pensions; (ii) the desire to leave a bequest; (iii) the need to defray unexpectedly large expenditures; (iv) the perceived costliness of annuities entailed by adverse selection; and (v) the shortsightedness of potential annuitants or their lack of understanding of an annuity’s properties. A sixth explanation, which may have more applicability in emerging market than advanced economies is a lack of trust in financial institutions.

These are plausible explanations. The provision of retirement income by employers or by the state, which typically takes the form of a life annuity could be expected to reduce the demand for annuities from an insurance company. In countries like France, Germany, and Japan, the combined annuity provided by state and employer currently replaces most of the retired person’s earned income. The dominant role played by these two pillars of retirement saving leaves little room for individual annuities, although scope undoubtedly exists for growth in the annuities market. The share of the wealth of the average retired person that is annuitized may be smaller in the United States than it is in these European countries. However, a recent study found that annuitized income, including estimated private pension and social security accrued benefits, accounted for 52 percent of average net wealth of a sample of Americans aged 51–62. For the bottom 5 percent by wealth the ratio was estimated to be 195 percent, of which 179 percentage points was accounted for by social security (Gustman et al (1997)).

The wish to leave a legacy will may also deter annuitization. When an individual buys a simple life annuity, her premature death deprives her heirs of all or most of the wealth represented by the premium. It also leaves her less prepared (although not necessarily insufficiently prepared) to deal with unexpected but necessary outlays. Adverse selection does raise the cost of annuities for the typical person. The purchase of a simple life annuity, regardless of its implicit rate of return, may seem like a bad bargain, since it involves an irredeemable upfront payment of a large sum of money in return for a stream of income of uncertain duration. Finally, there is no question that annuities are a sophisticated financial instrument. Unlike a bond or money market investment, the annuitant gives up the principal. She does not have her cake and eat it too. Even if annuities are a good buy, the form of the transaction is not that of the typical investment.

Nonetheless, these explanations, with the possible exception of the first are not entirely satisfactory. In principal, the typical person should benefit from owning a private annuity unless his or her wealth is already substantially annuitized.14 A discussion of the merits of life annuities follows, since it is relevant for the issue, explored in Section III, of whether annuities should be mandatory if an IA system is adopted. Appendix I, from which this part of Section II is drawn, presents a more complete discussion of the economics of annuities.

The life annuity offers potentially valuable longevity insurance—protection against the risk of outliving one’s resources, whether because of unexpectedly long life, fecklessness, or bad investments. Starting from the assumption of perfect financial markets—costless financial intermediation—and making a few other simplifying assumptions, it is possible to show that in the absence of a bequest motive an individual will not only acquire annuities but will devote all his wealth to that end. The reason for this is that the return an annuitant gets on his investment, provided he lives (his conditional return) will always exceed the rate of interest.15

Adverse selection, costly financial intermediation and a desire to leave a legacy are facts of life, and would presumably reduce the amount of wealth an individual would devote to an annuity. On the other hand, they need not reduce the demand for annuities to zero. Adverse selection affects the market for life annuities because the life expectancies of annuitants exceed those of the general population, which increases the premiums life insurance companies charge for annuities, making them less attractive to the general population. Nonetheless, annuities retain their property as longevity insurance, even if the terms are not as good. The same is true when the costs of financial intermediation are recognized. As for bequests, the purchase of an annuity reduces the maximum bequest an individual could leave, but depending on its terms has relatively little effect on the expected value of the bequest.16 It is not irrational or foolish for a person to buy an annuity while at the same time she plans to leave a bequest, particularly if she has good reason to believe she will live a long life.

C. The Cost and Insurance Value of Annuities

Recent research, summarized in Appendix I, has tried to gauge the effect of adverse selection and other influences on the attractiveness of annuities in the United States, the United Kingdom, and a few other countries. This work aims to calculate what a given age-group gets back in present discounted value terms for every dollar, pound, or peso it spends on life annuities. Typically the ratio of this present discounted value to the premium—the money’s worth ratio (MWR)—is less than one.17 If it were not, the insurance company would be losing money (unless it could invest the premiums at a rate that exceeds the discount rate). Recent studies of the annuity markets in the United States, Canada, the United Kingdom, Chile, Australia, Singapore, and Switzerland confirm that MWRs for annuitants consistently exceed that of the general population. For example, Mitchell et al (1999) estimated that the MWR (using the government bond rate for the discount rate) for a 65-year old female from the general population in the United States is .829; the MWR for a 65-year old female annuitant is .893 (see Table 1).

Table 1

Money’s Worth Ratio (MWR) Values for a Group of Countries: Nominal Annuities

article image
Source: For countries other than the United Stales: James, Estelle and Dimitri Vittas, “Do Consumers Get Their Money’s Worth?paper prepared for World Bank Conference,New Ideas on Old-Age Security,Washington, September 1999; for the United States,Mitchell, Olivia, and others, 1999, “New Evidence on the Money’s Worth of Individual Annuities, American Economic Review, Vol. 89 (December), pp. 1299–1318.Note: All annuities are immediate single payment life annuities. U.S. data arc for 1995.

These studies shed great light on the importance of adverse selection. If the difference between the MWR for the general population and that for the annuitant population reflects the impact of adverse selection, then they show that a substantial share of the load on an annuity for the population at large results from adverse selection. Other influences may be at work, however, as Walliser (2000)and Palacios and Rofman (2001) have pointed out. In the United States, and probably other countries, the higher an individual’s income, the longer he or she tends to live. If the demand for annuities is normally related to income, annuitants will tend to be longer-lived. Walliser (2000) argues that this does not explain all of the observed difference between the life expectancies of annuitants and those of the general population.

In any case, and as Brown (2001) and Mitchell, Poterba, and Warshawsky (1999) point out, the MWR is not a good guide to the value of annuities as insurance. The basic reason for this is that the MWR calculation, which discounts future payments by the probability of survival and by the financial discount rate does not take account of an annuitant’s aversion to penury in old age. Thus, an American male aged 65, who has about a 50 percent change of living to age 80, is assumed to discount income (consumption) in his 80s by more than 50 percent, and will attach very little weight to consumption in his 90s. Anyone who is truly averse to the prospects of growing old in poverty or living on the sufferance of relatives will place a higher weight on the annuity’s provision of income regardless of age. Some one with a total aversion to the risk of outliving his resources would act as if he was assured of living to some very ripe old age (say 90 or 95). A life annuity is extremely valuable for such a person. See the subsection “Annuities versus Self-Insurance” in Appendix I for further discussion.

Whether everybody, or even most people, would in practice discount the future at as low a rate as this is debatable. As a normative account, or a guide to sensible behavior in

retirement, however, it appears to accord well with most people’s intentions and to reflect standard financial advice.18

D. Annuity Providers

Life insurance companies are typically the sole providers of life annuities in both the OECD and middle-income countries, since the regulatory framework applying to financial institutions normally restricts the sale of annuities to life insurers. Argentina, where separate retirement insurance companies specialize in the sale of life annuities to individual account holders who are withdrawing the funds from their IA, is an exception to this rule.

Even if the regulatory framework were more liberal, life insurance companies would enjoy some comparative advantage in the provision of annuities owing to the economies of scope obtainable by providing both life annuities and life insurance. The provision of both of these products requires the knowledgeable estimation of life expectancies and survival probabilities, and hence the services of actuaries. In addition, life annuities and life insurance are natural hedges for one another.19

A further and important rationale for joint provision is that the premiums for each type of insurance are invested in similar ways, with an emphasis on fixed interest instruments. This emphasis stems from the nature of the commitment that the insurance company enters into when it sells a fixed life annuity—the payments on an annuity are exactly like those on bonds—as long as the annuitant survives—in that nonpayment constitutes default. The need to honor the annuity contract places a limit on the share of the portfolio of assets funding the annuities that could be invested in equities or high-yield bonds. There may be some scope for investing a part of the portfolio in these higher-risk instruments, however, if doing so increases the mean return appreciably without increasing risk appreciably.

A life insurance company offering life annuities confronts a number of risks. First, it must deal with interest rate uncertainty (or equity risk if it invests in equities). The extent to which interest risk is a problem depends on the extent to which the company can match its expected payments stream with the interest and redemption schedule of its portfolio of bonds and other assets. This in turn will depend on how complete the bond market is, and on its length. Bond markets offering a full range of maturities from 30 days to 30 years are not common outside the industrialized countries.20

Supposing for the sake of example that an annuitant’s maximum life post-retirement is 30 years, and that Treasury securities are available over the whole maturity range up to 30 years, an insurance company could acquire a portfolio of bonds that would be perfectly matched. In other words, that part of the annuity payment that exceeded the interest earned on the portfolio during the same period could always be met by selling a maturing bond, thereby avoiding the consequences of a falling bond market. In practice, a 30-year bond is not quite enough, since many residents in both the advanced and emerging market countries will still be alive more than 30 years after they retire and buy a life annuity. In consequence, some interest rate uncertainty is unavoidable even in the countries with the most developed financial markets.

There will also be a trade-off between the average rate of return on bonds and interest uncertainty. The insurance company that holds only government securities will earn a lower rate of return on average than one that invests in corporate securities, which have a higher yield. However, corporate debt, apart from the obligations of utilities is not generally available with 30-year maturities. Since life insurance companies in most OECD countries do invest in corporate bonds (typically, but not always AA rated), they are exposed to more interest rate risk than they would be if they invested only in public debt. In addition, the higher yield of corporate debt comes at the cost of higher default risk. For the U.S. market in particular, the apparent risk of default for the highest-rated corporate grades is not significant, but undergoes a noticeable increase with “high-yield” (i.e., junk) bonds. Liquidity risk—the risk that in periods of generalized liquidity preference bond markets may effectively freeze up—is especially significant for high-yield bonds, but may also affect even more highly-rated paper at times. In any event, many of the countries with annuities markets that are developing in response to an IA reform lack a liquid and stable bond market, even for government paper.

Some simple simulations based on U.S. data suggest that the impact of interest rate uncertainty in isolation with a bond market with maturities of up to 30 years is comparatively minor, since the share of the present discounted value of the expected payments stream of payments to annuitants who retire at the age of 65 and survive more than 30 years of retirement will not be large. Interest uncertainty becomes more important with higher yield shorter maturity corporate bonds, and in countries where the maturity range for government debt is relatively short or incomplete. Interest uncertainty can also combine with the other risks an insurance company faces to inflict substantial losses.

The insurance company can choose not to match the maturity of its assets with its liabilities, and keep a substantial share of its bond portfolio at the long end when the yield curve slopes upwards on the expectation that the resulting gain in interest income may offset the capital loss when the longer-term debt is liquidated before its term. This strategy exposes the insurance company to capital loss, however, if the general level of interest rates increases by more than the company expected.21 Conversely, if maturities are not sufficiently long when annuity contracts are signed, an insurance company can suffer considerable losses by overestimating the level of future interest rates, a fate which recently befell one of the United Kingdom’s oldest life insurance companies (United Kingdom Parliament (2001)).

The inclusion of equities or other risky assets in the portfolio of assets increases the risk of capital loss, since maturity matching is not possible with equities. On the other hand, investing in equities can be justified if holding them for a long period reduces the risk of low returns to acceptable levels. As an example, it may be prudent to assume that the annuity payments made to those relatively few members of a group of retirees who reach the age of 85 or older could be financed from a specified share, invested in equities of the premiums the group paid when it retired at age 65.22 As Section IV discusses, most national regulatory agencies place fairly strict limits on the portfolio share that life insurance companies may devote to equities and other assets with an uncertain return.

The insurance company also faces some uncertainty if the number of its annuitants is not sufficiently large, since the law of large numbers may not then apply. This should not affect the larger companies, although it might affect the market for annuities to the very old, since their numbers will have declined. It may also affect companies just entering the business, to the extent that it takes time for them to build up a large policyholder base. More importantly, however, is the fact that life expectancies and survival probabilities are not predictable with certainty. As a result, life insurance companies will make relatively conservative assumptions about survival probabilities, increasing the cost of annuities above what they otherwise would be. A conservative approach is undoubtedly prudent, since underestimating the probability of survival can prove to be an expensive mistake (see the subsection “Longevity Risk” in Appendix I).

Summing up the foregoing, life insurance companies and other providers of life annuities incur a binding obligation to make a stream of fixed payments whose number is not completely predictable. The risk to them of this commitment can be minimized by holding government debt with an appropriate balance of maturities. However, few countries have markets for public debt with both long (e.g., 30-year) maturities and no gaps across the maturity spectrum. Some diversification of their portfolios to include equities and higher-yield bonds may increase the overall rate of return without increasing risk substantially, and allow life insurance companies to offer more competitively priced annuities. This strategy entails accepting both an increase in interest rate risk (because of the shorter maturities of corporate debt) and in default risk.

It may be hard to gauge the point at which an increased share of higher risk instruments increases risk unacceptably. Given the fact that it takes some time to determine whether the sale of annuities to a given cohort of retirees has been profitable, the pressure of competition can induce life insurance companies to offer premiums that expose them to considerable loss if their interest rate assumptions prove to be wrong. This practice should, in principle, be inhibited if reserves are calculated conservatively, since low premiums oblige a company to add to its reserves.23 Nonetheless, to the extent that there are pressures on insurance companies to offer excessively good terms, there could be serious consequences for the taxpayer if life annuities are protected by a public guarantee.

The structure of the life or retirement insurance industry—specifically, the degree of its concentration and the number of firms involved—varies substantially across countries. In Chile and the other South American countries that have adopted versions of the Chilean system, the number of companies tends to be small, and the degree of concentration high. Argentina is a case in point, with 20 retirement insurance companies as of June 30, 2000, of which the four largest account for about two-thirds of the market.24 Concentration ratios are much lower, and the number of firms much higher, in the larger OECD countries. The particular structure of the industry in Chile and its neighbors has implications for regulation (see below).

E. The Regulatory Framework

Life insurance companies in the countries where annuities markets have developed or are developing—basically, most of the OECD, and those countries in Latin America and Eastern Europe that have implemented an IA reform—are subject to a variety of regulatory regimes. Three areas are of particular interest for this study: regulations governing investment policy, the status of annuity holders in the event of bankruptcy or default, and the regulation of actuarial assumptions.

The investment policies of life insurance companies are normally subject to either a “prudent person” (or “prudent portfolio”) regulatory framework, or a framework of quantitative restrictions. Under the prudent person (PP) approach, the company’s investment policy is supposed to be guided by the test of what a prudent person would do in the circumstances. The PP approach could be interpreted as implying that a company’s investments should maximize the expected rate of return while keeping risk acceptably low, but in practice the assessment of portfolio quality may be based on the quality of individual investments. The quantitative restrictions approach typically sets ceilings, usually expressed in percent of total portfolio assets, on the share of the portfolio that may be invested in particular asset classes. The two approaches are not mutually exclusive, since the PP approach can and is often complemented by quantitative restrictions on portfolio composition. In particular, most countries prohibit or limit investments in the company’s own stock or the stock of affiliates.

The principal-agent approach to regulatory issues may shed light on the problem of regulation of life insurance companies, and the regulation of their portfolio allocation decisions in particular. The principals in this case comprise the policyholders and shareholders, and the agent is management.25 Principal agent problems arise with financial institutions because the principals, (i.e., the shareholder-owners or depositor or policyholder-creditors) are unable to exercise direct and effective oversight of the managers (the agents) of the institution. Such a failure of control can arise when managers have technical expertise or privileged information they can exploit to their own ends.

A standard criticism of the use of quantitative restrictions on asset holdings is that it may induce suboptimal or risky investments by even a prudent investor. This follows from elementary portfolio theory, which implies that quantitative restrictions will in general push a portfolio off its efficient allocation frontier (Shah (1998)).26 However, when applied to the investment decisions of managers, the validity of this criticism depends on the assumption that managers interests’ coincide with those of the principals. The principal agent approach suggests that this may not be the case. It does however imply that, provided effective market-based incentives for sound investment decisions exist, quantitative restrictions should not be necessary and may be harmful. These incentives are maximized when company management is well-trained; information on company performance is clear, reliable, and promptly disseminated; shareholders will punish bad management; and policyholders are financially savvy: in other words, when corporate governance is effective.

The putative superiority of the PP approach, at least in an environment of strong and effective financial regulation and supervision, has not made it the regulatory formula of choice in the case of insurance company regulation. With some exceptions, including The Netherlands and the United Kingdom, OECD members prefer to rely on quantitative restrictions rather than the PP approach in the regulation of life insurance companies (Davis (2001)). The same applies to life insurance companies in Argentina, Chile, and Peru. Some of the limits that apply to positions in equity and other risky assets might conceivably cramp a company’s asset allocation strategy. For example, Canadian life insurance companies are subject to a limit of 5–25 percent in real estate and equity investments combined; Japan limits equity holdings of life insurance companies to 30 percent and real estate holdings to 20 percent of total portfolio. In the United States, the states are responsible for insurance regulation, and they can and do place quantitative restrictions on the shares of asset classes in the portfolio. Nonetheless, and given the nature of life insurance companies’ obligations, these restrictions may not impose severe limits on their operation. Reliance on quantitative regulations may enhance the quality of regulation in countries where the regulatory agency is not yet ready to follow the PP approach (see below).

Three aspects of the legal status of annuitants in the event of bankruptcy are of particular importance: the first is whether the assets that fund annuities are treated differently from the other assets of the insurance company; the second is the nature of annuitants’ claims on a company in the event of bankruptcy; and the third is the degree of protection, if any, offered annuitants by the insurance industry or the state. Typically, the assets that fund annuities are not treated collectively like a pension fund’s assets; i.e., they do not constitute a separate entity that is distinct from the company’s other assets, and they may be attached by the company’s creditors.27 As for the annuitants’ claim on the company, in the United States the annuitant is simply another creditor, and in fact his claim is subordinated to that of the bond holder. Holders of annuities from one moderately large U.S. life insurance company received only 70 cents on the U.S. dollar for a period of 13 months after state regulators seized control of the company (United States General Accounting Office (1999)). Life annuities are not protected by insurance arrangements as occupational pensions are by the Pension Benefit Guarantee Corporation, although annuitants whose policies are not being honored may be able make a claim against an industry-operated fund that is regulated by the particular state (e.g., Delaware, New York) in which the insurance company operates. In other countries, the state has been known to intervene on an ad hoc basis to compensate policyholders of troubled life insurance companies, but there is no explicit guarantee, nor has an implicit guarantee typically been in effect.

In some countries, the approach of quantitative restrictions has gone beyond portfolio allocation and has also been applied to the demographic assumptions that life insurance companies use in calculating their premiums. In Argentina, Chile, Colombia, and Peru, insurance companies are required to use the life table stipulated by the regulatory authority. The mortality rates of the regulatory table are significantly lower than the rates in the national tables. In calculating the present value of their reserves, the companies are required to use interest rates determined by the government (Palacios and Rofman (2001)). The retirement insurance companies in Argentina must also rely on a stipulated interest rate in pricing their annuities (see Table 2).28 This means that price competition in the Argentine market is expressed solely through variations across companies in the commission they charge on their annuities.

Table 2

Required Rates of Interest for Annuity Reserve and Quotation Calculations in Argentina, Chile. Colombia, and Peru

article image
Source: Palacios and Rofmen (2001).

Regulation of the discount rate used to calculate reserves and premiums may have some unexpected consequences. In the case of Argentina, for example, it appears that an increase in market interest rates will not be reflected in the premiums annuitants pay, whether the increase is temporary or permanent. In the case of the other three countries, the limit on the interest rate used to value reserves means that, if an increase in market interest rates is reflected in lower premiums, it will be necessary to establish additional reserves to ensure that provisioning for annuities is adequate. The relatively low rate of interest used to determine reserves combines with the mortality assumptions to produce a conservative estimate of reserves (i.e., more than is likely to be necessary).

III. Potential Problems with Reliance on Private Sector Annuities

The previous section reviewed some of the basic features of the market for private annuities, and argued that some annuitization of wealth at retirement enhances welfare, especially the welfare of the improvident. The longevity insurance annuities provide should be valuable, despite the fact that the discounted present value of the payments the average annuitant receives per dollar of premium can be well below one. This section addresses the potential consequences of relying on private sector annuities to fill the void created by the reduction or elimination of the state pension. It concludes that there are some substantial drawbacks to such a strategy, particularly if the purchase of at least a minimum annuity is not mandatory.

A. Price Differentiation

One potentially problematic feature of the annuities market is the practice of price differentiation on the basis of sex, income and other personal characteristics. Life insurance companies in the United States and other countries charge women a higher premium for a given income stream than they charge men, because women live longer. In the United States, a life annuity would cost a 65-year old women about 8 percent more than it would cost a man (a marked differential is also found in the UK annuity market). The potential exists for similar differentiation on the basis of race in the United States, since life expectancies of African Americans are lower than those of the general population.29 State pension systems, however pay women the same pension they pay men with the same earnings history.30 Similarly, private pension plans in most countries may not differentiate on the basis of the sex of the plan member—only salary history and length of participation in the plan determine the value of the pension. In consequence, the average lifetime benefit a female receives from both the state and an occupational pension will be greater than that of a male with the equivalent earnings history.

The possibility of premium differentiation by sex raises the issue of the competing claims of social solidarity and individual equity. The traditional approach of solidarity can be justified on the grounds that the burden on women of their need to accumulate more savings by the time of retirement that is imposed by their greater longevity should be shared by society at large. The individualist view, in contrast might argue that the fact that women typically enjoy longer life means that equal pensions increases their lifetime income and consumption, creating an unjustified disparity between the way they and men are treated. If the purpose of a public pension system is to achieve some minimum income level during the retirement years, however, then what is relevant is annual income, not the expected value of the post-retirement income stream. The point needs to be addressed, since reliance on private sector annuities has some potentially undesirable consequences. For example, a women will have to save more during the same working life or work longer and retire later than a man to achieve a given income level in retirement. Conceivably, a reduced state pension, which would be the same for both sexes, could prove a basic level of support, complemented by an annuity financed by the funds in the IA, whose premiums would differ between the sexes.

Forbidding the use of the sex of the applicant as an index in annuity pricing would mean that male and female annuitants of the same age would be charged the same premium.31 This restriction would pose a problem for life insurance companies as well as one for public policy. Life insurance companies that are successful in obtaining as their clients a share of men higher than the share of males in the retiring population at large will have higher profits than those companies with a disproportionate number of women. If the premium life insurance companies charge per dollar of lifetime income stream is high enough to allow them to earn a normal profit on annuities for women, it will be more than enough to earn normal profits on the business of male annuitants, since the competitive premium for male annuitants would be less than that for women. The break-even unisex premium will be somewhere in between.

Hence, life insurance companies would have an interest in discouraging female applicants, if they can do so legally, and in encouraging male applicants instead. Life insurance companies could be required to accept applications of annuities from women, but any insurance company with a share of women higher than the average would, other things being equal, earn below average profits.32 It is hard to tell how great a problem this would be in practice. It might lead companies to adopt a passive attitude to female annuitants, while competing hotly through nonprice means—if that were legal—for male annuitants.33

Apart from sex (gender) and race, the development of genetic tests raises the issue, as Walliser (1998) points out, as to the limits to be placed on life insurance companies’ use of personal information in pricing its products. The issue is particularly important for life insurance, but could also become important for life annuities.

A related issue is differentiation in pricing by size of annuity. Small annuities could be more expensive per unit of income, since some administrative costs are a function of the number of policies rather than their total value. Little evidence on pricing practices is available for any country, however. If price differentiation were a problem, a law forbidding it might also lead insurance companies to discourage the applications of small annuitants.

B. Interest Uncertainty and the Annuitant

Another issue that needs to be addressed is the unpredictability of the income that a given annuity premium can buy. Annuity premiums vary with the general interest level. Even if an individual account holder can predict quite accurately the balance in her account at retirement age, she will not be able to predict the ratio of income to premium accurately until very close to the date at which she locks into her annuity. This is a problem with life annuities. Its practical significance depends on the variability of the interest yield curve. Fluctuations in short-term interest rates should not have a marked impact on the price of life annuities, unless they are associated with fluctuations all along the maturity spectrum. The impact of the problem can be mitigated by making the rules governing withdrawals from IAs more flexible. One approach would be to allow the deferral of the date at which an annuity is to be purchased, with programmed withdrawals from the IA during the period between the date when the funds in the IA were first available to purchase an annuity and the date at which purchase actually takes place. This would at least allow the IA holder to benefit from an increase in interest rates in the years immediately following the date when she would normally annuitize. The strategy would not guarantee her any given premium rate, however. Another strategy would be to allow the purchase of one or more deferred annuities some years before retirement. Blake (1999) notes that deferred annuities in the United Kingdom have proved to be very expensive.34 Neither strategy, singly or in combination, would ensure that different cohorts of retirees would all obtain the same income per dollar, pound, or peso per unit of premium, however. The regulated interest rates that characterize the Argentine system, while they may limit competition between annuity providers, would at least ensure this uniformity.

C. Inflation and Indexed Annuities

A fixed income life annuity would be a pretty close substitute for the typical state pension if it were indexed. The evidence implies that indexed annuities, where they are available, come at an extra cost, however. Studies of the MWR for indexed annuities in Chile, Israel, and the United Kingdom find that indexation can reduce the MWR substantially (see Tables 3 and 4).

Table 3

MWR Values for a Group of Countries: Indexed Annuities

article image
Source: James, Estelle and Dimitri Vittas, “Do Consumers Get Their Money’s Worth?” paper prepared for Bank Conference, “New Ideas on Old-age Security”, September 1999Note: All annuities are immediate single payment life annuities.
Table 4

United Kingdom: MWR for Nominal and Inflation-Indexed Annuities

article image
Source: Brown, Jeffrey R, Michell, Olivia S., and Poterba, James M., 1999, “The Role of Real Annuities and Indexed Bonds in an Individual Accounts Retirement Program” NBER Working Paper 7005 (March).Note: The estimates are based on a sample of 14 large insurance companies. The data were provided By Annuity Direcg, Ltd. The reference date is 8/21/98. All annuities have a five-year guarantee period.

The difference between the MWRs for indexed and nominal annuities could reflect differences in the size of the markets and the completeness of maturities for the securities that fund these instruments.

Indexation is an important property of a pension, since even a comparatively low rate of inflation can substantially lower the real value of a pension or a life annuity over the course of retirement. When inflation is fairly predictable, which is more likely when its average rate is low, part of the annuity payments can be saved for later years. The rate of return to such saving will be less, however, than the implicit conditional rate of return of the annuity (assuming the annuitant survives). This home-grown approach to inflation insurance requires some financial expertise as well as self-discipline.

More importantly, the approach remains vulnerable to inflationary shocks. A highly variable rate of inflation can result in substantial unforeseen declines in real income. For example, if an individual plans for an annual rate of inflation of 2½ percent, and the actual rate averages 4 percent, over a 15-year period the real value of the annuity payment is reduced by 20 percent. An indexed annuity, whether provided by the state or the private sector, thus provides valuable protection.

There is in principle no reason why the private sector cannot provide an indexed annuity, provided the state supplies indexed bonds. In doing so, the state is assuming the risk of inflation.35 In addition, the need to match the maturity structure of liabilities with assets holds for indexed debt just as much as it does for fixed income (nominal) debt, so that indexed debt needs to be available over the whole range of maturities. In the case of the United States, nominal debt is available at maturities that range continuously from short term to almost 30 years. There are gaps in the maturity spectrum for indexed debt because it has not been issued for long, or at many maturities. The indexed debt issued by the UK Treasury spans the same maturity spectrum as nominal debt. Price (1997) reports that in practice the longest maturity of government indexed bonds in Chile is 20 years.

Advocates of privatized pension systems maintain that privatized systems are not subject to political risk—or to the same degree of political risk—as state systems. One of the arguments made in support of this view is that the real value of state pensions in many countries has been highly variable. This was especially true of Latin American pension systems in the 1980s. The bouts of hyperinflation that beset Argentina during that decade, for example, caused huge fluctuations in the real value of the average pension despite its being officially indexed.

Nonetheless, indexed debt is not entirely free of political risk. Governments that have difficulty indexing state pensions may well have difficulty honoring their commitment to indexed debt. Conceivably, a government in the throes of a serious fiscal crisis that has issued indexed bonds might not be able to honor them.36 The issue of indexed debt, which appears to be necessary for the private sector to offer inflation insurance is not a sure guarantee, in other words that the insurance obligation will be honored in every circumstance. This point needs to be borne in mind in the case of countries where a tradition of sound money is not well established.

D. The Consequences of Shortsighted Behavior

A basic argument for social security, and one that applies to any system, is that some individuals—perhaps a significant share of the population—need to be protected from their own shortsightedness. Left to their own devices, people will not save enough for their old age.37 A further argument is based on the state’s presumed obligation to support the destitute elderly. In these circumstances, a public pension program is a means of avoiding a free rider problem. Libertarians challenge this defense of social security on the grounds that it entails a paternalistic interference by the state in the economy and in citizens’ lives.

Whatever the merits of the latter view, there is plenty of informal evidence that people have difficulty understanding how to plan for a secure retirement. Even the well-educated can be naïve about financial matters, and an IA reform will involve everyone, including functional illiterates with no real understanding of basic personal finances.38 Even in the advanced economies, some social security recipients will have never bought a house, or taken out a mortgage. Some will be marginally competent, even if they have been able to hold down a job.

The diminished role of the state pension requires retirees to decide whether or not they wish to compensate for the resulting drop in the degree of annuitization of their wealth. Persons whose wealth was overannuitized to begin with will choose not to, and the decision will enhance their welfare. Others, with little or no annuitized wealth may also choose not to, and there is a danger that their decision will jeopardize their welfare. This risk is present even if the level of wealth upon retirement is adequate to maintain an individual’s habitual standard of living. By luck or by design, a person may have enough wealth (including wealth in the form of annuities) to finance an adequate rate of expenditure in retirement even if he lives to a ripe old age, and still encroach on it too quickly.

The argument can be made that in countries with generally well educated populations and competitive and well developed financial markets that individuals will have no shortage of good financial advice, and that competitive pressures and appropriate legislation will minimize the risk of the unsophisticated being persuaded to make unsound decisions. That said, there is no shortage of bad advice and little or no regulation of financial advisors. In Chile, where the limited options at the distribution stage should aid sound decision-making, there have been instances of collusion between the salesman and the IA holder to increase the share of the accumulated funds in the IA that the holder could withdraw in a lump sum. The IA holder in one instance received a sum that was worth far less (measured at any reasonable discount rate) than the income stream he gave up (Palacios and Rofman(2001)). In any case, even prudent advice need not stop someone from spending at an excessive rate.

E. The Pros and Cons of Mandatory Annuitization

Both the potential benefits and the costs of mandatory annuitization will depend on the share of the wealth of retiring persons it affects, and the generality of the policy’s application—the flexibility with which it is applied. The main drawback of annuitization is that it reduces the welfare of individuals whose wealth is highly annuitized to begin with. This potential welfare loss is caused in part by an individual’s lack of full control over the degree to which his wealth is annuitized. A participant in a company pension plan, for example, may not be able to choose a lump sum or programmed series of distributions over the standard life annuity. Some arbitrage is possible. For example, a retired person on a pension planning to leave a bequest can acquire life insurance. However, in general, it is unlikely that everybody will be able to reverse the impact of mandatory annuitization in this fashion.

Mandatory annuitization could be very harmful for the terminally ill. If the administrative complexities of such an exception are overlooked, a case can be made for making an exception to a general rule in their favor.39 It can also be argued that annuitization is disadvantageous for those groups in a population whose life expectancy is significantly lower than that of the general population, unless the premiums of the annuities that members of these groups buy reflect their lower life expectancies. The redistributive income of social security on the lifetime incomes of the poor may be offset to some degree by their lower life expectancies.40 However, calculations of the implicit rate of return of annuities for different classes of the population do not capture the insurance value of annuities, as Section II explains. (The argument can be also be made, as already discussed, that the standard form of the state pension discriminates against men in favor of women because the difference in life expectancies is not reflected in the value of the pension.)

Mandatory annuitization could also confer important benefits. First, by bringing the general population into the pool of annuitants, it reduces adverse selection, and lowers average life expectancies of the pool. This should lower premiums. Walliser (2000) has estimated that the drop in adverse selection resulting from social security privatization would lower premiums by 2–3 percent in the United States. However, a simple comparison of the MWRs calculated by Mitchell et al (1999) and James and Vittas (1999) for the general population and those calculated for the population of annuitants suggest the gains might be greater than this (see Table 1). The gains would be greater still if a system of group annuitization could be put in place. This idea is discussed further in Section V. Second, it should reduce the risk of poverty among those elderly whose wealth would otherwise not be annuitized, and who lack the discipline, foresight or good luck to make their resources last through old age.

Although the issue of mandatory annuitization can be seen in black and white terms—make the purchase of an annuity mandatory, don’t make it mandatory—it is more constructive to make the question one of degree, and to address questions such as: what share of the funds accumulated in an IA should be annuitized; and what exceptions should be allowed. Many of the arguments advanced against mandatory annuitization are really arguments against excess annuitization. The lower the share of the funds accumulated in an IA that must be used to purchase an annuity, the less the risk of over-annuitization, the higher is the minimum bequest an individual leaves to his heirs, and the easier it is for him to finance exceptional but necessary expenditure. On the other hand, the lower the share of annuitization, the less the amount of longevity insurance an individual enjoys, and the more he has to rely on self-control and discipline to avoid improvident spending. A decision as to the degree of annuitization should be informed by reliable information on the level and composition of wealth upon retirement, which will reveal the extent to which individuals already have annuitized income. Countries lacking reliable information on the role of annuitized wealth in the total wealth of individuals of retirement age and planning or implementing an individual accounts reform should place a high priority on conducting a survey of individual wealth.

The administrative demands of a policy that permits even very limited exceptions to a rule of mandatory annuitization should not be overlooked. It would be important not to snarl up the exemptions process in red tape. At the same time, safeguards would have to be built in to guard against the abuses to which a policy of this sort can be subject.41 This and other administrative aspects of a IA reform with mandatory annuitization are discussed further in Section V.

If mandatory annuitization is necessary, how much is enough? One possible answer is that the share of the funds in an IA that should be annuitized should be whatever is necessary to buy a life annuity with payments that when combined with the (reduced) state pension will equal or exceed some stipulated minimum, related perhaps to the poverty line (in the United States) or some measure of subsistence income. Funds exceeding this amount would be freely withdrawn. Under this approach, mandatory annuitization would not aim at achieving some stipulated ratio of income in retirement to income in working life (a minimum replacement ratio), but instead would aim at ensuring a minimum retirement pension in absolute terms. This approach, which might be complemented by an exemption for the terminally ill if that were feasible administratively would be a way of addressing the problem of excess annuitization.

Requiring the purchase of an annuity that generates a minimum income level may not, however, provide the needed amount of longevity insurance for a large share of the population. For individuals whose working life income was well above subsistence, the minimum pension might replace only a small part of that income, and conceivably provide too little longevity insurance (as well as protection against improvidence). This problem could be addressed by requiring that the share of the funds in an IA devoted to an annuity, in combination with the state pension, be enough to achieve both a floor on income and, over a certain range of income, a minimum replacement ratio. Typically, this is what state pensions do. In Chile, the minimum amount of the funds accumulated in the IA that must be annuitized is that amount which would guarantee a pension equal to the greater of 120 percent of the minimum pension and 70 percent of the account holder’s average salary over the last ten years of his working life. Any excess may be withdrawn at the account-holder’s discretion.42

Both of these approaches have to contend with the problem of shortfalls in the accumulated funds in an IA, particularly if the government guarantees a minimum income level. Any guaranteed income level creates a potential moral hazard problem, especially an unconditional and full guarantee. The potential seriousness of the problem would depend on how high the minimum income level (in absolute terms) and/or minimum replacement ratio were set and on the income range over which the latter would apply. In the case of the minimum replacement ratio approach, a full guarantee—i.e., a guarantee that the account holder would be entitled to an annuity with annual income satisfying the minimum replacement ratio— might be very costly, since it might stimulate excessively risky investments by the account holder prior to retirement.43

The need to contain the risk of moral hazard might call for only a partial guarantee. The cost to society of an offer of an minimum pension would be reduced by making the terms of the guarantee consistent with the existing social safety net. Integration would imply that there would be no separate minimum guarantee for the minimum annuity. Instead the rules of the social assistance program would apply, and state assistance would be provided only if the sum of the annuity funded by an individual account and other income fell short of the minimum income under social assistance.

A further decision needs to be made regarding indexation: should the annuity that account holders are required to buy be indexed, or should the account holders have the choice of a fixed or an indexed annuity? Given the problems of self-insuring against inflation, a strong case can be made for at least allowing the purchase of an indexed annuity, if not requiring it. If the purchase of an indexed annuity is required, it will be important that such annuities be available on competitive terms. Inter alia, this requires an active market for indexed public debt covering the full maturity spectrum. Developing a deep market for indexed debt could take some time. Moreover, as long as the flow of new indexed debt issues is large relative to the existing stock, changes in the public sector borrowing requirement could have a substantial impact on the rate of growth of the market, and might cause significant fluctuations in the yield of indexed debt.

Indexed annuities may appear unattractive to prospective annuitants because their initial nominal values have to be well below the fixed value of a nominal annuity. Notwithstanding this deceptive difference in starting values, the indexed life annuity that most governments provide is the only such annuity available to most people in most countries. If the choice of one or the other is allowed, the minimum annuity income requirement has to be adjusted, in the case of an indexed annuity, to take account of the lower initial income. There are precedents for an indexation requirement: Australia, the United Kingdom, and Germany require that private pensions be indexed.44

Some other issues of design arise from the nature of annuities and the annuities market. Since the income stream generated per dollar, pound, or peso of premium will vary with the level of interest rates, so will the value of the premium required to purchase an annuity that conforms to a minimum annuity income requirement. Similarly, the balance in the IA of a woman needed to acquire a minimum annuity will exceed that of a man with an identical earnings history. These consequences of annuitization might be considered problematic, and could be avoided by expressing the minimum requirement in terms of the amount of funds in the IA. The rationale for this approach would not be obvious, however, since security in retirement depends on the level of income that may be sustained during retirement, not on the amount of wealth as of the start of retirement. Consequently, using a measure of income as the target variable is preferable.

IV. Implications of Increased Annuitization for Regulation of Annuity Providers and the Public Information Function

A. Regulatory Issues

Whether or not annuitization is made mandatory, the increase in the role of the private annuities market that an IA reform will foster increases a society’s stake in sound and honest financial management of life insurance companies and other annuity providers. With mandatory annuitization, effective financial regulation and supervision becomes essential, since otherwise the purpose of annuitization will be defeated.

The sound regulation of the annuities market requires the sound regulation of the insurance market. Annuities are normally sold by life insurance companies in both advanced and emerging market economies, but arc only one of a number of financial products they sell. The assets that back them are typically not segregated or managed separately from the other assets of the company. This means that regulation must promote the sound financial management of the life insurance business as a whole.45

Life insurance companies in OECD member countries have typically been subject to both a prudent person rule and to quantitative restrictions on their investment choices. The role of quantitative restrictions on the asset allocation decisions of insurance companies has been more significant than it has been with the portfolios of pension plans. As Section II explained, there are some differences of view as to whether such restrictions are appropriate. One difficulty may be the fact that a given set of restrictions may not be appropriate for all life insurance companies, since the mix of their business may vary in ways that calls for different asset allocations. For example, some insurance companies in OECD countries offer what are essentially investment vehicles—e.g., variable annuities—where it is understood that the client shares both the gains and the losses from the securities in which the company invests. Others emphasize more traditional life insurance products, and the asset allocation that is appropriate for the second group of companies may not be right for the first. This issue needs to be investigated. It is hard to tailor a set of limits on asset allocations that will be optimal for even one insurance company, let alone all of them, unless the limits are so broad that they cease to be binding. Broad limits can nonetheless serve as guidelines, signaling to insurance companies that a large concentration of assets in an asset class like common stock will generally be inappropriate.

Competent regulation of the life insurance business is no easy matter in any country. Life insurance is a technically demanding business, and the regulatory office would have to be staffed by individuals with a wide range of skills, including actuarial science and financial analysis, and ideally with some relevant industry experience. Recruiting a skilled staff might prove challenging in very small countries with a dearth of skilled financial professionals, particularly if the supervisory agency is unable to pay its skilled employees salaries that are comparable to what they could earn in the private sector or abroad.46

The regulation of life insurance in the smaller emerging market countries may pose special challenges. If the experience of the Latin American countries with defined-contributions systems is anything to go by, the number of insurance companies in countries following their lead will be small and the degree of concentration high. The small number of firms will facilitate communication, and reduce the workload of the regulatory agency. The relatively high degree of concentration may increase the cost of premiums unnecessarily, however.

The question arises as to whether the case for quantitative restrictions is greater in such countries than elsewhere. The key question is: do such regulations facilitate the task of solvency regulation? The answer is not clear cut. The goal of solvency regulation should be to maximize the chances that a life insurance company’s portfolio has the right maturity composition and quality to keep interest rate risk at acceptable levels, and to ensure that reserves are large enough to minimize longevity or mortality risk. In principle, and provided the regulatory authority has sufficient clout, this can be done by analyzing the composition of an insurance company’s assets directly to see whether the degree of immunization to interest risk is adequate, whether provisioning reflects the riskiness of the company’s portfolio, etc. Such an approach will require a skilled if not a large staff. Currie (undated) notes that many off-the-shelf software products are available to both companies and regulatory agencies to test the strength of insurance companies’ financial positions.

The use of quantitative restrictions and the restrictions that apply in some countries on the choice of life tables and the interest rate used to calculate reserves can be interpreted as devices to ease the task of regulation by minimizing the risk of default and by simplifying financial decision-making (i.e., if a company cannot invest in equity or real estate the task of its investment department and that of the regulatory agency is that much easier; if there is no choice regarding the life table the demand for actuarial services by both industry and the regulatory agency is reduced). By the same token, these restrictions discourage healthy competition. It is possible, for example, that the premiums for annuities may be set unnecessarily high, both because the life table that the industry is required to use has lower mortality rates, and because future annuity payments must be discounted at a relatively low rate in pricing them. If reserves held against the risk of default and other contingencies are relatively high, and the interest rate applied to them is also low, the premium is further increased, which could result in super-normal profits. Whether it does or not depends on how commissions are determined, since this will be the only element of the price that will be free to vary. Conceivably there may be limits on how low commissions will go, since the role of individual salesmanship is important in the defined-contributions systems of South America.

This brief discussion suggests that quantitative restrictions entail a trade off. They may simplify the regulatory task and reduce the risk of insolvency, particularly in the early stages of the development of the life insurance business, or the annuity component of that business. Once the regulatory function matures, they may become onerous, and the trade off becomes unfavorable. Interestingly, regulation in Chile has come to rely less and less on quantitative and other restrictions. It may be that there is a natural evolution, as Doyle and Piggott (2001) suggest, away from an initial dependence on quantitative restrictions.

The increased importance of annuities as a source of retirement income may be of no particular consequence for the regulatory function in advanced economies, if financial regulation there may be safely assumed to be effective. On the other hand, mandatory annuitization could conceivably be coupled with a state guarantee on annuity payments in the event that the annuity provider is unable to fulfill its contractual obligations. If there is no explicit guarantee, political pressures may lead to an implicit guarantee.47 Among countries with IA systems that have reached the annuitization stage or have made a decision on the matter, Chile guarantees the life annuity payment in the event of the insolvency of the insurance company for 100 percent of its value up to the minimum pension, and for 75 percent of the difference between the annuity payment and the minimum pension up to a specified ceiling. Argentina guarantees 100 percent of the value of the annuity payment up to a limit of five times the so-called “maximum basic pension,” which amounts to about 1.6 times the average economy-wide wage (Palacios and Rofman (2001)). Peru has no guarantee. Poland, however, was planning a full guarantee.

Just as a state guarantee of an annuity of some minimum size at retirement creates an incentive for risky investment behavior by the holder of an IA, so a guarantee on the income an annuity provides creates an incentive for risky investment behavior by the insurance company, and thus a potentially serious risk of moral hazard. The moral hazard risk created by an annuity guarantee is similar to that created by deposit insurance. In particular, life insurance companies could offer lower premiums on their annuities, which they would finance with investments with a relatively high mean return and risk. If all goes well, the investments are successful, and the annuity payments are honored. If things go badly, the state becomes a partner with the insurance company, and the general taxpayer shares the loss. There is little evidence on the point. Mitchell et al (1999) in a study of the U.S. market found no relationship between low premiums and poor ratings of the insurance company offering the premium.48

The design of deposit insurance schemes tries to strike a balance between the goals of preventing banking system panics and avoiding creating a stimulus to the kind of behavior that contributed to the Savings and Loan debacle in the United States. As far as deposit insurance itself is concerned, the imposition of a cap can afford some basic protection to the smaller saver and business, while discouraging the holding of larger sums that would normally be invested in other financial instruments. However, a deposit insurance scheme should be only one part of a regulatory framework that creates appropriate incentives for sound lending by bank management and effective surveillance by shareholders, while encouraging depositors to hold accounts with the better managed banks.49

The provision of annuities, even when these are mandatory, may not be a public good in the strict sense of the term. On the other hand, it can be argued that the state would have a moral responsibility—and would certainly be under political pressure—to ensure sound investment and management by annuity providers. As with deposit insurance, the provision of a limited and not an open-ended guarantee might best balance the interests of the annuitant and the taxpayer. That said, a guarantee with a ceiling (like those offered in Argentina and Chile) might be seen by the public as inequitable in a system that did not limit in a consistent manner the amount of the funds in the IA to be subject to mandatory annuitization.50 The combination of a full guarantee with mandatory annuitization that targeted a comparatively high replacement ratio would be very unwise.

B. The Public Information Function

The debate over an IA reform has emphasized the dangers of ill-advised investment choices by IA holders during the accumulation phase. Similar dangers lurk at the distribution phase. The growth in the market for annuities will bring a great many people into the market with little or no knowledge of financial issues in both advanced and middle-income economies. The standard of financial literacy may be higher in the former group, but the range of financial products on the market and the sheer volume of the information available on retirement planning is bewildering. Nor will there be any shortage of sales representatives eager to sell annuity products when an IA reform increases demand for them. Is there a case for regulation of the provision of this information, and of the sales function, or for the public provision of information on annuities and financial planning?

The basic argument for the state’s provision of information on privately produced and marketed goods and services is that either the quantity or quality (including coverage) of information provided by the private industry is inadequate. Casual inspection of websites and other sources suggests that there is no problem with the sheer quantity of information available on annuities sold in countries with well-developed markets. As for quality, it is certainly possible to obtain clear and comprehensive discussions of annuities at various websites. What is less certain is how many potential annuitants will have access to or will visit them. There may be a role for the state in the direct provision of similar information to individuals approaching retirement, since many may lack access to or be unaware of the relevant sources.

If annuitization is mandatory, and the form the eligible annuities may take is restricted (for example, to life annuities with a spousal right of survivorship), then the information to be provided could be relatively simple, particularly if the goal of the information were simply to help the prospective annuitant make an informed decision regarding the options mandatory annuitization allowed. In the extreme case of group annuitization, (see below), where the individual was not the negotiating agent, there would be no compelling need for advice and council on the choice of annuity. A statement of the rationale of the policy of group annuitization as well as general information on how the annuity would fit into an individual’s finances might nonetheless be helpful. It would also be helpful to explain, in the case of partial annuitization, how the share of the balance in the IA to be annuitized is determined.

If annuities are to be sold individually as well as on a group basis, it will be important to see that annuity salespersons confront the right incentives when they promote their companies’ products.51 In some countries, salespersons are required to acknowledge that they work on commission, and to prepare an account of their discussion of the clients’ needs and product characteristics with a copy to the client (Currie (undated)). Typically, the sales and other marketing functions are not regulated by the agency overseeing the financial position of the insurance companies.52

V. Family and Tax Issues and Administrative Issues

A. Family and Tax Issues

The paper has stressed the state pension’s role as an indexed life annuity, arguing that the longevity and inflation insurance it provides are of great value. Nonetheless, the typical state pension has also been designed to benefit vulnerable members of the contributor’s family and to redistribute income in favor of the low-paid. For example, in most OECD and middle-income countries, the state pension provides an additional benefit to the widow/widower and surviving children of a contributor, but in the form of an additional pension, and not an actuarially fair joint survivors’ annuity. Benefits will also often be paid to the homemaker spouse of an insured worker. In the United States, for example, if the spouse has never worked outside the home she is eligible at age 65 for a pension equal to one-half of the social security pension her spouse receives. If she has worked outside the home she is eligible for the greater of 50 percent of her spouse’s benefit or the benefit to which her own contributions would entitle her.

Critics of the U.S. system’s treatment of the spouse point out that it was conceived at a time when most social security contributors were men with a homemaker wife. The increase in the divorce rate and in the role of two-earner and single or divorced female households means that the spousal benefit is ill conceived for the current family. Whatever its merits, the provision means that the implicit rate of return to contributions from a married worker can be much higher than the implicit return to contributions from a single worker with exactly the same earnings history. Similarly, the return to the contributions of a two earner household will be less than that of a single earner husband and wife household with the same income. Similar departures from actuarial fairness characterize the pension systems of many other countries (Thompson and Carasso (2002)). A decision would have to be made whether to retain the spousal benefit or phase it out under an IA reform.

Social security systems also aim to protect women from the economic consequences of divorce. An issue to be addressed is whether the annuity or other income funded by the IA system should be accorded special status, or treated like other marital property. If it is treated like other marital property, then its disposition in the event of divorce is covered in most countries by the provisions of the separation agreement. An alternative treatment is to stipulate that that part of an annuity payment that can be attributed to the increase in the balance of IA over the period of the marriage (or perhaps less fairly, but more simply, the entire annuity payment) be split in half in the event of the divorce, regardless of the legal form of the title in which the IA is held. Several OECD countries now apply a “splitting” system. It is particularly important in the case of a spouse without a long history of work outside the home, and without a substantial pension of her own.

An IA reform will also need to address the issue of the tax treatment of social security benefits. Typically, the state pension receives favorable treatment. In the United States, 15 percent of the benefit is exempt from tax regardless of the overall income of the recipient. The benefits are totally exempt for recipients whose incomes fall below a certain threshold: US$32,000 in the case of a recipient taxpayer filing jointly.53 This treatment increases social security’s redistributive character, and might also be justified by the absence of income tax relief for employee payroll tax contributions.

Broadly speaking, in taxing annuity or pension income, a country can follow either the consumption tax model or the income tax model. Under the income tax model, contributions are taxed as part of current income, and the income component of the pension benefit is taxed when it is paid. The income that contributions earn during the accumulation phase is generally not taxed. Under the consumption tax model, either contributions or the pension benefit is taxed, but not both.54 To be consistent with the consumption model, if contributions to an IA are not taxable, the income from the annuity or other distribution should be.

The taxation of pension and annuity income is itself not straightforward, however, since the periodic payment the pensioner or annuitant receives includes both a return of principal and accumulated earnings. The share of principal repayment in the payment is small to begin with, and rises over time, like the share of principal repayments in a monthly mortgage payment. In addition, the amortization period (the period over which the principal is repaid) varies depending on the age and sex of the annuitant, and its length is uncertain. Consequently, the average share of income from an annuity that should be taxed should vary with both the age and the sex of the annuitant. In the United States, the IRS Publication 939, which sets out the general rules and tax schedules for taxation of pensions and annuities runs to more than 75 pages (United States Department of the Treasury (1997).

This brief discussion has simply flagged the great importance of these two issues. They are worthy of separate papers in themselves.

B. Administrative Aspects of Reform

The growth expected for the market for annuities in countries undertaking an IA reform will require a review of the current regulatory and supervisory framework. This review should take place even in the laissez-faire case of no guarantees and no mandatory annuitization. It will be important to ensure that the regulatory and supervisory framework applying to life insurance companies adequately protects annuitants.

If the purchase of a private sector annuity becomes mandatory, if complicated restrictions apply to the form that the annuity may take and if a minimum guarantee on its payment is provided, then the preparatory work for reform will be much more substantial. Specifically, successful reform will require the establishment of new systems and procedures for the monitoring, control and regulation of the annuities market and its participants. Setting up these systems and procedures will pose demands on both the financial institutions engaged in offering IA accounts and annuities, and on the government itself. The following would be the main tasks of the new systems:

  1. An IA holder education system to educate IA holders as to their rights and responsibilities regarding annuitization. To the extent that a choice was allowed among different forms of annuities, the merits and drawbacks of each form would have to be very carefully explained. A decision would need to be made as to the role played by annuity providers in providing prospective annuitants with information.55 The educational campaign would have to begin some years before annuitization took place.

  2. A notification system (coordinated with the educational program) to inform prospective annuitants of the deadlines for important decisions well in advance of the deadlines.

  3. A registration system for mandatory purchases of annuities, and—depending on whether the payments process could be self-policing or not—a system to record and monitor payments to annuitants. Procedures will also have to be worked out to deal with deferred annuity purchases, and the consequent partial withdrawals.

  4. A system to ensure that annuities were purchased only from eligible institutions, and a means of vetting these institutions. The vetting could be handled automatically by the agency responsible for regulating life insurance companies, if they were to be the only providers of annuities.

  5. A system to verify that annuities conformed to the model prescribed by law.

  6. Arrangements for financing payments by the government to increase the balance in an IA at annuitization to the minimum guaranteed by law.

  7. Arrangements for the financing of the fund that would guarantee the payment of annuities by insurance companies unable to honor their commitments (if there were to be such a guarantee).

  8. Rules to determine the conditions under which an individual could be exempt from the requirement to purchase an annuity, and procedures to vet applications for exemptions to mandatory annuitization (if exemptions were allowed). These procedures could perhaps piggy-back on existing procedures for determining eligibility for a disability pension.

Chile was the first country to adopt an IA system, and one with substantial restrictions on the form and amount of the withdrawals that may be made from an IA. Its experience might suggest that the demands of good administration are not overwhelming. On the other hand, the characteristics of the Chilean annuity market would argue for caution in drawing lessons for other countries. The number of firms involved in the Chilean market is small (there are only seven pension fund companies and about 27 life insurance companies); and the variety of annuity products from which the choice is made is essentially limited to two. Moreover, as noted, life insurance companies must all use the same life table and the same interest rate in calculating reserves. The small size of the market, the small number of institutions involved and the restricted choice allowed prospective annuitants in Chile all facilitate the administration of the system. Interestingly, some observers of the Chilean system maintain that prospective annuitants have not always been fully aware of the opportunities open to them (Palacios and Rofman (2001)).

All reforms involve issues of sequencing. In the case of an IA reform, it appears that all of these components of the regulatory apparatus would have to be in place before a single IA holder withdrew his funds. The speed with which the market for annuities would grow would depend on the way the reform was implemented, however. There could be a substantial interval between the inception of an IA system and the signing of the first annuity contract. The length of time will depend mainly on the treatment of middle-aged contributors to the state system, and on the treatment of their accrued pension rights under that system. For example, if workers who are close to retirement do not have the choice of moving to the new IA system, their retirement cannot create a demand for annuities.

The treatment of accrued pension rights is also important. Under a Chilean style reform, participants in the old state system receive a “recognition bond” that is intended to represent the present value of the claims they had accumulated under the state system. When such former participants retire from the new IA system, the bond is credited to their account, and can be used to fund an annuity (or a series of programmed withdrawals). The issue of these bonds makes the private annuities market grow more rapidly than it otherwise would. The recognition bond approach is not the only way to deal with accrued pension rights, however. Participants in the old system could simply have been paid a pension by the state whose amount was determined by their earnings history under the old system. In general, the speed with which the annuity market can be expected to grow is of relevance to decisions regarding the budgets of those governmental agencies involved in regulation. Nonetheless, the administrative apparatus needs to be in place from the time the very first annuity contract is signed.

The complexity of the annuitization process might be greatly reduced, along with the cost of annuities, if a system of group purchases of annuities could be established.56 Group annuitization might work as follows: the government could group annuitants of a given age cohort in subcohorts of some minimum number, using a selection procedure that would insure that different sub-cohorts would have the same average life expectancy.57 It would then invite bids from life insurance companies. The combination of the economies of scale entailed by grouping and the elimination of adverse selection would reduce the cost of annuities substantially. This could be particularly important for small annuities. The administration of the annuity contracts could be carried out by the life insurance companies who had bid successfully for the business, or might even be stripped off, and undertaken by the government agency responsible for administering the state pension (i.e., the Social Security Administration in the United States). In this latter case, the life insurance companies would bear interest rate and longevity risks, but not the costs of administration. They would, however, be responsible for making periodic (presumably monthly) payments to the administrative agency (which would in turn have to advise the life insurance companies on the size of the payment, and monitor deaths of annuitants).58

Finally, given the importance of sound regulation and the other prerequisites for the successful provision of annuities by the private sector, it is conceivable that in countries where the conditions for private sector provision were not yet opportune, the state should continue to be the sole supplier of mandatory annuities. Even in countries where private sector provision is deemed to be feasible, continued public sector provision may be the best way to address the various concerns this paper has explored.