Republic of Djibouti Country Economic Memorandum

eotN.315D Republic ofDjibouti Country EconomicMemorandum Volume II 33115-DJ No. Report Report No. 33115-DJ Republic of Djibouti Country Economic Memorandum

Unlocking Authorized Authorized Disclosure Disclosure Public Public Djibouti’s Growth Potential: The Road Ahead

(In Three Volumes) Volume II: Main Report

August 10, 2006

Social and Economic Development Group Middle East and North Africa Region Public Disclosure Authorized Authorized Disclosure Disclosure Public Public Public Disclosure Authorized Authorized Disclosure Disclosure Public Public

Document of the World Bank Public Disclosure Authorized Authorized Disclosure Disclosure Public Public

REPUBLIC OF DJIBOUTI Country Economic Memorandum Table of Contents

PART I. GROWTH ANALYSIS ...... 1 CHAPTER 1: GROWTH PATTERNS. SOURCES. AND CONSTRAINTS ...... 2 1.1 Growth Patterns ...... 2 1.1.1. Overall Growth Performance: Stylized Trends ...... 2 1. 1. 2. Volatility of Growth ...... 3 1.1.3. Patterns of Growth and Recent Developments ...... 5 1.2. Sectoral composition of GDP growth: the supply side ...... 8 13 .Demand composition of real GDP growth...... 10 1.4. Sources of and constraints on economic growth ...... 11 1.4.1. The roles of capital accumulation, labor and total factor productivity ...... 11 1.4.2. Long-run sources of growth: incorporating a cross-country perspective ...... 13 1.5. Bottlenecks to growth and private investment: a closer look ...... 17 1.6. Conclusion and policy implications ...... 26

PART I1. DJIBOUTI’S GROWTH REFORM AGENDA: THE ROAD AHEAD ...... 27 CHAPTER 2 .MACROECONOMIC AND FISCAL STRATEGY ...... 28 2.1. Medium-term economic outlook ...... 28 2.1.1. What growth can be expected in the coming decade? ...... 29 2.2. Reforming government pay and employment policies: evaluating options for reform...... 32 2.2.1. The Public Sector Wage Bill: The Magnitude ofthe Problem ...... 33 2.2.3. Policy instruments for reducing the wage bill ...... 39 2.2.4. Reforming government pay and employment policies: options for reform ...... 41 2.3. Conclusion and policy implications ...... 58

CHAPTER 3. MAKING THE LABOR MARKET AND THE INVESTMENT CLIMATE WORK FOR GROWTH AND JOBS ...... 59 3.1. Main features of Djibouti’s labor market...... 59 3.2. Labor market outcomes: stylized facts and recent trends ...... 63 3.3. Explaining poor labor outcomes ...... 73 3.3.1. Wage Distortions in the Labor Market: Do Public Sector Employees Earn a Rent? ...... 73 3.3.2. Investment climate and job creation ...... 81 3.3.3. Labor market regulations ...... 88 3.3.4. Weak role ofpassive and active labor programs for the unemployed ...... 93 3.3.5. The supply-side: the poor quality of labor ...... 96 3.4. Conclusion and policy implications ...... 98 CHAPTER 4 . MOBILIZING FINANCE FOR INVESTMENT ...... 101 4.1. The currency board arrangement ...... 101 4.2. Djibouti’s financial sector ...... 105 4.2.1. Main features ofthe banlung sector ...... 105 4.2.2. Bank lending trends ...... 108 4.2.3. Overall soundness ofthe banking system ...... 110 4.2.4. The money changers ...... 111 4.2.5. Informal finance mechanisms ...... 112 4.2.6. Microfinance ...... 113

i 4.3. Financial sector reforms ...... 115 4.4. Monetary and banking developments ...... 1 18 4.5. Conclusion and Policy Implications ...... 123

CHAPTER 5 .TOWARD AN EFFICIENT INFRASTRUCTURE ...... 125 5.1. Transport Sector ...... 125 5.1.1. The Port ...... 126 5.1.2. Railway ...... 138 5.1.3. Djibouti International Airport ...... 141 5.1.4. Roads ...... 143 5.2. Power and Water Sectors ...... 151 5.2.1. Electricity ...... 15 1 5.2.2. Water ...... 153 5.3. Telecommunications ...... 156 5.4. Conclusion and policy Implications ...... 158

LISTOF FIGURES Figure 1.1 Djibouti vs Comparators: Real GDP Growth Rate ...... 2 Figure 1.2 Djibouti vs Comparators: Real GDP per Capita in constant 2000US$. 1980-2004 ...... 3 Figure 1.3 Djibouti vs Comparators: Real GDP per Capita Ratios. 1980-2003 ...... 3 Figure 1.4 Djibouti vs Comparators: Growth Volatility. 1981-2004 ...... Figure 1.5 Real GDP Growth Rate vs Volatility. 1986-2004 ...... 3 Figure 1.6 Real GDP Growth in Djibouti and Yemen ...... 4 Figure 1.7 Response ofDjibouti’s real growth rate to a one std. dev shock in Yemen’s growth rate...... 4 Figure 1.8 Response ofYemen’s real growth rate to a one std . dev shock in Djibouti’s growth rate...... 4 Figure 1.9 Actual vs . Potential Real GDP Growth. 1981-2004 ...... 4 Figure 1.10 Sectoral Shares in Real GDP. 1990-2004 ...... 8 Figure 1.11 Average Annual Expenditure Growth. 1990-2004 ...... 10 Figure 1.12 Savings vs Investment as Share ofGDP ...... 10 Figure 1.13 Labor and Capital Productivity Growth. 1990-2002 ...... 12 Figure 1.14 Evolution of Total Factor Productivity. 1990-2002 ...... 12 Figure 1.15 Contribution to Average Growth: Djibouti vs Sub-Saharan Africa ...... 15 Figure 1.16 Contribution to Average Growth: Djibouti vs Mauritius ...... 15 Figure 1.17 Contribution to Average Growth: Djibouti vs Tunisia ...... 16 Figure 1.18 Growth That Could Have Been. 1970-1999 ...... 16 Figure 1.19 Growth That Can Be...... 17 Figure 1.20 Political Stability. Growth and Economic Volatility ...... 18 Figure 1.21 Foreign Direct Investment Share in GDP vs Political Stability ...... 18 Figure 1.22 Private Investment Share in GDP vs Political Stability ...... 18 Figure 1.23 Public Investment Share in GDP vs Political Stability ...... 18 Figure 1.24 Real Effective Exchange Rate and Real GDP Growth ...... 19 Figure 1.25 Exchange Rate Developments (1980 - 2002) ...... 19 Figure 1.26 Real Effective Exchange Rate and Fiscal Balance as Share of GDP ...... 19 Figure 1.27 Real Effective Exchange Rate and Current Account Balance as Share of GDP ...... 19 Figure 1.28 Average Monthly Civil Servant Wages ...... 19 Figure 1.29 Average Hourly Wages by Job Type ...... 19 Figure 1.30 Interest Rate Spreads ...... 20 Figure 1.31 Domestic and Foreign Interest Rates (percent). 1990-2002 ...... 20 Figure 1.32 Djibouti vs Comparators: Average Telephone Rates ...... 20 Figure 1.33 Djibouti vs Comparators: Average Water Tariffs ...... 20 Figure 1.34 Djibouti vs Comparators: Average Electricity Tariffs ...... 20 Figure 1.35 Fiscal Impulse and Output Gap ...... 21 Figure 1.36 Components ofOverall Fiscal Balance ...... 21 Figure 1.37 Components ofPrimary Fiscal Balance ...... 21 Figure 1.38 Components ofAdjusted Fiscal Balance ...... 21

.. 11 Figure I.39 Composition ofGovernment Total Expenditure. period average. 1990.2004 ...... 21 Figure 1.40 Current and Capital Expenditure as % of Total Expenditure. 1990-2004...... Figure 1.4 1 Capital Expenditure. 1990-2004 in millions DF & trend line...... 22 Figure 1.42 Revenue composition ...... 22 Figure 1.43 Revenue (including grants). as a percent of GDP. 1990-2004...... Figure 1.44 Djibouti vs Comparators: Government Effectiveness...... 23 Figure 1.45 Djibouti vs Comparators: Rule of Law ...... 23 Figure 1.46 Growth Diagnostics: Low Private Investment ...... 24 Figure 2.1 Wage Bill. 1990-2004...... 33 Figure 2.2 Average Public Sector Wage vs . GNI per capita. 1995-2004...... 34 Figure 2.3 Relative Weight of Public Sector Wages. 2004-2005 ...... 34 Figure 2.4 Ratio of average public sector wage to per capita GNI...... 34 Figure 2.5 Monthly Wage Distribution in Public Sector. 1995 ...... 34 Figure 2.6 Wage share bill as percent of total wage by Ministries. 2002-04 ...... 36 Figure 2.7 Age Structure of Civil Servants. 1996 ...... 39 Figure 2.8 Age Structure of Male Civil Servants. 1996 ...... 40 Figure 2.9 Age Structure of Female Civil Servants. 1996 ...... 40 Figure 2.10 to 2:49 Figures related to Impact on Public Wage Bill ...... 45 Figure 2.50 Income distribution. 1996 ...... 55 Figure 2.51 Share of Public Wages to Total Wage Income Against Per-Capita Expenditure. 1996 ...... 55 Figure 3.1 Labor Force Participation Rate by Age Groups. 1996 ...... 64 Figure 3.2 Labor Force Participation Rate by Age Groups. 2002 ...... 64 Figure 3.3 Male Labor Force Participation Rate by Age Group. 1996-2002 ...... 65 Figure 3.4 Female Labor Force Participation by Age Group. 1996-2002...... 65 Figure 3.5 Labor Participation Rate by Location. 1996-2002...... 65 Figure 3.6 Male Labor Participation Rate by Location. 1996-2002...... 65 Figure 3.7 Female Labor Participation Rate by Location. 1996-2002 ...... 65 Figure 3.8 Education and Labor Market Participation. 1996 ...... 66 Figure 3.9 Unemployment Rate by Location. 1996-2002 ...... 66 Figure 3.10 Male Unemployment Rate by Location ...... 66 Figure 3.11 Female Unemployment Rate by Location ...... 66 Figure 3.12 Unemployment Rate by Age Groups. 1996 ...... 67 Figure 3.13 Unemployment Rate by Age Groups. 2002 ...... 67 Figure 3.14 Average Male Unemployment Rate by Age Groups. 1996 vs 2002 ...... 67 Figure 3.15 Average Female Unemployment Rate by Age Groups. 1996 vs 2002 ...... 67 Figure 3.16 Unemployment Rate vs Years of Education. 1996 ...... 68 Figure 3.17 Unemployment Rate vs Education Level. 2002 ...... 68 Figure 3.18 Average Monthly Wage. 1996 ...... 69 Figure 3.19 Average Monthly Wage by Sector of Employment. 1996 ...... 69 Figure 3.20 Gender Wage Gap Ratio by Educators Level Completed. 1996 ...... 70 Figure 3.21 Labor Market Rigidity and Informality ...... 90 Figure 4.la Djibouti and Trading Partners: CPI Indices. 1990-2004...... 102 Figure 4.lb Djibouti and Trading Partners: CPI Indices. 1990-2004...... 102 Figure 4.2 Exchange Rate Development. 1980-2004...... 103 Figure 4.3 Total Bank Deposits and Credits. 1990-2004...... 108 Figure 4.4a Credit Distribution. December 2004. by Economic Activity ...... 109 Figure 4.4b Credit Distribution. December 2004. by Maturity ...... 109 Figure 4.5 Nominal GDP and Broad Money (log scale). 1990-2004...... 118 Figure 4.6 Inflation and Broad money. 1990-2004 ...... 119 Figure 4.7 Inflation and M2 Velocity. 1990-2004 ...... 119 Figure 4.8 Demand and Time Deposits ...... 119 Figure 4.9 Currency Composition of Deposits. 1994-2004...... 119 Figure 4.10 Foreign Currency Deposits Ratios. 1995-2004...... 120 Figure 4.1 1 Foreign Currency Deposits. 1995-2004...... 120 Figure 4.12 Non-performing loans. 1994-2004...... 121 Figure 4.13 Deposit and Credit. 1995-2004...... 121

... 111 Figure 4.14 Djibouti: Domestic and Foreign Interest Rates. 1990-2004 ...... Figure 4.15 Djibouti: Interest Rates Spread. 1990-2004 ...... Figure 4.16: Broad Money and Net Foreign Assets (NFA). 1994-2004 ...... 122 Figure 5.1 Container Throughput. 1994-2003 ...... 128 Figure 5.2 Major Cargo Flows. 1994-2004 ...... 129 Figure 5.3 Import and Export Cargo Flows. 1994-2004 ...... 129 Figure 5.4 Djibouti’s Port Productivity in Comparative Perspective (2003) ...... 131 Figure 5.5 East Afiican Port Tariffs (USD per full TEU). 2004 ...... 132 Figure 5.6 Average Port Transit time Compared with Other International Ports. 2004 ...... 133 Figure 5.7 Port Traffic (metric tons). 2002/04 ...... 136 Figure 5.8 General Assessment of the Current Condition of the National Priority Road ...... 144 Figure 5.9 MET’SResources for the Road Sector ...... 145 Figure 5.10 Financing Needs and Resources Collected on the Road Corridor ...... 150 Figure 5.11 Average Electricity Tariffs. 2001 ...... 152 Figure 5.12 Average Water Tariffs. 2001 ...... 155 Figure 5.13 Number ofTelephone Lines per 100 inhabitants: Djibouti’s vs . Comparators ...... 156

LISTOF TABLES Table 1.1 Djibouti vs Comparators: Real Per Capita GDP Annual Growth. 198 1-2004 ...... 2 Table 1.2 Djibouti vs Comparators: Real GDP Annual Growth. 1981-2004 ...... 2 Table 1.3 Cbmposition of Real Gross Domestic Productivity by Sector of Origin 1985-2004 ...... 8 Table 1.4 Sectoral Real Growth Patterns. 1985-2004 ...... 9 Table 1.5 Expenditure Share to GDP. 1980-2004 ...... 10 Table 1.6 Determinants of Growth. 1990-2002 ...... 11 Table 1.7 Djibouti vs Comparators: Period Per-Capita GDP Growth Rate ...... 13 Table 1.8 Panel Growth Regression (Fixed Effects) ...... 14 Table 2.1 Summary ofMedium-Term Macro Economic Scenarios, 2005-2015 ...... 32 Table 2.2 Characteristics ofCivil Servants Monthly Wages, November 2000 ...... 35 Table 2.3 Characteristics ofCivil Contractuals Monthly Wages, November 2000 ...... 35

Table 2.4 Summarv ofResults fiom Public WageI Reform Scenarios ...... 43 Table 2.5 to 2.16 Impact on Public Wage Bill (2005.15), ...... 45 Table 2.17 Potential gains and trade offs ofcivil service reform scenarios ...... 57 Table 2.18 Reforming Government pay and employment policies: Winners and Losers ...... 57 Table 3.1 Costs of Joining the Formal Private Sector, 2001 ...... 60 Table 3.2 Distribution ofthe Employed, 1996 ...... 68 Table 3.3 Employment by Gender, 1996 ...... 69 Table 3.4 Level of Education Completed by Wage Earners, 1996 ...... 69 Table 3.5 Monthly Wages by Education Level Completed, 1996...... 70 Table 3.6 Demand for Employment and Registered Offers at the National Employment Service ...... 71 Table 3.7 Employment Evolution ofWorkers Registered With Social Security ...... 72 Table 3.8 Public Private Wage Differential, 2005 ...... 74 Table 3.9 Annual Wage. Income, 1996 ...... 74 Table 3.10 Monthly and Annual Wages. 1996...... 74 Table 3.11 Descriptive Statistics ofDjibouti Wage Earners. 1996 ...... 75 Table 3.12 Public and Private Sector Participation Equation...... 76 Table 3.13 Log Wage Equations for the Public and Private Sector ...... 77 Table 3.14 Mean Log Wage Differential between Public and Private (Formal) Sector ...... 78 Table 3.15 Private Returns to Education in Public and Private Sector ...... 79 Table 3.16 Selectivity Corrected Wage Equations ...... 80 Table 3.17 Selectivity Corrected Private Rate ofReturn to Schooling per Year. 1996 ...... 80 Table 3.18 Djibouti vs Comparators: Costs ofDoing Business. 2004 ...... 81 Table 3.19 Average Tariff Rates. Djibouti vs Comparators ...... 86 Table 4.1 Selected Financial Indicators. 1990-2004 ...... 107 Table 4.2 Sectoral Distribution ofCredits. 1997-2004 ...... 108 Table 4.3 Deposits Structure. 1990-2004 ...... 110 Table 5.1 Total Throughput in Djibouti, Mombasa, and Dar-es-Salaam, 2003 ...... 127

iv Table 5.2 Container Throughput and Transshipment in East African Ports. 2003 ...... 127 Table 5.3 Evolution of Port Traffic. 2003/2004 ...... 130 Table 5.4 Comparative Cost of Stay in Port. 2001...... 132 Table 5.5 Port Revenues Due to Ethiopian Traffic in Transit in 2003 ...... 137 Table 5.6 Estimate of Effect ofPort Activity on Employment ...... 137 Table 5.7 The Road Network in Djibouti...... 143 Table 5.8 Road Traffic and Maintenance. 2001-2003 ...... 145 Table 5.9 Capital Works Program, 200612020 ...... 149 Table 5.10 Road Maintenance Program, 200~2020...... 149 Table 5.1 1 Production and Consumption of Electricity. 1999/2004 ...... 151 Table 5.12 Production and Consumption ofWater. 1999/2004 ...... 154 Table 5.13 Information technology: Djibouti vs . Comparators. 2003 ...... 157

LISTOF BOXES Box 1.1 Recent Economic Developments (2002-04) ...... 7 Box 3.1 The informal Sector in Djibouti: Main Features ...... 61 Box 3.2 Labor Market Dynamics ...... 63 Box 3.3 A Complex Legal Framework for Investment ...... 84 Box 3.4 Labor Market Regulatory Reforms ...... 92 Box 4.1 Conditions and Policy Implications for Currency Board Arrangement ...... 104 Box 4.2 Three Types ofInforma1 Finance Arrangement ...... 113 Box 4.3 Current Status of Microfinance in Djibouti ...... 115 Box 5.1 Djibouti’s Regional Competitors ...... 134 Box 5.2 Djibouti’s Port and Its Ties with Ethiopia, 1998-2004 ...... 137 Box 5.3 The International Road Corridor ...... 146

PART I.GROWTH ANALYSIS

Part I (Chapter 1) reviews growth patterns and performance in Djibouti and identiJes the main sources of and constraints on growth. After assessing past economic pe$ormance, the chapter looks at patterns of supply, demand, and factor decomposition of GDP growth over the period 1980-2004. It then assesses the long-term sources of growth, drawing on two empirical exercises. The first is based on a neoclassical growth model (Solow’s model of growth accounting). The second is based on an expanded version of Barro’s endogenous growth model (including the respective roles of private and public investment in the growth process), using modern panel data econometrics. The last section of the chapter complements the empirical analysis with a more qualitative assessment of the main bottlenecks to growth and private investment, drawing on a recent analytical framework developed by Ricardo Hausmann et a1 (2004).

CHAPTER 1: GROWTH PATTERNS, SOURCES, AND CONSTRAINTS 1.1 Growth Patterns

1.1.1. Overall Growth Performance: Stylized Trends

1.1. Djibouti’s economic growth over the past two decades has been lower than that of regional comparators and of other countries at similar income levels. Djibouti experienced average growth of less than half a percent between 1980 and 2004. After growing at about a percentage point per year during the first years of independence (1977-1986), GDP declined sharply during most of the 1990s. The civil conflict that erupted in 1991 triggered a period of political turmoil that led to negative real GDP growth until 1998. Real output growth picked up thereafter, reaching more than 3 percent by 2004, but Djibouti’s growth performance remains below Sub-Saharan African levels. Djibouti’s brief periods ofeconomic recovery depended heavily on exogenous factors. They include an increase in external support (foreign aid and military support), intensified transit to Ethiopia in the late 1990s that led to a strong resumption of port activities, and stronger demand for services by foreign expatriates since 2001. At the same time, significant efforts were also made by the authorities to improve public finances since 1998. These fiscal reforms included: a freeze on nominal salaries in 1993, and a demobilization program from 1996 to 2004 and a 20 percent civil service salary cut in 1998.

Figure 1.1 Djibouti s. Comparators: Real GDP Growth Rates (percent) m1980s 1990s 7.00%, 0 2000-04

-2.00% J Djibouti Yemen Ethiopia Morocco Tunisia Mauritius Middle- Sub- Developing Emerging East & Saharan countries economies North Africa Source: WO, 2004 Africa

iddle-East & North Africa b-Saharan Africa

ase Year. Source: WEO, 2004

2 1.2. With a fast-growing population, Djibouti has experienced a continued sharp decline in per capita income, by about 3 percent, over the past two decades. Income per capita in Djibouti was well above that in comparator countries during most of the 1980s. This situation reversed starting in 1991, owing to the continued decline in Djibouti's per capita GDP and a steady increase in the other groups (see Figure 1.2 below). Although per capita GDP converged rapidly toward that of Sub-Saharan Afncan countries (Figure 1.3), it fell well below per capita GDP the average of Middle-East and North Ahca, developing and lower-middle-income countries.

Rgure 1.2 Djibouti vs. Conparators: Real GDPper Capita (1980-2004) Fgure 1.3 Djiboutivs. Conparatom: Real GCPper Capha Patios (in constant 2000 US$) (1980-2003)

3.5 7

29002400 1

Source: WDI, 2004 Source: WDI, 2004

1.1.2. Volatility of Growth

1.3. Djibouti's economic growth has been highly volatile-surpassing the average volatility of output in Sub-Saharan Africa-although lower than that of its neighbors (Yemen and Ethiopia). In Djibouti, increases in income have been difficult to sustain, and declines in income have not been quickly reversed. Djibouti's volatility of real GDP growth (defined as the standard deviation of annual growth rates observed during the decade) was higher than the MENA average, although not higher than its neighbors (Figure 1.4). Djibouti's real GDP growth is negatively correlated with volatility (Figure 1.5). This is consistent with the findings of recent cross-country studies on the relationship between volatility and growth (Pinto et al, 2004).

Figure 1.4: Growth Volatility (1981-2004) Figure 1.5 Djibouti: Real GDP Growth Rates.Volatility(%) (period standard denations, in percent) (1986-2004) 9,001 - 5% 8 88Ds 8.00% 8- -.5% 0 200004 .. 4% 7.00% I 0 88'2024 6.0096 -- 4% 5.0096 .. 3%

4.00% --3% 3.0056 -- 2% 2.00% -- 2% 1.00% .. 1% 0.00% .-1% Djlboutl Yemen Ethiopia Morocco Tunisia Mauritius Middle- Sub Devvelopin~ E&& Sshmm Co~ntne?~ North Afnsa Afllca

Source: WEO, 2004 Source: WEO, 2004

3 1.4. The volatility of output in Djibouti is explained by external and internal factors. The period of highest economic volatility occurred in the 1990s, owing to regional instability and domestic conflict. Moreover, Djibouti’s heavy dependence on fluctuating demand by foreign military expatriates and for port transshipment receipts has rendered the country vulnerable to persistent current account deficits. Since 2000 external demand experienced positive growth, mainly owing to the strong resumption in port activities.

1.5. Business cycle fluctuations in Djibouti’s main trading partners do not affect Djibouti’s economic growth, with the exception of Yemen. Results from pair-wise Granger causality tests’ show that with the exception of Yemen, the business cycle fluctuations of Djibouti’s main trading partners (Ethiopia, Saudi Arabia, France) do not granger cause Djibouti’s growth. Yemen’s past business cycle movements help predict Djibouti’s current business cycles. Similarly Djibouti’s growth granger causes Yemen’s growth.

1.6. Yemen’s business cycle fluctuations negatively affect Djibouti’s growth rate, but not vice versa’ Figures 1.7 and 1.8 show the response of Djibouti’s growth rate to a (one standard deviation) positive shock in Yemen’s growth rate and vice-versa. A positive shock to Yemen’s growth negatively affects Djibouti’s growth in the second year after the shock, which it recovers from after five years. The impact ofa shock to Djibouti’s growth on Yemen’s output growth is less statistically robust, as reflected in the broad confidence interval bands. This difference could be explained by structural differences in the economies of Djibouti and Yemen. Yemen 1 Figure 1.6: Djibouti 6,Yemen: Real GDP Growth (percent) depends mainly on its oil sector, which is 25.0% unrelated to movements in Djibouti’s -tjibouti 20.0% \ economic growth. By contrast, Djibouti’s -Yemen, bp. ~ 15.O% ‘‘‘s economy relies heavily on port activities and 1 competes with Yemen’s port of Aden. This is illustrated by the impact of the decision by Pacific Shipping Lines to shift its transshipment business from Djibouti to Aden -10.0% 1 in 2004. This led to a slowdown in port 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 activities and translated into lower than expected economic growth. Source: WEO, 2004

Figure 1.7: Response of Djibouti’s real growth rate to a one std. dev shock in Yemen’s growth rate. ___---,--.-.._____.___ .-.- .-. ,,,..‘---____,,.I+-e-.- .01 - .. ...-.*’ ., ,

-.w ! 123 4 5 6 7 8 9 10

Source: Authors’ calculation using WEO, 2004 Source: Authors’ calculation using WEO, 2004

’ Ify is said to granger cause x, that does not imply that x is the effect or result ofy in a conventional sense. Rather it implies that lagged values ofx help predict y. Granger causality measures precedence. ’ We draw on impulse response functions based on an underlying vector auto regression ofthe real growth rates ofthe countries in question.

4 1.1.3. Patterns of Growth and Recent Developments

1.7. Three main patterns emerge from a historical analysis of Djibouti's real GDP growth: one showing a declining but positive GDP growth during 1980-86, another characterized by volatile GDP growth during 1987-96; and a third showing a relatively sustainable recovery since 1997. Djibouti's actual growth rates remained below potential for most of the 1990s, on average (Figure 1.9). Potential output growth3 reversed its downward trend in the second half ofthe 1990s, and since 2001 the economy managed to grow slightly over potential.

Figure 1.9 Djibouti: Actual vs. Potential Real GDP Growth, 1981-2004 (percent)

6 Economic decline High volatility of growth Economic recovery

-2

-4

-6 Actual real GDP growth growth (HP filter)

-a

Source: Authors' calculations using WEO, 2004.

1.8. Declining butpositive growth (1980-86). Between 1980 and 1986, real GDP grew at about 0.7 percent per year. The respectable growth record (about 3 percent a year) achieved in the few first years after independence (between 1978 and 1982) was flattened by weaker performance between 1983 and 1986. As a result of the end of the construction boom and the tapering off of public investment projects, fixed capital formation was brought down from more than 25 percent of GDP in the early 1980s to about 19 percent in 1986. Domestic savings were negative and the country continued to rely heavily on external resources.

1.9. High volatility of growth (1987-96). Between 1987 and 1996, real GDP declined by about 1.6 percent a year. Regional instability made Djibouti one of the few ports in the region where trade could be safely diverted, so transit trade rose in 1988. This economic recovery was short lived. In 1989 growth declined further because of serious floods and the discontinuance of food aid shipments for Ethiopia through Djibouti. The resource gap widened because of continuing high levels of domestic consumption, which averaged 113 percent of GDP between 1990 and 1991. Djibouti went through a period of domestic armed conflict in 1991-94. This was accompanied by declining foreign official assistance, weaker demand for services by expatriates, and a contraction

Potential output growth is calculated by drawing on a time series of past output data and using the Hodrick-Prescott filter of real GDP-level data for 198C-2002. Lack of reliable time series data on labor and capital stocks impairs a more sophisticated analysis of the production function and the determinants of growth. The potential growth rate, as measured here, reflects the constraints posed by existing economic, fiscal, and structural policies. A higher growth rate can likely be achieved by reforming such policies and improving the quality of institutions.

5 in port activities owing to stiffer competition from neighboring ports. The adverse effects of these combined factors led to a strong contraction of GDP (by about 20 percent on a cumulative basis) and a substantial deterioration in macroeconomic balances. Investment fell sharply, from about 19 percent in 1991 to nearly 9 percent in 1996, in part as a result of an accumulation of domestic payment arrears toward the private sector and public enterprises that were used to finance the overall fiscal deficit.

1.lo. Post conflict economic recovery (1997-2004). During 1996-98, Djibouti’s external aid flows increased. Inflation declined sharply, averaging 1.5 percent annually (against an average 5 percent inflation rate during 1993-95). The economy’s real growth performance was relatively flat in 1996 and 1997, but a recovery took hold in 1998 owing to Ethiopia’s decision at midyear to channel its transit trade through the port ofDjib~uti.~ A sharp upturn in port operations’ more than offset the negative impact of the reduction in France’s military presence and led to a positive real GDP growth. Since the late 1990s, stronger growth performance reflected continued political stability, strong port performance, increased donor support; and fiscal reforms carried out by the Djibouti government that led to substantial reduction in the budget deficit. As a result, in 2001 real GDP reached 1.9 percent then grew by 2.6 percent in 2002 and 3.5 percent in 2003. The recent upward growth trend also reflects a surge in foreign direct investment (FDI) flows owing to investments in the new port, military aid from the US and France (hovering around 5 to 6 percent of GDP over the past two years), and continued donor support. See box 3.1 for more details on recent economic developments.

1.1 1. Djibouti’s current level of economic growth, hovering around 3 percent, will not be able to generate enough jobs and reduce poverty. Under an estimated population growth rate of 2.6 percent per year, economic growth will not be enough to reduce unemployment and poverty. Growth performance has not been able to keep pace with a growing labor force, leading to an unemployment rate estimated at about 60 percent of the labor force. In addition, the incidence of extreme poverty (with a poverty line defined at US$1.8 per day) has deteriorated from 34 percent to an estimated 42 percent.

The increase in transit trade to Ethiopia offset the negative effects of the decline in electricity production following the fire that destroyed about 35 percent of production capacity in that year. ’Total port traffic increasing by 25 percent in 1999, largely reflecting increased transit trade with Ethiopia.

6 Box 1.1 Recent Economic Developments (2002-04)

Real GDP growth, which had reached 3.5 percent in 20D3, fell slightly to 3.4 percent in 2004, largely owing to a slowdown in trade activity. The recent slowdown was caused by a combination of factors including bottlenecks at the port, a decrease in trade with Ethiopia, and the decision by a Singaporean company to move its transshipment business to Aden in late 2003. The transshipment business had previously represented almost 25 percent of the volume of material handled, with that figure now being estimated at only 4 percent. Economic growth was otherwise supported by the demand for goods and services by military troops in Djibouti, and by continued growth in public investment.

As investments have continued to rise in the past three years, imports have continued to pick up, widening the trade balance. The trade deficit has been partly offset by a parallel inflow on the services account, which has helped contain the current account deficit (averaging 9.5 percent of GDP over the past three years). The current account deficit was financed largely through FDI flows and external borrowing. The surge of foreign direct investment is due to the development of a new port complex in Doraleh.

The overallfiscal (cash) deficit improved from Spercent of GDP in 2003 to 3.1 percent of GDP in 2004; however, the basic fiscal balance6 has deterioratedfrom 4.2 to -5.4percent over the same period. The worsening ofthe overall fiscal balance reflects increased expenditures, particularly capital expenditures and transfers. Total revenues have increased slightly as a percentage of GDP, with an increase in military revenue and grants offsetting a decrease in tax revenue collection (most recently due to a decline in receipts from the excise duty levied on petroleum products).

Inflation rose in 2004 compared with the previous two years, owing largely to high oil prices. However, it was contained to 3 percent through prudent monetaiypolicy. The main causes of the rise in inflation were the higher prices ofoil and food. External debt has grown over the reference period, from 56 percent of GDP to 64 percent of GDP, leading to a drawdown ofreserves in 2004 (to US9.5 million).

qibwti, and Price Djibouti, Fiscal Balances Output Dsvelopments ~. BO, I

-8.0 { / -8.0 1 I -10 o i. ___.__ -- - __ - - __ --.-. - -J 1Sffi 1888 1887 1888 1999 2OW 2001 Zw2 2W3 2004 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 -Real GDPGrowth -Real GDPpercspita -CPI -m~~est~c mcaI Balance --Overail Balence (paymnt baslo)

Djlbwtl, Money and Credit Djiboutl. Balance of Payments 20, 25.0

20.0

E 15.0 lo p10.0 b 'i 1: ::: s -5 -5.0 -10.0 -10 .- e95 s9e ea7 ea8 e99 moo ZWI Mo2 2003 2004 895 888 637 888 630 ZWO 2001 2W2 2W3 2004 -Broad MOney mNet Foreign Assek +Net Domestic assets -Overall balance ICurrentaccount -Capital account

Source: Djibouii authorities and KDI, 2004.

The basic fiscal balance, an indicator of the government's fiscal effort, is defined as the overall balance on a commitment basis excluding grants, military-related transactions, and foreign-financed expenditures.

7 f .2. 1, 1.13. With poor natural resource endowments, the primary sector's contribution to GDP barely exceeds 3 percent. Agricultural production has remained very small, averaging 3 percent of GDP over the reference period. Its limited role reflects mainly the lack of arable land (less than 1 percent of the total land area), the scarcity of pasture land (about 10 percent of the total land area), insufficient rainfall, and the small size ofthe domestic market. Domestic production ofhits, vegetables, and fish covers only part of domestic consumption. Most agricultural output is related to livestock, namely, producing milk and slaughtering cattle for meat and hides. Livestock israised mainly by nomads in the interior of the country. The expansion of this sector is hampered by the lack ofrainfall.

1.14. Other economic activities in the primary sector, including Bshing and salt extraction, contributed a minor amount to GDP. The considerable potential offered by the waters near Djibouti has yet to contribute to the development of the fishing industry. This industry has been hampered by inadequate fishing infrastructure, poor management ofthe Maritime Cooperative, and the insurgency during most ofthe 1990s. Recently, salt extraction has expanded; still, this activity accounts for less than one percent ofGDP.

1980s 1990s 1985 - 2002 ZOO0 - 2002 2002 - 2004 Primary sector 1.3 0.3 2.8 3.9 3.6 Of which Agriculture, livestock, and fishing 1.1 -0.5 2.0 2.0 Other 3.1 9.9 10.1 14.2

Secondary sector -0.8 -5.8 -0.9 2.9 5.0 Of which Construction and public works -4.4 -6.8 0.7 2.7 5.3 Water and electricity -0.4 -4.8 -1.8 3.6 4.8

Tertiary sector 0.4 -1.0 -0.3 1.2 1.8 Of which Commerce and tourism 1.4 -3.1 -0.1 2.4 4.2 Transport and communications -0.1 3.2 3.9 1.9 3.2 Banking and insurance a/ 2.3 -1.4 -1.2 2.1 3.1 Public administration -1.1 -3.0 -2.8 -1.1 -2.7 Indirect taxes less subsidies b/ -4.2 -2.0 -2.2 1.0 7.6

GDP at market prices -0.4 -1.8 -0.5 1.5 3.0

Notes: a/ Net of imputed chargefor banking services. b/E :luding stamp and licencej S. Source: Djibouti authorities; and Fund stafl'estimates

1.15. Manufacturing has remained minimal and contributed only about 3 percent to GDP throughout most of the 1980s and 1990s. Its value added declined at an average rate of4 percent in real terms, mainly because of sluggish demand conditions in the economy. The manufacturing sector exports virtually nothing. This sector consists mainly of a few bakeries, a flour mill, a few slaughtering units, a few units supplying building materials, and some repair units. Other industries include the production of bitumen, oxygen, and acetylene; confectionery; printing and handicrafts. Despite the government's promotional efforts, manufacturing has been hampered by the small size of the market, the high cost of labor and utilities, and the lack of entrepreneurial capacity and shlled workers.

9 1.3. DEMANDCOMPOSITION OF REAL GDP GROWTH

1.16. A different way of loolung at the main drivers ofDjibouti's real GDP growth performance is to concentrate on the demand side of the economy. The analysis below corroborates earlier findings: it reveals a pattern ofgrowth driven mainly by consumption demand.

1.17. Domestic demand has contributed to most of Djibouti's real GDP growth over the period 1980-2004. Domestic demand, particularly consumption, has been the main driver of Djibouti's real GDP. The end ofthe civil conflict in the late 1990s brought a recovery in private consumption, which reached 80 percent of total consumption on average over the past four years. External demand has not played a critical role in Djibouti, owing to structural weaknesses in the economy. Djibouti suffers from a structural trade deficit because of the narrow export base for locally produced goods and the need to import almost all consumer and commercial goods.

:I II.

Figure 1.11 Djibouti: Average Annual Exqenditure Growth (X) I I 1980s 1990s 2000-2004 4.0 1 1990-2004 Total Consumption 98.0 107.2 104.8 3.0 = Nongovernment 61.6 75.1 80.1 2.0 Government 36.4 32.1 24.7 1.0 0.0 Gross Investment 18.7 13.9 13.3 Nongovernment 5.3 8.0 8.7 -1.0 Government 13.4 6.0 4.6 -2.0 Internal demand .External demand -3.0 OReal GDPgmwth Exports of goods & services 62.0 47.3 45.4 -4.0 1 Imports of goods & services -78.8 -68.5 -63.5 1990-94 1995-99 2000-04 Source: WEO, 2004. Source: WEO, 2004

1.18. Low levels of investment, high levels of domestic consumption, and a pattern of negative domestic savings have led to a persistent resource gap. As explained above, investment rates averaged 19 percent ofGDP in the 1980s owing to a construction boom. Through most ofthe 1990s investment followed a downward trend, mostly because of a sharp decline in public capital expenditures. The investment rate declined Figure 1.12 Djibouti: Savings Vs. Investment as Share of GDP (%) from about 19 percent in 1990 to 14 25.00% - percent in the 1990s and averaged 10 20.00% - percent during the period 2000-02. Since 15.00% - 2003, the investment rate has picked up, 10.00% - reaching 20 percent of GDP in 2004. The 5.00% - recent investment boom is driven mostly by increased FDI flows directed toward new inhstructure investments in the new port at Doraleh. Domestic savings remained negative over the past two decades, leading to a persistent resource gap- Source: WDI, 2004

10 1.4. Sources of and constraints on economic growth

1.19. The previous section argued that Djibouti’s growth was below potential during the 1990s and above potential in the aftermath of the civil conflict, but presented no estimate for the sources of the long-run growth rate. This section explores this issue and provides estimates for the determinants oflong-run growth.

1.4.1. The roles of capital accumulation, labor and total factor productivity

1.20. Djibouti experienced a sustained fall in per capita income over the past two decades. This indicates that the capital-labor ratio has been falling together with productivity. This section explores this issue, presenting a factor decomposition of Djibouti’s GDP assuming a standard neoclassical Cobb-Douglas production function. Although poor data and a relatively short time series prevent strong conclusions, this section reveals that GDP growth was dnven primarily by total factor productivity growth.

1.21. The objective of the growth accounting exercise is to measure the contribution of employment, capital accumulation, and total factor productivity (TFP) to Djibouti’s growth over the period 1990-2002. For this purpose, a simple growth-accounting exercise was conducted, assuming an aggregate production function of a Cobb-Douglas type and constant shares of labor and capital in output. In addition to this simplified representation ofthe way the economy operates, a number of assumptions were made to cope with data gaps and deficiencies. The approach and assumptions are explained in detail in annex 1.1

1.22. Djibouti’s growth during the reference period was caused mainly by TFP. A negative contribution of TFP during 1990-1997 (coinciding with the civil conflict that devastated the country) offset the increase in the contribution of labor and reinforced the negative contribution of capital, leading to a negative real GDP growth ofmore than one percent. Since 1998, the increase in the contribution of TFP has largely made up for the negative contribution of capital, leading to a positive real GDP growth ofabout 1.5 percent.

1990-2002 1990s 2000-02 998-2002 Growth rates (in percent) GDP -1.02 =I= -1.83 1.48 1.47 Labor -0.31 0.57 -0.69 -0.08 Capital -1.45 -1.42 -0.91 -1.16 TFP -0.27 -1.20 2.30 2.13 Contribution to GDP growth (in percentage points) Labor 0.13 0.23 -0.27 0.03 Capital -0.87 -0.85 -0.54 -0.69 TFP -0.27 -1.20 2.30 2.13 Productivity growth rates (in percent) Labor productivity I! -1.33 -2.39 2.18 1.38 Capital productivity 21 0.44 -0.41 2.41 2.65 Capital per worker -1.76 -1.98 -0.22 -1.24 Source: WEO, Djibouti Statistical Yearbook; Authors’ calnrlatiom I/Defined as output per employee: 2/ Defined as output per unit of utilied capital

11 1.23. Capital and labor productivity experienced negative growth during most of the 1990s, a period of high political instability. Although this trend reversed in 1997-98, labor and capital still contribute very little to growth. The investment rate declined from about 19 percent in the 1980s to about 14percent in the 1990s and averaged 10 percent during 200042. An increase in unemployment (estimated by the 1996 and 2002 household surveys to have risen from about 44 percent in 1996 to about 60 percent in 2002) and low levels of human capital reflect a lower contribution of labor to GDP growth during that period. The decline in labor productivity, especially during the 1990s, did not translate into a commensurate decline in real wages, leaving actual wages above the labor economic price. The difference between these two indicators can be seen as an economic rent, which was protected by existing labor regulations and government pay policies. This issue is discussed in detail in chapter 3.

Figure 1.13 Djibouti: Labor and Capital Productivity Growth, 1990-2002 (percent)

1.24. Totalfactorproductivity growth remains low. Although TFP growth was negative over the first half ofthe 1990s, it has improved since 1998-partly because of greater political stability and partly owing to the technological spillovers associated with more intense trading activity, as witnessed by port activity, and the high profitability of that activity. However, productivity growth still remains low. A number of reasons may explain this: low human capital expenditures, weak private investment, rigidities in labor and product markets, and weaknesses in the investment climate.

Figure 1.I4 Djibouti: Evolution of Total Factor Productivity, 1990-2002 6

(Rrrnl!

1990 1991 1592 1991 1591 1595 1996 1597 I598 1999 2Wn iWI 2W2

12 1.4.2. Long-run sources of growth: incorporating a cross-country perspective

1.25. International evidence on factors of long-term growth can provide useful insights on policy options for meeting Djibouti’s growth challenges in the coming years. Djibouti’s growth performance has not been promising, compared with that of other countries in Sub-Saharan Afixa. Table 1.7 shows that the country has experienced negative growth rates on average for the past two decades except until recently. Hence an analysis to uncover factors that hinder its progress or that could promote growth is instructive.

Middle-East Sub-Saharan Period Djibouti Tunisia Morocco North Africa Africa 1975-79 -6.28 3.70 3.88 2.66 -1.03 1980-84 -6.90 2.00 0.59 -1.69 -1.12 1985-89 -7.22 0.07 2.76 -1.77 -0.30 1990-94 -6.16 2.90 1.29 1.75 -2.07 1995-99 -3.64 3.67 0.47 1.01 0.83 2000-03 I 0.131 3.021 2.271 0.131 1.111 Source: Penn Tables, WDI 2004.

1.26. This section provides estimates of the sources of long-run growth, incorporating the role of policies and institutions. A cross-sectional perspective is incorporated into the time series analysis, to derive implications from international evidence on the patterns and determinants of growth. This section follows the most recent strand of the empirical growth literature, which links a country’s economic growth rate’ to economic, political, and social variables. A large number of variables can be used as determinants of economic growth. This report focuses on variables that have received the most attention in the economic literature: macroeconomic stability, structural policies, quality of institutions, and human capital. Following Khan and Kumar (1 997) and Serven and Solimano (1990), the model also incorporates the roles of public and private investment in determining long-run growth.

1.27. The analysis of long-term determinants draws on a panel regression of 39 developing countries, spanning jive-year periods from 1970 to 1999. The list of countries is presented in annex 1.2. The regression explains average growth of per capita GDP over each five-year period. The explanatory factors of growth included in the regression are the following:

Level ofper capita GDP at the beginning of each period, measuring growth effects owing to convergence. Macroeconomic stability, (MS) measured by an indicator comprising a) inflation b) the ratio of external debt to GDP c) the ratio of the external current account balance to GDP d) the ratio of the budget deficit to GDP Openness in trade andjinance, (Open) measured by an indicator that includes a) the ratio of private sector credit to GDP b) the ratio of trade flows (sum of export and exports) to GDP Institutional quality, measured by an indicator (le> composed of

’ The estimated model can be used to compute expected growth rates that are conditional on the observed policy determinants, and a disturbance element that is orthogonal to the different determinants and purely transitory. In this regard, one could use expected growth as a measure of the long-term growth underlying actual or observed growth.

13 a) an index of government effectiveness b) an index of regulatory quality c) an index measuring the rule of law d) an index measuring the degree of corruption Human capital accumulation, measured by an indicator (HUM) composed of (a) the infant mortality rate, as a proxy of the health conditions of the population (b) the number of years ofprimary schooling of the population . Ratio of private investment to GDP (priinv). Ratio ofpublic investment to GDP (pubinv).

1.28. Principal component analysis was used to obtain the composite indicators MS, SR, IQ,and HUM; detailed results of their construction appear in annex 1.2. The regression was estimated using county-specific fixed effects (table 1.8).*

regressors Estimates Std. Errors

Lagged GDP (laggdp) -0.0947 *** (0.01) Macroeconomic stablity (MS) 0.0162 *** (0.00) Institutional quality (IQ) 0.0106 ** (0.00) Financial & trade openness (Open) 0.0074 (0.00) Private investment to GDP (Priinv) 0.0031 *** (0.00) Public investment to GDP (Pubinv) 0.0043 *** (0.00) Human capital (HUM) 0.0140 *** (0.00) constant 0.6531 *** (0.07) IR2 0.54 Source: Authors' calculations using WEO, WDl, Penn Tables, 2004. Notes: 1. Dependent variable is period average real GDP per capita. 2. Panel consist of 39 countries. 3. "'denotes significance at 7%, ** at 5% and at 70%.

1.29. The estimated growth regression was used to compute the contribution of the different factors to the average growth of Djibouti and ofcomparatorsg (Tunisia, Mauritius, and Sub-Saharan Africa) over each five-year period from 1970 to 1999. As a benchmark, all other countries in the sample were used excluding Djibouti and its comparators. Using the coefficients of the regressions, we decomposed for each five-year period the growth differential to the average of the remaining countries into seven components: macroeconomic stability, institutional quality, openness, private investment, public investment, human capital, and other factors captured by the country fixed effects and growth residual. The contributions to growth for Djibouti versus Tunisia, Mauritius, and Sub-Saharan Ahca are shown in Figures 1.15, 1.16, and 1.17.

The fixed effects specification fitted the data reasonably well, accounting for 54 percent of the variation of growth across countries and over time. All independent variables were highly significant at the 1 percent level, with the exception of Institutional Quality (IQ), which was significant at 5 percent, and Openness (Open), with a 10 percent significance. The explanatory variables also possessed the expected signs: better macroeconomic stability, stronger institutional quality, greater openness, and higher levels of private and public investment are all associated with stronger growth. 9 Lack of comparable time series data on Ethiopia and Rwanda did not allow the inclusion of these two countries in the sample.

14 1.30. Djibouti’s growth performance, afer controlling for income levels, is below average across four main fronts: macroeconomic stability, institutional quality, human capital, and private investment. Of the four, the low levels of private investment had a much larger negative average contribution to economic growth. To its credit, Djibouti has performed better than the average and its comparators in the degree of finance and trade openness. Relative to Sub-Saharan Africa, Djibouti perfoms similarly on macroeconomic stability and institution quality but is lagging much farther behind in its investment efforts, notably in terms of private investment and human capital development (see Figure 1.15). The country fares slightly better in public investment, though its contribution appears to be volatile. A wide gap is evident when comparing Djibouti with better-performing countries such as Mauritius (Figure 1.16) and Tunisia (Figure 1.17).

Figure 1.15 Contribution to Average Growth: Djibouti vs. Sub-Saharan Africa (in %per year) I 2 1 0 -1 -2 -3 -4 -5 I :; 1 Mxrostabillty 1 lnstiiutlonalquality 1 Open

E 1970-74 W 1975-79 0 1980-84 0 1985-89 E 1990-94 E 1995-99

Source: Authors’ calculations based on regression results of Table 1.8

Figure 1.16 Contribution to Average Growth: Djibouti vs. Mauritius (in %per year) 4 3 2 1 0 -1 -2 -3 -4

E 1970-74 E 197579 0 1980-84 1985-89 W 1990-94 E 1995-99 Source: Authors’ calculations based on regression results of Table 1.8

15 Figure 1.17 Contribution to Average Growth: Djibouti vs. Tunisia (in YO per year)

01970-74 I1975-79 0 1980-84 0 198589 w 1990-94 I199599

Source: Authors’ calculations based on regression results of Table 1.8

Growth that could be.. .. 1.31. How much higher would growth have been if Djibouti had matched the performance of comparator countries (relatively small and poorly endowed economies) along the main policy dimensions identified above in the past three decades? Based on the results of the regression presented in the earlier section, Figure 1.18 provides an answer. If Djibouti had matched the investment climate and policy environment of Tunisia, its real income per capita growth would have reached 7.7 percent. If it had equaled that of Mauritius, it would have increased to about 9.5 percent.

1.32. The analysis corroborates earlierfindings: higher private investment and human capital appear to have been the areas with the highest growth potential payof$ Additional growth enhancing payoffs are found in stronger macroeconomic stability and better institutional quality.

Fgure 1.18 Djibouti: Growth That Could Have Been. 1970 - 99 14% 1 Openness i2% w Macrostability 10% 0 hstitutiinal Quality 0 Public Investrent 8% w Humn Capital 6% Priiate Investrent 4%

-4% J Jordan Tunisia Mauritius Mahysia CostaRka

Source: Authors’ calculations based on regression results of Table 1.8 Note: Each colored box bar shows the additional change in per capita growth rate ifDjibouti’s performance had matched the performance of benchmark countries between 1970 and 1999.

16 Growth that can be.. ..reforms that holdpromisefor higher growth 1.33. Figure 1.19 uses the regression results from Table 1.18 to assess the growth dividends if key growth-promoting policies and institutions are improved (or worsened, in some cases) to equal those of the comparator countries. For instance, if Djibouti’s investment climate performance across the five structural areas had equaled that ofTunisia, its future per capita growth would rise to almost 9.4 percent; if it had equaled that of Mauritius, its future per capita growth would rise to more than 11.8 percent.

1.34. The highest payoff areas for growth are stronger private investment and human capital accumulation. Other reforms with significant growth dividends include improving macroeconomic performance and institutional quality.

I Faure 1.19 Ribouti: Growth That Can Be 14% Openness 12% Macrostability 10% 0 lnsttutonal Quality 0 Fublii Investment 8% Human Capital Private Investment 6%

-2% J Jordan Tunisia Mauritius Mabysia Costaka

Source: Authors’ calculations based on regression results of Table 1.8. Note: Each colored box bar shows the additional change in per capita growth if Djibouti’s perjormance between 2000 and 2004 had equaled that of a benchmark country.

1.5. BOTTLENECKSTO GROWTH AND PRIVATE INVESTMENT: A CLOSER LOOK

1.35. This section takes a closer look at the structural areas identifled as the main bottlenecks to Djibouti’s economic growth and private investment, drawing on Favaro (2005). The private sector in Djibouti contributes only about 20 percent to GDP and ensures about 30 percent offormal employment. Most of the 2,000 enterprises reported by the Chamber of Commerce are small or medium size.

1.36. There are several explanations one could offer for the lack of a vibrant private sector in Djibouti political instability, price distortions, poor macroeconomic performance, and institutional weaknesses. This section takes a closer look to each ofthese factors.

Political instability

1.37. Political instability deters growth and private investment. In Djibouti, political instability was associated with lower growth, a decline in investment, and higher volatility (Figures 1.20- 1.23). After independence from France in 1977, Djibouti was left with a government that enjoyed a balance between two ethnic groups, the Issas and the Afars. But the country’s first president, Hassan Gouled Aptidon, soon installed a one-party state dominated by the Issa community. Afar resentment erupted into a civil war in the early 1990s and though Gouled, under French pressure,

17 introduced a limited multiparty system in 1992, the (predominantly Afar) Front for the Restoration of Unity and Democracy (FRUD) was not allowed to participate in the first multiparty elections. The conflict ended in 1997 with a peace agreement that brought the moderate faction ofFRUD into government. The country's economic recovery in the aftermath of the civil conflict was associated with greater political stability and peaceful elections. In May 2001, the government signed a second peace accord in exchange for a power-sharing deal. Since then, the country has enjoyed political stability.

Figure 1.20 qibouti: blitical stability, growth and econonic vobtility figure 1.21 Djibouti Foreign Direct hveshnt Share in GOP versus PDPicai Stability

10 7 T4 ...... T4 -- 2 .i8 -- 0 -- -2

L -4 -- -6 -- -8 ...... -8% 1 1-10

1-rea GWgrowth rate -Pdity (CEM)1

Flgure 1.22 Djibouti: Rivate Investrent Share in GWversus Flgure 1.23 Qiboue: Wbiic hvestment Share m ODPversus Political Stability Political stability

12 -

2 lo- 0 -2 4 a- % 6- -4

-6

2- -8

-Rhrate hv. To GDP -polity (m Source: WDL 2004 and Polity IV dataset. Note: Political instability is measured using the polity variable from the Polity IV dataset (Marshall and Jaggers, 2000) which ranges from -10 (mot autocratic regime) to 10 (most democratic regime).

Price distortions

1.38. Price distortions cause a wedge between the demand and supply value of goods and production factors, and affects efficiency of resource allocation. Particularly damaging price distortions for growth and private investment fall on the exchange rate, interest rate, and unit labor costs.

1.39. Exchange rate. As a result of depreciations in the currencies of Djibouti's main trading partners, Djibouti's exchange rate appreciated by about 60 percent in real effective terms between 1980 and 2001 (Figure 1.25). Despite the recent depreciation of Djibouti's real effective exchange

18 rate (reflecting the weakening in international markets of the U.S. dollar, to which the Djibouti franc is pegged), the exchange rate remains largely overvalued. Movements in the real exchange rate do not seem to have much effect on real GDP growth and the current account balance. This could be explained by the structural features of Djibouti's economy, with its almost nonexistent tradable sector. The improvement in the fiscal balance since 1998 seems associated with the depreciation of Djibouti's real exchange rate. This could reflect the predominant role of official grants (mostly denominated in U.S. dollars) and U.S. military aid in the country's revenues.

170 170 Fgure 1.24 Qhuti: RBal Effedke Exchange Rate (ER)ard Real GDP IM' Figure 1.25 Djibouti. Exchange Rate Developments, (1980-2002) \ . Grw th (15%=100) /',

6% 140 4% 2% 130- 0% Nominal effective exchange rate -2% 120' I20 40 -4% 110' 20 .6% r . Rclative Riea I mmmmmmmmmmmmmmmmmmmooooo mmmmmmmmmmmmmmmmmmmooooo ------NNNNN

80 80 -tREER -real GDP growth rate

~ 70 70 1980 1532 1534 1986 1988 1990 1592 1994 1996 1998 ZOW ZWZ

Figure 1.26 Djibouti: Real Effective Exchange Rate (REQ) ard F6Cal Fgure 1.27 Qibouti: Real Bfeche Exchange Rate (m)and Current Babnce as Share of GOP Account Babnce as Share of GOP

10 160 3 T 15

5s ob CI -5 60 : 40 -10 f 20 ~r~~t~w~mmo-~~t~mhmmo-~~tmmmmm~mmmmmmmmmmmmmmooooo-15 ~~~~~~~~~~5"~"""""""~~~~~

:!! :!! -t REER -Fiscal Balance to GDP

Source: WDI, WE0 2004

1.40. Wage rate. Wages in Djibouti are extremely high compared with neighboring countries, amounting to five to seven times the level in Yemen and Ethiopia. Average salaries for public sector employees are more than 10 times the level in Ethiopia and about 5 times the level in Yemen. According to a World Bank survey camed out for the purposes of this report in Djibouti, Ethiopia and Yemen, the wage differentials for Djibouti are higher for slulled employees than for unshlled workers (manual laborers, hotel cleaning staff, and domestic servants). See Annex 3.3 on details ofthe survey to managers ofprivate sector firms carried out in March 2005.

Figure 1.28 Djibouti: Average Monthly Civil Servant Wages (US$) Figure 1.29 Djibouti: Average hourly wages by job type (US$)

900 3 I_ 12 800 -- 3.0 700 -- 2.5 600 -- -:" f 2.0 Multiplee 500 -- of i 3 \verage / Djiboutian --6 21.5 400 -- 1 e 1.0 300 -- 4 200 -- Monthly Wag- 0.5 2 100 -- 0.0 01 0 Djibouti Ethiopia Yemen Djibouti Y- Ethiopia Hotel cleaning staff 0 carpenter$ (skilled) carpenters (unskilled)

Source: World Bank Survey, 2005 Source: World Bank Survey, 2005

19 1.41. Interest rate. Domestic capital costs are high-despite a low cost ofmobilized resources- owing to the underdeveloped financial sector, which has not been able to mobilize domestic savings and has restrained credit availability to the private sector (see chapter 4 on the financial sector). The high lending rates offered by commercial banks have discouraged the development of credit to the private sector. Lending rates average 13 percent on secure loans, and between 18 percent and 25 percent on cash facilities. Interest rate spreads-the gap between lending and deposit rates-has risen since 1995, to nearly 11 percent by the end of 2004. This is consistent with local banks’ increased provisioning of nonperforming loans. It also reflects declining competition among operating banks (which have fallen in numbers from five to three). The absence of concrete measures to reduce lending risks is an important factor behind the high level ofinterest rates.

Fgure 1.30 Djibouti: Interest Rate Spreads (%) Figure 1.31 Djibouti: Domestic and Foreign Interest Rates (percent), 1990- 112 2002 1 14.0 , 3-month LIBOR t 12.0 D. l6 -5 10.0 O. 8.0 nE -E: 6.0 4.0

2.0 USDiscount rate lo 895 W96 1997 BQ6 WQQ 2000 2001 2002 2003 I I O.O

L 1 Source: Djibouti authorities 1.42. Although localfinancing is arguably a binding constraint, the same cannot be said about access to international finance. International rates are low and accessible, and Djibouti is able to tap into favorable lending terms (on average, 1 to 2 percent interest rates on multilateral lending, 10-20 years grace periods, and long maturity periods, ranging from 20 to 40 years). Almost 84 percent ofcapital expenditures in Djibouti are financed through foreign concessional lending.

1.43. High production costs affected by utilities prices. Electricity costs industries about US$0.35/kWh, and telephone calls abroad range from more than US$1 per minute to neighboring countries to almost US$3 per minute to Europe and US$4 to the United States. Water tariffs are two to three times higher than comparator countries. Utilities are expensive mainly because of high wages, a large and increasing tax burden, and poor management practices. The main public enterprise accounts show that the wages and salaries ofpublic enterprises range from 3 1 percent of value added for Djibouti-TClCcom to more than 50percent for the airport. The wage bill of the electricity company increased from 37 percent of value added in 1993 to more than 50 percent in 2000. Poor collection rates and fraud are other problems faced by some public enterprises, including the electricity company and the water company, that result in high prices.

Figure 1.32 Djibouti vs. Corrparators: Rgure 1 33 Qiboub vs Conparators Fylure 1 34 Qibouti vs Conparators Average Telephone Rdtes (LiSW3 mn). 2001 Average Water Tartfs (USIYcubic meter), 2W1 Average Eectrlcty Tariffs (WkWh). 2001 0.25 030 7 Industry 0.25 rn m l-busehold 0.30 0.20 0.15 0.20 0.10 0.10 0.05 0.00 0.00 qibouti Yemen Ethiopia Mauritius Djibouti Yemen Ehopia Mrocco Maurlius Djibouti Yemen Ethiopia Mrocco

Source: WDI, IrU, 2001 Source: ONED ,2001 Source: EDD, 2001

20 1.34. ~~~croeco~io~~~~cstabiltty has been broadly ~ai~t~~ne.~~.'I'he currency hoard man, wmcnt has provided financial stability and low domestic ~~~~~a~~on~a~era~~n~2 percent in recent years). EIowcser, it did riot prcvcnt large structuraf fiscal deficits. ~~~i~o~~i'sfiscal policy stance has been pro-cyclical and the deficits the.mselve.s, far from being merely cyclical, have. hccome more structural (Figures 1.35-1.383."" Wecause the currency board system excludes creating money to finance. the budget deficit, iiscal deficits have been financed through the acc~im~~~ati~~nof dorncstic payment arrears, li hich amo~~~~ttdto 28 percent of GDP sn 2002. At more than 63 percent or GDP, the. burden of foreign and dom~s~~cpublic debt in ~~i~o~~~its reiativelp high for a ~~e~rel~~pin~ con^^^^ ~ulnerabl~to shocks. The. ~r~i~r~~~~e~~~c~~~in the country's fiscal posttion m 2003 and 21103 IS explained by a ~~~~ndfai~of iritl ~~~~-rcIated revciiues, rathtr than a strong fiscal c~~ns~)~~da~~on e ffort .

21 during the same penod (to 11.7 percentage paints of C;I>P). The steady decline of tndrreet taxes reflected a fall 111 lrnpclrt taxes and duties on acemint of an ~~~~a~~r~blee~~~~u~~~i~ en~~ronmen~I This reflects the s~iet~ral features of ~~ib~}u~I~se'crsnomy: Imparts represent more than 90 percent of total c~~nsum~~ti~~1~.In the absence of a VA'L import duties and taxes are de Fieto domestic c~~ns~~~p~iontaxes, re~r~sen~i~~~ more than SO percent of totali tax revenues, Locally produced goods from the manu~ac~~~ringand a~i~~il~~r~sectors represent 1 percent of CKP. Weak institutional quality andpoor investment climate

1.47. A key aspect of the quality of the business climate is the rule of law-the security of property rights. Main concerns expressed by investors include the unpredictability of the judiciary (because of the lack of autonomy of judges) and the slowness of the court system in resolving disputes. A good illustration of the effect of these governance weaknesses on the business climate and on the cost of doing business in Djibouti is the lengthy liquidation of two commercial banks (A1 Baraka Bank and the Banque de Djibouti et Moyen Onent), which has yet to be finalized. The two banks were placed in liquidation in 1998 and 1999, respectively. But the recovery oftheir debts and the gradual reimbursement of depositors have been slow, owing to judicial delays and difficulties in liquidating real state collaterals.

1.48. The attractiveness of the local environment to private investment is also hindered by a Jawed judicial system. The opacity of decisions in cases of business conflicts, lengthy appeal procedures, insufficient separation ofpowers, and a lack of independence in judges feeds suspicion in the business community and discourages investors. The recent improvements in the judicial system's physical environment, as well as fairer compensation of judges, have added some credibility to the system, but there is still a way to go to institute a state of lawfulness and make private investors feel secure.

Figure 1.44 Djibouti vs. Comparators: Figure 1.45 Djibouti vs. Comparators: Government Effectiveness Rule of Law

U Djibouti Sub Middle East Asia Latin OECD Djibouti Sub- East Asia Middle Latin OECD Saharan East & America Saharan East & America Africa North & Africa North & Africa Caribbean Africa Caribbean

Source: Kaufmann, Kraay and Mastruzzi (2003). Source: Kaufmann, Kraay and Mastruzzi (2003)

1.49. Investors in Djibouti are also concerned about high transaction costs, triggered by administrative delays and cumbersome procedures for starting and exiting a business. This concern has hampered the development of small and medium enterprises, leading a growing number of small entrepreneurs to operate in the informal sector. The National Agency for the Promotion of Investment (ME) is intended to gradually assume the task of promoting private investment-a function currently split between various state institutions-and offering a one-stop shop for any new investor.

1.50. Another obstacle for businesses operating in Djibouti is the unpredictable application of rules and regulations, which impedes analysis of investment returns. Entrepreneurs have expressed concerns about bureaucratic arbitrariness in determining taxable profits, inconsistent application oftax exemptions, and the lack of a commercial court for the resolution ofcommercial disputes (see chapter 3).

1.51. The complex tax system penalizes economic activities and is weighted against small enterprises. The tax burden weighs heavily on producers (high profit tax, fuel tax, and social contributions). The top marginal personal income tax rate is the same as the top marginal corporate

23 income tax rate (25 percent), which is considered good practice. But serious imbalances exist in the taxation of corporate income. The tax system offers generous tax incentives to foreign investors, but domestic firms face high taxes on wages and salaries. This imbalance is exacerbated by high social security contributions (amounting to almost 20 percent ofgross nominal wages). In addition, the complex corporate tax system increases transaction costs for the private sector, particularly for small businesses that cannot bear extra costs to pay experienced tax advisors (Djibouti Public Expenditure Review, 2004).

Summing up... identijjing the main bottlenecks to growth andprivate investment

1.52. This section briefly summarizes the main bottlenecks to private investment in Djibouti, drawing on the analytical framework developed by Hausmann et a1 (2004). This framework uses a decision tree to identify the most binding constraints on growth and private investment (Figure 1.46). The first question is whether private investment is low because returns to investment are low or because costs offinance are high.

Figure 1.46. Djibouti: Low Private Investment

1.53. In Djibouti returns to investment are low, as evidenced by the weak productivity of capital and labor over the past two decades. This report argues that low returns to investment reflect mostly government failures rather than market failures. The small size of the country and the fact that more than 80 percent ofthe population and economic activity is concentrated in the capital city mean that information externalities are not likely to be a binding constraint on growth. In addition, the government plays a predominant role in Djibouti’s economy-directly, through the budget, and indirectly, through its rules and regulations, its pay and employment policies, and its management ofkey public utilities.

Among the main constraints deterring growth and private investment in Djibouti, these are the strongest candidates:

24 1.54. Low private appropriability of returns. In Djibouti an unfriendly business environment leads to high transaction costs. Administrative red tape, weak enforceability of contracts, and regulatory uncertainties owing to lack of transparency and predictability of rules and regulations increase the costs of doing business (see chapter 3). &gid labor market regulations and high taxes on labor also help explain the low private appropriability of returns, particularly for the domestic private sector (see chapter 3). High public sector wages, which are a benchmark for private sector wages, drive up unit labor costs and public utility prices, leading to higher production costs (see chapter 3).

1.55. Low social returns to factors ofproduction. This is explained by insufficient investment in human capital and inefficient provision of infrastructure. While Djibouti has made substantial efforts during the past five years in investing in its population's education, educational attainment of the labor force in Djibouti remains low (the primary education completion rate is less than 60 percent). For this to be a binding constraint Djibouti should have high returns to education, which is the case. High returns to education mean that given the scarce supply of educated people, the economy is willing to pay high returns to the few who are educated. Quantitative estimates of returns to education in Djibouti from standard Mincer regressions (see chapter 3) show that returns to education are high. Private returns to university education are close to 14 percent for graduates workmg in the public sector. A university-educated worker in Djibouti earns on average US$328 per month more than a worker with basic vocational training." In addition, unemployment (or more accurately, underemployment) is higher among uneducated workers, which suggests that current growth and private investment trends are constrained by inadequate human capital. A similar analysis can be applied to infrastructure. Costs ofproviding infrastructure service remain very high across key infrastructure sectors: energy, water, telecommunication, and transport (see chapter 5).

1.56. High cost of local finance. Access to foreign savings is not a binding constraint for investors, owing to the country's financial openness. However, the high costs of local finance remain a problem for domestic private investors. The high costs of capital and the interest spreads are explained by Djibouti's undeveloped financial system, which is characterized by poor financial intermediation and diversification. This affects access to finance for small and medium enterprises (see chapter 4).

Afinal note on statistics...

1.57. Caution must be exercised when interpreting the findings of this chapter given the serious data limitations. Economic and financial statistics suffer, although unevenly, from deficiencies with regard to their quality, coverage, periodicity, timeliness, and dissemination. The situation is particularly critical for the national accounts and to some extent the fiscal data. Only public sector employment and wages data are available for the labor market. Employment and wage data for public enterprises and the formal private sector, which were made available in the 1980~~are no longer provided by the authorities. Furthermore, no efforts have been made to quantify the impact of the informal sector on employment or GDP (see annex 1.3 for details on outstanding statistical challenges).

l1In the private formal sector, the premium accruing to university education over vocational training using average monthly wages is US$43/month-far lower than the premium in the public sector, whch is US$33 l/month.

25 1.6. CONCLUSION AND POLICY IMPLICATIONS

1.58. This chapter reviewed the main patterns of Djibouti’s economic performance over the past two decades. It showed that output growth fell sharply through the 1990s and displayed high volatility. Declining output per-capita was caused by falling productivity which in turn was due to political instability. This drove down the marginal productivity of capital and reduced private rates of return. To reverse these trends, productivity increases are needed which can be induced by a stable environment that is conducive to investment. The economic recovery that has taken place since the late 1990s seemed to relate more to the postconflict reconstruction that tends to take place after peace is achieved than to any fundamental change in policy. In addition, a surge in FDI and port receipts, associated with the strong resumption in port activities, and official capital inflows, stemming from the increased foreign military presence, led to a boom in domestic consumption that drove economic growth.

1.59. Despite Djibouti’s recent economic recovery, growth fell short of providing jobs and reducingpoverty. As the population grew, per capita income experienced a sustained fall and was negative during most of the past two decades. Against this background, this chapter assessed the main sources ofeconomic growth and bottlenecks to growth. The chapter points to four interrelated areas of underperformance: low levels of human capital and private investment, macroeconomic instability (particularly fiscal policies that have not been conducive to growth), and poor institutional quality. These weaknesses offset to some extent the advantages offered by Djibouti’s strategic location, its current political and financial stability, the relatively advanced port infrastructure, and the openness ofits trade and financial policies.

1.60. To accelerate growth in the short term and sustain it over the long term, Djibouti policy makers will need to rise to a multifaceted challenge in the years ahead by: Furthering fiscal consolidation efforts to support growth, reduce poverty, and create jobs. Reducing the wage bill by reforming public sector pay and employment policies will help attain a sustainable fiscal path over the next 10 years, while making fiscal room to finance human capital investments (that is, to attain the education MDGs) and other poverty-reducing expenditures. It will also increase budget flexibility so the government can better manage economic volatility. And it will lower unit labor costs in the economy, because public sector wages serve as a benchmark for private sector wages. Pressing ahead with structural policies needed to energize private investment and support human capital accumulation, by strengthening the investment climate (lowering transaction costs associated with regulatory uncertainties); enhancing labor productivity through stronger labor market flexibility; attaining the education MDGs and expanding the formation of high-skilled labor; mobilizing finance for investment; and improving the efficiency and competitiveness ofbackbone infrastructure services.

Part I1 (chapters 2-5) explores these policy priority areas in detail to pave the way for Djibouti’s growth reform agenda in the years ahead.

26 PART 11. DJIBOUTI’S GROWTH REFORM AGENDA: THE ROAD AHEAD

Part I showed that Djibouti’s weak growth is mainly explained by underperformance across four main fronts: low private investment, low human capital accumulation, weak macroeconomic performance (and particularly limited $seal space), and poor institutional quality. nese factors are interrelated. Part 11 (Chapters 2-5) elaborates on each of these priority areas and identifies options for reform. In particular, it shows how government pay and employment policies, the labor market, education and training policies, the investment climate and the availability ofjnance and infrastructure services might be affecting the quantity and quality of private investment and the country’s growth potential in diflerent ways.

27 CHAPTER 2. MACROECONOMIC AND FISCAL STRATEGY

Chapter 1 used cross-country regressions to estimate the growth potential of Djibouti’s economy. Notwithstanding the limitations ofthis approach, the exercise highlighted the main policy areas that influence growth prospects. The key message from the empirical analysis is that growth and job creation in Djibouti will depend on successfully implementing key policy reforms that promote private sector-led growth and human capital accumulation while maintaining macroeconomic stability.

This chapter complements the empirical analysis with a review of Djibouti’s medium-term economic outlook and a discussion of prospects for growth over the medium term. The chapter discusses medium-term macroeconomic scenarios based on different policy reform assumptions. The final section discusses options to reform current government pay and employment policies with the aim ofreducing labor costs and lowering unit costs ofproviding public services. The measures proposed are meant to promote private investment and human capital (through the attainment ofthe education MDG by 20 15) while preserving fiscal sustainability over the medium term.

2.1. MEDIUM-TERM ECONOMIC OUTLOOK

2.1. Djibouti’s short- to medium-term economic outlook looks favorable, but in the absence ofpolicy reforms, growth will remain insufficient to reduce poverty and create new jobs. Djibouti will benefit from the windfall of military-related revenues, the demand for services by foreign military expatriates, and new FDI flows in the new port. In addition, growth could be stimulated by the public investment program to support the implementation ofthe PRSP.

2.2. The Djibouti authorities have committed to launch an ambitious public investment program as apillar of their strategy to reduce poverty. The country’s PRSP was endorsed by the Boards of the Bank and the IMF in June 2004. One of the key goals is to achieve the MDGs by 2015. The implementation of these poverty reduction objectives will increase pressures on public spending. Two recent developments will bring additional sources of revenue: Dubai’s plans to finance the new port and the military cooperation agreements signed with the United States.

2.3. The Emirate of Dubai is planning to finance a new port in Doraleh with an industrial and commercial free zone that could spur growth in the medium term. This project will cost about US$400 million (about 60 percent of GDP) over five to seven years. It will include three main phases: a new oil terminal, a container terminal, and an industrial zone. The construction ofthe new oil terminal started in 2003. A large part of the total cost will likely be spent on importing equipment and other inputs for the project, as well as foreign labor. The regulatory framework under which the industrial zone will operate is still under discussion. However, the spillovers ofthis FDI on the rest ofthe economy could be limited without concomitant policy reforms to improve the investment climate, lower unit labor costs, and enhance labor market flexibility.

2.4. Foreign military personnel stationed in Djibouti will bring additional revenues but the net impact on growth depends on how they are used. The foreign military forces stationed in Djibouti will continue to bring additional revenues in the next few years. The challenge will be to use these resources as efficiently and transparently as possible to support growth, reduce poverty, and reduce the large stock ofbudgetary arrears.

28 2.5. Without policy reforms, Djibouti’s fiscal position is likely to deteriorate over the medium term and growth will remain insufficient to reduce unemployment. Military revenue is expected to remain fixed in nominal terms after 2006. Even if domestic revenue rises and expenditure grows moderately, the overall deficit (on a commitment basis, including grants) is expected to deteriorate, averaging about 3 percent of GDP over the next 10 years, and the basic fiscal deficit (excluding grants and military-related revenues) to average about 6 percent. With limitations imposed on domestic financing under the currency board arrangement and without increasing net foreign borrowing, a fiscal financing gap is likely to be closed by a net accumulation of domestic arrears, barring further consolidation efforts. As a result, average GDP growth of 3.2 percent would not be sufficient to reduce unemployment.

2.1.1. What growth can be expected in the coming decade?

2.6. This section presents two medium-term macroeconomic scenarios over the next 10 years (2005-15).

0 The base case scenario, which may be considered as a “low case” scenario for some countries, assumes a status quo situation in which the authorities do not undertake further structural policy reforms. Under this scenario, growth is assumed to be maintained at current levels, hovering around 3 to 3Spercentper year during the coming decade.

0 The high case scenario assumes a nominal wage bill reduction and the achievement ofthe education MDG by 2015. This scenario assumes higher growth, ranging from 4 to 4.5 percent per year, based on improved investor confidence and stronger donor support. Table 1.1 presents the results ofthe two scenarios, comparing key variables over period averages.

Base case scenario

2.7. In the first half of the projection period, Djibouti would benefit from a relatively improved growth rate, compared with historical levels, owing to the higher investment in the port. The investments associated with building a new port in Doraleh and the public investments in support of the government’s PRSP program are expected to maintain current growth levels during 2006-09, with growth growing on average at about 3.6 percent, and leveling off to 3.5 percent over the second halfof theperiod. This rate is still lower than the growth anticipated in the PRSP. Inflation is expected remain stable over the projection period, averaging 2.1 percent, and the exchange rate is expected to remain fixed at DF 177.72 to the U.S. dollar.

2.8. In the absence offiscal consolidation efforts, the wage bill would decline only minimally from its current level of almost 14 percent of GDP to about 12 percent of GDP by 2015, contributing to an increasing fiscal deficit. On the expenditure side, higher expenditures in the latter part of2004 and early 2005, because ofthe presidential electoral period, are assumed to level off (reaching 33 percent of GDP by the end of the period). But because of growing demand for public services, public expenditures are expected to remain high on average during the period. On the revenue side, direct tax revenues would remain stable over the period. Indirect tax revenues (mostly import taxes) would decline after pealung in the early part of the period. With military- related revenue expected to remain fixed in nominal terms after 2006, the overall fiscal deficit,(on a commitment basis, including grants) is projected to rise by 3 percent over the period. However, with expenditures exceeding domestic tax revenues, the basic domestic deficit will remain in deficit, hovering around 6 percent ofGDP. The overall fiscal deficit is expected to reach 3 percent.

29 The authorities would not be able to pay off arrears, resulting in sustained high public debt relative to GDP.

2.9. Current account pressures would intensify financing needs, leading to an increasing debt-to-GDP ratio over the projection period. Under the base case scenario, total financing needs would double, from US$119 million in 2005 to US$228 million by 2015. The current account deficit would be partially financed though external assistance in the form of official grants, though additional new lending would need to be negotiated to close the financing gap.

High case scenario

2.10. In the high case scenario, growth is expected to hover around 4.5 percent over the projection period. Investment and imports would continue to drive growth, with investment as a share ofGDP rising to 22 percent by 2015, compared with 15 percent in 2003. Inflation is expected to remain stable, at about 2 percent ofGDP, and the exchange rate policy is expected to remain the same.

2.1 1. The high case scenario assumes higher investment rates driven by structural reforms aimed at energizing private investment. In particular, the high case scenario assumes that the authorities will simplify the regulatory framework for investment, rationalize the tax and customs administration systems-lowering transaction costs associated with tax compliance-and further enhance labor market flexibility. In addition, the high case scenario assumes reforms in government pay and employment policies aimed at lowering labor costs. Reducing public sector wages would also help reduce unit costs of providing public services. Coupled with greater private sector participation in managmg key infrastructure services, this reduction will help reduce production costs, spurring firm productivity and growth.

2.12. The high case assumes a reduction in public sector nominal wages that would improve Djibouti’sfiscalposition. In the high case scenario the wage bill is expected to be reduced from 14 percent of GDP in 2004 to 10 percent of GDP by 2015. Investment expenditures are expected to pick up owing to greater import needs driven by the PRSP implementation and stronger private investment. These pressures on public spending would be partly offset by increased import taxes and net fiscal savings deriving from improvements in tax and customs administration,I2 allowing the fiscal deficit (on a commitment basis) to remain stable, averaging 2.4 percent of GDP over the projection period. Military-related revenue is assumed to remain constant in nominal terms, representing on average 6 percent of GDP. Over the same period, on a cash basis, the fiscal deficit would average 3.5 percent ofGDP and the primary fiscal balance would average 4 percent ofGDP, which is consistent with a medium-term fiscal sustainability path (see the Debt Sustainability Analysis carried out by the World Bank in the Djibouti Public Expenditure Review, 2004). This reduction in the wage bill would allow repayng the domestic stock of arrears with an annual payment of DF 2.2 billion. And as a result of these fiscal consolidation efforts, the public debt (foreign and domestic) would be reduced from 83 percent ofGDP in 2004 to 54 percent ofGDP by 2015.

2.13. The reduction in nominal wages in the public sector would also lower unit labor costs, making the economy more competitive and allowing public services to be provided more efficiently. Reducing public sector wages would bring down private sector wages, bringing them l2In the high case scenario it is expected that the revenue effort would be slightly strengthened. Because the tax effort is already high, the main hstof fiscal policy reforms is expected to come from the expenditure side, particularly from efforts to reduce the wage bill.

30 closer to the level of imported labor (expatriate workers) expected to be hired in the new port and free zone13. This would increase the chances of higher FDI spillover effects on the rest of the economy. With this reduction in unit costs ofproviding public sector services, there will be fiscal room to finance the recruitment of the additional teachers needed to attain the education MDG by 2015. This policy reform could therefore help increase human capital accumulation and promote growth in the medium run.

2.14. The current account balance will continue to be pressured by high import demands; however, increased FDIflows and official transfers rejlecting stronger investor confidence and donor support would allow the debt-to-GDP ratio to remain stable. Financing needs would be supported by higher inflows of FDI and donor support. Reserves (as measured in months of imports) are expected to remain stable. The debt-to-GDP ratio would average 62 percent of GDP over the projection period.

13 The government implemented a 20 percent reduction in nominal wages for the civil service in 1998. The absence of comprehensive wage and income data for that period covering public and private workers in various sectors of the economy across the income distribution prevented an assessment the impact of the nominal wage cut on private sector wages.

31 Real GDP growth (YO)

2006-09 2010-1 5 Gross domestic investment (YOof GDP) 2004 21 .o 21 .o 2005 22.0 22.4 20 0 6 - 0 9 20.8 22.4 2010-1 5 19.7 22.1 Fiscal balance (YOof GDP, commitment basis, incl. grants) 2004 -2.3 -2.3 2005 -2.7 -2.5 2006-09 -3.2 -2.4 2010-1 5 -3.8 -2.5 Basic domestic balance’ (YOof GDP) 2004 -6.7 -6.7 2005 -4.8 -6.0 2006-09 -4.2 -5.5 2010-1 5 -4.0 -3.3 Stock of domestic arrears (YOof GDP) 2004 20.5 18.7 2005 19.3 16.0 2006-09 16.8 10.2 2010-1 5 12.8 2.0 Wage bill (YOof GDP) 2004 13.8 I3.8 2005 13.0 13.0 20 0 6 - 0 9 12.7 12.4 2010-1 5 12.0 11.0 Current account balance (YOof GDP) 2004 -10.6 -10.6 2005 -12.9 -13.5 2006-09 -14.0 -13.8 201 0-15 -15.0 -13.0 External debt (YOof GDP) 2004 62.6 62.6 2005 63.3 64.0 2006-09 61.4 62.4 2010-1 5 71.5 62.2 Note: a. Domestic balance is defined as domestically financed expenditure less domestic revenue.

2.2. REFORMING GOVERNMENT PAY AND EMPLOYMENT POLICIES: EVALUATING OPTIONS FOR REFORM

2.15. Djibouti faces a turning point where, unless sustained efforts to stimulate private investment are placed at the core of the reform agenda, the windfall of military revenues and new investments in the port may not fulfil their development promise. Prospects for growth and job creation will depend on successfully implementing key policy reforms that promote private sector-led growth and human capital accumulation while maintaining macroeconomic stability.

2.16. The high public sector wage bill slows growth. In the standard Keynesian model, fiscal expansion boosts aggregate demand and leads to an economic expansion. Recent empirical evidence shows that changes in public spending have a bigger impact than tax changes do (Alesina et al. 2001). Particularly important are changes in the public wage bill and government transfers. In

32 Djibouti, high wages cut into profits, reducing investment. The country has a very high level of public sector entitlements and yet a very low level ofservice delivery.

2.17. This section focuses on policy reform options for reducing the public sector wage bill. These reforms aim at reducing labor costs and lowering the unit costs ofproviding public services. The objective is twofold: to attract private investment and to create fiscal room to finance additional human capital investments.

2.2.1. The Public Sector Wage Bill: The Magnitude of the Problem

2.1 8. Despite considerableprogress over recent years, Djibouti still has one of the largest wage bills among comparator countries, particularly considering the low level of service delivery. It represented nearly 14 percent ofGDP in 2004, 60 percent of total fiscal revenues, almost 50 percent of total current expenditures, and more than 40 percent oftotal government expenditure^.'^

Flgum 2.1: Djibouti: Wage 8111,1S90~2004

04 ...... ,,.,.,.. 1993 1091 1992 1993 1994 1995 1996 1097 1995 1999 mW 801 m02 803 mC4 -In percent of GDP -In percent of current exp In percent of total revenue In percent of Total exp I - I Source: Djibouti authorities

2.19. In the early 1990s, the public wage bill represented on average 19 percent of GDP, before rising to about 24 percent of GDP during 1993-95, mainly as a result of the mobilization of security personnel in the face of ethnic strve. This trend started reversing in 1995 and the government wage bill was reduced (from 24.2 percent of GDP in 1995 to 16.4 percent of GDP in 1998), because of several measures: partially demobilizing 1,500 soldiers, eliminating irregularities in employment and cash allowance surveys, cutting back benefits and salaries of embassy personnel, and eliminating overtime work. The expenditure curtailments, a major component of Djibouti’s stabilization efforts since 1998, were instrumental in bringing down the share of wages and salaries to GDP to 14 percent by 2003. Measures included: cutting by 10 percent nominal wages; and by another 10 percent nominal wages for civil servants whose salaries exceeded DF 80,000 per month; freeze in nominal wages since 1993; hiring freeze in the civil service except for the social sector, and not replacing civil servants eligible for retirement (amounting to 800 employees).

2.20. The driving factor behind the exceptionally large wage bill is the high level of civil service pay. The annual average pay of a civil servant is about 10 times the country’s GNI per capita. Over the period 1990-2004, the annual salary of civil servants excluding the military

~~ ~ l4The high level and the structure of civil pay were inherited from the French colonial administration and has changed little since independence. By virtue of Djibouti’s currency board system (under which the nominal exchange rate has not changed since 1971), inflation is only about 2 percent, which does little to adjust real wages down as in other developing countries.

33 averaged about US$8,000. By 2004 it was more than US$9,OOO-that is, 10 times the GNI per capita of the country (estimated at US$949). Including the military reduced the differential to 7 times GNI per capita. The average monthly civil servant wage was also 7.3 times higher than the minimum monthly wage of DF 18,850 in 2004.

Figure 2.2 Djibouti: Average PuMic Saor Wage vs. GNI per caws. 19952CO4 (USD) Figure 2.3 Djibouti: Relatie Weight of Public Sector Wages, 2004

12 984

Average hbli Average hbli Average hbltc Average hblic 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Sector Sector Wage Sector Wage Sector Wage/ WageIAverage (average)/GNI per (including the Minimm Wage Frvate Sector capita military)/GNI per DAvg. Yearly Public Sedor Salary (USD) +GNI per capita (USD) Wage capita Source: Djibouti authorities and WDI, 2004 iource: Djibouti authorities and WDI, 2004

Figure 2.4 Djibouti: Ratio of average public sector wage to per capita GNI, 2004

Source: Djibouti authorities; WDI, 2004

2.21. Djibouti’s civil service is characterized by a highly skewed wage distribution (Figure 2.5). Data availability limits our analysis to 1996, using a representative household survey. The modal monthly wage is approximately US162 for men and US$224 for women, which is still large relative to neighboring countries, including Yemen.

Fgure 2.5 Djibwti: Monthly Wage Ustribution in Fublc Sector, 1996

0.0025

0.0020

0.0015 2. fl cv) 0.0010 m n 0.0005

0.0000

-0.0005

Monthly Wages (USD) I Source: Authors’ calculations using EDAM 1996

34 2.22. There is also a wide dispersion in wages within the public sector. The ratio of the highest earning worker to the lowest is 10.3 for permanent employees and 15.6 for contractual employees. From Table 2.2 and 2.3, A rough pattern also emerges within categories: lower-earning categories (Cl, C2, D, and 1, 2) see larger within-category dispersion compared with higher- earning categories (A and 6,7).

Dispersion Ratio Note: The dispersion ratio is defined as the ratio of maximum wage to minimum wage in each category. The overall dispersion between the largest and lowest civil servant's monthly wage is 10.3.

Dispersion Ratio1 1.91 1.71 1.41 1.31 1.41 1.31 1.01 1.71 Note: The dispersion ratio is defined as the ratio of maximum wage to minimum wage in each category. The overall dispersion between the largest and lowest contractual's monthly wage is 15.6

2.23. Evaluating civil service wages and salaries entails examining three key dimensions: the total wage bill, civil service pay, and civil service employment. The earlier discussion revealed anomalies pertaining to the size of the wage bill and average pay. But anomalies in civil service employment are also present, mainly because the government continues to be the employer of last resort (with public wages being more attractive than private sector wages) and also because informal quotas are determined along ethnic lines.

2.24. Determining whether there is excess or surplus civil service employment ultimately depends on the appropriate role of government within and across sectors. l%s involves identifying the major programs in which civil servants are employed, evaluating whether government spending on these programs is justified, and assessing the appropriateness of the wagehonwage balance within these programs. A key challenge in such analysis is to get accurate data on the growth and breakdown of employment. Weaknesses in the quality of budgetary inf~rmation'~impair a more detailed analysis of the wage bill and civil service employment by sectoral programs. A review of the distribution of the wage bill across sectors is provided in Figure 2.6, drawing from budget laws since 2002 to 2004. The ministries with the highest wage bills are the Interior, Defense, and Education. These three ministries took up 70 percent of the total wage bill in 2002-04.

l5Lack of detailed and consistent data on the number of employees and their salaries by sector and subsectoral programs impairs an accurate computation of civil service employment by sector.

35 Figure 2.6. Djibouti: Share of Gowmment Wage bill by Ministry, (in percent, period awrage. 200244) 30

F >

Source: Djibouti authorities

2.25. An obvious manifestation of excess employment is the existence of unproductive workers. The total number of unproductive workers can be identified only through a proper accounting of civil service employment with payroll lists and the employment register. There is an urgent need to compile data on civil service employment by ministries and sectoral programs. A functional review of ministerial portfolios can help identify surplus employment in specific areas. Civil service pay scales are typically determined centrally and therefore cut across sectors and programs. Yet, scarcity of data on civil service compensation and private sector wage comparators hinder a more detailed analysis of civil service pay scales.

2.26. Since there is no ‘optimal’ size for a country’s civil service, its actual size should be measured against expected outcomes. Djibouti’s public sector wages seem high when considering the low service delivery levels. Let’s take the education sector as an illustration, since the ministry of Education has the highest wage bill (excluding defense and the Ministry of Interior), accounting for almost one fourth of the total public sector wage bill in the country. Let’s then compare Djibouti’s public sector wages with the level and relative efficiency of current education expenditures on relative to two other middle-income countries in North and Sub Saharan Afnca: Mauritius and Tunisia. 16. The total public sector wage bill in Djibouti, close to 14 percent ofGDP, doubles than Mauritius’, about 6.7 percent. Total public expenditure on education is about 5 percent of GDP in Mauritius, whereas it was about 6 percent of GDP in Djibouti. Yet, the composition of education expenditures differs between the two countries. The education wage bill accounts for about 30 percent of total public education expenditures in Djibouti compared with 10 percent in Mauritius. The average salary of a teacher is about 9 times GDP per capita in Djibouti, but it is only about 1.3 per capita in Mauritius. In spite ofDjibouti’s higher spending on wages and salaries, educational outcomes in Mauritius are much higher than in Djibouti. Adult literacy rate is about 84 percent in Mauritius, compared to about 53 percent in Djibouti. Gross enrollment rates in primary and secondary education are 102 and 85 percent in Mauritjus, but are only 52 and 22 percent in Djibouti.

2.27. Comparing Djibouti with Tunisia also strengthens the point that Djibouti’s high education expenditures are not commensurate with the achieved outcomes, even afer controlling for income levels. Tunisia’s public sector wage bill amounted to 12 percent of GDP in 2004 whereas Djibouti’s wage bill was close to 14 percent of GDP. Djibouti’s public education spending, l6As requested by the Djibouti authorities, the report includes a comparison with Tunisia and Mauritius.

36 about 6 percent of GDP in 2004, was comparable to Tunisia’s level of public spending, about 8 percent of GDP in education. Djibouti’s public spending on education is considerably higher than would be expected given its income level, yet, as the figure below illustrates, by such measures as the adult illiteracy, Djibouti is not doing well as its income comparators. Djibouti’s public education expenditure has been inefficient (spent more but achieved less), compared to Tunisia (it spent slightly more but achieved much more). Djibouti’s education expenditure levels bear no strong relationship to adult literacy, completion rates and schooling standards. This contrast to Tunisia’s efficiency of education service delivery: gross enrolment rates in primary and secondary education are 109 and 81 percent respectively, compared with 52 and 22 percent in Djibouti. Moreover, primary completion rate is less than 60 percent, compared to 95 percent in Tunisia.

Djibouti: Adult illiteracy and Public Expenditures on Education

Undrashi-vr. L-.t flsi-nt Adult illiteracy rates higher than expected Adult illiteracy rates higher than expected - Educetim spending lower then expeded Education spendng higher then expected I LIC tl

A LMC

A MNA

Tunisia ~

M0.t rmcimt Ovenohl-vrr. Adult illiteracy rates lower then expected Adult illiteracy rater lower then expected Education spendlng lower than expected Health spending hlgher than expected I Ikvktionstrorn Ex- PublicI= Expenditures on Edu-tion Source: World Bank staff calculations from WDI data. 2004. Note: The x-axis measures differences between actual public expenditures on education and predicted expenditures. Predicted rates are derived from world trends. They represent the public expenditure rates that would have predicted given the countries’ income levels. The distance from the country point to each one of the axes measures the size of these differences. Accordingly, the northwest quadrant reflects the ‘zone of greatest efficiency’, followed by the north-east, then the south-west, and finally the southwest quadrant- the ‘zone of greatest inefficiency’. MNA: Middle East and North Africa countnes (except Iraq, Libya and West Bank and Gaza). LIC: Low-income countries (World Bank classification). LMC: Lower middle-income countries.

2.28. In spite the progress achieved by recent educational reforms, about 50 percent of children are still out of schools, and more than 20percent of those enrolled do not complete six years of primary education. While Djibouti’s gross enrollments increased from 39 percent in 1990 to 52 percent in 2004, owing to the substantial investments by the budget and by donor’s support, these rising rates of school enrolment are distorted by glaring inefficiencies and inequalities of access and quality in educational provision, particularly in lower and higher secondary education. The Djibouti Public Expenditure Review (2006) estimated that the overall internal efficiency is about 0.55 in the primary cycle and 0.33 in the secondary cycle. Using data on repetition and dropouts the PER estimated that these inefficiencies result in a waste ofresources amounting to 58 percent of total current education expenditures (excluding tertiary education). Unit costs of educating primary and secondary school students in Djibouti tend to be significantly higher than in other middle-income countries, owing to the large share of the wage bill in the total education budget (over 90 percent oftotal current education expenditures are wages and salaries.

2.29. Djiboutipolicy-makers are caught in a dilemma. On the one hand, they would need to maintain macroeconomic stability and fiscal sustainability over the medium-term to sustain growth and poverty reduction efforts. On the other hand, growing demographic pressures and the current

37 low levels of service delivery (with a current gross enrolment rate of 53 percent and a primary completion rate of less than 60 percent), it is clear that enrollments have to growth sharply in the future. Djibouti will require a substantial increase in public expenditure on education to meet the Millennium Development Goal ofachieving 100 percent primary completion rates by 20 15. Yet, as highlighted above, the size ofthe public wage bill in Djibouti is already high, mostly driven by an average high wage of about 9 times the country’s GDP per capita

2.30. As illustrated by the Djibouti Public Expenditure Review (2006) a reduced wage bill could create fiscal room to reallocate expenditures and contribute to achieve the education MDG. If the government is to attain the education MDGs, at present levels of financing and efficiency, the primary education sector would face an average annual deficit of about DF 2411 million over the plan period. Only with a decline in teacher salaries, the financing gap would be eliminated. The financing gap corresponds to a reduction of the wage level in the education level by about half. Educational reforms would need to put particular emphasis on decreasing the attrition of students within the three school cycles and during transition between them. In conclusion, the MDG goal of reaching primary completion rate of 100 percent by 2015, under current levels of education financing and internal efficiency, is not fiscally sustainable. The achievement ofthis goal would require a more cost-effective education structure in Djibouti; one of the ways in which this could be reached is through a reduction in the average salary of civil servants. But these simulations do not take into account other policies that could improve the internal efficiency ofpublic service delivery and lead to further improvement in cost-effectiveness (See Djibouti Public Expenditure Review for more details on these estimations).

2.2.2. High government wages affect growth, employment and poverty reduction

2.3 1. Why does this redistributive system exist? Redistribution through public employment is less visible than direct transfers and therefore politically less costly. In addition, public employment is used to correct labor market imperfections. In Djibouti, the welfare system is skewed in favor of retirees from the formal sector and does not efficiently protect the temporarily unemployed. Moreover, public sector jobs in Djibouti come with significant non-wage compensation (almost 50 percent of total public sector compensation) through worker protections and allowances (indemnities) linked not to productivity but to personal characteristics. These non-wage allowances contribute to the structural rigidities in the government budget.

2.32. High government wages affect unit labor costs and crowd out productive investments, hindering the country’s competitiveness and growth. High public wages and salaries also affect negatively the overall competitiveness of the economy and hinder growth. Downward rigidities of public wages have limited the government’s room to respond to fiscal pressures arising from the growing population and the volatility ofrevenues, and have led to increased fiscal imbalances over the years. Because Djibouti’s currency board system does not allow for creating money to finance budget deficits, financing of deficits has generally been accommodated by accumulating arrears or increasing taxes and surtaxes. High government wages have also affected the efficiency of public utilities. Wages and salaries represented between 35 and 50 percent of total output generated by Djibouti’s public utilities (see chapter 5). The wage bill of the electricity company (Electricitt de Djibouti, EDD) represented almost 30 percent of the wage bill of all public enterprises combined, about 8 percent ofthe total wage bill (government and public enterprises combined), and 2 percent of GDP. Djibouti electricity tariffs are among the highest of comparator countries, averaging US$0.20/kWh. High prices are partly explained by technical reasons (higher cost of producing electricity because of a reliance on imported diesel fuel to generate power and inefficiencies owing

38 to network losses), but they are greatly affected by public sector pay and employment practices. As will be discussed in chapter 5, public utilities suffer from high administrative overhead costs due to overstaffing and high wages.

2.33. High government wages negatively affect poverty reduction by crowding out other government spending on basic social services. Considering the very low starting level in education and health service coverage and the increasing population, achieving the MDGs will require additional hiring ofteachers, doctors, and nurses. There are few options for meeting this new hiring but through nominal wage or civil service restructuring or both. The next section will explore these options in detail.

2.2.3. Policy instruments for reducing the wage bill

2.34. Various instruments are available to Djibouti policy makers to reduce the wage bill, induce a reduction in the unit labor costs, and reduce public sector ~verstaffing'~.These objectives could be achieved through natural attrition or a concurrent hiring freeze, accelerated attrition through substantial wage adjustments or benefit cuts, or retrenchment. These measures could be taken separately, sequentially, or simultaneously. This section presents a brief overview of the range of reform options for reducing the public sector wage bill. The next section provides further details on the effects of selected options on the overall wage bill and on unit labor costs.

Attrition

2.35. Thepotential for accelerating attrition depends partly on the existing age structure of the public sector. A relatively larger cohort of older workers would facilitate measures to increase the process. Though current data are lacking, EDAMS 1 (a representative household survey conducted in 1996) provides some insight. It reveals that the public sector, excluding the military, has a relatively young workforce.

2.36. Djibouti's civil service is relatively young, which limits the scope of reforms options such as early retirementpackages. The majority of civil servants are between 25 and 29 years old, with women in this age bracket forming 27 percent of all women employed in the public sector, compared with 19 percent of men in the same category. The current age structure may take on a bell-shaped or normal distribution pattern as public sector workers age.

Figure 2.7 Djibouti: Age Structure of Civil Servants (%) 1996

I 1 Source: Authors' calculation using EDAM 1996

l7Another option is the use of progressive taxes with revenues channeled towards an MDG fund.

39 Figure 2.8 Djibouti: Age Structure of Male Civil Servants (%) Figure 2.9 Djibouti: Age Structure of Female Civil Servants (%) 1996 1996 30 1 I251 25 20 15 10 5

Source: Authors' calculation using EDAM 1996 Source: Authors' calculation using EDAM 1996

2.37. The process of attrition can be accelerated by reducing remuneration, inducing a smaller wage differential relative to the private sector. Options include nominal wage cuts. Allowing real wages to decline through negligible or delayed nominal wage increases (at less than prevailing inflation rates) is typical government practice worldwide.

2.38. There are alternative measures to adjust the civil service pay scale. Government grading systems typically set base salary compensation according to the qualifications and experience of new hires, implying some link to work productivity. In Djibouti, the civil service code defines the level of salaries according to qualifications. Since 1993, the government adopted a freeze in nominal salaries.

2.39. Efforts to reform civil service pay scales risk being undermined by large nonwage benefits that are not linked to productiviQ. Nonwage benefits represent the only variable that can be manipulated to reduce total compensation, other than nominal wage cuts and pay scale reforms. In Djibouti, the nonwage benefit premium represents a significant share of total public sector compensation (close to 50 percent). Reducing public sector employment through attrition by cutting nonwage benefits may elicit opposition, especially from teachers (the most numerous within the civil service categories). An alternative may be to reduce subsidies of available services (interrupting the current practice in which the government pays the bills of public utilities employees).

Retrenchment

2.40. The wage bill can also be reduced through quantity mechanisms-public sector downsizing through layoffs or voluntary separations.

2.41. Mandatory vs. voluntary. Retrenchment system design varies according to the voluntary or involuntary nature of job separation. In Djibouti, the main retrenchment scheme that the government has implemented is the demobilization program for ex-combatants in the aftermath of the civil conflict. The military demobilization program is voluntary, but the duration, level, and modalities ofthe compensation program are determined by the government.

2.42. Duration, level, and modalities of compensation. Compensation to retrenched workers can be made in the form of severance pay delivered in a lump sum at the moment of separation, or as periodic benefits paid over time (unemployment benefits). In Djibouti there are no unemployment benefits. Layoffs are allowed only for economic reasons and severance payment takes the form ofa

40 large, one-time, lump-sum payment. Labor Code provisions regulating severance pay are the only mechanism to help workers mitigate income loss because of unemployment. All workers whose labor contract has been terminated for economic and technical reasons are entitled to severance pay, paid by the employer as a lump sum upon contract termination.

2.43. Thefinal amount of severance pay depends on the size of the employer and remains flat, irrespective of the employee’s length of service. The Labor Code mandates severance pay equivalent to two months’ wages for employers with fewer than 11 employees, four months’ wages for employers with 11 to 49 employees, and six months’ wages for employers with more than 50 employees. Entitlement is not automatic. The amount is determined by court decision within the limits set by the law regarding the size of the benefit. This setup reduces the role of severance pay as an income replacement mechanism, because the process of litigation can significantly delay access to benefits.

2.44. Theflat benejit formula and the linking of the ceiling to the size of the enterprise reduces the role of severance pay as an income replacement mechanism during unemployment. A comparison of the cost of severance pay in Djibouti with the cost of an unemployment insurance scheme that has a 2 percent contribution rate reveals that severance pay in Djibouti is as expensive as unemployment insurance but offers much less in terms of coverage and benefit levels (Robalino et al. 2004). Workers in small and medium enterprises are at a particular disadvantage.

2.2.4. Reforming government pay and employment policies: options for reform

2.45. Reforming government pay and employmentpolicies would help spur growth and poverty reduction. First, as discussed in the previous section, reducing the wage bill will create the fiscal space needed to attain the education MDG by 2015. Second, reducing the unit labor cost in the public sector can also help lower labor costs in the economy (public sector wages being a benchmark for wages in the private sector) and reduce unit costs ofproviding public services. This will help improve firm productivity, attracting private investment and spurring growth.

2.46. This section draws on a wage and employment model that allows policy makers to examine various scenarios for reducing the public sector wage bill. Hypothetical scenarios highlight strategies that the government may consider in devising public sector reforms. Its appeal lies in its ability to be interactive and the relative ease ofexpanding it further. The results should be read with caution. First, the model is necessarily a simplification and the outcomes of any scenario indicate the general magnitude of fiscal consequences rather than a definitive assessment. Second, the model was built from scratch on the basis ofdata provided by the Ministry ofEmployment and the Ministry of Finance. The lack of detailed data on Djibouti’s civil service structure hinders a more accurate assessment.

2.47. There are digerent ways to reduce the public sector wage bill. The scenarios in this section propose alternative options, either through a price effect (reducing nominal wages), or a quantity effect (reducing overstaffing), or a combination. In addition, the scenarios discuss alternative ways to meet the education MDG by 2015 (depending on whether quality improvements are factored in the projections).

41 Main assumptions

2.48. The fiscal implications of civil service reforms depend on a number of factors. Each scenario presents the fiscal impact as well as the effect on the unit monthly labor cost for public sector workers. The following parameters were used to identify the possible scenarios and estimate their fiscal implications during the period 2004-20 15 18:

Macroeconomic assumptions: annual real GDP growth and inflation rate during the period of reference (2004-15) Magnitude of pay reform (cuts in wages): measured as the amount ofthe reduction in nominal wages and the subsequent reduction in the public-private wage differential Magnitude andpace of employment reform (cuts in staffing): options to reduce public sector employment through assumptions based on attrition rates, a hiring freeze, and military demobilization Financing needs to meet the education MDG by 2015: assumptions based on the number of additional teachers needed to meet the MDG of full primary school enrollment, and additional teachers that would bee needed for secondary school and vocational training. The scenarios cover alternative assumptions about whether the quality of education is improved during the

reference period (assumptions about teacher-to-student ' ratios and non-wage recurrent expenditures for education). Two targets are considered. The first assumes a student-primary teacher ratio of30: 1, and the second relaxes this requirement to 40: 1.

Summary of scenarios

2.49. Scenario 1 (de facto baseline) assumes no civil service reform. The baseline scenario assumes a modest real GDP growth rate of 3.5 percent, on average, during the period (2004-15). It does not factor in additional expenditure needs stemming from the attainment of the education MDG. Under this scenario, and in the absence of any price-quantity adjustment, total nominal wages are expected to grow by 56.5 percent between 2004 and 2015. As a share of GDP, the wage bill would fall only slightly, from 13.8 percent to 12 percent. More womsome is the impact of inaction on the unit labor costs in the public sector (proxied by the average public sector wage). In the baseline scenario, unit costs in the public sector would rise by 24.3 percent from US$551 to US$685.

2.50. Scenario 2 (MDG baseline) discusses the fiscal implications of attaining the education MDG with no adjustment in thepublic sector wage bill. Assuming a ratio ofstudents to teacher of 30:1, and the same average real GDP growth rate as the baseline scenario (about 3.5 percent on average), by 2015 the wage bill would remain very high, at about 13.4 percent of GDP. Nominal wages would expand by 78 percent. As a result, the fiscal position would deteriorate and unit labor costs in the public sector would increase to 29 percent.

2.51. Scenarios 3 to 5b discuss alternative ways to attain the education MDG with an adjustment to the public sector wage bill. The remaining scenarios assume a higher annual real GDP growth (on average 4.2 percent), which is consistent with the high-case medium-term macroeconomic scenario. As Scenario 2 illustrated, attaining the education MDG while maintaining fiscal sustainability calls for an adjustment to the public sector wage bill. The adjustment options consider a mix of quantity effects (such as a hiring freeze or demobilization program) and price

'*The base year for the calculations is 2004. The baseline figures were all provided by the Ministry of Finance. 2005 is a projection, and not based on actual data as these data were not available at the time of drafting this report.

42 measures (wage cuts). Scenarios 3a to 4b assume that the education MDG would be attained with a student-to-teacher ratio of30: 1.

0 Scenario 3 combines a freeze on hiring non-teachers with allowance for medical staff while maintaining constant nominal wages. The wage bills falls by 4.3 percentage points by 2015 though the unit labor cost rises by 7.7 percent. 0 Scenario 4a proposes a nominal wage cut of 10 percent across the board, to be implemented by end of 2006. The extent of the cut cannot be accurately determined owing to the lack of detailed information on Djibouti's civil service structure. The proposed wage cut is put forward in the model simulations to illustrate the fiscal implications of a nominal wage reduction of such magnitude. Under this scenario, the wage bill falls by 4.5 percentage points by 2015, and the unit labor cost (average public sector wage) declines. 0 Scenario 4b considers a combination of a nominal wage cut and a freeze on additional recruitment in the higher-earning categories ofcivil servants. 0 Scenario 5 considers an alternative wage bill adjustment through staffing cuts based on demobilizing the military and security forces. Demobilization as the only fiscal adjustment measure lowers the wage bill as a share of GDP. But it puts upward pressure on the unit labor cost (average public sector wage), which would rise by 35 percent over the period. 0 Scenario 5a combines military demobilization with a nominal wage bill cut. This measure reduces the wage bill by 6 percentage points as a share of GDP by 2015, with a concomitant reduction in unit labor costs in the public sector. 0 Scenario 5b assumes selective hiring freezes on new recruitments in addition to the 10 percent cut in nominal wages. This enables a greater decline in unit labor costs than the reduction observed in Scenario 4a.

2.52. Scenarios 6, 6a, and 6b consider similar adjustments to the wage bill, but they assume lowerfinancing needs to meet the education MDG by 2015. Scenarios 6 through 6b assume that a lower number of primary teachers is needed to attain the education MDG by 2015, based on a larger student-to-teacher ratio of 40: 1. Scenario 6a assumes a cut in nominal wages of 10 percent, leading to a reduction in the wage bill of 5.5 percentage points by 2015 and a lowering of the average public sector wage by 8.2 percent. Scenario 6b combines the cut in nominal wages with a selective employment freeze ofhigher-wage earners in the public sector. As a result, the unit labor cost falls by 11.5 percent over the period, while the wage bill share declines by 7 percentage points.

cenarios IDescription

1 De facto baseline -2.01 56.51 24.31

2 MDG baseline -0.4 77.8 29.0

3 Freeze on hiring and wages -4.2 34.3 7.7

4 Wage Cut (32%) -3.6 41.0 2.3 4a Wage Cut (10%) 4.5 31.1 4.9 5b Wage Cut (10%) & Employment Freeze -5.1 21.7 -7.8 5 Demobilization (DM) -3.1 51 .O 34.9 5a DM &Wage Cut (10%) -5.9 11.5 -0.4 5b DM 8 Wage Cut (10%) & Employment Freeze -6.5 1.6 -3.7

6 Alternative MDG baseline -1.9 57.3 24.3 6a Wage Cut (10%) -5.5 16.0 -8.2 6b Wage Cut (10%) & Employment Freeze -6.2 6.6 -1 1.5 dote: The change in wage/GDP share is measured in percentage points. Source: Authors' calculations using data provided from Ministrybf Employment and Ministry of Finance

43 Baseline scenario assuming no policy reforms (status quo)

2.53. The base year is 2004. The interval of interest for the simulations is a 10-year horizon between 2005 and 2015. The following assumptions underlie the base case:

Annual real GDP growth and inflation rate averaging 3.5 percent and 2.1 percent, respectively, between 2005 and 20 15 (consistent with the baseline medium-term macroeconomic scenario discussed earlier). Based on Ministry of Health projections, the number of medical professionals (general practitioners, specialists and pharmacists) increases by 6 members each year after 2004 while the number ofnurses and midwives increase by 15 members annually. This scenario assumes that the government ofDjibouti increases nominal wage by 2 percent per year over the intervallg. Number of civil servants and contracted personnel (contructuels) in addition to the annual wage bill as a &action of GDP, expected recruitment and attrition rates, and monthly wages are presented in Annex 2.1. In Table A 2.1.1, which shows the size of the public workforce, separate breakdowns are provided for civil servants with open-ended contracts and for contracted personnel (contructuels). The highest-paid scale is grade A and pay declines corresponding with each letter grade through D. Educators are categorized separately with primary teachers belonging to Category B 1 while secondary teachers and vocational professors are in Al. Assistant teachers (TA) and assistant professors are also provided separate categories. Military personnel. In 2002, the government committed to demobilize 3,983 military personnel over a 10-year period beginning in 2003. In 2004, only 411 personnel were demobilized. It is assumed that 3572 military personnel remain to be demobilized over the period ofanalysis. Growth of the workforce based on the recruitment and attrition rates. For example, primary educators are recruited at an average annual rate of 3 percent, secondary educators at 2.8 percent, and vocational educators at 2.5 percent. The rates from 2004 onward are the average of observed rates in 2002 and 2003.

Scenario 1: No adjustment to thepublic sector wage bill (status quo)

2.54. Under the baseline scenario, based on the projected rates of recruitment, attrition, and expansion of medical personnel, the public wage bill share is expected to decline steadily from 13.8percent in 2004 to 11.8percent in 2015. The path ofnominal wages by categories is provided in Figure 2.11. It is obvious that education and defense consumes a large portion of the wage bill. The nominal wage bill is higher by 56.5percent in 2015 relative to2005, and the average monthly public sector wage or unit labor cost (ULC) rises by 24.3 percent over the interval. Leaving aside reforms needed to attain the education MDG, government commitment to reducing the wage bill to 10 percent of GDP by 20 15 requires lowering the nominal wage bill from its current level by DF 3.9 billion, or 1.8 percent ofGDP :

l9 Even if we assume that the government continues to implement the current freeze in nominal wages of civil servants, this will only affect the level ofthe baseline projections, but it does not affect the magnitude of the changes driven by the policy reforms. The messages emerging from each ofthe policy reforms would remain broadly the same.

44 Table 2.5 Djibouti: Impact on Public Wage Bill (2005-15), Scenario 1 Categories 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 No. of employees 13,804 14,052 14,309 14,576 14,853 15,140 15,438 15,748 16,069 16,403 16,880 17,384 wage bill (DFmil) 16,218 16,493 17,171 17,882 18,628 19,409 20,228 21,088 21,990 22,937 24,121 25,382 shareofGDP(%) 13.8% 13.2% 13.0% 12.8% 12.6% 12.5% 12.3% 12.2% 12.0% 11.9% 11.8% 11.8% Mth. unltcost (DF) 97,907 97,809 100,003 102,237 104,512 106,829 109,189 111,591 114,036 116,524 119,080 121,675 UnltcostinUSS 551 550 563 575 588 601 614 628 642 656 670 685

FQure 2 10 Qlboub NuntEr Of RJbk Sector 6Tpby-S by Grade Foure 2 11 Qbub RJblK Wage Btil by Errpbyee Grade

~ 20.000 Mlltary8 NSF 30.000 1 8 Dembillzatwn Cost I D Mltary 8 NSF 18,000 Contractuals 25,000. mctors 8 Nurses 16.000 OD = Teachers 14.000 C2 20000- Contractuak 5 OD 12,000 c1 t 15.000. * ma 10.000 .w 82 C1 ~10000- 8.000 0 B1 82 a 0 81 6.000 5000- UP2 0 A2 4,000 AI n. AI 2.000 .A 0 U-)(ObmmOr

Source: Authors ’ calculation using dotafrom Djibouti authorities Source: Authors ’ calculation usina data from Djibouti authorities

Rgure 2 13 Qibauti Wage 81 Share to GF (percent) FQure 2 12 QibouU Average RJblK Sector Mnthiy Labor Cost, DF

16 00% 130.000 8 Dermbiliuatwn Cost 14 00% 0 Mltary 8 NSF 120.000 12 00% 0 mctors 8 Nurses Teachers 110,000 10 00% Consultants OD 800% 100,000 DQ 600% C1 90,000 82 4 00% 0 81 80,000 2 00% 0 A2 8 AI 70.000 0 00% BA

Source: Authors ’ calculation using datafrom Djibouti authorities Source: Authors ‘ calculation using data from Djibouti authorities

Scenario 2: No adjustment to the public sector wage bill, coupled with increasedpublic spending to attain the education MDG

2.55. This scenario is identical to the base case scenario except that additional teachers are recruited at a rate necessary to attain the education MDG. The target is dnven by fulfilling demand for primary teachers, which by 2015 would need to be at 3,667 assuming the MDG-recommended student-teacher ratio of 30: 1. Anticipating higher demand for secondary and vocational education requires corresponding increases in their recruitment rates. The assumptions for this alternative scenario are identical to the base case except that in targeting the education MDGs, primary teachers are recruited at a rate of 5.79 percent per year, secondary educators at 7 percent, and vocational educators at 8 percent. By 2015, the projected wage bill is 13.4 percent of GDP. The nominal wage bill of DF 28.8 billion is higher by 77.8percent relative to the baseline wage bill in 2004. The average unit costperpublic sector employee rises by 29percent over this interval.

Table 2.6 Djibouti: Impact on Public Wage Bill (2005-15), Scenario 2 Categories 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 No.ofernployees 13,804 14,152 14,517 14,902 15,308 15,735 16,185 16,659 17,158 17,685 18,334 19,025 wageblll(DFrnil) 16,218 16,663 17,535 18,464 19,454 20,512 21,641 22,848 24,138 25,518 27,115 28,834 shareofGDP(%) 13.8% 13.4% 13.3% 13.3% 13.2% 13.2% 13.2% 13.2% 13.2% 13.2% 13.3% 13.4% Mth. unltcost(DF) 97,907 98,123 100,653 103,247 105,907 108,633 111,429 114,294 117,231 120,240 123,241 126,303 UnltcostlnUSS 551 552 566 581 596 611 627 643 660 677 693 711

45 Fgure 2.14 Djibouti: Mrrber of hblic Sector En'pbyees by made Fgure 2.15 Q~boub:hbk Wage Blll by -bye Grade

1 ~ MIitaryB I -.,...znm, 8 NSF DembllQabn Cost 16,000 8 Conbactuals Mltary & NSF 0 Dactors 8 Nurses 16.000 OD =. E 25.000 Teachers 14.000 8 C2 - m Contractuak 20,000 OD 12,000 6 c1 6 15.000 mCZ 10,000 8 82 c1 6.000 0 B1 3 10,000 82 6.000 0 81 oP2 5.000 4,000 0 A2 8 A1 0 AI 2,000 .A 0

Source: Authors 'calculation using datafrom Djibouti authorities Source: Authors ' calculation using datafrom Djibouti authorities

Fgure 2 16 QibouU Average hbkSector Mnthly Labor Cost, DF Fgure 2 17 QibouS Wage 81 Share to GCP (percent)

Dembikabn Cost ~ 16.00% 1 0 Mttary 8 NSF 14.00% mctors 8 Mrses

Source: Authors 'calculation using datafrom Djibouti authorities Source: Authors ' calculation using datafrom Djibouti authorities

Adjustment scenarios: Lower public sector wage bill, attainment of the education MDG, and higher real GDP growth

2.56. In devising policies to reduce the wage bill, we need to distinguish two components: the quantity effect (the excess number of public jobs), and the price effect (the excess level of real wages paid to public employees). This section develops alternative scenarios that play with the key variables ofthe model, enabling assessment of the fiscal gains and possible tradeoffs of alternative policy options.

2.57. All subsequent reform scenarios assume a higher real GDP growth rate, averaging 4.2 percent between 2005 and 2015. This change is motivated in part by the anticipated favorable response of donors and foreign investors to reform initiatives by the government. Additional current expenditure needs to meet the education MDG by 201 5 are also factored in.

2.58. The following assumptions are shared by the remaining scenarios unless stated otherwise.

0 Annual real GDP growth and inflation averagmg 4.2 percent and 2.1 percent, respectively, from 2005 to 2015. 0 Nominal wage increases of2 percent per year over the interval. 0 An additional 6 medical doctors and 15 nurses per year from 2005 until 2015, based on the recommendations by the Ministry ofHealth. 0 Primary teachers recruited at a rate of 5.79 percent per year, secondary teachers at 7 percent, and vocational educators at 8 percent between 2005 and 2015.

46 Scenario 3: Hiringfreeze and constant nominal wages

2.59. Scenario 3 assumes that in addition to keeping nominal wages constant from 2006 onwards, there is a freeze on hiring additional public workers other than replacing retirees - with the exception of teachers needed to meet the MDGs and medical civil servants (doctors and nurses). The resulting higher real GDP growth rate allows the wage bill share of GDP to fall to 9.5percent by 2015. However, unit labor costs per public sector employee rises by 7percent over the period.

Table 2.7 Impact on Public Wage Bill (2005-15), Scenario 3 Categories 2,004 2,005 2,006 2,007 2,008 2,009 2,010 2,011 2,012 2,013 2,014 2,015 No.ofemployees 13,804 14,152 14,455 14,774 15,110 15,465 15,838 18,232 16,648 17,086 17,413 17,760 Wageblll (DFmll) 16,218 16,644 17,051 17,481 17,935 18,414 18,920 19,454 20,018 20,615 21,178 21,774 ShareofGDP(Y4 138% 134% 129% 124% 120% 116% 112% 109% 105% 102% 98% 95% Mth. unit cost (DF) 97,907 98,009 98,304 98,605 98,912 99,226 99,546 99,873 100,207 100,548 101,350 102,168 UnltcostInUSS 551 551 553 555 557 558 560 562 564 566 570 575

Djibouti: Nurber of Erpbyees by Group Djbouti; Nuder of Errpbyees by Group 20.000

1- ^^^ Military& NSF lo,""" 1 ~Contractuals Milnary NSF 16,000 20 000 a 14,000 OD mctors a Nurses -B Teachers 12,000 C2 g 15000 rn Contractvak 10 000 H c1 e OD 8,000 82 j I0000 ma rn c1 6.000 0 B1 a rn 82 5 000 4,000 OM 0 El 2,000 AI UA2 0 0 DAI Dmh""o-&4gZ$ EA 2SSSSSZZoOooNNNNNNNNNNNN

Djibouti: Average PdbliK Sector Mnthty Labor Cost. DF Qiboub Wage Bill as Share of CD

1C6,oOO 16.00% D Cenubilizatbn Cast 100.m 14.00%] D Miiarv & NSF

:IL w-,, 95.m rn Teachers 1000% Consultants 9o.m 8 00% 85.m

Bo.000 75.m

70.000 S88_0ZS?T?f?

'ource: Authors ' calculation using data from Djibouti authorities Source: Authors ' calculation using data from Djibouti authorities

Scenario 4: Moderate cut in nominal wages

2.60. Scenario 4 assumes a one-time nominal wage cut of 3.2 percent in 2006 across all categories ofpublic workers and a freeze on nominal wage increases as of2006. With a higher real GDP growth rate, averaging 4.2 percent over the 10-year interval, the wage bill as a share of GDP is reduced to lopercent by 2015. Unit labor costs in thepublic sector rise by 2.3percent over the period.

Table 2.8 Impact on Public Wage Bill (2005-15), Scenario 4 Categories 2,004 2,005 2,006 2,007 2,008 2,009 2,010 2,011 2,012 2,013 2,014 2,015 No. ofemployees 13,804 14,152 14,517 14,902 15,308 15,735 16,185 16,659 17,158 17,685 18,334 19,025 Wagebili (DFmil) 16,218 16,644 16,621 17,158 17,724 18,321 18,950 19,614 20,315 21,054 21,933 22,866 ShareofGDP(Y4 13.8% 13.4% 12.6% 12.2% 11.9% 11.5% 11.2% 11.0% 10.7% 10.4% 10.2% 10.0% Mth. unitcost(DF) 97,907 98,009 95,408 95,946 96,486 97,028 97,572 98,116 98,662 99,208 99,688 100,160 UnltcostInUSS 551 551 537 540 543 546 549 552 555 558 561 564

47 1 Fgure 2 18 Qiboub Nunber of Enpbyees by Group 1 I Fgure 2 19 Qlboub NuWr of -by- by Group I 20 000 Mlitarya NSF Cost 18,000 25'000 m Dembliuatan Contractuals 20,m ~irtary NSF 16.000 a OD - 0 mctors 8 Nurses 14,000 z Teachers 12,000 C2 - 15.000 m Contractuab 10,000 E c1 E OD 8000 rn 82 3 lo.m mi2 c1 6.W 0 61 a rn 82 5.030 4.000 ow 0 81 2W E A1 0 A2 0 0 AI DA

Source: Authors ' calculation using datafrom Djibouti authorities Source: Authors 'calculation using datafrom Djibouti authorities

Figure 2.20 Djibouti: Average RbkSector Mnthiy Labor Cost. DF Figure 2 21 D/lLvu11. Wage Bill as Share of GDP

~ Irn.W, Dembiluatan Cost 0 WrV 8 NSF

NNNNNNNNNNNN

Source: Authors ' calculation using datafrom Djibouti authorities Source: Authors ' calculation using data from Djibouti authorities

Scenario 4a: Bold wage reduction

2.61. A variation of Scenario 4 is a larger cut in nominal wages, which will help provide additional fiscal space to pay for quality improvements in the education sector. Scenario 3a assumes a nominal wage cut of 10 percent across the board to be implemented by the end of 2006, with no further wage increases until 2015. This allows the wage bill as a share to GDP to fall to 9.3 percent by 2015. The unit labor cost in the public sector is reduced by 4.9 percent over the period.

Table 2.9 Djibouti Impact on Public Wage Bill (2005-15), Scenario 4a Categories 2,004 2,005 2,006 2,007 2,008 2,009 2,010 2,011 2,012 2,013 2,014 2,015 No. of employees 13,804 14,152 14,517 14,902 15,308 15,735 16,185 16,659 17,156 17,685 18,334 19,025 Wagebili (DFmil) 16,218 16,644 15,453 15,953 16,479 17,034 17,619 18,236 18,888 19,575 20,392 21,260 ShareofGDP(%) 13.8% 13.4% 11.7% 11.4% 11.0% 10.7% 10.4% 10.2% 9.9% 9.7% 9.5% 9.3% Mth. unitcost (DF) 97,907 96,009 86,706 89.206 89,708 90,212 90,717 91,224 91,731 92,239 92,685 93,124 UnitcostInUSS 551 551 499 502 505 508 510 513 516 519 522 524 I I/ I Fgure 2,ZDjibouti: Nunher of Enpbyees by Goup /I Fgure 2.23 D/lbouti hWr of Enpbyees by Group Mlitaary&NSF 25'W Dembleatan Cost W Contractuals Mlitary 8 NSF 20.m mctors a Nurses -F Teachers $2 15.m Contractuab E DD 3 1O.W ma CI a 82 5.0730 0 81 0 A2 0 .AI

NNNNNNNNNNNN 1 I, Source: Authors ' calculation using data from Djibouti authorities Source: Authors ' calculation using data from Djibouti authorities

48 Fgure 2 24 Qlboub Average Nble Sector Wnthly Labx Cost DF Fgure 2 25 Qlboub wage BJI as Share of GDP

I , Source: Authors ' calculation using data from Djibouti authorities Source: Authors calculation using data from Djibouti authorities

Scenario 4b: Bold wage reduction and employment freeze

2.62. Another variation builds on Scenario 4a by implementing a one-time cut of 10 percent in wages and a freeze on additional recruitment in higher-paid civil servant categories (Al, A2, B1, B2) and on contractual categories (5, 6, and 7). In these categories, recruitment of personnel takes place only to replace anticipated attrition. The outcome of this simulation suggests the wage bill share of GDP falls to 8.6percent by 2015 with the unit labor cost falling by 7.8percent.

Table 2.10 Djibouti: Impact on Public Wage Bill (2005-15), Scenario 4b Categories 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 No.ofemployees 13,804 14,099 14,410 14,738 15,084 15,448 15,831 16,236 16,663 17,114 17,649 18,218 Wagebill(DFmll) 16,218 16,535 15,252 15,644 16,057 16,493 16,954 17,440 17,955 18,499 19,102 19,743 ShareofGDP(Y4 13.8% 13.3% 11.5% 11.1% 10.8% 10.4% 10.0% 9.7% 9.4% 9.1% 8.9% 8.6% Mth. unit wst (DF) 97,907 97,727 88,201 88,453 88,710 88,973 89,241 89,514 89,793 90,078 90,195 90,307 UnitcostinUSt 551 550 496 498 499 501 502 504 505 507 508 508

Scenario 5: Military demobilization

2.63. In this scenario, a reduction in civil service employment is proposed as an alternative way to reduce the wage bill. This reduction can be achieved by demobilizing the military and security forces. In 2002, the government committed to demobilize 3,983 personnel over a 10-year period. In 2004,411 security force members were demobilized, which leaves an average of 397 personnel per year to be demobilized between 2005 and 2013.

2.64. Under Scenario 5, the wage bill as a share of GDPfalls by 10.7percent by 2015 but the unit labor cost rises by 34.9 percent. The large increase in unit labor cost can be explained by noting that the average cost of demobilizing is estimated at DF 75,608 per person in 2004. Even while accounting for a nominal increase of 2 percent yearly, this remains lower than the average public sector wage over the period.

Table 2.11 Djibouti: Impact on Public Wage Bill (2005-E), Scenario 5 Categorler 2004 2005 2006 2007 2006 2009 2010 2011 2012 2013 2014 2015 No. ofemployees 13,804 13,755 13,723 13,711 13,720 13,750 13,803 13,880 13,982 14,112 14,761 15,452 Wage biil(DFm1l) 16.218 16,635 17,101 17,610 18,163 18,767 19,424 20,140 20,921 21,772 22,855 24,490 ShareofGDP(X) 13.8% 13.3% 12.9% 12.5% 12.2% 11.8% 11.5% 11.3% 11.0% 10.8% 10.6% 10.7% Mth. unit cost (DF) 97,907 100,782 103.846 107,026 110,323 113,738 117,270 120,920 124,687 128,567 129,027 132,080 Unltm5tInUSS 551 567 584 602 621 640 660 680 702 723 726 743

49 Fgure 2 26 Qibouh Nunter of srpbyees by Group Fgure 2.27 Djibouti: Wage Bill by hployee Group 18,oOO Military& NSF l",W ,,-1 303w0 ^^^ DenabillzatlonCost H Conbactuals Law" 1 14.W ~iltarya NSF OD 20.000 Doctors 8 Nurses 12.m C2 E Teachers lo.m 15,000 m Contractuals R C1 OD 8,000 H 82 10,000 BQ 6.000 C1 0 B1 5.000 4.000 B82 OK? 0 81 2.000 0 0 A2 B AI 0 NO~V).A g 8 8 8 E. $. 5 5 5 E. 5 urce: Authors calculation using datafrom Djibouti authorities ;ource: Authors calculation using datafrom Djibouti authorities

Fgure 2.28 Qibouti: Wage Bi as Share of GCP 140,000 - 16 0096 Denubillzatan Cost 130,000. 14 00% 0 MbN 8 NSF mctors &Nurses 12 00% 120.000 - E Teachers 1000% E Consultants 110,000 - OD 800% BO 100.000 .~- c1 6 00% 82 90,000 4 00% 0 81 0 A2 80,000 - 2 00% S AI .A 70,000 7 , , , , , , , , , , , , 0 00%

Source: Authors ' calculation using datafrom Djibouti authorities ;ource Authors calculatron using datafrom Djibouh authorrhes

Scenario 5a: Military demobilization and cut in nominal wages

2.65. A variation of Scenario 5 is to combine military demobilization with a nominal wage cut of 10 percent in 2006, assuming no further wage increases until 2015. Under this scenario, the wage bill falls to 7.9percent as a share of GDP in 2015 and the unit labor cost drops slightly by 0.4 percent over the period.

Table 2.12 Djibouti: Impact on Public Wage Bill (2005-15), Scenario 5a Calegories 2004 2005 2006 2007 2008 ZOOS 2010 2011 2012 2013 2014 2015 No.ofemployws 13,804 13,755 13,723 13.711 13,720 13.750 13.803 13.680 13.982 I4 112 14761 15452 Wawblll (DF mil) 16,216 16.635 15,090 15,233 15.404 15,604 15,833 16,095 16,392 16,724 17,212 18,081 ShareofGDP(W 136% 13 3% 11 4% 108% 103% 9 6% 94% 9 0% 86% 6 3% 80% 7 9% Mth. UnitcoStlDF) 97,907 100,782 91,629 92,583 93,564 94,566 95,594 96,636 97,692 98,756 97,167 97,516 Unn c0.t in US$ 551 567 516 521 526 532 538 544 550 556 547 549

Fgure 2.30 Qibouti: Nu-r of tirpbyees by Group Fgure 2.31 muti: Wage B1 by WyeeGroup

18,OW 20,000, .... Denubiltatbn Cost H Military& NSF 18.W ,,-I,",""" UMW&NSF Conbactuals = 16,oOO 14.m 0 mctors & Nurses 14.000 OD E Teachers 12.m 12,000 B C2 e Contractuab lo.m E lo.m OD B C1 4 8,wo 8.m ECZ E2 6.W c1 6.000 fi E 0 E1 4.m 82 4,000 om 2.W 0 81 2,000 0 R AI 0

Source: Authors ' calculation using data from Djibouti authorities Source: Authors ' calculation using datafrom Djibouti authorities

50 Fgure 2.32 Qitmti Average hblic Sector Mthty Ccst. DF Fgure 2.33 Djibouti: Wage Bill as Share of GDP 16.00% 105,000 1 rn CermbillzatonCost 14.00% 1 0 Mltary 8 NSF 0 Doctors 8 &sea 12 00% rn Teachers 10 00% Consultants

800% 85.000 - 6 00% 80,000 - 4 00% 0 A2 75,000, . 2 W% AI

000% I , , , , I , , , , ,

Source: Authors ' calculation using datafrom Djibouti authorities Source: Authors ' calculation using datafrom Djibouti authorities

Scenario 5b: Military demobilization, hiring freeze, and nominal wage cut

2.66. Another variation of Scenario 5 is to combine three adjustment measures: cut nominal wages by 10 percent, demobilize the military, and freeze additional recruitment in the higher-paid civil servant categories Al, A2, B1, and B2 and in the contractual categories 5, 6, and 7. In these categories, personnel are recruited only to replace anticipated attrition. Under this scenario, the wage bill as a share to GDP falls to 7.2 percent by 2015 and unit labor cost in the public sector falls by 3.7percent over the interval.

Table 2.13 Impact on Public Wage Bill (2005-15), Scenario 5b Categorles 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 NO. ofemployees 13,804 13,699 13,609 13,536 13,480 13,443 13,426 13,429 13,455 13,503 14,031 14,592 Wagebill(DFmil) 16.218 16,522 14,882 14,914 14,968 15,045 15,146 15,273 15,428 15,613 15.880 16,514 ShareofGDP(W 138% 133% 11 3% 10 6% 100% 9 5% 9 0% 8 5% 8 1% 7 7% 74% 72% Mth.unltcost(DF) 97,907 100,506 91,125 91.816 92.528 93,260 94,011 94,777 95.558 96,350 94,313 94,311 Unit cost in US 551 566 513 517 521 525 529 533 538 542 531 531

Fgure 2.34 Djibouti: Nunter Of Enpbyees by Group Fgure 2.35 Qltouti: W-e 81 by bpbyee Group 18,000 I MlitarvB rn CermbiiEaton Cost NSF 16.000 14.000 8 Conbactuals = 14.000 Doctors & Nurses 12.000 OD 12000 Teachers 10,000 8 C2 10,000 ContracPlak 6,000 MC1 k 6.000 6.000 6.000 8 82 6 4,000 0 61 rn 82 4.000 2.000 UP2 2.000 0 8 AI 0

wee' Authors ' calcularion usina data from Djiboutl aurhoriries Source: Authors ' calculation usina data from Djibouti authorities

FQm2 36 9 mutl Average RAOLSector Nbnthb Cost, DF FQure 2.37 Djibouti: Wzge BB as Share of GF

1M000 - 16 W% rn CembiilzatbnCost 14 00% 13 Mbry 8 NSF 12 W% 0 Doctors 8 Nurses rn Teachers

Source: Authors ' calculation using datafrom Djibouti authorities Source: Authors ' calculation using data from Djibouti authorities

51 Alternative approach to meeting the education MDG

2.67. The previous scenarios assumed that attaining the education MDG would be based on expanding the education supply (hiring additional teachers), assuming a student-to-teacher ratio of 30: 1. The following sets of scenarios relax the assumption for the student-primary teacher ratio and allow an alternative ratio of 40:l. This revised assumption implies that 2,750 primary educators would be recruited over the next 10 years to attain the education MDG by 2015. This implies that the number of additional teachers to be recruited would be 27 percent lower than the initial target of 3,667 primary educators (the number assumed in the previous scenarios). Under the revised student-to-teacher ratio, primary educators would need to be recruited at an average rate of 3.06 percent between 2005 and 2015. To complement the growth in the corresponding demand for higher education, secondary and vocational educators should be recruited at a rate of 3 percent per year.

Scenario 6: No adjustment in the public sector wage bill

2.68. Scenario 6 assumes no adjustment in the public sector wage bill, and a moderate real GDP growth of 3.2 on average over the period. It also assumes additional public expenditure stemming from attaining the education MDG (under the revised student-to-teacher ratio of 40:l). Under Scenario 6, the public sector wage bill as a share of GDP remains high, at 12 percent by 2015. The average unit labor cost in the public sector rises by 24Spercent over the period.

Table 2.14 Impact on Public Wage Bill (2005-15), Scenario 6 Categories 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 No.ofemployeer 13,804 14,061 14,328 14,605 14,893 15,192 15,502 15,825 16,160 16,508 16,964 17,445 Wage bill(DF mil) 16,218 16,509 17,205 17,935 18,700 19,504 20,348 21,235 22,166 23,145 24,295 25,516 SharaofGDP(w 13.8% 13.2% 13.0% 12.9% 12.7% 12.6% 12.4% 12.3% 12.1% 12.0% 11.9% 11.9% Mth.unltcost(DF) 97,907 97,838 100,062 102,328 104,637 106,989 109,384 111,824 114,308 116,838 119,345 121,890 UnItcortinUSI 551 551 563 576 589 602 615 629 643 657 672 686

Fgure 2.38: Djibouti: &mer of 6Tpbyees by Group Fgure 2.39 Djibouti: Wage BHi by 6Tpbyee Goup 20,000 30.000 Military& NSF Denobikatun Cost 18.000 16.000 Contractuals 25.000 QMMary8NSF &tors 8 Nurses 14,000 OD E 20.000 leachers 12.000 c2 e Contractuab E 15,WO 10,000 .C1 3 OD ma 8.000 82 8 1o.ooo 6 GI 6.000 0 81 5,000 m 82 4.000 OM 0 B1 0 A2 2.000 m AI 0 0 AI .A .A

ure 2.40 Qjbouti: Average hbkSector Wnthly Labor Cost , DF Fgure 2.41 Djibouti: Wage Bia0 Share of GDP 130,WO 16 00% Dembkatan Cost 14 00% mhMary8NSF 120.000 bctors a Nurses Teachers 110,000 Consultants OD 100.000 800% ma ma 600% mB2 90.000 4 00% 0 B1 0 A2 80.000 2 00% AI .A 70.000 , OW% GZaaa::

Source: Authors ' calculation using datafrom Djibouti authorities Source: Authors ' calculation using datafrom Djibouti authorities

52 Scenario 6a: Cut in nominal wages

2.69. A variation of Scenario 6 would be to consider a 10 percent reduction in nominal wages, with no further increases until 2015. Real GDP now is assumed to grow at an average of4.2 percent between 2005 and 2015. Under Scenario 6a, the wage bill as a share of GDP falls to 8.2percent while the unit labor cost falls by 8.2percent over the next 10 years.

Table 2.15 Impact on Public Wage Bill (2005-15), Scenario 6a Categories 2,004 2,005 2,006 2,007 2,008 2,009 2,010 2,011 2,012 2,013 2,014 2,015 No. ofemployees 13,804 14,061 14,328 14,605 14,893 15,192 15,502 15,825 16,160 16.508 16,964 17,445 Wageblll(DFmil) 16,218 16,490 15,163 15,496 15,841 16,197 16,567 16,949 17,345 17,756 18,272 18,814 ShareofGDP(Y4 13.8% 13.2% 11.5% 11.0% 10.6% 10.2% 9.8% 9.5% 9.1% 8.8% 8.5% 8.2% Mth. unltcost (DF) 97,907 97,725 88,187 88,414 88,635 88.849 89,055 89,255 89,448 89,633 89,760 89,874 UnltcostInUSt 551 550 496 497 499 500 501 502 503 504 505 506

Fgure 2.42 Qibouti Nurrber of Errpbyees by Group Fgure 2.43 Qlbouti: Wage BI by -bye Group 20 m 20,000 Cemblluaton Cost rn MiitaryB NSF 18.000 demll",W 0 Miltary & NSF rn Contractuals 16.000 16.000 &tors a Nurses 14,000 OD 14,000 Teachers 12,000 m c2 e 12.000 Contractuak 10,000 OD 10,000 .C1 E s 8.m .a 8.m rn 82 6000 c1 6,000 0 81 4.000 82 2.000 0 81 4.m OA2 0 0 A2 2.000 m AI 0 _. gggBwgg6;;;; IA

rce: Authors ' calculation using datafrom Djibouti authorities Source: Authors 'calculation using datafrom Djibouti authorities

Fgure 2.44 tjibouti: Average hbik Sector Mnthly Labor Cost. CF Fgure 2 45 Qiboue Wage Bill as Share of GW 1w.000 - Cembiluabm Cost DMltary&NSF 1- ^^^ mctors a Nurses g5'000190,000 -L----- 85.000.

80.000 -

i$gggwgg&;;g; Source: Authors ' calculation using data from Djibouti authorities Source: Authors ' calculation using datafrom Djibouti authorities

Scenario 6b: Cut in nominal wages and hiringfreeze

2.70. Scenario 6b combines a one-time cut of 10 percent in nominal wages with a freeze on additional recruitment in higher-paid civil servant categories (Al, A2, B1, and B2) and in contractual categories (5, 6, and 7). In these categories, personnel are recruited only to replace anticipated attrition. The outcome: a lower wage bill share of GDP, at 7.6 percent by 2015, with average public wage falling by 11.5 percent over the interval.

53 Table 2.16 Impact on Public Wage Bill (2005-15), Scenario 6b Categories 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 No.ofemployees 13,804 14,009 14,221 14,441 14,669 14,905 15,149 15,402 15,664 15,936 16,279 16,638 Wagebili(DFrni1) 16,218 16,381 14,962 15,187 15,418 15,656 15,901 16,153 16,413 16,679 16,982 17,297 ShareofGDP(Y4 13.8% 13.1% 11.3% 10.8% 10.3% 9.9% 9.4% 9.0% 8.6% 8.2% 7.9% 7.6% Mth. unit cost (DF) 97,907 97,441 87,671 87,637 87,592 87,538 87,473 87,399 87,314 87,219 86,936 86,632 UnitcostinUSf 551 548 493 493 493 493 492 492 491 491 489 487

Flgure 2.46 Djibouti: Nuerof by Group Fgure 2.47 Djibouti: Wage Bill by Errpbyee Ooup

mmor .A

II 1 Source: Authors ’ calculation using data from Djibouti authorities Source Authors ’ calculation using data from Djibouti authorities

FQure 2 49 Q~bouBWage 81 as Share of Gop

100,m - 16.00% CenobilizatanCost 14.00% I Mltary 8 NSF h 0 Doctors 8 Nurses 12 00% iTeachers 9o.m 10 W% rn Consultants almzsm- OD 85.m : 80,m -

75.m -

gw!BBB,,,,,,ggzzzg Source: Authors ’ calculation using data from Djibouti authorities Source: Authors ’ calculption using datafrom Djibouti authorities

The distributional impact of civil service reforms

2.71. In principle the social impact of the proposed adjustment measures to reduce the public sector wage bill could be positive, because these reforms would make it possible to attain the education MDG by 2015. However, it is important to note that wage bill savings achieved by rationalizing the public sector or cutting wages may have negative aggregate demand effects, and they need to be accompanied by adequate compensation.

2.72. This section aims to assess the potential impact of a reduction in nominal public sector wages on household welfare. The 1996 representative household survey is used to estimate the implications of a nominal public sector wage cut. The results can be applied to the current circumstances by assuming that income distribution has not altered significantly since 1996. Figure 2.50 provides a graphic description of income distribution in 1996 based on the sampled households. The national or absolute poverty line at DF 216,450 (US$1.8 a day) lies slightly below the lower boundary for the third income quartile (represented by upper-middle-income households). This explains why nearly 50 percent of Djibouti households are below the national poverty line, including almost all lower-middle-income families.

54 Fgure 2.50 Djibouti: Income Dstributiin in 1996

0.6

0.5

0.4 P 0.3 Um Ndiordpomiylire 0.2

0.1

0 9 10 11 12 13 14 15 log per-capita expendture, 1996

Source: Authors' calculations based on EDAM 1996

2.73. A cut in nominal public wages will affect the real incomes of middle- and upper-income households (because richer households have a larger share of wage income earned from the public sector) and may potentially flatten the income distribution. Figure 2.51 plots a nonparametric regression of the wage earnings share from the public sector to total wage earnings from all sources against per capita expenditure, which serves as a proxy for per capita income. For our exercise, for all sample households with individuals who are employed in the public sector monthly wages are reduced by 10 percent.*' The outcomes are shown in Annex 2.2 (Tables A 2.2.1 and A 2.2.2).

Figure 2.51 Djibouti: Share of kblic Wages to Total Wage hcom against Wr-Capita Expenditure, 1996

1 .oo 0.95 0.90 0.85 Nationrl Poverty Line 0.80 0.75 0.70 0.65 0.60 0.55 0.50 9 10 11 12 13 14 15 log per-capta expenditure, 1996

Source: Authors' calculation based on EDAM 1996

2.74. This preliminary analysis reveals the following results for the potential distributional impact ofa nominal wage cut of 10 percent:

Because income is usually underreported, wages are adjusted upward together with other income sources so that total income equals expenditure for households where total income was initially below total expenditure.

55 . The impact on the poverty headcount appears to be marginal. A 10 percent reduction causes the overall national poverty line headcount ratio to rise by one percentagepoint.2' Assuming a similar income distribution also prevails in recent years and that the average household size is 4.2 as in 1996, this translates to 1,190 households falling below the national poverty line, assuming a population of 500,000.22 ..Income inequality falls slightly: the Gini coefficient declines from 40.8 to 40.3. . Of the poorest households, 1.5 percent fall below the extreme poverty line which is equivalent to DF 100,229 and is the minimum level of per capita expenditure needed to maintain subsistence. In the third income quartile, 3.2 percent of households fall below the national poverty line, affecting almost equally both male- and female-headed households. Assuming a population of 500,000, this implies that 952 upper-middle-income households dip below the national poverty threshold. . As far as the poverty gap goes, the households affected experience a fall in per capita income thatputs them on average 2.5percent below the national poverty line.

2.75. These results may understate the negative impact of a cut in public wages on informal safety nets provided to households. Rural households rely on remittances from extended family members in urban areas. They could suffer a decline in transfers from families whose members are employed in the public sector, thereby compromising their real incomes. In addition, it is important to note that the analysis is based on 1996 data. Civil servants experienced a 10 percent cut in nominal wages in 1998. This could place a larger fraction of them below the poverty threshold than described by the 1996 data. An additional 10 percent cut in 2006 may push some middle-income households with public sector workers below the poverty line. It could also aggravate the condition of those families with public sector workers who are already below the line. A more detailed social impact analysis would be needed to estimate the amount of fiscal transfers that would compensate households that fall below the poverty line as a result of the reduction in nominal public sector wages.

Thepolitical economy of reducing the public sector wage bill: identijjing incentives for reform

2.76. Implementing a reduction of the public sector wage bill is politically difficult, and it can also be administratively complex. This section summarizes the potential impact of the proposed policy reform scenarios (tables 2.17 and 2.1 8). This crude analysis allows Djibouti policy makers to examine the incentives for reform, based on the perceived value of benefits and costs. This section compares the scenarios discussed earlier according to the following criteria:

the nature of benefits (economic, fiscal, social), and to whom they accrue the costs (economic, fiscal, social) and who bears them the degree ofpolitical feasibility (who wins and who loses). The assessment of social costs helps determine the degree ofpolitical feasibility. the degree and administrative complexity of the reforms. The administrative complexity depends on the variety of changes introduced, the degree of technical sophistication in the changes proposed, and the institutional capacity needed to implement them. It also depends on

A larger reduction is naturally expected to create a more substantial increase in poverty incidence. A 20 percent nominal cut led to an overall increase in national poverty by 3 percentage points, with poverty among households whose wages are earned in the public sector rising by 10.5 percentage points. 22 Ifa population of 700,000 is assumed, the figure becomes 1,667.

56 the duration of the policy change process. Some reforms can happen from one day to the next (such as a reduction in nominal wages), whereas other reforms (such as demobilizing military personnel) take place over a number of years.

adjustment cut in nominal wages to the public nominal and bolder cut in sector wage wages staffing bill (Scenario 4a) (Scenario 5b)

(Scenarios

Strong

Positive and strong

a share ofGDP) ~ Political feasibility I high I moderate I low I moderate I low low I Administrative I moderate I moderate I moderate high I 5Imm No Moderate Bolder Cut Moderate cut Bolder cut in adjustment cut in in nominal in staffing: nominal wages nominal wages to the public nominal wages military and moderate cut and bolder cut in sector wage wages (Scenarios 4b demobilization in staffing staffing bill (Scenario 4a) and 6a) (Scenario 3,5) (Scenarios 5 and 6b) (Scenario 4b) (Scenarios 1,2,6) Civil Children Private Children Private investors Private investors servants without investors without access Unemployed Unemployed access to Unemployed to education Children without Children without education Children access to education access to education without access to education

Private Civil servants Civil servants Military Military personnel Civil servants and investors and their and their personnel unwilling to be their dependents Unemployed dependents dependents unwilling to be demobilized Educated workers Children demobilized ‘queuing’ for jobs without in the public sector access to education

57 2.3. CONCLUSION AND POLICY IMPLICATIONS

2.77. Each of the options to reduce the wage bill presents potential fiscal gains and potential tradeoffs that need to be assessed carefully. Furthermore, attaining the objectives of fiscal savings and lower unit costs of service provision to increase human capital accumulation and attract private investment does not automatically result from simply reducing the wage bill and retrenching public workers.

2.78. Three factors may affect the outcome ofreforms. First, reducing the public sector wage bill can have a negative social impact and the accompanying compensatory measures could come at a fiscal cost. Second, the political economy of these policy reforms can make the reforms politically unfeasible, calling for a careful analysis of the magnitude and sequencing of the desired reforms. Third, measures that reduce public employment or wages do not guarantee either that workers will be absorbed by the private sector or that the measures will have a direct impact on growth. Other government failures can still lead to imperfect labor markets and weak private investment.

2.79. Complementary policy actions need to be taken to address the distorted incentive structure and resulting labor market segmentation in Djibouti. Revamping Djibouti’s education and training system would expand the pool of trainable workers. Strengthening the investment climate to promote the creation and expansion of small businesses would directly address employment creation objectives. Allowing private competition in infiastructure service delivery can also lead to more efficient allocation oflabor between the public and private sectors. These issues are explored in the following chapters.

58 CHAPTER 3. MAKING THE LABOR MARKET AND THE INVESTMENT CLIMATE WORK FOR GROWTH AND JOBS

As discussed in chapter 2, Djibouti’s recent growth performance has been modestly positive. Real GDP grew 3-4 percent over the past few years, with real per capita GDP growth significantly lower owing to population growth (2.8 percent per year). This population growth also explains the higher unemployment rates since 1996. Djibouti’s economy faces a dual challenge: to accelerate growth and to promote human capital to create more and better jobs.

3.1. Djibouti’s economy is characterized by high unemployment and limited job flows. Examining the employment situation in Djibouti yields a complex picture ofthe demand and supply pressures at work and the resulting labor outcomes. The labor market is characterized by a large public employment share, a high degree of labor market segmentation along public-private lines, and a rising informal sector. The pace of job creation in Djibouti has been slow for reasons that include high unit labor costs, rigidities in employment protection legislation that limit labor turnover, and the paucity ofskilled labor.

3.2. The labor market challenges facing Djibouti’s policy makers are daunting, and there is an obvious need to reconsider traditional approaches to the country’s policy reform agenda. These challenges imply stimulating employment growth not by absorbing labor in the public sector, but by promoting a vibrant private sector; not by mandating job security, but by promoting labor market flexibility; and not by relying on informal arrangements to insure against income risks, but by promoting social insurance and carehlly designed social safety net programs.

3.3. The other pillar of the growth and labor market strategy proposed in this report is upgrading the skills and productivity of the workforce. Strong investments in human capital will also be needed to improve the quality oflabor supply and to foster labor demand.

3.4. This chapter takes a broad look at the labor market challenges observed in Djibouti and outlines policy options to make the labor market work for growth and jobs. It first describes the main features of Djibouti’s labor market and reviews recent labor market developments. It then evaluates legal and institutional issues regarding the investment climate and the flexibility of the labor market. The last part is devoted to the education and training strategy to strengthen Djibouti’s labor market by upgrading workers’ slulls and expanding the pool oftrainable workers.

3.1. MAINFEATURES OF DJIBOUTI’S LABOR MARKET

3.5. Djibouti’s labor market is characterized by a large public employment share, a high degree of segmentation along public-private lines, the paucity of slulled labor, and demographic pressures that manifest themselves through rising unemployment.

Dualistic and segmented labor market

3.6. Djibouti’s labor market is dualistic and segmented on account of remaining labor market rigidities and high labor costs (wage and nonwage costs) in the protected or formal sector. The formal sector comprises the public sector and a handful of large private enterprises mostly managed and owned by foreigners. Djibouti’s formal sector consists ofthree oil companies (Shell, Total, and Mobil), two banks, and two bottling corporations (Coca-Cola and Courbeche). Besides the formal sector, a growing informal sector is directed mainly toward customers with low purchasing power.

59 In addition to the segmentation between jobs in the formal sector and the informal sector, the informal sector is segmented between Djibouti employees and foreign employees. The latter usually accept lower wages and harder work conditions.

3.7. Workers employed in the public sector and in the formal private sector are better off; with those in the public sector doing better. Next come the urban self-employed, followed by the wage earners in the informal sector. Analysis of wage and earnings dispersion shows increasing returns to education (not decreasing ones, as is usually found in higher-income countries). The analysis camed out for this report shows higher returns for postsecondary education. As explained in the next section, this could reflect the scarcity ofskilled labor and a selectivity bias.

3.8. The informal sectorplays a key role in the economy, comprising about half of allprivate sector enterprises engaged in semi-organized, small-scale, unregulated activities. In principle the informal sector is the disadvantaged segment of the labor market in the sense that it is expected to expand by absorbing displaced formal sector workers in a downturn and contract during economic upturns when the queue for good jobs shortens again. However, the growth in employment in the informal sector has not compensated for the stagnation or negative employment growth in the formal sector. It may have led' to low-skilled production arrangements, thereby contributing to Djibouti's below-potential performance.

3.9. The formal sector offers better working conditions than the informal sector: it pays better, provides fringe benefits such as pension and health care, and offers virtually absolute job security. By contrast, workers in the informal sector (self-employed, wage earners, and unpaid family workers) are largely unprotected by labor legslation and outside the social insurance and social safety nets. See box 3.1 for a summary of the main features of the informal sector in Djibouti. 3.10. The high business start-up costs for small enterprises discourage them from joining the formal sector. It takes two to three months and can cost more than US$340 to register a limited liability company with a sole proprietor.

Table 3.1 Diibouti: Costs of Joining the Formal Private Sector 2001. (an nual US%) Average costs if Average costs if the Net increase in costs when a Average company the company Sectors company joins the company joins the formal sales revenues operates in the sector (percent) formal sector informal sector Manufacturing 77,927 44,860 52,234 16.44 Services 95,946 62,804 73,128 16.44 Commerce 1,994,936 1,148,419 1,337,194 16.44 Transport 12,944 36,135 36,809 1.a7 Construction 259.281 210,675 210,676 0.00 Total 2,441,034 1,502,804 1,710,727 13.83 Note: Costs ofjoining the formal private sector are proxiedby the payment of the corporate taxes and levies Source: Mahamoud (2001)

60 Box 3.1. The Informal Sector in Djibouti: Main Features

A recent survey of 150 firms operating in the informal sector in Djibouti presents interesting findings. Following is a summary of the main features of these firms.

Size. Djibouti’s informal sector is characterized by ease of entry, a low-resource base; family ownership; a small- scale, labor-intensive, unregulated but competitive labor market; and informal processes of acquiring skills. Financial capital investments tend to be low. As a result, firms operating in the informal sector are generally small, with 1 to 6 employees. More than 90 percent of the 150 enterprises surveyed were owned and managed by one individual. Age. Firms surveyed had been operating for about 6.5 years, on average, with the most recently created being engaged in transport for about 4 years and the more established units operating in construction for about 11.5 years. The units operating in trade have been in business for 7.5 years, on average, and those providing services for 4.5 years. The relatively young age of the units surveyed does not exclude the existence of an informal sector before the 1990s. Instead, it reflects the weak viability of these units, as evidenced by a high rate of entrepreneurs (42 percent of those surveyed) changing activities each time they faced difficulties. Activities. Djibouti’s informal sector is less heterogeneous in terms of the types of activities than in other African countries, owing to the minimal role played by the manufacturing and agricultural sectors. Enterprises operating in Djibouti’s informal sector concentrate on trade, services, construction, and transport. By and large, trade and services dominate. About 85 percent of these enterprises are engaged in trading activities, providing 79 percent of informal employment; 6 percent of the units surveyed operate in transport; 4.5 percent are in services, providing 10 percent of informal employment; 3.5 percent are involved in producing goods; and only about 1 percent are in construction. Most customers are low income earners, civil servants, and employees in the modem private sector. Sales to government departments are negligible and barely exceed 0.2 percent. Access to credit. Most informal sector entrepreneurs rely on personal savings as their primary source of start-up capital, followed by credit from family and/or friends. Only about 6 percent of these units claim to receive bank credit. Type of employment. Self-employment is the norm. In addition to self-employment and family labor, wage labor (causal or regular) is also an important feature of Djibouti’s informal sector. Wage labor appears to be more common in transport and construction activities. The informal sector is the primary source of employment for unskilled or low-skilled workers, female workers with little education, and older workers. More than 70 percent of firms employ family members and occasionally employ workers. The remaining 30 percent of firms surveyed employ up to six workers. As to the origin of the labor force, 25 percent of units surveyed claim to employ Djiboutian national workers exclusively. The others hire foreign workers, mainly Ethiopians, Somalis, and Yemenis. Skills and human capital. The majority of informal sector workers are low skilled or unskilled. Formal education levels tend to be lower than in the formal sector. On-the-job training is the norm and usually considered sufficient to obtain the knowledge and skills required for undertaking various economic activities in the informal sector. Typically, investments in human capital take the form ofapprenticeships under informal arrangements. Earnings. Wages tend to be below those offered in the formal sector and below legislated minimum wage levels. Notwithstanding, the earnings profiles of informal sector workers differ considerably by occupational status. In general, the self-employed have the highest earnings, followed in descending order by regular wage earners, casual wage earners, and apprentices. This order holds true regardless ofthe branch ofinformal activity. Legafity. Informal sector enterprises operate outside the institutional and regulatory framework. However, this notion does not hold uniformly. Compliance with the country’s laws and regulations varies. The scattered evidence available shows that a significant share of informal sector enterprises in Djibouti really operate on the margins of legality, that is, they comply with some regulations but not with others. The extent of compliance with various applicable legal requirements appears to correlate positively with firm size and age. Poor compliance obviously relates to the weak enforcement by Djibouti authorities. The study also explained that the rapid expansion of Djibouti’s informal sector is due to a combination of factors: the inability of the formal sector to create enough jobs for a growing labor force; the high costs of operating in the formal private sector; and a tendency of civil servants (in about 11.5 percent ofthe units surveyed and most over 50 years old) to operate in the informal sector through the late 1990s to make up for the decline in pension benefits, the 1998 nominal wage cut and the accumulation ofwage payment arrears.

Source: Djibouti, MinistBre de Finances (2002)

61 High labor costs despite high unemployment

3.1 1. As mentioned earlier, Djibouti’s average wages in the public sector are 10 times higher than those in Ethiopia, and 5 times higher than in Yemen. Wages in theformalprivate sector are 1 to 7 times higher than in neighboring countries. Large private sector companies in Djibouti offer wages that amount to twice as much as those offered by companies operating in the informal sector. Salaries for workers (at comparable levels of slulls and job description) are also among the highest on the continent. Salaries for a local cadre in a senior managerial position can reach up to US$1,600 per month in the formal private sector. The burden of labor costs (wages, salaries, and benefits) is also much higher than in neighboring countries. In Djibouti it ranges between 30 and 50 percent, whereas in Ethiopia and Yemen non-labor costs represent about 20 percent of total labor costs.

3.12. The apparent insensitivity of wage rates in Djibouti to the high level of unemployment reflects remaining rigidities in labor market regulations and the high level ofpublic wages. The efficient functioning of the labor market is hampered by the Labor Code-which was introduced in 1952, some 25 years before Djibouti’s independence-and various restrictions aimed at protecting expatriates working in Djibouti. Despite amendments introduced in 1997 and 2004, the Labor Code still contains legal provisions that are generous compared with regional standards and contributes to higher labor costs in Djibouti and to labor market segmentation.

3.13. Another factor explaining the inconsistencies in wages and unemployment trends could be the shortage of qualified manpower. It is estimated that each year, more than 4,000 young people who have had no access to a general secondary education or to any kmd of professional training or qualification enter the labor market. The number is expected to grow over the next few years, in particular in the capital, mainly reflecting rising dropout rates. About 10 percent of primary school students drop out for the last two grades every year, and more than 30 percent drop out at the end of the primary cycle. Only about 27 percent of those completing primary education gain a place in secondary school. Technical and vocational education in Djibouti begins at the secondary-school level. Very few opportunities (and usually of a very specific kind) are available for some students to pursue professional training.

3.14. Labor market dynamics in Djibouti’s informal sector are depicted in box 3.2. Labor supply is likely to be an upward-sloping curve with respect to real wages. It is relatively elastic to real wages but becomes less so beyond a certain level, owing to the shortage of qualified labor and various impediments to labor market functioning. Labor demand can be represented by a downward-sloping curve that reflects higher labor demand by firms at lower real wages. Labor demand in Djibouti’s informal sector is relatively inelastic with respect to real wages (labor demand 1). This could be attributed to the features of small enterprises in the informal sector: low investments in capital stock, low productivity, diseconomies of scale related to the small size ofthe domestic market, and the restrained demand for goods and services by expatriates (military and others) and for port services by Ethiopia. More than 76 percent of wage earners in the informal sector earn less than the minimum wage (set at US$122).

62 Box 3.2 Djibouti: Labor Market Features - D At the minimum wage, (F/P) and given the level of labor demand, the implied level of employment is L ,which is -- SD a non-market-clearinglevel. The corresponding level of unemployment is ( L - L ). If real wages were to decline to (~/p)~by any measure (through either direct cuts in nominal wages or higher domestic prices, that is, an increase in the nominal value of the domestic currency relative to some- foreign currency), the labor market would clear at an D employment level of L,, which would be higher than L . In such a situation, there would be no more cyclical unemployment and the observed unemployment would consist only of structural and frictional unemployment. With more flexibility in the labor market, in particular regarding hiring and firing policies and wage setting, the labor demand curve could shift out (labor demand 2). In this case, employment would increase to L, , higher than L, , which clears the labor market at a real wage (W/P)~. This real wage would be higher than the real wage in a less flexible labor market, (~/p)~ (Dridi, 2003)

Labor supply (household) %le I \ I \\ I Minimum wage

\ Labor demand 2 (firm)

.. i:.. : : Unemployment level (at minimum real wage) ij= :j I i: Employment

3.15. Djibouti’s labor market outcomes, with labor supply far exceeding labor demand, reflect the interaction of demographic and economic factors. Like most other developing countries, Djibouti is experiencing rapid population growth, which has generated a surge of new entrants in the labor market. Although these demographic forces could promote economic growth, absorbing them in the labor market requires sustained firm growth. Poor labor demand has been influenced by low real output growth. Inadequate human and private investment, along with weaknesses in the investment climate, has so far prevented economic growth and job creation. As a result, employment growth has been outpaced by the growth ofthe labor force.

3.2. LABORMARKET OUTCOMES: STYLIZED FACTS AND RECENT TRENDS

3.16. Djibouti is characterized by low labor force participation and high long-term structural unemployment. The structural unemployment is caused by a highly distorted dual labor market, characterized by a large public sector that pays wages and benefits well above the levels paid by the private (both formal and informal) sector.

3.17. Statistical challenges make the study of key labor market outcomes in Djibouti difficult. No labor market survey has been carried out in the country, and the last population census was conducted in 1983. This chapter relies on data provided by household surveys carried out in 1996 and 2002. It also draws on the findings ofrecent firm-level surveys (small-scale enterprises in

63 the informal sector and a field survey of large private firms in the formal sector). The analysis of unemployment is particularly challenging. Unemployment figures should be read with caution, because they may also include underemployment. Partly owing to weak formal insurance mechanisms, very few individuals can afford to be jobless in Djibouti. As a result, many workers undertake any type of work in the informal sector. Often, this work leaves them underemployed, workmg in a low-productivity environment. However, Djiboutians think of employment as a paid, regular job in the formal sector. As a result, household survey data tend to overestimate unemployment and underestimate the size of the informal sector. Notwithstanding these caveats, this section explores available household data to provide an overall picture of labor market outcomes and trends.

Labor force participation

figure 3 1 Qibouti Labor Force Participation Rate (Oh) by Age Figure 3 2 Qibouti Labor Force Wrticipation Rate (Oh) by Age G-oups,lS96 Gfoups,2oo2 100% - 100% - 90% - 80% - 90%80%-70% - 70% - vile 60% - 60% - 50% - 50% - 40% - 40% - 30% - 30% - 20% - .----- 20% - 10% 10% - 0% .. 0% 7 1519 20-24 2529 30-34 3539 40-49 50-59 60-65 15-19 20-24 25-29 30-34 35-39 40-49 50-59 60-65

3.19. Age is an important determinant of the Participation ratio in Djibouti. In both 1996 and 2002, a similar pattern can be observed: participation rates increase with age and then decline. The participation rate peaks at the 4049 age group and then starts falling. This inverse U-shaped curve (Figures 3.1 and 3.2) is commonly seen across both developed and developing countries. In Djibouti, the observed pattern is dictated by the behavior of the male participation rate. While male participation peaks at 86.4 and 96.7 percent in 1996 and 2002 respectively, in the 35-39 age group, female participation peaks in both years at the 20-24 age group and decreases with age, more noticeably so in 1996.

3.20. The life-cycle pattern of participation has changed significantly between 1996 and 2002 for both male and females. Male labor force participation rose in similar proportions until age 50, thereafter rising proportionally more (Figure 3.3). A similar pattern emerges for female labor participants (Figure 3.4), In 1996, the behavior of female labor participants follow a traditional

23 The labor force is defined as the number of employed plus the number of unemployed (1 5 to 65 years old) who have been actively searching for a job. 24 Labor force participation rates for ages 12-65 in 1995 in selected African countries: Burkina Faso, 83.5 percent; C8te D’Ivoire, 66.4 percent; Ghana, 81.9 percent; and Somalia, 75 percent.

64 pattern of female labor market activity, rising up to 20-24 years of age and declining steeply thereafter, partly reflecting marriage and child-rearing norms. However, in 2002, the rate of decline was considerably lower.

Figure 3.3 Djibouti: Male Labor Force Participation Rate (%) by Figure 3.4 Djibouti: Femle Labor Force Participation Rate (%) by Age Groups, (1996-2002) Age Groups, (1996-2002) 100% 100% 90% 90% E 1996 2002 80% 80% 70% 70% 60% 60% 50% 50% 40% 40% 30% 30% 20% 20% 10% 10% 0% 0% 15-19 20-24 25-29 30-34 35-39 40-49 50-59 60-65 15-19 20-24 25-29 30-34 35-39 40-49 50-59 60-65

3.2 1. The overall increase in labor force participation in Djibouti between 1996 and 2002 was driven by large gains in participation among older age cohorts. It could be explained by the net accumulation of pension arrears to public sector and private sector employees, which reached 28 percent ofGDP in 2001. This may have pushed elder people affected by the pension arrears to find complementary jobs.

3.22. Education is a key factor determiningparticipation in Djibouti's labor market. Figure 3.8 shows a positive correlation between years of education and the participation rate. The gender gap in labor market participation is large for low levels of education but decreases considerably when education rises.

3.23. Participation rates in the capital city are higher than in rural and other urban areas. The rural labor force participation rate rose by 15 percentage points between 1996 and 2002, largely because of increases in male participation. Female participation rate in rural areas remains extremely low (16.8 percent in 1996 and 28.6 percent in 2002) relative to that in urban areas. One explanation is that rural households require a greater amount ofhousehold work because they have more children on average, are isolated, and lack access to basic services-all ofwhich increases the burden on women in the household, who are traditionally also in charge of fetching water for drinkmg.

Figure 3.5 Djibouti: Labor ParticipationRate by Figure3.6Djibouti: MaleLabw Participat ion Rate by Figure 3.7 Djbouti: Femk Labor ParticipationRate by Location, (1996-2002) Location,(lQ96-2002) Locabon, (1996-2002) HI996 M2002 1996 2002 1996 D2002 - 80 m 80

60 80 60

40 40 40

20 zo 20

0 0 "

Source: Authors calculation based on EDAM 1996.

65 Fgure 3.8 tjibouti: Educatbn and Labor Mrbt Pawpation. 1996 loo%, 90% -

*, -Al 30% - - - - -Female 10%

0 12 3 4 5 6 7 8 9101112131415 Years of Mucatbn

Source: Authors’ calculation based on EDAM 1996 and 2002

Unemployment

3.24. Unemployment rate rose by 12 percentage points between 1996 and 2002 and remains one of the most signifcant economicproblems in Djibouti.” The unemployment rate in 2002 was 56.1 percent, compared with 44.1 percent in 1996. The increase was primarily due to a decline in the employment rate among males. Over this six-year interval, male unemployment rates rose by 14.7 percentage points while the increase in female unemployment was half as large, at 7.8 percentage points.

3.25. The rural unemployment rate is generally higher than the urban rate. In 1996, high rural unemployment could be attributed to high female unemployment rates. However by 2002, rural unemployment rates for males caught up with those for females, reflecting deterioration in employment opportunities for males. In contrast, the rate of female unemployment in rural areas fell by four percentage points.

Figure 3.9 Djibouti: Lherrploymnt Fate (X) by Figure 3.10 Djibouti: Mik? Vlerrpioymnt Rdte (“A) by Rgure 3.11 Djibouti: Femle Vlerrpioymnt Rdte (Oh) Location. (1696-2002) Location. (1696-2002) by Location. (1996-2002)

-..? _In 7” ~ 0 I696 2002 I” - I” - IV ~ 0 1996 2002 0 1696 rn 2002 -60- r;. 60 - &-60- - - 50- 50- 0 50- 4 K r 40- E 40- 5 40- is$30- -6530- is830- E 20 - 20- 20 - t E = 10- = 10- 10- 0, 03 07 Qiboutcvilk Other Vban Wral Djibouti-vilk Other Man Wrii Djibouti-vilie merVban Wrai

3.26. High unemployment rates, in conjunction with low labor participation rates, imply that only a small fraction of Djiboutians engage in market work. Less than 20 percent in 1996 and 30 percent in 2002 of the population aged 15 and above was employed. The situation is even more dramatic for females: less than 15 percent were employed in 1996, though their employment rate rose to 23 percent in 2002. However, this observation must be tempered with anecdotal evidence that suggests the data probably underestimate the participation rate and overestimate the unemployment rate, because they may not capture underemployment. This is particularly true for females; the 1996 Djibouti poverty assessment showed that in households where all men are unemployed, women resort to various activities to feed their families.

25 In 1996, the most disadvantaged economic groups had unemployment rates higher than 60 percent.

66 3.27. The unemployment rate is even higher ifdiscouraged workers are taken into account. The 1996 survey identified individuals who claimed to be unemployed and did not actively look for a job in the four weeks before the interview. These individuals formed 49.2 percent of all who declared that they were unemployed. Technically, these people are considered out ofthe labor force and therefore do not contribute to the unemployment rate. The 1997 poverty assessment states that if these discouraged workers were taken into account, the unemployment rate in Djibouti would have increased to 59.9 percent in 1996.

3.28. The life-cycle pattern of unemployment has altered between 1996 and 2002 for both males and females. The unemployment rate in both years starts out at a very high rate, reflecting the seriousness of youth unemployment. Almost 90 percent ofDjiboutians between 15 and 20 years old are unemployed (Figure 3.12 and 3.13). In 1996, the unemployment rate declined with age and could be explained by the limited slulls ofnew entrants into the labor market. Low unemployment rates for older females can be explained by their very low participation rate, discussed earlier. However, by 2002, a U-shaped pattern ofunemployment rate against age emerges-in part because ofhigher labor force participation by females and males in the older age brackets (Figure 3.14 and 3.15).

Fgure 3.12 Djibouti: Unemploymnt Rate by Age Groups, 1996 Figure 3.13 qibouti: Unemployment Rate by Age Groups, 2002

100% Male 100% Male 90% - 90% Tm -Female 80% -All 80% -All 70% 70% 60% 60% 50% 50% 40% 40% 30% 30% 20% 20% 10% 10% 0% 0% 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60+ 12-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60+

Source: Authors' calculation based on EDAM 1996 and 2002 Source: Authors' calculation based on EDAM 1996 and 2002

Rgure 3.14 Djibouti: Average Male Unemployment Rate by Age Fgure 3.15 Bibout!:. Average Female Ulenpbymnt Rate by Groups, 1996 vs. 2002 Age Groups, 1996 vs. 2002 100% - 90% - 0 199682002 0 19968 2002 90% - 80% - 80% - 70% - 70% - 60% - 60% - 50% - 50% - 40% - 40% - 30% - 30% - 20% - 20% - 10% - 10% - 0% T 0% 7 12-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60+ 12-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60+

Source: Authors' calculation based on EDAM 1996 and 2002 Source: Authors' calculation based on EDAM 1996 and 2002

67 3.29. A non-linear relationship exists between unemployment and years of education. Using 1996 data (which allowed measuring years ofeducation), the unemployment rate is seen to increase with education; pealung at six years of education and then decreasing (Figure 3.16). Female unemployment is almost always higher than male unemployment. A similar pattern of increase followed by decline in the unemployment rate can be seen using data for 2002 on education levels (Figure 3.17).

~~ FQure 3 16 Qibouti Unenpbyment Rate (%) versus Years Figure 3.17 Djibouti:Unenploymnt Rate (%) versus of Mucabon, 1996 Mucatiin Level, 2002

80% -

70% - - - All 60% - 50% - 40% - 30% - 20% - 10% 10% -

0 % "I" I 0 1 2 3 4 5 6 7 8 9 101112131415 None Rimry Lower Upper Higher Mu Years of Mucation Secondary Secondary

Source: Authors' calculation based on EDAM 1996 and 2002 Source: Authors' calculation based on EDAM 1996 and 2002

3.30. In all regions of Djibouti, the public sector is the main employer, but there are regional differences regarding the sector of employment (Table 3.2). The public sector employs the largest share of wage earners at 43.2 percent, followed by private informal sector (20 percent) and the private formal sector (13.3 percent). This is comparable with other developing countries. Lindauer and Sabot (1983) noted that in Afhca, it is not uncommon to find 50 percent or more of wage earners in the government or state enterprises. The private sector is more developed in Djibouti- ville-it is almost nonexistent elsewhere. Approximately 23.6 percent of the people who work are employers or self-employed and more than 19 percent ofthe employed work in the informal sector.

Private Formal 15.4 5.2 7.0 13.3 Private Informal 22.2 12.9 10.1 19.9 Self-employed 22.0 22.1 26.1 22.4 Employer 1.3 1.2 0.5 1.2

3.3 1. There are clear differences in the gender composition of wage earners where males are disproportionately employed in thepublic sector. More than 58 percent ofemployed males work in the public sector. In contrast, only 16 percent of employed females are civil servants; the majority (71.6 percent) work in the informal sector.

68 I Male1 Female Public 58.31 16.0 Private Formal 14.5 11.2 Private Informal 9.8 38.0 Self-employed 16.2 33.6 Employer 1.3 1.1

3.32. More than half of Djibouti wage earners have no formal education. Lack of formal education is disproportionately higher among female workers, at 7 1 percent, compared with males, at 40.8 percent. A larger fraction of males complete their schooling at each level compared with females.

Variable I All I Male I Female % with no formal education I 51.4 I 40.81 70.5 % with primary schooling 18.3 23.2 9.5 % with middle schooling 10.0 11.4 7.7 Yo with secondary schooling 3.7 4.4 2.4 % with vocational schooling 3.4 3.9 2.5 % with university education 3.2 4.6 0.7 Source: Authors'calculation using EDAM 1996

Earnings

3.33. Another key factor in determining earnings from labor is the wage structure. This section illustrates the average wages paid in different sectors for different types ofjobs using survey data from 1996-which, unlike the more recent survey in 2002, contained earnings information on individuals.

3.34. Wages are related to gender, location, education, and sector of employment. Figure 3.18 shows the average monthly wage by gender for different locations in Djibouti. On average, males and residents ofDjibouti-ville have higher wages. Public sector wage earners receive the highest monthly wages followed by the private formal sector. The narrow wage gap between public and private formal earnings does not fully capture the generous benefits and non-wage compensation that workers in the public sector receive. Informal sector workers have very low average wages. Females are overrepresented in this sector, which may explain the wage differential between men and women.

Figure 3.18 Djibouti: Aerage MonthlyWage. US$ (1996) Flgure 3.19 Djibouti: Aerage MonthlyWage by Sector of Employment US$ (1996) I 450.0, Total 450.0, n Total 400.0 4w'0 IMale 350.0 350.0 oFemale 300.0 3w,o 250.0 250.0 200.0 200.0 150.0 150.0 100.0 100.0 50.0 50.0 0.0 0.0 Fhrrai Fubk Rwate Form Rwate hforml Self-Enpbyed Enpbyer I

69 3.36. Wage earners with university Figure 3.20 Djibouti: Gender Wage Gap Ratio by Educatiw, education earn 2.7 times more than those Level Completed, 1996 with only primary education. Wage earnings correlate positively with higher education i:: attainment for both men and women. The 1.5 1.3 1.3 gender gap ratio-the ratio of male to female i;: monthly earnings-is highest for those 0.0 C - .Q c without formal education. Males earn more .- ES DS$g z 'a E 8 'E .$ than twice as much as females. The gender D a 8 8 w >83 gap narrows for those completing lower z" i $ secondary education but rises with upper -1 3B

Lower Secondary

Source: Authors' calculation using EDAM 1996

Recent labor market trends

3.3 7. Notwithstanding statistical deficiencies, data on registered demand and offers of employment over the past seven years show a persistently weak labor demand that helps explain the high unemployment rate (estimated at more than 50percent). Table 3.6 displays the following findings:

Labor supply has exceeded labor demand The registered labor supply has been, on average, about 2,500 persons per year over the period 1998-2004, while registered offers of employment (labor demand) hovered around 1,000 persons a year.

The demand for employment by men has been on average at least three to four times higher than women, while job offers to women have constituted about one-fourth that to men. These trends reflect gender disparities in education and cultural factors that affect women's participation in the labor force.

The majority of recruitment in recent years has been done through private arrangements between employers and employees, evidencing the limited role ofthe national employment service in promoting new recruitment.

70 ONMen 697 768 931 1,050 1,360

Government Service (NES) 102 56 28 133 423 Private Service 726 848 1,070 1,123 1,111

Difference -1,140 -1,864 -2,154 -1,515 -1,103 -3,862 -1,325

3.38. Available data from the Social Security Fund (Organisme de Protection Sociale, or OPS) reveal a modest net employment creation by the economy during 2000-04. Excluding new employment in the military sector (the United States’ Camp Lemonier and new jobs related to the construction of the camp in 2003), net job creation has been relatively modest over the past five years.

3.40. Positive net job creation has been occurring mainly in the service sector, particularly in the following subsectors: 0 building and construction, reflecting the boom in residential construction and donor-financed housing and public works projects 0 maintenance and repairs, offering services to the military, government sector, and households 0 small-scale retail trade, serving the emerging foreign military demand for consumer and equipment goods 0 hotels and restaurants, reflecting the presence ofthe large military contingents from the United States, France, Germany, and Spain. 0 consultancy services (including technical assistancehtudies), arising mainly from donor-funded projects.

3.41, However, other sectors have experienced little or no growth in net employment-notably transport, finance, manufacturing, and the domestic sector. The transport sector appears to be creating fewer jobs in net terms relative to its potential, particularly in cargo handling. One reason may be the increasing sophistication ofport operations, which requires skilled technicians, who are scarce in Djibouti. This may have led to a slight increase in imported foreign labor to perform some transport services, as well as some attempt at capital substitution by mechanizing some repetitive tasks. Also, the recent changes in the management of the port-the concession given to Dubai- could have led to better use ofexisting labor and less need for new recruitment. The financial sector has added practically no net jobs over the past five years, probably because of the limited bank competition. Both the industrial sector and artisans, shops, and the large wholesale markets in the commerce sector have seen some net job increases over the period, but they have been marginal. Public enterprises and pension funds have not had any meaningful impact on net employment creation. There is a downward trend in net job creation in the household sector, which in the past provided more jobs in terms ofdomestic work. These domestic jobs were offered mainly by foreign military personnel, diplomatic and international organization staff, and wealthy Djiboutians. The reduced numbers ofFrench foreign military personnel in recent years may be a contributory factor. Another reason relates to the massive outflow of refugees in September 2003, most of who were employed in the household sector.

71 147 196 253 103 150 223 155 68 135 294 159 94 326 232 23 147 196 253 103 150 223 155 68 135 294 -159 94 326 -232 23 3 29 88 10 78 18 31 13 71 28 43 29 14 15 152 0 9 64 0 6440 4 28 4 24 8 4 4 105 3 20 24 10 14 14 31 -17 43 24 19 21 10 11 47 17 120 106 40 66 90 17 73 304 78 226 151 68 83 568 17 120 106 40 66 90 17 73 304 78 226 151 68 83 568 634 59 45 14 49 34 15 274 146 128 237 161 76 267 0 22 43 25 18 6 14 -8 107 39 68 56 47 9 109 Commerce - Wholesale 2 6 1 15 -14 3 2 1 54 122 0 -6 Commerce - Specialized 4 6 15 5 10 40 18 22 158 103 55 178 112 66 159 Commerce - Muiti activities 40 10 11 81 48 55 7 129 114 15 280 216 64 288 I99 89 242 0 1 0 0 0109 1 96 3 12 36 -24 -19 0 14 8 40 -32 15 24 -9 92 77 15 147 49 98 86 Restaurants and Hotels 8 38 6 15 -9 48 41 5 119 74 45 44 53 -9 70 Other activities 3 30 23 0 23 58 40 18 58 55 3 79 58 21 95 Other activities (Leisure) 2 -2110 1100 024 -263 3 10 0 6990 9 42 7 35 22 20 2 41 20 21 136 0 000 0 40 4 20 2 24 -2 4 0 2540 4140 14 10 142 2 46 Cargo Handlingnrans 0 4450 5 24 7 17 19 20 -1 35 14 21 86 0 18150 1532 123 130 3 36 Social Organizations 0 18150 15 32 123 -1 30 3 36 0 I6 27 0 27 15 3 12 5 10 -5 28 0 28 78 0 16 27 0 27 15 3 12 5 10 -5 28 0 28 78 0 000 0 DO 0 00 013 2 2 Pnvate Orgs And Educ. 0 0 0 0 1 3 -2 -2 Cultural Organizations 0 0 0 0 0 0 0 6300 0 00 0 00 0 00 0 63 0 0 0 0 0 000 0 0 0 0 429 0 429 2 1 1 0 0 0 429 429 2 1 1 184 626 605 253 352 569 363 206 1,522 795 727 874 792 82 31,993 Note: a. From January lo September 2w4 Source: OPS, 2w4

3.42. Recent positive net job creation reflects short-term opportunities from the foreign military presence, donor-funded development projects, or private sector-funded housing or commercial activities. The short sample period and the mall number of observations4ata on sectoral net job creation are available only for the past five years (Table 3.7)-make it difficult to draw firm conclusions. However, it is evident from the data that despite the observed overall modest trend in net employment, significant net job creation has occurred over the past few years largely in construction and related services, small-scale retail and specialized trade, consultancy services, and hotels and restaurants. These developments have been linked to the recent expansion ofthe foreign military presence and increases in donor-financed housing and public works projects. Many of these job opportunities are short term, offered by small entrepreneurs using semiskilled workers and low technology to meet short-term business needs. Recent examples include small contracts awarded to a few domestic firms to hire employees on a short fixed-tern basis to do small repairs on refrigeration plants and air conditioners at the military base, and to build housing and restaurant facilities for troops on assignment.

72 3.3. EXPLAINING POOR LABOR OUTCOMES

3.43. The limited labor demand by the formal private sector in Djibouti could be explained by four broad sets offactors:

(1) distorting government pay and employment policies that have dnven up labor costs, under which public sector workers earn comparatively more than private sector workers, after controlling for other factors (education, age, gender)

(2) a weak investment climate that has stifled the entry of job-creating new firms and dampened incentives for existing firms to invest, expand, and hire;

(3) relatively onerous labor market institutions that have served as a disincentive for firms to hire new workers

(4) a paucity of skilled labor and poor work ethics and standards, which lead to laxity and worker absenteeism, negatively affecting labor productivity.

3.3.1. Wage Distortions in the Labor Market: Do Public Sector Employees Earn a Rent?

3.44. The public sector plays an important role in the formal labor market, employing roughly 56.4 percent of the labor force, compared with 17.6 the formal private sector’s 17.6 percent.26 The public wage bill is a significant component of the total government budget. Some portion of the wage bill can be attributed to excess employment, but some may be caused by wage levels that are more than competitive. A related concern is that this dominant public sector sets wages without taking account ofmarket forces, and that is having a spillover effect on the rest ofthe formal sector.

3.45. Public sector wages, until very recently, were a benchmark for private sector wages. Together with rigidities in the labor code, they help explain the paradox of downward inflexibility in wages and high unemployment. Table 3.8 uses public wage data by categories from the Ministry of Finance and data from a survey conducted by Bank staff in March 2005. The current ratio of public to private wages stands at 1.12, though this masks disparities across slull levels. Unskilled workers earn relatively higher wages in the private sector and more skilled individuals receive relatively greater earnings in the public sector, though the gap is reduced in managerial positions.

26 This estimate is based on a 1996 representative household survey of the sedentary population. The labor force includes public and private workers (both formal and informal). If only public and private formal workers are considered, the public share is 76 percent and the private share is 24 percent.

73 Overall 1.12 Unskilled workers 0.40 Low-skilled workers 0.74 Highly-skilledworkers 1.34 Managers 1.29

3.46. The wage differential ratio of 1.1 between the public sector and the formal private sector is corroborated by other data sources. Using data made available by the social security fund (OPS), the private sector wage can be inferred from the total contributions collected, the number of contributors, and the contribution rate (assuming the collection rate to be 100 percent-that is, no arrears in contribution payments). Using this methodology, the annual private sector wage for 2002 is estimated at DF 1,291,250. This amount is equivalent to US$7,266 in current US. dollars. It is lower than the annual average public sector wage for that year (close to US$8,000) and translates to a public-private wage differential of 1.1. However, this estimate must be read with caution, given the extent of evasion of social security contributions by Djibouti's private sector. In addition, the sample provided by OPS does not include all private sector wages in the informal sector.

Public 4,477.28 3,391.44 Private (formal) 4,073.66 3,182.97 Private (informal) 1,562.53 2,822.57 Source: Authors'calculafion using EDAM 1996

3.47. When annual public sector wages are compared with wage income in the private informal sector, the differential is 2 to 3 times larger. An alternative data source, the 1996 EDAMS household survey is used. Though dated, table 3.9 shows the wage differential between the public and private sectors: the average annual public sector wage in 1996 was nearly US$404 higher than the average annual private formal sector wage, resulting in a differential of 1.1. The gap widens when public sector wages are compared with the private informal sector, where the differential is 2.86. These numbers are likely to be underestimates because of the sample size and biases in the data sampling method, but they tell a similar story: the obvious discrepancy between the public and private sector average wages.

3.48. There is also a wage gap between Djibouti nationals and foreign workers: Djiboutians earn 4.1 and 4.3 times more than Ethiopians and Somalis, respectively, in annual wage income. In terms ofmonthly wages, Djiboutians earn 3.6 and 3.3 times more than Ethiopians and Somalis, respectively (Table 3.10).

Ethiopia 7.7 88.6 968.3 Somali 7.0 95.6 920.9

74 3.49. Using a simple average ofthe wage difference between public and private (formal) workers to determine whether there is a wage premium is misleading. Doing so does not account for differences in human capital endowment and other personal or household characteristic that influence earnings. The next section explores this issue through a three-step empirical analysis.

Estimating public-private sector wage differentials and returns to education

3.50. What factors affect workers’ decision to participate in public or private sector employment in Djibouti? Is there is a public sector wage premium (after controlling for human capital endowment and differences in private returns to education across sectors)?

3.51. Studies of public-private wage differentials are based on models of earnings determination developing by Mincer (1974). Variations in earnings are due to differences in human capital, measured by formal schooling and work experience. Other typical controls include location and gender dummies. In this section we estimate standard Mincer wage regressions separately for both sectors while correcting for selectivity bias. This helps account for factors that influence the sorting or queuing of workers into both sectors. A detailed account of the methodology and regression results appears in annex 3.1.

Descriptive statistics: key characteristics of the workers examined

3.52. The analysis is based on the 1996 individual and household survey (EDAM 1996), which makes it possible to estimate individual returns to schooling and other wage determinants across various categories, including men, women, public, and private (formal sector). Table 3.1 1 contains characteristics of the workers examined for this section, excluding employers. The sample is restricted to 2,011 wage earners between 12 and 65 years of age. Forty-three percent are employed in the public sector, 14 percent in the private formal sector, and 22 percent in the informal sector. Males make up 64.3 percent of overall wage earners. More than 75 percent of male workers are married and more than 58 percent work in the public sector. In contrast, 74 percent of female workers are unmarried and only 16.2 percent work in the public sector. Most ofthe workers in the informal sector are women, with nearly 75 percent having no formal education. A final observation pertains to income volatility. Income earned by informal sector workers is far more variable that that earned by workers in the public and private sector, as seen by the standard deviation in monthly earnings.

All Male Female Public Private (Formal) Self-Employed Variable Mean STD Mean STD Mean STD Mean STD Mean STD Mean STD monthly wagedeamings (US$) 288.2 (373.5) 354.5 (362.6) 168.8 (363.0) 375.4 (288.8) 348.5 (262.5) 240.0 (563.4: age 35.0 (11.0) 36.8 (10.7) 31.7 (10.7) 35.8 (10.4) 34.7 (9.7) 38.2 (11.7) female (%) 35.7 100.0 13.2 30.0 53.6 married (%) 58.4 74.6 29.2 69.9 63.3 58.5 % in public sector 43.6 58.9 16.2 % in private (formal) sector 13.4 14.6 11.3 % in private (informal) sector 20.0 9.9 38.3 % self employed 22.6 16.3 33.8 % obtained certificate 1.4 1.5 1.1 2.2 3.3 0.0 %with no formal education 51.4 40.8 70.5 47.1 47.1 74.8 %with primary schooling 18.3 23.2 9.5 23.4 26.3 11.9 %with middle schooling 10.0 11.4 7.7 15.2 13.7 3.3 % with secondary schooling 3.7 4.4 2.4 5.6 5.9 1.8 %with vocational schooling 3.4 3.9 2.5 5.4 4.8 0.9 % with university education 3.2 4.6 0.7 5.4 5.2 0.7

INo of. Uncensored observations 2011 1293 718 878 270 453 Source: Authors’ calculation based on EDAMS 1996. Note: Standaid deviations (STD) are not reported for dummy variables.

75 Labor market participation

3.54. This subsection examines how factors such as educational attainment, experience, gender, and parent's occupation influence workers' sectoral choices. Participation equations are estimated using the probit model, and the outcomes are shown in table 3.12. For each variable, the results provide the marginal effects or the probability ofjoining the public or formal private sector, calculated as the mean values of the variables. For comparison, the participation choices of self- employed workers are also included.

3.55. The relationship between age and labor force participation is not linear across all three sectors. For workers in the public sector, the probability of participation increases until an individual is 42.3 years and then starts decreasing. The turning point is similar for individuals in the private sector (at 40 years) and individuals who are self-employed (at 43.8 years).

3.56. Education positively affects participation in the public and private sectors and reduces the likelihood of participation in the informal sector. Workers with university education have a higher probability ofjoining the public sector. Age is a proxy for more work experience. The older a worker is, the higher the probability ofjoining each sector. Females are less likely to participate in the formal sector and more likely to participate in the informal sector.

3.57. Having a parent in the public sector raises the probability of Jinding employment in the public sector and reduces the probability of working in the informal sector. Income effects on participation, measured by unearned household income, are statistically significant and negative for public and private sector workers. This is expected because higher non-labor income reduces the need to seek formal employment.

Marginal Std. ' Marginal Std. Marginal Std. Variables Effect Dev Effect Dev Effect Dev age 0.011 *** (0.00) 0.004 *** (0.00) 0.007 *** (0.00) age squared/100 -0.013 *** (0.00) -0.005 *** (0.00) -0.008 *** (0.00) female -0.068 *** (0.01) -0.008 *** (0.00) 0.009 *** (0.00) married 0.009 '* (0.00) 0.001 (0.00) -0.009 *** (0.00) obtained certificate 0.048 *** (0.02) 0.039 *** (0.02) primary schooling 0.042 *** (0.01) 0.014 *** (0.00) -0.009 ** (0.00) middle schooling 0.106 *** (0.02) 0.020 *** (0.01) -0.014 ** (0.00) secondary schooling 0.134 *** (0.03) 0.028 *** (0.01) -0.004 (0.01) vocation training 0.144 *** (0.03) 0.016 ** (0.01) -0.014 (0.01) university 0.138 *** (0.04) 0.023 *** (0.01) -0.022 ** (0.00) parents work in public sector 0.010 ** (0.00) -0.002 (0.00) -0.011 ** (0.00) parents work in private sector -0.013 * (0.01) -0.003 (0.00) -0.019 *** (0.00) In-unearned income -0.005 *** (0.00) -0.002 *** (0.00) -0.001 (0.00)

0.15.~ IPseudo R square 0.33 0.19 I Note: standard errors are in brackets. *** denotes statistical significance at 1% level, ** at 5% and * at 10%. Source: Authors' estimates based on EDAM 1996.

Determinants of wage earnings

3.58. A wage regression using the Mincer framework is often used to calculate the private returns to education and the degree to which other individual characteristics may influence earnings. The model is estimated for public and private wage earners using ordinary least squares

76 with and without conditioning on selection choice. Inclusion of the selection term (constructed from the earlier participation equation controls) for the fact that workers do not randomly work in either sector allows us to test whether factors influencing the sorting of workers into these two sectors also affect wage earnings. The independent variables common to each wage regression are dummy variables for rural and urban locations, for completing education at various levels, for age, for females, and for marriage.

3.59. Experience exerts a positive influence on wage offers, particularly in the private sector (table 3.13). The coefficient magnitudes differ for both specifications and indicate the importance of correcting for selectivity. Referring to the selectivity-corrected earnings regressions, experience exerts a positive influence on wage offers while the quadratic term has the expected negative sign. The curvature of the wage-experience profile is steeper in the private sector than in the public sector, which is consistent with the general pattern in many countries.

3.60. Being female has a negative impact on wages in the private sector. The coefficient for female workers in the private sector is negative and statistically significant, implying that female workers face an earnings penalty in the private sector. No clear conclusions could be drawn about females in the public sector given the statistical insignificance of the female coefficient. The data for similar professions, such as biologists, teachers, and administrative agents, show that males earn more on average.

3.61. Marriage has a positive impact on wage offers in the private sector. This may be due to reduced absenteeism or lower monitoring cost associated with married workers. The coefficients for schooling attainment will be discussed in the next section.

3.62. Private sector workers have higher productivity than the average worker in the labor market. By contrast, public workers have below-average productivity. The coefficient estimates of the selection term for individuals were statistically significant, being negative in the public sector and positive in the private sector. The positive coefficient for the private sector indicates that individuals who select jobs in the private sector have higher productivity than the average worker. The negative coefficient for public sector workers implies that they have lower productivity than the average worker.

age 0.007 (0.02) 0.108 *** (0.03) -0.047 '* (0.02) 0.066 *** (0.02) age squaredl100 0.011 (0.03) -0.123 "* (0.04) -0.037 (0.02) -0.072 ** (0.03) female dummy 0.176 (0.11) -0.424 *** (0.10) -0.059 (0.06) -0.314 *** (0.07) married dummy 0.101 (0.06) 0.220 ** (0.09) 0.183 *** (0.06) 0.201 ** (0.09) primary school 0.141 ** (0.06) 0.402 *** (0.11) 0.233 **' (0.05) 0.280 *** (0.08) middle school 0.375 **' (0.10) 0.756 *** (0.14) 0.551 *** (0.07) 0.613 *" (0.13) high school 0.670 *** (0.12) 0.954 *** (0.19) 0.856 *** (0.07) 0.751 *** (0.10) vocational training 0.659 *** (0.13) 0.815 *** (0.18) 0.870 *'* (0.06) 0.667 *** (0.19) university 0.942 *** (0.12) 0.977 *** (0.18) 1.137 *** (0.09) 0.822 *** (0.14) constant 5.603 ** (0.57) 2.747 *** (0.93) 4.376 *'* (0.28) 4.164 *** (0.42) selection term -0.274 ** (0.11) 0.318 (0.18) --

Chi square 1167.13 31 6.20 0.42 0.27 Note: Standard errors are in brackets. *** denotes statistical significance at 1% level, ** at 5%, and at 10%. Source: Authors' calculation based on EDAM 1996

77 Public-private sector wage differentials

3.63. Using a simple average of the wage difference between public and private dformal) workers to determine whether there is a wage premium is misleading, because doing so does not account for differences in human capital endowment and other personal or household characteristics that influence earnings. It is possible to break down various components that contribute to the difference between predicted public and private sector wages using an Oxaca- Blinder decomposition.

3.64. After controlling for differences in education, experience, and selectivity bias, large public-private wage differentials still exist in Djibouti and they are primarily a rent or premium. Table 3.14 shows the Oxaca-Blinder decomposition results for Djibouti with and without conditioning on selection. Both methods clearly show that the most important factor determining a positive wage differential in favor ofpublic sector wages is a wage premium or rent. The premium is much larger when characteristics that determine participation in both sectors are controlled for, as captured in the second column.

3.65. Public sector wage premiums are also evident in other developing countries. Terrell (1993) found in Haiti that public administration workers earned a sizeable rent. Similarly, Lindauer and Sabot (1983) found that the wage premium was the most important determinant of public-private wage differential in Tanzania. In Djibouti, the positive rent earned by the public sector is more likely due to the legacy of French colonial rule, which paid high wages to the public sector and rewarded the patronage and loyalty of the elite class. Given that these differentials are estimated with wages net of fnnge benefits and given that the public sector has higher benefits (leave time, health benefits, and so on), the estimated rents or base wages can be viewed as conservative.

3.66. The results from the estimation of the selection process and its effect on wages seem to indicate that there are barriers to entry into the public sector. Public sector employees are more likely to be male and to come from the elite class, with parents in the public sector. The policy implications from this analysis are that the public sector hiring and wage-setting practices are not efficiently allocating labor, and hence are wasting resources.

Total log mean differential 0.406 0.038 Components attributable to: Wage premium 0.212 2.856 Human capital endowments 0.342 -0.065 Market returns -0.148 -1.800 Selection -0.953

Private returns to education

3.67. Aside from the higher base wages for government work, differences in returns to schooling for workers in the public and private sectors may also help explain the queuing for public employment. The results from the earnings functions can be used to estimate private rates of return to each level ofschooling.

78 3.68. Table 3.15presents the computedprivate wage returns per year of schooling for different education levels in the public and private vormar) sector using results from the previous wage regressions.

Selectivity-Corrected OLS Regression Regression Public Private Public Private Primary 2.4 6.7 3.9 4.7 Middle 5.8 8.9 8.0 8.3 Secondary 9.9 6.6 10.2 4.6 Vocational 10.5 2.2 11.8 2.0 University 13.6 1.2 14.0 3.6 Notes: Estimates for primary category is the earnings premium over no-formal education. Sirnilart)', middle school category shows earning differential over primary school. Secondary category shows differential over middle school, vocational shows differential over middle and university shows differential over secondary. ISource: Authors'calculations based on wage regressions in Table 3.13.

3.69. Workers in the public sector earn larger private rates of return to schooling than do private sector workers with secondary, vocational, and university education. For wage earners in the private sector, the private return to education peaks at 8.9 percent after completion of middle schooling, compared with 5.8 percent for those in the public sector. The private return to secondary schooling is 6.6 percent per year of schooling for wage earners in the private sector relative to 9.9 percent for those in the public sector. The returns for private sector wage earners with vocational training drop to 2.2 percent, compared with 10.5 percent for public sector workers with the same qualification. Returns to education for workers with university level education are much higher in the public sector than in the private sector. The private return of moving from high school to university education is 1.2 percent compared with 13.6 percent for graduates working in the public sector.

3.70. As illustrated earlier, higher private returns do not, however, signal that productivity is higher (and hence better rewarded) in the public sector. The higher returns to education in the public sector are explained by the distorting government pay policies, the legacy of rewarding educated workers for patronage and loyalty to the ruling elite, and the lack of viable job opportunities in the private sector for educated workers.

3.71. Average returns to education in Djibouti can also be examined for all workers and for males and females separately. Separate wage regressions corrected for selectivity are shown in table 3.16. Table 3.17 provides results using the worlung population engaged in various sectors, corrected for selectivity bias.

79 Variables AI I All (years of edu.) Male Female age 0.095 *** (0.03) 0.103 *** (0.03) 0.133 *** (0.03) age squared11 00 -0.104 *** (0.03) -0.109 *** (0.03) -0.144 *** (0.04) -0.131 * (0.08) female dummy -0.875 *** (0.09) -0.783 *** (0.09) married dummy 0.260 '** (0.05) 0.270 '** (0.05) 0.273 ** (0.12) 0.067 (0.23) primary school 0.597 *** (0.06) -- 0.492 *** (0.07) 0.908 *** (0.14) middle school 1.003 *** (0.09) 0.787 *'* (0.10) 1.533 *'* (0.20) high school 1.328 *'* (0.13) 1.184 **' (0.14) 1.795 *** (0.31) vocational training 1.348 *** (0.13) -- 1.166 *** (0.15) 1.860 *** (0.27) university 1.596 *** (0.14) -- 1.500 *** (0.14) 1.923 *** (0.45) constant 2.923 *** (0.67) -- 2.177 *** (0.79) 1.838 (1.55) selection term 0.345 ** (0.17) 0.367 *' (0.16) 0.377 ** (0.18) 0.470 (0.41) years of education --- 0.116 *** (0.01)

Chi square 2006.10 2027.49 1497.17 779.65 No. of uncensored obs. 201 1 1997 1293 718 Note: Standard errors are in brackets. *** denotes statistical significance at 1% level, *' at 5%, and at 10%. Source: Authors' calculation based on EDAM 1996.

Note: Estimates for primary category is the earnings premium over no-formal education. Similarly, middle school category shows earning differential over primary school. Secondary category shows differential over middle school, vocational shows differential over middle and university shows differential over secondary. ISource: Authors' calculations based on wage regressions in Table 3.13 and 3.16.

3.72. An additional year of schooling, corrected for the worker's sector allocation, raises earnings by 11.6 percent-higher than the world average of 10 percent.2' This figure may be overstated because it does not account for years ofrepetition, which is high in Djibouti, particularly for secondary education. Compared with other Ahcan countries such as Ghana, C6te d'Ivoire, and Kenya, returns in Djibouti to secondary and university schooling are lower (Schultz 2003).

3.73. In addition, the analysis of public and private wages in Djibouti revealed that private returns depend on whether the worker is employed in the public sector or in the private sector. Estimates on private returns to education should be treated with caution, and future work should provide more accurate estimates by sector as well as control for selection bias.

3.74. Female workers who have completed primary and middle schooling earn relatively higher returns than their male counterparts. By contrast, male workers who have completed secondary schooling and tertiary education have higher returns to education than women. Returns are almost double for females relative to males completing primary and middle school, at annual rates of 15.1 percent and 15.6 percent, respectively. However, this reverses with secondary schooling and higher levels of education. Males receive higher returns on having completed secondary schooling and tertiary education.

21 See Psacharopoulos (1994) for world averages.

80 3.3.2. Investment climate and job creation

3.75. The reasons for the poor labor outcomes observed in Djibouti are complex, and some lie outside the labor market. One reason is the high cost of doing business. Job creation largely depends on the ability ofnew firms to enter the market and hire workers. The entry of new firms, and the expansion of existing firms, is a prerequisite for employment growth and unemployment reduction. Firm entry depends on the costs of starting a business and also on the future expected benefits and costs of operating a business. Similarly, firm growth depends on the rate of return to investment which in turn is influenced by the costs of doing business. Institutions, policies, and regulations that affect firm entry, survival, and growth are referred to in this report as the investment climate.

3.76. Djibouti's overall business environment has improved over the past few years and the climate for trade and investment is reasonably good compared with that of many of its African neighbors. A number of positive aspects include a strategic geographic location, an open trade and investment policy, no controls on foreign exchange regime, a stable currency, substantial tax breaks, and other incentives. Investors do not cite the immigration process as being particularly lengthy, as in many other countries. The company registration process, while somewhat lengthy and complicated, does not take an inordinate amount oftime and does not present unreasonable delays. The goods clearance processes are considered reasonable by investors. Still, much work remains to be done to make Djibouti a premier investment destination.

3.77. Entrepreneurial activity remains limited. Investors still encounter multiple bottlenecks across the entire spectrum of conducting business (start-up, operation, and exit). This chapter argues that the quality ofbusiness regulations and the institutions that enforce these regulations are a major determinant of private investment and job creation in Djibouti. Cumbersome and unpredictable rules and regulations increase the costs of doing business and deter investors. Heavy and costly regulation also encourages entrepreneurs to operate in the informal economy. In Djibouti, entrepreneurs registering a business do not complain about the number ofprocedures or cost to register a business. However, land registration is a major bottleneck. It takes on average more than 30 days to regster property in Djibouti, compared with 21 days in Yemen. Land registration is also costly. Regstering property costs approximately 7.5 percent of the property value, compared with 3.9 percent in Yemen. High costs of dismissal are also reported as a main obstacle by investors. In Djibouti, on average, an employ,er must pay 20 weeks of wages to the worker to be dismissed, compared with 17 weeks ofwages in Yemen (table 3.18).

Sources: Djibouti - USAID Investor's Roadmap (2002),World Bank's Doing Business Indicators, (2004).

81 Employers’ views on the major obstacles tofirms’ operation and growth

3.78. One way to assess the investment climate is to consider the perceptions of entrepreneurs about the major constraints on running and expanding their businesses. Such information is gathered in the World Bank Doing Business Database and the Investment Climate Surveys conducted by the World Bank across the world. These exercises have not been conducted in Djibouti. This section draws on available information on the country’s investment climate.

3.79. Over the past few years, firm-level survey data have been collected on large, formal sector firms and on small, informal sector firms. These data also draw on entrepreneurs’ perceptions of obstacles to firm growth and survival. The analysis is based on a purposive sample ofbusiness managers oflarger private sector firms (employing 20 to 100 workers), and managers of small businesses (employing 1 to 10 employees) in the informal sector. Three surveys were used: a survey offive large private sector firms (USATD, 2002); a survey of 150 informal sector enterprises (Ministkre de Finances 2002); and a recent qualitative survey on obstacles to businesses and labor costs in Djibouti, Ethiopia, and Yemen, conducted by the World Bank in March 2005. (See annex 3.3 for details on the methodology and results ofthe survey.)

3.80. Following are the main findings from these surveys:

0 High unit labor costs. In Djibouti unslulled and slulled wages are 5 to 10 times higher than in neighboring countries. However, they are not generally higher in PPP dollars, reflecting the high cost of living in Djibouti. In other words, an unslulled laborer costs an employer more in Djibouti than in Yemen or Ethiopia but takes home much less purchasing power. As a result, the reservation wage is higher in Djibouti than in neighboring countries; 0 High nun-wage costs. Social security charges and allowances are higher in Djibouti than in neighboring countries. 0 Low capital and factorproductivity because of the high cost structure. Indirect costs such as utilities, transport, finance, and taxes, as a share of total costs, are two to three times higher than in neighboring countries. These high costs are lowering the competitiveness of firms, inhibiting firm expansion and job creation. 0 High transaction costs. Firms report obstacles associated with the unpredictability ofrules and regulations and administrative red tape. 0 Labor regulations and skill shortages. These are also high on the list of obstacles. 0 Lack of procedural transparency. This is the most recurrent problem mentioned by investors in almost every stage of the investment process. There are few formal, written guidelines. The success of many applications and requests hinges on the approval of the minister responsible for the particular portfolio. The relatively small size of the country and its government plays a large role in the intimacy of decision making by political leaders. However, the lack of systematic approval processes reduces transparency and increases regulatory uncertainties, raising the costs ofdoing business. 0 Regulatory uncertainties. Entrepreneurs these uncertainties as a major obstacle to expanding business. High taxation, burdensome tax administration, poor access to and high cost of capital, cumbersome regulations, and lack of transparency are top business concerns. A significant number of firms also consider contract enforcement and conflict resolution as major concerns.

3.81. See annex 3.2 for a comprehensive account of the procedures and costs of starting, operating, and closing a business inDjibouti.

82 Business start-up: company registration, entry visas, land registration

Company registration

3.82. Although the list ofprocedures for setting up a business is not extremely long, the company registration process remains unnecessarily burdensome. Investors report the following main obstacles:

The company registration process is dispersed cross several agencies with little or no coordination between them; moreover, there are numerous duplicative requirements- often multiple interactions with a single agency. Work has been done to establish a one- stop approach to registration but it has not been practically implemented.

The process of obtaining investment incentives is complicated by a vague investment code whose opaque application raises uncertainty for investors. The complexity of the registration process is exacerbated by the absence of a formal guide (box 3.3). This creates uncertainty in the process and results in unnecessary delays. Even with the aid of facilitators and local lawyers, it can still take foreign investors up to one month to complete the process offorming and registering a company. The tax incentive system is complex and inconsistently applied. Investors have not had access to the complete range of incentives offered by the investment code because of confusing procedures (see Box 3.3).

3.83. The surveys also revealed high business start-up and operating costs for small enterprises. It can take two to three months and can cost more than US$340 to register a limited liability company with a sole proprietor. In addition, a number of legal and regulatory constraints limit further development of small enterprises, hindering firm productivity and job creation. They include regulations governing the legal recognition of economic activities, a heavy taxation system, and discriminatory treatment regarding fiscal incentives regulated in the investment codes, limited access to finance, and limited access and high costs ofpublic infrastructure services.

Entry visa, and residency and workpermits

3.84. Visas and residency and work permits do not pose significant obstacles or costs for foreign investors or their expatriate employees. In general the systems function quite satisfactorily and in practice cause few problems. Nevertheless, private sector operators have identified several impediments that hinder the pre-investment phase:

The absence ofa business visa welcome system, which constrains investors who arrive in Djibouti willing and able to set up a business venture rapidly. Delays and a lack oftransparency in the process for approving expatriate employment contracts and issuing expatriate work permits. The redundancy ofrequiring holders ofvalid residency permits to apply for a new entry visa to re-enter Djibouti.

83 Box 3.3 A Complex Legal Framework for Investment

Djibouti suflers from both too little and too much legislation. Many texts overlap. Some are obsolete; others are :ontradictory, while still others have no implementing legislation. This results in a level of uncertainty that fosters %mbiguityand, ultimately, arbitrary decisions that undermine the credibility of the legal system.

The 1994 Investment Code was complex. It provided for two separate regimes (A and B), depending on the amount of the proposed investment. The legal and institutional system governing the creation of free zones is also too complex. A system of free zones and duty-free points would give rise to the establishment of duty-free companies known as offshore :ompanies, which could advantageously replace the duty-free industries and the incorporated free trade zones. The entire system must be simplified and unified to make it functional and more attractive to investors.

4pprovals issued under the Investment Code are not transparent. The procedure for granting benefits under the Investment Code has been criticized on many occasions. The time limits observed in applications for acceptance or approval often seems much longer than those stipulated in legislation. The approval procedure under regime B most often requires an authorization from one or more ministries, when it does not go all the way to the OEce of the President. Reasons for rejecting applications are not always given, and beneficiaries do not receive all the benefits normally associated with approval. There is no investors’ guide that might clarify administrative procedures.

The Code does not afford protection against the risk of expropriation. Among the risks for which coverage is thought particularly important for investors are, first and foremost, nationalization and expropriation. The risk of nationalization resurfaced in 2001 with the appearance of a measure, effective retroactively, applicable to the jobs of store man and goods handler. Under the terms of this provision, these jobs were reserved for Djiboutians or corporations whose parts or shares were held entirely by nationals of Djibouti. This development is all the more unfortunate in that the Investment Code does not mention such situations, which are far from conducive to foreign investment.

The 2004 Investment Code represents significant progress over the 1994 Code, but it is still far too complex for an administration with limited means. The Djibouti Investment Charter distinguishes between investment regimes on the basis of sector of activity, there being five in all: industry, services, agriculture and health, tourism and “new sectors.” For each sector several regimes are set out. Industry has six: export enterprise, pioneer enterprise (technology and skills above the level found in Djibouti), strategic local enterprise, modernization and expansion, industrial construction (buildings with an area of more than 1,000 m3), and small and medium enterprises (equipment valued at less than DF 10 million). Three regimes are listed for services: service-exporting free trade zone, offshore activities (promoting tertiary finance), and port and airport free zone. The 15 regimes have their own incentives, including tax abatement or exemption, loans or overdrafts at preferential rates, duty-free import of equipment, and the possibility of repatriating profits, dividends, or capital.

The Code has been simplifiedyet there is aplethora of regimes. The provisions of this bill have the merit of reflecting a desire to modernize and adapt the existing Code to economic realities and needs. It also clarifies the sectors in which the public authorities intend to focus their development efforts, at least through fiscal and customs-related incentives. The bill also has the advantage of combining in a single document the regime for investments and the regime applicable to the various types of free zone. Yet the bill is probably too complex to administer, not least of all for a service woefully short of manpower and equipment. Interviews conducted with private sector companies revealed the existence of substantial delays before the Board issued its approval or denial. And though the law states that a lack of response from the Board within the stipulated time limits signifies approval, investors were usually unable to benefit from the tax exemptions granted by the law. Specifically, without a decree (arGfe3 indicating approval by the Council of Ministers, investors are unable to benefit from the income, real estate, and import duties exemptions. Many nations have done away with incentives altogether, instead moving toward lower tax rates for all companies, or stepped corporate tax rates for manufacturing firms, mining firms, and commercial and trading firms. This reduces the ambiguity associated with the approval of incentives.

Land registration and construction permits

3.85. Investors consider the site location process to be one of the most difficult aspects of the overall start-up process. Because of scant natural resources and limited infrastructure, the cost of factors of production in Djibouti is relatively high. This high cost combined with a heavily centralized, multilayered, and complicated administrative system makes land location (and site

84 development) one of the principal constraints faced by the typical investor. Reforming the processes and institutions involved in land acquisition are critical prerequisites for attracting new investors. The following issues are reported by firm managers:

0 Acquiring available land is one of the most difficult undertalungs that an investor must face. The level of government involvement in land purchases is inappropriate. The process is time-consuming and lacks transparency. 0 There is an acute shortage ofreasonably priced land ready for investment. 0 The current process for acquiring, registering, and recording land is complicated, overly centralized, multilayered, and lacks standard application forms. The multiplicity of government agencies and commissions involved makes the process prone to abuses. 0 The process for obtaining construction permits is time-consuming and cumbersome. Multiple agencies are involved in the approval process. Some members of the private sector noted that the time limit imposed on government officials to issue building permits is routinely ignored.

Business operation

3.86. The costs of importing and exporting goods, banlung, public utilities, taxation and contract enforcement affect firm’s operation and expansion.

Importing and exporting goods

3.89. Import and export procedures are fairly simple: unlike in many other Afncan countries, there are no goods pre-inspection procedures, and the port ofDjibouti generally operates below full utilization. Port and customs processing times are not as lengthy as in other Sub-Saharan Afiican countries. Goods are generally cleared in 72 hours or less, and theft is not considered high relative to neighboring ports.

3.90. Customs clearance times of 72 hours are still somewhat competitive with other African ports. But to remain competitive the services offered by the Port of Djibouti will need to be provided faster. Modernization programs in Dar-es-Salaam, Beira, and Maputo will leave the Port ofDjibouti less than competitive in the future. Speedy offloading and clearing of goods is a key to the success of a port. Although Customs is not particularly well equipped technologically, it can clear goods within a day (as evidenced by the fast-track clearance of perishable items). Despite the introduction of the Single Bill of Entry widely used in the rest of the world, there has been no attempt yet to unify the various forms used by importers/exporters into a single form. Electronic Data Interchange systems are not used in Djibouti. In general, forwarders must first await the arrival of shipping documents (bill of lading, invoice, paclung list) from their foreign agent, and then submit them to the shipping agent to retrieve the manifest.

Importers also report periodic problems with Customs officials in goods valuation, particularly for goods originating from ports outside Europe. Because there is no formal process for resolving conflicts, disputes are often settled by alerting more senior customs administrators known to the plaintiff. Despite these concerns, the process as a whole remains manageable because most of the goods are transiting to Ethiopia and are not taxed at the local consumption rate. It must be noted that in Ethiopia clearing procedures are considerably lengthier, so goods bound for Ethiopia are allowed to remain in the port for up to two weeks without incurring charges.

85 3.91. Port dues are generally moderate, to remain competitive with other regionalports. Export clearance is not regarded as a problem for investors, mainly because of the low amount of Djibouti exports (mostly re-exports), and the ongoing government attempts to promote port activities.

Banking

3.92. Investors complain about high costs of capital and lack of access to long-term finance. Chapter 4 provides a detailed assessment of the finance sector challenges in Djibouti.

Public utilities

3.93. The majority of private sector entrepreneurs complain about the high cost of infrastructure services. Chapter 5 provides a detailed account ofthe challenges in the infrastructure services in Djibouti. The costs of electricity, water, telephone, and internet access are higher in Djibouti than in neighboring countries (table 3.19).

Internet Fees (US$/kWh)a (us$/rn3 lb (US$/3 min)' (US$) Industry Household Household Peak Off-peak Setup Monthly Djibouti 0.24 0.20 0.35 0.20 0.20 28.13 56.27 Yemen 0.09 0.06 0.25 0.02 0.01 15.61 14.44 Ethiopia 0.06 0.03 0.13 75.00 34.00 Morocco 0.07 0.10 0.20 0.14 0.12 5.31 Mauritius 0.12 0.18 0.07 0.07 12.06

Sources: a/ USA1D:DjiboutiInvestor's Roadmap, November 2001; Djibouti PER 2004; Mauritius figure is for 1991 b/ USA1D:Djibouti Investor's Roadmap, November 2001 which also states cost to industry at US$ 0.84 per cubic meter; Ethiopia from Djibouti Integrated Framework for Trade Diagnostics; Other countries - Djibouti PER 2004. c/ calculated from ITU World Telecommunication Indicators, (2001). d/ Various sources. Based on monthly usage of 10 hours.

Taxation

3.94. Although the tax structure in Djibouti is fairly simple, investors complain that interaction with the tax authorities is highly complicated. For each type oftax, investors must submit numerous forms with basically the same information and attachments (see annex 3.2). Taxes on labor use are high but the numerous tax incentives offered to foreign firms and the low tax compliance of local firms operating in the informal sector point toward an ambiguous effect on aggregate unemployment.

Contract enforcement

3.95. Property rights are secured by law. However, the formal establishment of property rights has little effect on investment without effective incentives to respect and enforce those rights. A high-quality and reliable judiciary reduces transaction costs for businesses and sends positive signals to investors that the rule of law will be protected and enforced. Of particular concern for businesses are the costs associated with litigation, as well as the long time that it takes to resolve a dispute through the formal judicial system. Investors report that commercial cases can take two to

86 three years to complete. The longer it takes to resolve cases and the lower the resolution rate, the higher the firms’ transaction costs.

3.96. Human and material resources allocated to the justice system are inadequate and judicial independence is lacking. The budget ofthe justice sector is not more than 1 percent ofthe total state budget. There are not enough judges to cope with the volume ofcases; the same is true of court staff. Moreover, the lack of specialization and training adversely affects the quality of the decisions rendered. The credibility of the work is undermined by the frequency with which cases are dismissed, the slow pace ofproceedings, and the frequency with which legally prescribed time limits are not respected. The backlog of cases contributes to partiality among judges. Although improvements have been noted, such as the recent provision of facilities and better salaries for judges, much remains to be done. There is no stature guaranteeing judges’ independence and impartiality.

3.97. There is a lack of harmonization with international arbitration law. Foreign investors generally prefer international arbitration to pursuing the matter in the local courts for reasons of speed, confidentiality, and impartiality. This method of settling disputes is favored by the international business community. However, Djibouti has opted for its own international arbitration code. International commercial arbitration was established by Act No. 79 of 1984. The 32 articles of this Act describe the terms of arbitration proceedings and provide for the establishment of an arbitration appeals board to rule on rejected or pending appeals. However, it appears that this Act has never been implemented and that the procedure for international commercial arbitration is not operational. Moreover, the treaty establishing the Common Market for Eastern and Southern Ahca (COMESA) has created a Court ofJustice with competence in matters relating to arbitration clauses, although within a limited scope. These initiatives seem to have yielded no results, because offoreign investors’ marked preference for international arbitration procedures.

Business exit

3.98. Cumbersome regulations burden the company liquidation process. The process is time consuming because of a lack of coordination between the various government agencies involved (Ministry ofFinance, Ministry ofJustice, Ministry ofLabor). In addition, record keeping is entirely paper based, and there is no official guide, regularly published and updated, to make the process details, fees, and delays more transparent.

Summing up ...toward an investment climate that promotes firm expansion and job creation

3.99. Transparency and predictability of rules and regulations. Less complex and more transparent regulations coupled with less discretionary power for the bureaucracy will help reduce regulatory uncertainties and bureaucratic red tape. In particular, the Djibouti authorities should aim to reduce the number of permits, simplify procedures, improve coordination between agencies interacting with businesses, and lower administrative barriers to firm formation and growth.

3.100. Contract enforcement. The judicial system needs to be restructured and provided with adequate resources. This entails the preparation and enactment ofa true statute for the judiciary, the recruitment of new judges with advanced training, and the provision of such training to judges already on the bench. There is an urgent need to improve the efficiency ofthe system ofjustice and to accelerate the pace ofjudges’ decisions. This step is necessary to bring down transaction costs linked to random judgments that are often difficult to enforce.

87 3.102. Import and export of goods. A typical clearance should not take more than 24 hours. Achieving this goal will probably require close coordination between the ports and Customs. This goal has been achieved throughout Ahca despite manual procedures. Mechanisms such as pre- clearance, bonding, and insurance can be established, and other steps can be taken to significantly reduce clearance time. Djibouti should also work with customs officials from Ethiopia to help reduce clearance times for goods bound for Ethiopia. A fast-track system will ensure that Djibouti remains the port ofchoice when other ports, including Asmara, come on line. In particular, customs procedures in Djibouti will greatly benefit from measures to streamline and standardize the clearance process, by consolidating the number of entry forms and publishing an official guide for Customs clearance; introduce automation, both in record keeping and in allowing importers/exporters to file electronically; work with COMESA on harmonizing customs procedures; and introduce a formal system for resolution of conflicts between customs and importers/exporters.

3.103. Taxation. The tax registration needs to be simplified. Such streamlining could be undertaken by tax authorities independently, without any change in the tax legislation. The ultimate goal would be to create a one-stop tax registration system whereby an investor interact only with the tax authority, and whereby a single number could be issued for all relevant taxes.

3.104. Banking. Developing the financial system to improve access to finance and lower the costs of capital will help mobilize savings for investment. Chapter 4 provides a detailed assessment ofthe main challenges.

3.105. Public utilities. Improving the institutional environment to better provide infrastructure services by completing regulatory reforms and improving the institutional capacity, transparency, and management of infrastructure service providers. Chapter 5 provides a detailed assessment of the challenges faced by infrastructure services.

3.106. Company liquidation. The government should attempt to streamline liquidation procedures and reduce documentary requirements. Much of this can be accomplished at the same time as company registration procedures.

3.107. In sum, policies to foster job creation need to go beyond the labor market. Perceptions from entrepreneurs suggest that critical constraints on job creation currently lie outside the labor market. They also mean that labor market regulations are only a part of a broader institutional and policy framework that constrains firms’ ability to grow and create new jobs. This has important implications for policies to foster employment and reduce unemployment. They cannot be limited to improving the workings of the labor market; instead they should aim to improve the overall investment climate and focus on the most severe obstacles to firm growth.

3.3.3. Labor market regulations

3.108. Labor marketpolicies can affect wage and employment adjustment in a variety of ways, including truncating the lower end of the wage distribution (with impacts on the level and structure of employment), affecting labor demand incentives (firm incentives for hiring and firing) and labor supply incentives by promoting (re-)entry to employment for laid-off workers and new entrants. Combined with relevant education and training policies, labor market policies can help upgrade workforce skills and expand the pool oftrainable workers.

88 3.109. The 2005 Labor Code”, introduced greater flexibility in the labor market. It makes it easier to hire workers. It decentralizes wage-setting at the firm level, allowing wage bargaining between the interested parties, and eliminates minimum wage provisions. It also follows international labor conventions to limit child exploitation and labor market discrimination against gender, race or religion (See Box 3.4 for comparisons for the 1952 labor code; and Annex 3.2 for details). In spite of this considerable progress, some rigidities remain. Dismissals are still difficult, full-time permanent employment is the norm, and workers are entitled to a wide range ofworkplace benefits and protection. In addition, the costs of court challenges to dismissals act as an additional disincentive to firing workers.

(4 Wage bargaining

3.1 10. The 2005 Labor Code strengthens the role of socialpartners in the wage setting process and wage setting in the formal sector, entirely centralized and controlled by the government, with a weak role for socialpartners. In the past, the government was a pervasive force in formal sector wage determination, directly and indirectly (through the wage pull of the state-owned enterprise (SOE) sector). The formal private sector closely followed the public sector’s lead in wage determination. Formal sector employers (state-owned enterprises) placed retention above productivity gains, commonly mimicking public sector wages. As a result, they often remained price takers fi-om government with respect to wages. At the same time, a large informal sector (estimated to account for more than 20 percent of employment) has provided very flexible wage determination. The new labor code strengthens the role of social partners in the wage setting process. It decentralizes wage-setting at the firm level, allowing wage bargaining between the interested parties. It also allows the establishment ofprivate employment recruiting agencies.

(iv) Employment protection legislation

3.1 11. Employment protection legislation is a key labor market institution for ensuring secure and decent work. It is important to balance the positive impact on employed workers with the needs ofthe unemployed and those entering the labor market, whose prospects ofbeing hired may be negatively affected by the strict job security provisions in Djibouti’s employment protection legislation. From an economic viewpoint, there is also a balance to be struck between the positive features of strict employment protection legislation (such as incentives to provide productivity- enhancing training to workers) and the need for sufficient ease ofjob turnover to promote efficient allocation oflabor and contribute to macro-level productivity improvements.

3.1 12. Past rigidities in employmentprotection legislation entailed significant costs to Djibouti’s economy. The adoption of the new labor code, that relaxes employment protection legislation, is a step in the right direction. First, strict employment protection legislation fosters informality, self-employment, and small firms. Figure 3.20 shows a positive correlation between the strictness of employment protection legislation and the share of the informal sector in total employment. A similar positive association holds between the strictness of such legislation and the share of self- employment. Cross-country regression analysis corroborates a number of these effects (World

** The new labor code was finalized in 2005. It was officially adopted by presidential decree in March 2006. A new feature of the labor code (the extension of the legal working hours from 40 to 48 at unchanged salary level) was proposed by the Djibouti authorities in early 2006 but it has not yet been implemented. Negotiations ofthe collective agreements to implement the new labor code started in some branches such as the oil, banking and transit sectors. The first implementing decrees (2 out of 30) ofthe new labor code are under preparation and relate to youth employment, which is the main preoccupation ofthe authorities.

89 Development Report, 2005). Second, greater job security for those with jobs appears to come at a cost for the unemployed and new entrants in the labor market, so that (re)entry into formal sector employment is negatively affected. This insider/outsider impact on formal sector employment rates would appear to undermine the Djibouti authorities’ claim that strict employment protection legislation is good for workers in the aggregate. The high protection of permanent workers may have contributed to the observed duality in the labor market (formal versus informal sector).

I 80 I I

Source: World Development Report, 2005

3.1 13. The 2005 Labor Code represents a considerable improvement over the 1952 labor code along a number of dimensions:

It introduces greater flexibility in the labor market. Greater flexibility in hiring workers, as regards salary and benefits, contract duration (open-ended, temporary and apprenticeship contracts); training duration, and overtime hours, provides incentives to firms to hire worker. The 2005 Labor Code simplifies the recruitment procedures by allowing workers to be hired without prior government authorization. It also allows the establishment ofprivate employment recruiting agencies.

Itfollows international conventions against discrimination in the labor market along gender, religious and race lines, and against child exploitation. The new code stipulates equal pay for equal jobs irrespective of gender, race or religion. It also stipulates maternity leave rights to promote women’s participation in the labor market. It also strengthens women’s rights to participate in trade unions. The new code also stipulates minimum age restrictions. The minimum age restriction (16 years of age) is also welcome so as to limit child exploitation.

It abolishes minimum wage restrictions. The minimum wage is an important element of wage determination, particularly given its possible impact at the low end of the wage distribution in providing a floor to wages and hence a constraint on hiring low-skilled workers. The monthly minimum wage for unslulled workers was set by law at US$106, representing more than 56 percent of the average wage for unslulled workers. This ratio placed Djibouti at the high end of the range of minimum wages by international standards. Average wages mask considerable wage dispersion across the formal and informal sector. Most employers in the formal sector pay well above the minimum wage. However, there is evidence of weak enforcement of minimum wages in the informal sector. More than 76 percent of wage earners in the informal sector earn

90 less than the minimum wage. Employers within the free zone are authorized by law to pay less than minimum wage, offer less annual leave than the rest of the country, and demand a longer work schedule.

3.114. In spite of theprogress achieved with these reforms, a number of restrictions on hiring and firing workers remain. Remaining restrictions on hiring include: impediments on contract types and contracts of foreign workers; limits on the duration of fixed-term contracts; restrictions on paid overtime work; and restrictions on the length of apprenticeships. While the new labor code simplifies the dismissal process and follows international practices (i.e. requesting 2 months of salary in advance). However, a number of firing restrictions remain. Employers are required to notify the Ministry of Labor before undertahng economic or technically induced terminations. Approval from the Ministry ofLabor is required for dismissals of 10 or more employees. The long period ofpre-notification for impending dismissals, can also discourage firms from firing workers. Apprenticeship contract terminations are also subject to prior approval by the Ministry of Labor. Finally, the 2005 Labor Code’s requirement subjecting employers to re-hire first those employees laid off during a crisis for economic and technical reasons-although helpful in enhancing job security-takes away the employers’ right to hire whomever they want based on market conditions. In addition, the costs of court challenges to dismissals act as an additional disincentive to firing workers.

3.1 15. In sum, dismissals are still difficult, full-time permanent employment is the norm, and workers are entitled to a wide range of workplace benefits and protection. Employment protection legislation is more strongly enforced in the formal sector. In the informal sector, employer surveys show weak enforcement. This may have contributed to the observed duality in Djibouti’s labor market in which vulnerable groups, including the youth, women, and the unskilled, churn from one job to another in the informal sector, while those in the formal sector who maintain a regular contract still enjoy a significant degree ofjob protection.

3.116. Non-wage costs of labor remain high. The financial obligations of employers under the 2005 Labor Code can be grouped in two areas: one concerns statutory leaves (annual, sick, maternity, and death), and the other relates to payroll taxes (taxes on labor have been already discussed earlier). Although the Code stipulates the legal worlung hours of 48 hours per week for each employee, it also provides for generous paid leave benefits, which negatively affects labor productivity (by reducing effective working hours). Some of these benefits are welcome (such as maternity leave that could improve women’s participation in the labor market). However, the sum of all these paid days (annual, other paid and statutory leave) has been estimated at some 50 days per year for permanent workers in the formal private sector. This amount could become quite onerous. Pension contributions by employers to employees’ retirement are also significant. Many entrepreneurs are required to pay 20 percent ofpayroll taxes (16 percent ofemployers’ contribution and 4 percent ofworkers’ contribution) to the pension fund.

91 Box 3.4 Djibouti: Labor market regulatory reforms

Until recently, the labor market was governed by the 1952 Labor Code, whose rules were primarily conceived for a colonial economy. Its main objective was to protect the French expatriates working in Djibouti. Realizing that many provisions of the 1952 labor code had become obsolete, the Djibouti authorities issued a new Labor market Act in 1997 (Loi 140, 1997).

The I952 Labor Law was considered obsolete and extremely rigid by domestic and foreign investors in Djibouti. Specifically, private investors pointed to: the high cost and inefficiency of local labor as a consequence of the inappropriate wage guarantee, severance pay and suspension of employment provisions (local labor is viewed as overly protected) the lack of a clear practical definition of skilled and unskilled workers the obligation to recruit local labor and the extensive procedural complications involved in terminating employees’ contracts even if for justified cause; the inadequate higher education system, which does not prepare the local workforce for the modern technical and managerial demands of the private sector employment marketplace the lack of consideration for the views of the Djiboutian Association of Employers in matters involving the revisions to the Labor Law.

The Labor Act No. 140 of 1997 was ajirst, and perhaps decisive, step toward liberalizing the labor market. Under this Act contracts are the result of a market mechanism operating between the supply and demand of labor. The Act also makes the conditions for hiring workers less stringent and abolishes direct intervention by the state. The second step, which was more ambitious, was the redrafting of the Labor Code. The new code, adopted in 2004 goes further in the effort to liberalize the labor market by emphasizing the modalities of fixed-term labor contracts, dismissals for economic reasons, notice, and the settlement of collective disputes.

The 2005 Labor Code represents a considerable improvement over the older labor legislatwn. It makes it easier to hire workers. It simplifies the recruitment procedures by allowing workers to be hired without prior government authorization. It introduces greater labor market flexibility. It decentralizes wage-setting at the firm level, allowing wage bargaining between the interested parties. It allows the establishment of private employment recruiting agencies. It also eliminates minimum wage provisions. Its provisions against child exploitation (setting the minimum age at 16 years of age) and against gender, race and religion discrimination are also welcome (Source: Loi N.l33/AN/05/5eme L Portant sur le Code du Travail, 2005; Note sure les apports du nouveau Code du Travail, ). For more details see Annex 3.2.

Source: Ministdre de I’Emploi et de la Solidaritd Sociale, ZOOS

(iii) Taxes on labor

3.1 17. Taxes. on labor may affect equilibrium wages, which in turn may affect equilibrium employment. The extent of the impact is affected by the level of competition in the labor market. In particular, the impact of high taxes on labor use depends on the extent to which there is a tax shift of labor taxation onto the cost of labor (and hence labor demand) as opposed to onto take-home pay (and hence labor supply).

3.1 18. Djibouti’s tax wedge is relatively high, at 34.3 percen?’ (compared with 17.3 percent in Turkey and 14.1 percent in Korea). In other words, the difference between a worker’s take-home pay and what it costs to employ himher in Djibouti is too great. The high tax wedge is partly responsible for long-term unemployment. Djibouti’s corporate income tax rate (at about 25 percent) is not very high. However, income taxes make up only a small sliver of the overall tax wedge, while payroll taxes make up the greatest portion.

3.1 19. Payroll taxes and other non-wage costs Vringe benefits) make the overallprice of labor higher than in neighboring countries. Social security contributions are retained at the source and total about 20 percent ofan employee’s salary, ofwhich the employee contributes 4 percent and the employer contributes 16 percent. This may lower the incentive, particularly among domestic employers, to offer jobs to low-income workers. By law, foreign firms operating in the Free Zone

29 The tax wedge on labor is defined as (labor cost - net pay) / labor cost . 100. Labor cost comprises gross wage plus employer payroll contributions on wage. Net pay is calculated as gross wage less employee payroll contributions and personal income tax.

92 are exempt from the government’s social security and medical insurance programs. Coupled with observed low collection rates of taxes on labor (about 46 percent), this implies that payroll taxes and social security contributions may not impose a severe burden on all employers. Hence their role in discouraging aggregate labor demand seems ambiguous.

3.3.4. Weak role of passive and active labor programs for the unemployed

3.120. High and persistent unemployment (most likely masking the presence of large underemployment) make a strong case for passive and active labor programs. The case is reinforced on equity grounds, given the significantly high poverty rates among households with unemployed heads. In 1996, 68 percent of households with unemployed heads were below the absolute poverty line compared with 44 percent of households with employed heads. The mean annual household expenditure in 1996 for households with an unemployed head was US$4,370- 3 8 percent lower than for households with an employed head.

Social security

3.12 1. Appropriate social security arrangements can play an important role in workers ’ risk management and can also affect labor market outcomes. By imposing additional costs and various constraints on employment relationships, social security arrangements influence the working of the labor market in Djibouti. The pension system and health insurance mechanisms affect the labor market directly (through longer-term incentives for older workers to increase labor force participation) and indirectly (through the large tax burden on labor imposed by social insurance programs.

3.122. The segmentation evident in Djibouti’s labor market is mirrored in differing social security arrangements, with civil servants and employees of state-owned enterprises and a few large private firms enjoying much more robust social security arrangements than workers in smaller formal private sector firms and in the informal sector. This exacerbates differentials in wages across the sub sectors, reinforcing the relative attraction of public sector employment in terms oftotal compensation.

3.123. Only workers in the public and private sectors are covered by the social security system. Pension and health insurance benefits of formal sector private employees, state-owned enterprise employees, and civil servants are provided by the OPS. The military and security forces are covered by a separate pension fund: the Military Social Fund (Caisse Militaire de Retruite).The uncovered population includes the self-employed and wage earners in the informal sector). These groups rely entirely on self-insurance and traditional social support networks to provide income smoothing and old age security. Social security for the informal sector workers remains an objective that Djibouti’s public policy makers need to consider soon.

3.124. The newly created Conseil National de Securite‘ Social (CNSS) is the governing body of the social insurance system. The CNSS houses the OPS and the CNR. The OPS provides health insurance, insurance against work-related accidents, family allowances, and old age, disability, and survivorship pensions to private sector employees and contractual workers in the public sector. The CNR provides pension benefits only to civil servants, parliamentarians, and the police. Coverage is limited to 15 percent of the labor force, and benefits are received by less than 11 percent of the population over age 50. Under the OPS, workers are given the option to retire after 55 years with less than 25 years of service. The aging of Djibouti’s population over the long run is expected to contribute to the deterioration of the fund’s financial position. The fund is already burdened with

93 high administrative costs, accumulated arrears (amounting to 8 percent of GDP), and an overgenerous benefit system.

3.125. The social security system in Djibouti faces mounting fiscal pressure from its pension scheme, particularly from civil service schemes, which are funded from general revenues. The recent parametric reform of the pension system has helped contain the contingent liabilities and fiscal pressures associated with the pension schemes. However, a number of remaining features reveal mounting fiscal pressures: the CMR is still in deficit, and though the OPS has managed to balance its accounts the low contribution rates-particularly from private sector employers-are worrisome. In addition, the significant use of lump-sum payments at retirement and the relatively low retirement age of55 help increase fiscal pressures and can encourage early exit from the formal labor force. Although not solely to blame, the fiscal pressures of the civil service pension system act as an additional squeeze on public investments in human capital. An important priority should be to get better estimates ofunfunded fiscal liabilities associated with the social security system.

Social assistance

3.126. Social assistance programs can play an important role in labor market adjustment. To the extent that they work effectively, they help protect workers who have lost jobs and exhausted other entitlements before finding a new job. A well-designed and adequately implemented social assistance system can ensure adequate work incentives and encourage productivity-enhancing labor turnover.

3.127. Djibouti is spending about 4percent of GDP in budget lines that are usually classified as assistance programs. From these, however, only 0.5 percent of GDP goes to programs benefiting the poor and vulnerable. Expenditures in cash transfers to households and institutions are highly regressive, and a review of current allocations is urgently needed. In 2003 the government spent US$26.9 million in transfers to household and institutions (3.14 percent of GDP). Of these, however, 40 percent represented transfers to various parts ofthe public administration (such as the National Assembly) and to finance Djibouti’s involvement in the peace process between Eritrea and Ethiopia. Housing subsidies for civil servants (36 percent ofthe total, or 1.14 percent ofGDP) and fellowships to study abroad (15 percent) captured an additional 50 percent. These last two items probably benefit individuals in the top 10 percent of the income distribution. Hence, transfers focusing on poor and population groups represented only 10 percent of the total or 0.2 percent of GDP (0.5 percent is expected for 2004). In 2003 half these transfers were pensions for demobilized soldiers and disabled war veterans. In 2004, under the PRSP, DF 230 million (US$1.3 million, or 0.2 percent of GDP) has been earmarked for poverty reduction activities and the development of a safety net, but specific interventions still need to be defined.

Active labor market programs

3.128. Active labor programs can strengthen the matching process in the labor market but in Djibouti the performance of these programs has been mixed. Active labor market programs include job placement, training and retraining programs, public works, and self-employment support. Public expenditure on active labor market programs in Djibouti has been a negligible share of GDP. It is important to stress that active labor market programs achieve far less when labor demand is depressed, and they are not a substitute for improvements in the business environment. Even in conducive circumstances, they are only part of a menu of policies to support employment growth. Nonetheless, though active labor market programs alone may not raise aggregate employment, redistributing job opportunities among different groups is also an important objective in Djibouti because ofthe stagnant pool oflong-term unemployed (or underemployed) workers.

94 3.129. Job placement programs. The effectiveness of job placement programs in Djibouti has been limited, owing to the growing role of the informal economy. Informal networks and private employment agencies seem to play a stronger role in finding employment than public employment agencies. Part of the reason lies in the limited institutional capacity of Djibouti’s public employment agency.

3.130. Vocational training programs. Training workers who have only primary education has tended to have a positive net employment impact, particularly for the low-skilled, though the impact on earnings is mixed. In Djibouti, vocational training institutions are mainly in the hands of the public sector, with only marginal participation by private centers, and the Vocational Training System suffers from a number of critical weaknesses. Training programs managed by the Ministry of Labor and Social Protection capture 0.2 percent of total public expenditures, or less than 0.1 percent or GDP. As is often the case, the vocational training system has been a safety net for students who drop out of school as opposed to training adults transiting between jobs or upgrading or acquiring new slulls. A recent assessment of the Vocational Training System (VTS) carried out by the Djibouti PER points to the following weaknesses: overemphasis on initial training relative to continuous training, undeveloped on-the-job training activities, weak monitoring and evaluation systems, ill-equipped and poorly financed training centers, and lack of involvement by the private sector, resulting in a disconnect between slulls supplied and the demands ofthe labor market.

3.13 1. Public works programs. Public works can act as effective safety net interventions. These programs have two major benefits: a transfer benefit to the beneficiaries, which is in effect the wage the participants get, and a stabilization benefit, the consumption smoothing that wage affords against temporary income shortfalls.

3.132. The main issues related to the public works programs in Djibouti are the extent and cost of benefits and whether they reach the intended participants. The way these programs are designed and implemented is critical to their success. Impact evaluations of public works projects financed by the Agence Djiboutienne d’ExCcution de Travaux d’IntCr& Public (ADETIP) show mixed results owing to a variety of design and implementation issues. Some of the reported implementation problems include delays in releasing funds, capacity constraints of the program administrators, and less employment creation than was targeted. The high share of wages in total costs (ranging between 30 and 40 percent) makes this kind of public works project an expensive way to channel benefits to the poorest. However, the infrastructure created is still an important component of the overall value and justification ofthese public works programs.

3.133. Djibouti’s public work effort has contributed marginally to generating employment. Between 2000 and 2003 the public works agency, ADETIP, implemented projects amounting to DF 1.4 billion. Though 83 percent of expenditures were associated with labor costs, the 114,000 worker-days generated during 2000-03 represent only a small fraction of the supply. Indeed, with an unemployed population estimated at 83,000 individuals in 2002, there would be a supply of at least 334,000 hours per day on average (assuming that the unemployed want to work only half time-4 hours). Yet in 2002 ADETIP generated an average of 1,136 hours per day, or less than 1 percent ofthe total supply.

3.134. Moreover, the average day of work is remunerated generously, suggesting that those who benejit are not in the lowest quintiles of the income distribution. Information about the distribution ofwages within and across projects was not available. The director general ofADETIP reports, however, that wages range between DF 1,500 and DF 3,500 per day (US$8.50-20.00 per day), or three to six times the relative poverty line (DF 1,225 per day, or US$3.36 per day). Yet

95 total labor expenditures divided by the total number of days of work created through ADETIP projects gives an average daily cost of labor of DF 10,000 or US$7 per hour.

3.3.5. The supply-side: the poor quality of labor

3.135. The aggregate level of human capital in an economy’s labor force is a major determinant of economic growth. Education and training of workers-as a fundamental form ofhuman capital-is a key source of labor productivity.

3.136. The key supply issue in Djibouti is not the quantity of labor but its quality. As indicated by the population growth rate (2.8 percent) and the high degree of urbanization (more than 80 percent of the population lives in the capital city), labor is abundant-particularly the supply of younger workers.

3.137. Low educational achievement combined with fragile health indicators and the consumption of khat are key factors behind the low productivity of the labor force. Djibouti has an education deficit and higher infant and maternal mortality rates than Sub-Saharan African countries and other countries at similar income levels. Despite progress in reducing this deficit, considerable challenges remain if Djibouti is to meet the MDGs by 2015.

3.138. Primary enrollment rates are low (the net rate is less than 45 percent) and access to upper basic and secondary education is still limited for various reasons, resulting in a gross enrollment rate of 9 percent and 4.4 percent, respectively. Out of every 100 boys and 100 girls in first grade, 80 boys and 79 girls reach the sixth grade (the end of the primary cycle). Only 23 boys and 26 girls reach the tenth grade; and just 7 boys and 11 girls reach the final year ofthe secondary cycle. Given the high level of dropout and repetition in the three education cycles and assuming the pattern of survival rate and intake rate persists over time, in future men and women will have, on average, only 3.9 and 3.3 years of formal schooling as a part of their human capital profile. These results have a significant negative impact on the quality of labor.

3.139. In addition, the considerable amount of time and resources devoted to chewing khat also affects the quality of labor. Many adult males in Djibouti chew khat in all-male gatherings for 5 to 6 hours a day, talung time from work. Lax work ethics and frequent absenteeism at the workplace are also factors that negatively affect the productivity ofDjibouti workers.

3.140. Improving the education base of the workforce has to be at the top of the agenda for enhancing labor productivity and stimulating labor demand. The education and training policy reform agenda has three main building blocks: expanding the pool of trainable workers; promoting the upgrading of worker slulls in the workplace; and developing a pool of highly skilled professionals linked to growth-promoting activities.

Expanding the pool of trainable workers

3.141. Basic education is a pre-requisite for producing trainable workers. A critical goal for the medium term is to meet the education MDG by 2015. As a result of current low levels of primary enrollment and high dropouts in secondary education, the average Djibouti worker has about half the years of schooling that hisher counterpart in other comparable African countries has.

3.142. The recent national education reform has been put into action along the lines of the recommendations made above. The government has set forth an ambitious but critical education

96 reform agenda: achieve universal primary education (MDG goal met by 2015), increase the number ofyears ofschooling ofthe labor force, and improve the quality of schooling. Improving education quality will require instituting stronger ,accountability for school staff and teachers, training teachers, and increasing the quantity and availability ofteaching materials.

3.143. The education reform agenda brings about financial and fiscal considerations. Both the government, and to a lesser degree, households play a critical role in financing education. Estimated total spending on public education averaged 6.3 percent of GDP over the past three years, much higher than in Djibouti’s neighboring countries (the Figure is 2.7 percent of GDP in Ethiopia). The share of education in the state budget is about 20 percent in the past few years, higher than in other countries at similar income levels. However, a deeper analysis carried out by the recent Djibouti Public Expenditure Review (2004) revealed the highly skewed allocation of expenditure toward wages and salaries at all levels of the education system. Teacher salary levels are high, at about US$650 per month, or about seven to eight times per capita income, as compared with five times per capita income in Kenya and Burundi. The share of recurrent expenditures (wages and salaries) accounts for more than 85 percent of the total. For the education system as a whole, personnel-related expenditures (wages, salaries, and other allowances) account currently for about 90 percent of recurrent spending, with only 10 percent allocated to goods and services (including textbooks, teaching materials, routine school maintenance) and 4 percent for bursaries abroad.

3.144. Djibouti faces a dilemma forfinancing its vision of education for all. On the one hand, satisfying the unmet demand for universal primary education calls for substantially increasing human capacity (recruitment ofteachers) and physical capacity (classrooms) to gradually enroll the 50 percent of. children currently out of school. On the other hand, given current internal inefficiencies and high unit wage costs, this type of push would absorb about half the national budget. Therefore, a scenario based on current trends and a business as usual attitude is simply not fiscally viable.

3.145. The government is well aware of the high unit labor costs and inefficiencies in education spending. The Ministry ofEducation has recently initiated a reform in the primary education cycle aiming to contain unit costs and improve efficiency. It consists of shortening the primary cycle by one year (from six to five) and implementing a nine-year compulsory basic education cycle. This action alone is expected to gradually lead to internal efficiency gains (less repetitions and dropouts) ofat least 50 percent (from 15 percent to 6 percent) over the next four years. Further cost reduction measures proposed by the Ministry ofEducation include the following:

Primary school level: Contain the increase in the number of teachers by a better use of those on board, systematically eliminating any potential unproductive employees and those on unjustified extended leave from the combined personnel and payroll files. Gradually increase the average class size from 30 to 45 by 2008. Maximize the use of teachers specializing in French (rather than Arabic) and thus contain their number (from the introduction of the Arabic language and Islamic teaching starting in the third year of the new nine-year fundamental cycle). Middle and secondary school levels: The current class size (pedagogic divisions) of 50 is already high because of the limited number of such schools, and therefore it cannot be further increased. Several middle schools are scheduled to be built in the next few years to expand the system while improving quality. The government’s reform agenda proposes to increase the minimum ofnumber ofhours that teachers work per week. For middle school, service hours would go up from 21 to 23, leading to 22 hours effectively taught. At the

97 secondary level, the number of teaching hours per week would also grow, from 18 to 20. Ths measure will help slow down the rate ofhiring new teachers. Higher education level: With the increased autonomy ofthe Pole Universitaire de Djibouti (PUD), the prospects for unit cost containment are better for higher education. First, the number of students locally trained is rapidly increasing, with a corresponding reduction in the number of bursaries abroad. Second, the recent construction of larger amphitheaters allows for a doubling ofclass sizes from 50 to 100 students for most subjects. Finally, more autonomy for PUD translates to reduced tuition and fees currently paid to French universities for partnership in registration and for videoconferencing.

Cost of school construction and need for maintenance: Competitive bidding will help reduce average unit costs. Further cost-cutting measures are envisaged with regard to differentiated designs by region (urban, rural), building styles, and physical requirements of the nine-year fundamental education cycle. A new maintenance strategy will be implemented with communities, for added commitment to and ownership of school facilities. Availability of textbooks: The lack of textbooks is an impediment to disadvantaged students and deters demand for education by poor households. The government plans to develop and produce textbooks locally and introduce cost-recovery measures for middle schools, while keeping textbooks free for students in the primary cycle. The average textbook unit cost could drop sharply to DF 300 compared with about DF 3,000 for imports, leading to budgetary savings of at least DF 90 millionper year.

3.146. Even full implementation of these measures will not be enough to achieve universal primary education by 2015 without wage adjustment. Recent World Bank estimates show that the full implementation of the measures indicated above could help achieve 67 enrollment targets by 2008 (Djibouti School Access and Improvement Program, Phase 11, 2004). It is unlikely that the 100 percent enrollment target can be attained by 20 15 without adjusting wages. Lower public sector wages, combined with other measures to restructure the civil service, would help create the fiscal space for building new classrooms, hiring and training additional teachers, and freeing resources for nonwage expenditures that will be needed to foster demand for education. Chapter 2 explored alternative wage bill reform scenarios (for reforming the current government pay and employment policies) to meet the education MDG while maintaining fiscal sustainability.

Upgrading worker skills: revamping vocational training policies

3.147. The secondpillar of the education and training reform agenda should be develop apool of highly skilled professionals linked to the port activities. New activities related to the port and the creation of a free trade zone could generate new jobs. Unfortunately, the majority of the unemployed are unshlled workers, who often will have to compete with more qualified workers from other counties. The efficient use of public resources then requires finding mechanisms to deliver this training that are reasonably cost-effective.

3.4. CONCLUSION AND POLICY IMPLICATIONS

3.148. Djibouti’s labor market offers limited employment opportunities in the formal private sector. Labor force participation is low and unemployment has been extremely high (particularly long-term unemployment). The unemployment rate has increased. The worrisome feature of

98 Djibouti's labor market is the shlls gap. Young, low-educated persons, especially women, face the highest risk of unemployment (or underemployment). There is also a pool of young and educated individuals who are currently unemployed because they have a higher wage reservation (or excessive wage expectations) than foreign workers from neighboring countries. Persistently high levels of unemployment would normally reduce wage levels, absorbing some of the unemployed. Instead, the high level ofwages paid by the public sector apparently has educated youth queuing for public sector jobs. Public sector wages and other labor market restrictions limit any downward adjustment in private sector formal wages, giving rise to the dual labor market and high unemployment rates.

3.149. Given the extremely limited size of its domestic market, Djibouti's long-term future depends on increasing its competitiveness. Because of the country's location, level of infrastructure, limited natural resources and lack of arable land, the service sector is the main source of growth-with particular focus on the Ethiopian market. In this regard, labor costs (both wages and social security payments) represent a major determinant of the competitiveness of Djibouti's private sector. Any strategy to increase competitiveness will need to concentrate on two key elements: reducing wages paid by the formal (especially public) sector, and implementing policies to facilitate entry in the labor market and creation ofnew jobs.

3.150. The gap between public and private wages and between wages of national and foreign workers is attributable to several factors, including differences in protective labor legislation and limits on firing workers, large nonwage benefits (which represent up to one-half of total compensation), and the missing link between public compensation and prod~ctivity.~'All these compensation features attract job seekers to the public sector. Private sector wages in the transport sector and in the informal sector are lower because they rely on foreign laborers, who are not covered by social protection legislation and benefits, and whose lower reservation wage reflects conditions in their home countries.

3.15 1. The result is a high degree of labor market segmentation, a large public employment roll and wage bill, and a private sector dominated by small, informalFrms and a few large firms that import foreign labor. This leads to the excess supply of labor for the public sector and the disproportionate private sector demand for foreign workers. To move toward a well-functioning labor market that is sufficiently flexible to absorb a greater number ofjob seekers in Djibouti (with ftlctiofial unemployment caused only by skill mismatch and job search) and to equate labor demand and supply at a market-clearing wage, wage distortions need to be reduced as much as possible.

3.152. Addressing wage distortions calls for more flexible labor market policies, adjustments in public sector pay policies and better quality in the labor force. The channels in the public sector include changes in labor market regulation and payroll taxes, cutting wages directly for new entrants or indirectly, reducing nonwage remuneration. In addition, education and training systems need to be strengthened to expand the pool oftrainable workers. All these public sector reforms aim at correcting price signals. These actions need to be complemented with labor market and education policies that ensure more efficient allocation oflabor.

3.153. Strengthening the role of passive and active labor market programs. These programs are less developed in Djibouti than in other developing countries at similar levels of economic development. Given the country's limited administrative capacity, large underreporting, and limited

30 Additional factors have distorted labor costs in the past, such as the requirement that public and private employers finance the military demobilization ofDjiboutians (taxe patriotique) during most ofthe 1990s.

99 financial resources, the donor community has to support the government’s efforts to develop effective social safety nets. Moreover, given the widespread informality, the need to reinforce social protection for those who lack access to formal insurance mechanisms should be considered. Particular attention should be paid to targeting female workers in the informal sector. Measures could include greater use of cash transfers and job placement schemes that are open to all those willing to work at a program wage rate-which is set at a level low enough to ensure self-targeting ofthe most needy.

3.154. Developing statistical capacity. Existing data sources are insufficient to provide solid and consistent evidence on labor market outcomes and their trends. Improving the statistical basis and deepening interpretation could be a useful start. Examples include carrying out labor market surveys to better understand youth and female participation and firm-level surveys to better understand the causes of weak labor demand. Finally, there is a need to better understand the political economy oflabor market reforms and, in particular, the political feasibility of government pay and employment reforms.

3.155. Improving labor market outcomes will also entail policy reforms to improve the investment climate and remove a wide range of constraints on business operation. Policies to foster job creation need to go beyond the labor market. Perceptions from entrepreneurs suggest that critical constraints on job creation currently lie ,outside the labor market. It also means that labor market regulations are just part of a broader institutional and policy framework that constrains firms’ ability to grow and create new jobs. This has important implications for policies to foster employment and reduce unemployment. They cannot be limited to improving the worlungs of the labor market; instead they should aim at improving the overall investment climate and focus on the most severe obstacles to firm growth.

3.156. The key message from this chapter is that maintaining the status quo, without significant progress in key structural reforms, will most likely lead to a modest rate of economic growth that will do little to alleviate current labor market pressures. Malung serious inroads into reducing unemployment and low labor force participation will require substantial investments in education and training and instituting structural reforms conducive to a friendly business environment to boost labor demand. The focus of this report is the employment and investment generated by the new private sector. For our purpose, the service sector employment rate is considered a better indicator ofjob creation potential than the total employment rate. Given the large size ofthe public sector, the total employment rate could be misleading, because it could reflect further increases in public employment with no changes in the business environment.

100 CHAPTER 4. MOBILIZING FINANCE FOR INVESTMENT

4.1. Financial development contributes to private investment and economic growth. A well- functioning financial system promotes economic growth by increasing the amount of financial savings through mobilizing financial resources; channeling these resources to the most profitable users, thus enabling an efficient allocation of resources; and ensuring efficient operation of financial intermediaries by enabling them to operate at the lowest costs. Furthermore, a sound financial system benefits growth indirectly by ensuring a better transmission ofmonetary signals.

4.2. In recent years, the Djibouti authorities have introduced financial sector reforms that have improved banking supervision, but the country ’sfinancial sector remains underdeveloped 31 Financial reforms introduced in 2005 have improved the legal framework for banking supervision, including regulation ofmicrofinance institutions and increased supervision ofmoney changers (that are now subject to annual controls). In spite of these reforms, Djibouti’s financial sector remains underdeveloped. The share of the financial sector in GDP remained at about 11 percent between 1990 and 2004,3’ while the assets of the banking system stagnated at about 11 percent of GDP. During this period, the size of the financial sector dwindled as the number of commercial banks declined from five to three, and the number ofinsurance companies declined from eight to two. At the end of 2005, Djibouti had 11 money changers, 2 active commercial banks, and two new foreign-owned banks (their respective shareholders are from Malaysia and Yemen) that have recently been established in the country and will be operational by the end of 2006. Djibouti’s banlung system is dominated by foreign-owned banks. Financial intermediation is still hindered by the lack of diversification in institutions, financial products, lending practices (which aim to minimize risk), and a weak judicial structure to enforce contractual and property rights.

4.3. This chapter assesses the role of Djibouti’s financial system in promoting private investment and growth and outlines reform options to improve its effectiveness. The first section reviews the currency board arrangement and its role in macroeconomic developments. The second section discusses Djibouti’s financial system and reviews recent monetary developments. The third section discusses the main financial sector reforms introduced since 1990 and explains why their impact has been limited so far. The last section suggests measures that could help improve the efficiency of financial intermediation and enhance the financial sector’s contribution to economic growth.

4.1. THE CURRENCY BOARD ARRANGEMENT

4.4. Djibouti has had a currency board arrangement (CBA) since 1949, when it still was a French territory. The CBA operates in a pure form, and maintains foreign reserves in excess ofthe value ofthe monetary liabilities issued by the Central Bank of Djibouti (BCD). The Central Bank does not function as a lender oflast resort, nor does it extend credit to the government, the banking system, or domestic companies (box 4.1). Since its creation in March 1949, the Djibouti franc has been pegged to the U.S. dollar. It was revalued twice, in 1971 and 1973. As a result, the exchange rate is now DF 177.721 for one US. dollar, down from DF 214.392 before 1971.

31 However, the development of the financial system is itself endogenous to growth, thus the lack ofdevelopment in not necessarily a cause of low economic growth. 32 The share of the financial sector is measured by the ratio of value added generated by banking and insurance activities to GDP.

101 4.5. The CBA has enabled Djibouti to maintain full convertibility into the US dollar at a fied exchange rate despite economic uncertainties throughout the 1980s and 1990s. It also allowed the country to emerge, on occasion, as a regional commercial andfinancial center. Over the years, the CBA has lent credibility to the country. While the government usually aimed to maintain the integnty of the system, it occasionally turned away from the budget deficit financing constraints that the CBA required. However, it remained consistent with external monetary developments, including monetary policy set by the U.S. Federal Reserve Bank. Operating under a 100 percent foreign reserve requirement and using rule-bound monetary policy, transparency, and protection fiom political pressure, the CBA has helped Djibouti maintain a relatively strong currency.

Figure 4.1 .a. Djibouti and Trading Partners: CPI Indices, 1990-2004 (2000=100)

__ - - Conposite CPI -Djibouti

~ 6o '1990'1991 '1992'1993'1994'199~'1996'199~'1998' 1999'2000'2001'2002'2003'200~'

Note: The composite CPI is a weighted average of the CPIs of DjiboutiS main tradingpartners (Ethiopia, France, Italy, Saudi Arabia, the UK and Yemen) using their trade share as weights Source: Direction of Trade, WE0 2004.

Figure 4.1 .b. Djibouti and Trading Partners: CPI Indices, 1990-2004 (2000 = 100)

0 ,0

Source: WEO, 2004.

102 4.6. By binding its supply of base money closely to its holdings of reserve assets, the CBA has also served Djibouti well in keeping inflation at a low level, as it has in most countries where such an arrangement has been adopted.33 The inflation, as measured by the consumer price index, declined fi-om about seven percent in 1991 to about 3 percent in 2004.34Djibouti's inflation evolved in line with that of its main trade partners (Figures 4.1 .a and 4.1 .b), with the exception of Yemen, which experienced much higher price increases4specially during 1995-2004, when inflation averaged 18 percent per year.

4.7. However, the hard peg to the dollar did not allow the economy to adjust as easily to internal and external shocks. A number of external shocks affecting the real and financial sectors unfolded in the 1990s. These included: oil shocks affecting the Gulf countries; business cycles in trading partner countries in Europe and in neighboring countries (especially Ethiopia and Yemen); and interest rate shocks in the United States. While other countries in the region, most ofwhich are potential competitors, experienced large exchange rate depreciations, Djibouti's currency appreciated in effective terns (Figure 4.2). The real effective exchange rate appreciated by about 60 percent and interest rates rose by more than five percentage points. These developments led a loss of Djibouti's external competitiveness and a worsening of the trade balance. During the same period, Djibouti also suffered from severe internal shocks. Sticky wages and prices prevented the necessary adjustments to cope with internal and external shocks. Real effective exchange rate misalignment, rigidities in the investment climate and in the labor market hindered foreign investment and economic growth. During the period from 1990 to 2002, foreign investment remained below 1 percent ofGDP though it rose to 5 percent in 2004. Real GDP declined by about half a percent on average per year, and unemployment soared with more than half ofthe labor force out of work, leading to declining per capita income and rising poverty.

Figure 4.2 Djibouti: Exchange Rate Development, 1980-2004 (2000 = 100)

---- 120.0I4O'O 1 -- FEER

Source: WEO, IFS, 2004.

33 Based on a large sample of developing countries, Beaney and Fielding (2002) found that the 52 pegged exchange rate countries averaged inflation rates far lower than those experienced by the 28 flexible-rate countries. The standard deviation of output growth is on average a little higher under pegged rates. The standard deviation of inflation is, however, much higher in the flexible-rate sample. 34 Although the Djibouti franc is anchored to the U.S. dollar, it is worth noting that consumer price indices (CPIs) for Djibouti and the United States have evolved at different rates, with inflation in the United States significantly lower and less volatile from 199H9and slightly higher than in Djibouti thereafter. The difference in inflation trends may reflect the difference in the mix of goods produced and consumed, as well as well demand or supply shocks in the two countries.

103 4.8. While Djibouti policy-makers have maintained adequate official reserves, this is not the only factor that ensures a successful CBA. Djibouti has long maintained a fully open trade system with free capital transactions. Also, the liquidity constraint implied by the lack of a lender of last resort did not pose problems to commercial banks, because, as subsidiaries oflarge foreign banlung groups, they have steady access to international financing3’. Additionally, improved bank supervision has enhanced standards of bank governance and management, adequate loan loss provisioning, and ongoing supervision of balance sheet risks (such as currency and maturity matching). However, maintaining adequate official reserves is only one critical element in preserving investors’ confidence (see box 4.1).

4.9. Macroeconomic and labor market policies have not been always consistent with the maintenance of a fixed exchange rate. In Djibouti, the labor market lacks the flexibility that should accompany a fixed-value domestic currency. Labor costs are high for the region, and they include onerous social security contributions, large severance payments, and other constraints that prevent efficient re-allocation of labor between industries. Furthermore, fiscal discipline was not always observed, as evidenced by the accumulation of large domestic and external arrears during the 1990s. Against background of civil strife that unfolded in the early 199Os, fiscal policy was expansionary. To finance its large deficits, the government accumulated substantial payment arrears.

4.10. Reflecting these developments, the CBA weakened in the 1990s -this trend was reversed since 1998. Between 1994 and 1997, foreign reserves declined, and as a result reserve adequacy ratios36increased by seven percent (from 0.73) and 15 percent (from l.l8), respectively. Foreign currency deposits increased by 19 percent and their share in total deposits increased by about 15 percentage points (from 48 percent). There were also marked private capital outflows, captured in the negative balance ofpayment errors and omissions (averaging 3 percent ofGDP per year). Since 1998, the government stepped up fiscal consolidation efforts that helped reinforced the CBA. In addition, increased transit trade with Ethiopia, resumption ofdonor support, and the military-related revenue windfall generated additional financial sources. Another factor that plays a significant role in maintaining the CBA has always been the deeply rooted respect for the currency board within Djibouti society and the political structure.

Box 4.1 Conditions and Policy Implications for Currency Board Arrangement

A CBA maintains a permanently fixed exchange rate backed by the firm promise of the government of the establishing country to issue or redeem the local currency at a fixed rate against a foreign anchor currency. As a result, the money supply adjusts to the level of official gold and to the inflows and outflows of foreign currency reserves, expanding or accelerating when there are net inflows and contracting or decelerating when there are net outflows. The trigger for these monetary changes is typically price or interest rate differentials with the anchor currency economy (see Enoch and Gulde [ 19971 and Hanke and Schuler [2000], among other studies dealing with CBAs).

Advantages and disadvantages

One of the main advantages of a CBA is that compared with a full-fledged central bank, it is a cost-effective way of managing monetary policy. It is also a strong, “double-barreled” device. Through the currency peg, it indicates a commitment to price stability, while refraining from being a lender of last resort or expanding domestic credit. However, a CBA prevents the government from using monetary and fiscal policy to help offset economic cycles, especially on the downside. It also effectively removes both exchange rates and interest rates as instruments of government policy, surrendering them to the objectives of external exchange rate stability. Thus, the exchange rate is no longer a policy instrument that may be used to stimulate exports when needed, or to limit imported inflation when it threatens to raise domestic interest rates.

35 The only two domestic banks ceased operations in 1999 after they failed to survive liquidity constraints and competition from the two foreign-owned banks. 36 Reserve adequacy ratios are calculated dividing the monetary base (Ml) and broad money (M2) by foreign reserves

104 Key factors for successful currency board operations

A successful CBA is predicated on five conditions. First, the peg should be appropriate and supportive of an acceptable level of external competitiveness. The currency involved in the peg should be appropriate from the point of view of the country’s external trading pattern. Ideally, the reserve currency should be that of the country’s largest trading partner and largest potential source of new foreign investment. Second, there should be adequate flexibility in goods and factor markets (in prices, wages, and economic structure) to keep the system from breaking. Laws and regulations that protect domestic labor and keep labor costs high, including through costly social security contributions and other constraints, prevent the efficient reallocation of labor among industries. Third, there must be a free flow of goods and services, in addition to capital, across international borders to allow the price level to adjust fully to changes in the flows of international reserves. Fourth, a CBA should have adequate financial market supervision, risk assessment, and management so that financial institutions will be able to withstand the consequences of external shocks on interest rates. At the same time, bank supervision should follow best international practices to ensure high standards of bank governance and management, adequate loan loss provisioning, and ongoing supervision of balance sheet risks (such as currency and maturity matching). In this regard, adequate official reserves are only one critical element of the preservation of confidence. Fifth, a CBA should maintain fiscal prudence, which implies a sound fiscal framework that does not require discretionary access to central bank financing by the general government, and maintains public debt within limits of serviceability.

Policy Implications

For a CBA to be successful, the fixed exchange rate must be the sole fixed point surrounded by highly flexible markets. Maintaining the currency board must be the top priority, with other policies designed to support it, not conflict with it. Any government reluctance or inability to conform to either the hard budget constraints or the free market principles that must accompany the adoption of a currency board would compromise the effectiveness ofthe arrangement.

4.2. DJIBOUTI’SFINANCIAL SECTOR

4.11. In addition to the CBA, the Djibouti financial sector consists of two commercial banks and several otherfinancial institutions. At the end of 2005, these other institutions include eleven money changers and two insurance companies dealing in car, business, and shipping-related insurance service^.^' Two new foreign-owned banks (the country origm of their main shareholders are Malaysia and Yemen, respectively) have recently been established in the country, and are expected to be operational by the end of 2006. In spite of these developments, the financial sector remains shallow and inefficient. There is no domestic capital market that could bring together entrepreneurs and investors who would seek placement opportunities and be prepared to take risks.

4.2.1. Main features of the banking sector

4.12. The Djiboutifinancial sector remains confined to providing short-term financial sewices andperforms weakly in comparison with other small states and neighboring countries. Table 4.1 presents comparative data on financial depth in Djibouti relative to that in other small states, neighboring countries (Egypt, Ethiopia, and Yemen), and four other African countries (Ghana, Kenya, Burundi and Rwanda). Djibouti’s performance was weaker than most of its comparators3* on most indicators between 1990 and 2004: it experienced a lower financial intensity as evidenced by lower broad money growth, a broad money-to-money ratio, the growth of credit to the private sector, the growth of total deposits, and the lower share of time deposits in total deposits. During the same period, real GDP growth was estimated at about -0.5 percent in Djibouti, while it reached an average of 3.5 percent in the other small states and even higher levels in the other comparator countries.

37 Eight insurance companies existed before December 2000, when a new regulatory law was introduced. Most of these companies fell short of meeting the new requirements (in terms of technical reserves, assets, liabilities, and tariffs) and were constrained to cease activities. 38 The exception where Djibouti was stronger was in relation to Burundi and Rwanda with respect to the ratio of broad money to GDP and the ratio of time deposits to total deposits. However Djibouti performed poor relative to the two countries in other indicators of financial intensity.

105 4.13. Financial intermediation in Djibouti is fairly weak and dominated by foreign-owned banks. Two commercial banks (Banque Al-Baraka and Banque de Djibouti et du Moyen Orient) closed in the late 1990s owing to continued liquidity problems that endangered their viability. The bankmg sector is now dominated by two large French-owned commercial banks, Banque pour le Commerce et 1’Industrie Mer Rouge (BCI) and Banque Indosuez Mer Rouge (BIS), which account for about 95 percent of total deposits and issue more than 85 percent of total credit. A weakened loan portfolio has limited activities of the third operating bank, the local subsidiary of the Commercial Bank of Ethiopia (CBE), to talung deposits and international transactions by Ethiopian customers. The CBE ceased operations in 2004. By 2005 it completed reimbursement of its deposits and is currently trying to recover debts owed.

4.14. Financial intermediation is costly to those soliciting funds, and lendingpractices aim at minimizing risks. For instance, the banks demand marketable collateral in a country where assets are few and illiquid. Also, the spread between lending rates and deposit rates is high-almost twice that prevailing in Ethiopia and a percentage point higher than spreads in Yemen and in other small states. As a result, the two large commercial banks operate very much like traditional savings banks (caisse d’dpargne), with a focus on collecting deposits and only very limited credit activity. Credit extended to the private sector has been declining because oflimited opportunities in the real sector, deficiencies in accounting standards, and a weak judicial system. Gradually, bank activities have shifted to financial placements with parent companies.

106 Table 4.1 Djibouti: Selected Financial Indicators, 1990-2004 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 ~l~~t-~~~4 'md money growth (annual percentage change)

Small~ ~ states ~~ ~~ 20.2 16.8 15.1 14.7 27.2 21.0 14.9 18.6 10.1 15.7 15.4 13.7 12.1 8.8 12.6 15.8 Djibouti 3.6 1.3 -6.2 -2.2 1.5 3.3 -10.0 -1.4 -2.8 5.2 1.1 7.5 15.7 17.8 13.9 3.2 Egypt 19.7 27.5 14.3 16.4 12.4 11.1 10.5 15.1 8.6 11.5 8.8 11.6 15.6 16.9 13.2 14.2 Ethiopia 18.5 10.7 13.3 18.9 10.7 21.7 10.8 3.4 12.7 5.9 14.0 9.5 12.3 10.4 10.9 12.3 Ghana 13.3 27.2 52.2 45.6 53.2 40.7 39.4 44.3 20.2 23.9 46.5 40.6 50.5 38.1 25.9 37.4 Kenya 20.1 19.6 39.0 28.0 73.8 14.1 15.9 11.9 3.2 5.4 3.9 2.4 10.0 11.5 13.4 18.1 Yemen 27.5 18.8 27.6 35.4 34.7 20.4 8.6 10.7 11.7 13.8 25.1 18.7 18.0 20.0 15.0 20.4 Burundi 14.4 2.33 2.9 7.78 21.6 5.831 23.1 9.42 -3.7 41.9 3.89 16.2 27 22.3 16.7 14.2 Rwanda 5.32 5.68 13.3 3.69 -6.5 73.69 8.2 47.5 -3.9 6.55 14.3 9.21 11.4 15.2 12.1 14.4 atio of broad money to GDP (percent) Small states 47.8 50.4 48.0 48.2 48.5 49.9 51.1 52.4 53.8 55.1 56.4 59.4 61.6 61.0 62.5 53.7 Djibouti 13.2 72.5 65.8 66.0 63.6 64.9 58.8 57.0 54.1 54.6 53.7 55.6 62.3 69.7 74.8 63.1 Egypt 74.7 82.3 75.2 77.5 78.2 74.5 73.2 75.2 73.1 76.1 74.9 79.3 86.8 92.0 89.7 78.9 Ethiopia 28.7 28.7 30.8 28.6 29.8 30.3 29.9 29.8 35.0 34.4 35.1 37.4 44.1 44.8 40.3 33.8 Ghana 12.6 12.8 17.8 19.8 22.5 21.3 20.3 23.5 23.1 24.0 26.7 26.8 31.1 32.0 33.4 23.2 Kenya 20.9 22.0 26.5 27.1 40.7 39.9 41.3 41.3 38.6 38.1 37.1 35.9 39.0 39.5 40.2 35.2 Yemen 59.4 58.9. 58.8 64.1 66.4 48.1 36.7 34.0 39.5 32.3 30.8 35.0 37.9 36.6 36.6 45.0 Burundi 17.6 16.5 16.0 17.0 20.2 20.0 23.4 19.6 16.2 20.2 18.7 20.2 24.1 27.0 27.7 20.3 Rwanda 14.8 14.0 14.0 13.8 22.2 18.8 16.3 18.3 15.8 16.2 16.9 17.3 17.6 18.5 17.8 16.8 atio of broad money to money Small states 2.9 3.1 3.0 3.1 3.2 3.2 3.3 3.3 3.4 3.2 3.3 3.4 3.3 2.9 2.8 3.2 Djibouti 2.1 1.9 1.6 1.5 1.5 1.6 1.4 1.6 1.7 1.7 1.9 2.0 1.8 1.8 I.8 1,7 Egypt 2.7 3.2 3.4 3.5 3.6 3.7 3.8 4.0 3.6 4.0 4.1 4.2 4.3 4.1 4.0 3.7 Ethiopia 1.3 1.3 1.2 1.4 1.3 1.5 1.7 1.5 1.8 1.6 1.7 1.8 1.7 1.6 1.5 1.5 Ghana 1.3 1.5 1.5 1.7 1.7 1.8 1.9 1.9 1.9 2.0 2.1 2.0 1.9 2.0 1.9 1.8 Kenya 2.1 2.2 2.1 2.1 3.4 3.7 3.7 3.6 3.5 3.2 3.0 2.9 2.7 2.3 2.4 2.9 Yemen 1.3 1.4 1.4 1.5 1.5 1.5 1.7 1.6 1.9 1.8 1.9 2.0 2.2 2.3 2.3 1.8 Burundi 1.5 1.4 1.3 1.2 1.2 1.3 1.4 1.4 1.3 1.3 1.4 1.4 1.4 1.4 1.3 1.4 Rwanda 1.9 1.9 1.7 1.6 1.3 1.6 1.5 1.8 1.8 1.8 1.9 2.0 2.1 2.0 1.9 1.8 redit to private sector growth (annual percentage change) Small states 18.3 17.5 15.8 16.3 15.8 22.0 16.5 23.3 14.6 10.3 10.5 9.6 15.1 13.8 15.7 15.8 Djibouti -1.7 -6.3 2.5 -7.6 3.3 10.3 1.1 -9.2 9.1 -34.7 13.9 -14.3 -4.7 -2.9 0.2 -2.7 Egypt 20.1 2.9 20.5 18.6 30.2 34.3 25.9 21.4 31.1 19.3 7.9 14.2 5.4 8.8 3.6 17.6 Ethiopia 21.2 1.4 -5.6 -30.7 38.6 38.0 73.4 21.7 1.4 27.9 9.1 4.4 -2.9 6.4 11.7 13.8 Ghana 20.9 -7.3 48.1 28.4 45.9 43.9 73.1 69.9 40.0 59.7 46.7 19.0 31.6 31.9 33.4 39.0 Kenya 20.2 17.3 18.9 5.0 16.0 37.3 -12.6 16.0 8.1 14.1 3.7 -6.5 4.1 4.0 22.6 11.2 Yemen - 13.2 15.9 26.2 13.0 66.8 -7.7 52.6 34.2 39.3 21.3 25.8 14.3 35.1 30.8 27.2 Burundi 32.7 38.8 6.7 35.1 8.2 -15.1 26.4 12.3 26.7 43.8 34.7 0.4 35.2 10.7 2.9 20.0 Rwanda -25.9 -17.3 25.9 16.7 -9.0 74.2 1.2 56.5 20.5 10.7 16.9 9.3 10.5 14.1 - 14.6 otal deposit growth (annual percentage change) Small states 12.8 14.6 19.5 14.7 20.5 20.1 18.6 19.4 11.1 13.1 15.1 11.9 16.5 11.9 12.5 15.5 Djibouti 2.8 1.8 -11.4 -0.6 1.2 7.0 -12.7 -1.1 8.5 -5.0 1.6 8.8 17.1 19.4 14.3 3.4 Egypt 30.7 23.3 21.6 12.7 10.7 10.1 11.1 10.5 7.0 7.4 12.8 14.3 12.5 21.8 14.6 14.8 Ethiopia 12.3 5.3 13.1 15.9 32.1 11.0 19.0 34.4 6.4 11.1 15.3 11.6 15.3 13.4 18.9 15.7 Ghana 20.0 55.8 42.1 39.2 47.4 40.6 42.6 48.2 16.9 16.2 38.0 55.8 47.8 32.3 29.4 38.2 Kenya 13.6 20.3 34.5 27.9 27.4 33.1 26.6 19.4 2.6 5.4 5.5 4.6 9.3 13.7 14.0 17.2 Yemen - 11.6 17.5 16.2 15.5 112.2 22.2 16.5 20.0 14.4 32.0 23.3 25.3 21.8 20.2 26.3 Burundi 12.9 27.9 2.6 2.6 37.0 -13.6 17.0 14.1 -2.9 58.5 8.9 20.6 31.7 19.8 14.6 16.8 Rwanda 1.2 10.9 10.0 1.6 -7.6 84.6 6.8 40.0 0.9 11.9 16.6 13.6 16.2 14.8 31.5 16.9 :atio of time deposits to total deposits (percent) Small states 70.7 71.1 71.1 70.5 71.2 71.0 70.7 70.3 71.4 69.4 69.3 68.9 68.5 66.0 65.7 69.7 Djibouti 63.7 58.7 51.6 49.9 51.4 55.9 55.3 57.8 55.2 50.9 56.9 59.2 54.2 51.9 53.5 55.1 Egypt 83.8 84.7 86.1 86.8 87.3 87.5 88.3 88.9 89.4 89.5 90.1 90.7 90.5 89.4 89.1 88.1 Ethiopia 44.7 47.4 47.8 50.5 48.9 55.1 61.7 54.8 53.7 54.3 55.5 57.8 66.0 55.6 55.4 53.3 Ghana 37.1 51.8 55.1 56.6 59.8 67.0 69.7 66.8 64.6 72.3 82.2 72.2 67.4 68.0 63.8 63.6 Kenya 64.6 66.4 65.3 65.6 72.0 78.2 80.8 82.2 81.9 79.2 77.3 74.8 73.7 65.6 67.8 73.0 Yemen 50.3 50.3 45.3 44.5 43.3 75.8 76.4 78.6 80.6 84.0 83.3 83.6 86.8 87.8 87.5 70.5 Burundi 43.7 47.2 44.2 40.8 43.9 36.3 47.6 43.4 41.8 43.3 47.1 44.3 46.6 42.4 36.5 43.3 Rwanda 65.3 63.4 55.2 50.9 33.3 49.5 47.8 48.4 53.4 51.2 57.4 61.5 61.0 60.9 64.4 54.9 mces: IMF, International Financial Statistics and World Economic Outlook 2004 Small states are Antigua and Barbuda, the Bahamas, Bahrain, Barbados, Belize, Bhutan, Botswana, Cape Verde, Comoros, Cyprus, Dominica, Equatorial Guinea, Estonia, Fiji, Gabon, The Gambia, Grenada, Guinea-Bissaq Guyana, Maldives, Malta, Mauritius, Qatar, Samoa, SPo Tome Principe, Seychelles, Solomon Islands, St Kim and Nevis, St. Lucia, St. Vincent & Grens., Suriname, Swaziland, Tonga, Trinidad and Tobago, and Vanuatu

107 Table 4.2 Djibouti: Sectoral Distribution of Credits, 1995-2004 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 (DF millions) Individuals 8,785 7,849 9,412 9,469 7,599 7,643 7,413 6,270 6,160 6,368 Commerce and Industry 33,912 31,713 28,235 32,070 31 ,I85 32,433 29,289 23,018 25,524 25,146 Government 189 653 1,591 2,285 2,621 1,544 1,373 1,442 1,337 1,575 Other services 547 838 902 1,450 673 361 440 269 21 8 102 Total 43,433 41,053 40,140 45,274 42,078 41,981 38,515 30,999 33,239 33,191

(percent of total credits) Individuals 20.2 19.1 23.4 20.9 18.1 18.2 19.2 20.2 18.5 19.2 Commerce and Industry 78.1 77.2 70.3 70.8 74.1 77.3 76.0 74.3 76.8 75.8 Government 0.4 1.6 4.0 5.0 6.2 3.7 3.6 4.7 4.0 4.7 Other services 1.3 2.0 2.2 3.2 1.6 0.9 1.I 0.9 0.7 0.3 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0

(annual change in percent) Individuals -10.7 19.9 0.6 -19.7 0.6 -3.0 -15.4 -1.8 3.4 Commerce and Industry -6.5 -1 1 .o 13.6 -2.8 4.0 -9.7 -21.4 10.9 -1.5 Government 245.5 143.6 43.6 14.7 -41.1 -11.1 5.0 -7.3 17.8 Other Services 53.2 7.6 60.8 -53.6 -46.4 22.0 -39.0 -18.8 -53.4 Total -5.5 -2.2 12.8 -7.1 -0.2 -8.3 -1 9.5 7.2 -0.1 Source: Central Bank of Djibouti

4.2.2. Bank lending trends

4.15. Bank lending has remained low, exhibiting a declining trend since 1999, despite noticeable increase in deposits. The deposit transformation rate3' , remained above 70 percent during most ofthe 1990s but has been declining sharply since pealung at 97 percent 1998, reaching about 41 percent in 2004 (Figure 4.3). This trend may be explained mainly by uncertainty about government policies, investment profitability in Djibouti, and a slowdown in trade activities.

Figure 4.3 Djibouti: Total Bank Deposits and Credits (DF millions), 1990-2004

90000 100.0 80000 .- 90.0

70000 .- 80.0 .- 70.0 60000 u) .- 60.0 .E= 50000 C 'E -- 50.0 g LL 40000 al D .- 40.0 30000 -- 30.0 20000 -- 20.0 10000 -- 10.0

0 7 0.0

I -Total Deposits 0 Total Qedit -Deposit transformation ratio (%) Source: Djibouti authorities.

39 The deposit transfokation rate is defined as the ratio of total credits to total deposits.

108 4.16. Bank lending distribution is highly concentrated in terms of beneficiaries, activities, and maturities, and the customer base remains modest. About 76 percent oftotal loans are extended to the commercial and trading sector and 19 percent to individuals (mostly for mortgages) (Figures 4.4.a and 4.4.b).40 Short-term lending represents about 53 percent of total lending, and long-term lending averages about 6 percent. It is also estimated that the two main commercial banks together have about 150 large customers with a credit volume exceeding DF 15 billion; their relations with the largest 50 customers play a key role in determining overall lending volumes.

Figure 4.4.a. Djibouti: Credit Distribution, December 2004

A. By economic actidty Public Others Individuals 0% 1 I

I 76% I Source: Djibouti authorities.

Figure 4.4.b Djibouti: Credit Distribution, December 2004 B. By maturity 1

Short- term 53%

term Medium- 6 Yo term

Source: Djibouti authorities.

4.17. Banks lend primarily to the trading community and loans are not easily available to other customers such as small businesses. Against the background of economic uncertainties, the lack of profitable projects, commercial banks are cautious when granting credit and impose stringent conditions when they do so. Small operators, particularly those engaged in the informal sector, are excluded from bank lending because they often fall short of meeting the stringent qualification criteria set by banks. Lacking alternative sources of financing, small projects that are more commensurate with the size of domestic markets and that can be labor-intensive are often abandoned.

40 The Djibouti Development Bank (BDD) was the only bank specializing in providing long-term credit and equity to the real sector.

109 Structure by matuny (OF mitlior!s)

15,828 17,824 20.5.u 22.ffi8 20,729 2l,ffi5 18,405 15.593 18,121 i8,mi 18.443 17.311 24,225 28.133 33,515 1,827 2 187 3.109 2 €38 3,085 2,852 3,OBl 3.258 3,388 3.506 3,783 4,324 4,651 4.675 4,810 22,910 19573 15,401 13.507 13,101 14,629 11,857 11,828 11,531 9.891 11,188 12,223 12.583 15,898 19.505 5417 6 181 5,589 5,843 8,014 12.105 11.058 11.604 13.168 13.324 15.542 17.888 18.089 x.833 22.939 47.083 45.878 44.642 45.388 44.808 50.75- 45.380 43.W 48.5m 48.Y)2 48,924 51.547 60.528 70,490 80,889

(pent of total deposits)

33 8 38 1 460 486 482 415 428 32 38 0 42 5 37 7 33 5 400 41 3 41 4 39 48 70 85 58 58 57 75 73 75 77 84 77 70 59 487 42 7 345 29 5 28 2 288 28 1 275 25 4 21 2 22 8 23 7 20 8 22 8 24 2 13 5 13 5 12 5 15 1 178 238 244 267 25 3 28 8 31 8 343 31 5 28 1 28 4

(annual shawe in prurnt)

12 5 14 5 74 41 18 -79 -145 92 98 -7 1 -8 1 38 8 20 3 15 0 18 7 42 2 45 33 -37 37 84 40 35 78 14 3 78 58 -2 4 -14 5 -21 3 -12 3 -3 0 11 7 -188 08 -0 8 -15 4 128 94 2.8 265 233 -3 5 .8 7 224 17 1 510 -87 50 13 5 13 188 13 8 78 75 11 7 .2 5 -2 7 17 -1 1 130 -105 44 72 02 50 53 174 15 5 14 7

Structure by Dep~ltor (DF millions)

18W 15883 18468 15184 15334 15,875 14.778 15,888 15,ffil 15 325 15,lffi 15.483 17.652 18,591 18885 8335 IC432 11701 12288 9921 8,678 8.445 8,725 10.747 12,759 21.w 23.780 14.6TT 19,887 37,825 10315 8150 5718 5880 6682 5,088 4,351 2,784 2.155 1.158 1.783 1,880 2.458 3.454 3 738 8786 1Mol 10755 12038 12681 17,758 15.80) 15.883 17.546 17.352 8,135 8,523 25,741 28.188 20,221 47083 451178 Me42 w 44908 fS,Y7 21.157 18,787 19,701 18.508 48,958 51,547 W.528 70.400 80,889

(pehant of lot4 depslts)

40 2 345 358 535 341 712 598 847 81 5 828 33 1 32 0 28 2 25 8 23 5 17 7 227 25 2 27 0 221 415 445 485 545 €88 448 461 24 2 25 2 488 21 3 20 0 12 8 13 0 14 8 255 205 148 108 52 35 32 41 48 45 20 5 227 24 1 285 28 8 745 784 852 88 1 e38 157 187 42 5 40 0 25 0

(annual chawe In phant) -15 2 37 -7 7 08 107 -128 75 11 45 57 18 71 75 00 25 2 12 2 48 -18.1 0.1 48 -75 23 1 I87 71 2 89 -35 3 35 3 81 0 -5 5 -37 5 30 13 1 -85 -255 -38.1 -22 5 454 52 5 -5 8 480 405 82 53 34 11 8 78 387 -54 48 85 -1 1 474 53 157 5 85 -28 3 -2 5 -2 7 17 -1 1 -488 -11 3 -11 3 5.0 41 1844 53 17.4 155 14 7 Source: Djibouti authorities

4.18. Current trends in bank lending are also determined by the structure of demand for credit. In addition to enterprises eligible for bank financing, there are two major groups in Djibouti that could need financing. The first includes operators performing low-income-generating activities (such as trading, handicrafts, and construction services). These small-scale operators respond to demands for goods and services by low-income and poor families, and they need flexible financing at competitive prices. They are excluded from bank credit because the low loan amounts they require incur high transaction costs that banks are unwilling to assume. The absorption capacity of these operators is low, given their limited activity, and because oftheir precariousness they cannot provide the collateral required by banks. The second group includes domestic investors with some financing capacities. Although they can apply to banks for credit, these operators are discouraged by banks’ stringent conditions and collateral requirements. Banks could increase their operations with these operators if the judicial system became more efficient and responded in a timely manner, leading banks to soften their conditionality; and if adequate financial products and an adequate regulatory framework are developed.

4.2.3. Overall soundness of the banking system

4.19. Thefinancial system is broadly sound; but the quality of bank assets needs to be further improved as reflected in the high share of non-performing loans to total loans. Capital adequacy and risk diversification ratios were generally-metby the two dominant commercial banks, BCI and BIS.41For the banking system as a whole, the share ofNPLs (including provisions for NPLs) have averaged 24 percent over the period 1995-2002. This trend was reversed since 2004 and currently

41 Both banks follow prudential regulations from their headquarters, which are stricter than those imposed by the BCD. Most domestic banks failed to observe prevailing regulations and fell victim to lax management, especially in screening and monitoring loan beneficiaries.

110 the share of NPLs averages 17 percent for the two active commercial banks (see section 4.4 reporting recent developments on non-performing loans).

4.19. In 2004 more than Sopercent of bank revenues came from net interest income. Limited competition in the small banking sector has helped banks manage their interest income and stabilize profit developments. Because of the small size of the banking market, the monetary authorities have been concerned for a long time about the risk of collusion in setting interest rates. In August 1999 the BCD instructed banks to cease coordinating interest rate structures and to publicize and report their structures to the BCD. However, recent interest rate developments suggest that banks may still use their market power to some degree. Minimum deposit rates declined from 3.9 percent in 2000 to half a percent in 2004, but minimum lending rates remained at 10.8 percent, thus further widening interest rate spreads. There is no evidence supporting the idea of a liquidity crunch in the Djibouti banking system. Limited lending opportunities, coupled with commercial banks’ very cautious lending approach, lead to a significant amount of net foreign assets placed by commercial banks with their parent banks.

4.20. Notwithstanding the improvements in recent years, the banking sector still exhibits a few vulnerabilities that need to be addressed. First, one of the commercial banks still fails to meet a number of prudential regulations and the other fails to meet most of them. Failure to quickly and decisively address this issue may ultimately have negative effects on the integnty of the banlung system. Second, the large share of interest income from foreign assets in revenues (more than 25 percent of total revenues) might involve some risks over the medium term, particularly if competition in the banking sector increases. Similar risks may exist if international interest rates remain low over a long period. If this were to occur, domestic interest margins might narrow (given the tight link of domestic rates to international rates) and banks would have to cut costs or find other sources of income to avoid losses. Recent experience shows that the two main commercial banks managed to more than offset the decline in interest income by increasing commissions and other fees for customers’ operations. Third, the liquidity and solvency ofthe banking system could also be threatened by a banking crisis in the parent institutions’ home countries or in a major trading partner. While not very likely, such a crisis would have a large impact on the Djibouti banking system.

4.2.4. The money changers

4.21. In addition to the three commercial banks, there are six money changers in Djibouti. Some have been in operation since 1948, beginning as small enterprises focused on serving the businesses of their owners. They often operate informally, and their operations are traditionally limited to the subregion and the Arabian peninsula. Their working procedures remain largely outdated despite the increasing volume oftheir activity.

4.22. The money changers offer currency exchange and international transfer services, even to clients with no bank account. Apart from the two commercial banks, Djibouti has 11 currency exchange and transfer agents. The currency exchange activity is limited to Djibouti and does not I involve transactions with other countries. Nevertheless, when foreign currencies are not available in the domestic banking system, the money changers call on regional commercial banks to satisfy their need or, in some cases, travel to purchase and repatriate foreign currencies. For currency transfers, the money changer issues a transfer order to the beneficiary or a subsidiary in the beneficiary’s country, or orders the bank in which it has an account to make the transfer.

111 4.23. The money changers play an important role in financial intermediation. Based on the 2001/2002 onsite supervision, the money changers handle about 17 percent of total financial transfers abroad and about seven percent oftransfers received in Djibouti. About 45 and 40 percent of receivable transfers originate from the United Arab Emirates and the Republic of Yemen, respectively; the rest come mostly from Saudi Arabia and Ethiopia. About 90percent of these transfers are in U.S. dollars; the rest are in Djibouti francs and Ethiopian birrs. The currency composition of transfers payable is about 66 percent in the U.A.E. dirham, 20 percent in the Saudi Arabian riyal, and 10 percent in the U.S. dollar. The money changers are appealing because they perform several useful functions: They operate on somewhat smaller spreads and charge lower commissions than the commercial banks. They deal with all types of customers and have more flexible workmg hours than banks. And they are more specialized in their operations than banks, which conduct a broader range ofactivities.

4.24. Although the BCD is responsible for supervising the money changers, their activities remained large& unmonitored until recently. After the events of September 11, and in line with the intensification of the international fight against terrorism and money laundering, the BCD launched in end-2001 and early 2002 a vast onsite program of supervising all money changers. To further enhance supervision, the authorities adopted in 2002 a law on money laundering, confiscation, and international cooperation with regard to the proceeds of crime. At the same time, they revised the banking laws to provide for more regular onsite supervision and systematic control ofall money changers.

4.25. Annual on-site supervision ensures that the money changers broadly operate in full compliance with the regulations. Most ofthe money changers rely on correspondent networks that impose sufficient financial guarantees and soundness. Nevertheless, the BCD inspection team recommends that the money changers do the following: computerize their data and improve their maintenance; gradually introduce modem accounting practices to keep up with their growing activities; and request, if necessary, technical assistance from the BCD regarding accounting and regulatory practices.

4.2.5. Informal finance mechanisms

4.26. Outside the formal sector, informal finance exists in the form of traditional arrangements. Informal finance is understood in this context as contracts or agreements conducted without reference or recourse to the legal system, to exchange (primarily) cash in the present for promises of (primarily) cash in the future?’ These arrangements are used by low-income participants to pool small savings and extend small credits. Box 4.3 provides an overview of the features and activities ofthe three main arrangements in Djibouti.

42 Informal finance can be seen as a positive reaction to the inadequacy and lack of flexibility ofthe banking sector with regard to low-income and irregular income earners.

112 Box 4.2 Three Types of Informal Finance Arrangement

Three types of informal finance arrangements are present. The first is revolving savings and credit associations, also known as tontines. The most widespread type is the mutual tontine, organized on the principle of solidarity between members who know each other well. This is a financial association created between people who decide to pool their savings and take turns in accessing the pool. Each member gets the pooled money according to an established order, which can be subject to change. Every member can also lend money or take out a loan. In principle, debts are not subject to interest. A mutual tontine primarily helps members accumulate their own savings on a daily basis, or over a period of one week or one month. Participants generally pay their share according to the time horizon they choose; with higher contributions for longer time horizons (daily contributions are lower than weekly ones, which are lower than monthly contributions). The average size of a tontine is nine participants, of which about 85 percent are women. The membership usually brings together neighbors (40 percent), friends (20 percent) and coworkers (60 percent). About 20 percent of the existing tontines are daily, 30 percent weekly, and half of them are monthly.

The second form of informal finance arrangement is a credit to wage earners (biil or masrouf), which is widely used in the capital city. Small loans are extended to participants for fixed fees to pay for their daily expenses. These loans are then reimbursed by participants at the end of the month when they (usually) receive their salaries.

The third type of informal finance arrangement is the retailer credit. In this situation, some traders entrust peddlers or retailers, who do not have working balances at their disposal, to sell some goods door to door or in their shops. The sale income is then paid to the original traders for a fixed commission.

Source: Djibouti authorities. a. Tontines are known in virtually all African countries. They are diverse and have different names depending on the country or even digerent regions within the same country: esu or aa'jo in Benin and Niger, for example. There are three types of tontines: the mutual tontine, thejnancial tontine, and the commercial tontine. b. In 1995 the shares paid by participants were estimated at US$3 for a daily tontine, US23 for a weekly tontine, and US$89for a monthly tontine.

4.27. The traditional arrangements have several advantages and weaknesses. The success of these informal systems stems principally from ensuring low transaction costs; a supply of loans, savings, and implicit insurance; services that are sensitive to constraints faced by women, in particular their inability to obtain formal financing; the substitution of confidence in character for physical collateral; and socially enforced or self-enforced contracts. These forms show how those excluded from the banking system fashion their own financial instruments in the absence of formal services. But they have several weaknesses that should not be overlooked, including alack of deposit insurance; operation within a limited group, solely on the basis of individual savings and inability to satisfy the financing demands for economic activities that exceed available savings; the limited proportion of savings invested and the short-term basis; and the impossibility of recourse to a legal system that enforces contracts.

4.2.6. Microfinance

4.28. Microfinance corresponds, in this context, to formal schemes designed to improve the well-being of the poor through better access to saving services and loans. Thus, both informal finance and microcredit serve the poor. However, informal finance derives from the need of those excluded from the formal finance for financial services and loans, while microfinance stems from a donor-driven supply of financing. Lessons drawn from informal finance are usually very instructive for designing microfinance.

113 4.29. Although microfinance in Djibouti started a decade ago, it is still at an early developmental stage. Microfinance started with the French Development Agency’s (AFD) program (Programmes Aide aux initiatives de Base), which benefited, between 1992 and 1995, 32 small and medium enterprises (hotels, fisheries, bakeries, and small manufacturing enterprises). The project failed for various reasons, including the absence of reliable information about the solvency of beneficiary firms; the AFD’s lack of experience in microfinance; the insufficient human and financial resources to manage and monitor the program; and a judicial system that could not enforce contracts, as evidenced by the fact that only two enterprises have fully repaid their loans.

4.30. A similar project was launched by Caritas Internationalis (CARITAS) between 1996 and 2000. The approach used was similar to that ofBangladesh’s Grameen Bank.43 It was based on the voluntary formation of small groups (four people) to provide mutual, morally binding group guarantees in lieu of the collateral required by conventional banks. The project benefited 800 women, who received small loans (on average US$600 over one year with a 10 percent interest rate) totaling about US$709,000. The project ended in 2000 for similar reasons as the AFD program. Following the CARITAS approach, a microcredit program was launched by the Social Development Fund (FSD) in 200 1. The program was financed by the African Development Bank and promotes income-generating activities by granting microcredits to women. The implementation of the program is entrusted to nongovernmental organizations (NGOs) and financed by the FSD, which also provides support on methodology, training, management, and computerization. The women who participated live in the capital and three other cities (Tadjourah, Obock, and Ali Sabieh). They each received between US$170 and US$560. The program benefited 800 women, who obtained loans totaling about US$113,000.

4.31. A more recent initiative was the creation in 200144of the Djibouti Development Fund (Fonds Djiboutien de Ddveloppement, or FDD), presumably to replace the Djibouti Development Bank, which was liquidated in 1999. The FDD started operations in 2005. In 2002 the International Fund for Agriculture and Development (FAD) conducted a study on the feasibility of rural microfinance. A project agreement was signed in 2003 for a total of USD 3 million aimed at supporting microfinance development and rural micro-enterprises.

4.32. The development of microfinance in Djibouti has been hindered by the lack of a regulatory framework, a weak judicial andfiscal environment, resource constraints, and the lack of institutional support. Microfinance activities remain largely unregulated. However, since the 2000 banking law alludes to adjustments necessary to regulate microfinance a~tivities:~ it is presumed that the BCD is aware of the need to develop a legal framework for microfinance institutions. The legal system is weak and often fails to enforce contractual and property rights. There is limited knowledge of judiciary rules, inadequate numbers of justice professionals, and fiequent political interference. In addition, most microfinance institutions in Djibouti are largely subsidized, operate with insufficiently trained staff, and have not yet defined clear development plans. Furthermore, their accounting and management systems are poorly developed.

43 The Grameen Bank provides credit to the poor in rural Bangladesh, without any collateral. Its banking system is based on mutual trust, accountability, and participation. 44 The decree instituting the FDD was issued in 2001 and modified in July 2002 (Decree no. 2002-0133). 45 Articles number 11 and 16 of the law indicate that any organization that starts a microfinance program needs to solicit an official operation agreement, even if the conditions for its financial viability are not entirely met or it is unable to provide reliable data in line with BCD’s regulations.

114 Box 4.3 Current Status of Microfinance in Djibouti

Microfinance is at a very early development stage in Djibouti. To date, no decentralized financial institution has been created and there is yet an institutional, legal or regulatory framework specifically for microfinance. To remedy these shortcomings, the Djibouti government approached the UNDP and the International Fund for Agriculture and Development (IFAD). An initial agreement was signed with IFAD for USD 3 million in 2003 with the aim of supporting microfinance development and rural micro-enterprises, through the setting up of a network of credit and savings institutions in urban and semi-urban areas in addition to perennial support structures for micro-entrepreneurs. The project launch took place in February 2006. The UNDP project has a budget of USD 112,600 to finance the implementation of a National Microfinance Committee as well as developing a national microfinance strategy in addition to microfinance legislation.

I Source: Note prepared by the Banque Centrale de Djibouti, 2006

4.3. FINANCIAL SECTOR REFORMS

4.33. To maintain the integrity and effectiveness of thefinancial system, the authorities took a number of measures in the last decade covering three main areas, namely the legal framework, supervision and prudential regulations, and financial intermediation. However, these measures were driven mostly by the requirements of donor-supported programs and were not really part of a comprehensive financial sector reform program. Furthermore, these measures were not very ambitious and their implementation was slower than initially envisaged. As a result, they have had only limited impact. Additional efforts need to be made to strengthen the financial system and improve Djibouti’s attractiveness as a regional commercial and financial center. These efforts should focus on the areas mentioned above but also should aim at improving contractual property rights and further developing the financial sector.

4.34. Revised banking laws and central bank statutes were adopted by parliament in June 2000, replacing laws dating back to 1985. The 2000 banking law introduced four main provisions. First, for a new banlung institution to operate, at least 30 percent ofits capital should be owned by a sound foreign banlung institution (only 20percent in the 1985 law). Second, each bank should conduct regular internal audits and report the results to the BCD. Third, the BCD should undertake annual onsite supervision of each bank. Fourth, disciplinary actions should be taken against banks’ managers in case of misconduct. As a result of this law, reporting requirements were strengthened. All banlung institutions must report to the BCD on their activities in However, the new laws contained a number of weaknesses, including inadequate rules for setting up new financial institutions; a failure to specify sanctions for banks not complying with banlung regulations; and operating rules for the central bank that would allow it to extend commercial and other credits to private, non-financial institution^.^' Legislative amendments to correct these shortcomings were prepared in 2002. The revised draft law, which also includes institutionalizing the external auditing of the BCD’s accounts mechanism, was submitted to the cabinet for approval in late 2002 but has yet to be adopted by the cabinet and the national assembly. Two new banking laws were adopted by the cabinet in January 2005 - the first with respect to the setting up of new financial institutions and the second concerned review of the BCD statutes.

4.35. In the same vein, December 2002 the government adopted a new law on money laundering, confiscation, and international cooperation with regard to the proceeds of crime. This law provides Djibouti with a sound legal framework that should be used to ensure continuous monitoring of the financial system’s integrity. A financial information service was created within

46 In the past, banks did not report to the BCD systematically. 47 The provision was in the previous law, but never used, and was inadvertently kept in the revised BCD statutes.

115 the BCD for this purpose. The authorities are also pursuing their efforts to increase the awareness in financial institutions of money laundering issues, updating the regulatory framework for financial operations, and enhancing the capabilities ofthe BCD’s financial investigations unit. The next steps are to expedite the adoption of the revised banlung laws to correct flaws in the financial system; and to implement measures to combat money laundering and fully comply with the UN resolutions.

4.36. The authorities have made headway in improving the supervisory capabilities of the BCD, for both offsite and onsite supervisory capabilities. This was achieved through personnel training and external technical assistance. For offsite supervision, conducted at least once a year, the Djibouti banlung regulations include a comprehensive set of prudential ratios for banks. The requirements take into account the country-specific context4* by being stricter than the generally internationally accepted standards for capital adequacy and risk diversification. Onsite supervision of commercial banks started in 1996 and aims to inspect each financial institution once a year. However, because of administrative capacity constraints, the BCD has supervised only one bank per year onsite. The BCD completed-for the first time-controls of all the money changers operating in Djibouti between end-2001 and early-2002. It plans to inspect them once a year, starting in 2003. To this end, the BCD started a program to train its staff, including recent recruits, with the support ofthe Bank ofFrance.49 In addition, the BCD is approaching other donors to raise additional financial resources to fund the training program.

4.37. To improve the transparency of the central bank, the monetary authorities have taken steps to regularly audit the annual financial statements of the BCD in accordance with International Standards on Auditing (ISA). A new external auditing firm audited the BCD accounts for 1999,2000, and 2001 in December 2002. The same firm will conduct the audit of the BCD accounts for 2003 and 2004, thus covering a five-year period. The BCD received technical assistance from the IMF on adopting the ISA in 2005 and is acquiring new software in 2005 to centralize their accounting data as mandated by the IAS standards.

4.3 8. Notwithstanding the progress achieved so far, the authorities are encouraged to pursue efforts aimed at strengthening the banking supervision. The BCD should extend its training program to include various banking issues” so as to improve the capacity ofits staff to keep abreast of all developments in the financial system. Money changers should also benefit from training courses organized by the BCD, especially courses on accounting procedures and standards, and should be kept abreast of developments involving their activities, such as changes in financial regulations and implementation ofthe law on money laundering.

4.39. Sound financial intermediation for the benefit of all operators remains a focus for further reform. To better assess some of the risks that banks are facing in their financial intermediation activities, a series of measures were implemented in 1991. A Risk Office (Centrale des Risques) was established, centralizing the accounting of about 80 percent of outstanding bank credits above a minimum level (DF 10 million for commercial banks and DF 5 million for the development bank). In addition, the BCD initiated the establishment of a data bank on outstanding checks without ~over.~’A government decree requiring the submission of such information was

48 Namely, a high level ofNPLs, Djibouti’s role as a regional commercial and financial center, and the preservation of the currency board. 49 This training program began in June 2002 and continued through December 2002. 50 Issues include recent developments in the Bale 2 agreement with regard to new prudential standards, new financial products, and various aspects of microfinance. 51 This ratio increased further subsequent to the reduction in 2001 of the minimum level of outstanding bank credits, which was brought to DF 3 million.

116 adopted at the same time. However, these initiatives were limited in scope and could not by themselves encourage the banks to be more active in the domestic market.

4.40. In 2001, in an effort to supplement the commercial banks in providing financing to investors, a presidential decree authorized the creation of a developmentfund (the FDD). Under the supervision of the Ministry of Finance, this fund is a public enterprise that is expected to use concessional lines ofcredit from bilateral, multilateral, and private sources to extend credit to small and medium enterprises, especially in the tourism, fisheries, services (commerce, handicraft, transport) and agriculture sectors. The FDD investment program for 2003-2010, estimated at about US$35 million, will be financed by the Djibouti government (5 percent), international organizations (70 percent), and the domestic private sector (25 percent). However, there was concern about the possible profitability ofthe fund, and the government emphasized the importance ofnot providing guarantee^.^' Furthermore, the FDD is not supervised by the BCD, which could generate difficulties similar to those that led the BDD to cease operations in 1996.

4.41. To broaden financial activities in support of the national poverty reduction strategv, the authorities have also taken initiatives to develop microcredit programs. These initiatives aim to establish institutions that grant direct loans, using their own funds and foreign borrowing, to private sector operators-especially to small operators who usually do not have access to bank lending. However, these institutions can fulfill their mission only if an adequate regulatory environment is put in place. It is also important that these institutions be regularly supervised by the BCD to establish healthy relationships with the beneficiaries ofmicrocredits.

4.42. In view of the difficulties faced by the banlung system in enforcing loan agreements, reflecting essentially weaknesses in the judicial system, in July 1999 the BCD established a notification system under which banks are prohibited from lending to defaulting borrowers whose nonperforming credits exceed a threshold amount and who are in arrears, until their loans are regularized. However, the steps undertaken to expedite the judicial process for bank recovery ofNPLs did not progress as envisaged.

4.43. Outstanding weaknesses in the enforcement of loan agreements combined with mismanagement hindered the viability of two commercial banks, the A1 Baraka Bank and the Banque de Djibouti et du Moyen-Orient, and they were liquidated in 1998 and at end-1999, respectively. The recovery of these banks’ debts and the gradual reimbursement of depositors have been somewhat slower than anticipated owing to a number ofjudicial delays and other constraints. In addition, the authorities experienced difficulties in liquidating assets, mostly those in the form of real estate.53

4.44. The weak enforceability of contracts and lack of secure property rights probably have a signijicant impact on the behavior of thefinancial sector.54The court system does not seem to enjoy the confidence ofthe private sector. Judges are perceived as lacking autonomy. Furthermore, the court system is slow to resolve disputes. A good illustration is the lengthy liquidation ofthe two banks just mentioned-which has yet to be finalized. Some of these problems are due to a lack of

52 The authorities indicated that they intend to take a small equity share of 20-30percent (as they do in a domestic commercial bank) and would share the profits and losses of the fund in proportion to its capital contribution. Moreover, the government’s contribution would be financed entirely with external concessional resources. 53 Djibouti’s real estate market is depressed, and it is feared that hastening the sales ofreal estate properties would lead to large capital losses. 54 Levine (1999) found that various legal determinants were significantly correlated with the indicators of financial development, having the expected signs despite a difference in their significance levels. He concluded that countries with legal and regulatory systems that assure high protection of creditors tend to have better-developed financial intermediaries.

117 financial and human resources; others result from political interference in the functioning of the judicial system. The authorities rightly recognize the importance of improving the legal system to ensure success in their national poverty reduction strategy, but they need to expeditiously take concrete measures in this direction.

4.4. MONETARYAND BANKING DEVELOPMENTS

4.45. During 1990-96 broad money and nominal GDP evolved with similar trends.55Broad money and nominal output displayed opposite trends during 1995-1007. While GDP grew moderately, broad money declined steadily (Figure 4.5). The decline most likely resulted from the erosion ofthe fiscal position. The rapid accumulation ofbudgetary arrears and forced borrowing to cover budget deficits during most ofthe 1990s weighed heavily on the financial situation ofprivate suppliers, public enterprises, and civil servants. Because deposits with commercial banks are the only financial instrument available in Djibouti, the worsening in the financial position of these economic agents led to the decline in broad money as well as a shift fi-om longer- to shorter- maturity deposits. At the same time, uncertainty about the economic situation and the lack of confidence toward the government appear to have contributed to some capital outflows, as evidenced by the high level ofnegative errors and omissions in the balance ofpay~nents.’~

Figure 4.5 Nominal GDP and BroadMoney (log scale), 1990-2004

5.1 - 5.0 - 4.9 - - / 0 0 4.8 - c # - zc-- \ 4.7 - \/ 4.6 - 4.5 - 4.4 , Norrinal GDP Broad mney 4.2 - - -

Source: Djibouti authorities

4.46. Between 1998 and 2004, broad money grew at about lopercent per year, while credit to the private sector declined sharply. Combined with a significant increase in deposits, this decline in credit to the private sector led to a sustained increase in the net foreign assets of commercial banks. Furthermore, the recent strong growth in broad money did not cause inflationary pressures, as evidenced by a CPI inflation rate that remained close to 2 percent (Figure 4.6) and a money velocity (as measured by the ratio of broad money to nominal GDP) that declined by about 28 percent (Figure 4.7). These elements could signal a recovery in the liquidity ofthe economy and appear to be a reversal oftrends observed during the second half ofthe 1990s.

” The only exception was 1992, when GDP increased at about 3 percent while broad money declined by about 2 percent, reflecting a fall in net official foreign assets. The ratio of net official foreign assets to the monetary cover decreased to about 130percent, from about 170percent at end-1991. The exceptional level of monetary coverage in 1991 is attributable to important transfers from Saudi Arabia to compensate Djibouti for its support to allies during the war to liberate Kuwait from the Iraqi invasion. Most of these transfers were used by the Djibouti authorities to finance the civil War. 56 Negative errors and omissions (including short-term capital flows) in the balance of payments were estimated at an average 2 percent ofGDP a year between 1990 and 2000, with a pickup to about 7 percent of GDP per year for 1996-98.

118 Figure 4.6 Inflation and Broad money, 1990-2004 Figure- 4.7 Inflation and M2 Velocity, 1990-2004 (annual percent change) (annual percent change)

20.0 8.0 1\ \ 6.0 - \ - -m Broad mney growth -Velocity 6.0 - 4 4.0 - "/ \ 0.0 - \

-4.0 -8.0 -

ONdfDODONd mmmmmooo ZZZZZWRR Source: Djibouti authorities

4.47. The stabilization and then reversal of the declining trend in broad money started in 2000, essentially because of the recovery in economic activity, mostly attributable to the transport sector (following Ethiopia's decision to channel its transit trade through the Port of Djibouti). In addition, with an improvement in fiscal management, the accumulation of budgetary arrears subsided and was halted in mid-2001. This undoubtedly contributed to a rise in deposits and broad money, which grew at a faster pace than nominal GDP. These movements also allow for the real money balance to catch up, toward a level more consistent with the level of economic activity in Djibouti. Large (unidentified) capital inflows (part of errors and omissions in the balance of payments) may also have contributed to money growth.

4.48. Deposits grew signijicantly in response to an improvement in the liquidity situation of households and the limited range ofjinancialproducts available. Furthermore, the composition of deposits changed in favor of foreign currency. The share of foreign currency in broad money (M2) rose from 36 percent in 1994 to 48.5 percent in 2004 (Figures 4.10 and 4.11). In the absence of other factors, this development appeared to reflect the growing preference for cash vouchers (an instrument denominated in U.S. dollars in bearer form), and an erosion in confidence in the economic system.

Figure 4.8 Demand and Time Deposits, 1990-2004 Figure 4.9 Currency Composition of Deposits, 1990-2004 @ercent ofbroad money)

45000 - 45000 - - - lrre deposits -Derrand depwns - - Time deposits -Demnd deposits 40000 - / 40000 - /

Source: Djibouti authorities

119 4.49. Currency substitution has increased. Foreign currency deposits represent only part of foreign currency holdings in Djibouti:’ and the data do not distinguish between domestic residents and foreign holders. With a substitution ratio (foreign currency deposits to broad money) of about 48.6 percent in 1995, Djibouti already stood among the highly dollarized economies examined in Balifio et al. (1999), such as Argentina, Nicaragua, and Croatia. The substitution ratio has remained around the same level since then. Figures 4.10 and 4.1 1 depict the trend of currency substitution in Djibouti by using the ratio of foreign currency deposits to both base money (Ml) and broad money (M2). The large gap between the ratios of foreign currency deposits to M1 and to M2 reflects the fact that broad money absorbs a large portion of foreign currency deposits, and hence a large part of changes in broad money is explained by changes in those deposits. Currency substitution may benefit the economy by enhancing financial intermediation and promoting financial deepening. However, by inducing a reduction in the real demand for domestic currency, it may involve the loss of seigniorage, the profits accruing to the monetary authority from its rights to issue legal tender. Although these negative impacts may be limited, their growing importance (as shown by the various ratios discussed above) raises some concerns. The evolution of these ratios may suggest that despite a fixed-exchange rate regime, progress achieved through financial sector reforms, a low level of inflation, and continued political stability, the Djibouti people are still not fully confident about domestic developments. This corroborates the declining trend of net domestic credit, which may suggest that banks still see investment abroad as a safer prospect.

Figure 4.10 Djibouti: Foreign Currency Deposits Ratios, 1995-2004 Figure 4.1 1 Djibouti: Foreign Currency Deposits, 1995-2004

Figure 4.10 Qibouti: Foreign Qrrency Deposits Ratios, 1995-2004 figure 4.11 Djibouti: Foreign Qlrrency Ceposits, 1995-2004 Forelgn currency deposiVMI 65.0 - + - Foreign currency deposiUM2 llO.O - 1 45 ,‘\\ 155.0

30.04 , I 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 1995 1996 1597 1998 1999 2000 2001 2002 2003 2004 -Foreian currencv deDoseS - - - - Foreian currencv dewsltlV.2 Source: Djibouti authorities

4.50. Owing to recent financial sector reforms, non-performing loans declined substantially to about 17 percent of total loans by the end of 2004 with an average provision rate of over 70 percent (See Figure 4.12). The substantial decline in non-performing loans is explained by the cease of operations of the CBE and also by the provisioning policies ofthe two commercial banks. There are substantial differences in risk exposure and performance between these two banks. By end of 2004, the share ofNPLs to total for BCI amounted to 23 percent whereas it was only 12 percent for BIS. At the same time, BCI’s loan loss provision rate is over 90 percent whereas is about 68 percent for BIS. The lower return on equity and on assets in BCI compared to BIS reflects in part the higher provisions for loan losses. Overall, the profitability of both banks has remained low because of the still substantial amount ofNPLs on their books.

57 A more comprehensive measure of the stock of foreign currencies held by Djibouti residents conventionally should be viewed in terms of the volume of foreign currencies circulating in Djibouti, foreign currency deposits in the domestic banking system, and cross-border deposits held at foreign banks. Because data on the last two components are not available, the only proxy for measuring foreign currency holdings in Djibouti is foreign currency deposits.

120 Figure 4.12 Non-performing loans, 1994-2004 Figure 4.13 Deposit and Credit, 1995-2004 (percent of total loans) (DF millions)

30.0 80,000 70.000 25 0. 60.000 20 0 . 50,000

150- 40.000 30.000 100- 20.000

50- 10.000

0 00, 1995 1996 1997 1998 I999 2000 2001 2002 2003 2004 1894 1695 1996 1997 1998 1989 2000 2001 2002 2003 2CC4 Total aedt aedit to private sectw -Total Deposits I/ - - I

4.52. Although commercial banks have a comfortable liquidity position, credit extended to the private sector has been low and declining since end-2000. These developments reflect commercial banks’ cautious approach in view ofthe high level ofNPLs, weaknesses in the judiciary system and overall uncertainty about the economic situation. Moreover, a large share of documentary credits (mainly trade credits) was gradually captured by other markets-particularly Dubai-after the Somali border was closed in early 200 1, and was not regained after the border reopened later that year (Figure 4.13).

4.53. On the demand side, high lending rates practiced by commercial banks discouraged the development of credit to the private sector. The spread between the domestic lending and deposit rates widened, and the lending rates have been only partially responsive to decline in international rates (Figure 4.14).

Figure 4.14 Djibouti: Domestic and Foreign Interest Rates (percent), 1990-2004 I

14 - ’>. 12 - 10 - 6- 6- 4- 2-

0 r N P v) h m 8 m m 8 m 8 8 6 N0 5 5 5 z z c z z c w w w z ...... 3-mnth LBOR - - - - US discount rate -ojibouti deposn rate Source: Djibouti authorities, US Federal Reserves.

4.53. From 1995 to 2004 the interest rate spread’ widened by more than six percentage points, from 3.4percent in 1995 to about 10 percent in 2004 (Figure 4.15). This widening is consistent with banks’ increasing provisioning for NPLs. It may also reflect declining competition among banks, as evidenced by the sharp increase in the spread in 1999 when the number of operating banks declined from five to three. The fall in international rates in the late 1990s and

’*The interest rates spread is measured by the difference between the minimum lending rate and the maximum deposit rate.

121 early 2000s was accompanied by a corresponding but partial reduction in domestic rates on deposits, consistent with indications of some increase in the risk premium and commissions charged to commercial banks’ customers in Djibouti. However, the recent increase in international rates has negatively affected Djibouti’s interest spreads. By the end of2004 the interest rate spreads hovered around 10 percent, further discouraging the demand for credit. As a result, banks deposited most oftheir liabilities abroad.

Figure 4.15 Djibouti: Interest Rates Spread, 1990-2004

10 9 8 7 6 5 4 3 2 1 0 0 N t Lo 0 N ;f 8 c c c z z I: Source: Djibouti authorities

4.54. As a result of opposite developments in deposits and credits, commercial banks accumulated large amounts of net foreign assets (Figure 4.16), which increased by at least US$45 million annually since 2001 and constituting 54 percent of GDP in 2004. The coverage ratio ofthe currency board, as measured by the ratio ofgross foreign assets ofthe BCD over the monetary base and government deposits at the BCD, has been fluctuating slightly around 115 percent since the end of2000.

Figure 4.16 Broad Money and Net Foreign Assets (NFA), 1994-2004, (DF millions)

100,000 @I @I Broad fvbney E Total NFA o NFA-ComricalBanks 80,000

60,000

40,000

20,000

0 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Source: Djibouti authorities

4.55. Aware of the need for a more active credit policy favoring private investment, and concerned about collusion among banks in setting interest rates, the authorities have recently introduced corrective measures to preserve a competitivefinancial market. These included several instructions to banks to cease coordinating interest rate structures and to publicize and report their

122 structures to the BCD. Parallel instructions were issued regarding bank fee structures, which were also subject to collusion. However, the effectiveness of these measures was limited by the BCD’s official stance of not interfering with the freedom of the commercial banks to determine credit policy.

4.5. CONCLUSIONAND POLICY IMPLICATIONS

4.56. The Djiboutifinancial sector has yet to achieve a more effective role in the economy and could be further strengthened. The level of financial intermediation in the economy remains relatively modest. Banlung soundness indicators portray a mixed picture: banks seem to be well capitalized and profitable, but NPL ratios continue to be high. There are no domestic banks to help extend banking system services beyond the major metropolitan areas or to small and medium enterprises.

4.57. To enhance the effectiveness of the financial sector and maintain its integriv, the authorities have made progress in the past 12 years in the implementation of a number of measures covering the legal framework, supervision and prudential regulations, and financial intermediation. The Anti-Money-Laundering Law adopted in 2002 provides Djibouti with a sound legal framework that should be used to ensure continuous monitoring of the integrity of the financial system. The revised banking laws, adopted in 2004, introduce clearer regulations for the operation and establishment of financial institutions and enhance transparency by institutionalizing the external auditing ofthe BCD’s accounts mechanism.

4.58. Financial products available in Djibouti are limited and consist mainly of deposits with commercial banks. Recently, one ofthe banks launched an initiative to encourage small consumer credits, focusing on credits to employees ofemployers with a good reputation, such as the port. It is too soon to draw firm conclusions from this initiative, but the bankers find the results encouraging. To ensure greater mobilization of domestic savings, another area of reform would be the development of new financial products. However, the success of such an operation hinges on establishing an adequate regulatory Eramework. Therefore, a well-sequenced work program would first identify a set ofproducts suitable for local customers and then tailor the regulatory framework to it. On both aspects of the operation, technical assistance must be sought.

4.59. Further measures are needed to help remove bottlenecks to the expansion of credit to the private sector, consolidate supervision practices, and encourage a generation of long-term savings. Such measures could include:

Putting in place a comprehensive training program for the supervisory staff at the BCD and the other financial institutions, thus reaping the full benefits of new banlung laws which go a long way toward ensuring a more efficient supervisory and regulatory environment.

Adopting an adequate legal and regulatory framework on microfinance. Expansion of microfinance activities is predicated on further reforming the regulatory and supervisory framework, while creating an environment conducive to small business development. In particular, it is important to streamline entry and exit procedures for institutions performing microfinance activities so as to ensure more regular financing of beneficiaries; reinforce capacity-building programs for the microfinance institutions and their supervisors; ensure

123 adequate capitalization of these institutions; and enhance the financial institutions law on microfinance activities and the supervision of microfinance institutions, in particular by developing special supervision procedures compatible with the ability of these institutions to monitor risks and with the quality oftheir structure and personnel qualifications. There is also a need to raise the awareness of beneficiaries about the importance of preserving the ability to take risks.

Strengthening the judicial system to enable the private sector to benefit from the comfortable liquidity position of commercial banks. The private sector has little confidence in the court system. It is important to alleviate the perception that judges lack autonomy and that the court system is slow to resolve disputes. By doing so, the great potential offered by the microfinance sector would be tapped and commercial banks would likely be more forthcoming in providing financing to the private sector.

Developing new financial products to ensure greater mobilization of domestic savings and address the financing need of small and medium enterprises, while establishing an adequate regulatory framework so as to ensure the success of this operation. The limited financial products available in Djibouti consist mainly of deposits with commercial banks. Small and medium enterprises that lack solid equity capital are unable to solicit banks for medium- and long-term financing. To finance their capital investments, they apply for short-term credits (cash facilities) that are usually renewable at high interest rates, which is detrimental to their cash flows.

Promoting Djibouti as a regional financial hub by expanding the choice of financial products by developing innovative financial products that could be needed by some regional operators; establishing agreements with regional banlung or insurances institutions, particularly in Ethiopia, including becoming a shareholder in these institutions; and promoting national banking products and services with neighboring countries and with regional economic institutions, such as COMESA.

124 CHAPTER 5. TOWARD AN EFFICIENT INFRASTRUCTURE

5.1. Djibouti’s infrastructure development is an integral part of the growth agenda. Modernizing infrastructure services will favorably affect aggregate demand; more importantly, it will contribute to strengthening the supply side of the economy. This chapter focuses on key infrastructure services with the greatest potential for private sector participation and for spillover effects to the rest ofthe economy: transport, water, energy, and telecommunications.

5.2. The efficient provision of backbone infrastructure services is key to promoting private investment. Efficient infrastructure creates opportunities for firms to connect with customers and suppliers and helps improve firm growth and productivity. Experience in many countries around the world suggests that the use of market mechanisms will, in many cases, be most conducive for developing efficient and competitive infrastructure services. This requires shifting the role of the government from direct provider to regulator ofservices.

5.3. This chapter considers the progress made so far in each infrastructure service, identifying main constraints and potentials for growth. It also outlines policy reform options aimed at promoting efficient and competitive infrastructure services.

5.1. TRANSPORTSECTOR

5.4. Djibouti’s strategic location on the Red Sea’s international shipping routes, as well as in relation to its East African hinterland, makes it an excellent regional transport hub-but its growth potential in the transport sector has not been fully e~ploited.’~Because of high transport logistics costs, a paucity of skilled workers, infrastructure deficiencies, and stiff competition from other regional ports, this potential has not been fully tapped. Still, although underperforming, the transport sector remains the mainstay ofthe economy. Port and transit services and associated road and rail links are the main activities of the services sector, which accounts for about 70 percent of GDP. Some 10,000 jobs are in a transport-related activity, including urban transport, 70 percent of them being permanent jobs.

5.5. The sharp increase in Ethiopian transit traffic, combined with the improvement of the road link to Ethiopia and the transfer of port and airport management to a private operator, dramatically stimulated port activity in recent years. As a result, port traffic increased from 1.7 million tons in 1997 to 4.5 million tons in 2002, of which 70 percent is Ethiopian transit traffic. This boom in Ethiopian transit traffic could not have been handled without improving the transport link between the port and Ethiopia (rehabilitating the International Road Comdor). Trucks on this road corridor cany more than 95 percent ofthe transit traffic to and from Ethiopia. The remaining 5 percent is carried by the almost bankrupt Chemin de Fer Djibouti-Ethiopien (CDE), a joint-national company operating the 800-kilometer rail link between Djibouti and Addis. The transfer ofthe port management to a private operator, namely Dubai Port International (DPI), in 2000 through a 20- year concession contract was a positive step toward improving the port’s efficiency. DPI has since then been modernizing the port infrastructure and the business environment by applying its

~~ ’’ This section draws on the World Bank Djibouti Transports Sector Review (2004) led by Jean-Jacques Crochet (MNSIF).

125 knowledge and expertise in port and free zone management. In 2003 the government signed another 20-year management contract with DPI to operate Djibouti International Airport.

5.6. The new private operator, DPI, has managed to improve the port’s competitiveness. Until its arrival, Djibouti had lost its transshipment traffic to competing ports in the region. In the period between the departure of the British from Aden until about 1998, the Port of Djibouti was the primary transshipment location for international trade in the area. However, the concession of the port of Aden and recent developments of new ports in the region greatly reduced Djibouti’s role. In early 1998 the ports of Aden in Yemen and Salalah in Oman opened large, modern deepwater facilities, and major shipping lines shifted their transshipment business away from Djibouti to those ports, where labor costs are about one-third of what they are in Djibouti and productivity is higher. DPI has managed to curb this downward trend. In 2002 transshipment traffic was back to its 1996-98 level of about 0.8 million tons (15 percent of total traffic) after a 10-year low of0.3 million tons in 1999-2000.

5.7. The Djibouti-Ethiopia corridor, a critical link in the port logistics chain, still suffers from inadequate maintenance, and the national road network remains inefficiently managed. The Road Maintenance Fund has been unable to ensure adequate, sustained financing for maintaining and managing the national road network. The urban transport system, concentrated in Djibouti City, where 80 percent of the population lives, is inefficient. Public transport services are scarce and of low quality. Urban transport is privately supplied in the informal sector, with an annual turnover estimated at US$60 million.

5.8. Improving the competitiveness of the port and its entire logistics chain will help mitigate the potential risk of a shvt of Ethiopian traffic away from Djibouti in the years ahead. The port is currently the only practical option for Ethiopian transit traffic. But if relations between Ethiopia and Eritrea normalize, Djibouti might lose its dominant share of the Ethiopian traffic transit market. Improving the quality and the competitiveness of the port and its logistics chain will help address this risk.

5.1.1. The Port

Current situation and performance

5.9. The Port of Djibouti lies in a favorable location along one of the busiest maritime routes in the world and at the crossroads of Asia and Europe, the Gulf region, and Africa. This location offers a strategic advantage for regional distribution and international trade. The port provides natural access to Djibouti and Ethiopia and serves as a base for both liner transshipment and regional feeder services. With 2,900 meters of quays and 14 berths, and covered and open storage areas measuring 35,000 and 380,000 square meters respectively, the port of Djibouti is among the largest on the East Afncan coast.

5.10. The facilities at the port have been designed to respond to the requirements of international trade. Its infrastructure accommodates all types of ships in total safety, at all times, and in all weather. Djibouti offers all the necessary services of pilotage, towage, ship repair (slipway), cargo handling, and warehousing (cold storage warehouses, freezing tunnels, storage yard). To the east, the new Djibouti Free Zone covers more than 14 hectares. The Port agency is responsible for providing and maintaining the port’s structures, channels, and navigation systems. It

126 also manages and operates the container terminal, but private operators handle cargo in the rest of the port.

5.1 1. The most important function of the Port of Djibouti has been and increasingly is to serve as the transit point for imports to Ethiopia and, to a very small extent, to Somalia. Djibouti has played this role for a long time. The railway connecting Djibouti with the Ethiopian capital, Addis Ababa, was constructed more than a century ago. More recently, Ethiopia lost its seaports, particularly Assab and Massawa, as a result ofthe conflict between Ethiopia and Eritrea.

5.12. Although the container throughput handled in Djibouti increased in recent years, it is still relatively small when compared with competingports in the region. Compared with container transshipment fi-om neighboring, competing ports, the throughput in Djibouti is relatively small. The boost in 2003 was mainly caused by the new transshipment service of PIL (Pacific International Lines), which moved from Aden to Djibouti in October 2002. This package was about 120,000 TEU (20-foot equivalent unit) on a years’ basis. However, when Djibouti started to experience operational problems because of the sharply increased throughput, PIL decided to return to Aden in October 2003. As a result, container throughput in the first months of 2004 was considerably lower than that realized in 2003. In 2003 Jeddah (Saudi Arabia) handled 1.78 million TEU and Salalah (Oman) 1.80 million TEU (forecast). Data on Aden for 2003 could not be retrieved, but the throughput in 2002 was 0.4 million TEU. A comparison with East African ports is presented in tables 5.1 and 5.2.

I Port Total I Djibouti 4,900,275 1,067,186 5,967,461

Mombasa 9,257,322 1.067,859 10,325,181 Dares Salaam 4,072,486 1,I 19,270 5,191,756 Note: Djibouti data are the same as in Annex 4.1 excluding “shifting and cover containers Source: World Bank, 2004

Transshipment Port Throughput Transshipment (%of throughput) Djibouti 241,122 105,361

Mombasa 380,353 49,605 13.0 43m7 1 Dar-es-Salaam 205,044 37,246 18.2

Port Louis 384,171 204,564 53.2

Port Reunion 173,208 9,103 5.3

5.13. Traffic in the port has long been under 1.5 million tons a year, with national traffic (excluding hydrocarbons) fluctuating between 200,000 and 400,000 tons. In 2003 the Port of Djibouti handled close to 6 million tons of cargo, but the 2004 throughput data show a decline of close to 19 percent-particularly in transshipment, which decreased by close to 80 percent. The main reason was the incapacity of the container terminal to handle the increased transshipment

127 cargo. This most likely explains why many shipping lines decided not to use Djibouti for transshpment containers. In 2004 transit cargo to Ethiopia also decreased by some 7 percent, mainly because of lower needs for food aid.

Figure 5.1 Container Throughput, 1994-2003

l9S4 1995 IS96 IS97 IS98 1S9S 2000 2001 2002 2003

Source: World Bank, 2004

5.14. Transit port for Ethiopia. In recent years, between 70 and 80 percent of the total throughput in Djibouti (containers, general cargo, dry bulk, and petroleum products) consisted of transit cargo destined for Ethiopia, the vast majority of it import cargo. Until 1996, transit traffic to Ethiopia was marginal-less than 150,000 tons, excluding oil. The installation of the first container gantry cranes and the subsequent development of open storage yards for the container terminal showed how right it was to focus development on subregional transshipment activities involving containerization. With the outbreak of war with Eritrea in 1998, virtually all Ethiopian traffic shifted to Djibouti, resulting in a significant upsurge in activity. From 1997 to 2002 total traffic increased by 160 percent, with transshipment activity reaching 800,000 tons in 2002 out of a total volume of 4.5 million tons, of which 69 percent was with Ethiopia. In 2003, it reached an all-time high of close to 6 million tons, with activity stimulated by the arrival ofPIL.

5.15. Transshipment port. From 1994 to 1996 transshipment grew rapidly, reaching a peak of more than 50 percent of total throughput in 1996. Since then it has fluctuated, ranging between 10 percent and 20 percent. In 2004 there was a considerable decrease. In that year transshipment cargo accounted for only about 5 percent ofthroughput.

5.16. Composition of traffic. Of the 3.5 million tons of imports, which accounted for 78 percent of all traffic in 2002, the main elements are liquids in bulk (41 percent), solids in bulk (17 percent), containers (27 percent), and bagged and palletized goods (14 percent). Of the 850,000 tons of exports, representing 19 percent of all traffic apart from fueling and watering, containers account for 84 percent, with bagged, boxed, and palletized goods accounting for 14 percent. Though already significant, containerization can be developed further. The current distribution oftraffic (47 percent in transshipments) and its share of total traffic in the port will develop as transshipment traffic and activity generated by the Doraleh project increases.

128 Figure 5.2 Major Cargo Flows, 1994-2004

1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Year

Source: World Bank, 2004

Figure 5.3 Import and Export Cargo Flows, 1994-2004

1994 1995 1996 I997 I998 I999 2000 2001 2002 2003 2004 Year

Source: World Bank, 2004

5.17. Chan ?s in the direction of tradeflows. Goods entering th I ort generally ome from the Middle East, mainly large imports of oil (54.9 percent), the Far East (14.2 percent) and the European Union (13 percent). Goods leaving the port are headed primarily for the Middle East (28.9 percent), the Far East (26.45 percent), Eastern and Southern Africa, and from Sudan to South Africa (22.25 percent). A general redirection of trade had been observed over time, with more traffic headed to the Middle East and less to Europe. In the late 1980s, trade with Europe accounted for half ofall traffic leaving Djibouti. Today it accounts for only 19 percent.

5.18. In June 2000 the government signed a 20-year management contrack' with DPJ which has had a positive impact on port performance. Although financial details are not available, it appears that the port is highly profitable. This is partly because container tariffs are high, although not out ofproportion to those of similar Afhcan ports. With about 22 moves per hour and per crane and an annual throughput of some 600 TEU per running meter of quay in 2003, the productivity of

6o The terms of the contract are as follow: DPI is paid a yearly fee to provide managerial and operational expertise. The performance obligation is that the Port of Djibouti produces a yearly financial result at least equal to that of 2002. Of the net proceeds (after a 25 percent government tax), the government and DPI are paid a certain percentage, identified as a dividend. Both parties have an obligation to reinvest 50 percent of their dividends in the port.

129 the container terminal is good. Dwell time (the time cargo can stay in port) for containers, 19 days on average in 2004, is exceedingly high, however. This is essentially because the 2003 transit agreement between Djibouti and Ethiopia allows a maximum dwell time for Ethiopian transit cargo of 180 days, of which the first 30 days are grace time for imports. This grace period is longer than that ofother ports in the region and lets Ethiopian importers use the port as a storage facility.

5.20. Port activity slowed down in 2004. There are two main reasons for the variation in cargo passing through the port. The first is the decline in the transshipment traffic owing to PIL’s transfer ofPIL its transshipment business from Djibouti to Aden in October 2003. Talung import and export activity together, this move cut Djibouti’s transshipment business from 377,559 tonnes in the third quarter of 2003 to 45,791 tonnes in the third quarter of 2004. The second reason is the decline in business other than transshipments, which fell by 18 percent. Most of this fall took place between the second and third quarters of 2004 and involved goods destined for Ethiopia, which account for 85 percent ofall cargo passing through the port.

2003 2004 % Change

Merchandise 4,791,245 3,267,886 -31.79

Djibouti 380,447 346,309 -8.97

Transit 3,061,272 2,691,936 -12.06

I Transshipment I 1,349,526 I 229,641 I -82.98 I I Petroleum I 1,544,255 I 1,485,070 I -3.83 I Total 6,335,500 4,752,956 -24.98

5.21, The government attributes the decline in port activity to bottlenecks resulting from the port’s rapid growth in recent years and its limited capacity. It hopes that the massive investment in new port facilities at Doraleh will begm to bear hitlater this year. However, the current problems illustrate the vulnerability ofthe port and the whole Djibouti economy to developments in Ethiopia unless new transshipment business can be won.

5.22. The impact of the decline in port traffic spread to other sectors and to the economy as a whole in 2004. According to the finance ministry, road transport movements fell by 18.5 percent between the third quarters of 2003 and 2004, having grown by 38.5 percent in the preceding 12 months. Freight traffic by rail between Ethiopia and Djibouti fell by 15 percent, owing partly to a reduction in Ethiopian imports of crude steel and vehicles but mainly to the cessation of food aid shipments to Ethiopia in the third quarter of2004.

130 5.23. Overall, the port is well managed. The port has been continuously modernized and is in relatively good condition. Its container terminal is well equipped for present traffic demand61 and the equipment appears well maintained. The rest of the port is better organized than comparable Afncan ports.

5.24. Nonetheless, a number of infrastructure deficiencies remain. The paving of roads, parking, and storage areas is generally poor, except in new dedicated areas. There is a lack of parlung and storage space. The aprons (the distance from the quay wall to the shed) are generally too narrow to permit efficient cargo handling operations and the circulation and parking ofvehicles. The unloading of dry bulk vessels, primarily bringmg food aid, is done with those vessels’ gear, because ofthe lack of quayside handling equipment such as grab cranes and pneumatic unloaders. Cargo that is unloaded is put directly into bagging machines on the quay. This time-consuming process makes the unloading time for dry bulk vessels long, which leads to higher shipping costs.62

5.25. The overall productivity of Djibouti’s port is fairly good by international standards. A measure of productivity for cargo handling is the number of moves per crane in the container terminal-about 22 moves per hour per crane in Djibouti, good by international standards (given the layout and dimensions of the terminal and the capacity of the vessels calling at the terminal). Another productivity indicator is the berth occupancy factor (BOF, the ratio ofthe time the berth is actually occupied by a vessel to the time the berth is available to receive vessels). In Djibouti, the BOF of the container terminal is about 0.50, which is acceptable by international standards. However, the BOF of the general cargo terminals is lower than 50 percent, indicating that the port is not used to full capacity. In 2003 Djibouti handled about 250,000 TEU across its 400 meters of quay, or some 600 TEU per running meter of quay. This is higher than other transshipment ports but still lower than regional competitors. In 2004, with the decrease in transshipment, traffic per running meter ofquay was about 500 TEU.

Figure 5.4 Djibouti’s Port Productivity in Comparative Perspective (2003)

I 1000 I

P5 800 Lc

600

0 ‘E 400 2 B 200

2 0 Djibouti Salalah Mombasa Limassol Piraeus Damietta (Oman) (Kenya) (Cyprus) (Greece) (Egypt)

Source: World Bank. 2004

5.26. Djibouti’s port tarvfs are higher than those of other international ports but lower than regional competitors. Port tariffs in Djibouti hovered around US$400 and US$500 per full imported TEU. Tariffs include port dues, stevedoring, and shore handling charges (Figure 5.5).63

61 There is a wide apron and a well-organized stack, including good markings. 62 If shore-based equipment was installed, the unloading process could be accelerated and bagging could take place at a later stage. 63 Any comparison of port costs should be read with caution. Reported official rates never correspond to the real cost of passage through the port. Shipping companies are given confidential rates that include substantial discounts, depending on

131 Table 5.4 compares the costs of a one-day stay in the ports of Aden and Djibouti. Djibouti charges about US$1,174, compared with US$1,507 in Aden. The same port services would cost US$1,468 at Jeddah and US$1,200 at Salalah, where towing and piloting are billed by the hour.

Figure 5.5 East African Port Tariffs (USD per full TEU), 2004

500 c 3 I +3 = 400 3 Lc L 300 Y)lo = 200 $ I-” 100 5 Po Djibouti Assab Reunion Mohassa Casablanca hkrseilles Genoa

Source: World Bank, 2004

Source: USAID Djibouti Investor’s Roadmap

5.27. The major technical constraint, particularly for import containers, is the long port transit time,64 which limits the port’s ability to handle further transshipment activities and increases transport costs. The reported average dwell time for import cargo in Djibouti is 19 days, which is higher than in competing ports in the region. In Bandar Abbas, Dubai (Jebel Ali), and Salalah, the free dwell time is 10 days. Ports in Europe and the United States report free dwell times of 3 and 4 days, respectively. Nonetheless, Djibouti’s average dwell time is shorter than that of other African ports: Dar-es-Salaam and Mombasa report 13 days as average dwell time, and Douala reports 18 days. A related concern is the lengthy grace periods65for imports (30 days) and exports (60 days) stipulated in the bilateral agreement signed between the governments ofDjibouti and Ethiopia. As a consequence, Djibouti’s container terminal was obliged to fill port staclung areas to full capacity.

the volume of goods processed and the number of containers handled. Moreover, not all ports have comparable features. They do not all have the same draught, which can limit their competitiveness by limiting the size of vessel they can accommodate. Dwell time, or port transit time, is defined as the average time it takes to move cargo through the port from the time it is unloaded to the time it leaves the port, and vice versa. This is the total time required to unload the cargo, to move and store it, to finalize the paperwork for administrative and financial procedures, and to clear customs. 65 The Agreement on the Utilization of the Port of Djibouti and Services to Cargo in Transit signed by the governments of Djibouti and Ethiopia in April 2003 stipulates that the dwell time of Ethiopian transit cargo is 180 days after the cargo outtum report is issued, of which the first 30 days are a grace period for imports. The grace period for exports is 60 days after arrival of the cargo in the Port of Djibouti.

132 Unable to handle further transshipment activities, Djibouti’s port has lost market share in the transshipment business.66

Figure 5.6 Average Port Transit time Compared with Other International Ports, 2004

u) 15

‘0P r 2 10 n

255

0 Qibouti Bandar JebelAli Salabh Bropean US ports Dares Mon-bassa Douala Abbas (Eubai) ports Salaam

Source: World Bank, 2004

5.28. If the average dwell time was reduced by 9 days Vrom the current 19 to 10)’ it would save about US$8 million of investments otherwise required for expanding the stacking area. Under the present system of stacking and with the equipment in the container terminal, the stackmg area requirement would decrease from about 25 hectares to about 12 hectares. Assuming a cost of US$60 per square meter of terminal (for paving, drainage, and lighting), this would represent an investment saving ofabout US$8 million.

5.29. The main reason behind the long dwell time is related to cumbersome and lengthy customs clearance procedures at the Ethiopian customs gate (Djibouti’s customs procedures are reported to be relatively fast). Importlexport companies have reported complains about: the complexity and large numbers of procedures; and the bill of lading reaching the consignee only after the cargo has been landed. A bill of lading cannot be broken up into smaller consignments, forcing importers to release the entire allotment covered by a single bill. Other complaints relate to complicated procedures for processing the letter of credit at Ethiopian banks. Duty surcharges of 120 percent levied at the Ethiopian customs gate on exiting the port require Ethiopian importers to first be sure that they have buyers for the cargo they import.

5.30. Among the region’s ports, Aden and Salalah are direct competitors as regional transshipment hubs. The development of a new container terminal at Doraleh may take many years-detailed feasibility studies are only now being conducted-and there are strong competitors in the region for container transshipment (particularly Salalah and Jeddah). During the probably long transition period, it is likely that the existing terminal could become congested and that the opportunity would be lost for developing the transshipment business, even if only on a modest scale. Box 5.1 briefly describes Djibouti’s strengths and weaknesses as compared with its two main competitors.

66A recent example is the business decision by PIL to reorient its transhipment activities from Djibouti’s port to Aden. PIL used to call at the Port of Aden but moved these activities to Djibouti because of security concerns in Yemen. However, in 2004 Djibouti’s insufficient stacking area made it impossible to meet performance standards, so PIL decided to return its transhipment activities to Aden.

133 Potential for growth

5.31. Djibouti has the potential to become a regional transshipment facility, but it lacks adequate capacity in terms of depth, quay length, and stacking area. If the throughput increased sharply, as when PIL shifted its Red Sea transshipment hub to Djibouti, new facilities would be needed. Djibouti is not really equipped to play an important role as a regional transshipment facility, for two reasons: the relatively shallow depth (maximum 12 meters) and the constrained handling facilities and storage areas.

5.32. The port of Djibouti, which has 1,143 employees and 3,000 dockworkers, can play a major role in East Africa. To do so, however, it must demonstrate an ability to capture and retain the transit and transshipment traffic that accounts for more than 85 percent of the 4.5 million tons processed in the region annually. The port facilities are becoming increasingly crowded, as indicated by a berthing occupancy rate of80 percent. Given this congestion, the pace of loading and unloading, which had reached the very good rate of 25 cycles per hour per berth, has fallen to 10- 12 cycles per hour. Better management of equipment and storage areas could lead to better anticipation of the movement ofvessels and more efficient functioning of the customs service. The customs service operates only 8 hours a day, whereas the port operates round the clock seven days a week. In the long term, the decline in productivity attributable to congestion shows that beyond the improved port management inherent in the institutional changes ofrecent years, there is an obvious need to make investments that fully justify the Doraleh project (discussed below).

Box 5.1 Djibouti’s Regional Competitors

Theport of Aden and its trafic Although currently penalized by additional insurance premiums for war, the port of Aden has managed to retain a loyal number of major shipping companies. The container terminal is operated by the Port Authority of Singapore Corporation, known in the business world for being a serious and aggressive player. A review of traffic recorded in the Aden Port Statistics is revealing. The port handled nearly 11,320,000 tons of cargo in 2001, an increase of 11 percent over 2000 (10,194,300 tons). The number of ships calling at the port was 2,332, as compared with 1,748 in 2000 and 1,782 in 1999. Lastly, the number of containers handled by the Aden Container Terminal in 2001 was 377,708 TEU, a significant increase over 2000 (248,177 TEU). These figures do not reflect the situation in 2002, when PIL left for Djibouti after the security problems it experienced with the attack on the Limbourg. It is feared that this incident, the impact of which has not yet been measured quantitatively, will affect the port’s attractiveness.

Theport of Salalah: A strong competitor Although the port of Aden is competitive in terms of container transshipments, the port of Salalah should not be underestimated. This port is operated by Maersk Sealand, the world’s largest shipper, through its affiliate APM Terminals. As of 1 January 2003 this enterprise had 326 vessels with a capacity of 818,850 TEU. Traffic in bulk goods exceeded 1 million TEU in 2002, a significant increase over 2001, and the number of ships calling in 2002 was 518, compared with 589 in 2001. During the first six months of 2003 some 279 ships called at Salalah. The port currently employs 830 people, as opposed to 930 a year ago. It is thus looking to increase its productivity.

Plans to develop the port at Salalah include the construction of two additional quays 400 meters long and a new 2,500- meter seawall. These projects will make it possible to increase annual capacity by 35 percent and accommodate much larger vessels. The goal of the port, and of Maersk, is to process 2.7 million TEU instead of the current 2 million and to be able to accommodate ships 400 meters long, capable of transporting 10,000 containers. Once completed, this project will make Salalah the largest container port in the region. Along with this expansion, a major free zone is planned; it will be open to distribution and storage companies and light industry (pharmaceuticals). These enterprises can be wholly foreign owned and will be exempt from taxes on profits as well as from import and export duties.

Run by an independent operator, the port of Djibouti thus constitutes an alternative that offers promising opportunities for development with the combined building of the new deepwater port and a free zone. With Dubai Ports International involved in the creation of the free zone and the adaptation of the transport chain (the podairport link) the project is becoming a reality. It offers good prospects for employment, provided that the transport and telecommunication sectors offer quality services at competitive prices and the state makes an effort to provide training adapted to the needs of the uroductive svstem.

134 5.33. To address the capacity constraints of the port, the government plans to construct a completely new container terminal at Doraleh. The plan is to construct a terminal 8 lulometers from the port, with a quay length of 2,000 meters (probably to be built in phases), a maximum depth of 20 meters in front ofthe quay wall, and a storage area of 700,000 square meters(a1so to be built in phases). The maximum capacity of this terminal would be 1.7 million TEU (compared with 250,000 TEU at the present terminal). The project will be built under a build-own-operate contract, and the estimated cost is about US$350 million. DPI has publicly announced that it will invest US$300 million. Negotiations with other potential shareholders to raise the remaining US50 million are reported to be taking place.

5.34. A new oil terminal is also being constructed in Doraleh. Emirates National Oil Company (ENOC) is investing US$30 million in this project. It consists of unloading facilities and a tank farm with a storage capacity of 110,000 tons in the first phase and 250,000 tons in the second phase. The terminal will also handle edible oils. PDI will contribute another US$20 million to the cost of this project. The terminal will be managed by Horizon Terminals, a joint venture of ENOC and the government of Djibouti.

5.35. The Doraleh master plan also includes an industrial and commercial free zone, which could be an asset. It could follow the successful pattern of the Dubai and Jebel Ali Free Zones. The Free Zone has an area of 400 hectares (in the first phase), to be extended to 1,000 hectares in the second. The initiative to include the Free Zone in the Doraleh project stems from the request of a number of Dubai-based companies that intend to enter the African market through this facility. The master plan also includes an area to be allocated to Asian companies wanting to penetrate the Afncan market from this location. Unlike its potential competitors, Djibouti is located on the continent of Africa, which eliminates the need for an additional sea leg from the Arabian Peninsula to the continent.

5.36. A transshipment facility in Djibouti would have to offer better services at lower costs than its direct competitors. Container shipping lines are footloose, unless they are offered a dedicated facility. The lines operate in a very competitive market and always search for the best price/quality ratio. There are already a number of transshipment ports in the regon, including Salalah, Jeddah, Aden, and soon Port Said East. Salalah and Jeddah are large players, handling millions of containers. In addition, these ports/terminals offer competitive services. They are operated by large international firms like Terminal APM International, which manages terminals in Salalah and in Port Said East. DPI is involved in handling containers in Jeddah. Aden does not presently have a large role. The Port of Singapore Authority (PSA) decided to withdraw from the management contract with the Port ofAden for political and security reasons.

5.37. Doraleh will be managed by an expert operator, DPI, which increases the chances of ensuring efficient transshipment services. The Dubai Port Authority is one of the ten best container port operators in the world. DPI has developed the port and all associated business very efficiently. DPI is an independent managedoperator, not linked to a shipping line. It tends to diversify its contracts among shipping lines and has shown interest in participating in the transit corridor and eventually in the creation of an inland container depot near Addis Ababa. This depot would complete the ports’ logistics chain. Because DPI also manages Djibouti International Airport, it also could develop the sea-air cargo business (as in Dubai).

5.38. The Doraleh project may be years away. In the short term, the Port of Djibouti needs to continue to serve Ethiopia’s growing demand for transit well. It is therefore essential to continue improving the port’s performance and capacity. Extension of the container handling facilities

135 (stacking areas and equipment, possibly new quays) may be needed if the transshipment business increases again in the Port of Djibouti and the new container terminal at Doraleh does not become available in the short term. If transshipment throughput does not increase significantly, there is less urgent need to extend the container handling facilities. The only reasons for extension then would be an increase in total throughput, in particular oftransit to/from Ethiopia and captive cargo to/from Djibouti.

5.39. For the port to serve Ethiopia’s growing demand for transit while increasing its revenues from transshipment, it will have to continue improving its performance and capacity, which requires significant investments. In March 2004, port management announced a new investment program of about US$15 million (financed from the port’s own resources), which includes additional container handling equipment, a new computer system, civil works to extend storage areas, and personnel training. In the medium term, other investments will need to be considered. The unloading ofdry bulk vessels, primarily bringing food aid, could be accelerated if shore-based equipment was installed. More importantly, additional container handling facilities (staclung areas and equipment), and, possibly, increased quay length will be necessary if transshipment demand increases.

Impact of Ethiopian transit traffic on Djibouti’s economy

5.40. Since 1998, Djibouti’s port traffic has tripled, due almost entirely to the growth of Ethiopian traffic in transit, for which Djibouti has a de facto monopoly. This has brought additional revenue and growth in employment. Indeed, GDP growth reached a peak after 1998, when port activity increased rapidly due to the surge of the Ethiopian traffic in transit. Port dues and cargo handling fees are the main direct revenues derived from this traffic (table 5.5). Conservative estimates indicate that Ethiopian traffic directly generates US$40 to US$45 million in port revenues, which is equivalent to 6.5 to 7.5 percent ofDjibouti’s GDP.

Figure 5.7 Port Traffic (metric tons), 2002/04

7,000,000

6,000,000 Total Traffic 5,000,000 7 4,000,000 w Of which Transit to fv Ethiopia Petroleum Products 3,000s000 c+ 2,000,000 Transshipment n5 I,OOO,OOO 0 2000 2001 2002 2003 2004

Source: World Bank, 2004

5.41. The rapid growth of Djibouti’s port traffic has contributed to the growth of GDP since 1998. Growth reached a peak after 1998 when port activity increased rapidly with the surge of the Ethiopian traffic in transit (Box 5.2). A substantial share of port revenues is paid into Djibouti’s budget.

5.42. Ethiopian traffic in transit brings in about 75-80percent of Djibouti’s port revenues and added value. Port activities (including domestic traffic, traffic in transit, and transshipment) are

136 estimated to contribute 10 percent of GDP. 80 percent of the output for the port is estimated to create value-added for the sector.

Handling TOTAL I Cargo Dues I Charaes I Transit Irn port 8.02 29.22 37.24 Transit Export 0.34 6.16 6.50 TOTAL 8.36 35.38 43.74

Box 5.2 Djibouti’s Port and Its Ties with Ethiopia (1998-2004)

The war between Ethiopia and Eritrea brought substantial economic benefits to Djibouti. In 1998, Ethiopia diverted to Djibouti all the trade it had previously sent through Eritrean ports, leading to a fivefold increase in goods transiting to and from Ethiopia. In addition to this greater employment and income for the port, numerous Ethiopian shipping and transit agencies opened in Djibouti-ville.

Several hundred Ethiopian trucks use the port each day, bringing additional income to the city and settlements en route to the border post ofGalafi. Improved transit trade and revenue have bolstered investments at the port. Djibouti’s plans to exclude Ethiopian and other foreign companies from stevedoring and handling activities strained relations in 200142. Although the threat was lifted in a ministerial agreement in April 2002, considerable unease remained.

Matters improved following a meeting of the Ethiop-Djiboutian joint commission in Addis Ababa in 2003, which resulted in a new series of accords for regulating the transport of goods and commercial relations between the two countries. A breakthroughwas reached in May 2004, when negotiators from the two countries, after six years of talks, agreed that Ethiopian Shipping Lines (ESL) should operate a system of door-to-door delivery, in which ESL would have a container terminal (dry port) near Addis Ababa to which it would bring cargo from Djibouti.

Source: Economist Intelligence Unit (2005).

5.43. Port transit-related employment is approximately 15percent of total employment in Djibouti. Traffic in transit is a major source ofemployment in the private sector. Employment generated by traffic in transit, including direct, indirect, and induced employment, is estimated to be between 12,000 and 17,600, as presented in detail in Annex 4.1 and in table 5.6.

I Twes of Effect on EmDlovment I NumbersofJobs I Direct 4,5004,600 Indirect 1,500-5,000 Induced 6,0004,000 I Total I 12,000-1 7,600 I Source: World Bank estimates

5.44. Three transport-related activities are particularly effective in generating employment in developing countries: trucking, port container handling, and transit-related activities (ship- and cargo-related services such as freight forwarding, ship repairs, and ship agents). In Djibouti the trucking industry generates a negligible amount of employment, because Ethiopian goods are transported almost exclusively by Ethiopian truckers. The main impact in Djibouti is from

137 handling, notably container handling, and transit-related services. This fact is important to take into account, because some of these activities could be carried out in Ethiopia instead ofDjibo~ti.~’

5.45. Overall, Ethiopia is better served by Djibouti than most other landlocked African countries are by their transitpartners. Especially important are the following advantages:

Shipping rates to Djibouti are lower than to Assab by US$20&US$300, and these rates are likely to continue because of the volume effect and the quality of port service. Port clearance time for Ethiopian goods is short. There is no customs inspection by Djibouti for Ethiopian goods, which are cleared by Ethiopian customs in the port, and there is no custom guarantee or escort system. There are also no heavy controls in the port or on the transit road by Djibouti authorities. There is no mandatory share of road transport to be supplied by Djibouti truckers (as is a general practice in Afixa). Djibouti has invested heavily to rehabilitate the transit road, which is now in very good condition for the almost exclusive benefit of Ethiopian truckers. There is no transit fee except a road maintenance fee, which is currently less than needed to cover the damage caused by road users and the opportunity cost of capital invested.

5.46. Beyond geopolitical factors, the quality of the transport infrastructure and service offered by Djibouti will be the determining factor in sustaining its current market share of Ethiopian traffic in transit.

5.1.2. Railway

Current situation and performance

5.47. The Djibouti-Ethiopia railway has experienced a continuous decline since the early 1960s. Its economic role is now marginal, and it faces severe technical andfinancial difficulties. In an attempt to revive the railway, the governments of Djibouti and Ethiopia have decided to jointly grant a concession to the private sector for operating and managing the railway.

5.48. The Etablissement du Chemin de fer Djibouti-Ethiopien (CDE) was created in 1981 to manage the railway as a binational public undertaking through a treaty6’ signed by the governments of Djibouti and Ethiopia in reference to the provisions of a General Transport Agreement ofthe same date.

5.49. The Djibouti-Ethiopia railway suffers from low technical standards and lack of maintenance. The railway is a 781-kilometer long, meter gauge, single-track railway linking the port of Djibouti to Addis Ababa. The railway line is partly located in Djibouti (106 kilometers) and partly in Ethiopia (681 hlometers). The technical design standards of the line are low (axle-load of 14 tons only, gradients up to 3 percent on some sections, sharp curves, low-weight rail). A program of €40 million, financed by the European Union, to upgrade the worst sections of the track is in preparation. After its completion (in 2006), the infrastructure will be suitable for normal (but by no means optimum) operations to again be possible.

67 It explains Djiboutians’ fears concerning the establishment of a dry port in Ethiopia. The treaty was later amended in accordance with proposals presented by the CDE Board ofDirectors.

138 5.50. The economic role of the railway remains marginal. During recent years, the volume of international freight transported by rail between the port of Djibouti and Ethiopia-the raison d’ctre of the railway-has been about 150,000 to 200,000 tons, representing only a 5-6 percent share of Ethiopian international freight. In the early 1960s, about 60 percent of Ethiopian international freight was carried by rail. Rail market share experienced a continuous decline after a paved road was built between Addis Ababa and the Eritrean port of Assab in 1962. The decline was accelerated by the opening of two roads linking Ethiopia to Djibouti (via Awash, Mille, and the Dobi junction and via Awash, Dire Dawa, and Debele). The role of the port of Djibouti has increased since Ethiopia stopped using the port of Assab after the war between Ethiopia and Eritrea in 1998. However, the railway has been unable to capture a significant share of the international freight market. The role of the railway in domestic freight traffic is also very limited, as is passenger traffic

5.5 1, The gradual loss of competitiveness of the railway to the road network has been caused by the lack of a commercial orientation, high turnover of the management team, and deficiencies in financial and human resource management. The bureaucratic style of management has been aggravated by rigidities in the public binational status of CDE. The low level of interest from the governments of Ethiopia and Djibouti in railway activities may also have played a role.

5.52. The productivity of the railway is very low, compared with other African railways. In recent years, for example, traffic density (150,000 traffic units per hlometer of line) was less than one-third that of the Abidjan-Ouagadougou railway. Staff productivity (about 42,000 traffic units per year and per employee) was barely 12 percent of that ofthe Abidjan-Ouagadougou railway. The ratio of staff cost to operating revenue (87 percent in 2003) is 2.5 times what it should be. Locomotive availability and reliability are also very poor. CDE generates heavy deficits (equivalent to more than 40 percent of its revenue in 2003) and is not financially sustainable. It faces a very serious shortage of cash. CDE is not able to purchase spare parts, notably for maintaining locomotives. This is leading to further reduction in rail transport capacity.

Growth potential and policy implications

5.53. The Djibouti-Ethiopia railway could regain an important economic role if it adopted an appropriate strategy and if its management became profit driven and efficient. The railway could indeed enjoy a competitive advantage over road transport, especially for long-distance (international), regular, large-scale freight traffic moved in unit trains. The competitive advantage of the railway in this segment of the market would be due to its ability to provide good quality service to customers and the lower operating cost of the railway compared with truckmg. The railway should concentrate its activity on this segment. Other traffic, including passenger traffic, should play a very limited role, if any. In any case, the hnd of traffic for which the railway is likely to be economically competitive with road transport would significantly exceed the current carrying capacity, given the railway’s low technical standards.

5.54. The main potential markets for the railway are petroleum products, container traffic, and other bulk commodities.

Petroleum products. The transport of petroleum products is the main potential market for the railway. Ethiopia’s consumption of petroleum products reached about 1.1 million tons in 2002. Except for petroleum products used in the northern part of Ethiopia (about 15 percent of consumption), which are likely to be imported through Port-Sudan, the remainder is and will be imported through the port of Djibouti. This traffic will be handled through the oil terminal being built in Doraleh (to be put in service by the end of2005). A

139 significant part ofthe market could be captured by the railway, provided that the railway is connected to the future oil terminal and adequate unit train loading facilities are in place; rail unloading facilities in Addis Ababa (and, possibly in Dire Dawa) are modernized to handle unit trains; and the railway operator or, preferably, petroleum marketing companies, acquire a fleet ofmodem tanker wagons.

Containers. A large part of Ethiopia’s general cargo imports and exports is containerized. However, there is almost no container traffic from the port of Djibouti to Ethiopia. Containers with freight for Ethiopia are emptied in Djibouti. Shipping lines do not allow containers to be sent to Ethiopia because they are not monitored there. For the railway to enter this market, a rail inland container terminal would have to be created in Addis Ababa, the management of which could include shipping lines. It would also be necessary for Djiboutian and Ethiopian customs to adopt adequate facilitation measures. Experience in other Afncan railways confirms that railways can be very competitive in this segment of the market. As an example, the railway is used to transport most ofthe container traffic on the Abidjan-Ouagadougou route, thanks to the excellent quality of service and despite aggressive competition from truckers. Performance on the Dakar-Bamako route is also excellent. Effective transport of traffic by the railway will mostly depend on the capacity and quality of the service offered (guaranteed transport time) and on the marketing and commercial skills ofthe railway operator.

Other bulk commodities. The railway could also play a role in transporting other bulk commodities imported by Ethiopia, notably grain, fertilizers, and building materials. These markets are, however, somewhat secondary in comparison with petroleum products and containers.

5.55. Passenger service. It is doubtful that rail passenger service can compete with bus transport where there are road connections. The railway will, however, continue have an economic role in providing basic transport service to communities in the Oromiya region of Ethiopia, which are not adequately served by road, until an acceptable road system is created for these communities. The operation of such service, if imposed on the railway operator, should be under a public service obligation (PSO) scheme (including a financial compensation system).

Policy implications

5.56. The governments of Djibouti and Ethiopia have agreed to launch a concession process for the Djibouti-Ethiopia railway. The present situation of the Djibouti-Ethiopia railway is quite similar to that of most railways in Western and Central Africa in the period from 1980 to 1990. Several of these railways (notably the Abidjan-Ouagadougou railway, the Dakar-Bamako railway, and the Cameroon railway) have since been revived through concession agreements with the private sector. The process is being conducted on behalf of the two governments by the Board of Directors of CDE. The concession process is well under way. Technical proposals were received at the end ofFebruary 2005 from RITES (India), a Comazar-Sheltham consortium (South Afnca) and an Ethiopian-South African consortium.

5.57. Concessioning the Djibouti-Ethiopia railway to a private operator is the only practical solution for revitalizing the railway’s substantial assets and providing an economic alternative to road transport on the Addis Ababa-Djibouti corridor. Indeed, based on experience in most developing countries, it is difficult to imagine that the railway would have any future as a state-mn company. In the past 15 years, railways have been concessioned in a number of countries in Afnca and Latin America. In general the experience has been positive. However, in the case of the

140 Djibouti-Ethiopia Railway, specific circumstances (low transport capacity, need for large investments, high risks, and so on) make it uncertain whether and under what conditions a concession would be feasible.

5.58. A number of economic, technical, and financial questions remain to be addressed to ensure a successful concessioningprocess:

Substantial increase in freight traffic. A main determinant of feasibility is the level of freight traffic reached after technical rehabilitation of the railway (about four years after takeover of operations by the concessionaire). The volume of traffic that will effectively be captured by the railway in its early years may be lower than presently anticipated. The concessionaire will therefore have not only to substantially increase the freight-canying capacity of the railway system but also to confront the powerful truchng industry, which is likely to defend itself against a new competitor and will benefit from planned road improvements.

Improved capacity of the railway. The capacity of the railway is about 800,000 tons per year. Increasing the capacity to 1.5 million tons per year (to ensure efficient operation) implies the creation of seven new crossing loops and the implementation of a new block system using a new telecommunication network.

Increased investments. A sizeable investment program will be necessary for the railway to reach a capacity that will allow it to operate efficiently and profitably.

Improving labor productivity. The concession company will be staffed with 1,500 employees to start (corresponding to a productivity of 208,000 traffic units per staff and per year). Staff size would gradually increase and reach 2,575 in 2010 (productivity of450,000). Based on experience with railway concessions in other parts of the world, these staffing levels may be excessive. Staff productivity of 500,000 could be reached after a few years of operation and should normally increase afterwards. Staffing levels of 1,000-1,200 at the beginning of the concession should suffice, and levels should probably not exceed 1,500 in the medium term. The growth (in constant terms) of staff remuneration also needs to be adequately estimated.

5.59. A donor-financed partial risk guarantee (PRG) scheme would enhance the feasibility of the concession. The PRG would mitigate some ofthe nontechnical, noncommercial risks incurred by the concessionaire and could prove essential for mobilizing funding. Risks that would be covered by the PRG have not yet been identified. As noted later in this report, an analysis of the main risks of the concession needs to be carried out as soon as possible. The following especially should be considered: the inability ofgovernments to finance infrastructure investment as specified in the concession agreement; traffic interruptions caused by security problems in some regions; and lack of enforcement by customs authorities ofthe customs regime of international freight transport as (and if) specified in the concession agreement.

5.1.3. Djibouti InternationalAirport

Current situation andperformance

5.60. Apart from military activities, airport traffic entails 5,000 to 5,500 movements of aircrafi: It generates a flow of 200,000 passengers, including 50,000 transit passengers, and 7,000 to

141 8,000 tons of freight. The decree of 8 January 2001 established “open slues” over the airport, lifting restrictions on the frequency of flights and facilitating the development of air links and the introduction ofservice to new destinations.

5.61. High wages increase production costs, hindering the productivity of the airport. The share of wages and salaries reached 77 percent ofvalue added generated by the airport in 2000. By contrast, the share ofwages and salaries in the value added of the port declined from 66 percent in 1996 to 36 percent in 2000 because of the sharp increase in transit activity and improved management practices.

5.62. In June 2002 the airport was privatized under a 20-year management contract awarded to DPI. The financial stability of the new operator was ensured through the allocation of facilities formerly leased to military missions, chiefly the German army, whose departure threatens to create problems for the new management. As in the case of the port, it is unfortunate that not even a minimum ofinformation was provided concerning the nature ofthe contract and the reasons for the exceptional length ofwhat would appear to be a standard management contract.

5.63. Total trafJc volume has increased since 2002. Freight rose from 13.6percent, of which more than halfconsists of daily imports of khat. Traffic is shared between the two national airlines under conditions of strict reciprocity. Passenger traffic increased by 8.5 percent while the number ofcommercial flights rose by 28 percent. This increase in commercial traffic is due to the policy of low fares pursued by Daallo Airlines and by the appearance of small airlines such as Yemenia and Ahca One. The airport has only one bonded warehouse, which is considered insufficient by the director. A review of aircraft movements in 2002 reveals the importance of military flights: there were 913 movements of military aircraft during the first quarter of 2003 and 3,332 commercial flights, or nearly 22 percent ofall first quarter traffic.

Growth potential and policy recommendations

5.64. The airport could become an important element of the multimodal logistical platform on which a regional trade center can be built. Stronger development of the airport will depend heavily on the success of the Doraleh complex. The establishment of a major free zone should attract industries and businesses from all over the region that can travel only by air and should be encouraged to relocate through a policy that would entail an increase in hotel capacity.

5.65. A satisfactory institutional link between the oversight authorities and the management of the port and airport appears necessary. It would then be possible to ensure the development ofair routes and increase links with the development ofthe regional trade center and freight movements. These efforts at institutionalizing and regulating could be undertaken jointly by all services and be headed by a single office within the Ministry of Transport. A multimodal authority that could accommodate the complementary or competing interests of the different modes of transport is a proposal that merits consideration. An advisory board that brings together all the major infrastructure users could be set up with a view to facilitating the convergence ofinterests between the private monopolies and users.

142 5.1.4. Roads

Current situation and performance

5.66. The main government institutions involved in road network management are the Ministry of Equipment and Transport (MET, MinistBre de 1 ’Equipement et des Transports); the Ministry of Housing, Urban Planning, Environment, and Spatial Planning (MHUEAT, Ministdre de 1 ’Habitat, de E’Urbanisme, de I’Environnement et de 1 ’Amdnagement du Territoire); and, to a lesser extent, the Ministry of Interior and Decentralization (MID, Ministdre de l’lntdrieur et de la Dkcentralisation), which is in charge of supervising the decentralization process initiated by the government in 2002.~~

5.67. The road sector in Djibouti consists of a national network totaling 1,193 kilometers, and a district network of 1,771 kilometers. The national network was created by a decree” in 1993 that classified 19 roads, including their urban sections. These roads are the backbone of the network, linlung the main cities and providing access to the neighboring countries of Ethiopia, Somalia, and Eritrea. In 1994, this network was divided into two categories: the national priority roads and the national secondary roads.

5.68. The national priority roads, which represent about 37percent of the network, support more than 98 percent of the traffic (in terms of vehicle-kilometers). Almost 60 percent of the priority roads arepaved, compared with none of the national secondary and district roads (Table 5.7). Current pavement works in Djibouti are concentrated on the national priority network, specifically the road between Tadjourah and Obock (62 kdometers) which should be fully reconstructed in the coming years. When that is done, about two-thirds of the national priority road network will be paved.

Classification Paved Roads Earth Roads Total National priority 430 31 1 741 National secondary 452 452 District 1,771 1,771 Urban roads 50 123 173 TOTAL 480 2,657 3,137

5.69. Average daily trafflc (ADT) volumes on Djibouti’s road network are low, except on the corridor between Djibouti and Galafi, which is used almost exclusively for international trade traffic to and from Ethiopia and carries more than 80percent of all road traffic within Djibouti Recent surveys show that ADT volumes on the national priority network (excluding the international corridor) were in most cases below 100, while ADT volumes on the national secondary network never exceeded 50. Because of the low traffic on most roads and the recent rehabilitation works on the North International Road Corridor, about 70 percent of the paved road network is in good to very good condition. Some 53 percent of the earth roads on the national priority network are in fair to good condition. The satisfactory condition of the network can also be

69 See Law 174/AN/O2/4L of 7 July 2002 decentralization and regions’ status. ’O See. Decree 93-007PWTP of 15 July 1993 modification of the national roads’ network classification.

143 explained by two favorable natural conditions: the country’s overall low rainfall; and the strength of the substructure and the availability ofgood construction material.

5.70. About 25 percent of the urban road network is in good to fair condition, while 75 percent is in bad condition without much-needed rehabilitation or reconstruction works. The length of the urban road network is 173 hlometers. These roads are mainly located within Djibouti-ville and the four regional capitals: Obock, Tadjourah, Dikhil, and Ali Sabieh. Most of this network is composed of earth roads, except in Djibouti-ville, where the main axes within the city are paved. The rapid expansion of informal settlements in the country’s urban centers, especially Djibouti- ville, has recently become a major issue for the country’s authorities. Large investments in road infrastructure and transport are now required to provide local population with adequate access to employment, education and basic health services.

Figure 5.8. General Assessment of the Current Condition of the National Priority Road Network

Paud Roads (#Oh-@ Earth Roads (311 Ian)

Rad 8%

Fat 20 %

Bad 17 ‘6

WOd 26% 38 %

Source: World Bank, 2004

5.71. With an average of 0.46 staff per kilometer of national priority roads, the MET’S workforce is excessive. The workforce grew from 224 in 1999 to 347 in 2003. Most of the new staff members were hired on contract (to circumvent civil service limitations) and may not have appropriate qualifications. MET’s employment issues in the future will be numerous. It will have to drastically reduce its current unslulled workforce and improve incentives for qualified supervisory staff. For these staff members, special attention should be given to potential conflicts of interest as they could be involved both in policy malung and procurement as government staff, and in commercial activities outside ofgovernment.

5.72. The public entity in charge or road maintenance, la Rkgie, accounts for more than 50 percent of all MET’S staff: Although the RCgie, in principle, performs only road maintenance works, it was granted one of the four construction contracts on the Dikhil-Galafi section of the International Road Corridor. Despite its extensive labor force and adequate equipment, it was unable to carry out works as required and was compelled to subcontract the most complex works to private contractors. The Regie is inefficient because the vast majority of its resources are spent on salaries rather than on works. It employs 188 workers, accounting for more than 50 percent of all MET’s staff. In contrast, recent experience has shown that the private road industry in Djibouti can carry out maintenance and construction works effectively. However this industry has too few business opportunities to allow it to develop properly, because most the resources fi-om the FER are provided to the RCgie. International evidence shows that such institutions are rarely efficient in less developed countries.

144 5.73. Resources allocated by the government to the MET for financing the road sector were more than halved (in current terms) between 1995 and 1999, from DF 343 million to DF 161 million (Figure 5.9). In 2000, however, the creation of the Road Maintenance Fund brought additional revenues to the sector, to compensate for lower government contributions. Consequently, in 2003 about DF 481 million was allocated to the road sector through MET, 40 percent more than in 1995. Figure 5.9 MET’SResources for the Road Sector

1395 1396 1997 1998 1939 2000 2001 2002 2003

Note: Donors’ contributions excluded Source:Wotld Bank ,2004

5.74. These contributions have traditionally constituted the major source of financing for the MET. Contributions from multilateral institutions (mainly the World Bank, the European Union, and the World Food Program) to the road sector also resumed after 2000, with almost US$29 million invested in new road construction and rehabilitation programs since then. These contributions have traditionally constituted the major source of financing for the MET. This situation is unlikely to change in the medium term. The donor community plans to allocate a further US$48 million to the road sector in the next few years. Almost all these funds have been or will be targeted to the national priority road network, with an emphasis on international axes.

5.75. Taken as a whole (including MHUEAT’s contributions to the urban roads but excluding all donor contributions), the resources allocated annually by the government to the road sector can be estimated at about DF 668 million, or about 50percent of all revenues (DF 1,329 million) collectedfrom road users. Fifty-six percent of such revenues come from gasoline taxes, 24 percent from the collection of the transit fee on the International Road Corridor, and the remainder from other taxes (such as for driving licenses and car registration). Although the overall budget for the road sector has increased since the inception of the FER, there are still serious imbalances in spending. Large parts of the network, notably the urban road network, still do not have sustainable financing mechanisms. The current government’s policy that gives priority to social and administrative expenses over actual road maintenance works is inefficient, especially given the increased international transit traffic and continued expansion of the road network. This situation needs to be reviewed to optimize road expenditures to meet future needs.

Truck traffic Revenue 282.1 304.6 Net profit Source: MET and Fonds d’Entretien Routier (FER)

145 5.76. The Road Maintenance Fund has not been fully operational. The Road Maintenance Fund (FER) was created by decree” in August 1999 as a special account within the government’s budget. Its mandates are to collect the road transit fees charged to heavy trucks using the International Road Conidor and to finance all maintenance works on the national road network. At its inception it was decided that priority for investments would be given to the roads that mainly support transit traffic and that expenses other than for maintenance works would not receive support from the Fund. All Fund operations on the road network were to be contracted out to private companies through competitive bidding processes. Issues with the management and operation ofthe Fund include the following: 0 Overstaffing. The technical and administrative staff currently stands at 52. This is excessive in relation to the annual budget it administers and the overall size ofthe national road network. High wage costs. Sixty percent ofrevenue is spent on salaries. The FER not only finances its own staff‘s salaries but also those of other administrations, notably the Regie. As a result, the Fund’s operating costs account for almost 77 percent of its total revenues, which is excessive. Road fund administrative expenses in Kenya and Ethiopia represent between 1 percent and 3 percent ofthose funds’ total income.72 Shortfall in revenue collection. Based on recent surveys conducted on the corridor, the FER should have been collecting annual revenues of about DF 440 million (based on an average of 450 trucks per day per direction and a standard fee ofUS$15 per truck). However, revenues in 2003 amounted to DF 305 million, or only 70 percent of expected revenue. The shortfall in revenue is explained by an ineffective toll collection system. Due to the absence ofweighbridge stations on the international road corridor, the Fund is unable to charge the haulers the load- based tariff set by the government. Between 2000 and the end of2004, a standard fee ofUS$15 per truck (based on a unit price ofUS$0.5 per ton and an average load of 30 tons per truck) was collected for each truck leaving Djibouti’s port. Although this fee was raised to US$20 in 2005, it remains low. With excessive operating costs and expenditures for activities that are not directly related to road maintenance (such as providing the Regie with new equipment), the Fund does not have the resources to fulfill its mission. Inadequate financial management system. Although the Fund should be managed as a corporate entity with detailed balance sheets, it has not yet provided complete data on its revenues and expenses.

Box 5.3 The International Road Corridor

The International Road Corridor is the backbone of the road network, linking Djibouti’sport to Addis Ababa. Within Djibouti’s temtory, it is composed of a common section of about 85 kilometers between Djibouti and the turnoff to Ali Sabieh at Doubalala, and of two main axes: the North Corridor to Galafi, at the Ethiopian border, through Dikhil, totaling about 134 kilometers, and the South Corridor through Ali Sabieh and Gu61i16 (22 kilometers). The North Comdor was recently rehabilitated under the IDA-funded InternationalRoad Comdor Rehabilitation Project.

Following the deterioration of relations between Ethiopia and Erhea in 1998, Djibouti’sport became the main gateway for traffic to and from Ethiopia. Ninety-five percent ofthis international traffic is now supported by the International Road Corridor, compared with only 5 percent by the Djibouti-Ethiopian railway. The volume of freight traffic on the InternationalComdor has fluctuated, depending on Ethiopia’s demographic and economic growth and the World Food Program’s emergency operations to Ethiopia. In the future the potential development of transit operations through the port of Assab in Eritrea, the evolution of containerization at the port of Djibouti, and the planned rehabilitation of the Djibouti-Ethiopian railway are likely to further affect the growth of freight traffic on the Corridor. Taking into account these additional external variables, this growth could be 3 percent per year in the next five years, and 2 percent thereafter. with a maior transfer of traffic from road to rail.

Decree 99-0128PlUMET of 14 August 1999 creation of the FER l2 Data from the Discussion Paper: Road Funds Revisited: A Preliminary Appraisal of the Effectiveness of “Second Generation ” Road Funds, World Bank (TWU Series), January 2002

146 Road transport services

5.77. Although the trucking industry in Djibouti is completely deregulated, it remains small. Djibouti’s truchng industry is small owing to the limited size ofthe national freight market (mainly provided for by small local firms) and the domination of Ethiopian trucking companies in the international freight traffic to and from Ethiopia. Only eight trucking operators are officially registered in Djibouti. The market leader, Djibouti 2000, operates 45 trucks and accounts for roughly 50 percent ofthe Djibouti share ofinternational trucking services. The Ethiopian operators’ competitive advantage in the market stems from several advantages, described in the following paragraphs.

5.78. The Djibouti trucking industry has higher operating costs than its Ethiopian counterpart. Despite a few fiscal advantages (such as the absence of duties on imported trucks), labor laws in Djibouti require all Djibouti trucking companies to employ local drivers, despite a limited supply ofdrivers and their overall poor qualifications in mechanics.

5.79. At the current rate of exchange, the wages of a Djibouti driver are DF 7O,OO&more than three times what an Ethiopian driver earns, (roughly DF 20,000). To these wages must be added the significant gap in employer contributions for benefits. Yet the productivity of Djibouti workers is relatively low, lower than of for Ethiopian workers. Djibouti companies also have a greater amount of downtime because drivers are not able to undertake emergency repairs along the road when necessary.

5.80. Djibouti trucks are subject to an additional tax of about US$20 at the Ethiopian border for using the country’s road network, although each transit truck (whether Ethiopian or Djibouti) leaving the Port ofDjibouti is required to pay a standard transit fee ofUS$20 per trip.

5.8 1. Customs clearance procedures at the borders and destination points in Ethiopia are reported to be more difficult and time-consuming for Djibouti truckers than for their Ethiopian competitors. This is despite a common policy of the COMESA member states (Article 85) to gradually reduce and eventually eliminate all nonphysical barriers to road transport within the Common Market, and to ensure that common carriers from other member states have the same opportunities and facilities as local carriers undertalung interstate transport operations. This situation often results in excessive interruptions and insufficient frequency of round-trip services. Djibouti 2000 reported that each of its trucks made two round-trips per month to Addis Ababa although they could make twice as many.

5.82. The vehicle jleet offering passenger services, both urban and interurban, is estimated at 2,000 (800 taxis, 500 minibuses, and 700 buses). Annual revenues generated by the passenger transport industry are estimated by the government at DF 10 billion (US$56 million). Although a law was issued in late 2002 to organize the whole sector and regulate transport services, the industry is still mainly made up of freelance passenger services that operate with little regulation. Vehicles are usually driven by drivers with inadequate training, poor driving behavior, and a high record ofroad accidents (676 out of 1,446 accidents recorded in 2004).

5.83. The old age and service life of most vehicles lead to frequent breakdowns, low operational availability, and high vehicle operating costs. Because of a lack of resources (both financial and human), the MET has not been able to enforce the regulations and provide accreditations to bus companies on the basis ofdefined routes and schedules, as it wishes to do. For

147 this reason, the government is actively seeking foreign assistance to reorganize the sector within the institutional and regulatory framework set up in 2002. Growth potential and policy recommendations

5.84. The road and urban transport sectors are key elements to make Djibouti a regional trade center and multimodal logistical hub. Both sectors could provide important job opportunities. The government should take action to increase the efficiency of these sectors. This would include reforming and restructuring the administration to better meet the sector’s needs and cope with the current budget constraints; optimizing road expenditure and giving priority to road maintenance and rehabilitation; promoting private sector participation in the international corridor; setting the financial contribution of the trucking industry to the road sector in relation to its impact on road infrastructure; and supporting the national trucking industry relative to its foreign competitors.

5.85. Clarify the institutions’ roles and duties regarding the urban road network, The roles of the MET and the MHUEAT for the urban road network need to be clarified. Ideally the MET’S responsibilities should be limited to those sections of the national roads in urban areas. Management of all other urban roads should remain the responsibility of the MHUEAT, which should contract the works to other agencies, including the MET (when the works are not part of a broader sanitation program). Should MET be given all the responsibility for the urban road network, specific agreements should be agreed upon with the MHUEAT to ensure that the ministries work closely together. This would lead to more coherent policies in urban areas and would increase the opportunity for common and cost-effective investments. Whatever the option chosen for the urban road network, it is crucial for the government to mention in the legslation the roles and duties ofboth ministries.

5.86. Reform and restructure the existing road administrations. The road sector in Djibouti is plagued by overstaffing, especially considering the limited size of the network. More than 55 percent of public spending in the road sector73 is currently devoted to operational and administrative expense, with less than 45 percent available for works. This situation has worsened as more staff have been recruited and fewer resources allocated to road maintenance. In parallel, the MET has not been able to increase its capability to manage the road network. To break with this unsatisfactory situation, the road institutions should be completely reorganized in a way that not only dramatically reduces the number of employees but also increases the level of expertise within the administration. Better work opportunities and salary schemes should be proposed to attract or keep high-qualified staff, and thus strengthen the government’s institutional, planning, and technical capacities.

5.87. Optimize the country’s road expenditures and give priority to road maintenance. The road network is a key asset for Djibouti, and the government will need to ensure that adequate resources are expended for maintenance and improvement. The government will particularly need to break the pattern of underhnding road maintenance. However, the country’s special circumstances have to be taken into account. They include severe budget constraints, very low traffic on most of the network (except the corridor) the population is concentrated in the capital city and lack of agriculture resources. The government will therefore need to assess the benefits of its road expenditures and limit them to those that are clearly justified. A preliminary analysis of expenditure needs on the national road network appears in tables 5.9 and 5.10. The budget requirements for road maintenance over the period 2006-20 represents a total amount of about DF 13.2 billion, ofwhich DF 5.8 billion is for the international corridor.

73 Based on CDE and FER’Sbudgets

148 Projects to be Financed by IFIs Projects not yet funded Classification Rehabilitation I Surfacing I Rehabilitation I Surfacing I Bridges I International 2,190 (EU) 200 2,390 Corridor

Other Paved Roads 270 (KF) 210 - 480

Earth Roads 3,100 90 3,190 TOTAL 2,460 3,100 210 290 6,060

Length’ Routine Periodic Classification TOTAL Maintenance Maintenance International Corridor 241 1,303.1 4,549.6 5,852.7 Other Paved Roads 248 1,111.5 2,762.3 3,873.8 Earth Roads 746‘ 1,809.5 1,650 3,459.5 TOTAL 1,283 4,224.1 8,961.9 13,186 I

5.88. Promote private sector participation in the Corridor. The government’s strategy to position Djibouti as a major regional hub entails maintaining the international corridor in good condition. This objective could be attained by delegating the management and maintenance of the corridor to the private sector, under a lease ~cheme.’~Another innovative option to involve the private sector in managing the Corridor could be a long-term performance-based contract. Both options could provide substantial benefits to the government and ensure the long-term preservation of the international corridor.

5.89. Performance-based maintenance and management of roads (PMMR) contracts can lead to satisfactory outcomes only if some key conditions are fully respected. Specifically, the government needs to make sure that clear and simple performance indicators are selected and defined before launching the bidding process; contract financing is secured for the entire contract period; and timely payments are made to the contractor. Considering that such contracts would be innovative in Djibouti, additional measures would also need to be taken. The Government would need to develop capacity in the private sector, through early dissemination and targeted training sessions, and increase public expertise in the management of PMMR contracts.

5.90. Support the development of micro-enterprises. Restructuring the road administration and phasing out in-house operations will allow the government to introduce sound business practices in the road sector and get better value for the resources spent on maintenance. The government should encourage the development ofmicro-enterprises for routine maintenance and the use of small-scale

74 In the National Strategyfor the Road Sector, the generic term “concession” describes the private sector participation in managing and maintaining the corridor, although the form of participation is a lease arrangement under which capital investments are shared by the government and the private operator.

149 contractors for periodic maintenance of earth roads. International e~perience'~has shown that such a system can be implemented in a relatively short period of time and bring significant benefits to the local population, because micro-enterprises use labor-intensive methods. Initial trial contracts could also be offered as a way to stimulate and encourage small-scale contractors and micro- enterprises. In Peru, outsourcing ofmaintenance works to micro-enterprises has been used for about 15 years and has led to very satisfactory outcomes.

5.91. Secure funds collected on the international corridor. Experience has shown that most funds collected through the FER were not spent on road maintenance. The Fund has been used by some administrations to finance their own projects or operational expenses. In the future, the risk of leakage will be high, except possibly under a concession arrangement. Funds from transit fee payments would accumulate but no major maintenance works would need to be conducted on the recently rehabilitated corridor for many years. Figure 5.10 summarizes the financing needs for maintenance on the corridor and the estimated revenues collected at the port (on the basis of a US$20 toll fee). Revenues would significantly exceed the financing needs on the corridor in the years when no major maintenance works are required, but would be far below needs when maintenance would be required.

Figure 5.10 Financing Needs and Resources Collected on the Road Corridor

IFlnanchg Nseds (Cumtatlve) 803 1824 5778 6958 1191 2 23575 24375 25203 2884 1 3387 1 3811 7 45354 57126 57827 58527 IB"e"Ye8 (Oumtatlve) I 585 7 I 1155 2 j 1697 8 1 2212 3 1 2699 4 I 3159 7 1 3594 1 1 4003 3 14388 0 I 4788 1 I 5167 9 I 5594 3 1 8045 8 1 6523 2 1 7027 2

Source: World Bank. 2004

5.92. Promote development of the trucking industry. For transit traffic to and from Ethiopia, the market share of Djibouti's truclung industry is only 1 percent. As noted earlier, this results mainly from an important disparity in operational cost factors with Ethiopia. Inequities in transit charges and customs clearance procedures also play a role. Considering the importance of services in Djibouti's economy and its potential for growth, action should be taken to ensure fair competition between all truclung companies operating in the corridor. The government should also pay greater attention to the current shortage of professional drivers in Djibouti, because it limits the potential development of the national truclung industry (especially because Djibouti's haulers can only hire Djiboutian dnvers) and leads to higher wages. A program is needed, in association with the Ministry of Education, to train professional drivers. Such a program would help reduce the labor unit costs in the trucking industry and increase its productivity.

75 Transport Note: "Why should Iand how Imake sure Iget the Road Maintenance implemented?" by Sally Bumingham and Natalya Stankevich - World Bank.

150 5.2. POWER AND WATER SECTORS

Electricity and water services in Djibouti are provided by public utility companies: ElectricitC de Djibouti (EDD) and the National Water Office76(ONED).

5.2.1. Electricity

Current situation and performance

5.93. Djibouti has neither fossil resources nor hydroelectric potential. This situation leads to a strong dependency on imported oil products (mostly from Saudi Arabia and Dubai), and to a high cost of energy, mahng access to modem energy particularly challenging, especially in rural areas and for the poor. The gnd provides electric power derived from thermal sources and supplied to the capital and three rural district headquarters. As of December 2001, the enterprise had 36,080 customers, served by a staff of 937 plus roughly 60 temporary staff, a moverof DF 8 billion, and a physical output of235,000 megawatt-hours.

Source: Electricit6 de Djibouti; 2004 figures are estimates

5.94. Commercial use of electricity is a rising share in overall consumption, increasing almost 10 percentage points from 1999 to 2004. Energy consumption has grown at a faster rate, rising by 58.3 percent between 1999 and 2004 compared with a 38.7 percent in energy production, thereby narrowing the gap between supply and demand. Cost of electricity in 2004 was US$0.25 per kilowatt-hour.

5.95. Fifty-seven percent of the country’s urban population has access to electricity, but only a third of the population does in the rest of the country. Access is almost nonexistent in rural settings, with the exception of a few small towns and some villages that have financed their own generators. A large majority (99.5 percent) of the urban population that has access to electricity uses it as their primary source of lighting. Though expensive, kerosene is commonly used for other domestic needs (such as coolung). The primary use ofelectricity in the summer is air conditioning. In fact, demand for electricity in Djibouti-ville almost doubles in the summer (in 2004, the peak demand in September reached 48.1 hlowatts-hours as opposed to 27.5 kilowatts-hours in February). The average household’s electricity bill is estimated to range from 25 percent (in the winter) to 50 percent (in the summer) of its average income. EDD counts about 36,000 clients, 90

76 Ofice National des Eaux de Djibouti

151 percent of them in Djibouti-ville. This is an overestimate, because a fair amount of clients are not active.

5.96. EDD equipment, most of which is obsolete, breaks down frequently. Daily multiple blackouts occur frequently, especially during the summer. A USAID study indicates that daily power cuts by EDD are a constant source of irritation for economic operators and that, on average, 25 percent ofdaily hotel room rates goes to cover the cost ofelectricity alone. Most businesses and upscale residences in Djibouti use individual stand-alone generators to compensate for EDD’s electricity generation shortages. This has meant that producers have had to invest in generators with an output that is disproportionate to their needs. Despite improvements in service reliability over the recent past, important clients such as the Doraleh Port, expected to start operations in 2006, are still opting to generate their own electricity to avoid relying on EDD’s supply for their refigeration and air conditioning needs.

5.97. On average, electric power in Djibouti is far more expensive than in comparator countries. The average tariff is four times higher than in Ethiopia. Relative to Yemen, Djibouti’s average tariff is nearly 267 percent higher. Rates are set by presidential order and published in the OfJicial Gazette. Unchanged since February 1983, they were readjusted by 11.6 percent in June 200 1.

Figure 5.11 Average Electricity Tariffs (USWkWh), 2001 $ 0.3 1 0.25

0.2 E Industry 8 0.15 ’ 0.1 0.05 @o 4 Djibouti Yemen Ethiopia Morocco

Source: EDD

5.98. High operational costs, in particular fuel cost and wage bill, contribute to high energy taryfi. Fuel purchases accounted for 30 percent of operating expenses in 2003 while wages accounted for 20 percent. Overstaffing appears to be significant. Given the nature and volume of production and the concentration of the distribution network in Djibouti-ville, many African enterprises would be able to function with half the personnel.

5.99. Low laborproductivity despite high wages. One worker per megawatt is the current norm in the most efficient power networks of industrialized countries. Djibouti’s power sector currently uses 10 workers per megawatt. Measured in relationship to number of connections, the EDD is inefficient: it currently has 3 1 connections per employee, whereas international good practice ranges from 70 to 100. The operational surcharge created by insufficient productivity is all the more costly in that Djibouti wages are relatively high.

5.100. Despite being a commercially run, state-owned enterprise led by a Board of Directors, EDD sees little benejit of autonomy: Tariffs are set by the Council ofMinisters and any personnel reduction is literally prohibited by the government’s social policy. EDD is, however, in charge of

152 its own investment planning. The Ministry of Energy is in charge of the overall sector policy, including promoting new and renewable energy sources, awarding operating licenses, and taking part in nationally, regionally, or donor-financed renewable energy initiatives. The last National Energy Plan, whose scope covered 15 years, dates back to 1987; its update is long overdue.

Potentialfor growth

5.10 1. The international complementarities of Djibouti’s and Ethiopia energy networks is well known to policy makers and provides the prospect of international trade in electricity. In summer Ethiopia produces a surplus ofenergy that could be provided at lower cost to Djibouti, because this period coincides with peak local consumption. Conversely, Ethiopia’s energy needs go unmet in winter while the limited power available in Djibouti is not entirely consumed. An international trade in energy is thus foreseeable, with each country being an importer or exporter, depending on the season. The only major upcoming change on Djibouti’s energy horizon is a 283-lulometer electricity interconnection line between Ethiopia and Djibouti, financed by the AfDB. This 230kilovolt line is expected to be completed by 2009 and will allow Djibouti to purchase Ethiopia’s large supply ofhydropower-generated electricity, especially in summer months, at reasonable rates.

5.102. Djibouti shows good potential for renewable energy development, especially in geothermal, solar, and wind power. Exploratory feasibility projects have been undertaken to various degrees of advancement, but no operational use ofthese renewable sources has been made possible and no viable short-term application has been assessed or presented. The areas with geothermal power potential are plagued with high levels of salinity and would therefore require special maintenance equipment to avoid severe scaling problems for plant and equipment.

5.2.2. Water

5.103. The quantity and quality of water resources determine Djibouti’s economic prospects and social werfare. The country’s sustainable water resources are estimated at 50 cubic meters per capita per year, compared with an average of 1,000 cubic meters per capita per year for the (water- stressed) Middle East and NorthAfiica region. The acute scarcity has become more critical because most known aquifers are overused, and quality is rapidly deteriorating. Total dissolved solids average 2,100 milligrams per liter in Djibouti-ville, compared with the WHO standard of 800 milligrams per liter. As a result, even young children have been diagnosed with kidney stones. Groundwater will have to be replenished in the medium term at costs that are estimated to be about 10 times the present production costs. Water fi-om desalinated water will likely cost on the order of US$2-US$3 per cubic meter at the production site and US$4.50 at the consumer level because of ONED’s high distribution losses and the country’s high energy costs.

5.104. Djibouti’s small population suffers from serious health problems resulting from poor quality and coverage of water supply and sanitation services. Water and sanitation-related disease explains a large share ofthe high infant and child mortality rates and a life expectancy that is below 50 years. Cholera, diarrheas, hepatitis A, and typhoid fever are all endemic.

5.105. Almost all of Djibouti’s water supply is sourced from underground wells, and most of these wells are old and close to exhaustion. Water availability in Djibouti-ville is becoming a matter ofincreasing concern, because ofthe gradual decrease in the level ofunderground water and the deteriorating quality ofthe water. Water is being pumped from deeper underground where salty seawater lies. Water salinity stands at 1,900 milligrams per liter, whereas by WHO standards it

153 should be 1,500 milligrams per liter. Water production amounts to about 39,000 cubic meters per day, which is sufficient for the country’s demands during winter; however, consumption increases to 45,000 cubic meters per day during the summer and supply is not always sufficient. Annual water consumption appears to be volatile, on the basis of 1999-2004 data. Production has declined since 1999 from 15.4 million cubic meters to 13.5 million cubic meters in 2003, though it recovered in 2004, rising to 13.8 million cubic meters.

1999 2000 2001 2002 2003 2004’ Average Production (millions of m3) 15.44 15.46 14.21 13.57 13.50 13.81 14.3 - Djibouti-ville 13.54 13.30 12.07 11.44 11.50 11.5 12.2 - Districts 1.90 2.16 2.14 2.13 2.00 2.3 2.1 Consumption (millions of m3) 8.91 9.57 9.00 9.72 8-87 9.19 9.2 - Djibouti-ville 7.62 8.29 7.77 8.46 7.76 8.1 8.0 - Districts 1.29 I 1.28 I 1.23 I 1.26 I 1.11 I 1.1 I 1.2

Number of users (‘000) - I 14.90 I 15.06 I 15.07 I 15.63 I 15.17 I

5.106. Half the population of Djibouti-ville is connected to the public water supply, and the other half draws water from connected neighbors or from public standpipes. Water service is intermittent during summer and low bacteriologxal quality makes water unsafe all year long. In rural areas it is estimated that less than 30 percent ofthe population (and their herds) have access to water within a reasonable distance, further weakening the already precarious food security. About 6 percent of the population in Djibouti-ville is connected to a partially inoperative sewerage system, and the rest to individual septic tanks or soakage pits. The rural population has no organized sanitation systems.

5.1 07. Djibouti’s public water supply systems suffer from inefficient organizational choices and a flawed institutional environment. Water losses are considerable. Physical and commercial losses were estimated at 33.5 per cent in 2004 on account ofpoor maintenance and illegal tapping into the I water supply network.

5.108. To these two types of loss must be added the lack of income inherent in the nonpayment of water bills, particularly by many government bodies. Using their capacity for political and social pressure, ministries and certain autonomous government entities resist attempts to cut off their water supply. Public workers, freed from budgetary constraints, engage in wasteful behavior. The lack of any credible sanctions for nonpayment and the absence of any sense of responsibility for ministries’ water consumption contribute to the problems faced by ONED. Centralizing payments in the Ministry of Finance has kept the originally anticipated budgetary discipline from being instituted, even though the overruns of spendthrift ministries have been somewhat curtailed. The ministries are in fact encouraged to consume more than their budget allows.

5.109. Water is more expensive in Djibouti than in its comparator countries. The price ofwater plays a role in competitiveness, but this economic good also has an obvious social dimension (it is absolutely impossible to find a substitute for water). It is therefore necessary to create conditions for universal access at reasonable prices. In addition, ONED employs 10 times as many staff per connection as the best-managed water supply and sewerage companies elsewhere in the world.

154 Figure 5.12 Average Water Tariffs (US$/cubicle meter), 2001

0.40 , I 0.35 ! 0.30 3 0.25

0.05 p 0.00 Djibouti Yemen Ethiopia Morocco Mauritius

Source: ONED

Potentialfor growth

5.110. The PRSP calculates that Djibouti would need to invest US$lOper capita each year over the 2003-15period to reach the targetsfor water supply and sanitation adopted under the MDGs. Djibouti lacks precise sector development targets in terms of service coverage and quality. However, some ofthe PRSP targets require better performance ofthe water sector: to reduce infant mortality from the 103 per thousand live births to 75 in 2006, and to reduce child mortality from 124 per thousand live births to 85 in the same year.

5.1 11. Even if it could obtain the financing, the water sector is poorly prepared to attract and invest efficiently the US$80 million necessary to meet the MDG targets and the country’s other development goals. In general, the sector has invested practically nothing in urban and rural areas in recent years because of the precarious financial situation in the public sector. Only ONED charges consumers for water supply services. However, cost recovery is low because of the poor quality of service, lack of incentives for ONED to collect tariffs, the negligence of government consumers in paying, and the poverty oflarge sections ofthe population. The budget appropriations that other sector agencies receive are barely sufficient to cover payroll, and maintenance and investments are left out.

Conclusion and policy implications

5.1 12. The high costs of public utilities coupled with the unreliability and deteriorating quality of the electricity and water services deter business operation and expansion. While the principal utility providers (EDD, ONED) arguably have adequate equipment (with the notable exception of EDD) and personnel, their management and their tariff rate structures are excessive and insufficiently flexible to accommodate the particular needs of commercial and industrial business operators. EDD’s financial condition is precarious, especially because of the large outstanding receivables owed to it by the state, public enterprises, and others. As a result the quality of infiastructure services has deteriorated. Electricity blackouts occur daily, and, on average, the monthly cost of electricity corresponds to almost one-quarter ofthe price of a single night’s stay in a hotel. Water availability is also becoming a matter of increasing concern. Technical and managerial inefficiencies in ONED have led to a gradual decrease in the level ofunderground water and deteriorating water quality.

5.1 13. Lowering electricity costs could also bring down the costs of providing water. The current high cost per kdowatt-hour makes Djibouti uncompetitive, especially when combined with the country’s large dependency on imported fuels whose prices are high and volatile. There are several

155 possibilities for addressing this situation. One is to bring transmission lines in from Ethiopia, which has lower hydroelectric power costs. Private investors might also simply develop their own electrical power generation using imported hydrocarbons under a separate policy regime. Providing energy costs at an internationally competitive rate could be critical to the future of all Djibouti companies. The tie-in with DPI should also be investigated to assess the feasibility ofusing locally generated electricity to serve the port and free trade zone.

5.1 14. A number of actions can be taken to reduce the financial constraintsfaced by EDD and ONED. The government can require EDD and ONED to bill large consumers monthly and other consumers bimonthly; enforce a policy of discontinuing supply to any customer in arrears for more than three months; and adopt a program to limit the illegal appropriation of electricity or water (system losses). 5.3. TELECOMMUNICATIONS

Current situation and performance 5.1 15. The quality and price of telecommunication services are a key factor in an economy’s competitiveness. Djibouti TLlLcommunications (DT) is responsible for providing both domestic and international telephone service. SociLtL des TLlLcommunications Internationales de Djibouti (STID) continues to provide international communications services. STID has a staff of 600, responsible for a small network of fixed lines (90 percent ofthem in the capital city), 13,000 mobile telephone subscribers (whose number has been growing rapidly since the awarding of a GSM license in May 2001), and 1,700 Internet service customers.

5.116. Djibouti enjoys one of the most modern telecommunication systems in Africa The country is connected by satellite and cable to the international automatic networksfor telephone, telex, and facsimile. Both the Sea-Me-We 2 (relaying between France, the Mediterranean basin, India, Singapore and Hong Kong) and the Sea-Me-We 3 (relaying between Germany, the Mediterranean basin, India, Korea, and Australia) fiber optic telecommunication lines pass through Djibouti.

5.117. The level of development of Djibouti’s telephone local network, although low by international standards, is higher than that of its neighbors (Yemen and Ethiopia). Telephone line density is particularly low, as is the number of portable telephone owners; in fact, a broader distribution ofmobile phones would help offset the inadequate size ofthe fixed network.

Figure 5.13. Number of Telephone Lines per 100 inhabitants: Djibouti’s vs. Comparators

@ 30 I I I I il 25

3e 20 I1998 15 I2003 10

0 I Djibouti Ethiopia Yemen Mauritius Tunisia Source

156 Notes: Hosts are the number of compufers directly connected to the worldwide internet network, PC stands for personal computer Source: ITU World Telecommunication Indicators

Potential for growth

5.1 18. The Djibouti telecommunications network could become the regional distribution hub for large volumes of data transmission, mirroring the country’s potential as the major transportation hub for the region. Making a telephone call from Djibouti can be prohibitively expensive. Access to the Internet can be slow, costly, and problematic. Yet this need not be so. One ofthe world’s principal optical fiber submarine cable systems runs through Djibouti, where there is a critical maintenance point.77 Fees are earned by Djibouti for this station but the potential for Djibouti access goes unused. Djibouti could receive revenue by buying bulk minutes and negotiating termination prices with countries such as Yemen or Ethiopia. Djibouti could take advantage of links between the undersea cable and its two earth stations but has t not done so. Djibouti could serve other AfYican countries, such as Uganda, by piclung up traffic off the satellite and feeding it into the cable. Currently there is no cable running up the eastern coast of Afixa. If there were such a cable, it would still require inland connections to landlocked countries. Therefore, Djibouti could be a regional hub for carrying digital traffic. This would require additional investment efforts.

5.1 19. Djibouti could also conceivably provide abundant bandwidth and telephone communications at very lowprices to allpotential local users. With the fiber optic cable, two earth stations, and fiber optic cable being strung around Djibouti, one can envision a smaller version of Dubai’s “Media City.” Should the Doraleh port and free trade zone become a reality, Djibouti could offer world-class access to abundant bandwidth and very low prices, as a means of attracting foreign investment and creating a cluster of firms. Making abundant bandwidth available to Djiboutians outside the zone could stimulate new economic opportunities for Djiboutian firms.

5.120. Upgrading telecommunications calls for greater private sector participation and an adequate regulatory framework, to shifl the role of the government as an independent regulator. Tahng into account the development of international competition and the corresponding reduction in prices, development of this sector cannot be financed with the profits realized by the telephone company alone. Moreover, Djibouti has the advantage of a modem infrastructure for access to international markets with the underwater fiber-optic cable. Talung advantage of this access will require political leaders to think soon about the conditions for optimizing such development. Given the level ofpublic financing available for this sector, it is unlikely that the state will retain the total control it now enjoys. It will have to define the modalities of a privatization that will also require

77 The SeWeMe3 cable runs from the United Kingdom to France and Portugal, connects with North Africa, continues through the Suez, and links Muscat and Dubai before heading onward to Pakistan, India, and Sri Lanka where it connects via multiple links to Singapore and China. The cable comes up out of the sea and into the Djibouti Telecom facility, where Alcatel engineers provide maintenance and support service.

157 establishing an independent public regulatory agency to provide oversight of all interactions in the sector.

5.4. CONCLUSION AND POLICY IMPLICATIONS

5.12 1. Djibouti’s economy could serve as a regional hub for transportation, energy, and digital information and communications. DPI may soon invest between US$300 million and US$400 million in the container port and free trade zone at Doraleh. Access to electricity in Ethiopia and Djibouti is now to be increased through regional cooperation in the energy sector. Power trade will give Ethiopians access to the relatively large electricity coverage in Djibouti, while it will give Djiboutians access to much cheaper electricity from Ethiopia. Private investment could also provide abundant and low-cost telecommunication services for Djibouti and the larger region by talung advantage ofthe strategic SeMeWe3 fiber optic cable, Djibouti’s two earth stations, and the current Chinese investment that is ringing Djibouti-ville with fiber optic cable. More efficient and cheaper electricity, telecommunications, and transportation logistics can stimulate private investment and spur growth.

5.122. However, Djibouti will not be able to tap the growth potential in its infrastructure sectors unless it addresses a central issue: the high labor costs and low productivity of the workforce. High wages and benefits for employees in the public utilities, overstaffing, and low productivity of unskilled workers are common concerns that cut across infrastructure sectors including railway, roads, energy, and water. For example, wages consume 87 percent of operating revenue in the railway sector and 60 percent in the road sector. High wage bills also contribute to large tariffs in the energy and water sectors. The wage bill of the energy company represented almost 30 percent of the wage bill of all public enterprises combined, about 8 percent of the total wage bill (government and public enterprises combined), and 2 percent of GDP (tables Iand I1 in annex 5.3). Yet, electricity tariffs are among the highest of comparator countries, averaging US$0.20 per kilowatt-hour. Overstaffing by workers who lack the necessary training to operate effectively is also a serious concern. Djibouti has 10 times as many public workers per megawatt in the energy sector as international best practice norms. Similarly, its public water utility enterprise has 10 times as many staff per connection as international best practices would dictate.

5.123. Concessioning the management of public utilities to the private sector could bring efficiency gains in terms of management, pay, and employment policies and could favor a decline in tarvfs, thereby enhancing external competitiveness. Such effects have been visible in the port and the airport. However, and despite reductions in recent months, prices for water, electricity, and telecommunications are still considered very high by producers. The poor state of telecommunications in Djibouti reflects an inability to compete at the international level that is caused by inadequate services. The same concern is reflected in the obsolete equipment operated by the state energy enterprise, which results in frequent blackouts. Any reduction of tariffs hinges on productivity gains, but also on the timely settlement of utility usage bills by the government and by public enterprises. The government can also introduce greater flexibility in the setting ofelectricity, water, and telephone tariff rates for commercial and industrial entities, and apply preferential rates for specific commercial and industrial sectors of the economy so as to encourage corresponding private sector investments.

5.124. Even if Djibouti takes advantage of opportunities for energy, bandwidth, and transport- related services and addresses the issue of high labor costs, the paucity of skilled labor raises an

158 important challenge for firm productivity. The lack of skilled workforce contributes to low labor productivity across infrastructure sectors. In the roads sector, for example, truck drivers are poorly trained in mechanics and therefore cannot undertake emergency repairs along the road. The Doraleh project calls for revitalizing the educational and training system if Djibouti workers are to benefit from job opportunities linked to port-related services. As discussed in chapter 3, the Ministry of Education should seek to promote vocational education opportunities and provide fiscal incentives for firms that offer on-the-job training opportunities.

159