Chile: Selected Issues

This Selected Issues paper on Chile assesses the long-term outlook of Chile’s private pension system. The paper provides an overview of Chile’s recent experience with public–private partnerships (PPPs), focusing on the design of its institutional framework. The paper discusses that Chile’s experience with PPPs, currently covering 44 projects, has been successful. It also analyzes bank profitability and competition in Chile from an international perspective. It also compares Chile’s external private debt across industrial and emerging market economies and analyzes implications for external vulnerability.

Abstract

This Selected Issues paper on Chile assesses the long-term outlook of Chile’s private pension system. The paper provides an overview of Chile’s recent experience with public–private partnerships (PPPs), focusing on the design of its institutional framework. The paper discusses that Chile’s experience with PPPs, currently covering 44 projects, has been successful. It also analyzes bank profitability and competition in Chile from an international perspective. It also compares Chile’s external private debt across industrial and emerging market economies and analyzes implications for external vulnerability.

I. Addressing the Long-run Shortfalls of the Chilean Pension System2

A. Introduction

1. Over the past twenty-five years, Chile has undertaken an ambitious and pioneering pension reform, based on individual control and responsibility over retirement. In the late-1970s, the old public pension system had become in serious imbalance and was acting a hindrance to the economy. In 1981, Chile introduced a new retirement system based on personal accounts that allow workers to accumulate personal assets to fund retirement. At the same time, it reduced the role of the State, so that these personal accounts would become the primary means of support for most retirees. This reform was also aimed at preventing the costs of the system from exploding as the population aged.

2. This chapter reviews the history of reform and current experience of the pension system. The new pension system has helped lift national saving, promoted the development of the domestic financial market, and removed significant barriers to growth in the economy. Nevertheless, it is facing challenges, including from a lack of competition in the market for managing the retirement accounts and, more seriously, from infrequent contributions by workers.

3. The system will likely meet the needs of the average retiree over the medium term, but benefits will subsequently gradually decline in the long term. According to estimates presented here, the average worker’s pension income would replace about 60 percent of the final salary over the medium term (2030–40). However, over the long term, the average replacement rate would substantially decline to just over 40 percent, well below the level required for a healthy standard of living during retirement. These replacement rates are significantly lower than the 80 percent level promised at the time of the reform.

4. The system will risk leaving many workers without enough for retirement, especially among women. The average replacement rates mask considerable variation across retirees; nearly one-half of women and one-quarter of men would have replacement rates lower than 20 percent. Also, it is estimated that close to one half of the retired-age population would not qualify for the government’s main safety net program, because they would not have contributed frequently enough to the pension system. These trends will likely force changes in the future, and this chapter concludes with suggestions to improve the current system and potentially help relieve social pressure.

B. History of the Chilean Pension Reform

5. By 1980, Chile’s old, state-supported pension system had become unsustainable. The old system was a partial pay-as-you-go system and was intended to be self-funding, due to high contribution rates, equivalent to about 22 percent of wages. However, the system fell into chronic deficits, reflecting poor returns and overly generous benefits to selected groups (Cheyre, 2001). By 1980, the public pension scheme required annual government support of 2¾–3 percent of GDP (SAFP, 2003), and there was a recognition that the ageing of the population would further weigh on the system.3 Estimates prepared at the time suggested that the system could require annual state support of around 20 percent of GDP by the year 2000 (Wagner, 1983). Since public debt was about 90 percent of GDP at the time, urgent action was required to reduce the fiscal cost.

6. The old pension system was also imposing constraints on the labor market. Workers faced a very complex pension system, with over 100 different plans offered by 32 different pension funds. Benefits varied significantly from one regime to another; some workers were able to retire with inflation-indexed pensions at age 42, while many blue-collar workers had to wait until age 65 to retire (Edwards, 1996). The heavy payroll tax also weighed on job creation, and it also became difficult to switch from one industry to another, as workers competed for jobs in industries with generous retirement schemes.4 Finally, there was widespread dissatisfaction associated with the fact that pensions covered only 62 percent of the labor force.

7. There was also a need for the economy to generate more saving, to increase the pool of domestic capital for investment. By 1980, national saving had fallen sharply—to under 10 percent of GDP (Figure 1). In addition, the financial system was also quite small, with total assets under management equivalent to only around 50 percent of GDP. Reformers hoped to raise the saving rate, which would help reduce domestic borrowing costs. More importantly, it was thought that higher private saving would also stimulate the demand for financial products, thereby encouraging development of the financial industry.

Figure 1.
Figure 1.

Trends in National Saving

Gross national saving, in percent of GDP

Citation: IMF Staff Country Reports 2005, 316; 10.5089/9781451951615.002.A001

8. In 1981, Chile introduced a pioneering pension program, which replaced the government-backed system with a system based on individual retirement accounts. The new system was designed to allow workers to accumulate assets during their working life into individual private accounts. The law required that workers contribute 10 percent of wages (along with fees, that have totaled 2–4 percent of wages), each month, into a private account, effectively halving the contribution rate for pensions. Reflecting many systems at the time, the retirement ages for men and women were set differently, at age 65 and 60, respectively.

9. Private administrators were created to collect pension contributions and manage the individual accounts. These firms (Administradoras de Fondos de Pensiones, or AFPs) were designed to collect and manage the funds, but with strict regulatory oversight over their investments.5 The system allowed the AFPs to charge contributors a fee for collecting the fees and managing the funds. Since workers could freely select which fund to join, it was hoped that competition among pension funds would keep costs down. Initially, however, the fees, including the costs for the disability insurance, were quite high (around 4 percent of wages), reflecting in part startup costs.

10. When the new pension system was created, the old one was closed to new contributors, and the government offered incentives to switch.6 Workers who had j oined the labor force before 1982 had five years to decide whether to join the new system or to stay under the old one. The government put in place three important incentives to encourage workers to move. First, the government lowered contribution rates dramatically, and workers saw substantial pay increases under the new scheme—averaging around 12 percent. Second, all contributions to the new system were made tax deductible. Third, the government left the benefits levels under the old regime largely unchanged, so that many people who participated in the less generous plans switched to the new one (which also makes comparisons of benefits under the old and new systems difficult).

11. To compensate the transferees from the old to new system, the government put in place a system of recognition bonds that partially reflected past contributions. The recognition bonds, designed to represent accumulated contributions to the old system, carried a real interest rate of only four percent, well below market rates (real rates on 90-day bank deposits averaged nearly twice that level during 1975–1990). As a result, the effective value of the bond did not represent the full value of the workers’ contributions. These bonds will play an important role for those retiring in coming years, representing 50–70 percent of the retiree’s pension account balance—with the remaining coming from contributions to the new program (Arellano, 1985 and Arenas de Mesa and Marcel, 1993).

12. To help ensure that workers do not outlive their account balances, the program also required that workers withdraw retirement income gradually, through either an annuity or a series of programmed withdrawals. Plan designers realized that retirees typically underestimate how much income they will need for retirement. As a result, they required that, at retirement, contributors use their individual accounts to purchase an annuity from insurance companies or draw down on their individual accounts in pre-specified amounts, to ensure that their balances are not drawn down too quickly.

13. Those retirees who choose to annuitize their account balances thus depend on the development of a well-functioning annuities market. Because workers are self-funding their retirement, they need protection against the risk that they may live longer than expected. By purchasing an annuity, it is possible to share the risk across many individuals, with those who die early compensating those who live longer. However, for this market to work well, insurance companies must operate competitively, and they must also be well regulated to guarantee that they will be able to honor the contracts with their annuitants.

14. The plan designers also foresaw that the programmed withdrawal option could raise the risk that retirees would outlive their pensions, so the reform also included a safety net—the minimum pension guarantee (MPG).7 The designers of the pension reform also realized that some contributors might not accumulate sufficient assets in their individual accounts to fund an adequate pension. As a result, the government would provide a safety net to workers who have contributed at least twenty years (240 months). After drawing down their accounts completely, qualifying pensioners would then receive the MPG income, to guarantee that their real income remains at fixed level.

15. For workers remaining under the old system, the government was committed to providing ongoing fiscal support to fund any revenue shortfalls. The old programs were consolidated under one government agency, Instituto de Normalización Previsional (INP), and the retirement ages gradually raised to 65 for men and 60 for women across all pension regimes. With the underlying imbalances between contributions and benefits persisting, the government was committed to financing the deficit of the INP. The government has also been providing a small assistance pension to the indigent.8

C. Trends in the Operation of the Pension System

Coverage and Participation

16. Initially, there was significant interest in the new system, and coverage expanded rapidly. Within five years of the start of the system, nearly half of workers were contributing (Table 1). These high rates, attained in such a short period of time, reflected the high level of interest in the new system. Employers liked the elimination of payroll taxes, and employees took advantage of the opportunity to earn higher wages to shift to the new system. In addition, respondents in the First Social Protection Survey report that just over half of those workers that switched did so because it was set as a pre-condition by their employer.

Table 1.

Pension System Participation

article image
Source: Haver Analytics, SAFP, and staff estimates.

17. Pension reformers created the new system on the assumption that workers would contribute most of their working lives. The designers of the new pension system thought that, eventually, most workers would have a high contribution density (the ratio of years spent contributing to number of working-years lived). They expected this density to be at about 0.6–0.7—suggesting that workers would contribute for nearly two-thirds of their working lives. As a result, they designed the minimum pension guarantee (MPG) so that the minimum contribution density needed to qualify for it was significantly lower—around 0.44 for men and 0.5 for women.9 Based on these assumptions, the reformers forecasted that retirees would see dramatic increases in their income—with the new system providing pensions that would replace 80 percent of final salaries.

18. A recent, broad survey of households reveals that the actual contribution density of workers has been only 52 percent, with considerable variations across socio-economic categories (Table 2).10 As highlighted in Arenas de Mesa, Behrman, and Bravo (2004), the contribution densities vary significantly by income, gender, education, and employment opportunities. In general, the poor, women, the self-employed, workers without a labor contract or with little education all contribute less often to the system. On average, most socio-economic groups—except wage earners and those with labor contracts—do not contribute close to the density levels assumed when the regime was put in place.

Table 2.

Contribution Densities

by selected socio-economic variables

article image
Source: Arenas de Mesa, Behrman, and Bravo (2004), Tables 8 and 10.

Densities of those who have completed schooling.

19. The data also suggest that women tend to contribute infrequently, and that many will not qualify for the MPG program. The average contribution density for women is only 44 percent, below the minimum level needed to qualify—suggesting that around one-half of women will not qualify for the MPG. This, in turn, reflects the low labor market participation rate for women—about 35 percent, compared with over 70 percent for men.

20. For women, low contribution patterns in their young and late years seriously affects their retirement savings. Many Chilean women are not in the labor force during their 20s and 30s because they are raising children or taking care of the household.11 In a system of individual pension accounts, this period is the most important one to accumulate assets. By missing the first ten years of contributions, they fail to take advantage of the benefits of compound interest, which effectively cuts their savings in half (Bernstein, Larraín, and Pino, 2004). At the other end of their working lives, retiring at age 60 encourages them to stop contributing sooner and, thus, they accumulate fewer assets. Estimates in Bernstein and Tokman (2005) show that, if women contributed until age 65, their retirement fund would be 30 percent higher.

21. The self-employed is the other major group that fails to contribute regularly. Under current rules, the self-employed (about 25 percent of employment) are not required to make social insurance contributions for pension, health, and disability. In all, these contributions amount to 23 percent of wages.12 According to surveys, the self-employed do not wish to contribute out of concern that they would not be able to access their pension savings during months when income is low. Surveys also show that they tend to think—mistakenly—that, if they contribute to the system one month, they have to pay into the system every month. As a result, over three-fourths of the self-employed contribute infrequently or not at all.

22. Finally, records from the pension system administrator suggest that contribution densities may be even lower than reported in the survey. The household survey relied on individuals to self-report the frequency of contributions to the system on a monthly basis from 1980 to 2002. The regulator of pension funds has, in coordination with the Ministry of Labor and Social Welfare, matched the respondent’s answers to actual AFP data and found that respondents systematically overestimated how often they had contributed.13 As a result, actual contribution densities may be even lower than reported here.

Rates of Returns and Fees

23. Gross returns to the pension fund system were strong initially, due to exceptional circumstances in the financial markets. Gross real returns averaged around 12 percent from 1981 to 1997 (Figure 2). The system managed produced such strong returns because the pension funds experienced two windfalls. First, pension funds held a large stock of bonds, whose prices surged as inflation fell and interest rates declined. Second, they realized a large capital gain on equity shares bought during the privatization drive of the 1980s. Since 1997, however, gross returns have fallen sharply, averaging only around 6½ percent in 1998–2003.

Figure 2.
Figure 2.

Pension System Returns

Real returns by AFPs, 5-year rolling average, both gross and net of administrative costs and insurance premia.

Citation: IMF Staff Country Reports 2005, 316; 10.5089/9781451951615.002.A001

Sources: SAFP (2002).

24. Net returns have been significantly lower than gross returns, due to high management fees. Net returns averaged only around 3 percent into the late 1980s, reflecting fees equivalent to 6 percentage points of returns.14 High start-up costs were to be expected, because there were significant fixed costs, including the establishment of complex computer systems to manage millions of individual accounts. The AFPs also conducted expensive marketing campaigns to convince contributors to switch to the private pension system. Net returns have increased during the 1990s and in more recent years, averaging just over 8 percent a year from 1991–1997 and just over 5 percent in recent years.

25. Although management costs have declined substantially in recent years, they remain high, at around ¾ percent of assets. Since the mid-1990s, management fees have fallen, thanks in large part to regulatory pressure to limit wasteful marketing.15 The current fee (of around 2½ percent of wages), equivalent to 0.6-0.8 percent of total assets under management, is still high when compared to low-cost providers in other countries. For example, the U.S. federal civil servant Thrift Savings Program, which offers only five funds (similar to the current regime in Chile), costs around 0.07 percent of assets (CBO, 2004). Low-cost private providers, such as Vanguard and Fidelity, charge approximately 0.2-0.3 percent of assets for their broad index funds to the individual retail investor, and fees are even lower for large account holders. Current Chilean costs are at about the same level as U.S. mutual funds, which offer a very wide range of choices in investments.16

26. There is scope to lower costs somewhat further, even though the average account is small. The median account balance is approximately US$2,900. With such small accounts, even the cheapest retirement fund in a low-cost U.S. provider would impose a low balance fee that would bring the expense ratio closer to 0.5 percent of assets. This suggests that, although the current system may not be able to reach the lowest costs of the U.S. industry, costs could be reduced by around 0.2 percent of assets, which would represent substantial savings of around 25 percent.

27. Fees could also be lowered by increasing competition within the AFP industry and taking advantage of economies of scale. Currently, there are only six AFPs, which are the dominant players in the financial system. The 2004 FSAP for Chile has recommended steps to increase competition between these AFPs. It also pointed out that the AFPs are responsible for some administrative functions (such as collecting contributions) that have natural economies of scale, and the report suggested further cost saving by centralizing operations (IMF, 2004).

Replacement Rates

28. Currently, the system seems to be performing relatively well against the ultimate measure, i.e. generating sufficient income for its retirees. According to the World Bank (1994, p. 75), the minimum target replacement rate should be 60 percent of lifetime wages, which—given the lifecycle of Chilean earnings—is equal to 50 percent of final salaries. By this measure, the system appears to be performing well for the average retiree. Aggregate data from Primamérica (in Chile) on the average value of pension income suggest that the 2002 replacement rate was around 60-65 percent. However, more disaggregated data using the household survey suggest that replacement rates may be somewhat lower, at around 50-55 percent. While these average replacement rates may be sufficient to maintain a standard of living in retirement, they are also substantially below the 80 percent promised at the time of the reform in the early 1980s.

29. The overall average replacement rate is thought to mask sharp differences arising from the wide disparities in incomes and contribution densities. As Table 2 indicates, there is some correlation between a worker’s income and contribution density. Thus, the pension system is reinforcing an already-high level of income inequality in Chile. Preliminary estimates using household survey data suggest that about one-third of the population may have very low replacement rates of 10–20 percent. Replacement rates would be so low because a large number of retirees cannot annuitize their income. Instead, they must draw down their balances using the programmed withdrawal option, which would mean that after several years, their pension accounts would be exhausted. At the other end, well-off contributors would likely experience replacement rates of 80 percent or above.

Fiscal outlays

30. The transition costs for the government have been roughly constant, at around 4 percent of GDP (Table 3). In Chile, a conscious decision was made to front-load most of the transition costs, and the transition to the new regime has been expensive, especially by Latin American standards (Figure 3). On average, it has required more support in Chile than in the rest of Latin America by about 3 percentage points of GDP. The primary component of the cost has been support to the operating deficit of the old plans. In recent years, following a decline in the number of retirees under the old system, this cost has shrunk by around 1 percent of GDP, to 2¾ percent in 2004. The second component of the transition cost has been the redemption of recognition bonds by retired workers who had switched from the old to new system. This cost has grown by around 1¼percent of GDP, offsetting the decline in support for the old regime.

Table 3.

Government Spending on Pensions

in percent of GDP, by selected years

article image
Source: Arenas de Mesa (2004).

Net cost (benefits less contributions).

Figure 3.
Figure 3.

Spending on Pensions in Latin America, 2000-2020

Annual fiscal deficits of the pension systems

Citation: IMF Staff Country Reports 2005, 316; 10.5089/9781451951615.002.A001

Source: Mesa-Lago (2000). Estimates may not agree with tables because of methodological differences.

31. The cost of government support to the remaining pension programs has remained moderate, at 1½-2 percent of GDP. The system of the armed forces has undergone reforms, to help reduce the cost to the government. During the early 1980s, spending averaged around 2 percent of GDP but, after the reforms, it declined to around 1¼ percent of GDP. The cost of assistance pension programs is only around ½ percent of GDP. Finally, the cost of the minimum pension guarantee has been relatively small (around 0.1 percent of GDP in 2004) because there are still only few retirees under the new pension system.

Pension System Costs in an International Perspective

32. By international comparisons, Chile is in an enviable position, as its pension program is not facing explosive costs. According OECD estimates prepared by Dang, Antolin, and Oxley (2001), Chile’s total public spending on old-age pension programs compares favorably with current spending by OECD countries (Figure 4). The average OECD economy spends around 7 percent of GDP on support for old-age pension programs, or about 1 percentage point more than in Chile. Furthermore, unless OECD countries reform these programs, their costs could nearly double over the medium term. Meanwhile, reflecting the front-loading of the transition, Chile’s peak outlays are now likely in the past.

Figure 4.
Figure 4.

OECD and Chile Old-age Spending

Gross current expenditures in 2000.

Citation: IMF Staff Country Reports 2005, 316; 10.5089/9781451951615.002.A001

Source: Dang, et. al. (2001), Haver Analytics and staff est.

33. Chile’s average replacement rates are also in line with international experience. According to recent estimates, the average replacement rate is 65 percent for retirees in OECD economies (OECD, 2003).17 Furthermore, if only the public pension programs are considered, the average replacement rate for OECD countries is only 48 percent. Chile’s replacement rate of around 50-60 percent is in between these averages, suggesting that the current system may not be far outside of international norms.

34. Nevertheless, the Chilean system lacks the progressivity that exists in other pension programs. The OECD estimates suggest that there is a considerable degree of progressivity in most OECD pension regimes. Indeed, poorer workers in the OECD can expect a replacement rate of over 70 percent, while wealthier workers can expect a rate of around 40–50 percent. In Chile, the situation is broadly reversed—and the system considerably less progressive.

35. To address some of the pending issues, in 2002 the authorities introduced a number of reforms aimed at encouraging competition and making contributing easier (Box 1). In particular, the multi-fund system has encouraged workers to take a more active role in managing their investment portfolio. Also, higher foreign investment limits (30 percent) should help ensure maintenance of a high rate of return to the system. Finally, the voluntary savings system may contribute to an increase in contributions by the self-employed.

2002 Reforms to the Pension System

The goversnment introduced significant reforms to the pension system to help improve its operation. By the late 1990s, there was a recognition that the system needed additional competition and that the AFPs should be more accountable to their members. The changes included: (i) introducing a multi-fund system of accounts (similar to the index funds in a mutual fund); (ii) raising the foreign investment limits; and (iii) enhancing the role of tax-deferred voluntary saving accounts. Details on the measures are presented in SAFP (2002), Chapter 8.

The multi-fund system was introduced to increase the number of investment options. Previously, contributors could only contribute to one of two types of funds. The lack of options made it difficult for the AFPs to compete on the basis of the rates of return, given the strict limits on the types of investments and amount of shares that the AFPs can hold. At present, the AFPs can offer five different types of funds, with a variety of risks (ranging from one fund that is primarily equities to one that is primarily bonds). Thus, workers can tailor their investment choices based on their willingness to tolerate risk. Workers older than 55 are limited to the more conservative funds, so as to limit their exposure to a sudden market decline.

Foreign investment limits have gradually been raised to 30 percent. Prior to 2002, foreign investment could make up no more than 16 percent of an AFPs assets. This limit was gradually lifted in stages, to 30 percent, by March 2004. Each time the limit was lifted, the AFPs increased their foreign investment quickly, to near the maximum limit.

In order to stimulate voluntary savings, reforms made it easier to contribute to, and draw upon, voluntary savings accounts. In the past, it was possible, but difficult, to contribute to a voluntary savings account; furthermore, it was not possible to draw down the account until retirement. Now, workers can draw down on these accounts, subject to a 3-7 percent surcharge, helping to ameliorate concerns of workers that they could not use the money if they became unemployed. In addition, the new law also made it easier for the self-employed and INP members to contribute to these accounts.

D. Medium-Term and Long-Term Outlook

Medium-term outlook for Replacement Rates

36. Through the medium term, the system is set to likely to perform well, with the average rate of return drifting upward at least until 2020. For those retiring in coming years, the average return on their funds is likely to be high, reflecting the strong asset growth registered through the 1990s. The gross returns for those retiring in the medium term could plausibly increase by one percentage point between 2000 and 2020 (Figure 5).18 On that basis, replacement rates could trend upward by 10 percentage points, reflecting larger account balances, and peak at about 60-70 percent percent for the average retiree by 2020.19

Figure 5.
Figure 5.

Long-run Returns of the System

Average annual gross real returns, 40-year moving avg.

Citation: IMF Staff Country Reports 2005, 316; 10.5089/9781451951615.002.A001

Source: SAFP and staff estimates. Assumes a 4 percent return before 1981 (reflecting real recognition bonds) and a 5 percent return in the medium term. (after 2004).

37. Over the long term, however, replacement rates are projected to fall. The ncrease projected for the medium term would begin to fade after 2020, when the proportion of those who contributed during the periods of exceptional returns would decline. By the end of the medium term, the real rate of return to retiree’s account balances would decline gradually and, by 2030, replacement rates would have gradually drifted back down to current levels.

Medium-term fiscal costs

38. Total old-age programs could decline through the medium term, due to substantially lower-than-expected costs for the MPG program (Table 4).20 Estimates provided by Arenas de Mesa and Marcel (1999) and Schmidt-Hebbel (2001) had suggested that spending on the MPG would increase to 1¼ percent of GDP over the medium term. However, these estimates assumed a constant contribution density of around 0.7 percent, while actual contribution levels are much lower. As a result, government spending on the MPG is expected to be only around ¼ percent of GDP in the medium term.

39. The low spending on the MPG suggests that it will not constitute the broad safety net that it was intended to be. As noted, workers need to contribute a minimum of 20 years to qualify for the MPG. Table 4 shows the large gap between the percentage of the retired-age that would qualify for the MPG and the percentage that would actually receive it. While nearly half of total retirees would qualify for the MPG, only around 10 percent would draw on the program, almost exclusively women. The gender differences arise because men who would contribute for twenty years or longer are forecast to have accumulated sufficient assets in their individual fund not to need the MPG. Furthermore, the contribution levels are so low that men who do not qualify would have contributed, on average, for just 11 years, and women 10 years. This suggests that nearly half of the population, mostly women, are contributing much less often than was assumed when the MPG was designed as a safety net.

Table 4.

Medium-term Pension System Outlook

Based on current contribution densities

article image

40. While the MPG cost estimates are subject to some uncertainty, it is unlikely that they will be much higher. The pension system has an option for workers to retire early, if their account balances are large enough. Workers can also withdraw money in a series of programmed withdrawals that tend to be higher than an annuity during the initial years. Both of these options raise the risk to the government that pensioners will outlive their pensions and will need to subsequently rely on the MPG. The fiscal impact of these options may be limited, however, given that about 85 percent of early retirees annuitize their accounts, reducing the risk that they will outlive their pensions. Furthermore, only about one quarter of all retirees choose the programmed withdrawal option (James, 2004).

41. Other old-age spending could also increase moderately. Spending on the armed forces and assistance pensions could increase slightly, each by around ¼ percent of GDP between now and 2030.21 The assistance pension has broad coverage, reaching around one-third of the elderly in 2004. However, its costs have not increased substantially over time because payments are relatively small (around US$65 a month in 2004). Furthermore, it is only available to retirees once they have exhausted all other pension income.

The long-run outlook

42. In the long run, however, the ageing of the Chilean population raises concerns that spending on the minimum pension guarantee (MPG) could increase significantly (Figure 6).22 The over-75 population is set to double (as a percent of the total population) between 2030 and 2070, when it stabilizes at around 12 percent of the population. Since this group of elderly retirees is the one most likely to outlive their pensions, pressure could build for the government to improve coverage.

Figure 6.
Figure 6.

Demographic Trends, 1980-2100

Ratio of workers to retired-aged, and share of retired-age population in the total population.

Citation: IMF Staff Country Reports 2005, 316; 10.5089/9781451951615.002.A001

Sources: Haver Analytics and staff estimates.

43. Long-term estimates were built using a simple actuarial model of workers’ accounts. The model assumes that account balances grow, predictably, as a function of the net rate of return to the pension funds, contribution density, and wages. The model accounts for differences suggested in the First Social Protection Survey in income inequality, contribution densities, and gender. It also includes estimates of the impact on contribution densities if the self-employed were brought into the system. Economic variables grow in a deterministic manner, as described in Table A.1. These estimates also consider the effect of broadening the coverage of the assistance pension, currently targeted to the indigent.

44. These estimates suggest that, in the long-run, replacement rates could fall substantially, to about 40 percent (Table 5). The average replacement rate in the long-run would thus be significantly below the 50 percent goal needed to sustain living standards during retirement. Furthermore, the gender differences in contribution densities would result in substantial disparities, with men having replacement rates 20-30 percentage points higher than for women, who would have a replacement rate of only 30 percent.23

Table 5.

Long-run Pension System Results

Based on current contribution densities

article image

2020-2030

Results when the economy in long-run equilibrium

45. This fall in replacement rates is due in large part to the decline in the rate of return on financial products, which would adversely affect both asset accumulation and the cost of annuities when workers retire. This decline in replacement rates, forecast to start between 2020 and 2030, would mirror an increasing share of workers who did not contribute to the pension system during its period of exceptionally high returns. It would lead to smaller account balances, which would thus purchase smaller annuities and lower incomes during retirement. In addition, the decline in the rate of return on financial assets would also affect the annuities market because, with lower rates of return, insurance companies would need to charge more for the annuities to cover their costs. As the cost of annuities increases, replacement rates would be driven down.

46. These results could be mitigated if workers increased the frequency of contributions. The estimates assume that the contribution densities are fixed at current levels. It is possible that some workers could begin to contribute more to the system, when they realize that they will not have enough for retirement. However, a decline in the rate of return on financial investments may not be properly assessed by contributors, in part because it may be difficult to identify against normal fluctuations in the financial markets. Furthermore, the effects would compound over several decades, making it difficult for workers to judge whether to they need to take action until it is too late.

47. In the face of these trends, in the long-term government spending on old-age programs would return to its 2004 levels, of around 5 percent of GDP (Table 5). The increase projected to take place after 2020–30 would mostly reflect an increase in payments under the MPG and assistance pension programs (with increases of ¾-1 percent of GDP per program from 2030 onward). The costs to the government of the old system would stop shortly after 2030, just as the long term costs of the MPG rise.

48. MPG spending program would only increase gradually, because many workers have too low contribution densities to qualify.Over the long-run, more workers would need the MPG because the account balances would be significantly lower than those who retired before 2030. About one-fourth of retirees would receive some form of MPG payment. This represents an improvement of the medium term coverage of the system, but many of the poorest retirees would still receive no government support. Given that spending on assistance pensions would remain higher than MPG spending, the assistance pension could become the default safety-net, even though it is not necessarily designed to do so, and the program does not have a natural connection with the pension system.

E. Possible Next Steps

49. Given the concern regarding the prospects for replacement rates, this section considers possible policy changes to increase participation and broaden benefits. Some of the adjustment will have to come from individual workers, in the form of higher contributions. However, it will be important also to phase-in any changes gradually, to give workers time to adjust their working and retirement plans. At the same time, any reform effort should also aim at broadening benefits by using some of the current fiscal space due to low anticipated fiscal costs and meager benefits under current policies. Additional resources could come from government “pre-funded” accounts that help reduce old-age poverty.

Reforms to increase participation

50. Steps could be taken to address the low participation of the self-employed in the social security system. Currently, few self-employed participate in the system, and this could be addressed by requiring them to contribute (at a minimum) to the pension system. According to estimates from the First Social Protection Survey, well over 50 percent of the self-employed have never contributed to the social security system; this represents nearly 15 percent of the Chilean work force. Nevertheless, many workers are avoiding making a payment, because doing so would mean that they would need to pay into the entire social security system (around 23 percent of wages). Instead, partial exemptions could be granted, requiring the self-employed to contribute into the AFP system and not the health system. Such a step could raise the replacement rate for men by 10–15 percentage points, but by less for women, because fewer women are self-employed.

51. Requiring all self-employed to contribute could be challenging, since many do not file taxes and could take informal jobs. Only two-thirds of the self-employed file tax returns, making it difficult for the authorities to know what their actual incomes are (Fuentes, 2004). Alternatively, many self-employed could avoid coverage by taking up informal jobs.

52. Replacement rates could also be increased if the retirement age for women were raised to the same level as men. By forcing longer contribution periods, more women would likely qualify for the MPG. Furthermore, a later retirement age would increase asset accumulation, lifting replacement rates for women by around 10 percentage points.

53. Nevertheless, addressing the gender differences in the coverage of the pension system will be more difficult and would likely require broader changes to labor and social policies. Changing the retirement age would likely be politically difficult. Furthermore, women’s low labor market participation reflects social decisions, along with economic incentives for women to work. Raising the replacement rates, and the benefits of the system, to women would likely require a number of social policies to encourage women to work, but that are not strictly related to the pension system (such as childcare). However, it is conceivable that some steps could be taken to enhance benefits, such as introducing a system of government contributions to the primary care giver’s retirement accounts during early child-rearing years.

54. Simulations using the actuarial models suggest that it may be possible to improve replacement rates by switching a system of asset-based fees. Currently, AFPs are prohibited from charging fees linked to the size of the assets of account holders. According to simulations conducted using the actuarial model, an asset fee set at 0.65 percent of assets would generate a similar level of income for pension funds as a wage fee of 2¼ percent (Table A.2, in the appendix). However, it would substantially change the incentives of the AFPS, because they would have an incentive to maximize the value of their assets—as opposed to the current system where they want to maximize the number of high-wage contributors.

55. With such a system, safeguards would be needed to ensure that contributors have a very clear idea of the costs. While an asset-based fee system can align the incentives of the fund managers with the contributors, there are possibilities for abuse. For example, pension funds could compete in a costly attempt to attract contributors with high balances (as opposed to high wage earners). In addition, asset fees can be more difficult to understand, so contributors would need to be educated on the impact of asset fees, not just on the current cost, but also on the future impact of higher fees on lower account balances. Statements sent to account holders would have to specify clearly how much money is being withdrawn from their accounts. To encourage competition, contributors could be told how much they would save if their accounts were at the cheapest AFP.

Possible steps to improve benefits

56. Since workers could face a substantial decline in their replacement rates, future benefits could be pre-funded with a special account. The pressures to the replacement rate are not likely to be fully felt for another twenty years, at a minimum. As a result, there is more than enough time to pre-fund an account that could provide benefits to retirees to help make up the difference.

57. One option would be for the government to deposit money into the accounts of young workers. This would act as a system of forced savings, and take advantage of the fact that the pension funds can diversify their portfolio internationally, and generate higher rates of return than the growth rate of the economy and the tax base. However, there could be intergenerational equity concerns with limiting the program to the future young, especially since the current cohort of workers (25–55 years old) is paying for the transition cost of the old system as well. Mitigating this, however, is the fact that many in the current generation benefited from exceptional returns to the pension funds that are unlikely to be repeated. Also, rules would need to be set up to prevent the grant program from discouraging saving for retirement and to ensure that the subsidy complements workers’ contributions—and does not substitute for it.

58. The old-age safety net could also be improved, at a moderate fiscal cost, by reducing the time to qualify for the MPG. This reform would substantially boost the numbers of pensioners receiving the MPG, which would help relieve forthcoming pressure, as more retirees realize that they may not qualify for the MPG. If the time to qualify was set at 15 years, about one-third of retirees (given current patterns in contribution densities) would still not qualify. As a result, it would cost only an additional ¾ percent of GDP a year to the government to expand the program. The relatively modest fiscal cost suggests that coverage would still not be broad enough for the MPG to serve as a broad safety net.

59. There is scope for redesigning assistance pensions so that they serve as an effective anti-poverty old-age program that complements the private account system. The current assistance pension program could be redesigned with higher benefits, since the current subsidy is only around one-third of the minimum wage. Unless this level were raised, it would be unlikely that the assistance pension could serve as an effective anti-poverty device. However, the program would need to be modified, so that the subsidy does not serve to discourage participation in the formal pension system. For example, benefits could start later than the traditional retirement age.

60. To limit the fiscal costs, any enhancements to assistance pensions would need to be flexible and respond to changing demographics. Over the next 50 years, the Chilean population will age quite sharply. As any enhancements to the pension system could become very expensive in the long run, new benefits could be linked to the increase in life expectancy. For example, many OECD economies are beginning to consider rules that would tie the retirement age to life expectancy.24 Furthermore, the program would need to be aimed explicitly at reducing old-age poverty, and spending priorities would need to be adjusted if poverty rates in other age categories increased above the old-age poverty rates. This program should be designed and explained as an anti-poverty program, and workers encouraged to continue to contribute to their own accounts—to ensure that they have more than just a minimal retirement income.

F. Conclusions

61. The results from the previous sections suggest that Chile’s pension reform has served some retirees well, but many others will face a shortfall that will grow ever more pressing over the next several decades. In the short-run, the average replacement rate is adequate, but it masks considerable variation between groups, so that 20–30 percent of the retired age population has very low replacement rates. While the replacement rates are projected to improve in the medium term, they will subsequently decline to levels at which they would not be high enough to ensure that retirees can maintain their standards of living.

62. From a fiscal perspective, future pension costs appear on a declining trend, but estimates may be deceiving because social pressures could force sharp changes to the old-age programs. The low coverage of the MPG indicates that it would not be, as initially intended, a broad based anti-poverty program for the elderly. Under current policies, many elderly would have retirement incomes that would last for only several years, because their account balances would be small, leading many to empty their accounts after only 5-10 years in retirement. Current policies would thus make many elderly look for work during retirement, even though many may not be able to do so. As a result, social pressures would likely build to provide broader anti-poverty programs for the elderly, but without sufficient time to pre-fund their costs.

63. Action should be taken now to pre-fund the system and design a proper anti-poverty program for the aged. The assistance pension was not designed to work with the current system. However, given the expected increases in life expectancy, Chile cannot likely afford a broad anti-poverty program at current retirement ages (65 for men, 60 for women). In particular, any debate regarding the expansion of the anti-poverty program would need to consider the long-run costs and demographic trends, and action should be taken as soon as possible to provide for retirement programs over the long run.

64. Ultimately, prudent macroeconomic management is vital under a system of individual retirement accounts. Since workers must contribute during their entire working lives, they need to be able to find work, and the financial markets must be stable. If the economy is poorly managed, and unemployment or inflation jump sharply, workers would face permanently lower retirement incomes. Also, since there is limited intergenerational risk-sharing from macroeconomic shocks in a system of individual accounts, economic policies must remain sound. Finally, the labor market needs to function well in order to ensure that workers can move from one job to another, without losing too much time. With a system of private accounts, any time spent outside of the labor force means that workers are not contributing to their accounts, and hence, that they will ultimately have lower pensions.

APPENDIX I: Modeling of the Results of the Chilean Pension Fund System

The most important element of the model used in the paper is that it is deterministic, so that account balances grow smoothly over time, depending on the: (i) contribution amount and fees, (ii) contribution density, (iii) time spent working, and (iv) rates of return. In this model, workers contribute a fixed percentage of their wages to a pension account, and fees are deducted from their contributions (in the case of income-based fees) or account balances (in the case of asset-based fees). However, this model assumes that workers will not contribute to the system every month of their working life, so the contribution density, or ratio of the number of months of contributions to the months during their working lives, is less than one. All workers retire at the statutory retirement age or stop contributing to the pension system. The account balances grow at a fixed, deterministic rate, based on the assumed returns to the pension fund system.

Heterogeneity is introduced in the model by allowing contribution densities, income, and macroeconomic and demographic variables to vary. These parameters can be changed, so as to produce a range of estimates and to test the sensitivity of the results, including by:

  • Contribution densities, that vary across the working population, using an empirical estimate of the probability distribution derived from the 2002 First Social Protection Survey of households. The contribution densities were divided into one of 101 cohorts (ranging from zero to one, in 0.01 increments), and a kernel density estimator was used to predict the probability mass at each density. The distribution of contribution densities is assumed to be constant across time, but differs by gender.

  • Self-employment, which is modeled as a period where the cohort is not contributing; however, using the survey. In addition, Table A.2 considers the impact of including a certain percentage of the time spent self-employed is covered by the pension system.

  • Income that is conditional on the contribution density, as estimated from the 2002 Social Protection Survey. The income distribution was calibrated to match to the income distribution calculated by the SAFP.

  • Macroeconomic and demographic variables, which are allowed to change over time. These are largely driven by a sharp decline in population and labor force growth during the forecast horizon. As a result, real GDP growth also falls–to keep labor productivity growth constant. In order to maintain economic consistency, wage growth, financial returns also fall.

As a result, the account balance of a worker accumulates in a predictable manner. Because of the assumption on contribution rates and income, these vary by the cohort of worker (and across time), with 101 cohorts of workers in each year. Following work by Arenas and Gana (2001), the real balance (Ki,t) of each worker within cohort (i) grows each year (t) by:

Ki,t=(1+rtfa)Ki,t1+12di(1+rtfa2)(cfw)Wi,t ,(1)

where r is the return on the pension fund, fa is an asset-based fw, is an income-based fee, di is contribution density for cohort i, c is the contribution rate, and Wi,t is the real wage for cohort i at period t. This paper follows their assumptions—that real wages (Wi) grow from age 18 (when a worker enters the labor force) until age 50. Then, wages plateau, remaining constant in real terms from 50 until retirement. See their paper for the complete derivation.

At retirement, the entire balance of the pension fund is converted to an annuity, if the retiree can afford a annuity income greater than the MPG. The annual annuity income depends on the gender of the worker, life expectancy, and the discount rate. The life expectancy data for estimating the annuity income for each cohort is taken from the SAFP (using table RV-2004). Nevertheless, the population data (used to calculate the size of a retiree cohort) is taken from the United Nations population estimates, which assumes that life expectancy is drifting upward slowly. Finally, the discount rate is assumed to drift downward, reflecting the decline in the real rate of return to pension funds. However, this annuity option is only available to those workers with account balances that are large enough.

For those workers whose cannot afford a MPG annuity, they draw down their account balances so that their pension income is equal to a MPG, until their account balance is exhausted. Strictly speaking, a retiree must draw down their income using the programmed withdrawal option, which generates a higher, initial income that declines over time. However, this formula is complex, so this simple approximation is used. For those cohorts who have contributed for 20 years or more, the state then funds a pension equal to a MPG for the rest of their lives. For everyone else, their pension income is zero.

These estimates are reproduced by contribution density-cohort, and when a cohort reaches the retirement age, this model assumes everyone retires. Using the empirical distribution estimated from the Social Protection Survey, it is possible to calculate the fraction of the population in each cohort, by gender. Ideally, the contribution density would vary by age (as well as by gender), but this aspect is not considered here.

As the retiree cohort ages, the mortality tables (RV-2004) are used to allow the cohort to “age”, so that for every year, it is possible to compute the total spending on the minimum pension guarantee. This step is necessary, because as each cohort retires, the retirees receive an annuity, which is fixed in real terms.25 However, the minimum pension guarantee grows (in real terms), so it is necessary to track the “gap” between the each retiree-cohort’s annuity income and the minimum pension guarantee. Furthermore, the minimum pension guarantee is different for those under age 70 than those over age 70, so retirees are tracked by age.

Because data on individual account holders are unavailable, the estimates through 2020 are only approximate. From 2005–2020, the model uses highly aggregated data to estimate the balances of retirees. Furthermore, it treats all retirees as part of the private pension system; however, most retirees will still be part of the old system until the early 2010s. As a result, until 2020, the estimates for the minimum pension guarantee are “too low”.

In addition, these assumptions are likely to be optimistic, since workers typically fail to save enough and usually underestimate how long they will live. As such, the deterministic nature of this model is likely to over-predict pension income and replacement rates. In particular, Chilean workers have a propensity for self-employment, to avoid mandatory social contributions. As a result of avoiding the short-term costs, the long-run fiscal costs are likely to be higher (and replacement rates, lower) than estimated here.

Furthermore, to keep the model compact, several elements of the current pension program are excluded from the model. First, all fees are assumed to be based on either income-based or asset-based; the fixed commission is assumed to be zero, and the costs associated with the disability and health insurance programs are ignored. Second, all workers are assumed to invest in only one fund, which provides a common return across all investors. Third, all workers convert their accounts to an annuity at the time of retirement—so that no workers take advantage of the programmed withdrawal or early retirement options.

The base case also includes the effect of the assistance pension, and assumes that the program would be available to every retiree whose account balance has been depleted. The assistance pension is meant to be the income support program for the very destitute, and as a result, the benefit levels are traditionally quite low (around one-half of the MPG, or about US$70 per month in 2005). However, these benefits are universal, once the retirees have exhausted all other retirement income. Since contributors with low densities (20–30 percent) can fund a pension for a number of years, the impact of this program would not be large, because a fraction of the retiree population would have died before becoming eligible.

Finally, this model ignores, for the sake of simplicity, several important strategic aspects of the pension program:

  • Options for Converting Pension Assets to Retirement Income: By assuming all pension assets are annuitized, workers cannot take advantage of the two possible alternatives: (i) either cashing out the pension fund (as long there is enough remaining to fund a minimum pension), or (ii) take a series of programmed withdrawals to roughly approximate an annuity. If workers attempt to strategically “game the system,”, these options could result in higher state support.

  • Employment Decisions to Avoid the Pension Contribution: The mandatory social contributions (pension, disability, and health) is approximately 20 percent of wage income, so many workers attempt to avoid paying these contributions by seeking self-employment. This has resulted in a lower-than-anticipated contribution densities, and a significant number of people who contribution less 10 percent of their working life.

  • Disability Coverage. Disability insurance and survivor benefits are included in the AFP system, as a form of a charge that is included in the AFP management fee. However, when the retirement benefits are very low, contributors find a way to qualify for disability coverage. For example, in Australia, disability payments have increased sharply in just twenty years. Here, in contrast, it is assumed that disability payments are assumed to be a constant 0.1 percent of GDP.

Table A. 1.

Assumptions Used to Estimate the Base Case 2005-2100

article image
Sources: Haver Analytics, 2002 Social Protection Survey, and Superintendency for Pension Funds (historical data) and staff estimate

Real GDP growth is based on staff estimates for the WEO until 2010. After then, the growth rate declines in line with the decline growth rate of the working-age population (a proxy for labor force growth). Implicitly, this assumes constant labor productivity grow around 3¼ percent.

Nominal GDP growth is set so that GDP inflation is roughly constant at 2-2¼ percent.

This is set at the mid-point of the BCCh target band.

Real wages grow so that nominal wages (which are equal to real wages + inflation) grows at the same rate as nominal GDP.

Gross average annual returns of the system (historical data for for returns since 1982). During the forecast horizon, the real return is set as a 100bp mark-up on real GDP growth. Normally, this would violate internal consistency; however, these higher returns are justified because international diversification should allow for higher returns.

This is taken as a simple, 50bp mark-down on pension fund returns.

In percent of wages. Includes only pension fund management fees (excludes the disability and other social insurance contributions

The life expectancy data are used to calculate the annuity factors (from table RV-2004). However, the UN Population Projections (the basis for the population forecasts) assume that the life expectancy continues to grow.

Taken from the kernel density estimate of the contribution densities for households (from the First Social Protection Survey ).

Table A. 2.

Long-run Outcomes of the Pension: Various Policy Alternatives

Base case results and deviations from baseline

article image

Excluding disability spending. Present discounted value, using in percent of current GDP, using annuity discount rates.

Replacement rates in percent of final salary.

Percent of retired-age population receiving a MPG subsidy (or “top-up”) to their pension.

This scenario is very similar to the assumptions in Arenas de Mesa and Marcel (1999) and Benet and Schmidt-Hebbel (2001).

References

  • Acuña, Rodrigo, and Iglesias, August, 2001, “Chile’s Pension Reform After 20 Years,” Social Protection Discussion Paper, No. 0129, December (Washington, D.C.: World Bank).

    • Search Google Scholar
    • Export Citation
  • Angel, Ricardo, 2003, “Minimum Pension Guarantees: Fiscal expenditure and the Chilean social security system,” prepared for the Fourteenth International Conference of Social Security Actuaries and Statisticians, September 23-25 (Mexico City: ISSA).

    • Search Google Scholar
    • Export Citation
  • Arellano, I.P, 1985, Politicas Sociales y Desarrollo: Chile 1924–1984 (Santiago: CIEPLAN).

  • Arenas de Mesa, Alberto, 1999, “Proyecciones de Déficit Provisional Chileno,” prepared for 11th Conference on Regional Fiscal Policies (Brasilia), January 2628 (Santiago: CEPAL).

    • Search Google Scholar
    • Export Citation
  • Arenas de Mesa, Alberto, 2004, “El Sistema de Pensiones en Chile: Principales Desafíos Futuros,” in El Sistema de Pensiones en Chile en el Contexto Mundial y de América Latina: Evaluación y Desafíos, prepared for Ponencias del Seminario Internacional, April 2223 (Santiago: ILO).

    • Search Google Scholar
    • Export Citation
  • Arenas de Mesa, Alberto Jere Behrman, and David Bravo, 2004, “Characteristics of and Determinants of the Density of Contributions in Private Social Security System,” University of Michigan Retirement Research Center, Working Paper 2004077 (February).

    • Search Google Scholar
    • Export Citation
  • Arenas de Mesa, Alberto Jere Behrman, David Bravo, and Pamela, Gana Cornejo 2001, “La Reforma de los Sistemas de Pensiones y los Desafíos de la Dimensión de Género,” ECLAC Series on Population and Development, No. 18, (October).

    • Search Google Scholar
    • Export Citation
  • Arenas de Mesa, Alberto Jere Behrman, David Bravo, Pamela, Gana Cornejo and Mario Marcel, 1993, “Proyecciones de Gasto Previsional 1992–2038,” Documento de Trabajo, February, (Santiago: Ministerio de Hacienda).

    • Search Google Scholar
    • Export Citation
  • Arenas de Mesa, Alberto Jere Behrman, David Bravo, Pamela, Gana Cornejo and Mario Marcel, 1999, “Fiscal Effects of Social Security Reform in Chile: The Case of the Minimum Pension,” Mimeo (Santiago: Ministerio de Hacienda).

    • Search Google Scholar
    • Export Citation
  • Benet, Herman, and Klaus Schmidt-Hebbel, 2001, “Déficit Previsional del Sector Público y Garantía de Pensión Mínima,” Notas de Investigación, Vol. 4, No. 3, 8795, December, (Santiago: Central Bank of Chile).

    • Search Google Scholar
    • Export Citation
  • Bertstein, Solange, Guillermo Larraín, and Francisco Pino, 2004, “Cobertura, Densidad y Pensiones en Chile,” prepared for the Joint CEP-SAFP Seminar: Desafíos del Sistema Chileno de Pensiones: Competencia y Cobertura. November, 11-12

    • Search Google Scholar
    • Export Citation
  • Bertstein, Solange, Guillermo Larrain, Francisco Pino, and Andrea Tokman, 2005, “Are Women Relatively Better Off When Old? Gender Income Gaps during Active and Passive Life,” mimeo. (March).

    • Search Google Scholar
    • Export Citation
  • Dang, Thai Than, Pablo Antolin, and Howard Oxley, 2001, “Fiscal Implications of Aging,” OECD Economics Department Working Papers, No. 305, September (Paris: OECD).

    • Search Google Scholar
    • Export Citation
  • Fuentes, Roberto, 2004, “Propuestas para aumentar la Cobertura del Sistema de AFP,” prepared for the Joint CEP-SAFP Seminar: Desafíos del Sistema Chileno de Pensiones: Competencia y Cobertura. November 11-12.

    • Search Google Scholar
    • Export Citation
  • CBO (Congressional Budget Office), 2004, Administrative Costs of Private Accounts in Social Security, Report 5277, March (Washington, D.C.: GPO).

    • Search Google Scholar
    • Export Citation
  • Cheyre, Hernán, 1991, La Previsión en Chile Ayer y Hoy: Impacto de una Reforma (Santiago: Centro de Estudios Públicos).

  • Edwards, Sebastian, 1996, “The Chilean Pension Reform: A Pioneering Program,” NBER Working Paper, No. 5811, (November).

  • ICI (Investment Company Institute), 2003, “The Expense of Defined Benefit Pension Plans and Mutual Funds,” Investment Company Institute Perspective, Vol. 9, No. 6, (December).

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, 2004, “Chile: Financial System Stability Assessment, including Reports on the Observance of Standards and Codes on the following topics: Monetary and Financial Policy Transparency, Banking Supervision, and Securities Regulation,” Country Report No. 04/269, August (Washington, D.C.).

    • Search Google Scholar
    • Export Citation
  • James, Estelle, 2004, “Private Pension Annuities in Chile,” NCPA Policy Report, No. 271, December (Dallas: National Center for Policy Analysis).

    • Search Google Scholar
    • Export Citation
  • Mesa-Lago, Carmelo, 2000, “Estudio Comparativo de los Costos Fiscales en la Transición de Ocho Reformas,” ECLAC Series on Development Finance, No. 93, (March).

    • Search Google Scholar
    • Export Citation
  • OECD, 2003, “Economic Survey—Chile,” (Paris).

  • OECD, 2005, “Pensions at a Glance: Public Policies across OECD Countries,” (Paris).

  • SAFP, 2003, “The Chilean Pension System,” (Santiago).

  • SAFP, 2005, “Informe Estadístico Trimestral de Afiliados y Cotizantes,” March, (Santiago).

  • SAFP, 2005, “Razones para no cotizar,” mimeo, May, (Santiago).

  • Subsecretaría de Previsión Social, Chile, 2002, First Social Protection Survey, (available at www.proteccionsocial.cl).

  • Wagner, Gert, 1983, “Estudio de la Reforma Previsional y Reforma, Efectos en la Industria y en el País,” Institute of Economics, (Santiago: Universidad Católica de Chile).

    • Search Google Scholar
    • Export Citation
  • Wagner, Gert, 1991, “La Seguridad Social y el Programa de Pensión Mínima Garantizada,” Estudios de Economía, Vol. 18, No. 1, 3591, (Santiago: Universidad Católica de Chile).

    • Search Google Scholar
    • Export Citation
  • World Bank, 1994, “Averting the Old Age Crisis: Policies to Protect the Old and Promote Growth, (Washington, D.C.: Oxford University Press).

    • Search Google Scholar
    • Export Citation
  • Zurita, Salvador, 1994, “Minimum Pension Insurance in the Chilean Pension System,” Revista de Análisis Económico, Vol 9, No. 1, 105126, (Santiago: ILADES).

    • Search Google Scholar
    • Export Citation
1

This report has received valuable research assistance from Maria Fernanda Pazmiño.

2

Prepared by Chris Faulkner-MacDonagh (WHD). I would like to thank Alberto Arenas de Mesa, Solange Bernstein, Pamela Gana, and Gonzalo Reyes—along with all of the participants in the 2005 Joint BCCh/IMF Seminar Series—for their very helpful comments. This study has used information from the 2002 First Social Protection Survey, and the author is indebted to the Under Secretary of Social Security for access to this data. All results of this study are the sole responsibility of the author.

3

The deficit of the civilian system was around 1¾ percent of GDP, and the system for the military the police ran a deficit of around 1-1¼ percent of GDP.

4

Total payroll taxes (pension, health, and other social insurance) amounted to 35 percent of wages in 1980 (Cheyre, 1991).

5

Until 2002, investors only had two investment options that were nearly identical across AFPs. Since 2002, the AFPs have offered investment options ranging from a fund with 80 percent equity weight (type A) to a fund with no equities (type E). Investors could then choose among a mix (subject to age restrictions) of funds.

6

Members of the armed forces and the police were able to keep their defined-benefit system.

7

The MPG income has averaged around 75 percent of the real minimum wage, although the government has flexibility in setting the precise level (any changes are retroactive).

8

See Arenas de Mesa (1999, 2004); Participation is limited (to around 410,000 persons in 2004) and benefits are small, equivalent to about 50 percent of the minimum pension guarantee. The program’ss fiscal cost has been constant at around ½ percent of GDP over the past twenty years.

9

These rates are based on the assumption that all workers enter the labor force at age 20, with men retiring at age 65 and women at age 60. Thus, men would have a working life of 540 months and women of 480 months. The minimum contribution density to qualify would be 0.44 for men (240/540) and 0.5 for women (240/480).

10

These findings are from a 2002 household survey—First Social Protection Survey—undertaken by the Chilean government with about 17,000 persons (retired, working, unemployed, and out of the labor force).

11

See SAFP (2005). For women, family duties account for nearly half the time in inactivity compared with only ½ percent of the time for men. Studying accounts for another 15 percent for women and 39 percent for men.

12

According to SAFP (2003, Chapter 1), the total social insurance bill includes: 10 percent for pension contributions; 2½-3 percent for AFP management fees and insurance costs; 7 percent for the health system; and 3 percent for unemployment insurance (about 2½ percent is paid by the employer).

13

This matching process was done in a way to maintain the confidentiality of the information of individual account holders.

14

Net return is a somewhat artificial measure; it reflects the difference between the five-year moving average of the return on the fund and the total fees paid to the AFP (in percent of assets).

15

SAFP (2002), Chapter 5.

16

In 1998, the cost for full-service U.S. mutual funds averaged 0.71 percent of assets (ICI, 2003).

17

Canada, Denmark, United Kingdom, and the United States have private supplemental pension systems that boost replacement rates by 20–40 percent. For example, the U.S. Social Security system provides only a 39 percent replacement rate for the average retiree (but as high as 50 percent for low income workers); however, it exceeds 70 percent when income from private plans is included.

18

Assumes the rate of real return is 5 percent from 2005–2030, similar to other forecasters (Angel, 2003, Table 6). A 5 percent real rate of return is consistent with a ½ percentage point markup over the medium term trend growth rate for the economy (4½ percent), owing to portfolio diversification.

19

The higher replacement rate is from Acuña Iglesias (2001); the lower estimate is from a model of individual contributors, using household data.

20

These estimates exclude spending on two other programs that could increase more sharply as the population ages, but are not strictly “old-age” spending. First, they exclude the disability programs, because this program is self-financed and the qualification rules are complex. Second, these estimates exclude the costs for the public health program. As the population ages, demand for medical services increases rapidly, but unpredictably.

21

Based on the assumption that the military’s deficit is assumed to stay constant as a share of real per capita GDP, while the assistance pension is available to all retirees without a pension income.

22

Early estimates (from Wagner, 1991 and Zurita, 1994) place the cost at between ½–3 percent of GDP. More recently, OECD (2003) notes that the MPG costs could be twice as high as forecast.

23

These replacement rates are calculated for individuals and assume that men choose not to use the survivor options available to them, which would lower their retirement income. Thus, these results should be seen as an “upper bound” on the replacement rates.

24

One such possibility would limit the time spent in retirement to be no more than one-fifth (say) of the average life expectancy. For Chile, this rule would result in an increase in the retirement age to 71 for women and to 66 for men. For equity considerations, it may be better to harmonize the retirement age, which would suggest that retirees could draw on the public assistance program until around age 69 (for both men and women).

25

Men can purchase an annuity with a survivor’ss pension, which lowers their replacement rate significantly. However, this option is not considered here; instead, the results are considered for just men and women, individually, without considering the impact of retirement on their spouses.

Chile: Selected Issues
Author: International Monetary Fund
  • View in gallery

    Trends in National Saving

    Gross national saving, in percent of GDP

  • View in gallery

    Pension System Returns

    Real returns by AFPs, 5-year rolling average, both gross and net of administrative costs and insurance premia.

  • View in gallery

    Spending on Pensions in Latin America, 2000-2020

    Annual fiscal deficits of the pension systems

  • View in gallery

    OECD and Chile Old-age Spending

    Gross current expenditures in 2000.

  • View in gallery

    Long-run Returns of the System

    Average annual gross real returns, 40-year moving avg.

  • View in gallery

    Demographic Trends, 1980-2100

    Ratio of workers to retired-aged, and share of retired-age population in the total population.