OFFICE OF THE CHIEF ECONOMIST, LATIN AMERICA AND CARIBBEAN REGION, THE WORLD BANK BACKGROUND PAPER FOR REGIONAL STUDY ON SOCIAL SECURITY REFORM

Transcription

1 OFFICE OF THE CHIEF ECONOMIST, LATIN AMERICA AND CARIBBEAN REGION, THE WORLD BANK BACKGROUND PAPER FOR REGIONAL STUDY ON SOCIAL SECURITY REFORM The Performance of the Funded Pension Systems in Latin America by Juan Yermo Organization for Economic Cooperation and Development 1

2 The Performance of the Funded Pension Systems in Latin America by Juan Yermo Organization for Economic Cooperation and Development Abstract: This paper analyses the performance of private pension systems with respect to two key indicators: replacement rates and the return to pension contributions. While returns have been high since the inception of the systems, some deficiencies in the risk management design of the system may put performance into jeopardy in the near future. Further, the expected cost-cutting nature of competition has not materialised and most countries have simply opted to take away the little freedom that individuals had to make choices. Only Chile has begun to diverge from this trend, liberalizing foreign investment and introducing some choice over investment during the accumulation stage. Reforms are still needed during the retirement stage, in order to permit more flexibility in the timing and form of annuity purchases. 2

3 Introduction Pension reform of the kind implemented in some Latin American countries brings both benefits and new challenges. These reforms constitute a long learning process that started with the pre-funding of pension benefits and that will continue for many years to come as many practical aspects of the systems are refined and readjusted. This paper aims to assess the validity of various hypotheses about the potential economic benefits of pension reform of the Latin American kind. The basic features of these reforms have been the phasing out of publicly managed, defined benefit plans in favor of defined contribution, personal pension plans managed by specialized financial companies. Unlike the previous system, which was run on a pay-as-you-go (PAYG) basis by the state, the new individual accounts are fully-funded and include some degree of investment diversification. One exception to this general trend is in Brazil, where a large public system has been maintained and where occupational pensions are being promoted on a voluntary basis as a complement to state pensions. The performance of these private pension systems can be gauged from several different perspectives. Ultimately, the two key variables are the pension that the elderly receive when they retire (often measured as a percentage of their salary the so-called replacement rate) and the return to pension contributions. A pension system should provide high enough benefits so that workers can maintain their standard of living after retirement but should do so for the lowest possible cost (by minimizing market distortions and any undesirable income redistribution). The first part of the paper describes the private pension systems of Latin America and explains how risks are managed. We observe that participants in the individual account systems bear the investment risk but have no say over how their pension contributions are invested. Participants also face significant investment and longevity risk after retirement. In most countries, however, individuals are offered more product choice at retirement than during the accumulation stage. We then compare the individual account model with the occupational model that is in place in Brazil. In Brazil, defined benefit plans are still the norm in the voluntary occupational system, but defined contribution plans are rapidly gaining ground. The second part of the paper evaluates the performance of the private pension system from the perspective of the plans participants. We focus on the 3

4 administrative costs and investment performance of the plans while contributions are accumulating in the worker s pension account and on the cost of annuities during their retirement. We find that the returns to contributions in the individual capitalization models are significantly reduced by the plans administrative costs. Also, replacement rates are unlikely to be high for the large percentage of workers who do not contribute on a regular basis. The third part of the paper focuses on the roots of this performance and analyzes the governance and operation of pension funds and insurance companies and the effectiveness of the regulatory and supervisory framework. We analyze three types of risk agency, investment, and longevity risks. Agency risks arise from conflicts of interest and the asymmetry in information among the three main parties involved in the private pension system workers, providers, and policymakers (including regulators and supervisors). We first examine the agency risks that affect the operation of pension funds. We then evaluate how successful pension funds have been in managing investment risk. We find that regulatory restrictions on foreign investment have compromised investment returns, but that restrictions on domestic equities have not been as costly because of the volatility of these markets. We also look at the role of insurance companies in managing both investment and longevity risk. The lack of suitable financial instruments and mortality tables appears to hamper the efficiency of the insurance business. Regulations are also generally more lax than those applied to pension funds. In Part IV, we evaluate the nature and extent of competition in the provision of private pensions. We assess the efficiency of pension funds in managing individual accounts, and we look at ways to encourage cost-reducing competition in pension provision. We also look at the pros and cons of occupational and personal pension plans and the main policy choices that arise from this debate. Our results from this paper call into question the claim that the introduction of private sector management fosters cost-cutting competition and innovation and helps to insulate pension systems from political interference. We also question the soundness of the risk management features of the private pension systems. The practical absence of individual choice during the accumulation stage, low ceilings on foreign investment, and restrictions on the timing and formality of product choices at retirement are putting into jeopardy the performance of private pension systems. The paper concludes by proposing some possible venues for reform. 4

5 I: The Design of Private Pension Systems in Latin America The two most important risks that an individual faces when saving for old age are investment and longevity risk. A person who saves for retirement through financial instruments does not know the exact value of her accumulated savings when she retires. Similarly, no one knows the number of years of life that he or she has left after reaching retirement. In defined contribution plans, the investment and longevity risks are borne by the individual. In defined benefit plans, on the other hand, these risks are in principle shifted away from the individual. In public defined benefit pension systems, these risks are shifted to the state, which in turn allocates them across generations. In occupational defined benefit plans, plan sponsors, and hence, ultimately, shareholders, bear these risks. Defined benefit plans, however, are not risk-free. Both governments and employers can default on their pension promises. Governments may be forced to reduce pension benefits to contain fiscal deficits while companies that sponsor pension plans may go bankrupt. Some personal pension plans also offer some degree of protection against investment or/and longevity risk. Annuities for example, are a life insurance product that can guarantee a fixed benefit until death in exchange for a lump-sum payment. Life insurance companies are able to offer such contracts because they can pool the longevity risk across many individuals. Ultimately, it is also the life insurance company s shareholders who are responsible for meeting any promises made by the company. In addition, any private pension system is laden with a variety of agency risks or conflicts of interest that arise from the delegation of functions by an interested party to another party who is better informed or has superior skills. Agency risks arise in the context of the management of a retirement plan and in the regulation and supervision of private pension systems. The Individual Account Systems of Latin America In nine Latin American countries (Argentina, Bolivia, Chile, Colombia, Costa Rica, El Salvador, Mexico, Peru, and Uruguay 1 ), publicly managed defined benefit plans have been curtailed and replaced by personal pension plans that are managed by specialized financial intermediaries. These 1 Paraguay is undergoing a similar reform. 5

6 specialized financial institutions are called pension fund administrators or a similar name 2. The new plans are based on a defined contribution formula. During the accumulation stage, plan members bear fully investment and longevity risks. Members can choose a provider (a pension fund administrator) but have no say in the investment allocation of their savings. Except in Chile, which introduced a second fund in May 2000, pension fund administrators may offer only one fund. 3 Moreover, the asset allocations of this fund are constrained by quantitative investment limits, and the performance of each fund cannot stray too far from the industry average. As a result, the funds offer similar risk-return trade-offs. To the extent that plan members differ in risk aversion, such a onesize-fits-all defined contribution formula can be rather costly. Individuals who are highly risk averse may be frightened by the volatility of the returns offered in the new system. Indeed, in countries where the public s confidence in the ability of the government to maintain macroeconomic stability is low, workers may choose to avoid the new system altogether and invest their retirement savings abroad. Chile has been the first country to start tackling this deficiency in the system by requiring the pension fund administrators to set up a second fund (the Fondo 2) that is invested exclusively in domestic fixed-income securities. Only men older than 55 and women older than 50 are permitted to switch their accumulated balances to the second fund. In April 2002, a law was approved that extends the number of funds that AFPs manage to five. 2. Administradoras de Fondos de Pensiones (AFPs) in Bolivia, Chile, El Salvador, and Peru, Administradoras de Fondos de Pensiones y Cesantías (AFPCs) in Colombia, Administradoras de Fondos de Jubilaciones y Pensiones (AFJPs) in Argentina, Administradoras de Fondos de Ahorro Previsional (AFAPs) in Uruguay, and Administradoras de Fondos para el Retiro (AFOREs) in Mexico. 3. This is the case for the mandatory system. Workers are also able to make additional voluntary contributions to the individual accounts managed by the pension fund administrators. In some countries such as Chile and El Salvador, these voluntary contributions must be deposited in the same fund as the mandatory contributions. In others such as Colombia and Mexico, there is a separate fund for voluntary contributions, which is subject to a more flexible regulatory regime. In Colombia, fiduciary societies are also able to manage these voluntary pension funds. 6

7 While workers are fully exposed to investment and longevity risk during the accumulation stage, they are fully insured against the risks of disability and death. The premiums for these policies are paid by the pension fund administrators to the insurance companies, except in Mexico, where the social security institute has retained a monopoly over these services. As shown in Table 1, workers face more choice during the retirement stage than during the accumulation stage.. In all Latin American countries except Bolivia and Uruguay, participants may choose at retirement between drawing down their accumulated balance as part of a program of scheduled withdrawals, purchasing an annuity. Some countries also permit a combination of these two options (deferred annuities). The annuities sold are of the traditional type (fixed, as opposed to variable). This type of annuity involves the transfer of (nominal) investment and longevity risks to an insurance company. In some countries such as Chile, Colombia, and Peru, annuity benefits must be indexed to a measure of prices, which also transfers inflation risk to the insurance company. In other countries such as Argentina, there is no indexation requirement, but benefits may be denominated in US dollars, which offers protection against devaluation (and hence at least partial protection against inflation). No Latin American country as yet offers the possibility of buying variable annuities, in which investment risks are borne by the pensioner and longevity risks are borne by the insurance companies. Such annuities may be particularly attractive for high-income workers or for workers in countries where the public pension system still offers generous pensions. Country Argentina Bolivia Chile Colombia El Salvador Mexico Peru Uruguay Table 1: Form of Retirement Benefits Benefit form Annuity, scheduled withdrawal (up to five years after retirement) Only annuity Annuity, scheduled withdrawal, deferred annuity Annuity, scheduled withdrawal, deferred annuity Annuity, scheduled withdrawal, deferred annuity Annuity, scheduled withdrawal Annuity, scheduled withdrawal, deferred annuity Only annuity Note: Lump sums are permitted in all countries except Bolivia and Uruguay, but there are substantial constraints. Relevant data can be consulted in Devesa, Martínez, and Vidal (2000). 7

8 The scheduled withdrawal option lays all investment and longevity risks on the pensioner. This option may be preferable when interest rates are very low and are unlikely to increase within the time frame permitted for purchasing a deferred annuity. However, pensioners are not free to choose the amount that they wish to withdraw as a benefit. Instead, the pension fund administrator recalculates the scheduled withdrawals every year as a function of the pension fund s return and the life expectancy of the worker and his or her family members. In Colombia, Chile, and Peru, workers can also buy deferred annuities and in the meantime draw down part of their accumulated balances as part of a scheduled withdrawal. Deferred annuities are often attractive when interest rates are expected to increase. Nonetheless, the problem of mistiming the annuity purchase is rife, given that workers are not able to buy deferred annuities before retirement and only deferrals of one to three years are permitted at retirement. In countries such as Bolivia, Mexico, and Argentina where deferred annuity purchases are not permitted, the investment risks that workers face are even greater. Some countries have restricted the amount of choice that plan members have among these three main options. In Bolivia and Uruguay, participants are required to purchase annuities at retirement. In Chile, workers who want to draw their pension (and retire from the mandatory pension system though not necessarily from the workforce) before the official retirement age are required to purchase an annuity, which must exceed 110 percent of the stateguaranteed minimum pension. If the annuity exceeds 70 percent of his or her average wage, the rest of the accumulated balance can be taken as a lump sum. A similar rule is in place in Argentina and Peru. On the other hand, workers in Chile and El Salvador who have accumulated funds that are insufficient to generate annuities above the minimum pension are not given the option of purchasing an annuity. Instead, they must choose the scheduled withdrawals option and draw down a pension equal to the minimum pension. After their funds run out, the government pays the minimum pension. The Brazilian Private Pension System The Brazilian private pension system, being voluntary, has not been subject to the constraints that are present in other Latin American countries. Plan sponsors and plan providers are free to design their plans as they please. However, some current features of the system can still be costly for some workers. The defined benefit plans that have been at the core of the Brazilian 8

9 occupational system are based on back-loaded formulas that encourage long tenure. Therefore, workers who leave a company after only a few years receive much lower benefits per year of service than they would have done if they had stayed for a longer period. Private pension plans in Brazil form a voluntary complement, or second pillar, to the social security system. 4 This second pillar consists of the so-called Complementary Pension System (Sistema de Previdência Complementar), established in 1977 by Law 6435 and the Fundos de Aposentadoria Programada Individual (FAPIs), which are long-term investment accounts managed by mutual funds. The complementary system itself consists of closed pension funds (Entidades Fechadas de Previdência Privada) and open pension funds (Entidades Abertas de Previdência Privada). The closed funds are constituted as employer-sponsored non-profit organizations covering the employees of a particular firm or group of firms. Closed pension funds support occupational plans that have traditionally been of a defined benefit nature. Increasingly, however, defined contribution plans are being promoted. Occupational pension plans have historically operated under a lax regulatory and supervisory framework that offered little protection to plan members against the bankruptcy of the sponsor. Regulations lacked transparency, the tax treatment of pensions was uncertain, and the supervisory authority was understaffed and had limited intervention powers. The new regulatory framework, which derives from Complementary Laws nos. 108 and 109 that were approved in , open up a new chapter in the development of occupational plans. They ensure more effective control of funding and investment of pension funds, more transparency and effectiveness of supervision, and greater disclosure of information to plan members. The open pension funds are constituted as insurance companies that cover any worker who chooses to enrol. Open pension plans were until recently structured as defined benefit (DB) schemes and took the form of inflationindexed deferred annuities, though there were some defined contribution 4. The first pillar consists of programs under the Regime Geral da Previdencia (RGPS), which covers workers in private firms and public sector employees who were hired under the Consolidated Labor Code and the Federal, state and municipal Regimes Juridico Unico (RJUs), which covers tenured government employees in the executive, legislative, and judicial branches and the military. A Constitutional Amendment in November 1998 allowed the establishment of complementary funds for the RJUs. 9

10 (DC)plans. Hence, the main players in this market are insurance companies, which carry out the four main services of a pension system: collecting contributions, administering accounts, managing assets, and paying benefits. The DB plans offer a guaranteed 6 percent real rate of return and between 50 percent and 75 percent of any actual excess return. 5 At retirement, the investor has the option of receiving the entire accumulated balance in full or in part. Open pension plans also offer other benefits, such as life, survivor, and disability insurance. In 1998, a new form of plan, the Plano Gerador de Beneficio Livre (PGBL), was created. The PGBL is a DC scheme with flexible contribution and investment options and without return guarantees. Companies can contract PGBL plans for their employees, similar to the 401(k) plans in the United States. Investors may alter the contribution rate and may choose between three different funds: a sovereign fund (government securities), a fixed-income fund, and a mixed-income fund. The PGBL administrator can invest contributions in only one of these three funds, which are managed exclusively by mutual fund companies. 6 When the worker retires, the PGBL administrator takes his or her accumulated assets and buys an inflation-indexed annuity. II: The Performance of the Private Pension Systems Policymakers main social concern about private pension systems is to ensure that these systems provide workers with enough retirement income that, together with other income sources, will lead to sufficiently high replacement rates (for example, 70 percent). At the same time, policymakers should be concerned about the efficiency of the system, specifically with the costs imposed on workers and other economic agents by these private pension systems. The Individual Account Systems of Latin America The actual amount varies between funds. Normally, the maximum that can be transferred is only achieved after a worker has been in the plan for a few years (about five on average). The excess return accumulated in a year can be retrieved or can be left to accumulate in the fund. Each exclusive fund may receive contributions from more than one PGBL plan as long as they have similar characteristics. The funds, however, are only open to investment from PGBLs. 10

11 Latin American countries differ in the extent to which they reduced benefits from the PAYG social security system. Correspondingly, these countries differ in the importance of the individual account systems as a source of retirement income. A full evaluation of retirement benefits and administrative costs would therefore necessitate a comparison of benefits from both the public and the private systems. Replacement Rates. One key measure of the performance of private pension systems from the perspective of plan members is the level of retirement income relative to their salary before retirement (the so-called replacement rate). Because the individual account systems of Latin America are based on a defined contribution formula, replacement rates depend on the following factors: The contribution rate Salary growth The worker s contribution record Returns to pension funds Annuitization rates (or life expectancy if the worker chooses the income draw-down option). The sensitivity of replacement rates to these variables can easily be modeled. In the following exercise, we assumed that salary growth is constant at 0 percent in real terms per year. The base model assumed a 10 percent contribution rate, a 5 percent real rate of return, 35 years of contributions, and a 5 percent annuitization rate. Table 2 shows the effect on the replacement rate of a 10 percent increase in each of these parameters. The contribution record is clearly the most important parameter. The low ratios of contributors to affiliates observed in most countries (about one in two in most countries, see Packard (2002) are, therefore, a threat to the income security of a large segment of the population. Table 2: Modeling Replacement Rates 10 % increase in: % Increase in replacement rate Contribution rate 10.0 Rate of return 11.0 Contribution record 22.7 Annuitization rate

12 Source: Authors calculations. The rate of return, net of asset and return-based commissions, is also an important determinant of the replacement rate. Since in all countries except Chile, workers have access only to a single fund, regulations and the investment choices of the pension fund administrators are the only factors that influence the returns to pension funds. A worker's choice of pension fund administrator will also affect his or her replacement rate. However, investment and rate of return regulations have reduced the extent to which returns vary. The limited availability of liquid, domestic securities has exacerbated this effect. Srinivas and Yermo (1999) reported an average correlation between pension fund returns in Chile and Peru as high as In Argentina, the average correlation is lower at Further evidence of the similarity of pension fund returns is provided in Table 3, which shows average returns (both arithmetic and geometric) since the inception of the system across AFPs. The maximum amount by which any AFP s return deviates from the system s average return is approximately 3 percent. The similarity in returns means that investors gain very little by switching between pension funds in these countries, especially in those countries where returns are subject to floors and ceilings. Nonetheless, over a 40-year period (the typical length of a career), such a small difference in annual return can lead to a difference in the accumulated balance relative to the system s average balance as high as 15 percent. Table 3: Average Rates of Return by AFP in Chile: Arithmetic average Geometric average Cuprum Habitat Magister Planvital Provida Santa Maria Summa Bansander System Note: Returns are for Fondo 1. Source: Superintendencies of Pension Fund Administrators 12

13 Correlations between pension fund returns are even higher for countries such as Bolivia, El Salvador, Mexico, and Uruguay, where pension funds invest only in domestic fixed-income securities. The average return of the industry is, therefore, representative of the actual return that any individual member obtained on his or her pension fund account and can be used, with little risk of misrepresentation, for estimating replacement rates. Table 4 shows the annual returns obtained by the pension fund industry since the establishment of the private pension systems and over the last year. The rate of return is calculated net of any asset management fees (only permitted in Mexico) but is not adjusted for contribution-based charges. These commissions have no impact on the accumulated fund, since they are paid on top of the mandatory contribution that goes into the individual account. On the other hand, contribution-based commissions create a gap between the actual replacement rate and the potential replacement rate that could have been achieved had these commissions been also invested in the individual account. We analyze this gap in the next section and use it as a measure of the costeffectiveness of the private pension system. As shown in Table 4, the highest real return to date was obtained by the pension fund industry in El Salvador, a 12.9 percent annual average return in real terms. The lowest was Peru s at 5.3 percent. Over the last 12 months, returns have been much lower. In Peru, the real return was in fact negative in

14 Table 4: Pension Fund Real Annual Returns to December 2000 Country Since inception Last 12 months Argentina Bolivia Chile Colombia 9.9 N/a El Salvador Mexico Peru Uruguay Source: AIOS. Notes: Real returns are annualized cumulative values. Returns for Chile are for Fondo 1. Colombia does not report returns over last year. For Mexico, returns are net of asset management fees. When adjusted for risk, the performance of the Peruvian pension fund industry is even worse. As shown in Table 5, the pension fund industry generated returns that averaged 0.46 percent on a monthly basis, but the standard deviation was one of the highest in Latin America at 1.2. Hence, the ratio of return to standard deviation (in other words, the return per unit of risk) was by far the lowest in the region (0.36). Table 5: Pension Fund Real Monthly Returns (from Inception to December 2000) Country Average Standard A/SD (A) Deviation (SD) Argentina Chile Mexico Peru Uruguay Source: Pension fund supervisors. Note: Colombia does not report monthly returns. Returns for Chile are for Fondo 1. While these high returns are likely to generate high pensions (as long as contribution periods are also long), their volatility will lead to significant differences in pensions across different retirement cohorts. An example, based 14

15 on Chilean historical returns, will help to elucidate the impact of return volatility on the accumulated fund. Figure 1 shows the AFP-average cumulative annual return for 20 different cohorts. Each cohort represented in the figure, except the 1981 one, started contributing at the beginning of the year. The 1981 cohort started contributing in July, the month when the system was launched. As of December 2000, cohorts who started contributing in the 1980s fared significantly better than those who started contributing in the 1990s. The 1980s cohorts have earned cumulative returns that range from a minimum of 8.73 percent (1987 cohort) to a maximum of percent (1981 cohort), with an average of 9.44 percent. The 1990s cohorts, on the other hand, have earned cumulative returns that range from a minimum of 4.05 percent (1995 cohort) to a maximum of percent (1999 cohort), with an average of 6.99 percent. Such differences in returns will generate a gap between the average accumulated balance of the 1980s and 1990s cohorts of the order of 35 percent. Figure 1: In Chile, inter-cohort differences in returns have been large Chile: average AFP cumulative return, Source: Superintendencia de AFPs 15

16 Since 2000, however, Chilean workers close to retirement have been offered the opportunity to trade out of the so-called Fondo 1 (the original pension fund) into a fund invested exclusively in fixed-income securities (Fondo 2). This option has not been taken up by anyone except a handful of lucky workers (only 64 contributors as of September 2000), who have benefited since then from bonds' out-performing equities. For workers who choose to annuitize their accumulated balance, bond yields and life expectancy are the other two determinants of the replacement rate that they will achieve. Since only single-premium immediate annuities are permitted, the timing of the conversion of the accumulated balance into an annuity can have a dramatic impact on the size of pension benefit obtained. In Chile, the Superintendency publishes a measure of the annuity s yield that is closely linked to the 10-year bond yield. The higher the yield, the larger the pension benefit will be. While historical data for this measure are not available, we have constructed a synthetic index of the value of annuities based on the average yield on fixed-income instruments traded in the Santiago exchange. Figure 2 shows the value of the annuity that a premium, fixed in real terms, would have bought at the end of each year since The value of the annuity is expressed in terms of the replacement rate, with the replacement rate of 1988 set arbitrarily at 50 percent. The interest rate used to calculate the benefit paid by the annuity is the annuity yield for workers retiring at the official retirement age, as reported by the Superintendencia de Valores y Seguros. As shown in Figure 2, there is a lot of variation in the annuity value over time. The difference between the highest and lowest replacement rate is 22 percentage points, the average replacement rate over the period is 60 percent and the standard deviation 6 percent. These large differences in replacement rates across cohorts are caused by the volatility of interest rates over the period. In particular, the decline in interest rates in the late 1980s lowered the value of annuities for cohorts retiring in these years. In particular, workers retiring in 1998 and 1999 would have been better off deferring the purchase of the annuity for two and one years, respectively. 16

17 Figure 2: Annuities have yielded varying levels of retirement benefits Chile: replacement rate from annuities (1988=50%) Replacement rate (%) Source: Author's calculations, based on data from Banco Central de Chile and the Superintendencia de Valores y Seguros. Given the dependence of replacement rates on all of these volatile variables, forecasting the actual performance of the system in the future is an exercise in clairvoyance. Pension fund returns and long-term interest rates, in particular, will evolve according to macroeconomic conditions and investment legislation, over which neither pension fund administrators or plan members have any control. For the moment, Chile is the only country with a significant number of old-age pensioners from the private pension system. A recent study from an AFP showed that replacement rates for a sample of 6,000 old-age pensioners were as high as 81 percent for early pensions and 73 percent for normal pensions. These replacement rates appear to be significantly higher than those under the previous system but are largely driven by the high real returns of the early 1980s. Commissions in the Accumulation Stage. Replacement rates by themselves do not provide a full picture of the performance of a private pension system. If such replacement rates have been achieved at the cost of high charges 17

18 paid by workers to cover administrative costs 7, workers' life-time incomes will suffer as a result. There are two main measures of charges that may be used: the reduction in yield and the charge ratio. The reduction in yield shows the effect of charges on the rate of return given a set of assumptions about the rate of return, the time profile of contributions, and the term of the plan. The charge ratio, on the other hand, is defined as the ratio of the accumulated balance that the charges by themselves would have generated to the accumulation without charges (in other words, had charges been added to the other contributions and also invested). As discussed by Whitehouse (2001), the use of the charge ratio can provide a misleading picture of the cost of a private pension system when commissions are calculated as a percentage of the accumulated fund. On the other hand, the charge ratio is a more useful measure than the reduction in yield of the cost-efficiency of the system in countries such as those in Latin America (except Bolivia and Mexico) where commissions are charged only on contributions made. Commissions to cover administrative costs (account and asset management expenses) are calculated as a percentage of a worker s salary or contribution. Fixed commissions are permitted in some countries such as Argentina, Chile, and Mexico. 8 In Bolivia and Mexico, pension fund managers can set commissions as a percentage of returns on the invested funds. Both Argentina and Mexico also permit loyalty discounts (for remaining with the same administrator). Variable commissions (those calculated as a percentage of contribution/salary) vary significantly across countries, as shown in Table 6. For a worker earning an average income, the country with the lowest charge, measured as a percentage of salary, was Bolivia at 0.5 percent in December The country with the highest charge was Peru at 2.39 percent. The ratio of charges to contributions in December 2000 is shown in the last column of Table 6. For countries where commissions are set only as a percentage of the worker s contributions or salary, this measure is equivalent to 7 8 In addition to these commissions, workers must pay monthly premia for disability and life insurance. In Peru, fixed charges were permitted until

19 the charge ratio 9.. In the case of Peru, the charges create a gap of 23 percent between the potential and actual accumulated balance at retirement for the contributions made in December Table 6: Commissions (December 2000) Administration Contribution / Fee/total fee / Salary Salary contribution A b C = a/(a+b) Argentina 2,09 7,72 21,3 Bolivia 0,50 10,00 4,8 Chile 1,61 10,00 13,9 Colombia 1,63 10,00 14,0 El Salvador 1,83 8,53 17,7 Mexico 1,98 12,07 14,1 Peru 2,39 8,00 23,0 Uruguay 2,04 12,32 14,2 Average 1, Note: Administration fee includes only account and asset management charges. Insurance premiums are excluded. Information for Colombia refers only to the mandatory pension fund system. Information for Bolivia includes only the contribution charge (the asset management charge varies from percent, depending on the amount of assets in the portfolio). Source: AIOS, Superintendencia Bancaria de Colombia. Figure 3 shows the commission to contribution ratio at the end of each year in three countries. Commission levels have fallen over the past few years in Argentina and Chile, while they have increased in Peru and remained stable in Uruguay. The different evolution of charges in these countries can be largely explained by the nature of competition in the industry and regulatory policies that have aimed at containing costs in Argentina and Chile In Mexico the two measures are not equivalent because some providers charge commissions on assets and returns. In Bolivia, the commission is set on assets under management. In Chile the AFPs also charge fixed commissions. 10 These issues are discussed further in Part IV. 19

20 Figure 3: Commissions have not declined systematically over time 30 Commission / Contribution, Argentina Chile Peru Uruguay Source: Superintendencies of respective countries, author s calculations The evolution in charges in Argentina and Chile is clearly favorable to workers who joined the system at a later stage. The accumulated balance to date of a worker who contributed regularly to the system from its inception is reduced by a larger percentage (the charge ratio) than that of someone who joined later. These differences in charges create potentially undesirable income inequalities between different generations. Figure 4 shows the cumulative charge ratio 11 for Chilean male workers earning the average wage in successive cohorts, where each cohort is 11 The cumulative charge ratio measures the total impact of charges on retirement income over a person's career. In order to calculate this ratio for Chile, both fixed and variable charges need to be taken into account. It is assumed that the representative worker is charged the industry average commission, where the weights are the contributions collected by each AFP. It is assumed also that the participant contributes to the system on a regular basis. In the Chilean case, the cumulative charge ratio is an accurate measure of administrative costs for older participants that made contributions from the start of the 20

21 identified by the year in which they would normally retire, starting with those that retired in The salary for the cohort that retired in December 2000 was set at 285,000 Chilean pesos, the average wage of the contributors to the pension system (and above the minimum wage stipulated by the Chilean legislation). Wages are assumed to grow at 2% per year in real terms and the contribution for disability and life insurance is assumed to be a constant 0.7% of the worker s salary, its average level during the last ten years 12. The cumulative charge ratio was highest for the cohorts who retired soon after the inception of the new system and falls gradually for later cohorts. During the early years of the system, over three quarters of the total contributions that were paid into the system were consumed by management fees. Since the first workers to retire from the new system only started to do so in the second half of the 1980s, few workers suffered from such exorbitant fees. Nonetheless, for the first workers to retire from the new system, the new system has been very expensive. For a worker earning the average wage who retired in December 2000 and had contributed each year to the system management fees would have consumed approximately one half of her total contributions. 13 system and retired before the year However, we also calculate the cumulative charge ratio for workers that will retire after this date. It should be noted that in their case, the cumulative charge ratio only offers a partial picture of total administrative costs over the person's career. 12 Precise information on insurance premiums prior to 1990 is not available. 13 The contrast of these results with previous evidence (e.g. James, Smalhout and Vittas, 2001) stems from focusing on commissions at a point in time, instead of their cumulative effect over a worker s career (as shown by the cumulative charge ratio). 21

22 Figure 4: Half the pension contributions of the average Chilean worker who retired in 2000 went to management fees! 95 Chile: cumulative charge ratio by year of retirement, salary = 200,000 pesos in December Percentage Source: Superintendencia de AFPs, authors calculations For younger workers who only started contributing after 1982 and will therefore only retire after 2022, the cumulative charge ratio drops to significantly lower levels (25-35%). The earlier cohorts, therefore, have borne the brunt of the set up costs of the new pension fund industry: the evolution of the fee structure has led to a redistribution of income from early (older) to later (younger) participants. While expected, such large intergenerational transfers are not unavoidable. 14 It is also interesting to note that after a secular decline, the cumulative charge ratio has began to creep up again for the youngest cohorts (those who will retire after 2034). This increase is attributable to the increase in the fixed commission, since the variable commission has actually fallen somehow over the last few years (see Figure 3). 14 Bolivia and Mexico, for example, allowed asset-based charges. Uruguay's reform, which required contributions to the new savings pillar only for higher income individuals, has also ensured that the new industry is subsidized in its early stages by those most able to create thickness in the market and endure high charges. 22

23 In countries where fixed commissions are permitted, even intra-cohort income inequalities can emerge. In Chile, even in the best of cases where workers choose the combination of flat and contribution-linked commission most appropriate to their salary level (i.e., that which minimizes the total commission as a percentage of salary), the poor end up paying a higher percentage of their salaries in commissions than the rich. The regressivity of the commission structure is clear from Figure 5, which shows how much higher is the cumulative charge ratio of the cheapest AFP for a middle income worker (300,000 pesos in 1990, 2% real growth p.a.) than the cumulative charge ratio of the cheapest AFP for a high income worker (900,000 pesos in 1990, 2% real growth p.a.). The gap was greatest at the beginning of the period, at over 3 percentage points. By the end of the decade it had fallen to about 7/10 of a percentage point, but largely as a result of the sustained increase in the cumulative charge ratio for higher income workers. Figure 5: Participation in the Second Pillar Is Costlier for Poorer Workers Chile: Cumulative charge ratio for different salaries, Charge ratio (%) Cheapest AFP - 900,000 pesos Cheapest A FP - 300,000 pesos Source: Superintendencia de AFPs Of course, a regressive charge structure would be less worrying if those AFPs chosen by poorer households offered a better service or performed better in terms of gross rates of return than other AFPs. There is no evidence that this is the case. Indeed, there is no correlation between the level of commissions charged by and the performance of a pension fund. As for the service offered, giving a worker more frequent or more detailed 23

24 communications on his or her accumulated balance is unlikely to compensate him or her for a lower replacement rate or net salary. Moreover, it would appear that many low income workers may not be choosing the lowest option given their earnings level, as demonstrated by the low price elasticity of demand calculated by Mastrángelo (1999) for Chilean AFPs. For Argentina, where AFJPs could also set fixed charges before November 2001, Rofman (2000) has calculated that the average commission (including the insurance premium) would be less than 3 percent instead of 3.4 percent if each contributor chose the cheapest AFJP for his/her income level. Commissions in the Retirement Stage. the administrative costs at the retirement stage appear to be lower than those during the accumulation stage. In Chile, for example, pension fund administrators did not charge a commission on scheduled withdrawals until In that year, all but one of the pension fund administrators began to charge a commission on the monthly contribution, which in December 2000 ranged between 1 percent and 1.25 percent. 15 Since disability and survivors insurance (from which retired people are excluded) accounts for approximately percent of the monthly contributions, the commissions on scheduled withdrawals are somewhat lower than those charged during the accumulation stage. The cost of annuities is more difficult to calculate because it includes a premium for insurance against longevity and investment risk. James and Song (2001) estimated that the money worth ratio of annuities in Chile (the ratio of the expected present value of benefit to the net premium) ranges from 95 percent to 100 percent for workers who have low discount rates (equivalent to the risk-free rate of government bonds) and from 86 percent to 90 percent for those with higher discount rates (a premium of 1.4 percent percentage points on top of the risk-free rate). Given that marketing costs account for approximately 6 percent of premiums, insurance companies appear to be providing good value for money to policyholders. As shown in Figure 6, sales commissions have increased in recent years in Chile. The cause of these high commissions (though lower than the ones applied by pension fund administrators) can be found also in the marketing process. As for the pension funds, annuities are sold directly to individuals through sales agents and insurance brokers. Workers receive little information on the different options they face and the comparison between fees and annuity rate across insurance companies often lacks transparency. One tactic used by 15 The remaining AFPs charged a fixed commission. 24

25 insurance companies to attract clients is to offer cash rebates which is often a decisive factor in the selection of annuity provider. Figure 6: The cost of buying annuities has steadily increased 7 Annuity com m issions / Prem ium s (%), Source: Superintendencia de Valores y Seguros The Brazilian System In Brazil, funded pension plans are a complement to the state-run PAYG-financed pension system. Most occupational plans (the so-called closed pension funds) are based on defined benefit formulas, with the worker s final salary as the reference wage. Increasingly, however, private companies have been introducing defined contribution plans as complements to defined-benefit schemes. Benefits were traditionally much higher for public sector workers than for other workers in the PAYG pension system. Since the introduction of Complementary Law No. 108 in May 2001, however, the contributions of public sector employers (including governments at all levels) as plan sponsors are capped at the same level as their employees contributions. While benefits for workers who stay with the same employer until retirement are relatively high and certain, workers with shorter tenures have traditionally suffered substantial portability losses on leaving the plan. Complementary Law No. 109, approved by the Brazilian Congress in May 2001, aims to reduce these losses by requiring all complementary pension plans to meet certain vesting and portability rules. As yet, however, the pension 25

26 regulator, the Secretaria de Previdência Complementar, has not introduced these regulations. III. The Governance and Operation of the Private Pension Systems The individual account pension systems are administered by two types of financial institutions: pension fund administrators and insurance companies. Banks, investment companies and other financial companies only participate in the pensions industry as owners of pension fund administrators or insurance companies or as recipients of investments from these companies. Pension Funds Open pension funds are the only vehicles for accumulating mandatory pension contributions in the new mandatory individual account pension systems of Latin America. Open pension funds are managed by financial institutions that are exclusively dedicated to this activity. 16 In all countries except Colombia and Mexico, participants can make additional voluntary contributions to their mandatory accounts. In Colombia, voluntary pension contributions are channeled into separate, so-called voluntary pension funds. These pension funds may be managed by the pension fund administrators or by so-called fiduciary societies (investment management subsidiaries of banks). Colombian employers may also set up closed pension funds to pay pension benefits to their employees. Ironically, workers risk and time preferences play no part in these new instruments for retirement saving, which have been otherwise hailed as a triumph for individual responsibility. Only since April 2000 in Chile have plan members been able to exercise some choice over their investment portfolio in accordance to their preferences. The introduction of the second type of pension fund, invested exclusively in fixed-income securities permits older individuals to transfer their savings to more conservative portfolios. Mexican pension law also envisages the establishment of multiple funds, but so far only one type of fund is functioning in the mandatory system. Individual investors in all other countries are permitted to have only one fund account and, therefore, have no power at all over the allocation of their 16 In Colombia, pension fund administrators also manage severance funds, which, like the pension funds, have individual accounts for each participant. 26