Republic of Djibouti Pension System Reform
Transcription
Republic of Djibouti Pension System Reform
Report No. 22087-D3i Republic of Djibouti Pension System Reform Strategic Note December 2001 MidcIle EastancdNorth Africa Region Human Development Sector Document of the World Bank REPUBLIC OF DJIBOUTI Currency equivalents as of November 2000 Currency Unit - Djiboutian Franc (DF) US$1.00=DF 177.7 Main economic indicators Gross Domestic Product (1999) = DF 94.4 Billion Population (1999)= 653,130 ACRONYMSANDABBREVIATIONS CMR: Caisse Militaire de Retraite (Military Pension Fund) CNR: Caisse Nationale de Retraite (National Pensions Fund) GNA: Groupement Nomade Autonome (Autonomous Nomad Association) MOFP: Ministry of Finance in Charge of Privatization MLNS: Ministry of Labor and National Solidarity NCSS: National Council on Social Security OPS: Organisme de Protection Sociale (Social Protection Organism) Vice President: Country Director: Sector Director: Task Team Leader: Jean-Louis Sarbib Mahmood Ayub Jacques Baudouy David Robalino ACKNOWLEGMENT This document was prepared by David A. Robalino (MNSHD). The report benefited from comments and inputs by Jacques Baudouy (Director MNSHD), Pedro Alba (Sector Manager MNSED), Ide Gnandou, Kouassi Soman, Mark Dorfman, Anita Schwarz, Aniruddha Bonnerjee, Osman Ahmed, Denis Chemilier and Allexandre Collard (IEM-ACTUARIA), Dimitri Vittas (Pensions Advisor), and was supervised at different stages by Zafiris Tzannatos (Adviser to the Managing Director Human Development) and George Schieber (Sector Manager SP and Health, MNSHD). The main simulationspresented in the paper were conducted using the PROST model. Eduard Bos generated the necessary demographic projections and Montserrat Pallares and Alexandre Collard prepared the baseline data. The document has been prepared at the request of the Govemment of Djibouti and will guide the pension reform program within the context of the Structural Adjustment Credit. Special thanks to Djiboutian authorities for their support and active collaboration. Particularly, His Excellency Yacine Elmi Bouh, Minister of Economy and Finance and His Excellency Ohamed Barkad Abdillahi, Minister of Labor and National Solidarity. 3 Table of Contents EXECUTIVE SUMMARY ........................................................ 6 I. INTRODUCTION ......................................................... 8 II. CURRENT SITUATION OF THE DJIBOUTIANPENSION SYSTEM ..........................................8 1. GENERAL CHARACTERISTICS ................................... 2. SHORTANDMEDIUM-TERM FINANCIAL SITUATION .................................... ............. 8..................................... 8 12 3. INSTITUTIONAL CONSTRAINTS .......................................................................................... 4. THE PROBLEMOFARREARS......................................... .. 17 18 III. RECOMMENDATIONSFOR POLICY REFORM......................................... 19 19 9............. 1. SHORT-TERM REFORM PROGRAM ....................... 1.1. Improvinggovernanceand management: preparingthe mergerof OPSand CNR.19 1.2. Improving institutional capacity.20 1.3. Improvingfinancial sustainability.22 Options for parametric reform in the case of OPS.24 Options for parametric reform in the case of CNR.27 Options for parametric reform in the case of CMR.28 . 2............... 29 2. MEDIUM-TERMREFORMPROGRAM .............................................................................. 2.1 Consolidating the merger of OPS and CNR.............. 2.2 Promoting contractual savings .............. 2.3. Towards a thirdpillar .............. 29 29 29 IV. SEQUENCING AND MANAGING REFORM ACTIVITIES .31 V. POLICY MATRIX.33 VI. TECHNICAL APPENDIX .36 Tables Table 1: Demographic Composition of the Pension Funds Table 2: International variation of dependencyratios .10 Table 3: Basic structure of obligations and benefits.11 Table 4: International comparison of basic pension systems parameters.12 Table 5: Financial situation of Djibouti's Pension Funds.13 Table 6: Government arrears with OPS and CNR.18 Table 7: OPS financial impact of alternative parametric refors .26 Table 8: CNR financial impact of a reduction in the accrual rate.27 Table 9: Costs of implementing institutionalreforms.32 .9 Figures Figure 1: OPS evolution of new retirees.14 Figure 2: Deficits in OPS under alternative scenarios.14 Figure 3: CNR Age distribution of contributors and new retirements.16 Figure 4: CNR projected deficits.16 Figure 5: CMR projections of total wages and pensions (DF billion) .17 Figure 6: Organizational structure of the NCSS.20 Figure 7: The core management and information system.21 Figure 8: Fiscal cost of wages and contributionsto OPS and CNR.23 Figure 9: OPS: Current balances under alternative parametric reforms.26 Figure 10: CNR balance under alternative parametric reforms .. Figure 11: CMR total wage and pension expenditures under alternative reforms Figure 12: Evolution of age pyramids and the dependency ratio in Djibouti . 28 . .28 .30 4 Definitions Annual Accrual rate. The marginal increase in the replacement rate per year of contribution. The accrual rate times the number of years of contribution equals the minimum replacement rate (see ReplacementRate). Old age dependency ratio. In this document, this ratio is defined as the number of persons over 65 divided by the number of persons aged 15-64. Pay-as-you-go. Mechanism which finances current pensions on the basis of current revenues resulting from a payroll tax or an earmarked tax. Replacement rate. The value of a pension as a proportion of a worker's wage during some base period (e.g., last year or the average of the last 10 years). It also denotes the average pension of group as share of the average wage of the group. System dependency ratio. The ratio of persons receiving pensions from a certain pension scheme divided by the number of workers contributing to the same scheme during the same period. Vesting period: The minimum amount of time required to qualify for full ownership of pension benefits. 5 EXECUTIVE SUMMARY The Government of Djibouti has committed to reform its pension system in order to restore financial sustainability and improve management. To this end, it requested the World Bank to elaborate a Pension Reform Strategy that identifies the major financial and institutional constraints facing the pension funds and explores restructuring options. This policy note summarizes the major diagnosis and policy recommendations of the study. Current situation The Djiboutian pension system was set up before independence according to the French pay-asyou-go model. It currently comprises three pension funds: i) the Caisse Nationale de Retraite (CNR) that covers civil servants, police, ministers and parliamentarians and the ancient combatants of GNA (Autonomous Nomad Association); ii) the Organisme de Protection Sociale (OPS) that covers private sector employees, non-regular civil servants, and public enterprises]; and iii) the Caisse Militaire de Retraite (CMR) designed for the military personnel. The system covers 8% of the population older than 55 and 15% of the labor force. Since Djibouti's population remains relatively young, the overall system dependency ratio (21.4%) is low compared to other countries in the region. Today the funds face two major problems: - institutional capacity is weak; * !systemsare notfinancially sustainable Institutional capacity. A recent audit of current information systems and administrative and management practices within the funds identifies problems at three levels: i) administrative processes are inefficient and lead to the duplication of efforts and high administrative costs; ii) integrated and updated information systems are lacking; and iii) management and analytical capacity are weak and hamper the implementationof any reform program. Financial situation. All pension funds have an operational deficit and reserves have been depleted. The reasons behind the weak financial position of the funds are diverse. In the case of OPS, the problem is more in terms of a low collection rate for private sector and SOE workers, and to the Government's moratoria in the payment of contributions. In CNR, on the other hand, the problem results from high dependency ratios and overgenerous benefits. The financial crisis is also explained by the non-compliance of the Government to repay loans contracted with the pension funds. Preliminary estimates suggest that Government's arrears with the pension funds could amount to DF 7.1 billion or 8% of GDP. OPS. The deficit of OPS (including administrative costs) observed in year 2000 approximated DF 197.6 million or 0.2% GDP. These numbers do not take into account the 773 contractual workers who have been forced to delay retirement as a result of the Government's moratoria. Their pensions are estimated at DF 202.7 million per year (an average pension of DF 261,669 per year), while their late contributions for the period 1995-2000 amount to DF 553.6 million. OPS is currently covering its deficit with the surpluses of the health and family services branches which approximate DF 173.8 million and DF 241.7 million respectively. Actuarial projections of the pension branch show that under the status-quo, the present value of the deficit accumulated during the period 2001-2015 could approximate 4-6% of GDP, depending on the assumptions about coverage and economic growth. CNR. Only the Police regime is currently generating a budget surplus; both the regime for functionaries and parliamentarians have a deficit of DF 37.8 million (0.04% of GDP) and DF 26 'OPS has alsoa healthbranchanda familyservicesbranch. 6 million (0.03% GDP) respectively. Part of the financial problems of CNR are due to the current practice of parliamentarians and ministers to "accumulate" benefits simultaneously in their regime and in the regime for functionaries. This leads to replacement rates of over 100%. Projections of expenditures and revenues conducted in the case of the regime for functionaries show that its deficit would continue to increase reaching DF 124 million in 2005. The present value of accumulated deficits for the period 2001-2015 could approximate 1.5% of current GDP. CMR. This fund has an operational deficit of DF 271 million that is being financed by the Government. Thus, total transfers from the central budget (including contributions from the Government and military personnel) amount to DF 675 million (0.7% of current GDP). Given a small number of contributors per beneficiary, the financial crisis of the fund is severe. Actuarial projections suggest that pension expenditures would continue to increase and would have doubled by year 2010. The present value of accumulated deficits between 2001 and 2015 could reach 0.7% of GDP. Policy recommendations The problems facing the pension funds in Djibouti are by no means particular to the country. Similar issues are observed in other parts of the world. In fact, pension systems reform is likely to be one of the most important elements in the development agenda of governments in the years to come. In the case of Djibouti, a two-stages reform program is proposed. A first stage, which is the focus of the current Structural Adjustment Credit operation, would concentrate on consolidating a modest pay-as-you go system. This involves executing three major activities: - implementing institutional reforms to improve governance and management (through the creation of the National Council for Social Security, NCSS, and the preparation of a multi-year plan to merge CNR and OPS), and to generate efficiency gains that enhance service delivery and reduce administrative costs (by streamlining administrative processes, improving the allocation of human resources, and implementing an integrated management and information system); * addressing short-term financial needs by normalizing government contributions, defining a plan to refinance government arrears, and introducing a ceiling for the replacement rate in the regime for parliamentarians; * and improving the medium-term financial situation of the pension funds by aligning contributions and benefits while introducing a minimum pension to protect low-income workers. The NCSS would be jointly attached to the Ministry of Finance and the Ministry of Labor and National Solidarity. The NCSS would have authority over the two pension funds CNR and OPS, and in a period of 3 to 5 years it would constitute the new social security system that integrates OPSand CNR A second stage of the reform program would focus on: * consolidating the merger of OPS and CNR; * and introducing/reinforcingincentives to promote contractual savings. The latter would pave the way for a possible third stage that would focus on addressing long-terrn challenges for the current pay-as-you-go system. These include expanding coverage and addressing the financial problems associated with the aging of the population through the gradual introduction of a mandatory funded pillar with individual accounts. 7 I. INTRODUCTION 1. The Government of Djibouti has committed to reform its pension system in order to restore financial sustainability and improve management. The reform has been introduced as a key component of the Government's Structural Adjustment Program aimed at restoring fiscal discipline, promoting competitiveness, reviving growth and strengthening social protection systems. 2. The Government has requested the World Bank to elaborate a Pension Reform Strategy that identifies the major financial and institutional constraints facing the pension funds and explores restrucluring options. To this end, a World Bank team, in collaboration with the consulting firm IEM-ACTUARIA contracted by the Government to advise in the reform process, conducted a detailed study of the pension funds. The team has worked closely with the managers and technicians of the different funds and has benefited enormously from their institutional knowledge. 3. This policy note summarizes the major diagnosis and policy recommendations of the study, reflecting commentsfrom government officials and the Bank's technical experts during the internal review meeting held on March 15, 2001. As new data has become available, financial projections have been updated from previous versions of the note. 4. The note is organized in three sections. Section 2 is concerned with a diagnosis of major institutional and financial constraints. Section 3 presents reform alternatives and Section 4 discusses implementation issues. II. CURRENT SITUATION OF THE DJIBOUTIAN PENSION SYSTEM 1. General characteristics 5. The Djiboutian pension system was set up before independence according to the French pay-as-you-go model. It currently comprises three pension funds: i) the Caisse Nationale de Retraite (CNR) that covers civil servants, police, ministers and parliamentarians and the ancients combatants of GNA (Autonomous Nomad Association); ii) the Organisme de Protection Sociale (OPS) that covers private sector employees, non-regular civil servants, and public enterprises2 ; and iii) the Caisse Militaire de Retraite (CMR) designed for the military personnel. 6. The system covers 8% of the population older than 55 and 15% of the labor force, a relatively high percentage by international standards, given the country's level of income. As shown in Table 1, there are close to 35,000 contributors who represent 35% of formally employed workers or 15% of the labor force. This level of coverage is expected in countries with a per-capita income of US$2,5003 , more than two times the current per capita income of Djibouti (close to US$1,000). In terms of beneficiaries, 8% of the population older than 55 years of age are receiving a pension from one of the three funds. 2 OPShas also a healthbranchanda familyservicesbranch,financedrespectivelyby a 6 and 5.5%tax on wages. 3 SeePalaciosR. 1994. "International Patternsof PublicPensionCoverage."Workingpaper.WorldBank PolicyResearchDepartment.WashingtonDC. 8 Table 1: Demographic Composition of the Pension Funds Contractual Contributors 4,122 Average wage (DF)' 658,975 Total beneficiaries 983 523 Old age pensioners Survivors 460 Average old age pension (DF) 290,208 Average pension survivors(DF) 145,004 Old age dependency ratio 12.7% Overalldependency ratio 23.8% Functionaries 2,866 Average wage(DF) 1,445,851 rotal beneficiaries 1,190 Old age pensioners 546 Survivors 644 1,078,680 Average-old-age pension(DF) Average pension survivors(DF) 301,796 Old-age dependency ratio 19.05% Overalldependency ratio 41.5% Contributors OPS Public enterprises Private sector 4,834 19,338 1,190,256 987 1,268 653 914 334 354 469,652 611,062 241,352 255,432 13.5% 4.7% 20.4% 6.6% CNR FNP Parliamentarians 74 1,417 3,629,455 944,561 140 855 74 439 66 416 1,174,248 413,696 460,803 170,262 100% 30.9% 189.2% 60.3% CMR Military Total 1,758 2,381 Total 28,294 1,112,857 3,238 2,090 1,148 486,589 207,087 7.4% 11.4% GNA 0 Total 4,357 660 256 404 279,137 132,659 2,845 1,315 1,530 455,770 228,230 30.1% 65.3% n.a. Gendarmerie 623 Contributors I '002'442 Average wage(DF) I1,419 ITotalbeneficiareis 476,020 Averagepensions(DF) 59.6%/ Overalldependency ratio Totalpopulation 653,000 Totalcovered 42,534 6.51% Totalold-agebeneficiaries 4,405 7.7% Population>=55 57,129 35,032 15.30/o Laborforce 228,900 Totalcontributors a: Pensionsand averagewagesareexpressedin DjiboutianFrancsper year. b: Thereis no informationabouttheaveragewagein SOEsseparatedfromthe averagewagein the privatesector. Source:Individualrecordsof OPS,CNRandCMR,andMinistryof Finance. 9 Table 2: Internationalvariation of dependencyratios Country Pensioners/ Population 60+/ Pensioners/ Contributors Population 20-59 Total Population (percenta,ge) Selected countries in Africa Benin Burkina Faso 10.0 9.5 Burundi Djibouti Nigeria 23.4 21.4 0.4 Selected countries in Middle East and North Africa 34.0 Egypt Iran 21.7 31.0 Israel Jordan 29.7 Turisia 20.4 Turkey 50.3 46.9 OECD 11.5 12.5 11.4 11.5 10.2 0.2 0.1 0.4 1.0 0.0 14.6 6.1 15.4 21.7 10.7 14.7 14.1 34.4 1.9 10.1 3.3 2.9 3.4 19.7 Patternsof PensionProvision"by PalaciosandPallares-Miralles,1999. Source: "International 7. Djibouti's population remains relatively young and the overall system dependency ratio (21.4%) is low compared to other countries in the region. In Djibouti the ratio between the population over 60 and the population aged 20-59 is close to 11.5%, more or less in line with other countries in the region (see Table 2). On the other hand, the ratio between pensioners and contributors from all systems (21.4%) is relatively low. Nonetheless, as discussed below, the majority of contributors belong to OPS and therefore dependency ratios in CNR and CMR are above the average. Moreover, in OPS a low collection rate increases the effective dependency ratio. 8. By far, the largest system is OPS with close to 28,300 contributors and an 11.4% nominal dependency ratio (9 workers per beneficiary). Contributors in OPS are contractual workers hired by the central administration (15%), workers from public enterprises (17%), and workers from the private sector (68%). Not all these workers, however, are effectively contributing. In the case of contractual workers, while the Government retains their contributions, these are not remitted to OPS. As a consequence, retirements of contractual workers have been suspended. For workers in public enterprises and the private sector the collection rate is estimated at a low 46%. Thus, the effective collection rate for the overall system is only 42% and as a result the effective dependency ratio increases to 28.9%. 9. CNR is a relatively small fund in terms of the number of contributors and has a high dependency ratio (65%Y). In CNR there are approximately 4,357 contributors and 2,845 beneficiaries. The majority of contributors, 66%, belong to the regime for functionaries, 32% belong to the Policy regime, and only 2% to the regime for parliamentarians. In terms of beneficiaries, including old age pensioners and survivors, 42% are part of the regime for functionaries, 30% are in the Police regime, 5% are ex-parliamentarians, and 23% are ex-soldiers of GNA. CMR has also a high dependency ratio (60%) given that most militarypersonnel retire at a 10. relatively young age (35-45years). Indeed, the regime has flexible retirement conditions with only a minimum vesting period and no minimum retirement age. Today, there are 2,381 contributors and 1,491 beneficiaries. 10 11. The parameters used to compute contributions and benefits vary widely across and within funds and tend to deviate from international standards. Contribution rates vary from 8% in the case of OPS to 34% in the regime for parliamentarians within CNR (see Table 3). The latter is high when compared to the average contribution rate in Sub-Saharan Africa and the Middle-East and North Africa regions (10%) and even the 16% observed in OECD countries (see Table 4). In terms of the retirement age, it ranges between 40 years (in the case of the Police regime) to 55 years in the case of OPS and the regime for parliamentarians. The average retirement age for countries in Sub-Saharan Africa, where most pension funds are also facing financial difficulties, is 55 years, while the average for countries in the Middle East and North Africa region is 60 years. As for vesting periods, the shorter is 10 years in the case of the regime for parliamentarians, while the longest is 26 years for the Police. OPS has a contribution period of 15 years, which is similar to the average of the Sub-SaharanAfrica and Middle East and North Africa regions. Table 3: Basic structure of obligations and benefits OPS CNR Functionaries Ministers 17% 13% 17% 13% 340/o 10 55 15 Employers'contribution 40/o 6% 12% Total contribution Contribution period Minimum retirement age 8% 15 55 18% 25 50 20% 26 40 3%first10 3% first Workers' contribution Accrual rate per year of contribution 40/ 2 % years, 2/ after 10 years, ____________ CMR Police 7% 20/ after GNA 0 7% 3% first 4% 15, 2% after Reference salary Average 10 Last Last Last Last years Implicit minimum replacement rate 30% 60% 62% 40% 45% Collection rate 42% 100% 100% 100% n.a. 100% Effective overall dependency ratio 28.9%/ 41.5% 60.3% / 89.20/a 59.6% Observed average replacement rate 44% 45.5% 31.3% 23.1% (including survivors) Equilibrium replacement rate for a 27.7% 43.3% 33.10 180 non-funded system(a) Equilibrium replacement rate for a 15% 22% 25% 18%/a funded system (a): Adjustedby theeffectivecollectionrate. Source: Authorscalculationsbasedon infornationprovidedbythe pensionfunds. Methodsusedto deriveequilibriumreplacement ratesare describedin the TechnicalAppendix. 12. In general, observed and implicit average replacement rates are above the level that could be financed with current contribution rates and dependency ratios. In the current pay-as-you-go system, the affordable replacement rate depends on the number of workers supporting each retiree (the dependency ratio) and the contribution rate of each of these workers. In the case of OPS, with the current contribution and effective dependency rates (8 and 28.9% respectively), an average replacement rate of 28% could be financed. While the basic replacement rate (30%) is more or less in line with this level, the observed average replacement rate (close to 44%) is considerably higher, as a majority of workers retire after 18-20 years of work instead of the required 15. In CNR, the observed replacement for the police regime (31.3%) is below the equilibrium (33.1%), but for the functionaries and parliamentarians regimes within CNR, both implicit and observed replacement rates are above equilibrium rates. Not surprisingly, in all funds current contribution rates are below the level that would need to be observed to finance the targeted basic replacement rate in a system with individual capitalization. 11 Table 4: International comparison of basic pension systems parameters Retirement age Vesting period (years) Contribution rate (%) Middle-East and North Africa 59 5 13.1 311. 10.6 Sub Saharan Africa 56 13.8 9.1 Asia 56 10.4 13 Eastem Europe and former Soviet Union 67 25 25.5 and 59 13.9 10.5 63 18.3 16.3 Latin America Caribbean OECD Source: World Bank. 1994. Averting the OldAge Crisis 2. Short and medium-termfinancial situation 13. While little is known about the financial history of the pension funds, the financial situation seems to have started to deteriorate during the 90s; today all pension funds have an operational deficit and reserves have been depleted. The reasons behind the weak financial position of the funds are diverse. In the case of OPS the problem is more in terms of a low collection rate for the private sector and to the Government's moratoria. Indeed, with the current contribution rate (8%) and a nominal dependency ratio of 11.4% the system could afford to pay its retirees an average replacement rate of 70%! In CNR and CMR, on the other hand, the problem results from high dependency ratios and overgenerous benefits. In all cases, the mismanagement of reserves has aggravated the situation. OPS 14. In OPS the deficit (including administrative costs) observed in year 2000 approximated DF 197.6 million or 0.2% GDP. OPS has revenues of DF 1,068 million from contributions. With an average covered wage of DF 1.1 million per year, the system could receive revenues of DF 2.5 billion. However, the Government is not remitting contributions equivalent to DF 217.4 million and evasion in the private sector costs the system an additional DF 1.2 billion. Hence, the effective collection rate is only 42%. On the other hand, expenditures on pensions are close to DF 1.2 billion (DF 218.5 million correspond to retired contractual workers), leaving an operational deficit of DF 197 million (USD 1.1 million) after including administrative costs. These numbers do not take into account the 773 contractual workers who have been forced to delay retirement. Their pensions are estimated at DF 202.8 million per year (an average pension of DF 216,669 per year), while their late contributions for the period 2001-2015 amount to DF 553.6 million. 15. OPS is currently covering its deficit with the surpluses of the health and family services branches that approximate DF 173.8 million and DF 241.7 million respectively; given growing demands for health services (in part due to the HIV/AIDS epidemic), this financing strategy is unsustainable. The health branch (which also covers work accidents) and the family services branch are financed respectively by a 6.2 and 5.5% tax on wages. Thus, OPS receives contributions amounting to DF 2.6 billion, 40% of which concentrate on pensions. Because the health and family services branches generate a surplus, the overall balance of the system has averaged DF 286 million (US$1.6 million) between 1998 and 2000. Despite these relatively large surpluses, OPS has accumulated reserves (non-related to the pension funds) amounting to only DF 150 million. The managers explain that the current stock of reserves is smaller than the flows due 12 to the payment of debts that had accumulated up to 1998 in regimes such as work accidents. These debts were refinanced in 1997 (in that year the system generated a deficit of DF 387 million) and 1998. Moreover, in 1999 a large operation to reduce the number of employees was put in place; 4 close to 180 OPS workers received retrenchment packages4. Table 5: Financial situation of Djibouti's Pension Funds (DF 000) oPS Legal contributions Observed contributions Other income' Benefits paid Other expendituresb) Legal balance Observed balance Contractual 217,304 SOE 218,481 387,294 Private sector 2,301,669 1,068,002 Total 2,518,973 1,068,002 73,702 1,254,708 648,933 84,580 -1,177 -218,481 1,265,442 31,775 1,253,387 -197,584 CNR Observed contributions Other income Benefits paid Observed balance Functionaries Parliamentarians 745.521 91,317 783,316 -37,795 117,307 -25,990 Gendarmerie Military FNP 267,689 GNA 125.(53' 252,441 15,247 125,053 0 Total 1,263,044 36.175 1,296,664 -15,889 CMR Total Legal contributions 404,362 Observed contributions 0 Government subsidies 675,472 Benefits paid 675,472 Legal balance -271,110 Observed balance 0 a: Includespaymentsof latecontributions and revenuefromthe purchaseof earlyretirement. b: Includesreimbursement of contributions to workerswho leavethesystem. c: Thesearegovernmentsubsidies. Source:Contributions have beencomputedon thebasis of the averagewage. Pensionshavebeencomputedfrom individualrecords. Otherexpendituressuchas thoserelatedto thepaymentof disabilitybenefitshavebeenexcludedgivenlackof data. 16. Actuarial projections of the pension branch show that under the status-quo the present value of the deficit accumulated during the period 2001-2015 would reach DF 4 billion, depending on theassumptionsaboutcoverageandeconomicgrowth5. Projections have been conducted under the assumption that retirements are normalized. Thus, in year 2002, 870 workers would retire (including 773 contractual workers). Between 2003 and 2010, close to 250-400 new retirements would take place each year (50-100 from the public sector). After 2010, the number of new At the time of writing this report the documents supporting these transactions were not available. 5 We have evaluated the financial position of OPS in the case of two-scenarios: a low-growth/low-coverage scenario (we refer to this scenario as the low-case) and a high-growth/high-coveragescenario (the high-case). The low-case assumes that the number of covered workers grows very slowly, at 0.5% per year (below the population growth rate of 1.6%), and that GDP expands at a modest 2% per year. The high-case considers that the population of contributors grows at 2% per year (above the population growth rate) and the GDP growth rate converges to 3%. Differences in the GDP growth rate are given by differences in productivity growth (see the Technical Appendix for a more detailed description of the scenarios). 4 13 retirees would start to increase fast, as the generation of 35-45 years old, starts to retire (see Figure 1). It is expected that pension expenditures would double between 2001 and 2010, reaching DF 2.5 billion. Regarding revenues, the low-case and high-case scenarios suggest that contributions would not suffice to cover expenditures (even when government contributions are included). Thus, by 2015, the deficit of the system would have reached DF 1.2-1.8 billion, if the Government does not normalize its contributions,or DF 0.5-1.2 billion if it does (see Figure 2). Figure 1: OPSevolutionof new retirees 3000 2500 .! 2000- C 1500- E 1000 / 500 2001 2006 2011 2016 2021 2026 Source: author's calculations Figure2: Deficitsin OPSunderalternativescenarios 1.0 -199.0 4e -399.0-599.0 °-799.0 U LA _ -__ _ __ _ -999.0 -1199.0 - Stus-quo/Law Status-quo/GovConULow - 1399.0 _ Stabs-quo/High Status-quo/GovCont/High, l 159. -1799.0 2000 2005 2010 2015 Note: "Low" and "High" refer to the growth scenarios. "Gov. Cont" indicates that the Government pays contributions to OPS. Source: author's calculations CNR 17. In CNR only the Police regime is currently generating a budget surplus; both the regime for functionaries and parliamentarians have deficits of DF 37.8 million (0.04% of GDP) and DF 26 million (0.03% GDP) respectively. In contrast to OPS, the Government has been remitting contributions to CNR on a regular basis (DF 1.26 billion in year 2000). While delays are often observed, as a general rule by the end of the fiscal year 100% of legal contributions have been remitted to CNR. These contributions, however, are not sufficient to cover pension expenditures in the case of the regimes for functionaries and parliamentarians that have a joint deficit of DF 63.8 14 million. Part of this deficit is covered with the surplus of the police regime (DF 15.2 million) so that the overall deficit of CNR amounts to only DF 15.8 million. In GNA there are no contributions from workers (since all members of GNA are either retired or deceased) and the Government transfers the resources required to cover current pensions amounting to DF 125 million. 18. Part of the financial problem of CNR is due to the current practice of parliamentarians and ministers to "accumulate" benefits simultaneously in their regime and in the regime for functionaries, thus leading to replacement rates of over 100%o. Even if parliamentarians also contribute to both regimes, since current replacement rates are already unsustainable, cumulating benefits aggravates the problem. As an illustration, if a parliamentarian contributed during 15 years to the regime for functionaries and then 12 years to both regimes simultaneously, at retirement the replacement rate would be equal to: 3*15%+12*2% = 69% in the regime for functionaries, and 12*4% = 48% in the regime for parliamentarians. Thus, the total replacement rate would reach 117%! 19. Projections of expenditures and revenues conducted in the case of the regime for functionaries show that the deficit of CNR will continue to increase; the present value of the deficits accumulated during the period 2001-2015 would reach DF 1.4 billion6 . In the simulations we have assumed that within the next 5 years no new hiring in the Public Administration would take place, as part of the fiscal stabilizationprogram. At the same time, close to 70 workers would retire each year, thus reducing the total number of functionaries from 2,866 today to 2,489 in year 2006 (see Figure 3). The total number of contributors is assumed to stabilize afterwards. Under these assumptions, expenditures on pensions (old age and survivors) could go from 783 million today to 1 billion in year 2010, while revenues reach DF 906 million and the deficit of the fund increases to DF 178 million or 0.1 % of GDP (see Figure 4). Dueto the lackof data, actuarialprojectionsforthe parliamentariansand policeregimeshavenotbeen conducted. 6 15 Figure 3: CNR Age distribution of contributors and new retirements (Regime for functionaries) 0.14 100 0.12 90 0.1 X -__ 0.08- - . ___ _ ___ ** * 780 ~~~~~~~~~~~70 / ~~~~~~~~~~~~60 * * 0.06- __\ **.50 X0.04 6_* _------* 40 . G~~~~~~~~~~~~~E 30 0.02 * * ~~~~~~~~~~~~~20 Z 0 0 - 10 20 30 40 50 60 10 0 70 Age 2000 2005 2010 2015 2020 2025 2030 Source: Author's calculations Figure 4: CNR projected deficits 0.0 - -1_0.0 ° -100.0 -200.0 _ ____ ~ __ _____________ _._ ___ .2-200.0 -250.0 - _ _ _ _ -_ -300.0 -350.0- . -403.0 soO.o--450.02003 __\__ 2002 2034 2W8 \ ____ __ _,_ 20E 2010 _ 2012 2014 Source: Author's calculations CMR 20M CMR has an operational deficit of DF 271 million that is being covered by the Government. Workers and employer (the Government) contributions to CMR amount to DF 404 million. Total pensions paid, however, reach DF 675.5 million. The 271.1 million deficit is currently being financed by the Government. Thus, total transfers from the central budget represent 0.7% of current GDP. The actuarial analysis conducted by the consulting firm IEMACTUARIA indicates that pensions expenditure would continue to increase from DF 680 million in 2002 to DF 1.2 billion in 2015. The present value of these expenditures could reach DF 3.3 billion (see Figure 5). 16 Figure 5: CMR projectionsof total wages and pensions (DF billion) 3,0- - - - - - - - -- --- - - - - - - - - - - - - - - - - - - - - aD zo - 1,5 ~~~~~~~~~~~~~~~~~~PL-:iotn - 1,0 MasseSajuijale 0,5 net-ects 2mi 20M __0 209 ___ 211 2013 2015 Source: IEM-ACTUARIA 3. Institutional constraints 21. A recent audit of current information systems and administration and management practices within the funds reveals weak institutional capacity. A report by the consulting firm IEM-ACTUARIA identifies institutional problems at four levels: i) administrative processes are inefficient and lead to the duplication of efforts and high administrative costs; ii) integrated and updated information systems are lacking; iii) accounting and auditing practices do not have appropriate standards; and iv) management and analytical capacity are weak and hamper the implementation of any reform program. 22. Administrativeprocesses are inefficient. There are a series of administrative processes that operationalize the functions of the pension funds. These include: registration of new contributors and their employers, follow-up of life events, collection of contributions, liquidation of new retirees, payment of benefits, and management of funds reserves. IEM-ACTUARIA's report shows that the execution of these processes is highly inefficient as: i) procedures remain cumbersome; ii) there is little or no coordination between the different units in charge of implementation leading to the duplication of efforts; and iii) appropriate information systems that could reduce or eliminate manual processing are lacking. To date, databases of contributors and retirees remain outdated, there is almost no information on employers making it difficult to track contributions, and theoretically simple operations such as calculating the benefits of future retirees can take several days. In addition, revenues and expenditures are not properly monitored which often leads to cash-flows problems. 23. Appropriate information systems are lacking. While most units within the funds have access to basic information technologies, both in terms of hardware (computers, printers, faxes) and software (word processor, spreadsheet, database managers), these are not operating in the context of an integrated information system which supports management and streamlines administrative processes. Each fund, and within each fund different units, have designed spreadsheet or databases to support daily tasks. Unfortunately, without coordination, the resulting databases are considerably different in terms of structure and in the absence of a unique identifier code for each contributor and beneficiary, it is impossible to create links. Moreover, because the maintenance of the databases is not explicitly integrated into the execution of the different administrative functions, the majority are outdated. In terms of cash-management, a system able to generate short-termforecasts of revenues and expenditures (e.g., new retirees) is not available. 24. Administrative costs are high in part due to inefficient human resources management. Administrative expenditures reach 6% of total expenditures in the case of OPS, reflecting high 17 expenditures in wages. Indeed, preliminary evaluations suggest that the actual number of employees, particularly support staff, is high compared to real needs. In terms of management staff, they are highly motivated but a majority of them lack appropriate training in areas such as business administration, finance, and actuarial analysis. 25. While a detailed analysis of accounting and auditing practices has not been conducted, current systems do not have appropriate standards. All pension funds have developed their own computerized accounting system. Again, in the absence of an integrated infornation system, its maintenance is cumbersome and time consuming. Transactions such as the payment of benefits and the collection of contributions are processed and registered manually by the responsible units. The accounting department then receives hard copies of the registered transactions that are entered manually into the accounting system. Delays are often experienced, and reports are limited and are not generated on a periodic basis. While guidelines in terms of auditing procedures exist, these are not followed systematically. 4. The problem of arrears 26. The current financial crisis is, in part, explained by the non-compliance of the Government to remit required contributions and repay loans contracted with the pension funds. Preliminary estimates suggest that government arrears with the pension funds could account for 8% of GDP. The majority of these arrears (89%) are related to late contributions while arrears related to loans account for 11% (see Table 6). These numbers, however, vary depending on how interests and late fees are handled. For instance, in the case of OPS, total arrears excluding interests amount to DF 4,310 million and increase to DF 4,778 million or 4.8% of GDP when interests and penalties are included. Similarly, in the case of CNR, arrears without interests approximate DF 2,666 million and reach DF 3,038 million (3.9% of GDP) when interests and penalties are included. These estimates have not been validated. Moreover, information on arrears with the private sector is not available, yet given high evasion rates it is likely that these arrears are considerable. Table 6: Government arrears with OPS and CNR (DFMillion) Arrears Credit Net Arrears Interests Arrears including interests Late fees Arrears including interests and late fees 4,049 899 4,948 639 639 3,410 899 4,309 218 3,628 899 4,527 198 52 250 3,826 951 4,777 2,663 2 2,665 7,613 - 2,663 2 2,665 6,974 - 2,663 2 2,665 7,192 370 3,033 2 3,035 7,812 OPS Government Pubhc enterprises Total CNR Government Pubhc enterprises Total General total - 639 - 218 - 218 - 370 620 Source: Actuaria on the basis of numbers provided by OPS and CNR. These numbers differ from those provided by the Ministry of Finance. As discussed in the text, differences can be explained by the way interests and late payments are handled. 18 III. RECOMMENDATIONS FOR POLICY REFORM The problems facing the pension funds in Djibouti are by no means particular to the 27. country, similar issues are observed in other parts of the world. In fact, pension systems reform is likely to be one of the most important elements in the development agenda of governments in the years to come. 28. In the case of Djibouti, a two-stages reform program is suggested. A first stage, which is the focus of the current Structural Adjustment Credit operation, would concentrate on consolidating a modest pay-as-you-go system. This involves executing three major activities: * implementing institutional reforms to improve governance and management (through the creation of the National Councilfor Social Security, NCSS, and the preparation of a multiyear plan to merge CNR and OPS), and to generate efficiency gains that enhance service delivery and reduce administrative costs (by streamlining administrative processes, improving the allocation of human resources, and implementing an integrated management and informationsystem); * addressing short-term financial needs by normalizing government contributions, defining a plan to refinance government arrears, and introducing a ceiling for the replacement rate in the regimefor parliamentarians; = and improving the medium-term financial situation of the pension funds by aligning contributions and benefits while introducing a minimum pension to protect low-income workers. 29. A second stage of the reform program would focus on: consolidating the merger of OPS and CNR; * and introducing/reinforcingincentives to promote contractual savings. The latter would pave the way for a possible third stage that would focus on addressing long-term challenges for the current pay-as-you-go system. These include expanding coverage and addressing the financial problems associated with the aging of the population through the gradual introductionof a mandatory funded pillar with individual accounts. 1. Short-term reform program 1.1. Improving governance and management: preparing the merger of OPS and CNR 30. In order to improve governance and generate efficiency gains in management, it is recommended to create the National Council on Social Security (NCSS). The NCSS would have authority over the two pension funds, CNR and OPS, and in a period of 3 to 5 years it would constitute the new social security system that integrates OPS and CNR. The NCSS would be responsible for developing and implementing a multi-year plan to merge administratively and financially OPS and CNR (only the regimes for functionaries and parliamentarians), thus improving dependency ratios. The NCSS would also manage and coordinate the other aspects of the pension reform strategy (see Section IV). The NCSS would be jointly attached to the Ministry of Finance and the Ministry of Labor (see Figure 6). Its Board of Directors would include a representative from the Ministry of Labor, a representative from the Ministry of Finance, a representativefrom the Direction of the Treasury, the directors of OPS and CNR, 3 representatives from the private sector selected by the Chamber of Industry and Commerce, and 3 representatives 19 from contributors and beneficiaries of the pension funds. The Executive Director of NCSS would be elected by the Board of Directors on the basis of a list of candidates jointly presented by the Ministry of Labor and the Ministry of Finance. The law creating the NCSS would give the Board of Directors the power to replace the Executive Director when considered necessary. The budget of the NCSS would be financed from the revenue of the pension funds (OPS and CNR) in shares that would be determined jointly by the Ministry of Finance and the Ministry of Labor. Initially the NCSS could operate in OPS's premises. Figure 6: Organizationalstructure of the NCSS |Ministry of Laborand || NationalSolidarity l Ministryof Finance _ 0~~ LoP NCSS l X CNR Source: Author's design 31. The multi-year (3 to 5 years) plan to merge OPS and CNR would ensure that over time administrative processes, accounting systems, and the structure of contributions and benefits converge. The plan would first require that OPS and CNR coordinate the implementation of institutional reforms (see Section 1.2), so that they start sharing similar administrative and management procedures. While initially OPS and CNR accounts would continue to be separated, it is desirable that both move to similar accounting, auditing and reporting systems. Parametric reforms in OPS and CNR (see Section 1.3) would need to be coordinated as to ensure that contributions and benefits converge over time. 1.2. Improving institutional capacity 32. A necessary condition to improve the financial situation of the pension funds is to improve institutional capacity. To this end, 3 major interventions are required: i) streamlining administrative processes and developing an integrated information system; ii) adopting appropriate accounting, reporting and auditing standards; iii) rationalizing human resources management and reducing administrative expenditures; and iv) identifying mechanisms to ensure that reserves, if accumulated, are managed prudently. 33. Streamlining administrative processes and developing an integrated information system. As previously discussed, the Government is conducting a comprehensive study of current administrative processes in the pension funds7. For each process, the study evaluates tasks and information flows and recommends actions to improve efficiency. The study also presents the blue print of an integrated information system along with an implementation plan. This system would operate within a shared network, thus allowing different units (executing different administrative processes) to maintain simultaneously a single set of databases (see Figure 2). The structure of these databases has been defined and would need to be used across funds to facilitate future 7See:Actuaria2001. "Djibouti:InstitutionalReformof the PensionFunds". Paris. 20 exchanges of information. The challenge is now to implement the different recommendations of the report. Not all processes can be streamlined simultaneously and therefore priorities need to be established. It is recommended to concentrate on: i)finalizing the process of gathering information about employers, contributors, and retirees on the basis of the new database structures; ii) streamlining and computerizing the registrationprocess so that the new databases remain updated; and iii) given high evasion rates, improving the collection processes. The NCSS would be in charge of the implementation of the new system, thus allowing coordination across funds. 34. Improving cash-management. As part of the new management and information system, it is recommended to develop a simple cash-management tool that will allow funds accountants to perform short-term forecasts of revenues and expenditures. Given data on contributors and retirees, the system will generate short-term projections of new retirements and associated expenditures. A spreadsheet-based system would be probably the most cost-effective solution. The technical personnel have also requested instruments and training to perform medium-term actuarial analysis. A model such as the World Bank PROST could be installed in-house and personnel trained on its use. Figure 7: The core management and information system ..........-........................ ''...........''.... ... ...............-.-................... Registrationsof new affiliations Employers Collectionof Contributions ........................................................ .... ......................... t.I .. .-.............. /........ Followup of life events rContributors Contributions | Dependents | ............. ~~~~~~~~~~~~~~~~~~~~.......... .... ............................ Liquidations | Retirees Monitoringuse of familyservices Fail < b Payments ..................................................... .............................. ....... ...... Process LI Database - o Information flow Source:Author'sdesign 35. Rationalizing the use of human resources and reducing administrative expenditures. The new administrative processes will have implications for human resources policies. In particular, as productivity increases, the number of workers required to execute the different tasks could be reduced. The other components of administrative expenditures would also need to be evaluated. To this end, it is recommended to conduct an audit of administrative expenditures in OPS (including the health and family services branches). On the basis of this audit unit costs could be computed and eventually adjusted. 21 Improving accounting and auditing practices. There are three main objectives: i) to 36. promote transparency in the management of the pension funds' accounts; ii) to improve the quality of the information and the timing of the reports; and iii) to standardize accounting systems across funds in order to facilitate the future merger of OPS and CNR. To this end, it is recommended to conduct a detailed study of current accounting and auditing practices. The study should identify major weaknesses and propose recommendations along with an implementation plan. In the case of OPS, it is necessary to separate the accounts of the three branches (health, family services, and pensions), and to eliminate cross-subsidies. 37. Adopting mechanisms to ensure that if reserves accumulate they are managedprudently. International experience suggests that public pension funds reserves rarely generate positive rates of return8 . The recommendation in the case of Djibouti is to minimize the stock of reserves to levels required to cover liquidity needs. In terms of their management, the current practice in OPS of investing reserves in a savings account could be maintained. If higher levels of reserves are accumulated, an appropriate regulatory framework for their management would need to be put in place. The funds would need, for instance, to set-up an investment committee and define investment policies, asset allocation procedures, and appropriate reporting and auditing mechanisms. The management of reserves could also be contracted-out, but appropriate regulations are also required in this case. Areas that would need to receive attention include the selection and evaluation of fund managers, and guidelines to limit the range of possible investments as a function of their expected return and risk. Given the complexity of the issue, it is recommendedto command a study that identifies the best option(s) for reserves management in the case of Djibouti and the associated institutional requirements. 1.3. Improving financial sustainability 38. The proposed strategy to address the financial dfficulties of the pension funds considers actions that will have an immediate impact and actions that will mostly affect the finances of the funds over the medium-term. In the short-term, there are four key interventions: i) normalize government contributions to OPS (this will allow the retirement of the 773 contractual workers who despite verifying retirement conditions remain in the payrolls; ii) adopt a refinancing plan for government arrears with the pension funds that would ensure that the operational deficits of OPS and CNR are covered; iii) introduce a ceilingfor the replacement rate of CNR that applies to those workers who accumulate benefits in multiple systems; and iv) increase collection rates in the private sector through a better monitoring system and a reduction in labor charges. In terms of policies that will improve the medium term financial position of the funds, it is recommended to gradually adjust the parameters defining contributions and benefits. To protect low-income individuals from the impact of these adjustments, a minimumpension should be introduced. 39. Normalizing government contributionsand retiring the 773 contractual workers who verify retirement conditions. One of the main objectives of the fiscal program developed by the Government with the support of the IMF and the World Bank is to stop the accumulation of arrears. Hence, the FY02 budget includes allocations to cover the payment of wages and contributions to all pension funds. Its implementation is linked to the observance of a treasury plan that monitors weekly revenues and expenditures. In the case of OPS, over the next 10 years, government contributions are expected to average 0. 16% of GDP (including workers contributions). The normalization of government contributions would bring important fiscal s SeePalaciosR. "Avertingthe Old AgeCrisis:TechnicalAnnex". WorldBank.PolicyResearch Department, Washington, DC; and Iglesias A. and Palacios R. "ManagingPublic Pension Reserves. Evidencefromthe InternationalExperience",in NewIdeasaboutOld Age Securityeditedby HolzmannR. and StiglizJ. The WorldBank. Washington,DC. 22 savings. These savings result from the fact that total contributions are in the order of DF 217 million today and could decrease to DF 200 million by year 2005, while wages paid to the 773 contractual workers are in the order to DF 443.7 million (net of the 4% contribution that the Government discounts on wages and does not remit to OPS). In the case of CNR, current contributions represent 0.7% of GDP and could be reduced to 0.5% by the end of the decade (see Figure 8). Figure 8: Fiscal cost of wages and contributions to OPS and CNR |_8.00°h 7.009/ __-_-- 6 00% _- 5.00 _ _ 3.400°/ 1.00% - Wages contractual workers(netof confibubons) Vs-- Consributons fDncionaries _ __*-- Contibutons contractual workers . _ -_ 2.00% + ____ -b- Wages tanctionanes (excludes police iand military _0 *- -Total 0.00*% 2000) 2002 2004 2006 2008 2010 Source: Author's calculations 40. Increasing the collection ratefor private workers through a better monitoringsystem and a reduction in labor charges. We have seen that OPS is a young system with a nominal dependency ratio of only 11.4%. Its main problem is that the effective collection rate is too low given the Government's moratoria and evasion in the private sector. If in addition to normalizing government contributions the collection rate among private workers could increase from its current 46 to 55% within the next 10 years, the financial situation of the fund would improve dramatically (see Table 2, last scenario). This could be possible as a result of the implementation of the new information system that would facilitate tracking of workers and employers contributions. Collection rates are also expected to increase as workers' expectations about the long-term situation of the pension fund improve given the reform program. However, higher collection rates may be hampered by currently high social charges (19.7% of wages). It is therefore necessary to consider reducing the contribution rate for the health and family services branches, on the basis of an actuarial analysis. 41. Adopting a ceiling for the replacement rate in the regime for parliamentarians within CNR. The fact that high rank functionaries within the central administration can simultaneously accumulate benefits in two funds leads to high replacement rates (sometimes over 100%). It is recommended to introduce a ceiling that is as close as possible to the current equilibrium replacement rate of 40% (see Table 2). This measure is expected to generate savings of close to DF 40 million or US$250,000 per year (assuming that 50% of retirees in the parliamentarians regime accumulate benefits in both funds), over two times the current operational deficit of CNR. 42. Refinancing government arrears. This activity needs to be undertaken within the context of the global program to restructure arrears that the Government is implementing through the current Structural Adjustment Credit. The goal of the refinancing plan is not to provide pension funds with additional resources to invest, but simply to generate cash-flows which ensure that operational deficits are covered. The first step would be to validate the current estimate of the 23 value of government arrears (US$7,812 million). To this end, figures from the Ministry of Finance need to be reconciled with figures from the pension funds. In the case of loans that have appropriate documentation this should be a simple task. However, in the case of loans with incomplete documentation or in the case of arrears related to workers contributions, an agreement in terms of interest rates and late fees is needed. It is recommended to hire an internationally respected auditing firm to assist in this process. Once the total amount of arrears is known, the simplest refinancing mechanism would consist of registering the arrears as a formal Government's debt and agreeing on a payment schedule (e.g., quarterly payments) that takes into account the pension funds financial needs as well as the Government's payment capacity. More complex financial instruments, such as bonds, could be considered, but given the absence of a market for these bonds in Djibouti, the value added of these instruments is nil at this point. 43. Introducing parametric reforms and implementing a minimum pension. Improving the financial position of the different pension funds over the medium-term required gradually adjusting contributions and benefits within the context of a sustainable fiscal framework. There are four basic principles that should guide these adjustments: i) adjustments should be gradual; ii) acquired rights should be honored; iii) changes in contribution rates should not harm the competitiveness of Djiboutianfirms; and iv) mechanisms should be adopted to protect low-income groups. Extensive discussions with government officials have taken place to define plausible parametric reforms to improve the medium-term financial position of the fund while minimizing short-term social impacts. The focus has been on the vesting period, the marginal replacement rate (accrual rate), and the retirement age. Given the highly unequal distribution of pensions9 , authorities have also suggested examining the implications of taxing the pension of individuals in the top deciles of the pension distribution. However, because Djibouti's taxation rate is already high (close to 25% of GDP), this measure is controversial. Authorities have also agreed to introduce a minimum pension to protect low-income workers. This minimum pension could range between DF 150,000 (roughly equivalent to 2 dollars per day) to DF 170,000(the poverty line in Djibouti). Options for parametric reform in the case of OPS 44. In this section we summarize the results of the simulations of alternative parametric reforms in the case of OPS. The proposed reforms capture the range of plausible choices by combining different targets for the vesting period, the accrual rate, the option of a tax on high pensions, and the option of a minimum pension. These reforms are analyzed in the context of the two scenarios previously introduced: low-case (low labor productivity growth and low expansion of the system), and high-case (high labor productivity growth and fast expansion of the system). For illustration purposes, we have also simulated an increase in the collection rate among workers in the private sector. 45. The simulations show that the normalization of government contributions combined with a gradual increase in the vesting periodfrom 15 to 25 years, and a reduction in the accrual rate to 1.2-1.5% would generate important reductions in the deficits of OPS (betweenDF 3.8 and 4 billion in present valuefor the period 2001-2015, or US$21-23 million, depending on the scenario). The present value of the deficits for the period 2001-2015 could be reduced to DF 1.3-1.5 billion (US$7.8-9 million). These deficits could be covered by refinancing part of the Government's arrears with the pension fund. The normalization of government contributions alone can reduce the present value of the deficit by DF 2.4 billion. The parametric reform would generate additional DF 1 billion (US$5.6 million) in savings (see Table 7 and Figure 9 for a summary of the main 9In the caseof OPS,for instance,whilethe averagepensionin year2000 was closeto DF 486,000,the averagepensionof the highestdecilewasDF 1.4millionwhilethe averagepensionof the lowestdecilewas onlyDF 143,960. 24 simulations). We notice that within the next 15 years, the difference between a parametric reform targeting a 1.5 accrual rate and a parametric reform targeting a 1.2 accrual rate is minimal (in the order of DF 100 million in present value). Within a longer horizon (30 years), however, the difference can be significant (in the order of 3.5% of GDP in present value). 46. Within the context of the parametric reform, a minimum pension of DF 170,000 could be financed. Today, there are no retirees with a pension below DF 140,000. Ten percent of retirees, however, have a pension below DF 150,000(2.3 dollars per day). The poverty line in Djibouti is close to 3 dollars per day, or DF 170,000 per year. The calculations show that this level of minimum pension could be supported by the system. Its cost during the period 2001-2015 would approximate DF 280 million (0.3% of GDP). It is important to keep in mind, however, that under some circumstances (e.g., when workers do expect major increase in wages or when there are opportunities to work in the informal sector after retirement) the introduction of a minimum pension could provide incentives for retiring as soon as the vesting period has elapsed (see Technical Appendix. Section A4). Thus, individuals who would have continued to contribute in the absence of a minimum pension (today individuals contribute an average of 20 years when the vesting period is 15 years) would prefer to retire. This could increase pension expenditures in the short term. To avoid this situation, policymakers could consider linking the minimum pension to a higher vesting period and ensure that beneficiaries do not have other sources income in the formal sector. 47. A tax on high pensions could generate importantfinancial resources (and therefore reduce the need to refinance government arrears), but the social and political impacts should be carefully assessed. If a tax on pensions is introduced, it is important that it does not affect low and middleincome retirees. In the simulations, for instance, a 15% tax was applied only to pensions equal or above DF 500,000. For pensions between 500,000 and 800,000, only 50% of the pension was taxed. For pensions above 800,000, 100% of the pension was taxed. This type of arrangement could generate resources equivalent to 1.7% of GDP during the period 2001-2015. The present value of the accumulated deficits during this period would be reduced to DF 100-300 million (US$1-2 million), even after taking into account the cost of the minimum pension. If the tax is managed with prudence, then it could be a powerful mechanism to solve the short-term financial problems of the system. It is important to notice that the introduction of the tax reduces or eliminates the need to refinance government arrears. Policymakers therefore need to evaluate, on the basis of fiscal, political, and social considerations, which is the best combination of these two instruments. 25 Table 7: OPS financial impact of alternative parametric reforms Scenario a GovernmentVesting Accrual Tax' MinimumPresent %GDP Present %GDP Contributionperlodc rated Pension Value Value (FD000 (2000(2000-2030) Deryear) 2015) Low-case Low-case Low-case Low-case Low-case Low-case Low-case High-case High-case High-case High-case High-case High-case High-case withincrease inthecollection rateof theprivatesectorb No Yes Yes Yes Yes Yes Yes No Yes Yes Yes Yes Yes 15 15 25 25 25 25 25 15 15 25 25 25 25 2% 0% 2% 0% 2J1,8/1,2 0% 2/1,8/1,2 0% 2/1,8/1,2 15% 2/1,8/1,5 0% 2/1,8/1,5 15% 2% 0% 2% 0% 211,8/1,2 0% 2/1,8/1,215% 211,8/1,5 0% 211,8/1,515% 0 0 0 170 170 170 170 0 0 170 170 170 170 Yes 25 2/1,8/1,5 15% 170 ($5,466) -5.6% ($2,558) -2.6% ($1,581) -1.6% ($1,867) -1.9% ($345) -0.4% ($1,910) -1.9% ($384) -0.4% ($4,379) -4.4% ($2,395) -2.4% ($1,650) -1.7% ($107) -0.1% ($1,696) -1.7% ($148) -0.1% $834 ($27,659)-28.1% ($21,603)-21.9% ($9,923) -10.1% ($10,254)-10.4% ($8,507) -8.6% ($13,604)-13.8% ($11,849)-12.0% ($16,517)-16.8% ($22,785)-23.1% ($10,353)-10.5% ($8,577) -8.7% ($14,048)-14.3% ($12,263)-12.5% 0.8% ($8,171) -8.3% a: Refer to the Technical Appendix for a description of the scenarios. b: Assumes collection rate increases by 10 percentage points between 2002 and 2020. c: The vesting period reaches the new target within a period of 10 years. d: The accrual rate issetto 1.8%between 2002 and2007, 1.5%between 2007 and 2010, and 1.5 or 1.2% afterwards. e: The tax applies to pensions equal or above DF 500,000 per year. For pensions between DF 500,000 and 800,000 only 50% is taxable. For pensions above 800,000 100% is taxable. Source: Author's calculations. Figure 9: OPS: Current balances under alternative parametricreforms 400.0.. 200.0 0.0 -200.0 _ ,, ----600.0 _ Stbtus-quoIGov/Low - ReforrmGovMRRl.2ILow 800.0 -1000.0 ReJorm/GovMRR1.2Taxe/Low -*- -1200.0 Refrrm/GovMRR1. 2High ___ K- Reform/GovMRR1l2Taxe/Hihh__ _ a-a- Refrm/GovMRR1.5TaxePrivatatHigh 2000 2002 2004 2006 2008 2010 2012 2014 Note: "Low" and "High" refer to the growth scenarios. "Gov" indicates that the Govemnent pays contributions to OPS. "MRR" is the targeted marginal replacement rate (accrual rate). "Tax" indicates a 15 tax is introduced. "Private" indicates that the collection rate of the private sector increases by 10 percentage points ova a 20 year period. Source: Author's calculations. 26 Options for parametric reform in the case of CNR 48. Given data constraints, we have analyzed parametric reforms only in the case of the regime for functionaries. The range of options for parametric reform in this regime is more constrained than in OPS, since the vesting period is already high (25 years), the Government is remitting contributions, and workers are already retiring at age 55 instead of 50 as required by law. The focus is therefore on gradually adjusting the marginal replacementrate (accrual rate) to 1.2%. 49. The results show that reducing the accrual rate would not suffice to eliminate the operational deficits of the system during the period 2001-2015, nonetheless, these deficits would be significantly reduced After adjusting the accrual rate, the deficit of CNR would continue to increase, reaching DF 300 million in 2015 (0.15% of GDP, compared to 0.05% today). The present value of accumulated deficits, however, could be reduced from DF 1.4 to 1.2 billion, that is a reduction equivalent to 0.2% of GDP (see Table 8). During the period 2001-2030 savings generated by the parametric reform would approximate 2% of GDP. Table8: CNRfinancialimpactof a reductionin the accrualrate (Regime for functionaries) Present %GDP Minimum Value GovemmentVesting contributions period Yes Yes Yes 25 25 25 Accrualratea Tax pension (2000-2015) 3%/2% 0% 1.8%/1.5%/1.2%0% 1.8%11.5%I1.2%J6 15% 0 170 170 ($1,447) ($1,246) ($265) Present %GDP Value (2000-2030) -1.47% -1.27% -0.27% ($4,090) ($2,404) ($1,315) -4.15% -2.44% -1.34% a: The accrual rate is set to 1.8% between 2002 and 2007, 1.5% between 2007 and 2010, and 1.2% afterwards. Source: Author's calculations. 50. Like in the case of OPS, the short-term deficits could be financed by a tax on pensions and the partial repayment of government arrears. A 15% tax would be sufficient to eliminate the deficit of CNR until year 2007 (see Figure 10). This tax could generate resources equivalent to 1% of GDP. The present value of accumulated deficits could therefore be reduced to 0.27% of GDP. This deficit could be financed with the repayment of part of the Government's arrears. In the medium-term, the merger with OPS will also contribute to reduce financial pressures resulting from high dependency ratios. 51. While the regime for parliamentarians has not been analyzed in depth, it is clear that it is financially unsustainable due to a high dependency ratio; a merger with the regime for functionaries is recommended This implies that new parliamentarians would accumulate the same benefits as functionaries. Any additional benefits would be financed outside the public system. This would also, by default, solve the problem of accumulation of benefits in multiple funds. 52. The regime for the police is still generating an operational surplus, but given current trends, the financial situation of the fund is likely to deteriorate within the next 3 to 4 years. The data required for conducting an actuarial analysis of the fund need to be collected. On the basis of this analysis, corrective measures would be implemented. As suggested by the Ministry of Labor, an alternative to consider is the merger of this regime with the regime for the military (see next section). 27 Figure 10: CNR balance under alternative parametricreforms (Regime for functionaries) 50.0 0.0 -50.0 -100.0 -150.0 -200.0-250.0 -300.0 -_ 300.0- + Staus-quo/Low -ie RelornVMRR1.2 400.0RefornlMRR1 .2/Tax 450.0 - _ _ _ _ _ _ _ I 2002 2000 2004 2006 2008 2010 2012 2014 Note: "Low" indicates low-case scenario. "MRR" is the marginal replacement rate (accrual rate). "Tax" indicates that a 15% tax on pensions is introduced Source: Author's calculations. Options for Parametric reform in the case of CMR 53. The recommendation in the case of CMR is to include pension expenditures as an additional line in the budget, similar in nature to the line for wages. Indeed, given the special characteristics of its population, it is unlikely that CMR could operate as a regular pay-as-you go fund. A study conducted by the consulting firm IEM-ACTUARIA shows that to balance the finances of the system the contribution rate would need to increase by 10 percentage points, while a 30% tax on all pensions is introduced. Even in this case, the improvement in the financial situation would be temporary (3 years). By year 2005, the system would again generate an operational deficit that would need to be covered by the central budget. Under these circumstances, it is recommended to treat pension expenditures as wage expenditures. The study of IEM-ACTUARIA shows that under appropriate changes to the formula used to compute benefits, total wages and pensions expenditures could remain below DF 2.5 billion during the next 15 years (see Figure I1). Figure 11: CMR total wage and pension expenditures under alternative Reforms (billion DF) 2,9 Stotu q. 2RCO. 2,5-3 .30% R, 2.3------------ [.Pit ('oCtl S.i - 2~~~~~~~~~~~~~~~ 2,1 2000 i10% l t; 2002 2004 St.t. l 2008 quo Rferme l R~~~1,9 i~~~~~. R D rnit lR +10% R ,It% I 8,1%CoCIo, 2008 nor--lisaon .I 030% 2010 des co.ta Jimmuohun graduf.li§sopAts otis Riforn- SolO 3 0 % i dr s Rdtoron-d 9o 2014 uRils + pu's-gs hop IS RSr.es* Aql 2019 de I'EtSt de- 1-ux d n ut I RSForma.+inspossSion Aoqsi. (Cott, 2012 ian, D -- Insp6i 30% -+mp6t I5% (0 sc o.etobl.) dolMs axc ifs de7 7.,17'S S1(' de Impblb zI.iiUr.r box cot., S Source: IEM-ACTUARIA 28 2. Medium-term reform program 2.1 Consolidating the merger of OPS and CNR 54. As administrative processes, information, accounting, and auditing systems, and the parameters determining contributions and benefits start to converge, the administrative and financial merger of OPS and CNR could take place. Ideally, within a period of 3 to 5 years, the institutional reforms discussed in Section 1.1 would have prepared OPS and CNR for a merger. The merger would imply leaving NCSS as the head of the new Social Security system. This system would have three branches: i) health; ii) family services; and iii) pensions and disability. The latter would merge the administrative structure of the OPS's pension branch with CNR, as well as the respective accounts. A cutting date would need to be established after which all new workers in the public administration would join the new system. At the time of the merger, CNR contributors and beneficiaries would transfer their acquired rights. The financial and legal implications of the merger would need to be carefully assessed. 2.2 Promoting contractual savings 55. While in the case of Djibouti little is known about the development of occupational pension plans, these should be considered as important means to substitute earnings in old age. Both occupational and personal savings schemes enable workers who earn more during their active years to accumulate larger pension rights for their retirement years. 56. Even though promoting long-term personal savings and occupational plans will require structural reforms to create a more dynamic and diversified economy, along with an efficient and properly regulated financial sector, in the short-term, different types of incentives can be considered. For instance, countries such as Singapore, Korea and Malaysia have established postal savings systems that offer greater security and accessibility, highest interest (often tax-free), and lower transaction costs, especially, in the case of rural areas. The development of occupational plans can also be encouraged by making contributions tax deductible and deferring taxes on investment income. This type of incentives have been common to most countries and have made occupational programs suitable substitutes for higher wages, thus encouraging their diffusion. In the developing world, occupational pension programs have been growing in countries such as India, Indonesia, Mexico, Brazil and Zimbabwe. While there is a gradual tendency to make these programs mandatory, particularly in the case of OECD countries, Djibouti should initially adopt a flexible stance, given the still frail situation of the economy. Nonetheless, a minimum regulatory framework should be in place to require that the programs are appropriately funded and of sufficient portability. Incentives would need to be carefully designed to ensure that benefits do not concentrate on high-income households who would have saved in any event. 57. Within the context of the reform program it is recommended to conduct a study that evaluates the potentialfor the development of occupational pension plans and personal savings in Djibouti, and presents recommendations in terms of incentives and the necessary regulatory framework. This study would start by preparing an inventory of existing occupational plans and incentive systems to promote personal savings. It will then present recommendations to promote the developmentof these programs, on the basis of best practices at the international level. 2.3. Towards a third pillar 58. As in other countries, population aging in Djibouti will ultimately put pressure on the financial performance of the pay-as-you go system. To date, the support ratio in Djibouti (the ratio between the share of the population aged 60 and older to the population aged 20-59) is close to 11.5%, which is relatively low for international standards (as a reference, the support ratio in 29 OECD countries is close to 35%). Over time, however, this ratio is expected to increase reaching 20% by year 2050 and 35% by year 2075 (see Figure 12). 59. One alternative that the Government could consider to reduce financial pressures on the pay-as-you go system, is the introduction of a mandatory funded pillar with individual accounts. In a funded system, the ratio between the active population and the retired population no longer affects its financial position, as the benefits received by workers when they retire are a function of the savings they have accumulated during their working lives. For equity purposes, the current pay-as-you-go system would continue to operate, but it would not longer be the main source of income for retirement. Its function would be to redistribute income from high-income to lowincome workers by guaranteeing a minimum pension. The transitionprocess to a multi-pillar system, however, is a delicate process and should 60. be carefully designed and evaluated. Lessons from other countries suggest that the costs of the transition are lower when the pay-as-you go system is still relatively immature, while the benefits depend, in part, on the degree of development of capital markets. Indeed, high rates of return on savings will require efficient capital markets. The proposed system of incentives to promote voluntary savings (see Section 2.2) would smooth the transition process. Ultimately, the introduction of a multi-pillar system should be based on a thorough analysis of fiscal cost, and macroeconomic and social impacts. Figure 12: Evolution of age pyramids and the dependency ratio in Djibouti I ......... f~~~~~~~~~~~~ _ 0i4 - 035 0.3 0.25 0.2 j0.15 :0.1 100 2000 2010 2020 2030 2040 2050 200 27 Source: PROST demographicprojections 30 IV. SEQUENCING AND MANAGING REFORM ACTIVITIES 61. The first activity in the reform program would be the creation of the NCSS; the NCSS would be responsible for managing and coordinating the reform program at the technical and political levels. In the short-term it would focus on the following activities: Institutional reforms * Develop a multi-year (3 to 7 years) plan to prepare the merger of OPS and CNR. * Coordinate the design and implementation of the new information system (including a cashmanagement system) for OPS and CNR. This would imply: i) reviewing the recommendations of the study of administrative functions and information needs conducted by lEMACTUARIA; ii) preparing terms of reference for the procurement of consultants who will advise in the implementation process and preparing the procurement of information technologies; and iii) monitoring implementation. * Prepare terms of reference for an evaluation of current accounting and auditing systems. * Prepare terms of reference for an audit of the Government's arrears with the pension funds. * Prepare terms of reference for a study to improve the regulatory framework for reserves management. * Prepare terms of reference for an audit of the administrative expenditures of OPS and the actuarial analysis of the health and family services branches. * Prepare terms of reference for a study on the development of contractual savings in Djibouti. Reforms to improvefinancial sustainability * Coordinate with the Ministry of Finance the normalization of government contributions and the retirement of the 773 contractual workers who verify retirement conditions * Introduce a ceiling for the replacement rate in the regime for parliamentarians. * In coordination with the different stakeholders, prepare a final package of adjustments to the parameters of the pension funds, evaluate financial implications, and prepare necessary legal amendments'°. * Ensure that a minimum pension to protect low-income workers is introduced. * Coordinate with the Ministry of Finance the design and implementation of a plan to refinance arrears that would generate sufficient cash-flows to cover the operational deficit of the pension funds. '° Legaltextsfor OPS,CNRand CMRhavebeenreviewedandstreamlinedby IEM-ACTUARIA.Changes requiredfor implementingalternativeparametricreformshavebeenidentified. 31 62. The NCSS would also be in charge of informing civil society and in particular workers and beneficiaries about the objectives and the content of the reform program. Looking forward, a major challenge is to build consensus for reform. Some of the short-term and medium-term measures to be introduced (e.g., reducing pension benefits) may face high political resistance. The success of the reform process will depend on the Government's ability to promote dialog with different stakeholders, create awareness about the current problems of the system, and explain the pros and cons of alternative reform proposals. The major risk is that fixing the short-term financial position of the pension system delays medium-term reform. It is important, therefore, that shortterm reform is analyzed and discussed within a medium term planning horizon. The long run implication of current financial decisions should be made explicit to policymakers and stakeholders. Table 9: Costs of implementinginstitutional reforms Activity Design and implementationof the inforrnation system Update and audit of databases Resources DAYS Daily Rate 800 Consulting firn to conduct 75 audit of the system and prepare a development and implementation plan (in progress) Total fees 60,000 Travel Subsistence Total costs cost 6,000 3,000 69,000 I Information-systems specialist to program and test the system and train technical staff Infornation technologies 180 600 108,000 6,000 6,000 120,000 I Auditor 60 500 30,000 6,000 3,000 39,000 3 Technicians in charge of updating records 180 40,000 Review of pension funds I specialist on pension management regulations funds reserves management Study on the viability of I specialist on contractual Contractual savings savings 25 800 20,000 6,000 2,000 28,000 30 800 24,000 6,000 2,000 32,000 Audit of Government's Audit firm arears with the pension fnds Auditof administrative Audit firm expendituresof OPS Actuarialanalysis of the Actuary iealth and family services pranches 30 800 24 000 6 000 2 000 32 000 20 800 16,000 6,000 2,000 24,000 30 800 24,000 6,000 2,000 32,000 Grand total 416,000 Source: Authors calculations 63. Implementation costs. The Government has already committed US$69,000 for the current audit of the information system. Additional resources to complete reforms to improve institutional capacity could approximate US$347,000 (see Table 11). The resources of the PATARE project, prepared for technical assistance, could finance part (50%) of the activities. Mobilization of additional resources, preferable grants, would be necessary. The costs of the study to audit the current estimate of Government's arrears and define a fiscally sustainable refinancing plan could approximate US$100,000. The European Union (EU) has manifested its interest in financing this study. The EU is currently financing a study on Government's arrears with the private sector. In terns of the costs of evaluating and implementing parametric reforms, the Government has already spent US$150,000 in a study to retrieve and update necessary legal texts. The actuarial analysis of the impact of alternative parametric reform, on the other hand, has been in charge of the World Bank. No substantial additional financial resources are expected. The other activities (e.g., creation of the NCSS) will mostly require resources in terms of Government staff time. 32 V. POLICY MATRIX"1 Policy Objectives Actions Unit Date of Implementation Responsible A. Improvin Governance and Management Capacity * * Improve governance Improve institutionalcapacity Set up the National Committee November 2001 for Social Security (NCSS) in charge of coordinating the reforms across pension funds. MOFP/MLNS Prepare a multiyear plan to merge OPS and CNR January 2002 NCSS Implement new management and information system. June 2002 NCSS Update and audit databases of contributors and beneficiaries. June 2002 NCSS Study to evaluate appropriate level of reserves and identify mechanisms to improve reserves management. June 2002 NCSS Audit of the administrative expenditures of OPS and actuarial analysis of the health June 2002 NCSS and family services branches. B. Improve the short-term financial situation of the funds * * Adoption of a sustainable plan for refinancing arrears with the pension funds in the context of the global arrears settlement plan. Regularization of retirement and government's contributions to the pension funds. Inventory and audit of arrears and cross-debtsbetween the Government and the pension funds. January 2002 MOFP/NCSS Adoption of a refinancing plan of governmentarrears which allows to cover the short-term deficit of the pension funds. November 2001 MOFP/NCSS Retire all eligible workers (liquidations), in particular the 773 contractual employees who have delayed retirement. January 2002 MOFP/NCSS Ensure the regular payment of government contributions. November 2001 MOFP/NCSS The policy interventions/actionspresented in this matrix, are not necessarily board conditions for the StructuralAdjustment Credit that Djibouti is currently preparing. t 33 Policy Objectives Actions Date of Unit Implementation Responsible C. Improve medium-termfinancial sustainability Parametric reform to improve the medium-term financial sustainability of the pension funds. Adoption by the Government November 2001 of a plan to gradually adjust vesting periods, the retirement age, and the accrual rate of CNR and OPS, on the basis of proposed scenarios. MOFP/MLNS/ NCSS Introduction of a minimum pension. November 2001 MOFP/MLNS/ NCSS Introduction of a maximum replacement rate in CNR and elimination of the practice of cumulating benefits in multiple funds. November 2001 MOFP/NCSS D. Prepare the Ground for Medium-TermReform * Identify and implement incentives to stimulate the development of contractual savings Conduct a study to of current contractual savings arrangements and identify institutional and financial constraints to their development. Present recommendationsin terms of the required regulatory framework and incentives. December 2002 NCSS 34 Documents used to prepare the Pension System Reform - Strategic Note Averting the Old Age Crisis (A World Bank Policy Report) Budget de l'Etat pour l'Exercice 2000: Republique de Djibouti Country Assistance Strategy, Memorandum of the President - The Republic of Djibouti (World Bank Document) Education Sector Strategy Note - The Republic of Djibouti (World Bank Document) Interim-PovertyReduction Strategy Paper and Joint IDA-IMF Staff Assessment - The Republic of Djibouti New Ideas about Old Age Security toward Sustainable Pension System in the 2 1lt Century (The World Bank, Washington, DC) Recueil des Textes de Securite Sociale - Republique de Djibouti: R6forme des Regimes de Retraite Djiboutiens : Reforme Institutionnelle- Republique de Djibouti R6sultats de l'Enquete Demographique Intercensitaire : Republique de Djibouti: (Mars 1991) Un Carrefour dans la Come de l'Afrique - Evaluation de la Pauvrete 35 VI. TECHNICAL APPENDIX Al. A model of growth to ensure internal consistency in macroeconomic projections We operate along a growth path that maximizes inter-temporal social welfare. We postulate that social welfare can be approximated by a value function V(.) which depends on consumption per capita. Given a constant risk-aversion utility function and a Cobb-Douglas production function, the inter-temporal optimization problem to be solved can be written as: max: V(C,) (+ = r)T- {N 1 (N ,' } (1) s.t. Y = (A, N, ); 8K, K,+I = K,(1- 3k) +(Y - C,) logA, =logA,- + (2) Yaexp(-45.t)+o-YaEI where r is the discount factor, C is consumption, r captures consumers risk aversion (i.e., defines the curvature of the utility function), Y is GDP, K is the stock of produced capital, N is labor, gk is the depreciation rate of capital, A is labor productivity, Ya is the growth rate of labor productivity that decays at a rate S., E are random shocks, and ca is the variance of these shocks. The solution to this problem can be approximated by: K,+ - K,I -(1_k) i, [A ln(K,+, /N,+I)= a, , + a2 (3) (ln(K, / N,)- In(A) where a, and a2 are functions of ik, 0, the rate of time preference (r), and the coefficient of risk aversion (T). It can be shown that the coefficients al and a2 are given respectively by: a, = Ya -a 2 lnk and f. = exp(ya+ Yn) + a2 KCkk + Kkc(k f2 = Vexp(y. + v) + Kkk f3 = 2exp(y- + yn) Kk - f2 )/f 3 -1, where: / T) + KCC (Kk / T)- 4 exp(y. +,Y,)Kkk =-ck KCk = Kkk = (f (I + p)- 1 ( - 1)60-1 exp(- yar) = (I + p)exp(y,r) 36 c +(k (I -a - exp(yv. + YX K(1+ p)exp(ar.z)- (I -,5k ))10]0'- The model was first calibrated to reproduce an exogenously defined steady state growth rate of 3% per year (the current IMF/World Bank forecast). We then used the model to explore the growth impact of changes in parameters. Economic growth rates are more sensitive to changes in the growth rate of the labor force, labor productivity, and the elasticity of output with respect to capital. We notice that a given interest rate, r, can accept alternative steady-state growth rates, depending on the value of the other model parameters. For simulation purposes, two scenarios were retained: a low-growth scenario where GDP grows at 2% per year, and a high growth scenario where the economy expands at 4% per year (see Table 1). The difference in growth rates is given by expectations about productivity growth. In the case of the high growth scenario labor productivity would need to grow at 4% per year. Table 1: Macroeconomic Scenarios Parameters DeltaK DeltaA DeltaN LambdaA LambdaN R Thau Theta Alfal Alfa2 COR Implied average economic growth rate for the period 2002-2075 Source:Authorscalculations Scenario Change Low growth High growth in parameter 0.04 0.04 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.015 0.03 0.01 0.015 0.015 0.01 0.05 0.05 0.01 0.8 0.8 0.1 0.4 0.4 0.1 0.241237 0.25814 -0.07146 -0.11186 4 4 1 2% Sensitivity Change in growth rate -0.280% -0.110% -0.130% 0.730% 0.940% -0.280% -0.040% 0.860% -0.650% 3% A2. Simulatingthe financial position of the pension funds Core simulations were conducted using the World Bank PROST. To improve the short-term accuracy of the projections, flows of contributors and retires are based on transition probabilities between states, instead of population shares. In this mode, PROST requires the following data: i) the distribution by sex and age of the population of contributors and retirees of a given system; ii) the distribution of wages by sex and age; iii) the distribution of pensions by sex and age; vii) the average wage and the total pensions paid in the base year; iv) average number of years of contribution at retirement; v) mortality rates; and vi) retirement probabilities for different ages. The methods and sources used to generate the necessary information are described in this section 37 for OPS and CNR. We emphasize that the data received from the pension funds are of poor quality. Records remain incomplete and have not been audited. Assumptions at different levels have been therefore necessary. While the major recommendations of the analysis appear robust with respect to changes in these assumptions, the results of the simulations should be interpreted with caution. The case of OPS Data. The model for OPS was calibrated on the basis of three datasets: contributors from the private sector and state enterprises (18,116 records after checking for consistency); contractual workers (3,068 records after checking for consistency); and beneficiaries (2,878 records after checking for consistency). Sex and age distributions of contributors and beneficiaries. Since the databases are incomplete, we assumed that missing records were random and used available records to derive the distribution of contributors and retirees by sex and age (see Figure 1). These distributions were then applied to the "correct" populations (see Table 1 in the main text), derived from OPS accounting system. Figure 1: Age distribution of contributors and beneficiaries .1C2711.190402 0 la 20 3'0 4 50 60 - 70 e0 0 90 50 0Go 70 au 90 Source: Authors calculations givenOPS records Wage distribution. Since PROST does not allow for differentiation between workers from the private and pubic sector, individual records were merged to estimate a joint distribution of wages (see Figure 2). The average wage for contractual workers (DF 658,975 per year) represents 55%of the average wage for workers in the private sector and public enterprises (DF 1,190,256 per year). Current distributions are modified over time to ensure that the steady-state distribution is upward sloping in the support [15-65]. The implicit assumption is that in the steady-state, wages are proportional to labor productivity which increases up to age 65 and decreases afterwards. 38 Figure 2: Wage distribution by age and sex 1500/D . ............ ........ . ..... .......... 140%/- ~ 130%-.............. E 120% - o 110% 100% iE. Males(2000) __ - ___ a90% -a-t Mates (2075) _ R( I' 80% ________ ___ Females t _ (2000) _ ._______ --- - Females(2075) 70% a 6%50% -__ 15 35 -_- 55 75 Source: Authors calculations given OPS records Pension distribution. The database of beneficiaries was used to derive the distribution of pensions by age (see Figure 3) and income decile (Figure 4). The latter is needed to evaluate the impact of introducing a minimum pension. We observe that the distribution of pensions is highly unequal. The average pension received by the first decile represents 10% of the average pension received by I 0h decile. We use this distribution, to estimate the number of individuals who would have a pension below an exogenously defined minimum. The distribution is also used to compute revenues resulting from a tax on pension above a given minimum. The distribution is assumed to be constant over time. Indeed, attempting to project shifts in these distributions with the type of data available, would simply obscure the results of the actuarial analysis. Fi!ure 3: Pensions distribution by age 140%/ 120% __ i: - _ wo> 20% _ 55 __ 60 65 70 __ 75 80 85 90 Age Source: Authors calculations given OPS records 39 Figure 4: Pension distribution by decile 90.0% 80.01% 0 ! 70.0% _/ _ _ - _ 60.0%___,C 50-0% o 4000% * 30.0%/o 20.0% - - X 10.0% 0.0% 0% - - -_ _ 20% 40% 60% 80% 100% Populationshare Source: Authors calculationsgiven OPS records Total pensions paid and total contributions. Since the individual records are incomplete, they cannot be used to compute total revenues and total pensions paid. Thus, the income statement of OPS was used to derive the structure of revenues and expenditures. In the model, the collection rate of the system was adjusted to replicate the current level of revenue. With an average covered wage of DF 1.1 million per year, the system could receive revenues of DF 2.5 billion. However, the Government is not remitting contributions equivalent to DF 217.4 million and evasion in the private sector costs the system an additional DF 1.2 billion (the distribution of these arrears between state enterprise and private sector firms is unknown). Hence, the effective collection rate is only 42%. In terms of pensions, PROST internally adjusts the average pension to replicate observed expenditures. Demographic projections. Demographic projections have been developed by the World Bank Health Sector Anchor on the basis of past mortality data (the last census in Djibouti was conducted in 1990). Figure 5 displays life expectancies and the implied population growth rate, which decreases as fertility rates drop faster than mortality rates. New mortality rates and their future dynamics are supposed to take into account the impact of the FHV/AIDS epidemic. Nonetheless, given high levels of uncertainty in terms of the current HIV prevalence rate, the projections and associated results should be interpreted cautiously. Flows of contributors and beneficiaries. A critical determinant of the financial situation of the system are enrollment and retirement pattems. In the case of contributors, the population of each age group at time is given by: r C = (Cai (1 -ra) M(Ca-, ,a,t)) +±sg , NC, , (4) where r, is the share of contributors of age a who retire, M(.) is the mortality function, g is a parameter defining the growth rate of the stock of total contributors (NC), and sa is the share of new contributors of age a. 40 Figure 5: Life expectancies and population growth rate Males Females LibExpecancy: At B.......................... irth - 80.0 70.0 60.0 _ 50.0 .- *- UbExpectaicy:Al E& Al Age 20 AlAge60 000 AtAge60 AtAge65 50.0 tAger65 40.0 400 30.0 30.0 20.0 ............. ....................... ... .............. 70.0 -u-AtAge 2D _ - 80.0 20.0 - tO.O 0_ 2000 10.0' E< 2020 2040 2000 2060 2020 2060 2060 lAl 1.2% .4% 1.0% .2% 2001 2011 2021 2031 2041 2051 2061 2071 Year Source: Authors calculations given OPS records Similarly, in the case of old age pensioners, the population at each age group a is given by: Ra =(Ra,i - M(Rai-))+raCa-, (5) Unfortunately, in the case of Djibouti, there are no data available to rigorously estimate the parameters g,, s0 and r0 . In order to generate plausible enrollment and retirement patterns, we have operated under the following assumptions. Expanding coverage (g, ). We consider two scenarios. A low-case (which is associated with the low growth scenario, see Section Al) assumes that within the next 15 years the number of covered workers will grow\ very slowly, at 0.5% per year (below the population growth rate of 1.6%). Under this assumption, during the period 2001-2015 the share of the population aged 15 to 55 covered by the system decreases to close to 7%. This scenario is consistent with current low levels of coverage, which suggest that historically the system has expanded slowly. The high-case, on the other hand, considers that the population of contributors grows at 2% per year (above the population growth rate). Thus, the share of the population aged 15 to 55 covered by the system remains close to 8.5%. 41 Figure 6: Age compositionof new enrolled in OPS (Contractual workers) and average enrollment age by year 12 c6 4 0. 2 0 10 20 30 40 50 60 70 Age 1 00~~~0 Sl29.34?j 1980199 year Source: Authors calculations given OPS records Age distribution of new contributors. We also assume that the distribution of new retirees is triangular in the support [15,40] and uniform in the interval [41,50]. Moreover, we impose the constraint that the cumulative distribution in the interval [15,40] be equal to 85%. We base these assumptions on data for contractual workers. For this group, 86% of all workers enrolled before the age of 40. The mean enrollment age has been declining over time (see Figure 6), from 50 in 1980 to 30 in year 2000. In the case of workers from SOE and the private sector, data on age of enrollment is not available. Nonetheless, the individual records suggest that contributors are relatively young. Today 50% of contributors are less than 35 years old and 65% are less than 40 years old. Retirement. The individual records for private sector/SOE workers show that 95% of workers are bellow age 55 and that 99% are below age 60. This suggests that the majority of workers retire between ages 55 to 60. Retirement probabilities for these workers were approximated using 2000 data (see Figure 7). The data show that 65% of new retirees in year 2000 were aged 55 to 60. Moreover, 99% of the contributors are aged 65 or less. We assume that gradually that the probability of retirement after age 55 increases gradually, reaching 100% in year 2020. 42 Figure 7: Retirement patterns for private sector/SOE workers I 0.9 -_/ 0.8 0.7 L -_ _ 0.2 0.1 0 55 57 59 Age 61 63 _ 65 Source:AuthorscalculationsgivenOPSrecords In our simulation we also take into account the special case of contractual workers who have delayed retirement despite verifying retirement conditions. There are 773 workers in this situation (see Table 2). Table 2: Distribution of contractualworkers who have delayed retirement Yearly pensions USD 919,222 123,624 102,865 55 54 65,957 Total 1,211,668 Late contributions USD Total wages per year (1995-1998) 2,785,401 2,506,862 Total wages per year (1998-2000) Total contributions 3,127,728 Source: Authorscalculationsusingdatabaseof contractualworkerspreparedby Ministerof Finance. Age >56 56 Number 571 84 68 50 Retire 2001 2001 2001 2002 Vestingperiods. To explore the relationship between retirement age and contribution years we use the database of retirees, which keeps track of retirement age and vesting periods. The data show that among workers who retired in year 2000 there was an inverse correlation between age and contribution years (see Figure 8). While a 35 year old would have contributed, on average, 20 years, a 60 year old would have contributed 18. In the simulation we use the 2000 distribution of contribution years by age. Based on the data, we also assume that the average contribution period will increase by 0.5 each year (see Figure 8). We impose the constraint that vesting periods for all ages converge to 25 years by 2030. Since the retirement age remains at 55 and retirement probabilities converge to one for ages above 55, this implies that on average workers will enroll OPS at age 30. 43 Figure 8: Vesting periods and retirement age Y-bb ~~~~~~~~18.7625 FlftdVAM _ _ 15- 0 0 A.227 0 0 Source:AuthorscalculationsgivenOPSrecords When changing required vesting periods, retirement probabilities are affected as the share of workers meeting the new condition decreases within each age category. As an illustration of this effect, Figure 9 presents the reduction in the number of new retirees among contractual workers (the figure was derived under the assumption that all workers meeting retirement conditions retire) resulting from an increase in the vesting period from 15 to 25 years at an annual rate of one year and two years respectively. We observe that the percentage reductions are sizeable (in the order of 20-50%) and increase over time. Figure 9: Changes in the absolute number of new retirees resulting from an increase in the vesting period from 15 to 25 0.000% 1938 -10.000%- 200 2002 2004 2006 30.000% 2010 _I_ , -20.000% 20 e 2008 2 2 e 2 \ 1 year peryear speyear ____ -40.000% -50.000% -60.000% To handle this problem we assume that vesting periods across ages are normally distributed. The probability that a worker of age a ill retire is given by: pr(R, = 1) = D(v;'aF., oa )IP(a), (6) where R&=1indicates retirement, >(.) is the cumulative normal density with mean and standard deviation Va,or, that gives the probability that a worker has contributed at least v years (v is the vesting period), and T'(a) is the unobservable probability of retirement given that retirement 44 conditions are met. Under these assumptions, the growth rate of the probability of retirement can be expressed as a change in the vesting period and is independentof T (a): apr(R, pr(Ra =1) a(v;V,,aj =1) aT(a) D(V;V.a,a0a)'P(a) C-((V; Va"ga) (D(v;V,) ' (7) Using data on contractual workers (excluding those who delayed retirement) we estimated the standard deviation of vesting periods by age (see Figure 10). We can then infer for each year the share of workers who would have retired under the "old" vesting period, who cannot retire under the "new" vesting period. Figure 10: Standard deviation of contribution periods by age group 04.5 0 X s 01. *~0.5 0 <55 55-65 >65 Age groups Source: Authors calculations given OPS records The case of CNR Data. We used individual records for contributors and beneficiaries. Given the quality of the data, we limited our analysis to the regime for functionaries which accounts for 66% of CNR's total contributors and 42% of its retirees. We have included individuals within categories B to E and excluded individuals within category A equivalent to ministers and parliamentarians. Sex and age distributions of contributors and beneficiaries. As in the case of OPS, we assumed that missing records in the database of contributors were random and used available records to derive the age distribution (see Figure 11). Given that the field sex is not available in the database, the distribution by sex was not derived. In the case of retirees, the fields sex and age are missing. Hence, we used as a proxy of the current age distribution the steady-state distribution given past and current mortality rates and assuming that all new retires are 55 years old (this seems to be the case as indicated by CNR administrators). Figure 12 summarizes the resulting age distributionof the populations of retirees. 45 Figure 11: Age distribution of the population of contributors 0.14.-. 0.12 ~0.1 LI0.08 . _ _ __ 0.06 c*, 0.04- 0.02 0 0 10 20 30 40 60 50 70 Source: Authors calculations using individual records Figure 12: Age distribution of the population of retirees 0.05 0.045 0.04 LI 0.035 5 0.03 0.025 'CD 0.02 55 t 60 65 70 75 80 85 b~~~~~~~Ae usingindividualrecords Source: Authorscalculations Wage andpension distributions. In the case of contributors we merged the individual records to a database containing the index wage by category, class and grade. Since the nominal wage is proportional to the index, we simply looked at the distribution of the latter. The adjusted distribution is summarized in Figure 13. Given that the Government is in the midst of a fiscal stabilization program, we have assumed that nominal wages will not be adjusted within the next 5 years. Afterwards, wages are assumed to growth in real terms with labor productivity (see Section Al). In the case of retirees, in the absence of information about age, we used the same distribution of pensions by age than for OPS. The important element is to consider that the average pension is lower among older retirees, given that their wage at retirement is lower. Using a uniform distribution would have been another alternative, but since the mortality rates of old pensioners are higber, this assumption would have biased down-wards total expenditures. 46 Figure 13: Adjusted wage distribution by age for the population of contributors 200% _ 180% ¢ . 100% I - - : _ 160% ZD140%-_ 120% __ *___,_ 80% _ ___j '~60% _ 40%_ _- 20% 20 30 40 50 60 70 80 _ 90 Age Source: CNR individual records Total pensions and total contributions. Data on total contributions and total pensions paid come from CNR's income statement. The average wage of the functionaries was computed by dividing the average contribution per worker by the contribution rate (18%). The average pension for old age pensioners was computed by dividingtotal pensions by the total number of retirees. Flows of contributors and beneficiaries and vesting periods. We use the same model as in the case of OPS. The implementationin the case of CNR, however, is simpler. First, we assume that no new hiring will take place until 2006. From 2006 onward, the stock of functionaries remains constant. Second, we assume that new functionaries enter CNR at age 27. Finally, we assume that all functionaries retire at age 55 and that on average have contributed to the system 25 years. In the case of the current contributors older than 55 years, we assume that the probability of retirement is 30% (this probability was calibrate to preserve the trend in terms of new retirements, close to 80 per year). These simplifying assumptions are required given the quality of the data. Nonetheless, since in the case of CNR the parametric reform concentrates on changes to the basic and marginal replacement rates, these assumptions are appropriate to make a first evaluation of their impact. A3. Deriving equilibrium contribution rates and replacement rates in funded and unfunded systems Funded system This section proves the following proposition. There is a well defined linear relationship between equilibrium contribution and replacement rates which depends exclusively on labor productivity growth (g), the contribution period (T), the retirement period (L), and the interest rate (r). A higher contribution period, a lower retirement period and a higher interest rate will always reduce the contribution rate necessary to match a given replacement rate. The effect of laborproductivity growth, however, is ambiguous. 47 Derivation The total contribution accumulated by a given worker during a period of T years is given by: T (8) C = fawoet(rr)+r, ,=o where a is the contribution rate, g is productivity growth, r is the interest rate applied to the contributions and w0 is the worker's wage when he/she starts to make contributions. On the other hand, the present value of the total pension received by the worker when he/she retires is given by: L P= fpoe-r (9) ,=o where p0 is the pension that the worker receives and L is the number of years that the worker receives a pension. Let's assume that the pension is computed on the basis of a replacement rate, R, which applies to the salary in year z<T, so that (10) Po = tR.wOe, By integrating (1) and (2) and rearranging we derive the following relationship: a = O(g,r,T,L)R.eg, where 0(g,r,T,L)= ( g-r)e(rL reT(g-r)+rT - (11) 1i) rT A higher interest rate, a higher contribution period and a lower retirement period will always reduce the contribution rate necessary to match a given replacement rate. The effect of economic growth, -= ag however, Ra () e is + e 'z ambiguous. Indeed, we notice that while () ag < 0, can be positive depending on the value of z. Basically, if the Og) replacement rate applies to a wage which is close to the workers final wage (i.e., z is close to T), then labor productivity growth increases the expected pension and therefore the contribution rate. Unfunded system In the case of a non-funded system, the equilibrium contribution rate depends on the dependency ratio (number of retirees per contributors) and the average replacement rate (the ration between the average pension and the average wage). We have: 48 c =Bp = d.R, W.W (12) where c* is the equilibrium contribution rate, B is the total number of beneficiaries, p is their average pension, W is the total number of workers, w is their average wage, d is the dependency ratio and R the replacement rate. As a non-funded system matures, d increases and financial sustainability requires either a higher contribution (c*) or a lower replacement rate (R). A4. Changes in the accrual rate, minimum pensions, and incentives to retire One of the measures that would accompany the parametric reform is the introduction of a minimum pension. An issue to consider in this context is whether this minimum pension generates incentives for early retirement that may have negative impacts over the financial situation of the fund. In this section, in order to derive some basic insights, we formalize and solve the problem of deciding the optimal year of retirement within a defined benefit system (after the vesting period has been completed) and analyze how changes to the system parameters, including the introductionof a minimum pension, affect the decision. For simplicity, we postulate that at the end of the vesting period individuals choose the retirement year in order to maximize the present value of revenues during their remaining lifetime. The present value of this stream of revenues is given by: PV(R)= [wo(1+g)(1-c)p ]+ E[(W 0 (1+g) (b+aT)+rw 0 (1+g)')p'],(I3) where T is the number of years the individual will continue to work after the vesting period has elapsed, w0 is the wage at time 0, g is the expected average growth rate of the wage, c is the contribution rate to the pension fun, b is the basic replacement rate (the replacement rate at the end of the vesting period), a is the accrual rate, r is the share of the original wage that could be received in the form of a non-taxable income flow after retirement (for example by working in the informal sector), L is life expectancy, and p the discount factor. The first summation gives the present value of the revenue the individual receives while working (between time I and time T), and the second summation gives the present value of the income that the worker receives after retirement. This formulation assumes no ceiling in the replacement rate. Depending on the value of g, c, a, b, and r, the function linking PV(R) to T can take four basic shapes (see Figure 14). Shape 1: The function increases constantly so that the worker has never the incentive to retire (he/she would retire only if there is a physical impediment to continue work or a mandatory retirement age). Shape 2: the function decreases constantly so that the work would have the incentive to retire as soon as the vesting period has been completed. Shape 3 (U shaped): the function first increases, then decreases, so that there is an optimal retirement year. Shape 4 (inverted U shape): the function first decreases then increases so that the individual is better off by not retiring (as in shape I). 49 Figure 14: Total revenue and year of retirement ,19 e' 17 _- !15 13 g ___- 11 _ Shape1 (b=0.3,a=O.a2,c=O.20,r=O) 9g a _ _ -a-+ Shape2 (b=0.6,a=0.02,c=0.20,r=0.5) Shape 3 (b=0.3,a=0.02,c=0.20,r=0.5) 7 5 ___________________________ 0 5 _ 10 , 15 -Shape 4 (b=0.6,a=-0.02,c=0.20,r=0.5 20 25 30 Number of yearsaftervestingperod Wage at the end of vesting period (w) is set equal to 1. Growth rate of wages (g) is 2% per year. Source: Authors' calculations In the case of Djibouti, where b=0.30, a=0.02, c=0.20, and r is expected to be low, the curve will most likely adopt shape 1 (which implies that workers will have an incentive to delay retirement as much as possible), or shape 3 (where workers retire a few years after completing the vesting period). The impact of a minimum pension on the retirement date. Consider the case of an individual who, after the vesting period has elapsed, would receive a pension below a certain minimum P*. This implies that b*wo<P*. Moreover, assume for illustrationpurposes that P*/ wo =0.60 so that in order to receive the minimum pension this individual would need a replacement rate of 60%. Now consider the introduction of this minimum pension. The curve facing this individual will most likely adopt shape 1. However, if the expected growth in wages is low and the replacement rate to attain the minimum pension and/or the ratio between non-pension related income after retirement and income prior to retirement (r) is high, the curve will adopt shape 2. In this case, the minimum pension provides an incentive for early retirement (see Table 3). If individuals, however, expect higher wages, then the income curve will become S-shaped (see Figure 15) and again workers will have an incentive to wait. In the case of Djibouti (OPS), current retirement patterns and low wages in the infortnal sector suggest that the average r is low. If structural reforms stimulate growth, and thus expected wages, it is unlikely that the minimum pension would create incentives to retire as soon as the vesting period is over. However, in order to avoid the risk of an outbreak of low-income early retirees, the minimum pension could be linked to a higher vesting period. Table 3: Minimum pension, reservation income, and early retirement Reservation wage Replacement rate to attain a minimum pensior 0.2 0.4 0.5 0.4 Shapel hape. Ihope 0.5 Shape Sha el ha e 0.6 Shape IShape Shape2 0.7 Shope , hapeShape Source: Authors' calculations 50 Figure 15: Present value of income with a minimum pension and fast growth ! $20000 $10.052° $14000 $&000 $SO1.OOO $2000O 0 5 10 15 20 25 30 Tnme aftr vedngperiod Source: Authors' calculations 51