UNIVERSITA CATTOLICA DEL SACRO CUORE MILANO

Transcription

1 UNIVERSITA CATTOLICA DEL SACRO CUORE MILANO Scuola di doorao in Poliica economica Ciclo XXI S.S.D.: SECS-P/01; SECS P/02; SECS-P/05; SECS-P/06 Volailiy, Deb and Growh: he Role of Fiscal Policy in Low Income Counries Overview of heory and empirical evidence using case sudies Tesi di Doorao di: Luca Bandiera Maricola: Anno Accademico 2008/09

2 UNIVERSITA CATTOLICA DEL SACRO CUORE MILANO Scuola di doorao in Poliica economica Ciclo XXI S.S.D.: SECS-P/01; SECS P/02; SECS-P/05; SECS-P/06 Volailiy, Deb and Growh: he Role of Fiscal Policy in Low Income Counries Overview of heory and empirical evidence using case sudies Coordinaore: Ch.mo Prof. Luigi CAMPIGLIO Tesi di Doorao di: Luca Bandiera Maricola: Anno Accademico 2008/09

3 Foreword The auhor of hese research papers is an Economis he World Bank responsible for he Implemenaion of he HIPC Iniiaive, and he Task Team Leader for Fiscal Susainabiliy Analysis. The hree research papers presened in his disseraion were drafed during in relaion o he auhor s work wih colleagues in he World Bank. The paper iled Kenya s Ques for Growh Sabilizaion and Reforms Bu Poliical Sabiliy? (presened in Chaper 3) was published as World Bank Policy Research Working Paper No. 4685, and represens he background for he 2008 Kenya s Counry Economic Memorandum, a World Bank publicaion. The auhor s main conribuion o his paper focuses on sovereign risk and he governmen ineremporal budge consrain GIBC (secions 3.4 and 3.5). The paper How o of Fiscal Susainabiliy in Oil-Rich Counries: The Case of Azerbaijan (presened in Chaper 4) was prepared for a conference on Fiscal Susainabiliy: Analyical Developmens and Emerging Policy Issues, organized by he Bank of Ialy, in Perugia, in April This draf is also par of a broader World Bank projec aimed a designing a fiscal susainabiliy analysis ool o be used by boh professional economiss and counry auhoriies. The auhor s main conribuion o he paper focuses on he mehodology (secion 4.2) and he simulaions (secion 4.3). The paper Growh Deerminans in Fragile Saes and Heavily Indebed Poor Counries (presened in Chaper 5) was published in he book Deb Relief and Beyond. In his paper he auhor is responsible for he pars defining he objec of he analysis (secions 5.1 and 5.2), he inerpreaion of resuls, and policy implicaions (secion 5.3). 0

4 Table of Conens 1 Inroducion Policy implicaion for LICs Deerminans of Public Deb Dynamics in LICs Composiion of public deb A Framework for Public Deb Decomposiion in LICs Daa Aggregae public deb decomposiions for 17 LICs Case Sudies: Episodes of Deb Increases and Decreases in LICs Summary of Lessons...16 References Kenya s Ques for Growh: Sabilizaion and Reforms Bu Poliical Sabiliy? Inroducion Objecives and Background Main Findings and Organizaion of he Paper Analyical Framework Kenya s Economic Liberalizaion: A Snapsho Governmen Deb Dynamics Poliical Sabiliy and he Counry Risk Premium Challenges Looking Ahead: Concluding Remarks...39 References How o of Fiscal Susainabiliy in Oil-Rich Counries: The Case of Azerbaijan Inroducion The Fiscal Susainabiliy Analysis (FSA) Tool Fiscal Susainabiliy and Managing Oil Wealh in Azerbaijan Conclusions...56 References: Drivers of Growh in Fragile Saes Has he HIPC Process Helped Fragile Counries Grow? Inroducion Characerisics of Fragile Saes Daa and Mehodology Resuls Conclusions...67 References...71 Tables Table 2.1. Cumulaive deb decomposiion for 17 LICs Table 2.2 Public Deb Reducion Episodes in Low Income Counries, 1990 o Table 2.3 Public Deb Increase Episodes in Low Income Counries, 1990 o Table 3.1. Facors Explaining Falling Indebedness 1996/ / Table 3.2. Ineres Rae Decomposiion Table 3.3. Percen of Manufacuring Firms Perceiving Issue as Being a Major or Severe Consrain o Business Table 4.1. Relaive Peroleum Dependence for Seleced Oil Producing Counries (2005) Table 4.2. Permanen Income Approach o Oil wealh (in consan 2007 U.S. dollars)

5 Table 5.1. BMA Resuls: Pooled Daa wih Counry and Subperiod Fixed Effecs Table 5.2. Human Developmen Indicaors in Fragile HIPCS, Fragile Non HIPCS, and Nonfragile Compleion Poin HIPCs Figures Figure 1.1. Growh Volailiy by Income Groups... 3 Figure 1.2. Growh of Per-Capia GDP by Income Groups... 4 Figure 2.1. Average Public Deb Trends for 17 LICs... 9 Figure 2.2. Trends in Public Deb, Figure 2.3. Public Deb Dynamics of 17 LICs Figure 2.4. Conribuion o Deb Dynamics Figure 3.1. Lengh of Horizon and Opimal Invesmen Figure 3.2. Kenya: Public Deb Figure 3.3. Primary Balance and Ineres Paymen Figure 3.4. Revenue Mobilizaion Figure 3.5. Kenya s Poliical and Credi Risk Raings Figure 3.5. Impors of Machinery and Transpor Equipmen Figure 4.1. Fiscal Susainabiliy Framework for Oil-Rich Counries Figure 4.2. An Applicaion of he PI Approach o Oil Wealh in Azerbaijan Figure 4.3. (a) PI for Two Oil Price Assumpions in Consan 2007 Dollars; and (b) Acual Non-Oil Primary Defici Figure 4.4. (a) Illusraive Fiscal Sraegies and (b) Sress Tess for Ne Deb-o-GDP raio Figure 4.4. Disribuion of Fuure Deb Socks under he Base Line Scenario Figure 4.6. PI a (a) Hisorical Variances and (b) Hisorical Variance for Oil Prices and 50% of Hisorical Variance for he Real Exchange Rae Figure 4.7. PI Spending, a Hisorical Variance for Oil Prices and a 50% of Hisorical Variance for Real Exchange Rae: Feedback from Deb Surprises o Primary Surplus Correcion Figure 5.1. Groups of Fragile Saes and HIPCs Figure 5.2. Annual Growh of per Capia Income in Fragile Saes and HIPCs, Figure 5.3. Seleced Macroeconomic Indicaors in Fragile Saes and HIPCs, Figure 5.4. CPIA and Governance in Fragile HIPCs and Non HIPCs Boxes Box 2.1. Daa for he Low Income Counries Box 3.1: Vignees from Inerviews wih Privae Secor Firms, July Box 4.1. Public Deb and OSF Asses Dynamics Appendices Appendix 1. Summary of Case Sudies Appendix 2. Decomposiion of he change in he PV of public deb Appendix 3.Basic Rule of Deb Dynamics in an Oil-Rich Counry Appendix 4 Kenya s Ques for Growh Appendix 5 World Bank Oil Price Forecass Appendix 6 Public Deb Dynamics Appendix 7 Saisical Appendix for The Drivers of Growh in Fragile Saes

6 1 Inroducion Research on he negaive effec of deb on economic growh has mainly focused on middle income counries (MICs) and marke access counries (MACs), he laer a group of MICs wih unconsrain access o inernaional capial markes (Reinhar, Rogoff and Savasano, 2003). Lieraure on he effecs of macroeconomic volailiy on growh focusing on MICs was developed in he 1990 s (Fischer, 1993). Few sudies have aemped o analyze he nexus beween deb and growh in low income counries (LICs). This collecion of papers aemps o fill his gap, by examining he applicabiliy of he concepual framework linking volailiy, deb and growh o LICs. Specifically, he papers focus on he role of fiscal policy in LICs, is relevance for achieving and mainaining deb susainabiliy deerring economic crisis and excessive growh slumps. Resuls indicae ha LICs fiscal policies do no necessarily play a key role in ensuring respec of he governmen ineremporal budge consrain and deb susainabiliy. This resul is in sark conras wih exising economic research on fiscal susainabiliy in MICs leading o differen implicaions for policy conducs in LICs. Growh in developing counries (defined as MICs and LICs) is more volaile han in high income counries (HICs), while in LICs growh is more volaile han in MICs. 1 Volailiy in LICs is also associaed wih lower per capia growh (Figure 1.1 and 2.1) Figure 1.1. Growh Volailiy by Income Groups (Period average of sandard deviaions) Low Income Counries Middle Income Counries High Income Counries Source: Global Developmen Finance, The World Bank. 1 The definiion of HICs, MICs and LICs follow he classificaion of he World Bank Global Developmen Finance,

7 Figure 1.2. Growh of Per-Capia GDP by Income Groups (10-year rolling average; median of each income group) High Income Counries Middle Income Counries Low Income Counries Source: Global Developmen Finance, The World Bank. Long erm susained growh remains elusive for mos developing counries. The growh commission (World Bank, 2008) repors ha over he las 50 years only 13 counries have managed o grow no less ha 7% per year for an uninerruped period of 25 years, and double heir income every 10 years. None of hese counries belongs o he LICs group. No surprisingly LICs, especially in Sub- Saharan African small saes and resource-rich counries, are insead singled ou as sruggling o achieve durable growh. Undoubedly, macroeconomic sabiliy, defined as he absence of exernal and inernal disequilibria, is a necessary condiion for achieving susained growh (Fischer, 1993). Volailiy is a ax on privae secor invesmen (Rodrik, 1991), and evenually on growh. Macroeconomic volailiy increases uncerainy and reduces he horizon of privae secor invesmens, disors he allocaion of resources in favor of quicker reurns providing disincenive for long-erm plans (because he long-erm expeced yields are subjec o higher risk). Lowering volailiy conribues o privae secor developmen and ulimaely o growh (Aizenman and Pino, 2005). Lieraure poins o macroeconomic volailiy as deerminan of deb inolerance, ha is he recurrence of deb crisis in MICs a much lower levels of deb compared o developed counries (Reinhar, Rogoff and Savasano, 2003; Caao and Kapur, 2006). MICs end o experience deb servicing problems a raios of oal exernal, public and privae, deb as low as percen of GDP. The hreshold is even lower for public exernal deb. More recen evidence poins o he recurrence of domesic deb crisis associaed wih high level of inflaion (Reinhar and Rogoff, 2010). Even hough, counries wih volaile GDP would prefer a higher opimal level of deb o maximize ineremporal consumpion over ime (assuming agens adverse o risk), empirical evidence shows ha deb inolerance ends o dominae due o incomplee markes and negaive effecs of deb on growh (Caao and Kapur, 2006). Therefore, macroeconomic volailiy has a long lasing negaive effec on growh and i is associaed wih a higher probabiliy of defaul episodes, which, in urn, increase volailiy and furher depress growh. Hence, counries which, according o neoclassical heory, would benefi he mos from 4

8 borrowing and ha borrow exernally because heir domesic marginal reurn on capial is higher han heir ex-ane cos of borrowing from inernaional capial markes, suffer he mos from higher deb and would insead benefi he mos from mainaining no only deb levels considered susainable, bu from limiing macroeconomic volailiy ha increases he probabiliy of fuure deb defaul for any given susainabiliy hreshold (Schaber and van Wijnbergen, 2006). Fiscal susainabiliy is deermined by he governmen ineremporal budge consrain (GIBC), where fiscal policy and macroeconomic variables (growh, inflaion, exchange rae and ineres rae) deermine he pah of public deb. Ruling ou Ponzi schemes, fiscal susainabiliy requires ha he presen value of primary balances equals curren ne deb. However a more pracical definiion would also refer o he curren fiscal sance of a governmen. Fiscal policy is considered susainable if deb raios do no increase indefiniely wihou requiring major policy change in he fuure. A sufficien condiion for solvency requires primary balances o be posiively correlaed o deb surprises, or deviaions from he period average (Bohn, 1989), bu empirical research shows ha emerging marke economies as a group exhibi a lower average primary balance (adjused for cycle flucuaions) han indusrial counries a any level of public deb (IMF, 2003). The response of he primary surplus weakens as he deb-o-gdp raio rises in emerging marke economies, and his response ceases alogeher when deb exceeds 50 percen of GDP. This suggess ha on average he conduc of fiscal policy in emerging marke economies is no consisen wih ensuring susainabiliy once public deb exceeds a hreshold of 50 percen of (IMF, 2003). A highly indebed counry is more likely o experience a macroeconomic crisis in he face of shocks and severe crisis have he mos long lasing effecs on growh (Hnakovska and Loayza, 2004). Therefore mainaining a level of deb which reduces he probabiliy of a crisis would help developing counries o achieve long erm growh or reurn o growh faser afer a crisis even. To do so, all he deerminans of he GIBC and policies available o he governmen o influence hem should be considered. Lieraure on he deb crisis, poins o he negaive effecs of high deb on growh and singles ou he imporance of fiscal policy, which canno be subsiued for by deb managemen or financial engineering in order o mainain deb susainabiliy (Gill and Pino, 2005). Bu more recen sudies also prove ha boh low growh and high deb could be sympoms of poor qualiy of pas policies and insiuions, which have srong explanaory power in explaining defaul episodes (Kray and Nehru, 2006). This resul may furher broaden he scope of acion and responsibiliy of he governmen in mainaining a sable macroeconomic environmen. The lieraure considered above has been mainly developed in he conex of marke access counries, counries which face borrowing coss deermined in he inernaional capial marke, bu similar consideraions seem also appropriae for LICs, ha mainly rely on official crediors (counry governmens) or inernaional financial insiuions (IFIs) such as he IMF, he World Bank and regional developmen banks o secure he bulk of heir exernal financing requiremens. LICs appear o be more vulnerable o exogenous shocks han MICs. They also end o have less developed domesic deb markes and herefore o rely more on exernal deb. Their qualiy of policies and insiuion is on average poorer han MICs. As a resul, high deb and high vulnerabiliy o shocks would compromise macroeconomic sabiliy and hur growh more, or more ofen, in LICs han in MICs. World Bank research (Varangis e al., 2004), has found ha LICs, especially he poores, are disproporionaely affeced by exogenous shocks, especially flucuaions in commodiy prices. 2 The 2 Exogenous shocks can be defined as commodiy price changes, naural disasers, he sudden wihdrawal of aid, he imposiion of rade barriers in parner counries, or exernaliies ha resul from conflic in a neighboring counry. 5

9 frequency and severiy of shocks are higher in LICs han in counries wih higher income, and shocks are closely correlaed wih growh. The effec of shocks on growh is asymmeric and posiive shocks do no offse negaive ones parly because negaive shocks have irreversible effecs. Negaive shocks have imporan effecs on privae consumpion, and very ofen lead o increases in povery. LICs are disproporionally affeced by shocks, because hey lack exernal cushions (such as large foreign exchange reserves) and inernal sabilizers (such as well-developed credi and risk markes) o proec hem agains shocks. In he absence of a sabilizing mechanism, an inpu shock, such as a drough for insance, ranslaes very quickly ino an oupu shock and a fall in consumpion. However, exernal shocks are found o explain only a fracion of oal oupu volailiy in LICs (Raddaz, 2007). Inernal causes, resuling from conflics, poliical insabiliy, and economic mismanagemen, are he main source of flucuaions. In addiion, highly indebed counries and counries wih bad insiuions end o be more vulnerable o commodiy price shocks and he world business cycles han oher LICs. The possibiliy o differeniae he risk using financial markes seems consrained. Research shows ha consumpion does no appear o be less volaile in counries wih more developed financial sysem (Prasad e al., 2003). More so, capial flows are pro cyclical, and dry up in bad years, increasing he negaive effecs of exernal financing during crisis years (Kaminsky, Reinhar, and Vegh, 2003). This seemed also relevan during he curren financial crisis where evidence (IMF 2010) has poined o a generalized increase in domesic deb financing in LICs, facing reducion in fiscal revenues, o finance budge defici. Therefore consisen ineremporal governmen policies seems no only he firs line of defense, bu he main one agains devasaing shock o he economy. Fiscal policy should no only look a he expeced evoluion of he GIBC bu explicily consider uncerainy in he evoluion of key macro variables and se fiscal policy in a way which would avoid crisis oucome. Empirical research on deb dynamics in MICs shows ha mos of he episodes of large deb decreases are deermined by significan increases in fiscal surpluses (Budina and Fiess, 2004). While macroeconomic shocks o ineres raes and exchange raes, ogeher wih counry specific facors, such as he recogniion of coningen liabiliies, would deermine episode of rapid deb accumulaions.. In line wih IMF (2003) lieraure his sudy shows ha deb levels in MICs rise suddenly, and in large amouns following a crisis, while deb reducions are slow. In he following secions we show ha for LICs he role of fiscal policy is more nuanced. Pruden fiscal policy, on average, conribues o he reducion of deb levels in LICs. However, in exreme episodes of deb increases and decreases, he role of fiscal policy in lowering deb is downplayed: movemens in exchange raes are he main force explaining boh rapid increases and decreases in oal public deb in LICs. This resul implies differen conclusions of economic policy abou wha would help LICs achieve deb susainabiliy. 1.1 Policy implicaion for LICs The analysis of deb dynamics in episodes of exreme increases and decreases of public deb in LICs presens a sriking difference wih respec o he similar analysis applied o MICs. If in he laer fiscal policy, ogeher wih growh, ineres raes and real exchange rae dynamics, has had a key role in reducing he level of public deb over ime, in LICs, public deb has increased and decreased in he presence of primary fiscal deficis. Assuming ha he governmen plays a cenral role in he economies of LICs and MICs alike, is policies in LICs do no have o necessarily focus on posing primary surpluses o ensure he respec of he GIBC and herefore public deb susainabiliy, a necessary condiion for achieving prolonged growh. 6

10 Policies ha improve he business environmen, poliical sabiliy, and macroeconomic sabiliy could be more relevan in LICs for ensuring deb susainabiliy. Empirical research (see Kraay and Nehru, 2006) illusraes ha conemporaneous improvemens in he qualiy of policies and insiuions reduces he risk of deb disress in LICs and make oher possible deerminans of he risk of deb disress presen in he lieraure for Marke access counries (MACs) less relevan for LICs, because of heir differen deb srucure and heir greaer vulnerabiliy o shocks. The four chapers following below presen he resuls of empirical research applied o LICs a differen levels of developmen and wih differen characerisics of heir economy. Chaper 2 defines he roles of fiscal policy in LICs during episodes of large increases and decreases of public deb. The following hree chapers use differen mehodologies o presen resuls ha may help prioriize macro policies in LICs. These papers emphasize he role of perceived poliical risk, macroeconomic volailiy and qualiy of insiuions on he effeciveness of fiscal policy o achieve and mainain fiscal susainabiliy wih posiive implicaions for growh. Chaper 3 analyses he growh performance and he reducion in public deb achieved in Kenya since 1995, hrough microeconomic and macroeconomic links. The chaper argues ha key o achieve faser growh and lower deb was no he posing of coninued fiscal surpluses, raher he perceived decrease in poliical risk, before he 2007 elecions, coupled wih reforms in he composiion of fiscal revenues. The paper presens evidence from Invesmen Climae Assessmens and discusses policy changes relaed o he composiion of revenues and expendiures o show how Kenya achieved a dramaic reducion in public borrowing coss, compared o peers, increased growh and reduced public deb, bu a he same ime increased is primary defici. Chaper 4 considers he case of Azerbaijan, an oil- and gas-rich LIC, where naural resources will be depleed in he medium erm. As for many oher resource-endowed LICs, he conduc of fiscal policy is heavily dependen on fiscal revenues deriving from he exracion of naural resources and, consequenly, by he volailiy of he price of hose resources deermined on inernaional markes. The chaper argues ha in resource-rich counries, fiscal policy should minimize he negaive effecs of price volailiy on governmen revenues, expendiures, public deb and ulimaely he economy. A number of fiscal rules, including a permanen income rule, are discussed in a sochasic model o address he problem of uncerainy of fiscal policy relaed o he flucuaion of revenues from naural resources. The use of fiscal rules is imporan in resource rich LICs o suppor economic diversificaion, and o deermine he provision of public goods, including physical infrasrucure wih unclear financial rae of reurn. Finally, Chaper 5 focuses on growh deerminans on hose LICs where leaving condiions could be considered a he boom of any ranking of human developmen (Collier, 2007). The chaper applies Bayesian Moving Average (BMA) echnique o allow for model and parameer uncerainy in he deerminaion of growh covariaes. The analysis considers wo groups of parly overlapping counries, which are seldom sudied for lack of readily available daa: Heavily Indebed Poor Counries (HIPCs, see for example World Bank, 2009) and Fragile Saes (World Bank, 2007). BMA regressions find ha hose counries ha (i) graduaed under he HIPC Iniiaive and obained deb relief and (ii) improved he qualiy of heir policies and insiuions also benefied from a growh dividend resuling from furher deb reducion. Based on a limied sample, he analysis idenifies hreshold effecs relaed o he level of deb and he qualiy of policies for fiscal policy o become effecive in fragile heavily indebed LICs. 7

11 2 Deerminans of Public Deb Dynamics in LICs This secion examines he deerminans of public deb dynamics LICs over he period Deb dynamics in LICs is influenced no only by macroeconomic variables bu also by he composiion of deb and he provision of deb relief over ime. The deerminans of deb dynamics in LICs are described hrough a modified version of he exbook deb decomposiion formula 3 o ake ino accoun he average level of concessionaliy of he exernal deb porfolio in LICs and he erms of new borrowing 4. The paper also examines in deail he major shocks ha led o deb crisis in he LICs under consideraion. The sudy considers a sample of 17 LICs in Africa (13), Asia (3) and Lain America (1), for which daa on public exernal as well as domesic deb were available for he period Deerminans of exreme episodes of deb accumulaion and decreases are analyzed on he basis of 7 counry case sudies (Ehiopia, Ghana, Nigeria, Bolivia, Vienam, Laos and Uganda) 6. The average public deb-o-gdp raio in hese LICs amouned o 131 percen in 1990 and declined o 115 percen in 2002 (Figure 2.1). In presen value (PV) erms, he raio of public deb o GDP declined from 83 o 75 percen. Deb dynamics can be described by wo successive waves of increase and decrease in erms of GDP. The PV of public deb in percen of GDP increased o 86 percen in 1992 and declined seadily o 66 percen in 1996, before increasing again o 78 percen in 1998 and declining o 60 percen in In 2002, deb raios in he 17 counries included in he analysis has increased by 6 percenage poins of GDP o 66 percen. For he 7 counries for which deailed sudies were done, deb daa o end-2003 were available and signaled a coninued increase he deb raios, compared o The average rend hides ample differences among counries in he sample. Counries are presened according o decreasing raios of PV of deb o GDP in 2002, in Figure 2.2. In 9 counries ou of 17, he PV of exernal deb o GDP raio declined over he period , bu in 14 ou of 17 counries domesic deb increased from an average of 7.6 percen of GDP in 1990 o an average of 14.0 percen in 2002, iling he PV of oal public deb upwards in 9 counries ou of 17. Differen deb dynamics characerize HIPC and non-hipc counries included in he analysis. For he 11 counries ha obained relief under he HIPC Iniiaive by 2002, he PV of public exernal deb remained virually unchanged from 1990 o 2002 a abou 67 percen of GDP, bu he domesic deb increased markedly from 9 o 16 percen of GDP. However, in 6 HIPCs he PV of public deb decreased in In he 5 non-hipc LICs, he PV of exernal deb in 2002 was more han 20 percenage poins of GDP lower compared o he level in However, excluding Vienam, which has a PV of deb o GDP raio of 284 percen in 1990, he average raio for non-hipcs LICs signals a marked increased in he PV of exernal deb wih respec o GDP, from 46 percen in 1990 o 71 percen in Domesic deb remained sable beween percen of GDP and only 1 counry, Nigeria, achieved a subsanial reducion in he PV of public deb over he period. 3 See for example Burnside, The Appendix 2 develops he accouning framework used for he decomposiion of he changes in he PV of deb-o-gdp raio. 5 The 17 Low Income Counries include: Burundi, Ehiopia, The Gambia, Ghana, Lesoho, Madagascar, Malawi, Nigeria, Rwanda, Sierra Leone, Tanzania, Zambia, Bolivia, Vienam, Laos, Bangladesh and Uganda. 6 See Bandiera

12 Figure 2.1. Average Public Deb Trends for 17 LICs (In percen of GDP) NPV of exernal public deb Domesic public deb Sources: see Box Composiion of public deb The composiion of public deb in LICs is analyzed wih respec o he crediors, he level of concessionaliy and he share of domesic and exernal deb in he period In he 17 LICs, exernal public deb is composed mainly by concessional deb, owed o official bilaeral and mulilaeral crediors. Such deb has gradually subsiued commercial deb or deb exended by expor credi agencies ha used o dominae in he 70s. Over he period considered in he analysis, he share of concessional deb has increased from 54 percen of oal deb in 1990 o 76 percen in LICs owed more han 50 percen of heir deb o mulilaeral agencies and 45 percen o official bilaeral crediors in From 1990 o 2003, he share of mulilaeral crediors in oal LICs deb increased from 26 o 51 percen while he share of bilaeral official crediors decreased from 55 o 45 percen. Over he period, commercial crediors share has decreased from 18 percen o less han 5 percen of oal deb. Emblemaic is he case of Bolivia, where commercial crediors represened 46 percen of oal deb ousanding in Their share has consanly decreased unil i dropped below 1 percen in Domesic deb has doubled from 1990 o 2002, from 7 o 14 percen of GDP, bu, wih he noable excepion of Ghana and The Gambia, domesic deb has increased by no more han 10 percen of GDP over he period considered. Domesic deb is usually shor erm, as opposed o he exernal deb, and carries an ex-ane higher real ineres rae, compared o exernal concessional deb. In LICs domesic deb remains a secondary problem wih respec o exernal deb, bu he shor mauriy srucure and he high real ineres rae represen a risk for he governmen because of he frequen need o roll over he enire sock of deb. 9

13 Figure 2.2. Trends in Public Deb, (In percen of GDP) Domesic deb NPV of exernal publicd deb Eh Bdi Gmb Gha Nig Mlw Sle Zmb Lao Bol Mdg Tnz Ls Bng Rwa Uga Vn Average Sources: see Box 2.1. Noe: Counries in his figure are sored by oal deb o GDP raios in A Framework for Public Deb Decomposiion in LICs We analyze public deb rends beween 1990 and 2002 (2003 in he 7 case sudies) by decomposing pas changes in he PV of public deb-o-gdp raios ino a number of explanaory facors. Equaion (1) is he difference equaion for he PV of public deb o GDP raio, derived in he Appendix 2: g pv pd ndfs d (1 g ) dr (1 )( d f d f 1 ) ( 1 * d 1 i 1( ) [ ] * (1 g ) 1 (1 ) (1 ) 1 ) d f 1 ^ oher _ facors 1 RXR (1 )(1 RXR ) (1 g ) d 1 (1) where pv is he PV of public deb-o-gdp raio (defined as he sum of he PV of public and publicly guaraneed exernal deb and nominal domesic public deb), pd is he primary defici as a share of GDP, g is he real GDP growh rae, î is he weighed average of domesic and foreign ineres raes 7, is domesic inflaion rae (proxied by he percenage change in GDP deflaor), * is he US inflaion rae (proxied by he percenage change in US GDP deflaor), α is he share of foreign currency ( 1 ) i i (1 s ) iˆ is average nominal ineres on public deb. In pracice, i is calculaed as an effecive 7 d f ineres rae: he raio of ineres paymens on deb divided by he previous period sock of deb. 10

14 denominaed deb in oal public deb, RXR is he change in (bilaeral, US dollar per local currency uni) real exchange rae 8, dr is he PV of deb relief, is a funcion of he ineres rae, discoun rae, mauriy and grace period of deb and d f is he public and publicly guaraneed (PPG) exernal deb in erms of GDP. Equaion (1) decomposes he change in he PV of public deb o GDP raio ino he componens aribuable o he 6 facors shown in Table 2.1: (i) he primary fiscal balance ne of seignorage; (ii) real GDP growh; (iii) he implici real ineres rae; (iv) he real exchange rae; (v) deb relief; and (vi) he concessionaliy of new ne borrowing and he change in he erms of he ousanding sock of deb. The las erm, oher facors, is obained as he acual change in he deb o-gdp raio minus he sum of (i) o (vi). I includes: he recogniion of coningen liabiliies ne of privaizaion proceeds, he impac of grans, valuaion changes of oher foreign currencies vis-à-vis he U.S. dollar excepional financing, such as reserve accumulaion and measuremen errors. Below, we provide findings from he aggregae deb decomposiion based on equaion (1) for 17 LICs and from more deailed case sudies. 2.3 Daa The main challenge for a sudy on public deb dynamics in LICs is he qualiy of daa. 9 Significan effor was aken o consruc a consisen daabase over ime and across counries (See Box 2.1), ye he main concern lies in he coverage and definiion of public deb. A meaningful comparison across counries requires he use of a uniform measure of public deb. Public and exernal debs include long-erm deb of he naional governmen, sub-naional governmens, and public enerprises. However, here is a grea heerogeneiy in he coverage of domesic deb across counries. Mos of he counries repor only cenral governmen deb even hough sub-naional eniies can issue deb. Ohers have daa available for he consolidaed public secor including he cenral bank. Only a few counries repor comprehensive gross public deb figures, including public guaranees and pension liabiliies. A relaed saisical issue is he maching and comparabiliy of public deb repors wih he reporing of governmen finance saisics. Furher, long series of deb daa are no widely available for many LICs. Informaion on exernal public deb is generally more readily available han domesic deb. Conflicing deb figures 1 (1 s 8 RXR is defined as )(1 *) wih RXR>0 denoing a real exchange rae appreciaion. 1 RXR 1 9 See for example World Developmen Repor 2009, of he World Bank, which repors exernal deb daa for 120 of he 122 counries repored in he publicaion, bu domesic public deb daa for only 40 counries. All 40 HIPCs are repored wih missing daa and mos of he remaining 42 counries wihou available domesic deb daa are LICs. 11

15 may be found even for idenical coverage and definiions. 10 Box 2.1. Daa for he Low Income Counries This secion considers he public deb in 17 LICs. The deb daa used he mos comprehensive measure of public secor available. The sample includes daa from 1990 o Toal nominal sock and PV of public deb is consruced on a gross basis as he sum of : (a) he nominal sock or PV of exernal public and publicly guaraneed deb and (b) nominal domesic deb. Domesic deb includes only he ousanding, direc, fixed-erm conracual obligaions of governmen held by residens. The definiion does no include governmen guaranies (as opposed o he definiion of PPG exernal public deb), coningen liabiliies and floaing deb resuling from unpaid governmen obligaions, unless convered ino a fixed-erm conracual obligaion. The sources for public deb include IMF World Economic Oulook (WEO), various IMF Counry Repors, Global Developmen Finance (GDF), and he inerne sources of he counries cenral banks and reasuries. Daa on public secor fiscal balances (primary and overall) and ineres paymens are aken from IMF WEO and various IMF Counry Repors, and World Bank GDF. GDP, GDP deflaor, and nominal exchange raes are aken from IMF WEO, IMF IFS, and World Bank GDF. Daa on privaizaion and recogniion of coningen or hidden liabiliies are compiled from IMF Counry Repors. Deb relief is defined as he PV of he reducion in he sock of exernal deb as a resul of deb cancellaion and deb buyback and as he gran elemen of flow rescheduling of exising deb. 2.4 Aggregae public deb decomposiions for 17 LICs The decomposiion of changes in he average PV of deb-o-gdp raios for he 17 LICs during he period indicaes ha (Figure 2.3): (i) primary fiscal surpluses, real GDP growh, real ineres raes have conribued o a reducion in he average public deb-o-gdp raio hroughou he whole period; (ii) real exchange raes conribued o he increase in he PV of deb raios in he period considered; and (iii) he conribuion of deb relief o he decline in deb-o-gdp increased markedly in 2000, when mos of HIPC counries sared receiving deb relief. The aggregae cumulaive deb decomposiion was divided ino periods of decreases and increases in he PV of public deb-o-gdp raio (Table 2.1). The wo periods of increase in he PV of public deb-o-gdp raio lased only 1 year each and he real exchange rae depreciaion was he only idenified componen conribuing o he increase in deb raios in boh periods. The depreciaion in he real exchange rae was caused by a large depreciaion in he nominal exchange raes in 1992 and in In , a worsening of repaymen erms, as indicaed by a posiive value of he conribuion in he changes in he erms srucure, indicae ha he increase in he PV of public deb raios was parly caused by he accumulaion of arrears from 13 percen of oal deb in 1991 o 18 percen in Conversely, he increase in he PV of deb o GDP raio in was accompanied by a furher increase in he level of concessionaliy of new borrowing, indicaing ha he second period of deb increase was mainly caused by an increase in he volume of official concessional lending. 10 See also IMF (2003), which repors similar problems wih deb daa. 11 PV assumes ha he presen value of an increase in arrears is equal o he arrears nominal increase. A increase in arrears decreases he gran elemen of foreign deb. 12

16 Figure 2.3. Public Deb Dynamics of 17 LICs Primary Defici Real ineres rae Residual Deb relief Real GDP growh Real exchange rae change Change in he erms srucure Change in NPV of public secor deb Noe: The inerpreaion of his char is as follows. Each column represens he conribuion of each facor in he deb decomposiion of he year on year change in he PV of deb/gdp raio. Iems above he zero line conribue o an increase in he PV of deb/gdp raio, while iems below he line conribue o a reducion in he raio. As an example, a negaive sign for Conribuion from real exchange rae change in a given year indicaes ha real appreciaion during ha year conribued o a reducion in he PV of deb/gdp raio. On he same oken, a posiive sign for he Real exchange rae change indicaes ha a real depreciaion increased he PV of deb raio during ha year. Table 2.1. Cumulaive deb decomposiion for 17 LICs Public Deb Decomposiion Percen of GDP Toal Change in PV of Public Deb Primary Fiscal Balance Real GDP Growh Rae Real Ineres Rae Real Exchange Rae Deb Relief Change in Terms Srucure During and , real GDP growh was he main facor conribuing o he decrease of he raio of he PV of public deb o GDP (Figure 2.4). Fiscal surpluses increased significanly wih respec o previous periods, o an average of 2.6 percen per year in , compared o an average of 1.4 percen before (Figure 2.4). Deb relief conribued o 24.2 percen of decline in he PV of deb o GDP in wih he larges conribuion in when mos of HIPCs sared receiving sock of deb relief and he Paris Club agreed o gran deb relief o Nigeria. Real ineres raes remained low, given ha mos of he deb was no conraced a marke erms. 13

17 Figure 2.4. Conribuion o Deb Dynamics (In percen of GDP) Primary Defici Real exchange rae change Real GDP growh Deb relief 2.5 Case Sudies: Episodes of Deb Increases and Decreases in LICs The 7 individual case sudies 12 show ha mos counries experienced episodes of sharp increases and progressive reducions in deb-o-gdp raios beween 1990 and Based on he case sudies, we idenified 10 episodes of large deb reducions and 6 episodes of large deb increases. Table 2.2 and Table 2.3 rank he mos imporan facors ha conribued o changes in deb raios during hese episodes. For he episodes of large deb reducions, ranging beween 30 and 190 percenage poins of GDP, we found ha: (i) all episodes involved GDP growh as one of he main conribuing facors o he decline, (ii) no episode involved primary surpluses as main facor, while in 8 ou of 10 episodes of fiscal deficis weighed agains he decrease in public deb, (iii) 60 percen of deb reducion episodes involved real exchange rae appreciaions and (iv) 50 percen of episodes involved eiher deb relief or changes in he erms srucure of exernal deb or boh. For he 6 episodes of large deb accumulaions, ranging beween 18 and 109 percenage poins of GDP, we found ha: (i) 5 ou of 6 episodes involved primary deficis; (ii) 5 ou of 6 deb increases ook place in correspondence wih sharp real exchange rae depreciaions and; (iii) real GDP growh couneraced he increase in he raio of he PV of public deb o GDP. Even hough, on average, LICs ran fiscal surpluses during , fiscal surpluses were no a key facor in reducing he deb burden in LICs in exreme episodes. Real exchange rae appreciaion appears o have been a key facor in he periods of decrease or increase of he deb raios, because of he high proporion of public deb denominaed in foreign currency. Real ineres raes appear no o represen an imporan facor in he explanaion of deb dynamics, because mos of he exernal deb held by LICs had fixed ineres rae. Therefore, he ineres rae srucure of public deb was no significanly affeced by changes in domesic ineres raes. 12 The case sudies are Ehiopia, Ghana, Nigeria, Bolivia, Vienam, Laos and Uganda, see Appendix 1. 14

18 Table 2.2 Public Deb Reducion Episodes in Low Income Counries, 1990 o 2003 Counry, Time Period Toal Change (% of GDP) Iniial level (% of GDP) Main Conribuing Facors (Reducers; % of GDP) Main Conribuing Facors (Increasers; % of GDP) Bolivia, GDP growh ( 10.1) Primary defici (3.0) Change in erms ( 9.1) Deb Relief (-7.6) Ehiopia, GDP growh (-34.3) Primary defici (3.1) Real Ineres rae (-6.9) Oher facors (-6.6) Ehiopia, Deb Relief (-79.9) Change in erms (-18.1) Primary defici (17.4) Real exchange rae (11.6) GDP growh (-17.0) Ghana, Real Exchange Rae (-53.7) Oher facors (12.5) GDP growh (-20.6) Change in erms (-14.2) Lao PDR, Real Exchange Rae (-69.7) GDP growh (-51.2) Gran elemen of deb (88.9) Primary defici (17.9) Real ineres rae (-11.3) Nigeria, Oil revenues (-38.9) Nonoil Primary surplus (17.5) Real Exchange rae (-36.1) Real Ineres Rae (-15.6) GDP growh (-10.2) Nigeria, Oil revenues (-99.2) GDP growh (-17.4) Real exchange rae (-8.1) Nonoil Primary surplus (77.9) Oher facors (18.0) Real ineres rae (8.0) Uganda, Real Exchange Rae (-43.3) GDP growh (-25.0) Gran elemen of deb (17.9) Oher facors (16.2) Deb Relief (-4.8) Vienam, Real exchange rae (-199.5) GDP growh (-98.8) Gran elemen of deb (133.1) Primary defici (12.7) Real ineres rae (-24.6) Vienam, Deb Relief (-29.2) GDP growh (-14.2) Change in erms (-12.6) Off budge facors (15.2) Primary defici (9.9) 15

19 Table 2.3 Public Deb Increase Episodes in Low Income Counries, 1990 o 2003 Counry, Time Period Ehiopia, Toal Change Terminal level Main Conribuing Facors (% of GDP) (% of GDP) (Increasers, % of GDP) Real exchange rae (108.8) Primary defici (20.8) Main Conribuing Facors (Miigaors, % of GDP) Real Ineres rae (-22.8) GDP growh (-10.9) Ghana, Real Ineres rae (35.2) Primary defici (12.9) Oher facors (13.2) GDP growh (-9.8) Gran elemen (-7.7) Ghana, Real exchange rae (85.2) Gran elemen (-26.3) Lao PDR, Real exchange rae (127.2) Primary defici (12.4) Nigeria, Real exchange rae (59.9) Uganda, Nonoil Primary defici (37.4) Real exchange rae (24.7) Primary defici (17.0) GDP growh (-9.2) Gran elemen (-57.2) GDP growh (-18.5) Oil revenues (-64.5) Real ineres rae (-8.7) GDP growh (-23.3) Deb resrucuring (-15.6) 2.6 Summary of Lessons Deb dynamics over he 13 years considered shows a high degree of heerogeneiy across LICs, which underscores he imporance of counry-specific facors. These include he fiscal sance, he srucure of he economy, he exchange rae sysem, he composiion and srucure of he deb and deb managemen policies. The secions below highligh he main findings of he 7 case sudies. a) The Role of Fiscal Policy Primary fiscal surpluses are no a key facor in he reducion of he raios of he PV of deb o GDP in he LICs considered in he case sudies. No counries managed o run large fiscal surpluses for a prolonged period of ime during he period considered. The main reasons underlying mixed fiscal performance are he high level of planned public expendiure, he realizaion of coningen liabiliies, weak revenue collecion capaciy and he inabiliy o implemen counercyclical policies. (1) The public secor in LICs ofen finances a large share of domesic invesmens and social expendiures as per he povery reducion documens. For his reason, he imely delivery of commied foreign financing becomes key o avoiding budge overruns. (2) Coss associaed wih financing of inefficien SOEs or publicly owned commercial banks, as in he case of Bolivia, Ghana, Laos and Vienam, increase fiscal oulays ofen no covered by foreign financing. (3) Revenue collecion is usually weak, as in Laos and Ehiopia in he firs half of he 90 s.and collecion relies excessively on expor ariffs and non-ax revenues from raw maerial and SOEs. (4) Lasly, LICs are unable o run counercyclical fiscal policies and he level of expendiure is 16

20 ofen rigidly igh o commimens made during good years, as in he case of Bolivia s pension reform and expendiure commimens by regional governmens in Nigeria, where fiscal policy is linked o curren oil revenues. Consisen deb reducion in LICs were achieved also in presence of prolonged fiscal deficis, bu large deficis financed hrough deb were likely o significanly increase he deb burden, as in he case of Ghana and Nigeria over a shor period of ime. b) The Role of auomaic deb dynamics LICs public deb is mainly long-erm exernal deb a low fixed ineres raes. For his reason, he mos relevan auomaic deb dynamics depends on he real economic growh rae and he real exchange rae, while he real ineres rae, or more precisely, he difference beween he real ineres rae and he real economic growh rae, has a secondary effec on deb dynamics. Our individual case sudies show ha during he enire period considered all counries mainained posiive real growh and low ineres raes even in he mids of exogenous shocks. However, erms of rade shocks or he reducion in demand of expors by key rade parners has riggered a sharp real depreciaion ha increased he deb burden and offse, ofen in one-year period, he posiive resuls achieved hrough susained growh and macroeconomic sabiliy. The large negaive effecs of real exchange rae depreciaion on he raio of he PV of deb o GDP depend upon he excessive reliance of LICs on he expors of a few commodiies direced o a few rade parners. Reducion in inernaional marke prices or regional crises, as in he case of he Asian crises in 1997 or he Lain American crises in 1999, decreases he LICs access o foreign currencies hrough expors or FDIs and has a negaive impac on GDP growh and on fiscal revenues, increasing he need for addiional financing in he form of deb. As noed above, he role of real ineres rae is marginal for he composiion and erms srucure of LICs public deb. However, Ghana and Nigeria represen noable excepions. The frequen resor o budge financing hrough domesic deb increased he overall public deb burden, because domesic deb carries higher real ineres raes han foreign concessional deb and mos of domesic deb has mauriy of less han one year implying a high risk of rollover in he even of an increase of ineres raes. c) The Role of Deb Relief Each of he 7 LICs considered in he case sudies benefied from deb relief, wih Bolivia, Ehiopia, Nigeria and Uganda undergoing a number of cycles of deb relief and resrucuring evens. Rescheduling of deb service, buyback operaions or deb reducion, involving only a par of crediors, failed o achieve a permanen exi of he LICs from addiional deb relief. In general, he effec of deb relief appears o have had a smaller effec on deb burden han wha brough by coninued economic growh. However, he reducion in deb achieved by he 7 LICs as a resul of agreemens implemened in he second half of he 90s was in many cases he mos imporan facor in deermining he decrease in deb burden indicaors. In many cases, deb relief masked increasing rends in deb raios as a resul of real exchange rae depreciaion or primary deficis. Counries ha have had large deb resrucuring deals before 2003, including Bolivia, Uganda and Nigeria and Vienam, experienced upwards rend in exernal as well as domesic deb raios aferwards. Counries ha received deb relief in 2003 (Ehiopia, Ghana and Laos) experienced an increase in domesic deb. Therefore, deb relief, on he one hand, frees up resources ha can be used o finance fuure developmen expendiures and reduces he negaive effecs of deb overhang; on he oher hand, deb relief does no preven he build-up of new deb and he recurrence of episodes of deb disress. 17

21 d) The Role of Deb Srucures In LICs, deb srucures have changed over he period More concessional long-erm exernal deb represened more han 75 percen of oal deb in 2003, up from 54 percen in The level of concessionaliy of new deb lowered he deb burden, by spreading deb service over a long period of ime and mainaining he real ineres cos close o 0 or negaive. In 5 counries ou of 7 an increase in he gran elemen of exernal deb decreased he raio of he PV of deb o GDP afer Excepions are Uganda, whose level of concessional deb decreased afer HIPC relief and Nigeria, for which nominal deb, insead of PV deb was used in he analysis. 13 LICs deb is mainly denominaed in foreign currency which increases he risk of shifing deb burdens upward as a resul of exernal shocks affecing he real exchange rae. A predicable repaymen schedule and low ineres rae did no preven deb crisis in Ehiopia, Uganda, Bolivia and Ghana. Domesic deb played a marginal, albei increasing, role in he period considered. Counries wih high domesic deb, such as Nigeria and Ghana ended o have higher real ineres cos han counries ha limied he issuance of domesic deb. The ineres cos could sill rise furher in counries where forms of deb indexaion are inroduced, such as Bolivia. e) The Role of Coningen/Hidden liabiliies Pas, curren, and prospecive hidden deficis, due o he realizaion of coningen liabiliies can quickly and dramaically raise public deb. 14 Daa on coningen/hidden liabiliies are exremely sparse and no documened in formal daabases. For ha reason, he impac of coningen liabiliies on deb dynamics is likely o be underesimaed in he analysis, as mos counries idenify heir coningen liabiliies only ex-pos, e.g., he coss associaed wih banking failures are acknowledged only in he afermah of a banking crisis. Alhough some counries (Vienam, Bolivia) are racking coningen liabiliies, hey do no include provisions for probable losses from coningen liabiliies in he fiscal accouns. Mos counries also do no accoun for implici direc liabiliies like pension and social securiy obligaions. 13 However, he gran elemen on Nigeria is close o 0 or slighly negaive, implying ha PV and nominal value of deb are very similar over he period considered. 14 Coningen liabiliies are financial obligaions ha only become a direc liabiliy for he governmen if a cerain even occurs. A counry s financial sysem is ofen considered as is mos serious coningen liabiliy, as markes expec governmen suppor far beyond legal obligaions o ensure financial sabiliy (World Bank, 1998). 18