CHILE TECHNICAL NOTE: PENSION FUND SECTOR FINANCIAL SECTOR ASSESSMENT PROGRAM NOVEMBER 2004

Transcription

1 Public Disclosure Authorized This volume is a product of the staff of the International Bank for Reconstruction and Development / The World Bank. The World Bank does not guarantee the accuracy of the data included in this work. The findings, interpretations, and conclusions expressed in this paper do not necessarily reflect the views of the Executive Directors of the World Bank or the governments they represent. The material in this publication is copyrighted. Public Disclosure Authorized Public Disclosure Authorized FINANCIAL SECTOR ASSESSMENT PROGRAM CHILE TECHNICAL NOTE: PENSION FUND SECTOR NOVEMBER 2004 Public Disclosure Authorized THE WORLD BANK FINANCIAL SECTOR VICE PRESIDENCY LATIN AMERICA & THE CARIBBEAN REGIONAL VICE PRESIDENCY INTERNATIONAL MONETARY FUND MONETARY AND FINANCIAL SYSTEMS DEPARTMENT

2 Contents Page I. Introduction... 1 II. The Structure And Performance Of The Pension Sector... 2 A. The Growth of Pension Fund Assets...2 B. The Structure of the Pension Sector...5 C. The Portfolio Composition of Pension Funds: Historical Evolution...6 D. The Portfolio Composition of Pension Funds: International Comparison...9 E. Gross Rates of Return...14 F. Fees, Costs, and Profits...18 G. Net Rates of Return...24 H. Payouts and Replacement Ratios...25 I. Pension Funds and Capital Market Development...26 III. The Regulatory Framework A. Licensing Criteria...28 B. Governance Structures...28 C. Accounting and Auditing Rules and Practices...29 D. Custodian Rules and Arrangements...29 E. Disclosure...30 F. Investment Regulations...30 G. Outsourcing Regulations...40 H. The Voluntary Pension System...40 IV. The Supervisory Framework A. The Approach to Pension Supervision in Chile...41 B. Legal Status and Organization of the SAFP...42 C. Regulatory and Enforcement Powers of the SAFP...40 D. The Internal Structure of the SAFP and Main Supervisory Functions...44 E. Relationship with Other Supervisors...45 V. Recommendations A. The General Direction of the Reforms...45 B. Reforming the Investment Regime...46 C. Moving from Compliance to Risk-Based Supervision...49 D. Enhancing Competition in the Pension System...50

3 ii Tables Table 1: Pension Fund Assets (% of GDP) in Chile, Latin America and the OECD... 2 Table 2: Flow of Funds in the Private Pension Fund System, (% of GDP)... 3 Table 3: Coverage of the Labor Force (%) and Covered Wage Bill (% of GDP)... 4 Table 4: Portfolio Composition of Chilean Pension Funds (%), Table 5: Average Age, Income, Balance, and Size of Different Funds, December Table 6: Portfolio Composition of Pension Funds, by Type of Portfolio, Dec, Table 7: Portfolio Composition of Pension Funds (%): Chile and Selected Countries Table 8: Herfindahl-Hirshman Index of Exports Concentration, Chile and Other Countries Table 9: Financial Assets (% of GDP) and Pension Fund Participation (%), Table 10: Real Rates of Returns of Individual AFPs (% p.a.), Table 11: Fees, Operating Costs and Operating Profits Over Assets, Table 12: Costs and Fees of US Mutual and Pension Funds and Chilean Pension Funds Table 13: Average Gross and Net Rates of Return in Different Periods, Table 14: Simulation of Net Returns and Fees/Assets Resulting from 17% Fee on Contributions (%). 24 Table 15: Estimated Replacement Ratios Implied by Different Types of Annuities, Table 16: Investment Regulations in Chile Table 17: Investment Regulations in the OECD Table 18: Value at Risk of Different Types of Funds...40 Figures Figure 1: Covered Wage Bill (% GDP) and Per Capital Income (US$PPP)... 5 Figure 2: Number of AFPs (left) and Share of 3 Largest AFPs in Total Assets (right), Figure 3: AFP Holdings of Equity and Turnover...12 Figure 4: Real Returns, Real Wage Growth, Wage Bill Growth 10-year Rolling Average (%) Figure 5: Accumulated Real Returns, Real Wage Growth, Real Wage Bill Growth for Different Cohorts, Figure 6: Average Real Returns and Share of Equity in Pension Funds Portfolios, Figure 7: Real Return and Risk of Each AFP and Type of Funds, Sep. 02-Feb Figure 8: Fees/Contributions, , Total and Net of Insurance Premia (%) Figure 9: Fees and Operating Costs/Assets, , Net of Insurance Premia (%) Figure 10: Number of Sales Agents (left) and Transfers Across AFP (right), Figure 11: Operating Profits/Assets (left) and Return on Equity (right), Figure 12: Operational Expenses over Members and Assets (%), Figure 13: Pension Sector Equity Holdings, as % of Issuer s Equity and % of Free Floats, December, References... 55

4 1 I. INTRODUCTION 1. Chile has one of the most developed pension fund sectors among emerging countries and compares favorably with most OECD countries as well, as indicated by pension assets of 60 percent of GDP in The large size of Chile s pension sector is the result of an innovative reform initiated in 1981, that involved the switch from a public pay-as-you-go (PAYG) system to a fully-funded (FF) system operated by the private sector. The Chilean reform inspired a number of similar reforms in Latin America and other regions, and led to an intense international debate on pension reform. Proponents of the Chilean pension system have emphasized the potential strengths of a private FF system, including its insulation from political pressures, the higher returns on worker contributions, and the positive effects on labor markets, capital markets, and economic growth. Critics of the Chilean system have emphasized its potential weaknesses, including the absence of intergenerational solidarity, the high fees paid by participants, and the high exposure of workers to capital market volatility These strengths and weaknesses are observed to varying degrees in Chile and the countries that adopted the Chilean model. In the particular case of Chile, the strong aspects have outweighed the weak ones, resulting overall in a good performance of the pension sector over the past 20 years. Participants have enjoyed an average real rate of return net of fees of about 7.2 percent p.a. since 1981, well above the 4 percent average yearly growth of the wage bill in the same period (the implicit rate of return in a balanced PAYG system). The pension fund sector has proved remarkably resilient, despite episodes of international crisis affecting Chile and other countries, and significant changes in the number of licensed pension funds. The sector has also made a positive contribution to the development of Chile s capital markets, as indicated by the relatively large size of equity and bond markets, and the long maturities on debt instruments. Most importantly, the private pension system is generally regarded as reliable and credible. 3. Despite the positive performance overall, the review has identified a number of specific gaps in regulation that may need to be corrected, and two major issues that need to be addressed, in order to improve the performance of the system. The two major issues are the excessive complexity of the investment regime, and the high level of concentration of the industry. The review concludes that there is space for a selective relaxation of the limits, which would widen the room for asset managers to operate without causing a meaningful increase in the levels of risk. The concentration problem can probably be ameliorated by deepening and refining some of the programs and rules that are already being implemented, such as the tax-supported voluntary pension savings program, and the possibility for pension funds to reduce their costs by outsourcing basic functions to external service providers. Fund members could also be allowed to transfer their mandatory contributions to the voluntary system under certain conditions. If this gradual strategy does not produce substantive results, the authorities could envisage a more ambitious strategy involving the establishment of one 1 The literature is extensive. See, e.g., Barr (2000), Charlton and McKinnon (2001), Gillion (2000), Gillion et al (2000), Holzman (2000), Holzman and Stiglitz (2001), ILO (2002), Kotlikoff (1999), and The World Bank (1994).

5 2 single basic service provider, and the licensing of a larger number of asset managers operating under a blind quotation system. II. THE STRUCTURE AND PERFORMANCE OF THE PENSION SECTOR A. The Growth of Pension Fund Assets 4. Pension fund assets have grown rapidly in the past 2 decades, reaching 60 percent of GDP in The share of pension assets in GDP is much higher than the Latin American average and even the OECD average, as shown in Table 1. There are just a few OECD countries with comparable or larger pension systems. These include Australia, Canada, Denmark, Ireland, the Netherlands, Switzerland, the UK, and the US. Other financial institutions in Chile such as insurance companies, banks and mutual funds can also offer pension products, after the introduction of a new system of voluntary pension savings in These additional pension assets managed outside the pension sector have grown rapidly but still account for less than 0.5 percent of GDP. Table 1: Pension Fund Assets (% of GDP) in Chile, Latin America and the OECD Chile Latin America High Income OECD Sources: SAFP, AIOS, OECD 5. The rapid growth of pension fund assets is the result of a well-known 1981 pension reform, which involved the switch from a PAYG system to a FF system, and the imposition of a 10 percent contribution on wages to individual pension accounts, up to 60 UF (Unidad de Fomento), or the equivalent of 3 times the average covered wage. Workers pay an additional 2.4 percent of their wages in commissions. The new system was made mandatory for new entrants to the labor force and voluntary for all the existing workers. Most workers decided to join the new system, receiving a recognition bond for the amount of contributions paid under the former PAYG system. This recognition bond is redeemed by the Government at the time of retirement The rapid transfer of a substantial share of the labor force to the new pension system implied a large inflow of resources to pension funds. As shown in Table 2, by the late 1980s annual mandatory contributions (net of commissions) already amounted to 1.9 percent of GDP per year, or 2.3 percent of GDP per year including the redemption of recognition bonds. Net mandatory contributions grew continuously during the 1990s, reflecting the growing share of the labor force enrolled in the new system. By the early 2000s the flow of mandatory contributions stabilized at 2.8 percent of GDP, reflecting the end of the transition from the old to the new pension system at that time more than 95 percent of the 3.5 million active contributors were enrolled in the new system. 2 A detailed description of the 1981 reform is provided in Yazigi (2002) and Acuña and Iglesias (2001).

6 3 7. Asset returns constitute another important source of asset growth and have actually been more important drivers of growth than mandatory contributions, as shown in Table 2. The contribution of asset returns has tended to decline in recent years, reflecting the end of a period of exceptionally high rates of return (see below), but still constitute the largest source of asset growth, and should remain the driving source of growth in the future as well, given the large volume of assets. Table 2: Flow of Funds in the Private Pension Fund System, (% of GDP) Net Inflows ( ) Net Mandatory Contributions + Rec. Bonds Net Mandatory Contributions Recognition Bonds Asset Returns Mandatory Payouts Direct Benefit Payments by Pension Funds Transfers to Insurance Companies Voluntary Contributions Payouts Source: SAFP 8. The outflows from the pension fund sector are associated with retirement and the choice among the three basic retirement products: programmed withdrawals, temporary withdrawals followed by a deferred annuity, and an immediate annuity. The first two types of benefits are provided by the pension funds themselves, while the third involves a transfer of the final balance to insurance companies, for the purchase of life annuities. As shown in Table 2, mandatory payouts have increased rapidly from 0.5 percent of GDP in the late 1980s to almost 2 percent of GDP in recent years. This rapid growth of payouts reflects the maturation of the pension system, driven in turn by demographic factors and the relatively low average age of retirement in Chile a large number of workers takes the early retirement option at the average age of 55. Many of these early retirees remain in the labor force but receive their pensions and stop contributing to the mandatory system. 9. Finally, net voluntary contributions to the pension fund sector have remained relatively modest, at less than 0.1 percent of GDP. Total voluntary pension assets managed by pension funds account for about 1 percent of GDP, or only 1/60 th of total pension fund assets. There is an expectation that new rules on voluntary pension savings will accelerate the growth of pension assets but voluntary pension savings will probably remain substantially smaller than the mandatory flows, and in any case will be shared with other financial institutions, as noted above. 10. Based on the trends observed in the last decade and reasonable assumptions about key variables such as the rate of return, it is projected that pension fund assets will grow to about 80 percent of GDP in 10 years, and will stabilize at about 85 percent of GDP in the 2020s. The stability of pension fund assets at these levels reflects the full maturation of the pension system, projected to occur in the 2020s the period when payouts are expected to approach

7 4 the levels of inflows. 3 These projected levels imply a large pension sector, but are smaller than the levels foreseen by other researchers It is possible that recent amendments to the pension law restricting early retirement will lead to larger net inflows and faster asset growth in the future. The efforts to promote voluntary pension savings in pension funds and other institutions may also lead to a faster growth of the industry. At the same time, asset growth will remain constrained by the relatively low contribution base in Chile. As shown in Table 3, Chile has the highest covered wage bill in Latin America about percent of GDP, compared to an 18 percent of GDP average for the region as a whole. However, Chile does not compare favorably with Central and Eastern European countries (especially the more advanced ones) and with OECD countries. The relatively low contribution base is in turn due to a relatively low coverage ratio (contributors/labor force) and to a low share of labor income in GDP relative to these countries. 12. The contribution base may be increased to some extent by Government policies, such as increases in the contribution ceiling, the introduction of special programs for the selfemployed, expansion of the base of taxable income, and efforts at contribution collection. However, international experience indicates that increases in the contribution base are to a good extent a developmental problem. As shown in Figure 1, there is a close relationship between the covered wage bill and per capita income, and Chile is already above the international regression line, although by a small margin. This suggests that the contribution base may increase in the long-run, but at very gradual rates. The basic factors that determine the size of the base the coverage ratio and the labor share evolve very gradually and in line with the general level of economic development. This suggests that, if a faster increase in pension assets becomes a policy objective (designed to support larger pension benefits for future retirees), an increase in the contribution rate might need to be considered. 5 Table 3: Coverage of the Labor Force (%) and Covered Wage Bill (% of GDP) Chile and Selected Country Groups, Mid-1990s Chile LAC All CEE Advanced CEE OECD Covered Wage Bill (% of GDP) Coverage of the Labor Force (%) Labor Income/GDP Sources: Ministry of Planning (1999), SAFP, UN, World Bank 3 In the absence of an actuarial model, it was assumed that net mandatory contributions will remain stable at 2.8 percent of GDP (reflecting a stable contribution base of 28 percent of GDP and a contribution rate of 10 percent), the flow of recognition bonds will gradually decline to zero in 20 years, mandatory payouts will increase by 0.1 percent of GDP per year (the rate of increase of the past 12 years), and the rate of return will exceed GDP and wage growth by 1 percent p.a.. 4 See, e.g., Cifuentes, Desormeaux and Gonzales (2002). 5 By international comparison, a contribution ceiling of 3 times the average wage is high, but a contribution rate of 10 percent is low.

8 Covered Wage Bill (% of GDP) Figure 1 Covered Wage Bill (% GDP) and Per Capita Income (US$ PPP) Chile, LAC, CEE, OECD, Mid-1990s Chile 0 5,000 10,000 15,000 20,000 25,000 30,000 Per Capita Income (US$ PPP) B. The Structure of the Pension Sector 13. The large volume of pension assets in Chile is managed by only 7 pension management companies (Administradoras de Fondos de Pensiones AFPs). Each of these companies is constituted as a joint stock company with their assets segregated from the assets of the pension funds. The reduction in the number of AFPs to only 7 institutions is the result of an intensive consolidation of the industry during the 1990s, involving several mergers and acquisitions. As shown in Figure 2, the reduction in the number of institutions has been accompanied by an increase in the market share of the three largest institutions to more than 70 percent of total assets. At the time of writing the number of AFPs was being reduced further to 6, with the merger of the 2 smallest institutions, each holding a market share of about 3 percent. 14. Until 2000, each AFP was allowed to offer only one pension fund. In 2000 AFPs were allowed to offer 2 funds each the regular fund, and a second fund investing only in fixed income assets. In 2002 AFPs were allowed to offer 5 funds each, with funds being primarily differentiated according to the share of variable income that they can hold. Participants are allowed to switch across the family of funds managed by each AFP, as well as to switch across AFPs. Although some restrictions apply for switching across funds and across AFPs, the logic of the Chilean system is that workers should be able to compare costs and returns across funds and institutions and vote with their feet, thus exerting competitive pressure on AFPs. (In contrast to the occupational schemes of most OECD countries, which have boards appointed by members, through which workers vote with their voice ). Although the Chilean system is very transparent in many aspects, it never performed fully as originally intended, due to the high degree of industry concentration and the intensive degree of non-price competition among AFPs.

9 6 Figure Number of AFPs (left) and Share of 3 Largest AFPs in Total Assets (right), Number of AFPs Share of Three Largest AFPs 80% 70% 60% 50% 40% 30% 20% 10% 0% C. The Portfolio Composition of Pension Funds: Historical Evolution 15. The portfolio composition of pension funds changed significantly over the past 20 years. As shown in Table 4, in the early 1980s AFPs invested almost exclusively on bonds issued by the public sector and by other financial institutions, primarily banks. By 1990, portfolios were already more diversified, with smaller claims on the financial sector and larger investments in domestic equity and corporate bonds. This trend continued in the early 1990s, and by 1994 domestic equity holdings had reached a peak. Since the mid-1990s, the main portfolio shifts have included a reduction in public sector bonds and domestic equity and an increase in foreign assets. In recent years, AFPs have again increased their holdings of corporate paper, initially through larger holdings of bonds, but in 2003 through larger holdings of equity as well. However, holdings of corporate paper remain well below their peak levels of the mid-1990s. 16. These changes in portfolio composition have been to a good extent due to the evolution of the investment regime and exceptional events during the past 2 decades. In the early 1980s the investment regime was severely restrictive, allowing only investments in domestic fixed income instruments. Relaxation of the investment regime took place in various steps, starting in 1985, when AFPs were allowed to invest up to 30 percent of their portfolios in a selected number of shares, in order to be able to participate in the privatization of utilities. The privatization took place in the second half of the 1980s and was labeled as capitalismo popular the shares were primarily bought by pension funds and households at modest prices. Subsequent privatizations took place in the early 1990s with the participation of pension funds as well. The increase in the share of equity in the early 1990s reflects both increased volumes and price gains. By the mid-1990s pension funds realized large capital gains through sales of a large portion of their equity holdings to foreign strategic investors, who took control of many of these enterprises.

10 7 Table 4: Portfolio Composition of Chilean Pension Funds (%), Claims on the Public Sector Government Bonds Central Bank Bonds Other Claims on the Financial Sector Mortgage Bonds Time Deposits/CDs Other Claims on the Corporate Sector Shares Bonds Other Claims on the Foreign Sector Quotas of Mututal Funds Other Total Memo items: Total Variable Income Total Assets/GDP Source: SAFP 17. The AFPs started investing in foreign assets after other rounds of liberalization in 1990 and, especially, Most holdings of foreign assets take place through holdings of quotas of mutual funds abroad. These holdings are classified as variable income in the current asset classification system, regardless of whether the mutual fund is an equity or a bond fund. This results in a relatively high share of variable income in the portfolio 38 percent in The true share of variable income is probably closer to 30 percent, but it is clear that the increasing holdings of foreign assets involve to a good extent investments in foreign equity In recent years pension funds have again increased their holdings of domestic corporate paper. The share of corporate bonds in the portfolio almost doubled after 2000, reflecting the decision of major Chilean companies to take advantage of falling domestic interest rates and long maturities, and the more difficult conditions to raise funds abroad. On their side, pension funds were interested in good quality corporate bonds offering returns 200 basis points above Government and Central Bank bonds, and after a period of depressed equity returns. In 2003, holdings of domestic equity also increased, reflecting the rebound in the domestic equity market. 19. The recent increase in the share of equity also reflects to some extent the most recent round of reforms in 2002, involving the introduction of a multiple portfolio regime (multi fondos). In August of that year AFPs were allowed to offer up to 5 funds or portfolios, labeled as Funds A to E, and structured according to the share of variable income that they can hold. Thus, the investment regime includes floors and ceilings for variable income that are high for Fund A (40 and 80 percent respectively), and that decline progressively for the other funds, reaching zero in the case of Fund E. Members are allowed to place their 6 The same happened in the 1990s in the OECD see Annex Tables 1 and 2.

11 8 balances in up to 2 of the 5 funds offered by an AFP, except for those approaching retirement (55 and 51 years of age for men and women, respectively), who cannot select Fund A, and retired members, who cannot select Funds A or B. Members who do not exercise their right to choose are assigned to Funds B, C, or D according to 3 age brackets. The central age bracket is the widest (36 to 55 years of age), implying the assignment of a large number of non-choosing members to Fund C, the middle Fund. This construction was inspired by the lifestyle investment concept whereby young members should invest primarily in equity and older members should invest primarily in bonds, to avoid excess volatility in the period close to retirement As shown in Table 5, in December 2003 roughly 85 percent of total members were enrolled in Funds B and D. This outcome was due both to the default rules and to active choices out of the 1.5 million members that had exercised their right to choose a fund, roughly 50 percent had opted for Fund C and 25 percent for Fund B. The age distribution of members across the 5 different funds is generally consistent with the objective of the multi fund regime, with young members primarily enrolled in Funds A and B, middle age members in Fund C, and older members in Funds D and E. The average wage of Fund A members is much higher than that in Fund B, revealing a greater tolerance for risk among better educated, high income young workers. Interestingly, the average age of members of Fund D members is much higher than that of Fund E, although the larger average balance in Fund E suggests that there are several members close to retirement in that Fund as well. Table 5: Average Age, Income, Balance, and Size of Different Funds, December 2003 Fund A Fund B Fund C Fund D Fund E Average Age (years) Average Wage (1,000 Pesos) Average Balance (1,000 Pesos) 6,447 1,776 5,033 6,638 12,008 Number of Members (1,000) 224 2,983 3, Number of Active Contributors (1,000) 151 1,214 1, Source: SAFP 21. The portfolio allocation of different Funds is also generally consistent with the objectives of the multiple portfolio regime. Fund A, designed to offer the highest risk-return combination, effectively holds the largest amounts of domestic equity and foreign assets, while Fund E, designed to be the safe fund, holds the largest amounts of domestic fixed income assets, especially long-term bonds issued by the public sector, but also corporate and mortgage bonds. This outcome is to some extent the result of investment regulation itself, which directs investments towards variable and fixed income according to the type of fund. However, AFPs still have some margin to maneuver, and have used the greater room for investing in variable income in Funds A and B, particularly the first. As shown in Table 6, in December 2003 the share of variable income in Fund A was close to the ceiling of 80 percent of the portfolio, although this may be over-stated by the asset classification problem. 7 The are valid arguments for investing more in equity in early ages, but there has been a heated debate regarding the optimal share of equity over the working life. See, e.g., Samuelson (1994) and Jorion (2002) for a recent review.

12 9 22. Fund C was designed as the central fund, and is effectively the successor of the former system. As mentioned before, most members who did not exercise their option to choose were placed in Fund C, and roughly half of the members who made a choice opted for remaining in Fund C as well. As a result, in December 2003 Fund C still accounted for 55 percent of total assets managed by the AFPs. The portfolio allocation in Fund C has remained reasonably conservative, with a small share of domestic equity and a large share of short-term and long-term fixed income instruments. Table 6: Portfolio Composition of Pension Funds, by Type of Portfolio, Dec, 2003 A B C D E Total Claims on the Public Sector Central Bank of Chile Government Recognition Bonds Claims on the Financial Sector Mortgage Bonds Time Deposits Bonds of Financial Institutions Shares of Financial Institutions Forwards Claims on the Corporate Sector Shares Bonds Units of Investment Funds Commercial Paper Claims on the Foreign Sector Units of Mutual Funds and Shares Indirect Investments Abroad (*) Debt Instruments Other Other Assets Total Assets Total Assets (US$ million) 2,657 9,313 27,708 8,045 1,967 49,690 Memo Item: Variable Income D. The Portfolio Composition of Pension Funds: International Comparison 23. The average portfolio composition of pension funds in Chile looks conservative by comparison with pension funds in the OECD, reflecting to some extent the conservative allocation in Fund C and its dominant size. As shown in Table 7, Chilean pension funds hold a much larger share of cash and bank deposits (primarily bank CDs) than pension funds in other countries, and also hold more bonds and less equity than their OECD counterparts. The differences are probably larger than indicated, considering that the share of variable income in Chile is over-estimated by classification problems, and the share of equity in some OECD countries is under-estimated (the large share of other assets in some countries such as the UK reflects holdings of foreign equity not properly classified). Chilean pension funds also invest

13 10 much less in private equity less than 1 percent of the total portfolio in 2003, compared to 4 percent in the case of European pension funds A more conservative portfolio allocation for Chilean funds is probably justified, considering that the private pension sector is the core of the country s social security. In the other countries in the sample, the private pension system complements a public PAYG system that replaces percent of contributors incomes at retirement 9. This means that Chilean workers are more exposed to capital market volatility and replacement rate risk than workers in other countries. Dealing with this risk requires a smaller share of equity and a larger share of bonds relative to other countries. A higher share of liquid assets is probably also justified, considering that the Chilean pension system is defined-contribution and open, where participants can switch across funds, and where pension fund portfolios are marked-tomarket on a daily basis. For all these reasons, it is not surprising to see that Chilean funds hold more liquid assets, less equity, and more long-term bonds, relative to pension funds in other countries At the same time, the portfolio composition of Chilean pension funds can still be considered atypical in some aspects. Holdings of low-yield, short-term certificates of deposit look excessively large, even considering the larger liquidity needs of Chilean funds, and imply lower returns for members in the long-run. The share of equity looks somewhat low, particularly domestic equity, even considering the justification for a less risky portfolio in the case of Chile. The negligible amount of private equity after 20 years is somewhat disappointing, as this is an investment class that can produce attractive returns when properly managed, and that can also make an important contribution to corporate development and growth. Finally, the share of foreign assets also seems low, considering that Chile has a smaller and less diversified economy than the other countries. As shown in Table 8, Chile s exports became more diversified in the 1990s, but still remain substantially less diversified than most other countries with comparable pension sectors. 26. There are other aspects of the portfolio composition that also beg explanation. AFPs are allowed to invest in 91 listed shares approved by the regulatory authorities (the Credit Risk Committee), but only invest in 70, and the value of these holdings is well below the ceiling for the asset class. AFPs can also invest in about 140 other listed shares not approved by the Credit Risk Committee, up to a small percentage (3 percent of Funds A and B, and 1 percent for Funds C and D), but only invest in 20 of these enterprises and the value of these holdings is extremely small (0.2 of the portfolio). Investments in corporate bonds also seem restricted AFPs hold bonds of about 74 Chilean corporations, but 70 percent of these bonds is rated AA, 94 percent is rated A and above, and these are all issued by the largest corporations. Holdings of corporate bonds rated BBB and BBB+ are very small, even though the potential universe of well-performing large and medium issuers would seem to be significant. 8 Investment and Pensions Europe (2004). 9 Whitehouse (2001). 10 A large share of long-term bonds, especially bonds indexed to inflation, is also consistent with recent views of optimal portfolio allocation for long-term investors (Campbell and Viceira (2002)).

14 The introduction of the multi-funds has not changed in any fundamental way the AFPs portfolio strategies. Although Funds A and B have a larger share of domestic equity than Funds C and D, the universe of enterprises in these four portfolios is essentially the same. Investment in private equity is also very similar across different types of funds. In other words, all AFPs are meeting the demand from their members to offer a higher riskreturn combination in Fund A, but this is being achieved by simply holding more of the same group of domestic equities that they already held in Fund C before the reform. In the case of corporate bonds the picture is very similar there is little variation in bond ratings across funds. If there was an expectation that asset managers would be more innovative in Fund A and possibly Fund B, considering additional listed shares, or investing more in private equity, or going slightly down the credit curve, this expectation has not materialized. Therefore, if the multi-fund regime contributed to the larger equity holdings in 2003, that was essentially due to the larger share of members in Funds A and B, relative to Funds D and E, and not to different portfolio strategies. 28. The current portfolio allocation may be explained to a good extent by the relatively small size of the Chilean economy, and its relatively small, illiquid and concentrated capital market. The rapid growth of pension fund assets due to mandatory contributions and large asset returns have led to a situation of pension funds in search of additional assets that is not uncommon in other small countries. As shown in Table 9, pension funds already have a substantial participation in the market for many instruments, such as mortgage, corporate, and public sector bonds they hold roughly half of the stock of mortgage and public sector bonds and one third of the stock of corporate bonds. 29. The only major domestic market where the pension sector seems to be underrepresented is the equity market pension funds hold only 10 percent of the market value of the 90 equities in their portfolios. However, the low investments in domestic equity are also caused, at least to some extent, to capital market limitations. The 90 equities held in the portfolios account for more than 80 percent of total market capitalization and several companies have a very concentrated ownership structure discouraging participation by minority investors. The equity market is generally illiquid, particularly in the case of medium and small caps, reducing the attractiveness of the asset class. As shown in Figure 3, there is a clear relationship between the size of equity holdings and equity turnover. Although AFPs are not active traders, shares which are illiquid still prove problematic, not only because they make exit difficult, but also because they create valuation problems and may affect adversely rates of return. Moreover, the inclusion of additional small shares involve larger research and monitoring costs without improving significantly the risk and return characteristics of the portfolio. When pension funds invest in listed small caps these investments take place primarily through participations in closed-end investment funds, and are also small (around 1 percent of the portfolio).

15 Small Holdings (1,000 US$) Large Holdings (1,000 US$) Holding (1,000 US$) 12 Figure 3 AFP Holdings of Equity and Turnover 1, 000, , , , , Companies ordered by size Holdings by AFP Turnover AFP Large Holdings of Equity and Turnover 400, , , , , , , , , ,000 Turnover (thousand of US$) AFP Small Holdings of Equity and Turnover 12,000 10,000 8,000 6,000 4,000 2, ,000 10,000 15,000 20,000 25,000 30,000 35,000 Turnover (thousand of US$)

16 Although the limitations imposed by a small capital market are probably the most important cause of the current portfolio allocation, it is likely that the investment regime has also constrained the portfolio strategies adopted by the AFPs. A mechanical comparison of investment limits with actual holdings would suggest that most of the numerous quantitative restrictions are not binding, except for the ceiling of foreign assets, which was being raised from 25 percent to 30 percent at the time of writing. However, a closer examination of the regulatory and supervisory regime (provided in section III) indicates that it may more constraining than perceived. For example, AFPs adopt a very conservative policy in corporate bond investments, basically avoiding bonds rated as BBB and BBB+, because they do not want to sell bonds that fall below investment grade in an illiquid market. If they were granted some additional flexibility to manage below investment grade bonds, they would probably invest more in companies issuing bonds at the investment grade level, and more Chilean companies would be motivated to issue bonds as well. Also, small AFPs sometimes face constraints when investing in variable income and BBB bonds because the minimum size of bond issues would result in a violation of some joint limits on variable and fixed income. Table 7: Portfolio Composition of Pension Funds (%): Chile and Selected Countries Cash and Bills and Assets/ Country Deposits Bonds Loans Shares Other Total Foreign GDP Chile (2003) Australia (2002) Canada (2001) Denmark (2001) Ireland (2001) Netherlands (2001) Spain (2001) Switzerland (2000) United Kingdom (2001) United States (1998) Sources: SAFP, OECD, APRA, Irish Pension Fund Association, Danish Supervisor, Davis (1995 and 2001). Table 8: Herfindahl-Hirshman Index of Exports Concentration, Chile and Other Countries Year Chile Arg Aus Bra Can Dan Ire Ned Spain Swit UK US Source: UNCTAD

17 14 Table 9: Financial Assets (% of GDP) and Pension Fund Participation (%), Year Stock Market Capitalization (% of GDP) AFP (%) Corporate Bonds (% of GDP) AFP (%) Mortgage Bonds (% of GDP) AFP (%) Public Sector Bonds (% of GDP) AFP (%) Time Deposits (% of GDP) Sources: SAFP, BCC 31. More generally, the overall regulatory and supervisory framework, involving a complex set of restrictions and an aggressive supervision, may have led asset managers to adopt overly defensive strategies and avoid investments within the acceptable risk-return range, particularly in Fund A. This suggests that the optimal policy response will probably involve well coordinated actions in the two fronts a careful relaxation of some of the investment restrictions, combined with improvements in securities markets designed to address the limitations faced by pension funds, such as market liquidity (this issue is examined in detail in the capital market chapter of the FSAP). E. Gross Rates of Return 32. Chilean pension funds have posted reasonable returns on their assets since the creation of the new pension system in As shown in Figure 4, average real returns (measured on a 10-year rolling basis to smooth fluctuations) amounted to about 12 percent p.a. in the first 10 years of the new system, well above the average growth of the real wage bill and of real wages in the same period. The comparison with the growth of the real wage bill is important, because this variable proxies the implicit rate of return in a balanced PAYG system, while the comparison with real wage growth is also relevant because final replacement ratios depend on the difference between the two variables. Average real returns have declined to about 6.5 percent, but still have exceeded the growth of the real wage bill and the average wage by 2 and 3 percent, respectively. 33. Computing accumulated real returns and real wage growth for different cohorts yields a similar conclusion. As shown in Figure 5, the cohorts that joined the private pension system in the 1980s have enjoyed an average real rate of return around percent p.a., well above the average real wage growth of 2-4 percent p.a. The difference between real returns and real wage growth declined for subsequent cohorts but has never been lower than 2.5 percent. The difference has tended to increase again for very recent cohorts but this is due to shorter averaging periods and high returns in some of the recent years. Overall, the numbers show that real returns have declined but have also remained above the growth of real wages and the real wage bill. AFP (%)

18 Average Real Returns (% p.a.) 15 Figure 4 14% 12% 10% 8% 6% 4% 2% 0% -2% Real Returns, Real Wage Growth, Wage Bill Growth 10-year Rolling Averages (%), Real Returns Real Wage Growth Real Wage Bill Growth Figure 5 12% 10% 8% 6% 4% 2% 0% Accumulated Real Returns, Real Wage Growth, Real Wage Bill Growth for Different Cohorts, Real Returns Real Wage Growth Real Wage Bill Growth Figure 6 14% 12% 10% 8% 6% 4% 2% 0% Average Real Returns and Share of Equity in Pension Fund Portfolios, Denmark Sweden Switzerland Chile Netherlands Australia Canada 0% 10% 20% 30% 40% 50% 60% 70% Average Share of Equity (%) US UK

19 Curiously, the returns of pension funds in Chile were higher precisely in the period when investment restrictions were more severe and the average share of equity was lower. Indeed, the return performance of Chilean funds in the period seems to have exceed the international return-risk pattern by a wide margin, as shown in Figure 6. No country was able to match the 12 percent average real return that Chilean funds achieved in this period, and the few countries that posted real returns close to 10 percent, such as the UK, only achieved this result through substantial holdings of equity. 35. The impressive returns achieved by pension funds in the first 15 years were due to two exceptional, one-time factors. The first was a large capital gain on long-term nominal bonds held by pension funds in the very first years, and that were generated by a sharp (and possibly unexpected) drop in inflation rates. The second was a large capital gain on the equity purchased in the late 1980s and early 1990s, particularly the equity purchased at modest prices in the first period of capitalismo popular. In more recent years real returns have dropped to levels that are more consistent with the portfolio composition of pension funds, and with Fund C in particular. This also implies that future returns will tend to be lower than in the past, particularly if the asset composition of Fund C remains the same, as this Fund will probably concentrate most of members assets during their working lives. 36. In a defined contribution and open pension system such as the Chilean, asset managers should seek to maximize return for a given risk, so as to maximize final replacement ratios, but determining the risk tolerance of fund members is not a trivial task. 11 The multi-fund regime provides some guidance to asset managers, but there is still a wide room for determining the size of asset classes within the ranges provided by the regulations, and asset managers have decided to maintain a reasonably conservative portfolio strategy in Fund C. The positive aspect of a conservative strategy is a lower volatility of returns across different cohorts. The less positive aspect is a lower mean rate of return. It may very well be that a more conservative strategy is on balance preferable for the case of Chile. However, the implication of a decline in average returns and the average replacement ratio for future retirees needs to be considered in the formulation of pension policies. 37. Rates of return have been very similar across different AFPs since the start of the pension system, and this pattern has been maintained in recent years, as shown in Table 10. This similarity of returns across AFPs reflects the similar composition of their portfolios, or the herding behavior that has been well documented in the literature. 12 There is a prevalent view in analyses of the Chilean system that the herding behavior in Chile is due partly to the concentrated structure, and partly to the existence of the minimum relative return guarantee, which involves bands around the average industry return and the obligation for AFPs to bring any return to the minimum with its own compulsory reserves (the encaje). 38. There was possibly an expectation that the creation of multi-funds would open room for more portfolio differentiation and less herding, on the grounds that there would be more choice, and that a lower rate of return in one fund managed by one AFP could be offset by a higher return in another fund managed by the same AFP (thus increasing the probability of 11 These issues are examined in Davis (2001) and Blake (1997). 12 See, e.g., Queisser (1998) and Srinivas, Whitehouse and Yermo (2000).

20 17 members remaining in the same AFP even if the first fund under-performed). In order to allow more room for differentiation, the regulatory authorities also widened the bands around the industry average for Funds A and B. Herding indeed seemed to be less intense during the early stages of the multi-fund system, but during 2003 there was again a convergence in the composition of fund portfolios, leading to very similar returns in 2003 and for the whole period since the creation of the multi funds, as shown in Figure 7 and Annex Tables 3 and Although the minimum relative return guarantee may exacerbate herding behavior, extreme herding has also been observed in other pension systems, including developed pension systems with thousands of pension funds and very liberal investment regimes such as the UK. 14 Herding behavior in these cases happens because asset managers are pressed by boards of pension funds to demonstrate a satisfactory return performance, which usually involves comparisons with peer or industry benchmarks, and tend to adopt defensive strategies in order to avoid under-performing relative to the benchmark and losing their contracts. Moreover, although herding does imply a potential risk of more asset volatility, there is no clear evidence that this has happened to a significant extent in Chile. Table 10: Real Rates of Returns of Individual AFPs (% p.a.), Cuprum Habitat Magister Planvital Provida Santa Maria Summa Bansander System Annex Table 4 shows that the dispersion across major asset classes increased at the end of 2002, but declined again during The same result has been achieved by an analysis of the tracking error since the creation of the multi-funds (Summa Bansander (2003)). 14 The extreme herding behavior in the UK is well documented in Myners (2002) and Blake et al (2000).