92717 Review of the Actuarial Forecasts of the Proposed Contributory Social Security Regime in Timor-Leste World Bank Office, Jakarta September 14, 2014 This report briefly summarizes the results for our analysis of the design, financing and administrative processes for a national social security system covering all salaried workers in Timor-Leste, including analysis of the actuarial report (ACL Report) prepared for the Directorate General of Social Security by its consultant. I. Scope of Work In accordance with the TOR from the Ministry of Finance, we were asked to review and comment on the following issues:  Use the assumptions and data sources in the ACL Report to recreate the costing presented in that document. Any material differences between the costing in the ACL Report and those calculated by the WB should be noted. The costing should also be forecasted beyond 2045 (the date stated in the ACL Report) to 2050 and 2080. The financial stability of the scheme in the very long term (after 2045) should be analyzed.  Comment on the likely accuracy of each assumption and data source used in the ACL Report. In the event that assumptions are thought to be unrealistic and/or superior data is available, the WB should re-cost the pension scheme with more realistic assumptions and comment on these results.  In the event calculations show that the scheme is financially unsustainable in the long run, then the WB should: o Calculate and present the contribution rate that would be required to make the pension scheme financially sustainable with the benefits outlined in the ACL Report and SSR law o Calculate and present the change to the benefits required to make the scheme financially sustainable given a 15% contribution rate  Calculate and present the amount that the benefits outlined in the ACL Report would have to be reduced by given a total contribution rate of 6.75% (with the scheme still covering the private and public sectors). Please note that in our analysis, we analyzed the cost of old age, disability and survivor benefits only. We did not analyze the design or the cost of the maternity, paternity, adoption and funeral expense provisions of the proposed design. II. ACL Report Assumptions and Methods Comments Based on our review of the ACL report, we believe improvements could be made in the methods and assumptions used to estimate the revenues, expenses and long-term cost of the proposed national social security system, as many of these appear to differ from standard international practice and guidelines. These improvements would significantly improve the accuracy of the projections and result in a large increase in expected long-term costs as a percent of covered payroll. The following methods and assumptions used in the ACL appear to differ from standard international practice, and if so, should be adjusted. We wish to note that we did not have the opportunity to speak with the actuary who prepared the ACL report, so we were not able to clarify some questions regarding assumptions and methods, and relied on the contents of the report and our own modeling efforts to reproduce the results in the report.  Projection period. The projection period of 30 years is well below international standards and is particularly low for this plan because it is a new plan, does not grant credit for prior work history, and population aging will not begin to occur in Timor-Leste until many years in the future. International standards even for mature pension programs require a projection period of at least 75 years and given the characteristics of Timor-Leste and this particular program, an even longer projection period should be considered  Real wage increase. Real wages increase over the long term in every country due to inflation and improvements in productivity. In developing countries like Timor-Leste, large improvements in productivity are likely due to improving education, health, sanitation, infrastructure, technology, etc. In Western countries, productivity growth averages 1-1.5% per year. In the short run, rates of productivity growth of 5% per year or higher are possible in Timor-Leste, and in the long-term, productivity growth should grade down to global averages. Ignoring the likely growth of salaries significantly distorts the analysis  Use of average wage for projections of benefits. Benefits under this proposed program are based on the valorized average wage over the best 120 months of a workers’ career. The wage used to calculate pension benefits is therefore likely to be significantly higher than the national average wage, as workers’ salaries are generally highest late in their career and the highest years are used in the benefit calculation. Use of the average wage of all workers in benefit calculations will significantly understate old age benefit liabilities and cash flows. It is likely not as serious a distortion for disability and survivor benefits as these benefits normally commence at younger ages, but even for these benefits, the probability of death or disablement increases with age  Number of contributors and their age/sex composition. The report states that the number of contributors is assumed to be 8% of the population of working age in 2015, and that this percentage is assumed to grow by 0.6% per year due to economic development. Using the extensive statistics available in “Volume 12- Analytical Report on Labor Force” to more accurately determine the eligible workers by age and sex at the start of the analysis period produces a very different result. Analysis shows that 16% of the working age population is eligible to participate – double the rate shown in the ACL – and that the percentage varies substantially by age and sex. The difference in the percent of population covered will likely not significantly impact the estimated cost of the program as a percent of payroll since only those who contribute receive benefits. However, the ACL report implies that assumed employment rates as a percent of the population do not vary by age and sex. If so, this could significantly misstate expected plan cash flows and liabilities  Program Expenses. The assumption of no expenses is not reasonable. Under the proposal, a separate pension fund organization will need to be established, and it will need to collect contributions on a monthly basis from thousands of employers and workers scattered across Timor-Leste. These administrative costs will be significant and should not be ignored in the analysis  Contribution collection efficiency. The projections assume 100% contribution collection efficiency. This is highly unlikely, particularly in the early years of the program with a new pension fund. Even mature pension fund organizations administering plans that have been in existence for many years struggle to collect contributions from small employers and employers in remote locations. Appendix 1 to this report contains a list of Standards of Practice and actuarial guidelines for the preparation of actuarial valuations for national social security programs. In our opinion, the methods and assumptions used in this valuation do not meet those standards, and should therefore be adjusted. Appendix 2 contains a summary of the assumptions we used in our cost estimates and our understanding of the assumptions used by the actuary who prepared the ACL report. III. ACL Plan Design Comments The ACL was prepared based on an assumed plan design that is not contained in the law. The final plan design will be set by a law-decree within six months following enactment of the contribution social security regime. We have valued this plan design, but wish to note here that the design is non- standard in many respects and deviates from typical national social security system design in many important ways. We commented on many of these same issues in relation to the temporary regime for civil servants in our June 30, 2013 report entitled “Creation of a Reformed Pension System for Civil Servants in Timor-Leste”.  Target replacement ratio for old age benefits. No country that we are aware of has a target replacement ratio in its national social security system of 100% of final average pensionable earnings. It should be noted that ILO Convention 102 calls for a target replacement ratio of 40% and many countries find they are unable to afford even this level of benefit. The cost of the proposed program would be considered prohibitive in almost all countries, it would likely provide for a higher standard of living following retirement than the participant enjoyed during his/her working career, and it would provide no incentives for workers to save for their own retirement. It would also impose a high cost burden on employers and workers. For larger employers, it would significantly increase payroll costs and will likely reduce hiring of new workers. For small employers, the additional contribution burden may be high enough to cause bankruptcy or business shutdown. For workers, it will mean a significant reduction in their take- home pay. Evasion is also likely to be a serious issue in a system with such high contribution requirements.  Disability benefits. Typically disability benefits are expressed as a flat percent of salary regardless of age and years of contributions, or as a percentage of the projected benefit the participant would have received had he or she worked to the standard retirement age. The proposed design uses the accrued benefit at time of disablement, which often provides an inadequate benefit for a worker who becomes disabled at a young age. Generally, the target replacement ratio for disability is below the target for old age benefits. Otherwise it creates moral hazard as older workers will likely try to use the disability benefit provisions to retire prior to the standard retirement age. The disability benefit should be low enough to continue to provide an incentive to work.  Pre-retirement survivor benefits. The survivor benefit has an unusual design. The benefit payment period for the spouse can vary from as little as one year to life depending on the spouse's age and the age of any children when the participant dies, the benefit amount does not vary with the number of beneficiaries, and the benefit payment period depends only on the age of the youngest child. The benefit also does not reduce as each child attains age 17. The adequacy of the survivor benefit will also vary significantly with the age of the participant at death. If the participant dies young, the accrued benefit may be low, but it will likely be paid for a longer period of time as the participant is more likely to have young children. On the other hand, if the participant is older when he or she dies, the benefit may be high, but the period of payment will likely be quite short. Standard practice is to pay survivor benefits that are a percentage of salary at time of death or a percentage of the projected benefit the participant would have received had he or she worked to the standard retirement age. Benefits generally are independent of the number of years of contributions, vary with the number of eligible beneficiaries and reduce whenever a beneficiary becomes ineligible. Payment periods vary by country and culture. In countries like Timor-Leste, where large families are common and wives often do not work, survivor benefits are often paid to the spouse until the earlier of death or remarriage. Benefits to children are often extended beyond 17 if the child is still a full-time student to allow for completion of education. IV. PROST Analysis: ACL Report Design, Assumptions and Methods The information provided in the ACL Report and in the referenced data sources were not sufficient to completely reproduce the results in the report. We would need to speak with the author to get more information in order to replicate his or her results more precisely. Nonetheless, we were able to produce results that were slightly less favorable, but showed the same general pattern of costs as those shown in the report. In our analysis, using the plan design, data, methods and assumptions in the report, 15% was a sufficient contribution rate for the first 30 years of the new program. However, this rate is sufficient only because the program is new, the projection period is extremely short, there is no credit given for service prior to program establishment, retirement benefits begin only in 2025 and are only a fraction of the full target replacement ratio, and no benefits are paid to those currently over age 60 (the plan’s retirement age). We believe the finding in the ACL is misleading and significantly underestimates the true long-term cost of this proposed program, for the reasons outlined in Section II of this memo. International standards of practice for actuarial reports require full disclosure of these limitations and estimates of the true long term cost of the program. The time frame used for the analysis is too short and covers a particularly favorable time period. As previously discussed, we also believe many of the assumptions and methods used are inconsistent with international best practice, and the plan provides benefits which are too high and significantly out of line with international experience and plans for countries similar to Timor-Leste. If the ACL assumptions and methods and the proposed plan design are used (which we do not recommend), the table below shows the required contribution rate to balance the fund over various time periods. Please note that in preparing this analysis, it would not be reasonable to keep the assumptions in the ACL report for 2045 the same through 2100. For example, the 2045 fertility and mortality rates would clearly be too high for the year 2100, so we had to assume they continued to drop during this time period. Consequently, we needed to make some reasonable adjustments to the ACL report assumptions in order to prepare projections beyond 2045, but we did our best to follow the logic of the ACL valuation report. However, these assumptions may not be the same as the ones the ACL actuary would have selected if he or she were preparing analysis through the year 2100. As noted in Section 2, we believe an analysis period of at least 75 years is needed to give an accurate measure of the true long-term cost of the plan and for this plan, we would recommend an even longer time period. Table 1: Level Required Contribution Rates Over Various Time Periods, ACL Actuary’s Assumption with World Bank Extension to 2100 Required Analysis Period Contribution Rate 2015-2045 6.9% 2015-2050 9.0% 2015-2080 3.7% 2015-2100 33.7% As can be seen from Table 1, based on the assumptions used in the ACL Report (with our assumptions for the period beyond 2045), the cost of the proposed program as a level percent of payroll over the time period from 2015 to 2045 is likely to be less than 15% of covered payroll. We estimate the level cost over this time period will be about 6.9%. However, the estimated cost of the program through 2100, even using the ACL Reports methods and assumptions, is 33.7%. V. PROST Analysis: ACL Report Design, Bank Assumptions and Methods We have analyzed the expected cost of the proposed plan design using the World Bank’s PROST model, and using methods, assumptions and a projection period which we believe are reasonable and appropriate for this analysis, and take into account international guidelines. These assumptions are subject to revision based on further discussion with the government. However, we would not expect any revisions to substantially alter the results. If anything, costs are more likely to increase than decrease. We have extended the analysis period to 2100, as this is more consistent with international practice and is necessary for a new plan. It is particularly important to use a long time period for this plan, because the covered group and population are young no credit is given for periods of work prior to the start of the national social security system and no benefits are provided to those who are currently over the plan’s retirement age (age 60). Consequently, it will take 10 years before the first old age retirements, 30 years before the first participants will be able to retire with full benefits and close to 100 years for the system to fully mature. It is also necessary to use longer time periods in order to reflect the changes in demographics that will occur as a result of the gradual decrease in Timor-Leste’s birth rate and the gradual increase in life expectancy. Over time, this will cause Timor-Leste’s population to age and this will have a significant impact on plan costs. Based on our analysis, our best estimate of the likely long-term cost of this program is 31.2% of covered payroll. Figures 1 through 7 briefly summarize the most key results of our analysis and Appendix 2 lists the assumptions used for our analysis and a comparison with our understanding of the assumptions used in the ACL report. A more comprehensive actuarial valuation report can be prepared at the direction of the government of Timor-Leste. Figure 1 illustrates that the population of Timor-Leste will not begin to age significantly until about 2050. After that time, the cumulative effects of reduced fertility and increased life expectancy begin to have a significant impact. For example, in 2040, the percent of the population over age 60 is 6.7%, but by 2100 it is 24.1%. Figure 1: Population Composition 120.0% 100.0% 80.0% Ret. Age + 60.0% 15-Ret. Age 40.0% 0-14 20.0% .0% 2012 2013 2014 2015 2020 2025 2030 2040 2050 2060 2070 2080 2090 2100 Figure 2 illustrates population aging in a different way, by showing various population dependency ratios. The old age dependency ratio is the ratio of those 60+ to those of working age (15-59). As can be seen, this ratio decreases until about 2045 and then increases sharply. The child ratio is the ratio of children (ages 0-14) to those of working age (15-59). Due to declining fertility rates, this ratio decreases slowly until about 2030, and then decreases more rapidly thereafter. The support ratio is the sum of the old age and child ratios and shows how many citizens there are of non-working age for every citizen of working age. Today, there is almost one non-worker for every worker. This ratio declines through about 2070 and then begins to increase. A long projection period is needed, therefore, to show the impact of these demographic changes on the proposed new national social security system. Figure 2: Population Dependency Ratios 120.0 100.0 80.0 Old Age 60.0 Child 40.0 Support 20.0 Figure 3 shows the number of expected contributors and beneficiaries to the proposed national social security system. The number of contributors grows steadily throughout the analysis period due to an increase in the size of the population and the number of salaried workers. Lacking additional information, we assumed the percent of the population by age/sex that are salaried workers remains the same throughout the analysis period. The number of beneficiaries start out quite low as no one is eligible for an old age pension until 2025. Prior to that date, there are only a handful of disability and survivor pensioners. However, after 2025, the number of beneficiaries begins to grow more rapidly as the number of old age pensioners accumulate. The green line shows the plan dependency ratio (ratio of beneficiaries to contributors) grows steadily throughout the analysis period and would continue to grow beyond 2100. Figure 3: Contributors and Beneficiaries Contributors and Beneficiaries 500.0 80.0 450.0 70.0 400.0 60.0 350.0 300.0 50.0 Active Civil Servants 250.0 40.0 200.0 Beneficiaries 30.0 150.0 Dependency ratio 20.0 100.0 50.0 10.0 Figure 4 illustrates how the number of beneficiaries varies by type. There are three types of beneficiaries under the proposed plan design – old age pensioners, disability pensioners, and survivors (beneficiaries of those who die prior to retirement, and beneficiaries of those who die after retirement age). Figure 4 shows that in the long run, the vast majority of beneficiaries are old age pensioners, but in the early years of the program, disability and survivor pensioners are much more prevelant. Figure 4: Beneficiaries by Type Beneficiaries by Type 350.0 300.0 250.0 200.0 Survivor 150.0 Disability 100.0 Old Age 50.0 2015 2020 2025 2030 2040 2050 2060 2070 2080 2090 2100 Figure 5 shows the estimated plan expenditures expressed as a percent of covered payroll in each year. This is important because it illustrates the required contribution rate in each future year necessary to exactly cover expected plan benefit payments, administrative expenses, and investment expenses. This is often referred to as the pay-as-you-go cost of the plan. In the early years of the plan, costs are very low as there are only a few beneficiaries, and even those beneficaries receive relatively low benefits, since benefits are based only on years of contributions following plan establishment. However, expenditures as a percent of payroll grow each and every year and reach almost 50% of payroll by 2100. The high expenditures are due to the growing plan dependency ratio, as illustrated in Figure 3, population aging as illustrated in Figures 1 and 2, and the increasing replacement ratio of successive cohorts of beneficairies due to plan maturity. Figure 5: Expenditures as a Percent of Covered Payroll Expenditures as % of Covered Payroll 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 2015 2020 2025 2030 2040 2050 2060 2070 2080 2090 2100 Figure 6 shows the estimated level contribution rate to exactly balance plan contributions and expenditures over the specified time period. For example, to balance the plan over the period from 2015 through 2040, a level contribution rate of 6.7% would be required. Of course, in 2041, the required contribution rate would jump dramatically and would continue to increase thereafter, following the pattern illustrated in Figure 5. To balance the plan over the entire analysis period through 2100, a contribution rate of 31.2% would be required. Figure 6: Required Contribution Rate Over Various Time Horizons Required Contribution as % of Payroll 2100 31.2% 27.2% 2080 22.9% 18.5% 2060 14.0% 9.5% 2040 5.6% 2.4% 2025 1.1% .3% 2015 .0% 5.0% 10.0% 15.0% 20.0% 25.0% 30.0% 35.0% Finally, Figure 7 illustrates the patter of accumulated reserves, assuming a 15% contribution rate and a 2.5% real rate of return on program assets. This Figure illustrates that plan assets would accumulate to more than 25% of GDP by 2030 and then would be fully exhausted by 2063. Figure 7: Pension Fund Assets as % of GDP, 2.5% Real Rate of Return, 15% Contribution Rate Pension Fund Assets as % of GDP 15% Contribution Rate 30.0% 25.0% 20.0% 15.0% 10.0% 5.0% 2015 2020 2025 2030 2040 2050 2060 2070 2080 2090 2100 In analyzing these results, it is important to keep in mind the distinction between cost and how those costs are financed. Figure 5 illustrates the true cost of the plan as a percent of payroll. The cost is the estimated total expenditures in each future year, generally expressed as a percent of covered payroll. This is the percent of covered payroll that would need to be collected each year from employers and workers to exactly covered estimated plan benefit payments and administrative expenses. For any given pattern of costs, there are multiple ways that the government could choose to finance those costs. The only requirement for the financing methods is that the present value of the expected contributions collected over the time period is equal to or greater than the present value of the expected expenditures. We have used the level contribution as a percent of payroll to as a proxy for the cost of the program over time, because it facilitates an accurate way of comparing the relative cost of different program designs. In reality, the government could select a different method of actually financing the costs of the program over time. The government should also not underestimate the challenge of protecting fund assets in the event it chooses to finance the program with a level percent of contributions that is significantly higher than the actual initial cost of the program. In this instance, as illustrated in Figure 8, the accumulated assets may be a very large percent of GDP. If these assets are not carefully protected and invested, these assets may not be available to meet plans liabilities in the future when they are needed. For example, assets in pension funds may be lost due to theft, fraud or corruption, may be invested poorly, or may be used to justify unaffordable benefit increases. Consequently, careful consideration of all political and technical risks should be considered when determining the financing strategy. The government of Timor-Leste also asked the Bank to estimate the replacement ratio for a program with costs of 15% and 6.75%. With this particular proposed plan design, the percent reduction in the required contribution rate and the percent reduction in the target replacement ratio would be proportional. This is because the proposed plan design contains no minimum benefit, no maximum benefit (other than the 100% replacement ratio) and contains no cap on the amount of wages used to calculate contributions or benefits. If any of these features existed, then results would not be proportional. For this particular program:  If the desired long term contribution rate is 15%, the the target replacement ratio would need to be reduced from 100% to 48.1%  If the desired long term contribution rate is 6.75%, then the target replacement ratio would need to be reduced from 100% to 21.6% VI. Social Security Pension Fund Establishment The government of Timor-Leste asked for international experience regarding the establishment of separate statutory social security bodies to administer national social security programs. The focus of our analysis is on countries with defined benefit national social security programs financed by payroll contributions. Defined contribution programs always have either a State-run pension fund or use private pension funds. Social pensions are tax-financed pensions and these pensions are generally paid directly from the budget, particularly if the country does not have a contributory defined benefit plan as well. For purposes of this analysis, a social security administrator is defined as a separate statutory body established by law that is responsible for some or all aspects of social security program administration. Typically, the institution has its own governing body and its own field offices. The administrator’s typical functions could include registration of employers and workers, contribution collection, management of fund assets, approval of benefit applications, calculation of benefit amounts, maintenance of data history, reporting and disclosure to the government and participants, and payment of benefits. Not all administrators perform all functions. For example, in many countries the tax authorities collect the payroll contributions and remit them to the administrator. We used information from “Social Security Programs Throughout the World”, published jointly by the International Social Security Institute (ISSI) and the United States Social Security Administration for this research. This publication consists of four volumes – one each for Europe, Asia and the Pacific, Africa and the Americas. It contains information about the design and administration of social security programs for almost all countries in the world and is updated every two years. The most recent publications are from 2012 for Europe and Asia, and from 2013 for the Americas and Africa. Based on our analysis, the following countries have defined benefit national social security programs that are directly administered by the government without the establishment of a separate social security administrator. Asia Europe Africa Americas Bangladesh Guernsey Botswana Bermuda Brunei Ireland Gambia Jamaica Isle of Man Lesotho Jersey Asia Europe Africa Americas Malta United Kingdom This list contains a mix of smaller and larger countries. However, most of the countries that opt not to use a pension fund either have only a social pension or are smaller countries. VII. Conclusion In our opinion, the ACL report does not meet international standards for actuarial reports for national social insurance systems and misrepresents the long-term costs of the proposed Timor-Leste national social security system by using non-standard assumptions and methods and an unusually short analysis period. We also believe the target replacement ratios proposed for this system are significantly higher than international standards and will produce required contribution rates that are unaffordable for employers and workers and will impeded the growth of formal sector employment. Other aspects of the plan design, such as disability and survivor benefits, are also non-standard. We suggest the government of Timor-Leste carefully consider all aspects of design and funding before adopting the new system, as the decisions made will have very significant macroeconomic, labor market, budget, and social policy implications for the country for decades. Appendix 1: Standards of Practice for Social Security Plans  International Standard of Actuarial Practice 2: Financial Analysis of Social Security Programs, Approved by the IAA Council, 13 October 2013  Actuarial Standard of Practice No. 32: Social Insurance, Developed by the Committee on Social Insurance of the American Academy of Actuaries, Adopted by the Actuarial Standards Board January 1998, Updated for Deviation Language Effective May 1, 2011  Code of Professional Conduct, adopted by the Board of Directors of the American Academy of Actuaries on September 28, 2000  QUANTITATIVE METHODS IN SOCIAL PROTECTION SERIES, Actuarial practice in social security, Pierre Plamondon, Anne Drouin, Gylles Binet, Michael Cichon, Warren R. McGillivray, Michel Bédard, Hernando Perez-Montas, A joint technical publication of the International Labour Office (ILO) and the International Social Security Association (ISSA) isap2.pdf asop032_149 Social Actuarial Actuarial Practice in Insurance.pdf code_of_conduct.pdf Social Security ILO ISSA.pdf Appendix 2: Actuarial Assumption and Methods ACL Actuary World Bank Actuary Projection period 2010 – 2045 2010 - 2100 Social security start date 2015 2015 Population Projections TL data on population, fertility 2010 Census starting population, and life tables from 2010 Census UN projections of mortality and fertility from 2010-2080 Labor Force Projections 8% of working age population Labor force participation rates, (15-59), increasing by 0.6% per unemployment rates and year for shift to formal sector number of employees by age and sex from 2010 Census. No assumed shift to formal sector Average wage 230 USD in 2010 230 USD in 2012 based on information regarding pay of civil servants in December 2011 Wage distribution by wage None From December 2011 bands information regarding pay of civil servants Average wage variation by age None From December 2011 and sex information regarding pay of civil servants Wage cap None None Inflation 0% Projections through 2050 from Ministry of Finance and global averages thereafter GDP and GDP growth rates Not projected Based on Ministry of Finance projections of oil and non-oil GDP growth rates. Adjusted after 2030 to keep wage fund of contributors a constant percentage of GDP Productivity growth 0% 3.5% through 2016 and 2.5% thereafter based on discussions with Ministry of Finance and Civil Service Commission Investment earnings on program 0% and 1% Real interest rate of 2.5% reserves Retirement rates 100% at age 60 100% at age 60 Disability rates Based on Tables of Entrance into Based on analysis of disability Disability Situation, designated incidence and prevalence rates ACL Actuary World Bank Actuary as “EVK 1960” in comparable countries from World Bank data base and studies. Rates assumed to decrease by 10% from 2015- 2045 and an additional 10% from 2046-2080 Pension indexing None, though implicitly, 100% of 100% of inflation inflation Benefit accrual rate 1/30 per year of contributions 1/30 per year of contributions Minimum benefit None None Maximum benefit 100% replacement ratio 100% replacement ratio Collection rate 100% 90% Administrative expenses 0% 2% of estimated benefit payments each year