World Bank Pension Reform Primer

This briefing is part of the World Bank’s Pension Reform Primer: a comprehensive, up-to-date resource for peopledesigning and implementing pension reforms around the world. For more information, please contact SocialProtection, Human Development Network, World Bank, 1818 H Street NW, Washington, D.C. 20433; telephone+1 202 458 5267; fax +1 202 614 0471; e-mail socialprotection@worldbank.org. All Pension Reform Primermaterial is available on the internet at www.worldbank.org/pensions

AnnuitiesRegulating withdrawals from individual pension accounts

ension, to most people, implies a regularpayment from a specific age—such as

retirement—until death. Individual retirement ac-counts are a vehicle for retirement savings but theydo not become a pension in the conventionalsense of the word until they are converted to an‘annuity’. How much and what type of annuitiza-tion should be mandated are key policy questionsfacing reformers.

The value of annuitiesEconomists believe that annuities can make peoplebetter off. The intuition is straightforward. Lifeexpectancy is normally uncertain. So peoplewould have to spend accumulated wealth slowlyafter retirement to ensure an adequate incomeshould they live a long time. This kind of self-in-surance is costly because it increases the chancesthat people will consume less than they could haveif they knew when they were going to die. Thiscost can be reduced with annuities, which pool riskacross individuals.

An annuity is a kind of insurance against the riskof exhausting savings in old age. The benefit ofthis ‘longevity insurance’ depends on how conser-vative people are. More cautious individualswould spend less of their savings in the early yearsof retirement if there were no annuities as theysought to avoid any chance of running out ofmoney toward the end of their lives. The benefitalso depends on interest rates, life expectancy andhow much people plan for the long term. Under

reasonable assumptions about each of these vari-ables, an annuity has been estimated to be worth50-100 per cent of wealth at age 65.

Annuities can raise welfare 1

65 70 75 80 85 90 95 100 105 110

consumption

age

consumption pathwith perfect annuity market

consumption path with no annuity market

Demand for annuitiesGiven these impressive magnitudes, it would seemsafe to expect that there would be significant de-mand for annuities. But in fact, actual demand isquite limited. This is because the decision whetherto buy an annuity is affected by transaction costs,market imperfections and other factors, whichwere not considered above. Understanding thesereasons is important when thinking about howgovernments should intervene in the benefitwithdrawal stage of an individual accounts system.

P

Annuities 2

Adverse selectionOne possible explanation for low annuity demandis a market failure known as ‘adverse selection’.The potential for adverse selection is often used tojustify government intervention in annuitiesmarkets.

Adverse selection can occur when people knowsomething about their mortality risk that annuityproviders find costly or impossible to find out.This information asymmetry means that peoplewith higher mortality could expect to lose outfrom buying an annuity. The average lifeexpectancy of annuitants increases, so providersmust raise the price. This drives still more peopleout of the market. The market fails, because somepeople are unable to buy a fairly priced annuity.

A number of studies have documented annuityprices significantly higher than those that would becharged if insurance companies were to base theircalculations on the relevant interest rates and pro-jected population mortality. In other words,annuity prices were not ‘actuarially fair’. Depend-ing on the discount rate applied, the premium paidby annuitants in the United Kingdom and theUnited States was typically between 7 and 15 percent. This evidence, combined with the observa-tion that annuitants live longer than the generalpopulation, provides support for the market failureexplanation.

Other factors reducing demandBut there are many other potential explanationsfor underdeveloped annuity markets. These fallinto two categories: factors reducing the desirabil-ity of longevity insurance or means of providing aviable alternative to annuity products offered inthe market.

We begin with bequests. Standard life annuities are,by definition, exhausted when people die. Yetpeople often want to leave some of their wealth totheir family or even to charity. As well as concernfor their family’s well-being, bequests can be usedto encourage relatives to look after them in theirold age in exchange for the promise of the inheri-tance. Bequests, whether strategic or altruistic, canreduce the usefulness of annuities.

Precautionary savings can also reduce the demand forannuities. A sudden medical emergency requiresliquidity and flexibility that is impossible if wealthis fully annuitized. In the absence of health insur-ance, this motive can be a serious disincentive topurchasing an annuity.

There are at least two important substitutes forannuities purchased from private insurers. Thefirst is a public pension. In the United Kingdom andthe United States, more than half of the averagehousehold’s wealth is held in the form of a publicpension. This proportion is even higher in coun-tries with more generous benefits, such as France,Germany and Italy.

The second substitute—the family—can bedescribed as an ‘incomplete’ annuities market. Intheory, even a small family unit can make informalarrangements providing much of the benefit ofbuying an annuity. The advantages of keeping it inthe family include low monitoring and transactioncosts. And depending on the social sanctions thatare possible, enforcement mechanisms in thisinformal market may be very effective.Simulations have shown that intra-family arrange-ments could generate as much as three-quarters ofthe welfare gains from an actuarially fair annuitymarket.

Empirical studies have not found much evidenceof transfers within families that fit this model. Butthis is hardly surprising: the studies have focusedon industrialized countries with broad public andprivate annuity provision. In contrast, within-family provision may well be important intraditional societies and rural communities. Here,the transaction costs of buying annuities are high-est while informal contracts are common practice.

The desire for liquid assets or bequeathable wealthand the availability of substitutes for private an-nuities must be taken into account when designingbenefit rules in a defined contribution pensionsystem. Also, transaction costs and the state of theinsurance sector (including regulatory capacity)should be borne in mind.

3 Annuities

Why limit withdrawals?The fact that few people buy annuities voluntarilyposes a challenge for reforms relying on definedcontribution schemes. To reduce old age povertyand provide a reasonable degree of earnings re-placement in retirement, government interventionmay be warranted.

Mandatory provision for income in old age is usu-ally justified on two grounds. First, paternalism.People are myopic, and left to their own deviceswill not save enough. Others may be forwardlooking, but may lack the information needed tomake sensible savings choices. Secondly, there isthe phenomenon economists call ‘moral hazard’.People will not save enough if they expect gov-ernment to rescue them in their old age. Andgovernments in many countries cannot crediblycommit to leave pensioners destitute.

These same arguments apply to withdrawals in re-tirement savings systems. Myopic people mightspend their savings early in retirement. And publicsafety nets encourage even the forward looking tospend to use up their wealth and then rely on gov-ernment support. Lack of information—oninflation or life expectancy, for example—can alsomean people make choices they later regret.

Mandating annuitiesForcing people to convert the whole of their re-tirement savings into an annuity is an obvioussolution to the problems of myopia, lack of infor-mation and moral hazard. It also seems a sensibleresponse to the possibility of ‘adverse selection’mentioned earlier.

But we have already noted several reasons whypeople can find annuities unattractive, even whenthey have perfect foresight. Mandating annuitiescould reduce the welfare of these people, for ex-ample, by preventing them from leaving money totheir children. Moreover, public policy objectivescan be achieved without requiring fullannuitization of wealth.

Minimum annuity levelsThe tensions between individual preferences andpublic policy objectives point to the need to strike

a careful balance as opposed to a blanket mandateto annuitize. This balance will be different in eachcountry but a sensible starting point is to requirepeople to take out an annuity of a minimum level.No one will be left destitute as a result of myopia.And, if the minimum is set higher than the safetynet income, it mitigates the moral hazard problem.

A gap between the social safety net income andthe minimum annuity is advisable for two reasons.First, the social safety net might be uprated morerapidly (by earnings, for example) than the annuity.So after a long period of retirement, the annuitymight actually fall below the safety net.

Secondly, the safety net income is often set at alevel that is much lower than would be a reason-able replacement rate for an average wage worker.People with a reasonable level of accumulated re-tirement savings should not be permitted, throughmyopia, to dissipate this wealth and then fall to thesafety net level. Another way to avoid such asituation is to mandate not only the minimum an-nuity level but also a minimum replacement ratetarget based on the worker’s own pre-retirementearnings. Naturally, the higher this mandated re-placement rate, the greater the likelihood that thecertain individuals will, in their view, hold toomuch of their wealth in the form of an annuity.

Finally, in mandating the minimum annuity,policymakers must take the interests of schememembers’ dependants into account. Widows tendto be poorer than the rest of the elderly andwomen tend to live longer than men. If peoplecan tie their annuity to their own life alone, thenthe government might have to support many sur-viving spouses. Problems of myopia and moralhazard suggest that at least the minimum annuityshould be required to provide for survivors. Ofcourse, the stream of income required to maintainliving standards need not be as high as when bothspouses were alive.

IndexationThe purpose of mandating annuities will beundermined if the purchasing power of thepayment declines over time. Even low levels ofinflation can dramatically affect living standards.

Annuities 4

For example, 2½ per cent inflation over 25 yearswould nearly halve the value of a level (unindexed)annuity.

Inflation indexed annuities are not common.Even when they are widely available, as in theUnited Kingdom, take up is very low. This sug-gests another kind of myopia: people are unawareof the longer-term effects of inflation on theirbenefits. In economics terms, ‘money illusion’ isat work. Inflation protection should therefore berequired for at least the minimum mandatoryannuity and perhaps for all annuity products.

So that private insurers can offer inflation protec-tion, the government will probably need to issueindexed public bonds. These allow annuity pro-viders to insure their liabilities. But financeministries have often opposed indexed bondsbecause they legitimize inflation and inflationaryexpectations. If people are protected from infla-tion’s adverse effects, the argument goes, they willbe reluctant to support painful macroeconomicstabilization programs.

Broader macroeconomic concerns must of coursetake precedence over the narrower interests of theretirement income system. But, once expectationsof permanently high inflation are eliminated, thereare more effective means of ensuring stability andcredibility, such as an independent central bank.

Draw-downs and annuity optionsA draw-down is an alternative way of spreadingaccumulated retirement savings over time. Ratherthan purchasing an annuity, an individual with-draws his balance according to a preset formulathat takes into account average life expectancy andthe interest rate. The main problem with draw-down is the risk that people might outlive theirresources. A draw-down option could also exac-erbate adverse selection: people with shorter lifeexpectancy are able to opt out of the annuitymarket.

Scheduled withdrawals are useful for people whowant to share in the investment returns (and risks)of the provider. In contrast, a standard life annuitycontract is based implicitly on a fixed rate of re-turn. Since insurance companies assume all the

risk, the implicit interest rate is usually closer to theyield on government bonds with a similar duration.

An alternative product is a variable annuity. Thisis again an irrevocable contract, but the buyershares in the risk and the return of investing thefund. If returns are low, future payments can bereduced (and vice versa). In Argentina, for exam-ple, annuities must generate at least a 4 per centnominal rate of return. Above that level, annuitybuyers and sellers can agree to split the returns inany way they agree.

Many other variants that customize the level andduration of the annuity income stream and associ-ated risks can be offered. Some contracts allowfor a fixed period of payments, say 20 years, evenif the annuitant dies before the period is up. Someannuities allow for deferral of payments for severalyears. Limited inflation protection can be pur-chased at lower cost than a fully indexed annuity.An infinite number of combinations can bedevised.

Timing of withdrawalThe value of accumulated retirement savings can,depending on how funds are invested, be volatile.Annuity rates also vary over time with long-terminterest rates. In the United Kingdom, for exam-ple, an annuity for a 65 year old man fell from over15 per cent of the fund in 1990 to around 10 percent in 1998.

Variations in the fund value and annuity ratesmean the time at which retirement savings areconverted to an annuity can have enormous effectson pension income. So, for example, if people areforced to convert to an annuity at a set pension-able age, they will lose out if that coincides with,say, a stock-market crash. This ‘timing risk’ can bemitigated by allowing people to choose when theyannuitize drawing down retirement savings in themeantime. But even professionals fail to predictstock-market and interest rate trends.

There is a better solution to the problem of timingrisk. Annuitization can be thought of as a one-time portfolio shift, from a broad range ofinvestments to a narrow portfolio: the investmentsof the insurer backing the annuity, predominantly

5 Annuities

in bonds. Variable annuities are based on abroader portfolio. The insurer invests in a rangeof assets, and the annuity pay-out adjusts to reflecttheir value. This obviates the need for the one-time portfolio shift associated with timing risk.Variable annuities are also a better way of deliver-ing the flexibility of investments achieved by draw-down.

Early international experienceOnly two of the countries with mandatory, indi-vidual accounts—Australia and Hong Kong—allow members access to the whole fund balancewhen they retire. Australians generally take alump-sum pay-out at retirement. (What happensthereafter is complicated by the presence of anincome and asset tested public pension program.)Hong Kong will only begin collecting mandatorycontributions in late 2000, so there is noexperience of withdrawals yet.

Another dozen countries with individual accountschemes restrict withdrawals in one way or an-other. In the United Kingdom, for example,people can take out a lump-sum of up to a quarterof their accumulated pension fund. They can drawdown the rest of the fund gradually after retire-ment. But they must buy an annuity with theremainder by age 75 at the latest. Sweden willforce people to buy annuities with their mandatorypension funds. Sweden is the only country wherethe government provides all annuities. The newschemes in Hungary and Poland also requireannuitization but with private insurers.

Latin American schemes strongly encourage an-nuities but most allow for scheduled withdrawals.In Chile, about half of the quarter million pension-ers in the new private scheme have opted for someform of annuity.

RegulationsOnce the decision is made to restrict withdrawals,a series of difficult regulatory choices arise. Sev-eral have already been mentioned. For example,what are the specific types of annuities allowedand who can offer them? What is the minimumannuity that the retiring worker must purchase?The rules governing pricing and the way these

complex products are sold lead to additionalregulations. Finally, there may be implicit or ex-plicit guarantees which may necessitate furtherrules and a process for monitoring them.

The most basic decision is the benefit level belowwhich restrictions will be applied. In Latin Amer-ica, the minimum annuity level is usually set bothin terms of the worker’s own pre-retirement earn-ings and some absolute minimum specified by thegovernment. For example, workers in Argentina,Peru and Chile have the option of taking a lumpsum if the remainder of the balance would allowthem to purchase an annuity that provides areplacement rate of 70 percent.

In Chile, the minimum is determined according toa formula which states that if the individual canpurchase an annuity of value equal to or greaterthan the higher of 1.2 times the minimum pensionor a 70 percent replacement rate of the previousfive years’ average real earnings, the rest of the bal-ance can be taken in the form of a lump sum.Since the ceiling on taxable earnings is twice theaverage wage, this means that highest mandatedannuity is 140 percent of the average wage. Thistype of rule also provides flexibility with regard tothe retirement age.

Annuity providersDuring the accumulation stage, some countrieswith individual accounts have relied on specializedinstitutions. This is true for all of the LatinAmerican reforms and is also the case in Hungaryand Poland. In contrast, with the exception ofArgentina and perhaps Poland, most of these sys-tems allow annuities to be purchased from regularlife insurance companies and not only specializedfirms.

The problem with requiring specialized institutionsis that separate capital requirements, staff andother costs of doing business are increased. Thismay limit competition and is likely to result inhigher transaction costs for annuitants. On theother hand, weaker providers could lead to defaultand trigger expensive guarantees. A compromiseis to allow life insurance companies to participate

Annuities 6

but to require stricter standards for acquiring a li-cense to sell annuities in the mandatory system.

Regulating annuity pricesAnnuity providers might offer different annuityprices according to individual characteristics thatare related to life expectancy. Sex, marital status,income and parents’ longevity are all attributes thataffect people’s mortality risk. If insurers do nottake account of available information, they mightbe undercut by competitors offering better termsto better risks. They would face their ownindividual adverse selection effects.

However, differential annuity pricing raises someimportant public policy issues. For example, lowerannuity payments to a woman than to a man withthe same accumulated retirement fund is actuariallyaccurate. Even though people are aware thatwomen live longer on average, governments oftenrequire insurers to offer unisex annuity rates. Theredistribution from men to women that this im-plies is justified as a way of avoiding theperception of discrimination when women receivelower annuity rates. Some other issues may be-come even more important in the future. Forexample, the use of private medical informationand the potential for genetic testing are keysources of longevity information that will becomeeasier to obtain in the next decades.

In practice, most of the countries with individualaccount schemes impose strict regulations on theway annuities are calculated and sold. Govern-ments specify age-specific survival expectationsused in the calculations. These may differ fromnational mortality data as is the case in Argentina,Chile, Colombia and Peru where special tableswere sanctioned. All of these tables have signifi-cantly lower mortality rates than those found inpopulation-based tables. The difference persistseven compared with projected mortality, rangingfrom around 3 per cent in Argentina to almost 14per cent in Peru. The lack of reliable mortalitydata on potential annuitants poses a major chal-lenge to annuity providers and supervisoryauthorities.

The interest rates used in annuity calculations arealso regulated in Latin America. In Argentina,

insurance companies are required to use a 4% percent nominal rate for both reserves and pricing. InChile, reserves had to be discounted at 3 per cent ayear real until 1988. Since then, reserves are dis-counted at the long-term rate on the underlyingassets. The situation in Peru and Colombia issimilar, with a 4 per cent fixed interest rate for re-serves in Colombia and 3 per cent in Peru. Therate used to calculate the annuity is not stipulated.It is typically around 4 per cent in Colombia andalmost 6 per cent in Peru.

Figure 2 compares the monthly payment thatcould be purchased with $100,000 in Australia,Canada, the United Kingdom and the UnitedStates with quotes from four Latin Americancountries. The data are drawn from severalsources, but they refer to the same kind ofindividual and the same type of annuity. In thefour cases at the bottom of the chart, the annuityis price indexed. The five bars at the top refer tonominal annuities. Note that the Argentineannuity allows the holder to share in returns inexcess of four percent.

The pay-out from a nominal annuity lies between$700 and $880 a month. Inflation indexed annui-ties range from around $620 in the UK to almost$820 in Chile. Interestingly, the indexed annuity inthe United Kingdom pays a much lower amountthan the indexed Latin American products: 60 percent less than in Chile. Part of the explanation isthe fact that Chilean annuitants have life expectan-cies that are five percent lower than their(voluntary) counterparts in the United Kingdom.Real interest rates are also higher in Chile. Un-fortunately, because life expectancy of annuitants,interest rates and even the competitiveness of theinsurance industry vary, these figures do not tell ushow close these amounts come to providing a fairannuity.

This requires an estimate of the ‘money’s worth’ ofannuities sold. A widely used measure of this isthe ratio of the fair annuity price to the marketprice. Several studies have measured the money’sworth ratio in the United Kingdom and the UnitedStates. Typical results are in the 85-90 per centrange. But this does not measure the fairness ofannuity prices to people buying them. Using

7 Annuities

annuitants’ life expectancies, the ratio tends to bevery close to 100 per cent.

Annuity rates around the world 2

0 200 400 600 800 1000

UK

Colombia

Peru

Chile

Canada

US

Australia

Argentina

UK

monthly annuity from $100,000

nominal

indexed

But this calculation is problematic, especially indeveloping countries. First, many countries do nothave annuitant mortality tables or even projectedlife tables. So these have to be assumed. Sec-ondly, few countries have long-term bond marketsor, if they do, they are illiquid. It is difficult thento discount future annuity payments. Money’sworth ratios also ignore the risk that an insurer willdefault, which will affect cross-country compari-sons significantly. Finally, money’s worthcalculations implicitly assume that projected mor-tality is certain. In fact, demographers have oftenmade serious errors in forecasting mortality. Ifthis risk is taken into account, a significant part ofthe difference between ‘fair’ and observed annuityprices can be explained.

Mortality and wealthPerhaps the most difficult issue in annuity pricingis the potential for redistribution from those withlower lifetime income and wealth to higher incomeannuitants. This occurs when there is a positiverelationship between longevity and wealth. In-deed, this is simply the corollary of theredistribution to groups that are systematicallylonger lived in public pension schemes. Studieshave found such unintended redistribution in theNetherlands, Sweden, the United Kingdom, andthe United States.

Figure 3 shows the wealth-mortality relationshipfor older households in the United States based onthe Health and Retirement Study. People in thepoorest quarter of the population are on averagefour times as likely to die in any period than therichest quarter.

Wealth and mortality in the US 3

0

0.05

0.1

0.15

0.2

50-54 55-59 60-64 65-69 70-74 75-79 80+

age

poorest quartile

richestquartile

mortalityrate

This relationship has several important policy im-plications. First, it suggests that at least some ofthe observed differences between population andannuitant mortality rates can be explained bygreater demand for annuities among people in thehigher wealth quartiles. This casts some doubt onthe evidence of adverse selection and supports asimpler explanation.

Another implication is that national mortality ta-bles understate longevity in countries with partialpension system coverage, because people in theinformal sector tend to be poorer than average. Sotheir mortality rates are likely to be higher thanthose of members of the pension system.

Most important however, is the possibility thatmandatory annuitization will lead to unintendedredistribution away from workers with lower life-time incomes. Jeffrey Brown of HarvardUniversity finds that that these transfers couldamount to as much as 20 per cent of pension as-sets for low-income workers in an individualaccounts scheme. But he also suggests that thesetransfers can be reduced by allowing for guaran-teed payment periods, bequest options and joint-life annuities. Of course, these options lead to

Annuities 8

lower benefits for annuitants themselves sincethese options are more expensive than a standardlife annuity.

Transparency and supervisionEfforts to improve consumer financial literacy andto regulate and supervise new pension systemshave, naturally, tended to focus on the accumula-tion stage, as contributions and investment returnsbuild up in retirement savings accounts. In con-trast, there has been relatively little considerationof the conditions in the insurance sector and thesupervisory apparatus required for the benefitstage of the system. Early experiences, especiallyin Latin America, highlight the need for better in-formation and transparency in the new annuitiesmarkets. Parallel reforms in the insurance sectormay be necessary to ensure the success of thereform.

PENSIO

NREFORM

PRIMERre-form v.t. & i. 1. make (institution, procedure

etc .) better by removal or abandonment

of imperfections, faults or errors

pr imer n. 1. elementary book to

equip person with information

pe´nsion

n. 1. p

eriodic

payment

made on

retirement

or

above s

pecified

age

Further readingWalliser, J. (2000), ‘Regulating withdrawals in indi-

vidual account systems’ Pension ReformPrimer series, Social Protection DiscussionPaper no. 000x, forthcoming.

Valdés Prieto, S. (1999), ‘Risk in pensions and an-nuities: efficient designs’, Pension ReformPrimer series, Social Protection DiscussionPaper no. 9804, World Bank.

Brown, J.R., ‘Differential mortality and the valueof individual account retirement annuities’,presented at a National Bureau of EconomicResearch conference, October 1999.

Friedman, B.M. and Warshawsky, M.J. (1990),‘The cost of annuities: implications for savingsbehavior and bequests’, Quarterly Journal ofEconomics, vol. 104, no. 2, pp. 135-154.

Mitchell, O.S., Poterba, J.M., Warshawsky, M.J.and Brown, J.R. (1999), ‘New evidence on themoney’s worth of individual annuities’,American Economic Review..

Kotlikoff, L.J. and Spivak, A. (1981), ‘The familyas an incomplete annuities market’, Journal ofPolitical Economy, vol. 89, no. 2, pp. 372-391.

Conclusions and recommendationsq regulation of withdrawals in pension

systems based on individual accountsneeds to balance public policy objectivesand individual circumstances

q family arrangements can provide a largeportion of the welfare gains of annuities

q and preferences vary including the desireto bequeath wealth and take precautionsfor medical expenses

q at the same time mandatory annuitizationprotects pensioners against longevity riskand reduces government’s social safetynet liabilities, by ensuring people do notspend all their savings early

q balancing these different objectivesmeans that mandatory annuitisation ofthe whole of retirement savings isunlikely to be optimal

q the best strategy is to set a minimum,indexed annuity with adequate survivor’sprovision, with flexibility for anyremaining retirement savings