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

Transcription

1 OFFICE OF THE CHIEF ECONOMIST, LATIN AMERICA AND CARIBBEAN REGION, THE WORLD BANK BACKGROUND PAPER FOR REGIONAL STUDY ON SOCIAL SECURITY REFORM A Simulation Of Social Security s in Latin America: What Has Been Gained? by Asta Zviniene and Truman G. Packard

2 A Simulation Of Social Security s in Latin America: What Has Been Gained? by Asta Zviniene Social Protection Specialist, The World Bank and Truman G. Packard Economist, Social Protection, The World Bank

3 I. Introduction As in other regions, Latin America s population is aging. Advances in technology and healthcare have increased average life-spans dramatically in the last fifty years. Although the pace of the region s demographic transition varies widely, from relatively younger countries like El Salvador to relatively older countries like Uruguay, falling fertility rates combined with lengthening life expectancy are increasing the portion of the population in old age and lowering the number of new entrants into the work force. However, the blessing of longevity and an increasing relative share of retired elderly to active workers, have put social security institutions under enormous financial strain. This demographic change has been accompanied by economic liberalization and greater integration with the world economy. Growth in the share of elderly, and the push for greater efficiency and competitiveness, have forced policy makers in the region to re-examine labor market institutions first and foremost social (mainly retirement) security system - to accommodate these trends. Restructuring social security was made even more urgent by the fiscal deficits and mounting contingent liabilities of pension systems plagued by mismanagement and fraud. Structural s to social security in the region were initiated by Chile in 1981, and continued in the 1990s by Peru, Colombia, Argentina, Uruguay, Mexico, Bolivia, and El Salvador. Costa Rica, Nicaragua, the Dominican Republic and Ecuador enacted s between 2000 and Each involved a transition away from purely public pension systems - similar to those administered in Europe and the United States - to systems with explicitly defined multiple pillars administered and/or mandated by government. Most multi-pillar systems are designed so that the bulk of workers retirement income is financed from mandated private savings in individual accounts. These funds are invested by dedicated private pension fund managers, and while the activities of these fund managers and other financial service providers in the voluntary pillar are strictly regulated, the direct role of the state in the new model is often reduced to enforcing the mandate to save, regulating the new fund management industry, and guaranteeing a minimum threshold income to keep individuals from falling into poverty in old age. This paper presents the results of simulation analysis of the likely medium and longer term fiscal and equity outcomes of structural s to social security systems in Latin America. While several studies - conducted both prior to s and since - have presented the simulated fiscal impact of the shift to individual accounts in the region, rarely do existing studies extend beyond a single country case. The contribution of this paper is to evaluate the likely fiscal and equity impact of very different s in very different countries, using a uniform set of indicators and applying a single generic simulation model, the World Bank s Pension Options Simulation Toolkit (PROST). 1 These s, although passed into law, have yet to be implemented in Nicaragua, and Ecuador.

4 We should emphasize from the outset that we do not intend these results to be used to compare one country s with another s. Even where the multi-pillar models chosen by policy makers in different countries are very similar, circumstances directly and indirectly related to formal retirement security differ widely, invalidating any attempt at cross-country comparisons. Thus the results of the simulations cannot be used to determine whether Chile s was fiscally more successful than Peru s or Mexico s, or whether the impact of s on equity in Argentina were greater than in Colombia. Furthermore, since a number of assumptions we have imposed are very likely to vary, depending on who is performing the simulations and when these are performed, the results can only be used as indicators of the order of magnitude of various statistics and can only facilitate rough comparisons between countries. The sort of cross-country simulation analysis presented in this paper and elsewhere, has to be interpreted very carefully, and cannot replace country-specific analysis of s using tailor-made simulation models. This paper has five sections. Following this introduction, the principal features of the s we simulate for each country case are presented in Section II. This review of pension models countries have adopted is not intended to be comprehensive, but to give the reader a general idea of what has taken place in the region. 2 Section III presents the assumptions and methodology used in our simulation analysis with PROST. Section IV summarizes the main results dealing with fiscal sustainability of public pension promises and the equity impact of pension on formal retirement security institutions. Readers are referred to a more complete set of indicators in Annex I. Section V concludes with a note of caution about the methodology, and presents important caveats about our data and assumptions that have to be considered when interpreting the results. II. A Summary of Structural s to Pension Systems in Latin America The general character of structural s to public retirement security systems in the region has been similar. However, there is no single, cookie-cutter Latin American model. There are important differences that mark the introduction of the new pension systems from one country to the next. Table 1 shows some of the principal characteristics and differences between (structurally) ed 3 pension systems in the region. We have not included details on recent s in Costa Rica, the Dominican 2 For a comprehensive and detailed review of s in Latin America, readers are referred to Queisser, (1998) and Devesa-Carpio and Vidal-Meliá, (2001) 3 Although it introduced significant s to its retirement security regime for workers in the private sector, Brazil s does not qualify as structural among pension specialists since it did not introduce a system of mandated private saving. Brazil s is referred to as parametric since, although requiring changes to the country s constitution, ers adjusted contribution and benefit parameters within an PAYG financing framework.

5 Republic, Nicaragua, and Ecuador as we do not include these country cases in our simulation analysis. 4 In Chile, Mexico, Bolivia, and El Salvador mandated individual retirement accounts replaced public PAYGO institutions as the state s primary intervention to provide retirement income. In these countries, the (broadly defined) first pillar consists of guaranteed minimum contributory benefits, subsidies and other social assistance to the elderly indigent financed through general taxes. 5 In contrast Argentina and Uruguay (as well as Costa Rica and Ecuador) underpin pensions financed with accumulated individual savings with an explicit, redistributive first pillar still operating on a PAYGO basis that conditions benefits on a certain minimum number of years of contribution. 6 Another important dimension along which s in Latin America vary is the degree of choice afforded to workers. In Chile, Mexico, Bolivia, and El Salvador PAYGO systems were closed and workers were forced to take up individual retirement accounts. 7 In contrast, in Peru, Colombia and Argentina each new generation that takes up formal employment is allowed to choose between a significantly down-sized, earnings-related PAYGO pillar and individual retirement accounts as the primary financing mechanism for their pensions. Since the original s were designed to phase out the earningsrelated PAYGO systems in the longer term, if a new worker fails to choose they are affiliated to the system of individual accounts by default, and if they choose the remaining public pillar, they are always allowed to move to the private system at a later date. However, workers that choose private accounts cannot choose to move back to the PAYGO pillar. However, as a political concession to pass the, workers in Colombia were given the option to switch back and forth between systems every three years a feature of structural s in Colombia that has seriously compromised the new pension system. Largely in reaction against a policy of assigning all undecided workers to private individual accounts by default, draft legislation to allow similar switching has been 4 Although most of the data have been collected to conduct PROST simulations on these countries, two considerations lead to our decision not to include them. Structural s in Nicaragua and Ecuador have not been implemented yet this is to say that workers savings are not yet accumulating in individual retirement accounts either due to implementation difficulties (Nicaragua) or legislative problems (Ecuador). In the case of the Dominican Republic, prior to workers who earned above a certain wage ceiling were not allowed to participate in the pension system. This restriction was removed with the introduction of individual accounts. This feature is unique among the s in the region and prevents us from simulating uniform fiscal and equity indicators for Dominican Republic. Readers interested in the Dominican Republic are referred to Palacios and Pallares (2003). 5 For the sake of clarity, readers should be aware of an emerging distinction between contributory minimum benefits and non-contributory social assistance to the elderly. Since the publication of Averting the Old Age Crisis (World Bank, 1994) the terminology has been altered, the former contributory benefits being referred to as pillar one and the later non-contributory benefits designated pillar zero. 6 In most of these countries, some sort of non-contributory benefit targeted to the elderly poor also exists. 7 In Chile workers already contributing to the PAYG systems that existed prior to s were allowed to choose whether to move into the new system.

6 considered in Argentina. Similar attempts by the legislature to allow affiliates to return to the public system were recently defeated in Peru. In those countries where workers choose between public PAYGO and private savings arrangements for the earnings-related portion of their pension, the same first pillar (that is, minimum guarantees and basic poverty-prevention pension) arrangements usually apply. However, in Uruguay (and Ecuador, as yet not implemented), rather than pillars the new systems is best characterized as tiers, where participation in the new tier of private individual accounts is determined by the level of workers incomes. In the new privately managed second pillars, dedicated fund managers (Administradoras de Fondos de Pensiones or AFPs in Chile, Peru, Colombia, Bolivia and El Salvador otherwise named in the remainder of countries that mandate private savings for retirement) manage workers retirement accounts and invest their accumulated savings in tightly regulated portfolios. A portion of workers contributions to the private pillar pays for the services provided by the fund managers and covers the cost of premia for group disability and life insurance policies they are required to provide to contributing workers. 8 Workers are allowed to choose their fund manager from among the limited number in the closed industry, and those workers that have been with a fund manager for some minimum period of time, are allowed to switch to another (with restrictions on how frequently they can switch). Finally, there are significant differences in which groups of workers are affected by s to retirement security systems from country to country. Usually out of political considerations ing governments avoided making structural changes to the public pension systems benefiting the military and civil servants. After structural s to the social security systems for workers in the private sector, separate public pension systems remain for civil servants in Colombia, Mexico and El Salvador. Although nominally closed to new entrants, organized groups of workers are continuously trying to be covered by the generous PAYGO parameters of Peru s pension regime for civil servants (the cedula viva ), with mixed success. The retirement regimes for federal civil servants in Argentina were integrated into the ed national system, however, separate, relatively generous pension systems remain for civil servants in roughly half of Argentina s provinces. The military (and in many cases the police) retain separate, special retirement and other income security arrangements in all of the countries where ed systems are now in place. The simulation results presented in later sections of this paper pertain to structural s of the principal national PAYGO systems that covered mainly workers with formal employment in the private sector. 8 In Mexico and Costa Rica disability pensions are still paid by the public pillar.

7 Table 1. Principal Features of (Structurally) ed Pension Systems in Latin America Year of Earnings-related public PAYGOO system? Participation of new workers? Participation of self employed? Remaining separate system for civil servants? Chile Peru Colombia Argentina Uruguay Mexico Bolivia El Salvador closed to new remains remains remains remains closed closed closed entrants mandatory voluntary voluntary voluntary voluntary b mandatory mandatory mandatory voluntary voluntary voluntary mandatory mandatory voluntary voluntary voluntary no yes yes yes no yes no yes Dedicated fund managers AFP AFP AFP AFJP AFAP AFORE AFP AFP Contribution to individual subsidy retirement account (% of wage) Fees & insurance premia (% of wage) Switching between fund 2 x annually 2 x annually 2 x annually 2 x annually 2 x annually 1 x annually 2 x annually 2 x annually managers? Pay-out options Annuity or scheduled withdrawal Annuity or scheduled withdrawal Annuity or scheduled withdrawal Annuity or scheduled withdrawal Annuity only Annuity or scheduled withdrawal Annuity only Annuity or scheduled withdrawal Minimum return on investment? relative to average relative to average relative to average relative to average absolute no no d relative to average Minimum contributory pension? Social assistance pension? yes yes (but not for AFP yes yes yes yes no yes affiliates under 55) yes no - yes yes no yes - Source: Packard (2001), adapted from Cerda and Grandolini, (1998), Queisser (1998), Devesa-Carpio and Vidal-Meliá, (2001), AIOSS (2001), FIAP (2002) tes: a) Costa Rica introduced voluntary retirement accounts in 1996, but made private individual retirement saving mandatory in Fees are charged as a percentage of returns from investment. b) Participation in IRAs in Uruguay and Ecuador is determined by income level. Workers below a threshold level choose to split contributions between PAYGO or individual retirement accounts. c) IRA contribution net of fees and insurance premia as share of total payroll deductions (Packard, 2001) d) Guarantees required from the fund managers

8 III. Simulation Assumptions and Methodological Considerations Before reviewing the set of fiscal and equity indicators from our PROST simulations, it is important that we present the set of exogenous assumptions that can effect the results. Readers familiar with actuarial simulations will know how sensitive the results can be to assumptions made about economic growth, interest rates, and evolution of the covered wage-bill as a portion of a country s GDP, as well as other parameters. The long time horizon of the simulations typically over 50 years only compounds the sensitivity of the results we present to our assumptions about these parameters. We have chosen to impose a uniform set of macroeconomic and certain systemic assumptions, in order to isolate the impact of structural s. While this would seem a strange choice when simulating s in a group of countries as diverse as El Salvador, Mexico and Uruguay, it is not without precedent in cross-country simulation analysis (see Holzmann, Palacios and Zviniene, 2001). The assumptions that are common to all country cases are presented in Table 2. Details particular to each country case are presented in Annex II. Table 2. Assumptions Common to All Country Cases GDP growth = 3% Inflation = 0% Market Interest, discount rate = 5% Interest in deacumulation phase = 4% Interest on PAYGO investments = 3% Wage bill set at a constant portion of GDP Full compliance with retirement age increase everybody contributes if not allowed to retire Collection rate of contribution revenue = 100% For sustainable benefit, we assume that all benefits (old age, disability, survivors, etc.) are reduced by the same percentage. All demographic data is taken from World Bank population database (see E. Bos, M.T. Vu, E. Massiah, R.A. Bulatao, 1994) coverage expansion or contraction Recognition bonds are not included in implicit pension debt calculations Minimum pension guarantees are modeled where applicable PAYGO pillar benefits, income thresholds, contribution ceilings and minimum pensions are all indexed to nominal wage growth Fees paid by affiliates to funded second pillars, as in Devesa-Carpio and Vidal-Melia (2000)

9 In addition to the parameter assumptions documented in Table 2 (and those listed in Annex II), there are assumptions behind the simulated pension outcomes used in the equity analysis that have to be kept in mind when analyzing our results. We calculate internal rates of return from representative affiliates investments in the formal retirement security system an indicator that can be used to assess the likely impact of s to the formal pension systems on income equity. Details on the affiliate profiles used in this analysis are shown in Table 3. We use sexspecific average retirement ages and average years of contributions from PROST output tables. The average retirement age for men and women is applied to each of the three affiliate profiles that we have constructed for each sex (poorer affiliates; average affiliates; richer affiliates). When including years out of the formal labor market (particularly when calculating the internal rate of return for representative affiliated women) we assume that these years were evenly spread throughout the affiliate s adult pre-retirement life. Survivor benefits are not included in the internal rate of return calculations Table 3. Assumed Profiles of Representative Affiliated Men and Women (with respect to average affiliate of each sex) Profile Entry Age Years of Contributions Starting Wage Productivity Growth Mortality Poorer affiliate 16 Average % of average Average affiliate 19 Average 100% of average Richer affiliate 22 Average 3 150% of average 50 % of average 100% of average 150% than average Dies 2 years earlier than average 100% of average Lives 2 years longer than average Finally, in most of the country cases shown, structural s are more than several years old (see Table 1). In most cases, pension system data for the years prior to structural s are scarce, and when available, rarely can be found in the level of disaggregation required for our simulation model. In these cases (Chile, Peru, Argentina, Colombia) the simulations presented were done based on current data. To project a no- scenario the counterfactual to structural s - the current number of beneficiaries, wage distribution of contributors, etc. were used, and the parameters of the old single-pillar PAYGO system were applied. To simulate structural s particularly the introduction of individual retirement accounts we assumed that s were introduced in the base year - the year to which the data actually correspond). However, the switching pattern that is the distribution by sex and age of workers moving from single pillar to multi-pillar arrangements was set to reflect the actual distribution of contributors between the old PAYGO and new multi-pillar systems.

10 To take account of workers savings in individual accounts from the actual date of s to the base year, we simulated a one-off transfer of funds into the new private pillar. We did this by setting the contribution rate (as percentage of wages) to the new individual accounts in the base year equal to the actual statutory contribution plus a percentage of wages that reflected average accumulated assets in each country. In this way, we simulated a one-off contribution into individual accounts equal to the average accumulation of savings, divided by average wages. The value of recognition bonds was also adjusted for the actual date of s. While this technique is admittedly second best, past applications have rendered simulation results very similar to those of other authors. IV. Summary of Simulation Results iv.a. Fiscal Impact of Structural s Prior to s, pension liabilities are often one of the largest item on public budgets in developing countries. Governments inability to meet the growing liabilities - implied in the benefit promises of single-pillar PAYGO systems, and often unaccounted for in public sector balance sheets - is a source of policy risk to retirement income, and usually the driving force (and political selling point) of (Holzmann, et al, 2001, Holzman, 1998). Just as with the institutions that were replaced, the fiscal sustainability of public pension promises after s, can determine the credibility of the new multiple pillars of the social security system. Have structural s made governments pensions promises more fiscally sustainable? Figure 1 shows the value of governments implicit pension promises in each of the countries that undertook structural s to their retirement security systems. The figure also shows the value of this indicator had there been no s. The value of implicit pension promises is otherwise referred as implicit pension debt (IPD) and is defined as the present value of the stream of future benefits that a public pension system will have to pay current participants (contributors, beneficiaries and their survivors) according to the defined parameters of the system, to recognize their contributions up to the particular year in question. There are several different concepts and methods for calculating the IPD. The simulations presented in this paper employ a practical termination liability approach, described in detail in Holzmann, et al (2001), and proposed as the best method of calculating the IPD for cross-country comparisons. It comes as little surprise that where s phased out earnings-related pensions from the public PAYGO pillar Chile, Mexico, Bolivia and El Salvador that the simulated cost of governments pension promises (as a percentage of GDP) falls rapidly. However, even in countries where an earnings-related PAYGO pillar was retained in Peru, Argentina and Colombia - and/or where explicitly defined, first pillar benefits underpin pensions from a private second pillar as in Argentina, and Uruguay - s are likely to dramatically slow the growth of public pension liabilities, as measured by IPD.

11 However, a dramatic decrease in government pension promises is to be expected from s that partially privatized public pension systems. The simulated implicit pension debt falls in the wake of s as a portion of these obligations is converted into explicit debt or paid with transfers from the general budget. Although changes in the implicit pension debt reveal the extent of and how countries chose to spread the costs of transition from one regime to the next, a better measure of fiscal sustainability is the rate at which the total public debt for pensions is accumulating after s, compared with the rate of total debt accumulation had there been no s. The total pension debt shown in Figure 2 is that financed by government borrowing, and includes: (i) the current deficits of remaining PAYGO systems (the difference between pension payments and contribution revenues); (ii) payments to cover the minimum guaranteed pensions to workers contributing to private individual retirement accounts where such guarantees exist; (iii) government contributions to either a PAYGO regime or individual accounts in countries where these are made explicit in the law; and (iv) payment of recognition bonds to honor workers contributions to pre- systems. Even an analysis of governments simulated total pension debts accumulated after 2001 (Figure 2.) and the rate of it s accumulation, on the whole shows a dramatic improvement in fiscal sustainability brought about by s. The simulations show substantial savings from the introduction of individual accounts and accompanying s. These savings are most apparent in Peru, Bolivia, Uruguay, Chile and El Salvador. However, in Argentina our simulations show a considerable increase in the federal government s total expenditure on pensions in the scenario. 9 In addition to the loss of contribution revenue from workers who switched to individual accounts, these projections capture the increase in PAYGO deficits that arose from a policy of lowering employer contributions to the public pillar a (arguably failed) measure introduced after the 1994 to increase compliance with the mandate for employers and workers to participate in the system. 10 Another critical factor that caused Argentina s total spending on pensions to balloon since the in1994, was the federal government s policy of accepting the liabilities of overly-generous pension plans for civil servants at the provincial level. Provincial governments that agreed to close their pension regimes to new entrants, and to force contributing civil servants to join the national system along with private sector workers 9 Readers should note that our data and assumptions for Argentina were taken prior to the crisis and the devaluation of the peso. An account of the crisis and its impact on the ed pension system can be found in Rofman (2002). 10 The policy of lowering employer contributions was pursued to increase compliance. However, regulations in the product and factor markets that have little to do with the social security system, keep the cost of compliance and formalization in Argentina very high. Thus, lowering employer contributions to the public pension system had no substantial impact on lowering evasion, and has only served to deepen pension deficits (Rofman, 2002).

12 (either the publicly administered, earnings-related PAYGO branch, or mandated private individual accounts), transferred the obligation of paying the relatively generous benefits of retired provincial civil servants to the federal system. The sudden increase in total pension liabilities, and revenue reductions to the retirement security system stemming from lower contribution rates, widespread evasion and a long recession aggravated the fiscal stance of Argentina s multi-pillar pension system. However, our simulations only show the effect of lower contribution rates. Readers should note that the results presented here are simulations of the fiscal impact of laws, and thus (somewhat optimistically) assume that s where implemented correctly and that the new systems are then adequately administered. The simulations cannot capture probable administrative difficulties that could cause fiscal costs of to balloon unexpectedly. Fiess (2003) points out that the case of Bolivia is instructive in this regard, as the transition proved more costly than initially anticipated. When the Bolivian was designed and implemented, insufficient attention was paid to the institutions that were to govern the transition from the old to the new system. While a regulatory body was set-up to govern the new private pension funds, the system transition itself was insufficiently regulated, inviting fraudulent claims and a lax interpretation of the rules for and definitions of transition workers. This has contributed to higher than expected transition costs. Finally, the simulated results in this section are used solely to evaluate whether the sustainability of public pension promises has been improved by structural s in the region. This is to say, we are concerned with the reduction of policy risks to retirement income. Therefore, our analysis is solely focussed on whether the reductions of what were unsustainable PAYGO pension promises, are likely to reduce the risk of governments having to default on workers public pension rights. We cannot, with simulation tools employed in this paper, address wider macroeconomic concerns and possible risks that are raised when governments convert implicit pension debts into explicit debts with structural s to the pension system. These concerns are explained at length in Fiess (2003).

13 Figure 1. Simulated Implicit Pension Debts, With and Without Structural s, Percentage of GDP 400% IPD in 2001 IPD in 2020 IPD in 2030 IPD in % Government's Implicit Pension Debt as Percentage of GDP 300% 250% 200% 150% 100% 50% 0% Chile Peru Colombia Argentina Uruguay Mexico Bolivia El Slavador

14 Figure 2. Total Pension Debt (explicit and implicit) accumulated after 2001, With and Without Structural s, Percentage of GDP 400% Government's Total Pension Related Explicit Debt, as Percentage of GDP 350% 300% 250% 200% 150% 100% 50% 0% Chile Peru Colombia Argentina Uruguay Mexico Bolivia El Slavador

15 iv.b. Equity Impact of Structural s Although presented as a model of solidarity when they were introduced, most singlepillar social security systems in developing countries that operate on a PAYGO basis can be regressive in a number of ways. First, pension benefits are based on earnings rather than on need, and are often calculated to favor better educated workers with steeper earnings profiles. } Second, contributions from poorer workers with higher average mortality often subsidize benefits paid to longer-lived, higher income workers. Third, and related to the above, poorer workers tend to begin working and contributing earlier than those who are better off workers in higher paying jobs requiring more education tend to join the labor force later; since they start working sooner, the poor contribute longer during their active lives, for a relatively shorter stream of benefits in retirement. Finally, related to the implicit liabilities and borrowing requirements discussed in the previous sub-section, since the majority of workers in developing countries will not receive any benefits, the deficits of un-balanced public pension systems that are financed from current and future tax revenues can represent a transfer from uncovered workers to those covered by the systems a relative minority of workers already benefiting from more stable, better paying forms of employment. By replacing what were often regressive, single-pillar PAYGO systems that frequently paid overly generous pensions to a privileged few, with systems that diversify the risks to retirement income, s were expected to introduce a more effective system for lowering inequality. ed systems would offer minimum income guarantees, while freeing public resources for better targeted forms of social assistance. Have multi-pillar systems corrected the regressive impact of single-pillar public PAYGO systems? Prior to this report, little work has been done in this area, largely because the final impact of s on income inequality cannot be precisely measured until large segments of the population begin to retire with pensions from individual retirement accounts. This said, there are important equity aspects of the s that can be examined exploiting income distribution features in PROST. Our simulations show that the introduction of multi-pillar systems in the countries included in this analysis had a substantial impact on distribution between beneficiaries of different income levels. Figure 3 shows the simulated impact of on the internal rates of return earned by poorer versus wealthier beneficiaries. The red bars show the percentage point difference between the internal rate of return earned by a representative, wealthier-than-average worker, and that earned by a poorer-than-average worker of both genders in each country had there been no s to the social security system. The white bars show the same differential in internal rates of return, but taking s into account.

16 Figure 3. s Improve Equity by Lowering Regressive Transfers and Returns 4 3 (Percentage Point Difference Between Wealthier and Poorer Workers in Internal Rates of Return Earned from National Retirement Security System ) 2 Percentage Points Difference Men Women Men Women Men Women Men Women Men Women Men Women Men Women Men Women Chile Peru Colombia Argentina Uruguay Mexico Bolivia El Slavador As can be appreciated by the differences in internal rates of return without s, single pillar PAYGO systems in Latin America were notoriously regressive, conferring substantially higher returns to wealthier affiliated workers than to poorer affiliated workers. In every country (for which data were available for PROST simulations) social security s that introduced multi-pillar systems lowered the regressive impact of single-pillar systems. In Chile and Argentina, s went as far as reversing regressive returns, increasing returns earned by poorer men and women relative to wealthier workers of both genders. s have also had a notable impact on distribution of returns between genders. Figure 4 shows the percentage point differential between the internal rates of return earned by men and women of average income levels affiliated to the pension system in each country. In Chile and Colombia s increased the returns earned by women relative to those earned by men. Cox-Edwards (2000) claims that the impact of s in Chile favoring women can be attributed to the minimum pension guarantee underpinning individual private savings in that country. However, in every other country where the new retirement security model was adopted, women earn lower returns from the new systems relative to the returns earned by men. This said, James, Cox-Edwards and (2002) use survey data and simulation techniques to show that poor women in Mexico, Argentina and Chile have gained from s, receiving higher pension benefits than they would have under the single-pillar PAYGO systems. The authors attribute these gains to a better targeted first pillar.

17 In Mexico, Bolivia and El Salvador, s replaced systems that subsidized benefits to women with systems that favor men. This shift may partially reflect the relative greater importance of labor market participation and regular contribution in determining pension benefits in the new retirement security systems based on defined contributions (James, Cox-Edwards and Wong, 2002). The negative impact on the returns to women may reflect their relatively fewer years of employment and active contribution to the pension system. Men with fewer years of employment and those who work with fewer years of contributions would suffer a similar fall in returns with the introduction of individual accounts. To the extent that participation in the labor market is voluntary, the lower return earned by women in the ed systems should not be of too much concern. However, if there is still discrimination in the labor market against women in the form of lower wages, or limited opportunities other than uncovered employment the relatively greater importance of regular contribution in the ed pension systems could be a problem. Figure 4. s Can Hinder Equity Between Men and Women 4 Percentage Point Difference Between Men and Women in Internal Rates of Return Earned from National Retirement Security System ) 3 2 Percentage Points Chile Peru Colombia Argentina Uruguay Mexico Bolivia El Slavador However, the internal rates of return calculated with PROST shown in Figure 4, assume that participating men and women have the same work and contribution history. There is a more important factor aggravating the difference in returns from the new systems between men and women: the use of sex-specific mortality tables in the calculation of retirement annuities. Since women have relatively longer life-expectancy at retirement, annuity providers using sex-specific mortality tables will calculate annuity payments that are significantly lower than payments made to men retiring with similar levels of accumulated savings. Policies mandating married male affiliates to retire with joint annuities that will cover their female spouse, can improve the retirement security of

18 surviving widows. Mandates requiring private annuity providers to use unisex mortality tables can correct disparities in returns from systems based primarily on individual retirement accounts, but are highly controversial as they can hinder the functioning of insurance markets. Such attempts to meet equity objectives through further mandates in the operations of the private second pillar requiring insurers to use unisex mortality tables, in particular - can be detrimental to the development of private insurance markets, and are better pursued through whatever first-pillar arrangements countries have put in place. There is evidence to suggest that the new, first-pillar arrangements are succeeding in lowering the vulnerability of elderly women (James, Cox-Edwards and Wong, 2002). In the interest of brevity, we have only summarized some of the PROST indicators simulated for each of the country cases. Readers are encouraged to review the indicators presented in Annex I, which in addition to those already presented in Figures 1-4, include: the share of total pension benefits from the funded second pillar; (ii) average replacement rates earned from the funded pillar; and (iii) replacement rates earned by the poorest quintile of beneficiaries, among others. V. Caveats and Sensitivity of PROST Results to Data Errors and Assumptions The results of our simulations presented in the previous section, show that, with some notable exceptions, structural s to pension system in Latin America are likely to have a beneficial impact on the fiscal sustainability of more modest public pension promises, as well as an eventual impact on equity having eliminated an institutional source of inequality. However, as stressed by Holzmann, et al (2001) the sort of crosscountry simulation analysis presented in this paper, has to be interpreted very carefully, and cannot replace careful country-specific analysis of s using tailor-made simulation models. The PROST simulations presented in this paper must be treated with caution. Since a number of assumptions we have imposed are very likely to vary, depending on who is performing the simulations and when these are performed, the results can only be used as indicators of the order of magnitude of various statistics and can only facilitate rough comparisons between countries. As detailed in Holzmann, et al (2001), there are four sources of errors that can hinder the reliability of simulation results. First the data may be incomplete. Usually the problem lies in the insufficient disaggregation of the data. For example, only unisex average pensions are available for El Salvador, when there is a strong reason to believe that pensions of women are somewhat lower than those for men. Thus the same value of average pension was used for both sexes. This will slightly overestimate the current balance of the PAYGO and implicit pension debt, since women receive pensions for longer periods than men. The second source of error is introduced in cases where the data for a specific country are available but ignored since the particular data are not available in another country in out set of country cases. For example, we encountered this problem in trying to determine

19 benefit indexation rules which were obscure or ad hoc in the majority of cases. Therefore, we decided to use a uniform assumption on this very influential parameter. The third source of possible error could arise from making uniform assumptions across very diverse countries. Holzmann, et al (2001) explain that actuaries and economists often use anecdotal evidence to arrive at reasonable assumptions about the future path of the system that is being simulated. However, these assumptions are often difficult to support. For example, a macroeconomist may argue that the wage share of GDP will grow in El Salvador, while the same parameter will remain stable in Uruguay. However, it is not possible to determine the degree of future change in this parameter and assumptions have to be made. As Holzmann, et al (2001) point out, using different, only partially substantiated assumptions for different countries in a cross country study is particularly problematic. Thus, following the authors, we chose to make uniform assumptions for all countries where differentiated assumptions could not be firmly defended. The fourth source of possible error comes from the use of generic model like PROST in a cross country study. On the one hand, using the same method for across country cases adds to the comparability of indicators. On the other hand, there are limits to the detail that can be addressed by a generic model which can never substitute for a model tailor made for a particular country case. Thus, in countries with very complex pension systems, some very particular aspects of the system are not modeled. For example, PROST allows the user to model four benefit categories that usually are old age, disability, survivor and orphan benefits. If the country has five benefit categories we would average the two least significant ones or choose to ignore some financially insignificant benefit categories altogether. All of the above-mentioned factors can add up to significant departure of results from true pension system indicators. However, every attempt was made to avoid systematic errors and to document the assumptions that have been made.

20 References Kleinjans, Kristin J. (2001) The Colombian Pension System After the of 1994: An Evaluation CEPAL (1999), Encuesta Contigua de Hogares, vember Bos, E., M.T. Vu, E. Massiah, R.A. Bulatao (1994): World Population Projections , Johns Hopkins University Press, Baltimore, MD Jose E. Devesa-Carpio and Carlos Vidal-Melia (2000) ed Pension Systems in Latin America Pension Primer Series, World Bank Holzmann, Robert, Robert Palacios and Asta Zviniene, (2000) Implicit Pension Debt: Issues, Measurement and Scope in International Perspective Paper prepared for IAES World Bank (1994) Averting the Old Age Crisis Palacios and Pallares (2003) on DR` Fiess, rbert (2003) Pension or Pension Default: Macroeconomic Concerns Raised by Introducing Individual Retirement Accounts Background paper for a regional study on social security, Office of the Chief Economist, Latin America and Caribbean Regional Office, World Bank

21 Annex One. Full PROST Output from Simulations Table A.1. Current Pension Deficits (Benefit Expenditure Contribution Revenue) Financed by Government Transfers Uruguay Argentina Mexico Bolivia Colombia Chile El Salvador Peru 2001 (4.0%) (3.4%) (2.5%) (.1%) (.5%).2% (3.5%).8% 1.6% (.0%) (7.2%) (.1%) (1.4%) (.8%) (.7%).7% 2010 (2.6%) (2.2%) (2.2%).1% (.6%).0% (2.2%).0% 1.5% (.4%) (4.6%).2% (2.2%) (1.0%) (.9%).7% 2020 (2.1%) (2.2%) (2.3%) (.3%) (.7%) (.3%) (2.1%) (1.1%) 1.0% (1.5%) (3.4%) (1.0%) (3.2%) (1.4%) (.9%).4% 2030 (2.2%) (3.4%) (2.8%) (1.3%) (.7%) (.8%) (2.1%) (3.0%) (.7%) (3.3%) (1.5%) (2.7%) (2.9%) (2.1%) (.9%) (.1%) 2040 (2.5%) (5.0%) (3.6%) (2.8%) (.7%) (1.5%) (1.7%) (5.3%) (3.4%) (5.5%) (.5%) (3.9%) (2.6%) (3.1%) (.8%) (1.2%) 2050 (2.8%) (6.6%) (4.4%) (4.3%) (.6%) (2.3%) (.9%) (8.5%) (5.4%) (7.6%) (.8%) (4.0%) (.5%) (4.1%) (1.0%) (2.3%) Table A.2. Total Accumulated Explicit Pension Debt Since 2001 Uruguay Argentina Mexico Bolivia Colombia Chile El Salvador Peru % 34% 25% 10% 9% 8% 16% 29% 13% 30% 22% 19% 11% 21% 6% 6% % 83% 58% 35% 19% 23% 32% 84% 44% 82% 47% 51% 28% 52% 14% 19% % 142% 96% 72% 27% 43% 44% 164% 91% 151% 61% 92% 53% 94% 24% 40% % 213% 140% 119% 35% 68% 50% 269% 152% 235% 68% 134% 74% 144% 33% 69% % 290% 189% 172% 42% 96% 53% 392% 224% 330% 74% 177% 87% 202% 43% 104%

22 Table A.3. Implicit Pension Debt Uruguay Argentina Mexico Bolivia Colombia Chile El Salvador Peru % 196.1% 74.6% 87.3% 14.9% 22.0% 50.8% 108.0% 52.2% 55.5% 40.3% 130.1% 26.5% 46.3% 14.1% 32.8% % 207.9% 78.8% 101.0% 16.7% 31.2% 43.4% 139.6% 76.5% 83.4% 24.7% 152.3% 19.6% 59.0% 12.5% 44.1% % 235.0% 89.1% 125.1% 18.0% 44.0% 37.8% 188.3% 114.3% 125.7% 10.1% 179.6% 9.9% 78.4% 11.8% 60.4% % 266.7% 102.0% 154.2% 17.6% 58.1% 28.0% 247.5% 155.9% 170.8% 2.9% 201.1% 3.0% 102.4% 12.6% 80.3% % 295.2% 113.8% 179.5% 16.1% 71.5% 16.0% 309.9% 193.2% 210.3%.4% 208.7%.3% 126.2% 13.8% 100.8% % 313.3% 121.3% 195.6% 14.0% 79.6% 6.0% 363.4% 219.8% 237.9%.0% 210.8%.0% 148.3% 14.9% 118.1% Table A.4. Percentage of Pension Expenditures Financed From Contribution Revenue in 2050 Uruguay Argentina Mexico Bolivia Colombia Chile El Salvador Peru 74% 53% 43% 51% 86% 48% 89% 37% 35% 14% 87% 58% 87% 12% 80% 40%

23 Table A.5. Average Replacement Rates For Retiring Men and Women in 2050 Uruguay Argentina Mexico Bolivia Colombia Chile El Salvador Peru Men 69% 67% 43% 49% 62% 71% 45% 79% 53% 55% 73% 107% 49% 47% 52% 60% Women 43% 56% 35% 48% 46% 57% 23% 43% 49% 50% 39% 77% 25% 42% 40% 48% Table A.6. Average Replacement Rates Received by Poorest Quintile of New Beneficiaries in 2050 Uruguay Argentina Mexico Bolivia Colombia Chile El Salvador Peru Men 14% 14% 24% 24% 27% 27% 12% 24% 60% 60% 26% 41% 20% 23% 32% 32% Women 14% 11% 22% 23% 27% 27% 8% 15% 60% 60% 26% 26% 16% 34% 32% 32% Table A.7. Internal Rates of Return Earned by Retiring Men and Women of Different Income Levels Men Uruguay Argentina Mexico Bolivia Colombia Chile El Salvador Peru Richer 4.6% 6.0% 3.0% 4.7% 4.3% 6.4% 4.3% 9.4% 5.9% 9.9% 4.5% 6.2% 3.3% 12.1% 5.0% 8.3% Average 3.6% 4.5% 4.2% 4.4% 4.1% 6.2% 4.2% 7.9% 7.1% 10.6% 4.1% 5.9% 3.2% 10.5% 4.6% 7.3% Poorer 2.6% 2.8% 5.0% 3.4% 4.1% 5.5% 3.7% 6.7% 5.9% 8.7% 4.9% 5.7% 3.2% 9.2% 5.9% 6.2% Women Richer 4.7% 7.9% 4.8% 5.5% 4.3% 6.7% 4.3% 10.1% 7.9% 11.3% 4.3% 7.6% 3.5% 12.7% 5.0% 9.1% Average 3.7% 6.5% 4.6% 6.6% 4.0% 6.4% 4.1% 8.7% 8.0% 11.2% 5.7% 7.4% 3.0% 11.4% 6.3% 8.4% Poorer 3.0% 5.0% 5.8% 5.9% 4.9% 5.8% 3.7% 7.5% 6.9% 10.0% 6.7% 7.2% 4.0% 11.5% 7.1% 7.4%

24 Table A.8. Contribution Rate Necessary to Fully Self-Finance the Minimum Contributory Benefit (Percentage of Pay-roll) Uruguay Argentina Mexico Bolivia Colombia Chile El Salvador Peru Men 12% 19% 22% 96% 8% 8% N/A 42% 38.0% 61% 12.0% 12% 4.0% 9.0% 10.0% 10.0% Women 21% 33% 40% 87% 12% 12% N/A 44% 58.0% 84% 20.0% 20% 7.0% 14% 24.0% 24%

25 Annex Two. Country Specific Details, Assumptions and Sources of Data Country Case s Modeled Assumptions Data Sources Chile Introduction of funded second pillar Administrative costs of PAYGO pillar as % of contributions is assumed to be 5% (reliable data was not available) Average length will continue to be 23/38 Peru Introduction of a funded second pillar Increase in contribution rate from 9% to 13% in PAYGOO second pillar and 12.4% in funded second pillar Average length of service stays constant at 23.8 years Administrative costs of PAYGOO pillar as percentage of contributions is assumed to be 5% (reliable data was not available) Wage distribution, pension distribution, beneficiary numbers, current spending, current average length of service from INP and SAFP Wage distribution, pension distribution, beneficiary numbers, current spending from ONP and SBS Colombia Introduction of a funded second pillar Increase in retirement age from 55/60 to 62/57 in year 2014 Average length of service is 20 (mandated minimum). Only ISS is modeled (33% of all pension liabilities), other special regimes for civil servants make up the rest Beneficiary numbers, current spending, administrative costs of PAYGOO from local sources Wage distribution, switching patterns from Kleinjans (2001) Argentina Introduction of a funded second pillar Decrease in contribution rate for employers from 16% to 7% in PAYGOO second pillar Administrative costs of PAYGOO pillar as percentage of contributions is assumed to be 5% (reliable data was not available) Average length of service since 2008 is assumed to be 30 for those who retire at ages younger than 70 and 15 for those who retire later (mandated minimum) Wage distribution, pension distribution, beneficiary numbers, current spending from ANSeS and SAFJP Uruguay Introduction of a funded second pillar (more appropriately characterized as a tier since participation is determined by income) Equalization of retirement ages for both sexes at age 60 Average length of service is assumed to be 35 for those who retire at ages younger than 70 and 15 for those who retire later (mandated minimum). Bolivia Introduction of a funded second pillar Administrative costs of PAYGOO pillar as percentage of contributions is assumed to be 5% (reliable data was not available) Average length of service is 25 (mandated minimum) BONOSOL program is modeled Wage distribution, pension distribution, beneficiary numbers, current spending, administrative costs of PAYGO pillar, switching patterns from local sources Wage distribution, pension distribution, beneficiary numbers, current spending from local sources Gender differences in wages from CEPAL (1999)

26 Mexico Introduction of a funded second pillar Average length of service is growing from current 22 to reach 25 by 2012 and then stays constant (mandated minimum). Male/Female ratios in contributor and beneficiary population taken from Argentina, which has very similar gender structure of labor force to Mexico (Mexico 1980: 27% female; 1996: 31% female. Argentina 1980: 28% female; 1996: 31% female. Source: WBI). For working age disabled used contributor ratios and for old disabled old age pensioner ratios. INFONAVIT and social quota programs are included Wage distribution, pension distribution, beneficiary numbers, current spending, administration costs of PAYGOO pillar from CONSAR and IMSS Gender differences in wages taken from ENEU, 1998 El Salvador Introduction of funded pillar Average length will continue to be 25/27 Wage distribution, pension distribution, beneficiary numbers, current spending, administrative costs, current average length of service from local sources