Bank of Uganda. Volume 1 Number 1 The Bank of Uganda Staff Papers Journal by the Bank of Uganda

Transcription

1 Bank of Uganda Volume 1 Number 1 The Bank of Uganda Saff Papers Journal 2007 Jacob Opolo. A re-examinaion of he demand for money in Uganda: Naure and implicaions for moneary policy Yusuf A. Abdalla and Keneh Egesa. Trade & growh in agriculure: Uganda s radiional expor poenial wihin he evolving mulilaeral rading regime Kihangire D., Pos D J, and Cameron S. Has exchange rae variabiliy affeced ropical non-radiional expors? Evidence from Uganda s flower expors ( ) David Kihangire. Exchange rae misalignmen and financial liberalisaion: Empirical evidence and macroeconomic implicaions for Uganda, Kihangire David and Alber Mugyenyi. Is inflaion always and everywhere a non-moneary phenomenon? Evidence from Uganda Aingi-Ego Michael and Rachael Kaggwa Sebudde. Measuring efficiency of a marke in ransiion: The Uganda foreign exchange marke 2007 by he Bank of Uganda

2 The Bank of Uganda Saff Papers Journal Michael Aingi-Ego Chief Edior David Kihangire Depuy Chief Edior Charles Augusine Abuka Senior Edior Jimmy Apaa Edior The publisher, ediors and he Bank of Uganda assume no responsibiliy for he views expressed by auhors and reviewers in he Bank of Uganda Saff Papers Journal. Copyrigh. All righs are reserved. No par of his publicaion may be produced, sored in a rerieval sysem or ransmied in any form or by any means, elecronically, mechanically, phoocopying, recording or oherwise wihou he prior permission of he publisher. Bank of Uganda Publicaion Plo 37/43 Kampala Road P.O. Box 7120, Kampala, Uganda Tel: Fax: , Telex: 61069/ info@bou.or.ug Inerne: hp:// 1

3 Bank of Uganda Volume 1 Number 1 The Bank of Uganda Saff Papers Journal

4 EDITOR S NOTE The edior invies brief commens (no more han 1,000 words) from exernal conribuors ouside he Bank of Uganda on published aricles in The Bank of Uganda Saff Papers Journal. The commens should be addressed o he Edior, who will send i o he auhor (s) of he original aricle for reply. Boh he commens and he reply will be considered for publicaion by he Bank of Uganda Bank of Uganda 3

5 Volume 1 Number 1 CONTENTS 2007 Jacob Opolo. A re-examinaion of he demand for money in Uganda: Naure and implicaions for moneary policy Yusuf A. Abdalla and Keneh Egesa. Trade & growh in Agriculure: Uganda s radiional expor poenial wihin he evolving mulilaeral rading regime Kihangire D., Pos D J, and Cameron S. Has exchange rae variabiliy affeced ropical non-radiional expors? Evidence from Uganda s flower expors ( ) David Kihangire. Exchange rae misalignmen and financial liberalisaion: Empirical evidence and macroeconomic implicaions for Uganda, Kihangire David and Alber Mugyenyi. Is inflaion always and everywhere a non-moneary phenomenon: Evidence from Uganda? Aingi-Ego Michael and Rachael Kaggwa Sebudde. Measuring efficiency of a marke in ransiion: The Uganda foreign exchange marke 4

6 The Bank of Uganda Saff Papers Journal Vol. 1, No Bank of Uganda A RE-EXAMINATION OF THE DEMAND FOR MONEY IN UGANDA: NATURE AND IMPLICATIONS FOR MONETARY POLICY Jacob Opolo Absrac The exisence of a sable relaionship beween money and prices is generally regarded as a prerequisie for a moneary argeing framework. The sabiliy of such a relaionship is assessed via a money demand funcion. A sable money demand funcion ensures ha he impac of moneary policy is predicable and ha moneary policy acions are consisen wih he desired objecives of price sabiliy and long-run growh. This paper uses conemporary empirical mehods o model he demand for money. In paricular, coinegraion analysis is used o esablish he exisence of a long-run relaionship and he shorrun dynamics are modeled via he error correcion model. Using quarerly daa for he period 1990 o 2004, he sudy esablishes ha boh real base money and real M2 are posiively relaed o real GDP (wih he income elasiciy of 1.5 and 1.0, respecively), and negaively relaed o he relaive ineres rae and he nominal exchange rae (wih he ineres rae elasiciies of 0.01 and 0.2 and he exchange rae elasiciies of 1.2 and 0.9 for he real base money and real M2 models, respecively. The shor-run models vindicae he exisence of coinegraing relaionships. For he real base money model, 39.7 percen of adjusmen o an exogenous shock occurs in one quarer, while for he real M2 model, only 13.2 percen of he adjusmen occurs during he firs quarer. Tess on parameer sabiliy indicae ha he shor-run models do no reveal any sign of insabiliy. The Chow forecas es fails o rejec he null hypohesis of no srucural change in boh models. The recursive coefficiens and residual esimaes for boh he real base money model and he M2 also exhibi parameer consancy indicaing ha here is no evidence of a major insabiliy over he sample period. The implicaion of his resul is ha moneary growh is a good predicor of fuure inflaion and real oupu in Uganda. Thus, here is no sufficien evidence ha he moneary argeing framework is ineffecive. Keywords: Demand for money; income and ineres elasiciy; sabiliy; financial secor reforms; moneary policy; coinegraion; error correcion model 5

7 I. INTRODUCTION The demand for money serves as a condui in he moneary policy ransmission mechanism and also helps in deermining he welfare implicaions of moneary policy acions in he economy. Esimaes of he money demand funcion such as he income elasiciy helps in deermining he rae of moneary expansion consisen wih long-run price sabiliy, while he ineres rae semi-elasiciy aids in calculaing he welfare coss of inflaion. The presence of a well-defined and sable money demand funcion 1 is crucial for he reliable ransmission of he impac of changes in money supply o aggregae spending. A sable money demand funcion hus ensures ha moneary policy has predicable impacs on macroeconomic variables and ha moneary policy acions are consisen wih he desired objecives of price sabiliy and long-run growh. The rend of moneary aggregaes is imporan for moneary auhoriies, especially in he conex of moneary argeing frameworks, in which programme benchmarks are deermined by limis on he moneary base or on oher moneary aggregaes. In Uganda, where a moneary argeing framework is in use, esimaing a money demand funcion becomes a vial focus, as he empirical relaionship represened by he money demand equaion enables one o examine he ineracion beween moneary aggregaes and oher economic indicaors. In addiion, he Ugandan financial sysem has undergone fundamenal changes in he las decade or so. The implemenaion of various srucural reforms and deregulaions ogeher wih he inroducion of new financial producs has changed he srucure of he financial sysem. In he view of he fac ha broad money comprises a wider range of financial insrumens reflecing he consequences of he financial innovaions and he recen changes in he Ugandan financial sysem, he sudy invesigaes money demand using a broad range of moneary aggregaes. The objecive of his paper is hus o provide empirical foundaions for he conduc of moneary policy in Uganda by examining he naure and sabiliy of he money demand funcion. The res of he paper is srucured as follows. Secion 2 presens a brief overview of he naure of he financial sysem and moneary policy framework in Uganda. The heoreical and empirical lieraure on money demand is presened in Secion 3 while he specificaion of he empirical model, daa issues, variable characerisics and he ime series properies of he daa are discussed in Secion 4.The esimaion framework and empirical findings are presened in Secion 5, and finally, he conclusions and policy recommendaions are presened in Secion 6. 1 The sabiliy of he money demand funcion also improves is predicabiliy. The money demand funcion is considered sable if he esimaed demand funcion is predicable; is coefficiens are accurae; he ouof-sample esimaions are fair; i is a funcion of relaively few variables; and he explanaory variables in he esimaed funcion represen heoreically and economically plausible relaionships beween money and real economic aciviy. 6

8 II THE FINANCIAL SYSTEM AND MONETARY POLICY FRAMEWORK IN UGANDA 2.1 The financial sysem Uganda s financial secor has undergone considerable reform since he lae 1980 s. Before 1988, he formal financial secor was highly regulaed wih direc governmen conrols over credi, ineres raes and exchange rae policy. Major liberalizaion measures were inroduced in 1992, bu conrols in boh ineres raes 2 and credi allocaion were removed in several seps over he wo-year period, ill 1994 when ineres raes were fully liberalized. Since hen ineres rae managemen has been effeced hrough indirec moneary policy insrumens wih he Treasury bill rae as he anchor. While he main focus was on he removal of conrols on ineres raes, some insiuional reforms were also iniiaed. This process was complemened by parallel measures o srenghen bank supervision and foser financial discipline hrough new legislaion and regulaions, and policies o improve he efficiency and profiabiliy of financial insiuions. In addiion o freeing ineres raes and credi allocaion, liberalizaion of he financial sysem also allowed freer enry of insiuions ino he sysem, wih he main objecive of increasing compeiion and efficiency. This saw a fas expansion of he banking and non-banking secor afer The governmen has also reduced is role in he allocaion of credi and ownership of financial insiuions. 2.2 The moneary policy framework The Bank of Uganda has been implemening a moneary aggregae argeing framework, he Reserve Money Programme (RMP) 3, since 1993 when i assumed is primary responsibiliy of formulaion and implemenaion of moneary policy, following enacmen of he Bank of Uganda Ac (1993). Under he RMP, he overall macroeconomic objecives on desired real GDP growh, inflaion, and balance of paymens are defined. Broad money (M2) growh is hen projeced consisen wih hese macroeconomic objecives, given assumpions on velociy. The growh of he moneary base, he operaing arge 4, is hen projeced in line wih he broader moneary aggregae and inflaion. 2 The removal of ineres rae conrols affeced he Treasury bill raes as he Treasury bill marke changed from one of ad-hoc issues by governmen o a marke-based aucion for he deerminaion of ineres raes. From hen, he key ineres raes were linked o he weighed average of he Treasury bill rae as deermined in he four preceding Treasury bill aucions 3 The RMP ress on srong underlying economic relaionships beween base money, broader moneary aggregaes, economic growh, and inflaion. The relaionship beween broad money (M2) and he money base is relayed hrough he muliplier effec of financial inermediaion and he propensiy of people o hold cash. 4 Base money was adoped as he operaing arge since was available wih a shorer lag han daa on broader Moneary aggregaes. 7

9 The annual growh arge for base money is convered ino monhly arges ha reflec seasonaliy in he demand for money. From he monhly arges, he desired levels for daily movemens are compued. Oher hings being equal, he direcion of moneary policy sance is deermined by he gap beween he ouurn and desired levels of base money: ease if base money is below desired, ighen moneary policy if base money is above desired, and leave moneary policy unchanged/neural if base money is in line wih desired levels. III. THEORETICAL AND EMPIRICAL LITERATURE 3.1 Theoreical conribuions o money demand Money demand heories dae back o he quaniy heory of money, developed in he classical equilibrium framework by wo differen bu equivalen expressions. Fisher (1911) provided he famous classical equaion of exchange, which relaes he quaniy of money in circulaion (Ms) o he volume of ransacions (T), he price level (P ), and he ransacions velociy of money (V ) as shown in equaion (3.1) below. M s V = PT (1.1) In his formulaion, money is held simply o faciliae ransacions and has no inrinsic value. The alernaive paradigm, he Cambridge approach, is associaed wih he neo-classical economiss, Pigou (1917) and Marshall (1923). The Cambridge approach sressed he demand for money as public demand for money holdings, especially he demand for real balances, which was an imporan facor in deermining he equilibrium price level consisen wih a given quaniy of money (Sriram, 1999). Keynes (1936) buil on he Cambridge approach o provide a more rigorous analysis of money demand, focusing on he moives of holding money. Keynes s liquidiy preference heory emphasizes he role of ineres raes in he demand for money. He disinguished hree moives for holding money: he ransacions moive, he precauionary moive and he speculaive moive. All he hree moives influence a paricular person s holdings of money. Keynes argued ha he demand for money for ransacions and precauionary moives depends on he level of income, while speculaive demand for money depends on ineres raes. From Keynes perspecive, he demand for real money balances (m d ) is a posiive funcion of real income (Y) and a negaive funcion of ineres rae (r) as depiced by he liquidiy preference funcion given in equaion (1.2). ( Y r) m d = f, ; f y >0, f r < 0. (1.2) According o he liquidiy preference heory of he demand for money velociy is no consan, bu depends on ineres raes. The main proposiion of he Keynesian analysis is ha when ineres raes are low, economic agens 8

10 will expec a fuure increase in ineres raes; hus, preferring o hold whaever amoun of money is supplied. Therefore, he aggregae demand for money becomes perfecly elasic wih respec o he ineres rae (liquidiy rap). Following he Keynesian exposiion of he moives for holding money, a number of models were developed o confirm he relaionship beween he demand for real money and, income and ineres raes. Baumol (1952) and Tobin (1956) independenly developed similar demand for money models, which demonsraed ha even money balances held for ransacions purposes are sensiive o he level of ineres raes. The Baumol - Tobin model analyses he coss and benefis of holding money for ransacions purposes. The benefi is convenience (i.e people have money o make everyday ransacions) and he cos of his convenience is he ineres forgone. A person can hus hold a porfolio of moneary asses and non-moneary asses.if r is he difference in he reurn beween moneary and non-moneary asses and b as he cos of ransferring non-moneary asses ino moneary asses, such as a brokerage fee, hen a person minimizes he sum of brokerage coss and ineres income forgone. This leads o a well-know square-roo formula given in equaion (1.3) below. m d = ( by / 2r) (1.3) This saes ha he demand for real money balances m is direcly proporional o ransacions coss b and real income Y, and inversely proporional o he ineres rae r. The decision abou how ofen o pay he brokerage fee is analogous o he decision abou how ofen o make daily ransacions, ha deermines how much money o hold. In he ideal world of he Baumol-Tobin model, he elasiciy of money demand in response o income and ineres rae mus be 0.5 and -0.5 respecively. In 1956, Friedman developed he modern quaniy heory of money. He applied he heory of porfolio choice and posulaed ha he demand for money mus be influenced by he same facors ha influence he demand for any asse. Thus according o Friedman, he demand for money funcion is given by equaion (1.4) below. M P d = f e ( Yp rb rm ; re rm ;π rm ; ) (1.4) where (M/P) d is he demand for real money balances, Y p is a measure of wealh or permanen income, r m is he expeced reurn on money, r e is he expeced reurn on bonds, r b is he expeced reurn on equiy, and π e is he expeced inflaion rae. Friedman s heory implies ha a rise in marke raes raises he expeced reurn on money so ha (r b - r m ) remains fairly sable. Ineres raes herefore have lile or no effec on he demand for money. Permanen income is hus he primary deerminan of he demand for money. 9

11 Recen developmens in he heory of he demand for money incorporae scenarios where boh domesic and foreign monies provide liquidiy services in financing he exchange of goods and services, including non-radables, beween domesic residens. These are he currency subsiuion models of he demand for money. 3.2 Empirical issues in he esimaion of money demand funcions The empirical lieraure on modeling and esimaing money demand underscores wo imporan poins: variable selecion and represenaion, and esimaion framework. The proper specificaion of he scale and opporuniy cos variables, and he appropriae choice of an empirical framework ha is free of heoreical and esimaion problems are boh imporan for robus resuls o obain. Sudies ha have no accorded due consideraion o hese issues have ended o yield poor resuls (Sriram; 2001) Variable selecion and represenaion The general analyical framework, which derives from money demand heories, akes he form: M d = f ( S A P, ) (1.5) Where M and P represens he seleced moneary aggregae 5 in nominal erms and he price level, respecively. S is he chosen scale variable o represen economic aciviy 6, and A is he alernaive fore gone or he opporuniy cos of holding money. The opporuniy cos of holding money involves wo componens: he own rae of reurn and he rae of reurn on alernaive asses. Ericsson (1998) argues ha i is imporan o include boh ineres raes in he esimaion of a money demand funcion o avoid he collapse of he esimaed funcion. He furher saes ha he own-rae is imporan in economies where financial innovaion has been aking place. The reurn on real asses is usually proxied by he expeced rae of inflaion (Sriram; 1999). The real value of money falls wih inflaion whils he value of real asses is mainained. Therefore, here is a srong incenive for persons o swich ou of 5 Money is broadly classified ino wo groups, narrow and broad money. Narrow money consiss of hose asses, which are readily available for ransacions while broad money encompasses a wider range of asses. Empirical lieraure suggess ha he choice of an appropriae moneary aggregae is an empirical maer. As such, several definiions of money have been used in various sudies. Indeed, Sriram (1999) argues ha measures of money have been seleced based on he objecives of he researchers. 6 The mos commonly used measures are gross naional produc (GNP), gross domesic produc (GDP) and ne naional produc (NNP). Recen empirical work has however focused on oher scale variables involving more comprehensive measures of ransacions and he division of ransacions ino differen componens on he grounds ha no all ransacions are equally money inensive (see Goldfeld and Sichel, 1990). However his nowihsanding, Goldfeld and Sichel (1990) argue ha here is no firm evidence o he effec ha he caegorizaion of GNP ino various componens leads o an improvemen in he behaviour of he money demand funcion. 10

12 money ino real asses when here are srong inflaionary expecaions. Foreign ineres raes and he expeced rae of depreciaion usually represen he reurns on foreign asses (Sriram, 1999). The currency subsiuion lieraure provides he necessary suppor in choosing he appropriae variables ha accoun for foreign influence. Direc currency subsiuion concenraes on he exchange rae variable while he capial mobiliy or indirec currency subsiuion lieraure focuses on foreign ineres raes (see Levenakis; 1993) Esimaion specificaions Earlies empirical work on money demand primarily involved producing esimaes of velociy, characerizing is behaviour over ime and idenifying he insiuional facors responsible for long-run movemens in velociy. Before he 1970s, anoher popular approach, which ended o produce precise and consisen esimaes of he money demand funcion, was he parial adjusmen model (PAM). The PAM, a log-linear specificaion exensively used for esimaing money demand, was originally inroduced by Chow (1966) and laer popularized by Goldfeld (1973). The PAM augmens he radiional formulaion of money demand funcion by disinguishing beween desired and acual money holdings, and inroducing a parameer by which he acual money holdings adjus o he desired levels (Sriram, 1999). The PAM akes he general form depiced in equaion (1.6) below. M ln( P M ) ln( P 1 1 * M M 1 ) = γ ln( ) ln( ) (1.6) P P 1 where M */P is he desired level of real money balance, γ is a parameer ha governs he speed of adjusmen, (-1) is a subscrip denoing he previous ime period, and all oher variables as defined earlier. The acual level of money balances adjuss in each period only parly owards is desired level (Mishkin; 1998). According o Sriram (1999), his model faired well prior o 1973, bu failed o explain he insabiliy in he money demand - funcion experienced since hen, on accoun of boh heoreical and empirical problems. As an alernaive, he buffer sock models (BSM) and error correcion models (ECM) were developed. Unlike in he PAM, moneary shocks are explicily modeled in he BSM. This nowihsanding, here have been mixed resuls in he empirical applicaion of his approach (see Boughon and Tavlas, 1991 and Cuhberson and Taylor, 1987). Conemporary empirical work on money demand has hus ended o use coinegraion for esablishing he exisence of a long-run relaionship and error correcion model for shor-run dynamics Empirical findings on money demand in Uganda A review of he exising empirical work on he demand for money in Uganda is imporan, boh for comparison of empirical findings and idenificaion of 11

13 exising gaps. In wha follows, we review he empirical findings of hree sudies, Kaarikawe and Sebudde (1999), Nachega (2001) and Kararach (2002). Kaarikawe and Sebudde (1999) esimaed a money demand funcion for Uganda using monhly daa for he period January 1990 o December They used an income index compued as a weighed average of index of indusrial producion and coffee procuremen as a proxy for income 7. Price changes were found o be saionary in levels and hus did no ener he coniegraing equaion. Save for he exchange rae, all coefficiens had expeced signs and were saisically significan in boh cases. The nominal exchange rae was found o have a posiive elasiciy in he base money model, implying ha he depreciaion of he shilling has a posiive impac on he demand for base money 8. In he M2 model, exchange rae elasiciy was negaive, meaning ha as he Uganda shilling depreciaes, and agens shif away from he domesic currency. In general, he long-run money demand funcion was found o exhibi a sable relaionship. Using he parial and weak exogeneiy assumpions proposed by Johansen (1988), hey found ha while real M2 is srongly endogenous, real base money is no. Thus for base money, he shor-run relaionship was esimaed using nominal balances. The shor-run dynamics sugges ha while ineres raes and exchange raes affec base money wih a lag of up-o four and hree monhs, respecively, changes in income affec base money wih a lag of one monh. A one-period lag of base money and a hree-period lag of prices were also found o affec base money levels. The dynamic equaion for M2 indicaed ha he Treasury bill rae, deposi rae and inflaion affec base money wih a lag. Thus, here is no conemporaneous policy variable ha can induce movemens in M2. Nachega (2001) using quarerly daa for he period also used coinegraion analysis and error correcion modeling o examine he behaviour of he broad money demand funcion in Uganda. His resuls were consisen wih economic heory as all coefficiens had he expeced signs. The demand for broad money was found o be posiively relaed o real income and he relaive own rae of reurn 9. The income elasiciy of M2 was close o uniy and significanly differen from zero, which is consisen wih he quaniy heory of money. The semi elasiciy of he relaive own rae of reurn was also posiive and saisically differen from zero. On he oher hand, he LIBOR was found o exer a negaive influence on he demand for money. Overall, here was sufficien evidence o suppor he sabiliy of he money demand funcion. Kararach (2002) esimaed a demand funcion for Uganda using quarerly daa for he period The long-run money 7 The weighs used were 0.7 and 0.3 for coffee procuremen and indusrial producion, respecively. 8 This holds for he case of ransacions demand. A depreciaion of he shilling leads o a higher demand of domesic currency o purchase foreign currency for financing impors. 9 The relaive own rae of reurn was defined as he annual ineres rae offered on ime deposis minus he 91-day Treasury bill rae. 12

14 demand funcion was esimaed using he Cochrane-Orcu mehod. Income and ineres raes were found o be significan and heoreically plausible in explaining he demand for money 10. The behaviour of he inflaion variable was however conrary o heoreical expecaions, as he desire o hold money grew wih inflaion. Expeced inflaion, proxied by he previous inflaion, however had he correc sign. The error correcion model was also esimaed o capure he shor-run dynamics of he money demand funcion. The resuls of he ECM sugges ha he speed a which he economy reurns o equilibrium afer a shock is 2.5.The resuls also indicaed ha here is saisical evidence of insabiliy in he money demand funcion in Uganda. This resul is however conrary o he findings of Kaarikawe and Sebudde (1999) and Nachega (2001). IV METHODOLOGY 4.1 Specificaion of he empirical model Following he radiional approach, he demand for money is specified as a funcion of income, ineres raes, and exchange raes. The exchange rae capures he effec of expeced depreciaion on he demand for money. All variables ener he model in logarihmic form. The basic long-run model is hus specified as: d M (1.7) = β o + β1lgdp + β2ltbr91 + β3ldep+ β4lner + β5inf + ε ln P Where P is a measure of money divided by he CPI, ha is, real money GDP is real moneary Gross Domesic Produc TBR91 is he 91-day Treasury bill rae 11 NER is he nominal exchange rae. DEP is he average annual savings and ime deposi rae Inf Inflaion rae ε is he random error erm is a subscrip denoing ime. M d Furher, equal elasiciy (in absolue value) for he 91-day reasury bill rae (TBR91) and he average deposi rae on savings and ime deposis is assumed. Thus, equaion (1.7) is ransformed o equaion (1.8). d M (1.8) = α o + α1lgdp + α2( LTBR91 LDEP) + α3lner + β4inf + ε ln P 10 The ineres rae coefficien was however low. 11 The average ime and savings deposi rae (DEP) is used as he own rae of reurn and he 91-day reasury bill rae (TBR91) as he rae on alernaive non-moneary financial asses (i.e. he opporuniy cos of holding money) 13

15 Based on coinegraion analysis and weak exogeneiy ess, a shor-run money demand funcion is hen modeled in a single equaion conex using he error correcion model (ECM). The shor-run model shows how he adjusmen mechanism works o rever o equilibrium condiion when i is disurbed by exogenous shocks, which lead o deviaions from he long-run equilibrium. The order of he Auoregressive disribued lag (ARDL) in he shor-run model will be deermined by he order of he VAR of he appropriae coinegraing relaionship. Thus, he general shor-run model, he error correcion model, is given by equaion (1.9). L ( M P) = ρ 1 γ L ( M P) 1i i= 1 + ρ 1 4i i= 0 i λ LNER + i + ρ 1 2i i= 0 ρ 1 γ LGDP+ 5i i= 0 γ ( LTBR 91 DEP) + γ7dummy + µ (1.9) Where a dummy capures a policy change from conrolled o fully liberalized ineres raes in ρ 1 1 3i i= 0 i 6 1 i λ LCPI + γ ECT + CONSTANT 4.2 Variables and daa characerisics Variables used in he sudy As noed in he heoreical presenaion of he model, we hypohesize ha he fundamenal variables ha deermine money demand in Uganda are real GDP, ineres raes, exchange rae and he domesic price level. For esimaion purposes, we use he logarihmic ransformaion of quarerly daa for he period 1990: :04. The sudy of he behaviour of he variables is worhwhile before esimaion Money Esimaions are based on wo widely used measures of money in Uganda, namely, he moneary base and M2 12. The nominal money balances are deflaed using he Composie Consumer Price Index (CPI). Figure 1(b) shows he naural logarihm of he various real moneary aggregaes during he period 1990:01 o 2004:04. All moneary aggregaes display an upward rend during he period. By visual inspecion, i is apparen ha all moneary aggregaes are non-saionary. Consequenly, any furher analysis should be preceded by es of exisence of uni roos Real moneary GDP The sudy uses Hensridge s quarerly GDP series for he period , and for he subsequen period, Friedman s (1962) linear inerpolaion mehod was used. This involved adjusing a linear inerpolaion of he annual GDP series wih he weighed percenage errors beween acual quarerly daa on he capaciy o impor (defined as expors deflaed by he impor uni price 12 Moneary base include currency in circulaion plus commercial bank reserves a he cenral bank and M2 includes M1 (currency in circulaion and demand deposis) plus ime and savings deposis. 14

16 index) and he linear inerpolaions of he capaciy o impor. The resuling series were hen seasonally adjused o minimize he impac of he cyclical componens. Figure 1(a), plos he seasonally adjused quarerly moneary GDP agains ime. As shown in he figure, he seasonally adjused quarerly moneary GDP series GDP porray an upward rend, and by visual inspecion he series seem o be non-saionary Ineres rae spread The spread beween he 91-day Treasury bill rae and he average annual ineres offered on ime and savings deposi is used o deermine he relaive ineres rae elasiciy 13. This spread represens he reurn on domesic nonmoneary financial asses relaive o domesic moneary asses. An increase in he ineres rae spread will make non-moneary financial asses more aracive, hus leading o a decline in he demand for money. Trends in nominal ineres rae spreads are shown in figure 1(c) Nominal exchange rae The nominal exchange rae used in he sudy is defined as unis of Uganda shillings per Unied Saes Dollar. A priori, i is expeced ha a depreciaion of he domesic currency will induce economic agens o increase heir holdings of foreign currency since expeced reurns from holding foreign currency increase. Figure 1(d), plos he naural logarihm of he real exchange rae agains ime. By visual inspecion, one can conclude ha he series is non-saionary Inflaion Inflaion is compued from he composie consumer price index. Since he real value of money falls wih inflaion while he value of real asses is mainained, here is a srong incenive for economic agens o swich ou of money ino real asses when here are srong inflaionary expecaions. Thus a priori, inflaion is expeced o have a negaive impac on he demand for money. Trends in inflaion and he consumer price index are depiced in figure 1(e) and figure 1(f) respecively. By visual inspecion inflaion is expeced o be saionery Time series properies Mos macroeconomic variables have been found o have uni roos, which makes he esing procedures for causaliy complex. The ime series properies of he variables are examined using he augmened Dickey-Fuller (ADF) and Phillips-Perron (PP) uni roo ess. The null hypohesis under boh ess is ha uni roo processes generae he ime series. As shown in able 1, here is insufficien evidence o rejec he null of a uni roo for all 13 Following argumens by Sachs and Larrain (1993) ha nominal ineres raes are widely recognized appropriae deerminan of he demand for real balances, nominal ineres raes are used o deermine he spread. 15

17 variables in levels 14. However, using boh he ADF and he PP es, he null of a uni roo in he firs difference of all he variables is rejeced. The variables hus become saionery afer he firs difference. Hence all variables are I(1). V ESTIMATION AND EMPIRICAL RESULTS 5.1 Esimaion framework Having esablished he ime series properies of he variables, he I(1) variables are enered ino he coinegraing vecor o empirically es he hypohesis posulaed in equaion (1.8). Inflaion is excluded from he coinegraion analysis since i was found o be saionary 15. A dummy, o capure he full liberalizaion of ineres raes in 1994, is included as an exogenous variable. Johansen (1991; 1995) coinegraion es is used o esablish he exisence of a long-run relaionship beween he variables. The appropriae lag srucure of he VAR is deermined using he Akaike Informaion Crierion (AIC) and he Schwarz Bayesian Crierion (SBC). Afer deermining he lag srucure of he VAR, diagnosic ess of he preferred lag are conduced. Having esablished he exisence of a long-run relaionship, he shor-run dynamic relaion for real moneary aggregaes is specified, if he null of weak exogeneiy of he real moneary aggregae is rejeced. The dummies included in he VAR are also included in he error correcion model. A reducion based on Hendry (1987) general o specific simplificaion mehodology is used o sep-by-sep eliminaes he saisically as well as economically mos insignifican regressors. The las remaining equaion wih all variables being significan is hus repored. Finally, parameer sabiliy ess are conduced o ascerain he sabiliy of he money demand. In paricular, Chow breakpoin and forecas ess, and parameer consancy ess via he recursive esimaion of he coefficiens and residuals of he model are conduced. 5.2 Empirical Resuls The real Base Money model The coinegraion resuls for he base money model are presened in able 1.2 in he appendix. The maximal eigenvalue rejecs he null of no coinegraing vecor in favour of one coinegraing relaionship. However, he race es saisic poins o he exisence of wo coinegraing vecors a he 10 percen 14 We rejec he null of a uni roo agains a one-sided alernaive if he es saisic is less han he criical value. 15 Inflaion is however included in he ECM in order o capure he impac of inflaion dynamics on he money demand funcion. 16

18 level of significance. As he maximal eigenvalue and race es saisic do no unambiguously confirm he exisence of more han one coinegraing vecor, he coinegraing vecor consisen wih economic heory is adoped as a long-run money demand relaionship for base money. Figure 1.5 confirms ha his coinegraing vecor is saionary. The long-run money demand funcion for base money can hus be wrien as: LRBMON = 1.516LQGDP L(TBR91 DEP) 1.228LNER (1.10) Inerpreaion of he coefficiens All coefficiens in equaion (1.10) have expeced signs. The income elasiciy of he demand for money of 1.5 is posiive and is significan in explaining he demand for real base money 16. This is consisen wih economic heory, which posulaes ha an increase in real income increases he level of ransacions, and consequenly he demand for real money balances o finance he high level of ransacions in an economy. The elasiciy of he relaive ineres rae is negaive and saisically significan in explaining he demand for base money. An increase in he differenial beween he wo raes reduces he demand for base money as agens resrucure heir wealh porfolio in favour of he high yielding asses 17. However, he ineres rae has a low impac on he demand for base money as suggesed by he low relaive ineres rae elasiciy. The exchange rae elasiciy of 1.2 indicaes ha he exchange rae depreciaion has negaive impac on he demand for money. An increase in expeced nominal depreciaion will lead economic agens o subsiue domesic currency for foreign currency. Thus as he domesic currency uni depreciaes, agens shif away from domesic currency o foreign currency denominaed or physical asses. Table 1.2 also conains he adjusmen marix, which conains he weigh wih which he coinegraing vecors ener he equaion in he sysem. Each nonzero column measures he speed of shor-run response o disequilibrium in he endogenous variables of he VAR. Specifically, he firs column measures he speed a which LRBMON, he dependen variable in he firs equaion of he VAR moves owards resoring long-run equilibrium. From able 2, he coefficien 0.4 is he esimaed feed back coefficien for he money demand equaion. I shows ha 39.9 percen of he adjusmen is achieved in he firs quarer. The negaive coefficien signifies ha lagged excess money induces smaller holdings of curren money. The weak 16 When he uni income homogeneiy resricion was imposed, he x 2 -saisic was found o be wih a p-value of , which suggess ha he uni income elasiciy canno be rejeced for he longrun relaionship. 17 An increase in he Treasury bill rae increases he relaive ineres rae where as an increase in he deposi rae reduces he relaive rae assuming ha one rae does no change. 17

19 exogeneiy ess reveal ha real base money and he nominal exchange rae are no weakly exogenous, bu weak exogeneiy canno be rejeced for real GDP and (TRR91 DEP). However, a high adjusmen coefficien for real base money compared o he nominal exchange rae suggess ha real base money is more likely o be endogenously deermined ha he nominal exchange rae Error correcion model Based on coinegraion analysis and weak exogeneiy ess, he shor-run relaionship for real base money is modeled in a single equaion conex using he error correcion model. Consequenly, he shor-run dynamic relaionship for real base money is as shown in equaion (1.10), where he figures in parenhesis are -values. 18 DLRBMON = DLRBMON DLQGDP DLQGDP DLQGDP DLQGDP 0.026( TBR91 DEP) (2.158) DLNER 0.250DLCPI 0.011DLCPI DUMMY (2.452) 0.397ECM (4.134) (6.048) ( 2.480) ( 2.961) ( 2.676) (3.145) ( 2.105) 2 (2.451) (2.912) DLNER ( 2.621) (1.11) The resuls sugges ha GDP can affec base money wih a lag of up o hree quarers. Oher variables also affec base money bu wih varying lags as shown in equaion (1.10). The negaive coefficien of he ECM is significan, hus validaing ha he coinegraing relaionship beween he variables is valid. I essenially implies ha when an exogenous shock disurbs he equilibrium condiion, 39 percen of is effec is adjused in one period. Tess on parameer sabiliy indicae ha he shor-run model does no reveal any sign of insabiliy. The resuls of he Chow forecas es repored in able 1.3 fail o rejec he null hypohesis of no srucural change in he money demand funcion. Parameer consancy ess are also implemened via he recursive esimaion of he coefficiens and he residuals of he model, he resuls of which are presened in figure 1.3 and figure 1.4, respecively. As shown in figure 1.3, he model exhibis sable coefficiens wih he ± 2 sandard error inerval narrowing quickly. The recursive residual ess also confirm parameer sabiliy as he recursive residuals are wihin he band (see figure 1.4). Thus, he model exhibis parameer consancy, indicaing ha here is no evidence of a major insabiliy over he sample period. 18 The resuls of he diagnosic ess repored in able 3 show ha he model is economerically well specified. From residual ess, we are unable o rejec he nulls ha here is no serial correlaion in he residuals, and no auoregressive condiional heeroscedasiciy (ARCH) in he residuals. Whies Heeroscedaiciy es also fails o rejec he null of homoscedasiciy. The Jarque-Bera saisic also shows ha he residuals are normally disribued. 18

20 5.2.4 The real M2 model Long-run demand for M2 As shown in able 4, boh he maximal eigenvalue and he race es saisic rejec he null of no coinegraing vecor in favour of one coinegraing relaionship. The long-run money demand funcion can hus be specified as shown in equaion (1.12) below. Ln( RM 2 ) Ln( MGDP ) Ln( TBR 91 DEP ) Ln( NER ) = (1.12) All coefficiens are significan and have expeced signs as posulaed by economic heory. The income elasiciy of demand for money is 1.0, which is close o 1 as suggesed by he quaniy heory of demand for money. The relaive ineres rae elasiciy ( 0.2) is lower han he exchange rae elasiciy ( 0.9). Thus, in he long-run, when compared o he ineres rae, he exchange rae has a higher impac on he demand for real M2. In erms of he adjusmen marix, esimaed feed back coefficien of 0.2 shows ha 21.5 percen of he adjusmen is achieved in he firs quarer. The weak exogeneiy ess reveal ha while weak exogeneiy can be rejeced for real M2, i canno be rejeced for real GDP, (TBR91- DEP) and he nominal exchange rae. Thus a shor-run relaionship for real M2 can be modeled using and ECM Error correcion model As shown in able 1.4, weak exogeneiy is rejeced for real M2. Thus, he shor-run dynamic relaionship for real M2 is modeled as shown in equaion (1.13), where he figures in parenhesis are -values. 19 DLRM2 = DLRM2 (2.862) D( TBR91 DEP) DLCPI ( 2.047) (2.137) 4 (2.757) DLRM ( TBR91 DEP) 0.027DUMMY 0.132ECM ( 2.530) (2.461) ( 2.101) ( 3.348) DLQGDP DLQGDP 1 3 (2.965) 0.056DLNER ( 2.497) 2 (2.336) 0.035DLCPI ( 5.791) 1 (1.13) The resuls sugges ha GDP affecs real M2 wih a lag of up o only one quarer. Oher variables also affec base money bu wih varying lags as shown in equaion (1.13). The ECM coefficien is negaive and significan, hus vindicaing he exisence of a coinegraing relaionship. I essenially implies ha when an exogenous shock disurbs he equilibrium condiion, 13 percen of is effec is adjused in one period. Jus like he base money model, shor-run M2 model does no reveal any signs of insabiliy. The Chow forecas es repored in able 1.5 fails o rejec he null hypohesis of no srucural change in he money demand funcion. Recursive parameer coefficiens and residuals esimaes presened in figure 19 The diagnosic ess repored in able 5 show ha he model is economerically well specified. 19

21 1.5 and figure 1.6 indicae ha here is no evidence of a major insabiliy over he sample period. VII CONCLUSION AND POLICY RECOMMENDATIONS The coinegraion analysis indicaes ha here is a long run relaionship for boh base money and M2. In boh cases, he income elasiciy is found o be close o uniy, a resul ha is consisen wih he quaniy heory of money. The relaive ineres rae and exchange rae elasiciies are found o be negaive, which is also consisen wih economic heory. However in boh cases, he exchange rae elasiciy is higher han he ineres rae elasiciy, suggesing ha he exchange rae channel is sronger han he ineres rae channel in he moneary policy ransmission mechanism. The shor-run elasiciies are also o a large exen consisen wih economic heory. Furher, here is no sufficien evidence in suppor of he insabiliy of he money demand funcions, which implies ha moneary growh is a good predicor of fuure inflaion and real oupu in Uganda. 20

22 APPENDIX Table 1.1: Time series properies of he daa Augmened Dickey-Fuller (ADF) es and Phillips-perron (PP) ess Variable Levels Firs differences Wih inercep Wih rend and inercep Variable Wih inercep Wih rend and inercep ADF PP ADF PP ADF PP ADF PP LRBMON LRBMON ** ** ** ** LRM LRM ** ** ** ** LQGDP LGDP ** ** ** ** L(TBR91-DEP) L(TBR91-DEP) ** ** ** ** LNER LNER * ** ** LCPI LCPI ** ** ** ** The superscrips ** and * denoe rejecion of he hypohesis of a uni roo a 1 percen and 5 percen significance levels respecively Source: Auhors Table 1.2: Coinegraion analysis of he real base money model Coinegraion Analysis (rended case) 1 - Johansen maximum likelihood procedure Eigenvalue Null hypohesis 2 r=0 r=1 r=2 r=3 3 λ max λ race % criical value (63.00) (42.34) (25.77) (12.39) 90% criical value (59.16) (39.34) (23.08) (10.55) Sandardized eigenvecor β' Variable LRBMON LQGDP L(TBR91- LNER DEP) Vecor Vecor Adjusmen coefficien α LRBMON LQGDP L(TBR91- LNER DEP) LRBMON LQMGDP LRELINT LNER Tesing significance of a given variable Weak exogeneiy ess 5 Variable LRBMON LQGDP L(TBR91- LNER DEP) χ²(1) ** * P-value LQGDP L(TBR91- LNER CONSTANT DEP) χ²(1) ** * ** ** p-value Source: Auhors 21

23 1. Coinegraion wih real M2, real GDP, relaive ineres rae and he nominal exchange rae as endogenous variables. 2. R is he number of coinergraion vecors. 3. The saisics λ max and λ race are respecively he Johansen maximal and race eigenvalue saisics for esing coinegraion. The null hypohesis is in erms of r (he number of coinegraing vecors), and rejecion of r=0 is evidence in favour of a leas one coinegraing vecor. 4. The criical value in parenhesis is for he race saisic. The criical values for boh he maximal and race saisics are aken from Oserwald-Lenum (1992). 5. The Wu Hausman ess of weak exogeneiy. 22

24 Table 1.3: Esimaes and diagnosic ess for he Real base money ECM Dependen Variable: DLRBMON Mehod: Leas Squares Observaions: 56 afer adjusing endpoins Variable Coefficien Sd. Error -Saisic Prob. Inercep LRBMON LQGDP LQGDP LQGDP LQGDP (TBR91-DEP) LNER LNER LCPI LCPI DUMMY ECM R-squared Durbin-Wason sa Adjused R-squared F-saisic S.E. of regression Prob(F-saisic) Diagnosic ess Tess on residuals Breusch-Godfrey Serial Correlaion LM es Lag lengh=1 F-saisic Probabiliy Obs*R-squared Probabiliy Lag lengh=3 F-saisic Probabiliy Obs*R-squared Probabiliy ARCH es Lag lengh=1 F-saisic Probabiliy Obs*R-squared Probabiliy Lag lengh=3 F-saisic Probabiliy Obs*R-squared Probabiliy Whies Heeroskedasiciy es F-saisic Probabiliy Obs*R-squared Probabiliy Normaliy es Jarque-Bera Saisic Probabiliy Sabiliy ess Chow forecas es: forecas from 1997:4 o 2004:4 F-saisic Probabiliy Source: Auhors 23

25 Table 1.4: Coinegraion analysis of he real M2 model Coinegraion Analysis (rended case) 1 - Johansen maximum likelihood procedure Eigenvalue Null hypohesis 2 r=0 r=1 r=2 r=3 3 λ max λ race % criical value (63.00) (42.34) (25.77) (12.39) 90% criical value (59.16) (39.34) (23.08) (10.55) Sandardized eigenvecor β' Variable LRBMON LQGDP L(TBR91- LNER DEP) Vecor Adjusmen coefficien α LRBMON LQGDP L(TBR91- LNER DEP) LRBMON LQMGDP LRELINT LNER P-value Tesing significance of a given variable LRBMON LQGDP L(TBR91- LNER DEP) χ²(1) ** LQGDP L(TBR91- LNER Consan DEP) χ²(1) ** ** ** ** p-value Source: Auhors 1. Coinegraion wih real M2, real GDP, relaive ineres rae and he nominal exchange rae as endogenous variables and inflaion as an exogenous variable. 2. R is he number of coinergraion vecors. 3. The saisics λ max and λ race are respecively he Johansen maximal and race eigenvalue saisics for esing coinegraion. The null hypohesis is in erms of r (he number of coinegraing vecors), and rejecion of r=0 is evidence in favour of a leas one coinegraing vecor. 4. The criical value in parenhesis is for he race saisic. The criical values for boh he maximal and race saisics are aken from Oserwald-Lenum (1992). 5. The Wu Hausman ess of weak exogeneiy. 24

26 Table 1.5: Esimaes and diagnosic ess for he Real M2 ECM Dependen Variable: DLRM2 Mehod: Leas Squares Observaions: 56 afer adjusing endpoins Variable Coefficien Sd. Error -Saisic Prob. Inercep LRM LRM LQGDP LQGDP (TBR91-DEP) (TBR91-DEP) LNER LCPI LCPI DUMMY ECM R-squared Durbin-Wason sa Adjused R-squared F-saisic S.E. of regression Prob(F-saisic) Tess on residuals Breusch-Godfrey Serial Correlaion LM es Diagnosic ess Lag lengh=1 F-saisic Probabiliy Obs*R-squared Probabiliy Lag lengh=4 F-saisic Probabiliy Obs*R-squared Probabiliy ARCH es Lag lengh=1 F-saisic Probabiliy Obs*R-squared Probabiliy Lag lengh=3 F-saisic Probabiliy Obs*R-squared Probabiliy Whies Heeroskedasiciy es F-saisic Probabiliy Obs*R-squared Probabiliy Normaliy es Jarque-Bera Saisic Probabiliy Sabiliy ess Chow forecas es: forecas from 1997:4 o 2004:4 F-saisic Probabiliy Source: Auhors 25

27 Figure 1.1: Trends in variables used in he sudy 7.8 a) Logarihm of he seasonaly adjused GDP b) Logarihm of real moneary aggregaes M Base money c) Logarihm of (TBR91-DEP) d) Logarihm of he Nominal exchange rae (LNER) e) Inflaion Source: Auhors 26

28 Figure 1.2: Error correcion erm (Base money model) ECT Q1 1993Q3 1996Q1 1998Q3 2001Q1 2003Q3 Source: Auhors 27

29 Figure 1.3: ECM for real Base Money: Recursive coefficien esimaes for esing parameer consancy (Doed lines represen ± 2 sandard errors) Changes in real base money 3 2 Changes in GDP 2 1 Changes in GDP (one lag) Changes in GDP (wo lags) Changes in GDP (hree lags) Changes in (TBR91-DEP)- (hree lags) Chages in he nominal exchage rae Changes in nomonal exchahge rae (one Changes lag) in nomonal exchahge rae (hree lag Inflaion Inflaion (one lag) Inflaion (wo lags) Dummy Error correcion erm Source: Auhors 28

30 Figure 1.4: ECM for real M2: Recursive coefficien esimaes for esing parameer consancy (Doed lines represen 2 sandard errors) ± Changes real M2 (one lag) 2 1 Changes in real M2 (four lags) Changes in real GDP Changes in real GDP (one lag) Changes in (TBR91-DEP)-(one lag) Changes in (TBR91-DEP) - (hree lags) Changes in nominal exchange rae (wo lags) Changes in CPI Changes in CPI (four lags) Consan Dummy Error correcion erm Source: Auhors

31 Figure 1.5: ECM for real Base Money: Recursive residuals for esing parameer consancy (1995: :04) Recursive Residuals ± 2 S.E. CUSUM 5% Significance CUSUM of Squares 5% Significance One-Sep Probabiliy Recursive Residuals Source: Auhors 30

32 Reference Boughon, J. M. and Tavlas, G. S., Wha have we learned abou esimaing he demand for money? A mulicounry evaluaion of some new approaches. Inernaional Moneary Fund Working Paper 91/16. Baumol, W. J., The ransacions demand for cash: An invenory heoreic approach. The Quarerly Journal of Economics, Vol. 66, (4), pp Chow, G. C., Tess of equaliy beween ses of coefficiens in wo linear regressions. Economerica, Vol. 25, pp Chow, G. C., On he long-run and shor-run demand for money. The Journal of Poliical Economy, Vol. 74, (2). Cuhberson, K., e al., The demand for money: A dynamic raional expecaions model. Economic Journal, Royal Economic Sociey. Ericsson, N., Empirical modeling of money demand, Empirical Economics, Vol.23, pp Fischer, I., The Purchasing Power of Money. New York: Macmillan Friedman, M The Inerpolaion of ime series by relaed series. Journal of he American Saisical Associaion, Vol 57, pp Goldfeld, S. M., The demand for money revisied. Brookings Papers in Economic Aciviy, 3, pp Hendry, D.F., Economeric mehodology; a personal perspecive in advances in economerics, T. Bewley (ed.), Cambridge Universiy Press, M.A. Hensridge, N. M., The Reconsrucion of a macroeconomic daa se for Uganda, Cenre for he Sudy of African Economies, Working paper (67). Johansen, S., Saisical analysis of coinegraion vecors. Journal of Economic Dynamics and Conrol 12, (2/3) pp Johansen, S., Esimaion and hypohesis esing of coinegraion vecors in Gaussian Vecor Auoregressive Models. Economerica 59, pp Johansen, S., Likelihood-based inference in Coinegraed Vecor Auoregressive Models, Oxford Universiy Press, Oxford. Kararach, G., Evidence on he demand for money funcion in Uganda. UNICEF Zimbabwe Working Paper ( ). 31

33 Kaarikawe, M. and Sebudde, R. K., Is he reserve money programme sill a useful operaing framework for he conduc of moneary policy in Uganda? Bank of Uganda, Saff Papers, 1, (1), pp Keynes J. M., The General heory of employmen, ineres and money, Harcour, Brace and Company, New York; U.S.A. Levanikis J. 1993, Modeling money demand in open economies over he modern floaing exchange rae period, Applied Economics 25. Marshall A Money, credi and commerce, London: Macmillan Mishkin F., The economics of money, banking, and financial markes, Addison-Wesley, 1998, p Nachega, J.C., A coinegraion analysis of broad money demand in Cameroon. Inernaional Moneary Fund Working Paper 01/26. Nachega, J.C., Financial liberalizaion, money demand and inflaion in Uganda, Inernaional Moneary Fund Working Paper 01/118. Pigou, A. C 1917, The value of money, Quarerly Journal of Economics, 32, pp Sriram, S. S., Survey of lieraure on demand for money: Theoreical and empirical work wih special reference o error-correcion models, Inernaional Moneary Fund Working Paper 99/64. Sriram, S., A Survey of recen empirical money demand sudies, IMF Saff Papers, Inernaional Moneary Fund, 47. Tobin, J The ineres-elasiciy of ransacions demand for cash, The Review of Economics and Saisics, 38, (3) pp

34 The Bank of Uganda Saff Papers Journal Vol. 1, No Bank of Uganda TRADE & GROWTH IN AGRICULTURE: UGANDA S TRADITIONAL EXPORT POTENTIAL WITHIN THE EVOLVING MULTILATERAL TRADING REGIME Kenneh A. Egesa & Yusuf A. Abdalla Absrac Uganda s recen rade reforms Implemened under a comprehensive economic recovery program were parly inended o generae condiions ha would boos rade aciviies. This paper evaluaes poenial opporuniies for growh in Uganda s radiional agriculural expors, given he recen rade reforms. In evaluaing he poenials and effecs of rade policies ha may have affeced he conduc of agriculural aciviies in he pas, and in he fuure, given anicipaed improvemens in inernaional marke access opporuniies, and poenial domesic produciviy gains, we used price and oupu daa decomposiion. The idenified opporuniies o radiional expor producers and exporers would be based on hree major areas: Increased ransparency in pricing hrough promoion of compeiion, paricularly for ea and obacco, coninued invesmen in research o promoe produciviy in high yielding varieies, and disease and drough resisan varieies, he provision of adequae ransporaion and agro-processing infrasrucure, increased markeing o esablish brand ideniy, and invesmen in producion of value-added producs such as organic coffee. Keywords: Uganda; radiional expors; rading regime; recovery program. 33

35 I INTRODUCTION Agriculure dominaes economic and social aciviy in Uganda and employs abou 80 percen of he populaion. Over 70 percen of he rural populaion is direcly employed in agriculural crop producion. Even where agriculure may no be a major aciviy, i sill provides a higher source of employmen and income indirecly hrough agriculural processing, domesic rade, and ransporaion, among ohers. The agriculural secor accouned for nearly 35 percen of oal Gross Domesic Produc (GDP) as a 2003/04 (Uganda Bureau of Saisics Saisical Absrac, 2005). To dae, he secor coninues o be an imporan provider of inpus o muliple producion aciviies, including he manufacuring secor, which is radiionally heavily ied o agriculural secor developmen. Given ha over 85 percen of he populaion lives in rural areas and depend almos enirely on agriculure for heir livelihoods 20, he secor serves as a basic source of food self-sufficiency and securiy for he majoriy of he populaion. Agriculural oupu a he momen is esimaed o come from nearly 58 percen of he oal work force employed in agriculure (Governmen of Uganda, Povery Eradicaion Acion Plan, 2004/5 2007/8). Arable land for agriculural aciviy is fairly disribued hroughou he counry, wih an average landholding of abou 2 hecares per farmer. However, large differences in rural populaion densiies across he counry have ended o dicae he inensiy wih which communiies have had o engage in more organized agriculural aciviies across he populaion. The secor can be broadly caegorized ino moneary and non-moneary agriculure, which sems from he apparen differeniaion beween radable and non-radable agriculural producion pracices. Commercial agriculural farming is currenly no subsanially praciced, excep for sugar cane and ea producion as well as horiculural farming. Therefore, moneary agriculure by definiion basically comprises of cash crops, food crops and livesock producion, foresry and fishing 21 ; consiuing abou 58 percen of oal agriculural oupu, while non-moneary agriculure, which is devoed almos exclusively o subsisence agriculural farming consiue abou 42 percen of oal agriculural oupu. Over he pas decade, all agriculural sub-secors, excep cash crops, which was affeced by El Nino have performed relaively well, regisering growh raes of 3.5 percen and higher in 1997/98. Table 2.1 shows he disribuion of annual growh raes wihin he agriculural secor beween 1992/93 and 2003/ See Plan for Modernizaion of Agriculure: Eradicaing Povery in Uganda, A Governmen Sraegy and Operaional Framework, Minisry of Agriculure, Animal Indusry and Fisheries, Enebbe and he Minisry of Finance, Planning and Economic Developmen (MFPED), Kampala, Uganda. 21 Alhough he Agreemen on Agriculure does no explicily refer o fishing as an agriculural aciviy, i is referenced as such in Uganda. Therefore for purposes of complee coverage of he agriculural secor in Uganda, fishing is reaed accordingly as an agriculural pracice. 34

36 Table 2.1: Agriculural secor performances: Annual growh raes (1992/ /2004) Secor 1992/ / / / / / / / / / / /04 Moneary Agriculure Cash crops Food crops Livesock Foresry Fishing Non-Moneary Agriculure Food crops Livesock Foresry Fishing GDP growh rae Source: Uganda Bureau of Saisics (UBOS). In addiion o being a major employer wihin he economy, agriculure also is he main foreign exchange earner. Agriculural secor, which comprises of boh radiional and non-radiional expors consiue over 50 percen of oal expors. Tradiional expors include coffee, coon, obacco and ea, while he non-radiional expors comprises of fish, hides and skins, simsim, maize, beans, flowers, ec. Table 1.2 shows Uganda s expor performance over he period 1992/93 o 2003/04. 35

37 Table 2.2: Uganda s expor performance: 1992/1993 o 2003/04 in US$ millions 1992/ / / / / / / / /1 2001/2 2002/3 2003/4 Toal Expors o/w Agriculural expors Tradiional expors Coffee Coon Tea Tobacco Non-radiional expors Agriculural expors Fish & is prod. (Excl. regional) Fish & is prod. (Regional expors) Hides & skins Simsim Maize Beans Flowers Non-Agriculural expors Elecriciy Gold Oil re-expors Cobal Ohers Source: Bank of Uganda Besides buoyan growh in he agriculural secor, here has also been a concurren general increase in he diversificaion of merchandise expors since 1992/93, wih oal expor earnings seadily rising from US$ million in 1992/93 o US$ million in 2003/04. Of he main radiional expors, coon have rebounded eigh imes in value in 2003/04 compared o 1992/93, while obacco expors in 2003/04 grew six fold compared o is corresponding value in 1992/93. Non-radiional agriculural expors such as fish and fish producs, horiculural, hides and skins, maize and ohers have also increased, accouning o abou 30 percen of oal expors from only 16 percen in 1992/93. In addiion, oher new nonagriculural producs such as cobal and gold have furher boosed he conribuion of non-radiional expor earnings, which in a sense parly reflec he benefis of Uganda s ongoing expor diversificaion effors. The Expor performance has been underpinned by he governmen s conscious rade policy reform effors o promoe he expor growh. 1.1 Saemen of he problem and objecives of sudy The curren poliical economy of he world rading sysem demands ha all counries paricipaing in he design of he evolving mulilaeral rading sysem and rules hereof undersand and appreciae he reasons why naional auhoriies and 36

38 governmens ough o pursue liberal rade policies. The lieraure suggess ha rade policies are designed o improve and susain he welfare of ciizens. For his reason, rade policies are herefore usually implemened as an avenue for generaing revenues, for boosing domesic indusrial enerprise and o enable a shif in erms of rade owards improvemens in cerain economic and efficiency goals. Wihin he framework of regulariies peraining o membership of he world rading sysem under he WTO, governmens and naions may no necessarily be compelled o choose beween domesic or rade policies in order o achieve heir own objecives. However, hey are expeced o use insrumens of rade policy provided ha hese insrumens are founded on and accompanied by accepable disciplines. Thus, while counries are basically free o apply rade policies in pursui of heir own naional objecives, hey are equally encouraged o do so using he leas rade disoring measures, consisen wih wha economic heory would recommend in many circumsances. The quesion neverheless would be ha of wha rade policy insrumen o use. The curren lieraure ends o caegorize rade policy insrumens ino hree broad forms, namely hose measures ha affec quaniies hus resricing he volume or value of ransacions; hose ha affec prices (involving he imposiion of a moneary fee (ax) on supplies from abroad) and hose ha may indirecly affec quaniies and prices ogeher. This is because virually any policy or governmen inervenion may have an effec on he conduc of rade anyway (see Hoekman and Kosecki, 2001 for appropriae references o he discussion of he many ypes of policy insrumens ha have an effec on rade). The challenge for many governmens herefore lies in he need o successfully idenify and implemen he opimal combinaion of rade policies necessary for delivering inended objecives. This sudy specifically focuses on he analysis of he rade policies pursued so far and prospecs ha would allow Uganda o idenify and pursue addiional sraegies aimed a boosing he conribuion of he agriculural secor wih a bias owards radiional agriculural expors. Emphasis on radiional expors in his analysis is ideal because of heir imporance as major sources of income for many Ugandans, since hese are producs commonly grown in mos pars of he counry and provide he much-needed employmen for he rural populaion. In addiion o radiional expors linkages o agro-processing aciviies, boh coffee and ea are known o be perennial crops whose annual growing coss reduce once hey have reached heir final producion sages. The main objecive of he sudy is herefore o idenify he poenial opporuniies ha may be available for Ugandan producers and exporers of radiional agriculural producs. Subsequenly, inferences abou he benefis and pifalls of boh direc and indirec effecs of rade policy on he domesic prices and oupu of radiional expors are evaluaed o deermine wheher he producs could deliver cerain naional objecives ha ofen impac posiively on oal expor performance. The specific objecives include he deerminaion of he effecs of rade policies on: i) prices of radiional expors and ii) supply responses of seleced radiional expors o specific rade policy incenives. 37

39 The analysis is expeced o provide a broad assessmen of he counry s abiliy o compee efficienly in he world and regional markes for agriculural producs, by aking advanage of emerging mulilaeral rading rules hrough complemenary rade policies and heighened invesmen incenives paricularly in he more producive areas of radiional expors. Secion 1 of he paper discusses Uganda s experiences wih economic and rade policy reforms. Secion 3 provides a succinc clarificaion on impending mulilaeral rules on rade in agriculure as well as Uganda s specific ineress. Secion 4 evaluaes he impac of recen economic reforms on agriculural secor performance wih emphasis on radiional expors, and Secion 5 provides concluding remarks and recommendaions. II ECONOMIC AND TRADE POLICY REFORMS IN UGANDA By independence in 1962, Uganda s economy was arguably one of he fases growing economies in Sub-Saharan Africa, wih remarkable growh prospecs for he fuure. Economic performance was quie robus, aaining an average GDP growh of abou 6 percen beween 1963 and However, subsequen domesic poliical upheavals ha ensued during he 1970s led o unfavorable reversals of he gains aained earlier, due mainly o a seady slump in producion and economic aciviy. Agriculural producion in paricular slumped considerably, wih he major expor crops declining subsanially (excep for coffee, which a he ime araced high inernaional prices). As a resul, Uganda s balance of paymens, which before 1973 had observed seady surpluses, reversed ino defici posiions, which in urn resuled o a consisen running down of foreign reserves, while he Uganda Shilling became unrealisically overvalued. Access o foreign exchange was heavily raioned (Abdalla, Musinguzi, Shaw, Plunke, 2000). 22 Afer he miliary regime in 1979, Uganda began a new era of economic reorganizaion. A firs new phase of economic recovery, suppored by he Inernaional Moneary Fund (IMF) and he World Bank, characerized by srucural adjusmen programs, began in In his phase, measures were adoped o depreciae he Uganda Shilling in order o realign is exernal value wih marke expecaions. An aemp was also made o dismanle price conrols, while incenives o agriculural producion were insiued hrough upward adjusmen and realignmen of producer prices. The firs phase of he economic recovery was however shor-lived due o he governmen s failure o mainain he required disciplines ha would propel a relaively smooh recovery process. A second phase of economic reconsrucion was launched in 1987, again wih he suppor of he IMF and he World Bank. This phase among ohers was o implemen a comprehensive Economic Recovery Program (ERP), wih paricular focus on macroeconomic sabiliy; he liberalizaion of foreign exchange, rade, prices, and 22 The was a general collapse of manufacuring aciviy especially following he expulsion of Asians and he naionalizaion of key indusries by he sae. Subsequenly, he lack of appropriae echnical and compeen managerial skills led o exensive mismanagemen of hese indusries, wih adverse consequences for he economy. 38

40 markeing sysems; he review of incenive srucures and he business climae o enable he promoion of resource mobilizaion, invesmen and growh in expors; and he rehabiliaion of socio-economic infrasrucure and insiuional reforms. Trade policy was o be geared o complemen specific expor promoion measures so as o reduce ani-expor biases (see Abdalla and Egesa, 2004), and (Hinkle, Herrou- Aragon, Krishnamani, and Kreuzwieser, 2005) for a deailed review of economic reforms underaken since 1987). Among he key reforms implemened during his phase were: 2.1 Insiuional reforms and privaizaion Due o he dominance of agro-processing aciviy in he manufacuring secor, he privaizaion campaign recognized he role of he privae secor in propelling and susaining agriculure based processing aciviies. The major aciviies in agroprocessing, which were a poenial arge for growh under he privaizaion program included exile manufacuring, obacco processing, leaher and coffee processing. The fas and effecive dismanling of governmen monopolies in a number of agro processing indusries led o increased compeiion as numerous muli-naional corporaions began o impor capial and o se up firms 23. By 2001, he implemenaion of he privaizaion programme had conribued o exensive divesiures by he Sae of 108 ou of 148 public enerprises. 2.2 Inflaion conrol and he conduc of moneary policy in Uganda The aspiraions of he ERP dicaed ha inflaion be conrolled, mainly hrough he silencing of he prining press a he Bank of Uganda on he one hand, and resricing undue governmen borrowing on he oher hand. This was criical for purposes of amelioraing economic condiions ha allow farmers o obain realisic prices for heir agriculural produce and hence increased incomes. During he early 1990s, annual percenage changes in headline inflaion averaged abou 38 percen compared o an average of 3.5 percen observed during he early par of Liberalizaion of he exchange rae Afer launching he ERP in 1987, several bous of devaluaions followed in 1987, 1988 and 1989, accompanied by a one off currency reform, which demoneized he Shilling, while also aracing a 30 percen conversion ax. The measures were aimed a narrowing he differenial beween he official and parallel exchange raes. This episode of reform was laer followed by he adopion of a crawling peg and he legalizaion of he black marke in These developmens ogeher helped o rim down pressures on official channels of foreign exchange and provided addiional incenives o increase agriculural expors. The Figure 2.1 below depics he official and bureau mid exchange raes. 23 Cases in poin include governmens shares in he Rwenzori Highland Tea Company Ld (RHTCL) bough by RHTCL owned by Finlays Group, sale of governmens 20 percen sake in Briish American Tobacco Ld (BAT) o BAT and he lising of he remaining 10 percen on he Uganda sock exchange, Blenders Uganda Ld o Uniliver, Kakira Sugar Works, Kibimba Rice Co ld, Uganda Tea Corporaion ec. 39

41 Figure 2.1: Trends in Uganda's official & forex bureaux mid exchange raes ( ) 2,500 2,000 1,500 1, J ul- 90 J ul- 91 J ul- 92 J ul- 93 J ul- 94 J ul- 95 J ul- 96 J ul- 97 J ul- 98 J ul- 99 J ul- 00 J ul- 01 J ul- 02 J ul- 03 J ul- 04 Bureau Middle Rae Official Middle Rae Source: Bank of Uganda 2.4 Expor credis and guaranee schemes Subsequen o he removal of credi conrols and he suspension of direced credi lines hrough governmen programs such as he rural farmers scheme and crop finance in 1989, credi o he agriculural secor slumped. The removal of credi conrols herefore negaively affeced he agriculural secor as banks oped for less risky secors such as he services and manufacuring. Ugandan exporers of agriculural producs were generally consrained by lack of access o reasonably priced working and invesmen capial. Consequenly, he marke for credi o agriculural enrepreneurs was characerized by high lending raes, which ranslaed ino high coss of borrowing and doing business. The credi consrains o he agriculural secor led o a decline in he compeiiveness of Uganda s agriculural expors. Besides, he scany loans available o exporers a he ime, he mos available loans were of shor-erm naure. The hardly available long-erm working capial a reasonable prices herefore couldn enable invesors o pu up he necessary expor infrasrucure. To parly address he consrains, he Bank of Uganda (BOU) ogeher wih paricipaing commercial banks decided o iniiae a faciliy o provide medium o long-erm financing o poenial exporers hrough a number of credi lines and suppor schemes. These credi lines and schemes, suppored by exernal donor agencies, offered loans o small and medium sized privae enerprises hrough licensed banks The main credi lines under hese expor schemes include: he Apex Privae Secor Loan Scheme (Apex), which was designed o suppor small and medium scale privae secor enerprises in manufacuring, agro-processing, services, and ourism aciviies; he Expor Refinance Fund (ERF) designed o provide working capial o exporers of non- radiional expors; he Disressed Flower Projec Fund (DFPF) esablished o rescue flower farms in 1999/2000 o miigae working capial consrains; he Expor Promoion Fund (EPF)- par of he Apex credi line- esablished o assis exporers in promoing heir respecive expors and o idenify new markes; and he Developmen Finance Fund (DFF) se up in 1986 mainly o suppor producion in agriculural secor. This scheme in paricular is however being phased ou because of is poor performance. Oher expor marke credi schemes adminisered in he pas have included he Invesmen Term Credi Refinance Fund (ITCRF); and he Coon Sub -Secor 40

42 2.5 Liberalizaion of rade in agriculural inpus and oupus In erms of oupu, he governmen s monopoly on rade in coffee and coon hrough he Coffee Markeing Board (CMB) and Lin Markeing Board (LMB) respecively, resuled in farmers receiving low prices for heir oupu. Accordingly he paricipaion of hese parasaal organizaions in rade was derimenal o farmers. Farmers herefore were ofen forced o reac by swiching o oher crops. In 1991 and 1993, he CMB and LMB monopolies were however abolished. The aboliion of he parasaal monopolies o allow markes o se prices freely evenually led o increased paricipaion of he privae secor hus improving compeiion and beer prices for he farmers. In he case of inpus, impor secor reforms were inroduced, paricularly in cusoms. Quaniaive resricions were eliminaed. Impor licensing was abolished, excep for iems in he Negaive Lis 25. There are however resricions on impors of agrochemicals and pharmaceuical producs. In circumsances where he agro-chemicals and pharmaceuical producs have o be impored, approval mus be sough from he Agriculural Chemicals Conrol Board in respec of agro-chemicals, and from he Naional Drug Auhoriy (NDA) for pharmaceuicals. Imporers of agriculural chemicals mus normally apply o regiser and obain licenses, and o be cerified by he Agriculural Chemicals Conrol Board under he Agriculural Chemicals (Regisraion and conrol) Regulaions Similarly, he imporers of pharmaceuical producs mus apply for regisraion and cerificaion by he NDA under he Naional Drug Policy and Auhoriy Saue In addiion, ariff raes have been reduced, ofen significanly, and many non-ariff resricions such as quoas and impor bans have been convered ino ariff equivalens. The ariff schedule wih raes of zero, 10, 20, 30 and 60 percen in 1995 was reduced o a sandard schedule wih raes of zero, 7 and 15 percen in 2001, alhough some goods faced higher raes. New ariff raes are now a 0, 10, and 25 percen in line he implemenaion of he Eas African Cooperaion Agreemen. 2.6 Oher expor secor incenives and marke access opporuniies In 1993, only coffee exporers were required o be regisered, and shipmens of coffee were subjec o declaraions of "indicaive" prices and qualiy conrols on accoun of he special case naure of he commodiy in he inernaional marke. In hose circumsances, he governmen of Uganda reserved he righ o resrain expor volumes "in line wih demands of inernaional obligaions on coffee rade". In addiion, he Coffee Reenion Scheme esablished a he ime and agreed o by major producing counries in 1993 allowed exporers o reain a proporion of heir proceeds in foreign exchange abroad. A Coffee Sabilizaion Tax was inroduced in 1994 afer he inernaional coffee prices sho up significanly. The ax was boh a fiscal and policy insrumen used o ease upward pressure on he exchange rae resuling from inernaional coffee price increases and o improve condiions for oher expor or impor compeing secors. 25 A lis of producs declared resriced under mulilaeral rading rules for sraegic, saniary and phyo-saniary and securiy reasons. Uganda has a lis of hese producs declared. 41

43 However, he coffee ax was abolished, and o dae, he only expor duies ha coninue o apply relae o expors of hides and skins and a 1 percen levy colleced by he Uganda Coffee Developmen Auhoriy on coffee expors. Oher forms of expor incenives offered in he early 1990 s included compensaions o exporers for infrasrucure inadequacies and he cos of bureaucraic ineria. The prominen ones included: exporers being eniled o reaining 100 percen of heir foreign exchange proceeds from expors; and excep for expors of hides and skins, all forms of oher expors remained zero-raed for Value-Added Tax (VAT) in order o reduce coss of expors 26. Manufacurers and exporers of goods and services also have access o a duy drawback in order o enable manufacurers and oher exporers o compee in foreign markes wihou undue handicaps arising from duies paid on impored inpus. In a sense, his allows exporers o drawback up o 100 percen duies paid on inpus impored for making goods for expor. A bond scheme for exporers was esablished in order for hem o seek cusom licenses, o hold and use impored raw maerials, which are used for manufacuring goods for expors wihou paymen of axes. This enabled hem o redeem some amoun of working capial, which would oherwise be ied up due o paymen of duies on impored raw maerials. Wih regard o incenives drawn from numerous sources of marke access, especially for agriculural expors, Uganda is srongly commied o supporing parnerships in global economic inegraion. Currenly, he counry is a member of he Common Marke for Easern and Souhern Africa (COMESA) 27, a signaory o he ACP-EU Coonou Agreemen and he African Growh and Opporuniy Ac (AGOA). Wihin hese parnerships, Uganda s expors fully qualify for preferenial ariff raes under COMESA and EAC, and also ener he European Union and US markes duy and quoa free. III MULTILATERAL TRADE RULES, AGRICULTURAL EXPORTS AND UGANDA S SPECIFIC INTERESTS Prior o and following he onse of he WTO agreemen on agriculure, rade in agriculural producs has coninued o pre-occupy he ongoing negoiaions of he agreemen on agriculure. The core subjec of negoiaion relaes o he dispues among he world s rading parners on he absence of agreed mulilaeral rading disciplines. For a considerable period of ime during he GATT regime, which preceded he WTO, all counries mainained sric proecive rade policies on agriculure. However, he Uruguay Round Agreemen managed o esablish a srucured discipline for governing global liberalizaion in agriculure wih respec o border conrols and subsidies agriculure by various governmens. In many ways, he agreemen inends o resrain naional policies ha, on a global scale, creae disorions o agriculural markes by encouraging counries o liberalize heir 28 agriculural secors. 26 However, exporers are required o remain VAT regisered a all imes. This is inended o enable hem reclaim VAT expended on all inpus used in he process of producing and processing expors. 27 This is he major desinaion for Uganda s maize and beans expors. 28 A comprehensive review of he WTO Agreemen on Agriculure (AoA) and is relaed provisions is available in he WTO Legal Texs (2002). 42

44 3.1 Uganda s specific ineress in he Agreemen on Agriculure (AoA) The WTO agreemen on agriculure embodies a commonaliy of ineress for many developing counries ha expor agriculural producs. Uganda s specific ineress in he negoiaions have focused mainly on he following ideals: marke access opporuniies; a sraegy o reform domesic suppor measures o mee compaibiliy wih WTO guidelines; and developing counries ineress for he mulilaeral rading communiy o respec he meris of Special and Differenial Treamen (S&D) provisions accorded o Leas Developed Counries (LDCs) wihin he framework of he evolving Doha Work Programme (DWP), among ohers. On Marke access, boh ariff peaks and ariff escalaions have been major impedimen o producs of expor ineres o Uganda. Uganda s ineress have herefore included he call for he eliminaion of ariff peaks and ariff escalaions and he reenion of Special and Differenial (S&D) provisions allowed o developing counries o enable hem o benefi from exising rade preferences offered o hem wihin he framework of mulilaeral rading rules and oher regional and bilaeral rading arrangemens. Regarding he need o bring domesic suppor measures in line wih mulilaeral rading norms, Uganda has coninued o argue ha he curren provisions of domesic suppor measures remain unbalanced in favor of developed counries, in spie of relaed disciplines, which are being negoiaed wihin he Agreemen on Agriculure in he conex of he Doha Developmen Agenda. In is submissions o he negoiaing processes, Uganda has herefore called for he removal or simplificaion of domesic suppor measures, paricularly on exemp policies since hey have ended o disor marke access opporuniies for agriculural producs o he disadvanage of poor counries. Suppor for expor promoion measures consisen wih provisions under he Green and Amber Boxes for LDCs (alhough in par sill a subjec of furher review and negoiaions wihin he erms of WTO guidelines) consiue one of he erms of Uganda s ineress for furhering is marke access opporuniies. Regarding Saniary and phyo-saniary (SPS) Sandards, measures need o be aken o proec human, animal or plan life or healh based on scienific principles and, as far as possible, based on accepable inernaional sandards and guidelines. IV IMPACT OF REFORMS ON AGRICULTURAL PERFORMANCE This par explores he exen o which focused rade policy reforms can have a useful long-erm impac on agriculural expor performance. We evaluae changes in relaive prices and quaniies produced of four radiional expors as inermediae oucomes of specific rade reforms. The radiional expor commodiies seleced are coffee, coon, obacco and ea, based on he individual conribuion of hese commodiies o oal agriculural producion and expors and he imporance o he 43

45 livelihoods of heir producers. The oher crierion was based on he availabiliy of daa on hese expor commodiies. 4.1 Coffee Uganda produces abou 3.5 million (60 kg) bags of coffee annually from abou 300 million rees. Coffee provides he main source of livelihood o he majoriy of smallholder rural households especially in he cenral and wesern regions of Uganda. Abou 20 expor firms expor much of he coffee produced mainly in form of unprocessed green beans. Boh coffee producer prices and world prices rose seadily beween 1987 and 1999 wih he dispariy beween he world and producer prices increasing as well. From 1996 however, coffee prices rended downward wih a declining dispariy beween world and producer prices. The downward rend was parly on accoun of he governmen s policy o revive he coffee secor, which had been hi by he huge decline in prices beween 1996 and Figure 2.2 below shows he evoluion of coffee prices since Figure 2.2: Coffee world and producer prices World price (Ug shs/kg) Domesic price (Ug shs/kg) coffee world prices coffee domesic producer prices Source: Auhors 4.2 Tea Uganda s annual ea producion is esimaed o be abou 30,000 meric ones of made ea from 21,000 hecares. Abou 45 percen of ea oupu is realized from smallholder producion, while he res is from commercial ea esaes. The ea indusry generaes abou US$ 30 million per annum a an average price of US$1.1 per kg. The markeing of 85 percen of Uganda s exporable ea is done a he regional aucion marke in Mombassa, while 3 percen is sold by privae reay and abou 1 percen is disposed off a he London aucion. Producer prices for ea rose seadily in he lae 1980 s leveling off in 1992 and saying here for abou 5 years, afer which prices hen only rose minimally in 1998 and have sabilized a abou he same level since hen. In he lae 1980 s, ea prices rended ogeher wih world prices. However, beween 1992 and 1999, he dispariy beween domesic and world prices widened, as he world prices declined while domesic prices remained a he same level. 44

46 Figure 2.3: Evoluion of ea prices since Worl d pric e ( Ug shs/kg) g shs/kg) Producer price (U ea world ea domesic Source: Auhors 4.3 Coon Coon producion in Uganda reached a peak of 467,000 bales in 1969/70, bu has since hen dropped. For insance in 1993/94, producion had dropped o a mere 11,000 bales. However, afer liberalizing coon markeing in he counry, producion sared picking up and by 2003/04 was esimaed a abou 100,000 bales per annum. This is noneheless way below he ginning capaciy, which is currenly a abou 800,000 bales per annum. In erms of expors, Uganda earns abou US$20 million per annum from expors of lin. Only 5 percen of oal coon producion is processed beyond he lin producion sage. The prices for coon (boh producer and world prices) have followed similar rends since 1987, wih he excepion of he period beween 1997 and 2002 when domesic producer prices were on he rise conrary o declining world prices. Figure 2.4 below shows he evoluion of coon prices since Figure 2.4: Coon world and producer prices World price (Ug s hs/kg) shs/ Producer price (Ug Kg) Source: Auhors coon world coon domesic 4.4 Tobacco Abou one million people in Uganda derive heir livelihood from obacco producion eiher direcly or indirecly. Tobacco is also a major foreign exchange earner for Uganda; esimaed o earn abou US$20 million per annum. Beween 1987 and 1992, boh producer and world prices of obacco rose, before slumping and shorly recovering in The dispariy beween he wo prices has narrowed over he las five years. Figure 2.4 shows he evoluion of obacco prices since

47 ) Figure 2.5: Tobacco world and producer Prices Wo rld pri ce (Ug shs/kg obacco world obacco domesic Producer price (Ug shs/kg) Source: Auhors 4.5 Price decomposiion We price decomposiion is analyze he exen o which policy reform processes explain evoluion of marke prices of commodiies 29. However, i is worh noing ha domesic prices are affeced by oher facors, noably world prices, exchange raes, ariffs, domesic markeing margins and insiuional environmens. The effecs of hese oher facors can be measured by compuing percenage conribuions of hese various facors o changes in marke prices in order o pin down he relaionship beween changes in price incenives and he reform measures. We adop he mehodology by Quiroz and Valdés, (1993) o esablish he exen o which rade reform processes could explain he evoluion of producer prices for he four radiional agriculural commodiies in Uganda. According o Quiroz and Valdés, (1993), he assumed domesic price of a produc, for some period 0, can be deermined by: P d 0 = P w 0 * E 0 *(1+ 0 ) *(1+ c 0 ) (2.1) Where: P d is domesic price (e.g. a farm level) P w is world price E is he exchange rae (1+) is he ad-valorem ariff (1+c) is oher proporional coss (e.g. ranspor and markeing coss). A similar relaionship can be defined for some oher period (period 1 ) as: P d 1 = P w 1 * E 1 * (1 + 1 ) * (1+ c 1 ) (2.2) The linear represenaion of he above equaions, aking naural logs (ln) and subracing equaion ( 2.1) from (2.2): (lnp d 1 - ln P d (lnp w w 0)= 1 - lnp 0 )+(lne 1 - lne 0 )+(ln 1 - ln 0 )+(lnc 1 - l nc 0 ) (2.3) 29 FAO. Trade reforms and food Securiy: FAO

48 Since he firs-order difference of equaion (2.3) yields he (approximae) percenage change (afer muliplying by 100), he cha nge in he domesic price can be decomposed in a way ha can be inerpreed o reflec he individual conribu ions of he four facors, w hich if summed can add o 100. The daa for producer prices used in he analys is was obained fro m he Minisry of Agriculure, Animal Indusry and Fisheries, while daa on he exchange raes was from he Bank of Uganda, and daa on World prices from he IMF Inernaional F inancial Saisics he IMF 30. Equaion (2.3) herefore esimae s he percenage change in he domesic price beween periods represened on he lef hand side. This percenage change was hen decomposed ino percenage changes over he same period for he hree componens ha consiue he domesic price. These componens include prev ailing world prices, exchange raes and he effec of dom esic facors, including he policy environmen. The resuls of he price decomposiion for he key agriculural expor commodiies are in he Table 2.3 below. Table 2.3: Decomposiion of he sources of changes in producer prices of seleced agriculural expors Commodiy Period Change in Producer Change in World Change in Exchange Oher Price Price Rae Facors Coffee Tea Tobacco Coon Source: Auhors 30 Due o lack of sufficien informaion on he las wo iems we proceeded o combine heir effecs o represen he effec of domesic facors (policies and oherwise) 47

49 Policy inervenions in Uganda have ypically reversed he negaive effec s of domesic facors on producer prices of all radiional expors. Specifically, in he case of coffee, he conribuion of domesic facors has improved from a high negaive impac of 356 o a posiive impac of 2, while for ea; he improvemen was from 374 o 11. For obacco and coon, he conribuion of he domesic facors o producer prices improved from 327 o 18 and 377 o 23, respecively. There was a marked coninued improvemen due o he impac of domesic facors o producer prices, despie he fac ha heir conribuions remained negaive for coon, obacco and ea in he wo-reform periods-1993 o 2000 and 2000 o Table 3 shows ha some work sill needs o be done o ransform he negaive impac of he domesic facors ino posiive effecs, especially for ea, obacco and coon. In he case of coffee, i should increase as relaive sabiliy in he exchange rae and world prices have been prevailing o a large exen. Boh he exchange rae and inernaional price changes have coninued o impac posiively on producer prices, alhough in boh cases he posiive effec has been declining. Price ransmission and marke inegraion The magniude of price ransmission as in he conex of marke inegraion is useful in he analysis of world and domesic prices as well. The objecive of some policy reforms in Uganda was o increase he degree of marke inegraion during he pos reform period. We compue he correlaion beween he changes in world prices and producer prices during he respecive periods in order o esablish wheher shocks in world prices as a resul of inegraion were ransmied direcly o producer prices. Table 2.4 shows he correlaion coefficiens. Table 2.4: Correlaion coefficiens of producer prices and world prices for he radiional expor s, Period Coffee Tea Tobacco Coon Source: Auhors The correlaion coefficiens for coffee indicae an improvemen from 0.63 in he period o 0.97 and 0.99 beween and , respecively. Overall, he price developmens for coffee have followed he rend of world coffee prices. The reforms carried ou in he economic adjusmen period however, had a negaive impac on he ea indusry, while for obacco and coon, he posiive conribuion of he reforms owards inegraion were minimal. This mixed oucome maybe aribued o difficulies in designing muliple reforms simulaneously. Indeed, he resuls sugges ha coffee producion is being feaured as he key crop for reforms presumably because of is poenially high conribuion o overall expors. The resuls furher demonsrae, i is also parly a consequence of low compeiion in he obacco and ea markeing during he pos reform period, which has been lef o a few privae secor players whose aim a higher profi margins. 48

50 4.6 Markeing, ranspor and oher margins The margins have inherenly remained large in Uganda for a number of reasons. Firs, high ransporaion coss due o Uganda s landlocked siuaion, compared o oher compeing counries, which have direc access o pors and which are closer o he European and American markes. This is exacerbaed by he high coss of ransporing produce from he rural agriculural areas o he urban areas for processing and for onward disribuion o he inernaional markes. The high coss are a resul of high fuel coss and inefficien ranspor sysem marred by he poor rural feeder roads, among ohers. Noneheless, governmen has resu rfaced a number of rural feeder roads in an aemp o miigae domesic ranspor consrains. Oher effors have been direced owards liberalizing rade in agriculural expor commodiies o promoe compeiion and he decision o dives he Uganda R ailways Corporaion in order o ease ransporaion o seapors. 4.7 Supply response We analyze daa on aggregae commodiy producion o deermine he proporion of changes in oupu levels ha can be aribued o changes in he harvesed area and changes in yield. Oher lieraure have used linear programming mehods, paricularly for farm level dae in order o deermine how farmers would have responded hrough a wha if ype of analysis, raher han quanifying pas responses. However, i is quie challenging o obain he required deailed farm level daa. Moreover, good qualiy survey daa on crop producion, area harvesed and yield is equally no readily available in Uganda. Therefore, we use he mehod of oupu decomposiion for coffee and coon, whose daa was derived based on desk esimaes of yield and area harvesed, augmened wih he available producion daa o deermine he conribuions of changes in he harvesed area and changes in yield ha have led o changes in aggregaed commodiy producion. We exended he price decomposiion mehodology as expressed in equaion (2.3), and regressing he change in aggregae supply (as he dependen variable) on he area harvesed, yield and a residual, obain he following coefficiens: Table 2.5: Decomposiion of he sources of changes in producion of coffee and coon, commodiy period producion Area harvesed Yield Ohers Coffee Coon Source: Auhors 49

51 A one-uni change in area harvesed of coffee leads o approximaely 216 increase in coffee oupu during he period The coffee secor s expansion during his period suggess he impacs of a composie of facors such as non-price policy and insiuional reforms ha were underaken in Uganda a he ime. This couples dismanling of governmen s monopoly in coffee expor and markeing. The negaive conribuion of he area harvesed owards coffee producion during he period afer 2000, migh be explained parly by he decline in world coffee prices afer 1998, as farmers subsiued oher producs for coffee. The effecs of he coffee wil disease, which led o he burning of several coffee planaions o conain he spread of he disease o oher planaions also explains he negaive impac of he areas harvesed. The governmens expor secor diversificaion and promoion measures, which have resuled ino farmers venuring ino new agriculural commodiies such as flowers, vanilla, fruis, and vegeables, pepper, papain ec migh have also negaively affeced coffee expansion. However, he inroducion of a new high yielding variey of coffee such as robusa specie-clonal coffee migh have conribued o a general increase in he coffee yields. Figure 6 depics developmens in he yield and area harvesed of coffee. Figure 2.6: Coffee yeild and area harvesed Area Harv esed (millions of hecares) Yield (Tonnes per hecare) Source: Auhors coffee area harvesed coffee yield Changes in coon producion on he oher hand, have been driven by increases in boh yield and area of harvess in equal proporions over he enire period wih no conribuion from oher facors. The policy and insiuional reforms migh have played a minimal role in raising coon oupu, excepion during he period , when governmen monopoly in coon expor and markeing was dismanled. There has been a general increase in area harvesed afer he firs reform period ( ), mainly due o he increased emphasis on coon and exiles indusry as whole. Schemes such as he AGOA have eniced invesors in he exiles indusry; who have suppored farmers by providing chemicals, seed, equipmen and sarup capial. However, heir consumpion of coon has remained very low; abou 5 percen of oal producion. Figure 2.7 below shows rends of area harvesed and yield. 50

52 Figure 2.7: Coon yield and area harvesed (millions of es) d ar Area Harvese Hec ecare) per H Yield (onnes coon yield coon area harvesed Source: Auhors and UBOS The El Nino floods of 1999 migh have also conribued o he decline in area harvesed of coon, as he main coon-growing bel in Uganda was one of he areas affeced by he floods. This, nowihsanding, he conribuion of governmens rade policies owards simulaing higher oupu do no seem o have been as successful as price relaed governmen inervenions. V CONCLUSION The analysis generally suggess ha governmen s policy inervenions in rade have amelioraed he domesic facors consraining improvemens in producer prices of he radiional expors. The bigges improvemen in producer prices has been regisered for coffee followed by ea, obacco and lasly coon. However, even wih he general improvemen, he margins beween he producer prices and world marke prices have remained large paricularly for obacco and ea. These wo crops also happen o have domesic prices ha are leas sensiive o inernaional prices. This is parly explained by he small number of players paricularly in he case of obacco who decide on wha price o pay farmers. This price is usually fixed for some ime even when inernaional prices rise. In he case of ea, he reliance on he Mombassa aucion marke, which increases ransacions, has conribued owards he high margins. In addiion, he high ransporaion coss incurred for moving he commodiies o inernaional markes as well as from he rural areas o collecion and processing ceners for onward ransporaion have conribued o he high margins. Moreover, he lack of domesic branding of he produce o sui specific inernaional ases and qualiies has also negaively affeced producer prices. 51

53 The supply response o policies underaken by governmen appears o have been mos successful beween 1993 and 2000 for boh coffee and coon. However, oal oupu for coffee has declined in general on accoun of a decline in area harvesed. The declining area harvesed is due o farmers swiching o oher subsiue crops paricularly he non-radiional expors as well as he desrucion of planaions due o he coffee wil disease. In he case of coon, he inabiliy o adop new high yielding varieies has conribued o he slow growh in oupu. The idenified opporuniies open o radiional expor producers and exporers lies in hree major areas. The firs is increased ransparency in prices hrough promoion of compeiion paricularly for ea and obacco. This would have he advanage of exposing Uganda s radiional expors as hey become more compeiive in regional and global markes. The second opporuniy is in coninued invesmen in areas ha promoe produciviy such as research inended o provide high yielding and disease and drough resisan varieies and provision of adequae supporive infrasrucure paricularly for ransporaions and agro-processing. The research should also focus on improving qualiy and markeing channels for he commodiies if Ugandan expors are o penerae boh regional and global markes. In paricular, combaing he Coffee Wil Disease should remain high on he agenda of agriculural research insiuions. I is herefore imperaive for governmen o mobilize sufficien resources o fund he research aciviies, required infrasrucure developmens as well ensuring ha farmers are able o access he improved varieies. The oher opporuniy is in increasing Markeing effors o esablish brand ideniy and invesmen in producion chains for value-added producs such as organic coffee needs o be underaken. The success sories of coffee markeing in some European markes and China should provide models for pursuing his paricular opporuniy among he oher radiional expors paricularly in he case of ea which is a near subsiue for coffee. Emphasis by auhoriies should herefore be direced owards increasing access o credi faciliies o rural farmers, increased exension services, and marke informaion o ensure modernizaion of agriculure, rural elecrificaion o promoe agro processing and improvemen of suppor infrasrucure such as feeder roads and inernaional ransporaion faciliies ec. Furher analysis of wheher radiional agriculural expor producs are ruly compeiive on he global scene in ligh of he domesic consrains hey face is imperaive o ensure he efficacy of fuure invesmens in hose areas. Markeing effors o esablish brand ideniy and invesmen in producion chains for value-added producs such as organic coffee needs o be underaken. For coon, here is need for addiional emphasis owards exiles producion for expor wihin he region and o inernaional markes. This would help o increase he share of coon absorbed in he secor given he implied increase in value addiion involved in ha process. Zimbabwe s decline in obacco producion has opened a small window of opporuniy for Uganda on he inernaional marke considering he wo counries obacco expors are quie similar. Governmen should arac oher players in he marke besides Briish American Tobacco Ld o promoe compeiion in addiion o farm requiremens such as seed, pesicides and credi hus providing incenives for farmers o increase producion. IN he case of ea, Ugandan producers should consider developmen of a niche marke o ake advanage of he healh premium and higher prices perceived o be associaed 52

54 wih consumpion of organic (i.e. non-irradiaed) eas. Oher consideraions should include he developmen of a domesic ea aucion marke o avoid over reliance on he Kenya aucion marke wih is associaed high ransacion coss. References Abdalla, A. Y. e al., Opporuniies for Ugandan expors in he conex of WTO rade liberalisaion: Analysis and recommendaions. Prepared for he Uganda naional conference on rade policy, Kampala. Nahan Associaes Inc., Arlingon, Virginia (available on reques). Abdalla, A. Y., and Egesa, K. A., African Imperaives in he new world rade order: A case sudy of he agriculural secor in Uganda. African Economic Research Cenre (AERC). Bank of Uganda, Annual Repor Kampala: Bank of Uganda. P anagariya, A Trade and food securiy: Concepualizing he linkages. Food and Agriculural Organizaion, Rome. Governmen of Uganda, Plan for modernizaion of agriculure: Eradicaing povery in Uganda, a governmen sraegy and operaional framework, Minisry of Agriculure, Animal Indusry and Fisheries. Enebbe: Minisry of Agriculure, Animal Indusry and Fisheries. Hoekman, B. and Kosecki M., The poliical economy of he world rading Sysem: The WTO and beyond. 2 nd ed. New York: Oxford Universiy Press. Quiroz, G and Valdes, E, Agriculural incenives and African compeiiveness in four African counires: Governmen inervenions and exernal shocks. In Valdes, A. and Kay Muir-Leresche, K. (eds): Agriculural policy reforms and regional marke inegraion in Malawi, Zambia and Zimbabwe. Washingon, DC: IFPRI. Uganda Bureau of Saisics, Saisical Absracs Kampala: Uganda Bureau of Saisics. WTO, The legal exs: Resuls of he Uruguay Round of rade negoiaions. World Trade Organizaion, pp

55 The Bank of Uganda Saff Papers Journal Vol. 1, No Bank of Uganda HAS EXCHANGE RATE VARIABILITY AFFECTED TROPICAL NON-TRADITIONAL EXPORTS: EVIDENCE FROM UGANDA S FLOWER EXPORTS ( ) Kihangire David Bank of Uganda Pos D. J. and Cameron S. (b) Bradford Cenre for Inernaional Developmen (BCID) Universiy of Bradford BRADFORD, U.K. Absrac: This sudy examines he effecs of exchange rae variabiliy on Uganda s flowers expors during by esing he cenral hypohesis ha following he floaing exchange rae regime, Uganda s expors of ropical flowers are negaively and significanly correlaed wih exchange rae variabiliy. The absence of pure I(0) or I(1) in he daa, and lack of endogeneiy and simulaneous bias problems invies us o apply ARDL approach o coinegraion and OLS. The resuls sugges ha alhough Uganda s flower expors are negaively correlaed wih exchange rae variabiliy, he measured effecs are insignifican. 54

56 Keywords: Exchange rae variabiliy, flowers expors, Uganda, dependeneconomy 55

57 I INTRODUCTION: EXCHANGE RATE VARIABILITY AND TRADE Following he breakdown of he Brion Woods sysem in February 1973, several counries have adoped a new sysem of flexible exchange rae regimes, inducing periodic volailiies in he exchange raes of he floaing currencies. Theoreically, he effec of exchange rae variabiliy on expors is sill a debaable issue. One srand of lieraure (Friedman, 1953; and Johnson, 1969) argues ha flexible exchange raes are beneficial o he economy as hey promoe rade and overall macroeconomic sabiliy and ha shor-run flucuaions in exchange raes have no effecs on rade volumes. On he oher hand, some economiss argue ha parly because of marke imperfecions paricularly in LDCs, hedging is boh imperfec and very cosly as a basis for avoiding exchange risk (e.g. Mundell, 2000; Doroodian, 1999; Krugman, 1989). Hence, expors may be negaively correlaed wih exchange rae variabiliy in line wih risk-aversion hypohesis. Differences in he heoreical lieraure sugges ha alernaive approaches o invesigaing he effecs of exchange rae variabiliy on expors remain largely an empirical issue. Due o he ambiguiies of is effecs on expors a he heoreical level, several sudies have aemped o empirically invesigae he exen o which he uncerainy induced by high exchange rae variabiliy migh h ave been derimenal o world rade (Abbo, e.al. 2001; Doyle, 2001; Arize, 1995, 1996, Savvides, 1992; De Grauwe, 1988; Hooper and Kohlhagen, 1978). One srand of sudies empirically reveals ha exchange rae variabiliy is no an imporan negaive facor for exporers (e.g. Abbo, e.al. 2001; Doyle, 2001). However, anoher group of sudies provide evidence, which suggess ha exchange rae variabiliy is an imporan negaive facor among exporers (Kihangire, Pos and Cameron, 2005; Kihangire, 2004; Sukar and Hassan, 2001; Sekka and Varoudakis, 2000, 2002; Saurer and Bohara, 2001; Hassan and Tufe, 1998; Doroodian, 1999; Arize, 1995, 1996; Gruba and Grennes, 1993; Grobar, 1993; Savvides, 1992; Bahmani-Oskooee, and Laifa, 1992; Bini-Smaghi, 1991; De Grauwe, 1988). In spie of he above sudies, anoher group of sudies reveals ha he effec of exchange rae variabiliy on expors is ambiguous: posiive or negaive (Du and Xu, 2001; Klaassen, 1999). Alhough mos sudies have focused on indusrial counries wih a high proporion of manufacured expors, here are hardly any sudies examining he impac of exchange rae variabiliy on expors of specific agriculural primary commodiy dependen economies such as hose in SSA (e.g. see Klaassen, 1999). I is Klaassen s (1999) viewpoin ha in order o ry o answer he ambiguiy ha surrounds he effecs of exchange rae variabiliy on expors, furher sudies should focus on developing counries wih high ime-varian exchange raes. Deparing from Klaassen s (1999) sandpoin, his paper aims a conribuing o he lieraure by empirically examining he long-run effecs of exchange rae variabiliy on Uganda s expors under he floaing exchange rae regime. In Uganda, he governmen adoped a floaing exchange rae regime in November 1993, parly o improve he incenives o he exporers by removing he implici axaion o exporers and relaed ren-seeking aciviies (Reinikka, and Collier, 2001). 56

58 Since hen, he exchange rae for he shilling o he Unied Saes Dollar (Shs/US$) has been flexible and a imes very volaile. However, here are mixed (largely speculaive) views regarding he effecs of exchange rae variabiliy on Uganda s predominanly agriculural primary commodiy expors. From Hensridge and Kasekende s (2001) viewpoin, i is argued ha alhough subsequen periods of [exchange rae] urbulence have aken place, he overall performance of he foreign exchange marke as a way o deermine he exchange rae and allocae foreign exchange has been good enough for i no o have been called ino quesion (p. 56). Deparing from Klaassen s (1999) sandpoin, his sudy adops he risk-aversion hypohesis o invesigae he effecs of exchange rae variabiliy on Uganda s expors of flowers under he floaing exchange rae regime. We es he cenral hypohesis ha Uganda s flowers expors are invarian o exchange rae variabiliy. II THEORETICAL FRAMEWORK AND HYPOTHESES Consisen wih Klaassen (1999), we adap an analyical framework encompassing wo conceps: he mean-variance-analysis (Tobin, 1959; De Grauwe, 1988) and he augmened imperfec subsiuion model (Goldsein and Khan, 1978; 1985). The simplified model comprises five equaions, given by:- X S = f([p X /P]*REER, CU, TOT, V) (3.1) f 1, f 2, >0; f 3 ±0; f 4 < 0 X D = g([p X /P X *]*REER, Y*, TOT, V) (3.2) g 1, g 4, <0; g 2, g 3, >0 X S = X D =X... (equilibrium) (3.3) X S = (X S X -1 ) 0 (disequilibrium) (3.4) P X = f(x D X) 0 (disequilibrium) (3.5) Where, S X = Uganda s real expor supply value index for flowers (based on US dollar earnings); P x / P = The raio of Uganda s flowers expor price index, Px (approximaed by Uganda s overall expor index) o domesic price levels, P (i.e. domesic expor compeiiveness vis-à-vis non-raded goods); * Px / PX = The raio of Uganda s flowers expor price index, Px, o world * P X expor price levels, (i.e. Uganda s exernal expor compeiiveness vis-à-vis oher expor subsiues from he res of he world); 57

59 P = Uganda s domesic price levels (proxy of prices of non-raded goods); P x = Uganda s (f.o.b.) flowers expor price index ; CU Y* = = Represens capaciy uilizaion for Uganda s indusrial secor; Represens world income for Uganda major rade parners; REE R = Uganda s real effecive exchange rae, (an inernal compeiiveness facor); TOT = Uganda s ne barer erms of rade; and V = A measure of exchange rae variabiliy. Where V is defined in he form: V 1/ [ 1 2 ( e ) ] 2 1 i 1 = m e Is a moving sandard deviaion of he growh rae of he exchange rae (based on a moving average se of observaions), Kenen and Rodrick (1986). We chose his measure because: i has well defined properies; i s simple o compue and communicae; capures high frequency ER movemens. Alhough long persisence of ER shocks is inconsisen wih observed shor-erm persisence, (IMF (1984), Bailey e al, (1987), Savvides ( 1992), Chowdhung (1993) and Abbo e al (2001)). And X D and X are world expor demand and equilibrium expor supply respecively; P X * is world agriculural expor price index, and Y* is world income of Uganda s major rading parners, and oher variables as defined above. Equaion (1) suggess ha Uganda s expors supply of flowers is a posiive funcion of relaive expor price, real levels of exchange rae, and capaciy uilizaion; and a negaive funcion of exchange rae variabiliy. However, Uganda s expors supply of flowers is ambiguously correlaed wih erms of rade in line wih he Prebisch-Singer hesis. Equaion (2) suggess ha he world demand for Uganda s flowers expors is a negaive funcion of relaive expor price, he levels of exchange rae, and exchange rae variabiliy; and a posiive funcion of world income and erms of rade. Equaion (3) suggess ha he flowers expor marke migh be in equilibrium; while equaions (4) and (5) sugges ha due o some srucural problems, i migh be in disequilibrium. There are several heoreical reasons explaining why expors migh be negaively correlaed wih exchange rae variabiliy (Abbo, 2001, Akhar and Hilon, 1984, IMF, 1984). Since mos expor conracs in Uganda are priced and paid for in foreign currency, exchange rae variabiliy affecs expor earnings valued in domesic currency The general expor price index is used as a proxy due o lack of Uganda s commodiy expor price daa on hese commodiies. 58

60 Expor conracs may involve long ime lags due o producion delays, delivery lags and he acual selemen dae, all of which may increase he exen of uncerainy. Imperfecions regarding hedging faciliies may make i difficul o fully anicipae and conain uncerainy caused by exchange rae variabiliy. The exen of expor produc diversificaion, and marke power deermines a firm s abiliy o suffer or expor he risk. I can be inferred from he above ha exchange rae variabiliy migh be a major source of uncerainy regarding prices exporers receive measured in domesic currency erms. If Ugandan flowers exporers are sufficienly risk-averse, such variabiliy could be derimenal o he expors growh as agens reduce heir expor volumes in line wih he risk-aversion hypohesis (Baum, Barkoulus and Caglayan, 2002; De Grauwe, 1988; Hooper and Kohlhagen, 1978; Ehier, 1973). The above suggess ha i is possible o empirically invesigae he impac of exchange rae variabiliy on Uganda s expors under hree possible frameworks: a) Single equaion srucural expor supply, model (equaion 1); b) Reduced-form equilibrium expor quaniy economeric model (equaions 1, 2, and 3); or c) Reduced-form disequilibrium expor quaniy economeric model (1, 2, 3, 4 and 5) above. Since he world demand for Uganda s expors of ropical flowers may be assumed exogenous in line wih a small counry-assumpion, our analysis focuses on he srucural expor supply funcion, mindful of he endogeneiy and simulaneous bias consideraions among he various deerminans in he model. Equaion (5) recognizes he fac ha markes are never perfec paricularly from an LDC viewpoin. III METHODOLOGY AND RESULTS 3.1 General-o-Specific esimaions Ardl-Ecm esimaions and Bounds-Tess Consisen wih Pesaran, Shin and Smih (2001), we follow he corresponding ARDL-ECM srucural real expor supply economeric model re-represenaion for Uganda s ropical flowers expors may be expressed as:- log X S l 1 m 1 = β0 + β1 i log( PX / P) i + β2 i logcui + β3 i logreer i + β4 i logtoti + β5 i V i +.. i= 1 i= 1 S + γ logx 1 + γ 1log( PX / P) 1 + γ2 logcu 1 + γ3 logreer 1 + γ4 logtot 1 + γ5v 1 + υ (3.6) On he basis simplified esimaed model (6), we carried ou bounds ess using he calculaed F-saisic and esed he null hypohesis H0 of no long-run relaionship vis-à-vis H A as: n 1 i= 1 p 1 i= 1 q 1 i= 1 H 0 : γ = γ1 = γ 2 = γ3 = γ 4 = γ5 = H A : γ γ1 γ 2 γ 3 γ 4 γ (3.7) (3.8) 59

61 Suppor for exisence of a long-run relaionship prevails if here is insufficien evidence o accep H 0 in (12) and hence he corresponding ARDL-ECM long-run parameer esimaes are derived from (11) above in an equaion of he form: - log S X = ϕ 1 log( PX / P) + ϕ 2 log CU + ϕ3 log REER + ϕ 4 log TOT + ϕ5 Where γ1 γ 2 γ 3 γ 4 γ 5 ϕ1 = ; ϕ2 = ; ϕ3 = ; ϕ4 = ; ϕ5 = γ γ γ γ γ V (3.9) (3.10) 3.2 Empirical resuls The source of daa for he sudy, and he consrucion of he various series is explained in appendix (1). 3.3 Uni Roo consideraions The analysis of daa for uni roos revealed ha here is lack of pure I(0) or pure I(1) for all he series in he model. This suggess ha ARDL-approach o coinegraion ( Pesaran, Shin and Smih, 2001) may be an appropriae mehod for analysis. In addiion, we examined he various aspecs relaing o auocorrelaion, exogeneiy, simulaneous bias, heeroskedasiciy, and mulicollineariy before underaking any diagnosic ess in order o obain inerpreaions based on reliable, consisen, and efficien esimaions 32. In general, he resuls suggesed ha:- There is lack of pure I(0) and I(1) for all he series. Likewise here was insufficien evidence of endogeneiy and simulaneous bias problems. Hence he sudy adoped ARDL approach o coinegraion and OLS esimaion mehods; Esimaions from he corresponding auocorrelaion funcions from he coinegraing equaions revealed some evidence of serial auocorrelaion. This is expeced, given ha we are dealing wih ime series daa, and he fac ha here are no lagged dependen variables; There was insufficien evidence of heeroskedasiciy in he iniial coinegraing esimaions; Based Hausman-specificaion ess for exogeneiy, here was insufficien evidence of endogeneiy among he various explanaory facors, paricularly he exchange rae variabiliy as he focus variable; Based on Hausman-specificaion ess for simulaneous bias, here was insufficien evidence of simulaneiy beween expor supply and expor demand via he relaive expor price variable. The resuls of he various variance inflaion facors, VIFs, revealed ha mulicollineariy was no a worrisome problem, paricularly wih regards o he focus variable i.e. exchange rae variabiliy: he calculaed VIFs for 32 Deails of he resuls and heir corresponding inerpreaions are available from auhor on reques 60

62 exchange rae variabiliy was less han 2, agian he rule of humb of 10. Hence, ARDL approach o coinegraion was adoped as an appropriae esimaion mehod for invesigaing he effecs of exchang e rae variabiliy on flowers expors. The resuls of he iniial ARDL-ECM esimaions sugges ha all he iniial saic flower expor supply models pa ss all he diagnos ic ess relain g o serial correlaion, funcional speci ficaion, as we ll as homoskedasiciy. On his basis, reducions of variables o he correspondin g parsimoniou s models (Table s 3. 1 below) were possible, saring wih hose variables fo und o be leas significan. 61

63 Table 3.1: Resuls of general o specific expor supply esimaions, 1994:M1-2001:M12 Monhly lags C * (1.8986) T * (3.2695) DLFLO W * (2.9634) DSQLFLOW * ( ) DLCUIID * * * * * (3.2999) ( ) ( ) ( ) ( ) DLTOTU * DLPXCPI ** ** (1.8841) ( ) (2.9615) DLNEERO * DLVNEERO * LFLOW * ( ) LTOTU(-1) ( ) LCUIID(-1) * (3.3999) LPXCPIH(1) (0.5953) LNEERO(-1) * ( ) LVNEERO(-1) (0.1707) (2.0278) (2.6028) R2-Bar DW Sa isics RSSR Serial Correlaion (0.704) Funcional form (0.596) Normaliy (0.419) Heeroskedasiciy (0.734) Source: Based on daa colleced from Uganda, May-July 2002 (files FLOW99.OUT;RSLTCF.xls) An F-es for he null hypohesis ha variable deleion is daa-permissible agains he alernaive hypohesis ha variable deleion is no daa permissible suggess ha each of he corresponding calculaed F-saisic of ; ; ; and 62

64 respecively was much lower han he abulaed F (38, 84) saisic of 1.59 a he 5% level of confidence 33. Hence, here was insufficien evidence o rejec he null hypohesis ha variable deleion is daa permissible. The diagnosic ess also suggesed ha he parsimonious empirical models in Table 3.2 pass all he diagnosic ess relaing o serial correlaion, funcional specificaion, normaliy, heeroskedasiciy, as well as predicion. 3.4 Resuls of he long-run ARDL-approach o coinegraion On he basis of he parsimonious model, he sudy applied bounds-es analysis for coinegraion. The calculaed F-saisics (Table 3.2) were much higher han he abulaed criical values of I(0)=2.649; I(1)=3.805 (inercep, wih no rend) a he 5% level of confidence (See Pesaran and Pesaran 1997: 478). This suggesed lack of sufficien evidence o accep he null hypohesis of a no long-run relaionships beween Uganda s srucural expors supplies of flowers and heir major deerminans. Similar analysis based on each of he nine differen marke condiions menioned above yielded resuls leading o similar conclusions. Table 3.2: Calculaed F-bound ess (criical. bounds value: I(0) = ; I(1) =3.805) F-bounds saisic for flowers (1) F Saisic F(7, 72)= [.000] >I(1) (2) F Saisic F(7, 72)= [.000] >I(1) (3) F Saisic F(7, 72)= [.000] >I(1) (4) F Saisic F(7, 69)= [.000] >I(1) (5) F Saisic F(7, 83)= [.022] >I(0) (6) F Saisic F(7, 81)= [.000] >I(1) (7) F Saisic F(7, 72)= [.000] >I(1) (8) F Saisic F(7, 72)= [.000] >I(1) (9) F Saisic F(7, 72)= [.000] >I(1) Source: Based on Resuls of F-bounds saisic ess derived from ARDL-ECM GTS esimaions On his basis, he sudy concluded ha he long-run relaionship beween Uganda ropical flower expors and is deerminans was possible on he basis of he ARDL approach o coinegraion. The resuls (Table 3.2) are he basis of discussions in he subsequen secion. 3.5 Inerpreing he resuls for Uganda s flowers expors supply The sudy has esimaed Uganda s srucural expor supply funcion of fresh ropical flowers o invesigae he research hypohesis ha Uganda s ropical flowers expors are negaively and significanly correlaed wih exchange rae variabiliy. In general, he sudy found insufficien evidence o accep he research hypohesis. Alhough Uganda s flowers expors were negaively correlaed wih exchange rae variabiliy in all he eigh perspecives considered, he measured effecs were insignifican even 33 For deails, resuls in elecronic files FLOW99.OUT flowers 63

65 a he 10% level of confidence. Consisen wih he research objecives, he following observaions can be made: - Does exchange rae variabiliy maer for Uganda s expors of ropical flowers? The empirical evidence suggess he affirmaive: Uganda s flowers expors were negaively correlaed wih exchange rae variabiliy in line wih risk aversion heory under he floaing exchange rae regime. Does nominal exchange rae variabiliy maer for Uganda s flowers expors? The observed negaive bu insignifican elasiciy suggess ha alhough nominal exchange rae variabiliy in he iner-bank foreign exchange marke migh maer for Uganda s flowers exporers, he overall effec is an insignifican facor. Does real exchange rae variabiliy maer for Ugandan ropical flowers expors? The observed negaive bu insignifican relaionship suggess ha i migh be a less imporan negaive facor among Ugandan fresh flower exporers. Does exchange rae variabiliy in differen foreign exchange markes (i.e. iner-bank vis-à-vis bureau) maer for Ugandan exporers? Alhough he empirical evidence suggess a negaive relaionship exiss in line wih risk-aversion in each foreign exchange marke, he overall effec is insignifican. Overall, he resuls of he sudy sugges ha alhough Uganda s expors of flowers are negaively correlaed wih nominal effecive exchange rae variabiliy in boh he inerbank and bureau foreign exchange markes, he measured effecs are insignifican. The average elasiciy of responsiveness was under he floaing exchange rae regime. In paricular, Uganda s expors of flowers were found o be negaively correlaed wih exchange rae variabiliy in all he nine scenarios considered, alhough none was significan even a he 10 percen level of confidence. These resuls are parly explained by he fac ha expors of fresh ropical flowers are highly perishable. Hence, agains condiions of high and unsable exchange rae variabiliy, farmers have no opion apar from exporing oherwise hey face wha Doyle (2001) has called weaher hy sorm or face he consequences of loss of income or exi from he marke. 3.6 Evaluaing he consisency of resuls wih respec o heory For each esimaed flowers srucural expor supply funcion, he sudy esimaed he corresponding marginal coefficien (long-run elasiciy) of Ugandan flowers expors wih respec o exchange rae variabiliy using ARDL approach o he coinegraion. In each case, he measured elasiciy was negaive, and consisen wih risk-aversion heory. For example, wih respec o he conrol variables, Uganda s flowers expor supplies were found o be consisenly and posiively correlaed wih relaive expor price in line wih comparaive advanage heory, alhough he measured effecs were insignifican. The elasiciy of flowers expors wih respec o he erms of rade facor as well as relaive prices were found o be highly posiive (8.4 and 6.5 respecively), which may be explained by he comparaive advanage heory. The coefficien on capaciy uilizaion was found o be highly posiive bu insignifican. 64

66 This is parly explained by ven for surplus in Uganda s flowers expor marke. Following rade liberalizaion and he adopion of he floaing exchange rae regime, Uganda s flower indusry has emerged as one of he mos imporan non-radiional exporer of flower commodiies. I is likely ha his secor has no ye reached is full producion possibiliy fronier. I is imporan o noe ha here is insufficien evidence of sudies invesigaing he effecs of exchange rae variabiliy on flowers expors following adopion of he floaing exchange rae regime. This makes i relaively difficul o discuss in deail he exen o which he resuls of his sudy encompass or suppor ohers. From a broader perspecive, he observed elasiciy for Uganda s flowers commodiy expors which, for example, overall averages was generally consisen wih oher sudies focusing on specific expor secors in line wih risk-aversion heory. For example, in a sudy involving 22 African Counries, Sekka and Varoudakis (2000) observed elasiciies, which ranged beween 0.47 and 0.51 for exiles; and 0.36 for chemical producs; and invarian for meal producs. However, he resuls on Uganda also compare wih oher sudies ha found an invarian effec of exchange rae variabiliy on expors in suppor of risk-neuraliy heory (Abbo, Darnell and Evans, 2001; Bailey, Tavlas, and Ulan, 1986; Gour, 1985). However, i is worh noing ha he negaive bu insignifican relaionship observed for Uganda s expors of flowers migh also be explained by he perishabiliy of fresh flowers expors: once ready, hey mus be harvesed and sold off immediaely since he lifespan of he produc is shor. 3.7 Evaluaing he differen resuls wih respec o heory and implicaions for differen sub-secor expors in Uganda I mus be recalled ha he main objecive of his sudy was o invesigae he effecs of exchange rae variabiliy on Uganda s fresh flower expors, following he adopion of he floaing exchange rae regime. To achieve his, he sudy adoped a concepual framework ha conrolled for he main facors (i.e. he conrol variables) heoreically considered o be affecing he supply of expors so ha one migh obain inernally valid resuls. Hence, alhough i was beyond he focus of he sudy, i is inuiive for one o ask: wha were he effecs on Uganda s expors of flowers, of he main conrol variables and wha are he implicaions of he differen resuls for he fresh ropical flower expor indusry in Uganda? The following sub-secion examines his issue focusing on he esimaed long-run elasiciies for: (a) relaive nominal and real exchange rae, expor prices relaive o domesic price levels, (c) erms of rade, and (d) capaciy uilizaion. 65

67 3.7.1 Evidence of Uganda s sub-secor expor supply response elasiciies wih respec o levels of Nominal and Real Effecive Exchange Rae In secion II, i was hypohesized ha he supply of expors is posiively correlaed wih he levels of he exchange rae in line wih devaluaion heory (elasiciies approach o he balance of paymens). The empirical evidence in his sudy provides suppor for he devaluaion heory hypohesis: Uganda s expor supply elasiciies for flowers wih respec o he levels of exchange rae were found o be posiively correlaed wih he levels of exchange rae (elasiciy of response was more han uniy), consisen wih he devaluaion heory, alhough he measured effecs were insignifican. These resuls sugges ha alhough Uganda s aggregae supplies of flowers expors are posiively correlaed wih he levels of exchange rae, his may be an insignifican facor among Ugandan exporers of flowers, or ha he flower indusry is very compeiive a he prevailing levels of exchange rae so ha exporers do no see i as a problem. This is parly explained by he fac ha high world demand condiions for fresh flowers, which resul in high world prices for flowers, overshadow he influence of exchange rae. Implicily, i also suggess ha Ugandan flower exporers are able o compee more favourably a he prevailing levels of exchange rae Evidence of Uganda s sub-secor expor supply response elasiciies wih respec o levels of f.o.b. expor prices relaive o domesic price levels In secion II above, we also reasoned ha absracing from marke disorions, Uganda s expor supplies of flowers should be posiively correlaed wih realized own-expor prices relaive o domesic price levels in line wih comparaive advanage and dependen-economy heories. The empirical evidence in his sudy suggess furher suppor for he dependen economy heory in his respec. The elasiciy of supply response of Uganda s expors of flowers wih respec o expor prices relaive o domesic price levels was very high (average of +6.11). The implicaions are ha policies ha mainain expor price compeiiveness, by lowering domesic inflaion migh benefi Ugandan producion and supply of flowers expors in line wih he dependen economy heory Evidence of Uganda s sub-secor expor supply response elasiciies wih respec o Terms of Trade In secion II above, we hypohesized ha Uganda s expors of fresh-flowers, being primary commodiy expors of a primary-commodiy dependen economy, are posiively (comparaive advanage heory) or negaively (Prebisch-Singer Hypohesis) correlaed wih he erms of rade. The empirical evidence suggess ha he posiive elasiciy of response observed for Uganda s flowers expors (average of 8.6) is consisen wih he comparaive advanage heory hypohesis under he floaing exchange rae regime. The implicaions are ha policies o liberalize rade and exchange rae have benefied he flower expor indusry in Uganda. 66

68 3.7.4 Evidence of Uganda s sub-secor expor supply response elasiciies wih respec o capaciy uilizaion I was also hypohesized in secion II above ha Uganda s fresh flowers expor supply is a posiive funcion of capaciy uilizaion. However, in a dependen economy framework, heory suggess ha his relaionship depends on wheher or no he economy is below is full producion fronier (hence a posiive relaionship) or a is full producion possibiliy fronier (hence a negaive relaionship). The resuls for Ugandan expors of flowers indicae a posiive bu insignifican elasiciy of response. In general, hese resuls sugges ha Uganda s fresh flower expor indusry migh be close o is full producion possibiliy fronier; alhough some slack capaciy migh exis. The implicaion of his is ha o increase expors and overall oupu furher, compeiive relaive expor price and exchange rae incenive policies mus be accompanied by oher non-marke policy incenives o increase subsanial invesmens in produciviy and capaciy in flower expor indusry so ha Uganda s producion possibiliy fronier can be shifed furher ouwards. In summary, i can be concluded from he above discussions ha having conrolled for he main facors affecing flower expors supplies, Uganda s expors of flowers are negaively, alhough insignificanly, correlaed wih exchange rae variabiliy. This is parly explained by he perishabiliy of he fresh flowers, which once ready mus be harvesed and sold immediaely. In addiion, he elasiciy of response of Uganda s fresh flowers commodiy expors o is various deerminans is also broadly consisen wih he heory Evaluaing he resuls wih respec o mehodological issues The second aspec of evaluaion of he resuls perains o mehodological issues relaing o: (i) encompassing and parsimonious; (ii) daa-coherence; (iii) daa admissibiliy, (iv) exogeneiy; and (v) parameer sabiliy. Firsly, wih respec o daa-coherency, here were no main specificaion errors relaing o he randomness of he residuals. The resuls of diagnosic ess relaing o flowers expor supplies sugges ha each of he corresponding parsimonious expor model on which he long-run esimaions are based passes all he necessary ess relaing o serial correlaion, model specificaion, normaliy, homoskedasiciy and predicive abiliy. Secondly, based on Hausman specificaion ess for exogeneiy, he sudy did no find sufficien evidence o accep he null hypohesis ha here is a conemporaneous relaionships beween he various regressors, he focus variable and he error erm for each of he flower expor supply funcions esimaed for Uganda. These resuls sugges ha he various regressors, including he exchange rae variabiliy as he focus variable, are ruly exogenous. Thirdly, on he basis of he Hausman specificaion ess for simulaneous bias, he sudy invesigaed he exen of simulaneiy beween Uganda s srucural expor supply and demand funcions for Ugandan expors of flowers via heir respecive relaive expor price facors. The resuls revealed ha here was insufficien evidence 67

69 o accep he null hypohesis ha here is a conemporaneous relaionship beween he various regressors and he error erm. On hese bases, i can be concluded ha he corresponding resuls based on ARDL approach o coinegraion and OLS relaed esimaion echnique are consisen and efficien. Fourhly, he sudy conside red he issue of d aa-admissibiliy on he basis of F-ess for variable deleion for he general-o-specific parsimonious models. The corresponding resuls of he F-ess for he various flower expor models esimaed reveale d ha he bounds-ess for variable deleions were daa-permissible. Finally, he sudy examin ed he parameer sabiliy of he various parsimonious expor supply models for Ugandan flowers expors based on recursive ess. The resuls suggesed ha he underlying parsimonious ARDL-ECM models passed he necessary sabiliy ess. In general, he differences beween he resul s of his sudy vis-à-vis oher sudies in indusrial counries may parly be aribued o lack of well developed hedging markes in Uganda following he adopion of he floaing exchange rae regime. The perishabiliy of Ugandan ropical fresh flowers expors also parly explains he differences. IV CONCL UDING R EMARKS AND POLICY IMPLICATIONS This sudy aimed a invesigaing he effecs of exchange rae variabiliy on Uganda s expors of flowers by es ing he hypohe sis ha Uganda s fresh ropical flower expors are negaively correlaed wih exchange rae variabiliy. Having conrolled for he main facors in he heoreical debae of primary commodiy dependen economies, he resuls suggesed ha Uganda s fresh flowers expors were negaively, alhough insignificanly, correlaed wih exchange rae variabiliy: To some exen, he above resuls suppor he risk-aversion hypohesis, which, in line wih Doroodian (1999), may parly be aribu able o lack of well-developed hedging faciliies and insiuions in Uganda s foreign exchange marke. They may also reflec he view ha Ugandan exporers of fresh flowers are risk-averse alhough he measured effecs are insignifican. In general, alhough relaive prices, erms of rade, capaciy uilizaion, a compeiive excha nge rae and oher facors are imporan for Ugandan flower expors growh, exchange rae variabiliy may be an insignifican facor in r elaion o oher policies aimed a supporing he growh of his sub-secor caegory of Ugandan expors. 68

70 APPENDIX Table 3.1 below provides a sum mary of he main variables and he sources of daa. Appendix 3.1: Descripion of key variables used in he specific agriculural commodiy expor models Variable level Ln of level Diff. of ln Descripion of variable Source GRXPTS LGRXPTS dlgrxpts Gross expors, mn (US$) BOU/URA REXPTDX LREXPTDX dlrexptdx Real Expor Expendiure Index (1990=100) Consruced VCOF Value of coffee (mn, US$) BOU/UCDA VTEADx Value of ea (mn, US$) BOU RVCOF LRVCOF dlrvcof Log real value of coffee index 1995=100 Consruced RVTEA LRVTEA dlrvtea Log real value of ea index 1995=100 Consruced VFISH Value of fish expor (mn, US$) BOU VFLOW Value of flower expors (mn, US$) BOU RVFISH LRVFISH dlrvfish Real fish expor index Sep 1995=100 Consruced RVFLOW LRVFLOW dlrvflow Real flower expor 1995=100 Consruced FOUS VARIABLE VNEERo LVNEERo dlvneero Variabikliy of NEER, inerbank marke 1993/94=100 Consruced VNEERp LVNEERp dlvneerp Variabikliy of NEER, bureau marke 1993/94=100 Consruced VNEERoh LVNEERoh dlvneeroh Variabikliy of NEER, inerbank marke, based on Consruced headline CPI 1993/94=100 VNEERou LVNEERou dlvneerou Variabikliy of NEER, inerbank marke, based on Consruced underlying CPI 1993/94=100 VNEERph LVNEERph dlvneerph Variabikliy of NEER, bureau marke based on Consruced headline CPI 1993/94=100 VNEERpu LVNEERpu dlvneerpu Variabikliy of NEER, bureau marke based on Consruced underlying CPI 1993/94=100 VNEERimf LVNEERimf dlvneerimf Variabikliy of NEER, Uganda (IMF series, Consruced 1995=100 CONTROL VARIABLE GRXPTS LGRXPTS dlgrxpts Gross expors, mn (US$) URA/BOU VNEERo LVNEERo dlvneero Nominal Effecive Exchange Rae, inerbank marke Consruced 1993/94=100 VNEERp LVNEERp dlvneerp Nominal Effecive Exchange Rae, bureau marke Consruced 1993/94=100 VNEERoh LVNEERoh dlvneeroh Real effecive Exchange Rae, inerbank marke, Consruced based on headline CPI 1993/94=100 VNEERou LVNEERou dlvneerou Real effecive Exchange Rae, inerbank marke, Consruced based on underlying CPI 1993/94=100 VNEERph LVNEERph dlvneerph Real effecive Exchange Rae, bureau marke based Consruced on headline CPI 1993/94=100 VNEERpu LVNEERpu dlvneerpu Real effecive Exchange Rae, bureau marke based Consruced on underlying CPI 1993/94=100 VNEERimf LVNEERimf dlvneerimf Real effecive Exchange Rae, Uganda (IMF series, IMF(IFS) 1995=100 PxINDXu LPxINDXu dlpxindxu Uganda expor prices 1995=100 BOU/URA ToTu LToTu dltotu Uganda erms of rade (price of expors as a raio of BOU price of impors 1995=100 IIPWORLD LIIPWORLD dliipworld Index of indusrial producion in indusrial counries, UBOS 1995=100 PxLDCs LPxLDCs dlpxldcs Commodiy Expor Price Index in M&LDCs, IMF(IFS) 1995=100 PxWORLD LPxWORLD dlpxworld Commodiy Expor Price Index in World, 1995=100 IMF(IFS) PxAGRIC LPxAGRIC dlpxagric Commodiy Expor Price Index Agriculure, IMF(IFS) 1995=100 IID LIID dliid Index of indusrial producion, 1995=100 UBOS CUIID LCUIID dlcuiid Economic Capaciy, quarerly moving average rend Consruced* index less acual index of indusrial producion PxCPIH LPxCPIH dlpxcpih Raio of Expor Price index o Headline Consumer Consruced Price Index, 1995=100 PxCPIU LPxCPIU dlpxcpiu Raio of Expor Price index o Headline Consumer Consruced Price Index, 1995=100 PxPxAGR LPxPxAGR dlpxpxagr Raio of Expor Price index o underlying Consumer Price Index, 1995=100 Consruced 69

71 PXCFPXAG LPXCFPXAG dlpxcfpxag Raio of Expor Price index o world Agric. Raw Consruced Maerials expor price index, 1995=100 PXTEPXAG LPXTEPXAG dlpxtepxag Log price of coffee relaive o world agric expor Consruced price, 1995=100 PCOFCPU LPCOFCPU dlpcofcpu Log price of ea relaive o world agric expor price, Consruced 1995=100 PTECPU LPTECPU dlptecpu Difference of log of relaive coffee expor o Consruced underlying consumer price index, 1995=100 PCOFCPH LPCOFCPH dlpcofcph Difference of log of relaive ea expor o underlying Consruced consumer price index, 1995=100 PTEACPH LPTEACPH dlpteacph Difference of log of relaive coffee expor o headline consumer price index, 1995=100 Consruced Source: Series consruced are based on Daa colleced from Uganda during May-June

72 References Abbo, A Exchange rae variabiliy and rade. Lawler K and Seddighi H. (2001) (eds). Inernaional Economics: Theories, hemes and debaes. Essex. Pearson Educaion Ld. Abbo, A., Darnell, A.C. and Evans, L The influence of exchange rae variabiliy on U.K. expors. Applied Economic Leers. 8, pp Akhar, M.A. and Spence-Hilon, R The effecs of exchange rae uncerainy on German and U.S. rade. Quarerly Review New York. Federal Reserve Bank of New York. Spring, pp.7-16 Arize A.C Trade flows and real exchange-rae volailiy: an applicaion of coinegraion and error-correcion modelling. Norh American Journal of Economics and Finance, 6 (1), pp.37-51(15) Elsevier Science. Arize A.C Real exchange-rae volailiy and rade flows: The experience of eigh European economies. Inernaional Review of Economics and Finance, 5(2) pp (19), Elsevier Science. Bahmani-O, and Laifa Effecs of exchange rae risks on expors: Cross counry analysis. World Developmen, 20(8), pp Bailey, M J., Tavlas, G.S, and Ulan, M Exchange rae variabiliy and rade performance: Evidence for he big seven indusrial counries. Welwirschafliches Archiv. 122(3), pp Barkoulas J.T.; Baum C.F.and Caglayan M Exchange rae effecs on he volume and variabiliy of rade flows. Journal of Inernaional Money and Finance, 21(4), pp (16). Elsevier Science Bini-Smaghi, L Exchange rae variabiliy and rade: Why is i so difficul o find any relaionship? Applied Economics, 23, pp (Also an bridged version of his Ph.D. Thesis empirical chaper, Univesiy of Chicago, De Grauwe P Exchange rae variabiliy and he slowdown in growh of inernaional rade. IMF Saff Papers. 35 (1988), pp Doroodian K Does exchange rae volailiy deer inernaional rade in developing counries? Journal of Asian Economics, 10(3), pp (10), Elsevier Science. Doyle E Exchange rae volailiy and Irish-UK rade, Applied Economics, 33, (2), pp Rouledge, par of he Taylor & Francis Group. 71

73 Du H. and Zhu Z The effec of exchange-rae risk on expors: Some addiional empirical evidence. Journal of Economic Sudies. 28 (2), pp (16), MCB Universiy Press Ehier, W Inernaional Trade and he forward exchange marke. American Economic Review. 63(3), pp Friedman, M The case for flexible exchange raes. Essays in posiive economics. Chicago. Universiy of Chicago Pres, p Ghura, D. and Grennes T.J The real exchange rae and macroeconomic performance in Sub-Saharan Africa. Journal of Developmen Economics, 42(1993), pp Goldsein M. and Khan, M.S The supply and demand for expors: A simulaneous approach. Review of Economics and Saisics. 60 (1978), pp Goldsein M. and Khan, M.S The specificaion of goods and facor markes in open economy macroeconomic models. In Jones, R.W. and Kenen, P (1985) (eds). Handbook of Inernaional Economics. II. (Inernaional Moneary Economics and Finance). Amserdam. Elsevier Science Publishers B.V. Ch. 20, p Grobar L.M The effec of real exchange rae uncerainiy on LDC Manufacured Expors. Journal of Developmen Economics. 41(1993), pp Hensridge, M., and Kasekende, L Exchange rae reforms, sabilisaion, and Fiscal Managemen. In Reinikka, R. And Collier, P. (eds). Uganda s Recover: The role of farms, firms, and governmne. Washingon D.C. The World Bank. Ch.3, p Hooper, P. and Kohlhagen, S.W The effec of exchange rae uncerainy on he prices and volume of inernaional rade, Journal of Inernaional Economics. 8(1978), pp Inernaional Moneary Fund. (IMF) Exchange rae volailiy and World rade. A sudy by he Research Deparmen of he Inernaional Moneary Fund. Occasional Paper (28). Washingon D.C. IMF. Johnson, G.H The case for flexible exchange raes, Federal Reserve Bank of S. Louis. Monhly Review, June (b), pp Kenen, P. and Rodrick, D Measuring and analysing he effecs of shor-erm volailiy on Real Exchange Raes. Review of Economics and Saisics, pp Kihangire D.A The impac of exchange rae variabiliy on Uganda s expors ( ). Ph.D. Thesis. Universiy of Bradford. Kihangire D., Pos D J, and Cameron S Has exchange rae variabiliy affeced Uganda s coffee expors ( )? Discussion Paper 5, Universiy of Bradford 72

74 Klaassen, F Why is i so difficul o find an effec of exchange rae risk on rade? Economeric Sociey World Congress Universiy of Amserdam. Unpublished Paper No downloaded from he REPEC websie hp:// fmwww.bc.edu/repec/es2000/0133.pdf (forhcoming in Journal of Inernaional Money and Finance, July 2002) Krugman, P Exchange rae insabiliy. The Lionel Robbins Lecures. Cambridge. The Massachuses Insiue of Technology (MIT) Press Mundell, R. A Currency areas, exchange rae sysems and inernaional moneary reform, Journal of Applied Economics. III, pp Pesaran, M.H. and Pesaran B Working wih Microfi 4.0: Ineracive Economeric Analysis. Oxford. Oxford Universiy Press. Pesaran, M.H., Shin, Y. and Smih, R Bounds esing approaches o he analysis of level relaionships. Journal of Applied Economerics. 16, pp Reinikka, R. and Collier, P (eds). Uganda s recovery: The role of farms, firms and he governmen. Washingon. The World Bank. Sauer C. and Bohara A.K Exchange rae volailiy and expors: Regional differences beween developing and indusrialized counries. Review of Inernaional Economics, 9(1), pp (20) Blackwell Publishers Ld., Oxford, UK and Boson, USA. Savvides A Unanicipaed exchange rae variabiliy and he growh of inernaional rade. Welwirchafliches Archiv. 128(3), pp Sekka K. and Varoudakis A Exchange rae managemen and manufacured expors in Sub-Saharan Africa. Journal of Developmen Economics, 61(1), pp (17), Elsevier Science. Sekka K. and Varoudakis A The impac of rade and exchange-rae policy reforms on Norh African Manufacured Expors. Developmen Policy Review, 20, (2), pp , Blackwell Publishers Ld., Oxford UK and Boson, USA Suka A.H. and Hassan S US expors and ime varying volailiy of real exchange rae. Global Finance Journal, 12(1), pp (11) Elvier Science. Tobin, J. (1959). Liquidiy preference as behaviour owards risk. Review of Economics and Saisics. 25(4), pp

75 EXCHANGE RATE MISALIGNMENT AND FINANCIAL LIBERALISATION: EMPIRICAL EVIDENCE AND MACROECONOMIC IMPLICATIONS FOR UGANDA, David Kihangire 34 The Bank of Uganda Saff Papers Journal Vol. 1, No Bank of Uganda Absrac: This sudy empirically invesigaes Uganda s equilibrium real exchange rae (EREER) during 1993M1 o 2004M12. Using ARDL approach o coinegraion, we find ha a long-run relaionship exiss beween Uganda s REER and is deerminans, driven largely by rade balance, openness, fiscal deficis, and capaciy uilizaion. Consequenly, we esimaed a longrun EREER model. Comparing he acual REER and he EREER reveals ha Uganda s REER is overvalued over he recen period, The macroeconomic financial coss of his overvaluaion are found o be quie high for he Ugandan economy. Policy-wise, he resuls sugges ha appropriae policy mus aim a avoiding exchange rae overvaluaion in suppor of exernal macroeconomic sabiliy. Key Words: Exchange Rae Misalignmen, Expors, Uganda, equilibrium exchange rae, Overvaluaion. 34 Auhors acknowledge useful commens from members of he BOU MCPC, Ms Davis Fiona of MFPED, Charles Abuka and oher saff of BOU. All errors and omissions remain my own. The views expressed in his paper are my own and do no represen he official posiion of he BOU. 74

76 I INTRODUCTION In he lieraure, he concep overvaluaion of he exchange rae refers o condiions when he exchange rae is below is long-run equilibrium levels based on he fundamenal facors driving i (e.g. see Richaud, Varoudakis, and Veganzones, 2000). I is widely perceived ha one of he consequences of exchange rae overvaluaion (or rying o defend an overvalued exchange rae) is he frequen cause of balance of paymens crises. To overcome his, governmens usually pursue polices ha avoid overvaluaion. Since November 1993, Uganda adoped a floaing exchange rae regime aimed a supporing expor-led growh and povery reducion sraegy. Consequenly, afer an iniial appreciaion, he shilling depreciaed seadily (boh in nominal and real effecive erms) o January 2003 and helped o suppor expor secor growh. However, he exchange rae has since sared he appreciaion rends from mid-2003 and here are risks ha he underlying appreciaion and overvaluaion migh adversely affec fuure expor growh, and undermine macroeconomic sabiliy. Alhough he global decline in he value of he US$ agains major currencies migh have conribued o he appreciaion of he shilling, he recen appreciaion pressures since mid 2003 seems o be aggravaed by BOU s curren serilizaion policy, which aims a selling foreign exchange whenever opporuniies arise 35. The raionale for BOU s curren serilizaion policy is currenly undersood o mean assising BOU in is effors o mop up liquidiy. In his framework, i is difficul for one o exonerae BOU s acions o be adding o he appreciaion pressures, beyond ha caused by he weakening of he US$, and donor inflows. Alhough here is no apparen misalignmen beween he inerbank and bureau raes, a recen sudy (BOU and Governmen, 2004) of Uganda s exchange rae and is effec on expors performance suggess ha alhough Ugandan expors are hypohesized o break-even a Shs1755/US$ (Dec, 2004), he shilling has coninued o perform below ha level. This siuaion is no helped by he fac ha Uganda s recen headline inflaion rae of 8.4percen (January, 2005) is higher han ha of her weighed major rade parners, by a difference of abou 5-6 percenage poins, which suggess furher ha Uganda s nominal exchange rae should be depreciaing a approximaely ha differenial rae per year. The consequences of he recen exchange rae appreciaion on he Ugandan economy canno be underesimaed. Already, he Uganda Exporers Associaion have sough audience wih BOU o raise heir concerns abou he appreciaing shilling. This has invied he Minisry of Finance and Research Dep of BOU o iniiae a join sudy o invesigae he compeiiveness of he expor secor. Likewise, Uganda has also agreed o offer subsidies o he coon farmers in he Eas. A more recen adverse developmen perains o obacco producion and procuremen in Wesern Uganda, whereby BAT (U) Ld has declined from any more purchases of exra (and purpored surplus ) obacco supplies in ha region. BAT s argumen ha obacco producers mus be punished for overproducing appears o be counerproducive o 35 I is however, worh noing ha he shilling is no overvalued in he radiional sense where he BOU sells forex o defend a fixed level. Raher, because of he overvaluaion of he shilling, BOU is no able o coninue selling FOREX given he problems facing he expor secor. 75

77 governmen s sraegy of expor-led growh and povery eradicaion. Among oher raded goods, here are also complains from he sugar indusry in Uganda ha i is facing unfair compeiion from cheap impored Brazilian sugar 36. Alhough some expor secors such as flowers and ourism migh no be as vulnerable as ea, coon, obacco, or maize, he general early-warning signals sugges a need o invesigae he problem of expors and exchange rae overvaluaion furher. Agains his bac kground, i is inuiive for one o argue ha he appreciaion of he exchange rae hreaens o reard Uganda s fuure economic growh raes and overall macroeconomic sabiliy. This raises a fundamenal quesion: Is Uganda s exchange rae overvalued, and wha is he likely effec of his on he economy? This paper examines his quesion from four main perspecives. Firs, i highlighs he recen rends in Uganda s NER, NEER, and REER vis-à-vis expors and rade balance, in which i is shown ha he exchange rae has assumed is appreciaion rend in recen imes. Secondly, i highlighs he empirical evidence regarding he equilibrium REER in he conex of Edwards (1989) heoreical model. The resuls show ha Uganda s real exchange rae has been overvalued, paricularly from mid 2003 onwards. Thirdly, i examines he consequences of mainaining an overvalued NEER and REER on he budge, he expors, and moneary managemen. The resuls show ha i is oo cosly for Uganda o rely on forex serilizaion ha exacerbaes appreciaion and overvaluaion pressures, paricularly a a ime when governmen is a ne saver of foreign exchange and he expor secor is on a srong growh rend. On he conrary, increased use of domesic OMO and REPO insrumens appear o be more appealing for managing liquidiy han forex serilizaion. Hence, more weigh should be aached o domesic insrumens as ools of domesic liquidiy managemen as compared o serilizaion using forex. II RECENT TRENDS IN LEADING INDICATORS AND EVIDENCE FROM THE LITERATURE In he lieraure, several facors are known o influence a counry s real exchange. Key among hese include erms of rade, produciviy, governmen defici, rade balance, openness of he economy, and moneary shocks (Edwards, 1989). In he lieraure, four main sudies (Musinguzi and Muambi, 2002; Aingi-Ego and Sebudde (2000) Adam and Bevan (2002: cied in MacDonald, 2004); and McDonald (2004)) have aemped o invesigae he exen of exchange rae overvaluaion in Uganda. The sudy by Musinguzi and Muambi (2002) used monhly daa spanning 1993M8 o 2001M5. The main explanaory variables were erms of rade, produciviy, governmen deb, governmen consumpion of non-radables, and capial inflows. Alhough he auhors do no discuss he episodes of overvaluaion/under-valuaion, he resuls of heir sudy suggesed ha he exchange rae was overvalued during 2000, and slighly undervalued in early However, heir sudy does no explain he exen of overvaluaion and under-valuaion during 36 The complains from he domesic sugar indusry come a a ime when Governmen argues ha ariff barriers on impors already proec he secor. One migh argue ha such proecion is no pro-poor as domesic producers are forced o buy more expensive sugar and ha alhough proecion migh creae jobs in he sugar indusry, he benefis of hese jobs o he economy is offse by welfare losses o all oher consumers. 76

78 he recen period of , as heir sudy covers he period only up o 2001M5. The sudy by Aingi-Ego and Sebudde, (2000) examined Uganda s equilibrium REER based on annual daa ( ). The main deerminans were openness, erms of rade, Gov. expendiure and Capial flows. The resuling relaionship was obained in he form:- 1 reer = lopen lge lflow lo (4.1) Over he recen period ( ), he sudy by Aingi-Ego and Sebudde (2000) produced resuls, which suggesed ha REER was close o is equilibrium levels. However, alhough he coefficiens were consisen wih economic heory, he auhors did no repor he corresponding -saisics, which makes i difficul for one o ascerain he suiabiliy of he resuls of heir sudy. Secondly, he period covered spanned several economic policy regimes, he significan ones being varying exchange rage regimes, as well as conrolled vas liberalized rade regimes. Based on boh accouns, i is difficul for one o ascerain he reliabiliy of he resuls. Furhermore, since heir sudy was confined o he period 1970 o 1998, i does no help one o explain wheher or no Uganda s exchange rae is overvalued in he recen period ( ) In Adam and Bevan (2002), an aemp was made o invesigae Uganda s equilibrium real exchange rae (REER) using quarerly daa from 1993Q3 o 2003Q3. The auhors applied OLS esimaion mehods. The main deerminans of he REER in heir model were log of erms of rade (o), log of governmen expendiure as raio of GDP and log of ne capial flows. The auhors focused on he inernal REER as he main dependen variable, and obained resuls in he form:- REER = l o 0.33 lge 0.17 l flows (4.2) (1.77) (1.83) (3.93) Alhough he resuls appeared o be in line wih economic heory, he auhors did no proceed o repor he resuls of misalignmen. In general, boh sudies (Aingi-Ego and Ssebudde, 2000; and Adam and Bevan, 2003) do no help one o answer he quesion of wheher or no Uganda s exchange rae is overvalued over he recen period. Agains his background, his sudy re-examines he quesion of exchange rae overvaluaion in Uganda s recen imes. The presen sudy invokes he heoreical model suggesed by Edwards (1989) which is hen adaped o ake ino accoun boh fundamenal facors over he medium and long-run (e.g. erms of rade, produciviy, invesmen) as well as shor-run macroeconomic variables (such as moneary shocks, fiscal deficis/gdp raios, inflaion differenials, and exchange rae devaluaions). III THE MODEL, HYPOTHESIS AND ESTIMATION METHODS 3.1 The model and hypohesis Consisen wih he heoreical perspecives suggesed in Edwards, (1989) and Moniel, (1999), we adop an augmened analyical framework of exchange rae deerminaion for developing counries since i is considered o encompass he main 77

79 heoreical economic issues perinen o foreign exchange markes perinen o economically liberalizing LDCs such as Uganda. For example, we adap he Edwards (1989) model by aking ino accoun he effecs of inflaion differenial beween Uganda and her major rading parners. Secondly, we inroduce he erms of rade facor o ake ino accoun he small-open counry assumpions. This is because Uganda lacks he necessary marke power o influence erms of rade and oher prices on he world marke. Thirdly, we inroduce he capaciy uilizaion facor o ake ino accoun oher coss and srucural supply issues likely o affec he exchange rae (e.g. see supply of expors (Gour, 1985; IMF, 1984). Fourhly, we adap he public expendiure pressures ino he model o reflec he fac ha he influence of fiscal operaions may have a major effec on exchange rae. The composiion of fiscal expendiures beween raded and non-raded goods has been heorized o affec he exchange rae in Uganda (See Adam, 2004). Finally, we inroduced rade openness facor o capure he he influence rade pressures on he exchange. In line wih he heoreical perspecives suggesed in Edwards (1989), Uganda s real effecive exchange rae (REER) can be posulaed as 37 : - log REER= α + α logfisc+ α logtot+ α logcu+ α loglopen + α loglbasm+ αinfdif + αkapflo+µ' W Where, FISC (4.3) α α, α > 0, α > 0; α < 0; α < 0; α < 0; Μ α 0 0, < Is Uganda s public expendiure pressures/gdp, measured as fiscal defici o GDP raio; TOT Is Uganda exernal erms of rade index measured on he basis of radeweighed prices of expors in relaion o he rade-weighed price of impors? CU Is capaciy uilizaion, approximaed by he index of indusrial as a raio is rended oupu levels. LOPEN Is rade openness approximaed by he absolue sum of impors and expors as a raio of GDP. LBASM Is he moneary shock, measured as he raio of base money relaive o is rended levels. INFDIF Is he inflaion differenial, represened by Uganda s inflaion rae less he rae of inflaion of is major rade parners. CAPFLOW Is he capial flow, represened by Uganda s curren accoun balance as a raio of GDP (based on he principal ha developmens in he curren accoun mus be financed by developmens in he capial accoun). 3.2 The ARDL approach o coinegraion mehods of analysis On he basis of he above heoreical relaionships, we follow analyical mehods adoped in oher sudies (Kihangire e.al., 2005; Kihangire, 2004; Abbo, Darnell and 37 The heoreical basis of he Edward s model is generally consisen wih opimal currency heory framework. 78

80 Evans, 2001) o invesigae he deerminans of EREER based on ARDL-ECM approach o coinegraion. Conrary o oher mehods of analysis, we consider he ARDL approach o coinegraion appealing because i exends beyond he sandard error correcion mehods (which call for sric I(0) or I(1) series, by accommodaing oher series h a migh be fracionally inegraed (e.g. see Pesaran and Pesaran, 1997; and Pesaran, Shin and Smih, 2001). We esim ae he EREER based on he model (3) in he form given by:- log REER= β + l 1 m 1 n 1 p 1 q 1 2 log 3 l 4 5 log q 1 + α1 log + 6 g 0 i FISCi + α i TOTi α i ogcui + α i loglopen i + α i LBASMi α i lo INFDIF i.. i= 1 i= 1 i=1 i= 1 i= 1 i= 1 l 1 + α log CAPFLOWi + φ log REER 1 + log FISC + φ2 logtot 1 + φ3 log C + 4 lo N + φ log 7i φ 1 1 U 1 φ g LOPE 1 5 LBASM 1 i = φ 6 log INFDIF 1 + φ 6 log INFDIF 1 + φ 7 log CAPFLOW 1 + υ ( 4.4) On he basis of he esimaed parsimonious model in (4) above, we hen carried ou bounds ess using he calculaed F-saisic and esed he null hypohe sis H 0 of no lo ng-run relaionship vis-à-v is H A as: H φ = φ = φ = φ = φ = φ = φ = φ 0 (4.5) 0 A : = : φ φ1 φ2 φ3 φ4 φ5 φ6 φ7 0 H ( 4.6) Consisen wi h Pesar an, Shin and Smih, (2001), suppor for he ex isence o f a long- run relaionship prevails if here is insufficien evidence o accep H 0 in (5) and hence he corresponding ARD L-ECM long-run parameer esim aes are derived from (4) above in an equaion of he form:- log REER = ψ 1 log FISC + ψ 2 logtot + ψ 3 log CU + ψ 4 log LOPEN + ψ 5 log LBASM + ψ 6 log INFDIF + ψ 7 log CAPFLOW (4.7) Where; φ1 φ φ3 φ4 φ5 φ6 φ7 ψ 1 = ; ψ 2 = ; ψ3 = ; ψ4 = ; ψ5 = ; ψ6 = = ψ7 = φ φ φ φ φ φ φ (4.8) 3.3 Resuls and discussions T he daa for he various variables was obained from BOU s BOP, UBOS, and Minisry of Finance. Based on monhly daa over he period 1993 M1 o 2004M 4. The resuls of he general-o-specific AR DL-ECM (Table 4.2) revealed ha a long- run relaionship exis s b eween he exchange rae and is deerminans, driven largely by moneary shocks, inflaion differenials, capaciy uilizaion, openness, an d capial flows. A he 5% level, he calculaed F-saisics for bounds es wa s 4.18, compared o he abulaed F-saisic of I(0)=2.4 and I(1) =

81 Table 4.1: Resuls of general o specific ARDL-ECM EREER Esimaion, 1994M1-2004M12 Monhly lags C * (2.2749) DlogREER * * DlogFISC * * (2.1998) (1.6441) (2.5775) ( ) (2.2586) Dlog TOT Dlog CU * ( ) (2.3480) ( ) (2.2740) DlogOPEN * * ** ( ) (2.9402) (1.4668) (2.6444) (2.4345) Dlo glbasm ( ) ( ) ( ) ( ) DLINFDIF * * ** ** (7.2842) ( ) ( ) ( ) (2.3776) DLKAFLOW * DlogREER ( ) DlogFISC ( ) DlogTOT (1.2908) Dlog CU * ( ) DlogOPEN (3.5200) DlogLBASM (1.3678) DLINFDIF ( ) ( ) DLKAFLOW ** R2-Bar ( ) DW-Sa RSSR Serial Correlaion (0.118) Funcional Form (0.596) Normaliy (0.896) Heeroskadisiciy (0.880) Source: Auhor 80

82 The resuls of he ARDL-ECM model based on Akaike informaion crieria sugges ha:- log REER = log FISC logtot log CU log LOPEN ( ) ( ) ( ) ( ) ( ) loglbasm INFDIF CAPFLOW ecm ( ) ( ) ( ) ( ) (4.9) Whereby ECM is given by ecm= log REER * C logFISC 01244logTOT logCU logLOPEN log LBASM INFDIF CAPFLOW The resuls of he corresponding long-run ARDL relaionship reveal ha:- (4.10) log REER= * C logFISC logTOT logCU logLOPEN+ ( ) (1.3204) ( ) (0.5272) ( ) *logLBASM * INFDIF *logCAPFLOW * logreer ( ) ( ) ( ) ( ) R 2 = ; (4.11) (- values in brackes) Serial correlaion X2(12) =13.25(0.351); Funcional-form F(1, 90) = 0.360(0.550); Normaliy X2(1) =45.61(0.000) Heeroskedasiciy X2(1) = (0.827) The exen of exchange rae under-valuaion or overvaluaion can be examined by relaing he projeced values agains he acual values. The resuls are as shown in figures (4.3) and (4.4) below, and sugges ha during , Uganda s REER has largely been overvalued. 81

83 Figure 4.1: Acual and Equilibrium REER in Uganda, ( ) ACTUAL AND EQUILIBRIUM REER IN UGANDA, ( ) R REER,EREE Jul-94 Jul-95 Jul-96 Jul-97 Jul-98 Jul-99 Jul-00 Jul-01 Jul-02 Jul-03 Jul-04 MONTH/YEAR REER EREER Source: Auhors 82

84 Figure 4.2: Difference beween Acual and EREER (% of EREER) DIFFERENCE BETWEEN ACTUAL AND EREER (% OF EREER) DIFFERENCE AS % OF EREER Jul Jul-95 Jul-96 Jul-97 Jul-98 Jul-99 Jul-00 Jul-01 Jul-02 Jul-03 Jul MONTH/YEAR DIFFERENCE Source: Auhors IV THE MACROECONOMIC IMPACT OF EXCHANGE RATE OVERVALUATION ON UGANDA S ECONOMY The macroeconomic impac of exchange rae appreciaion and overvaluaion can be examined under five main perspecives as shown in Table 4.2 below. The analysis akes ino accoun changes in he exchange rae, using June 2003 as he base year 38. The following conclusions can be reached:- Tab le 4. 2: Impac of exchange rae appreciaion & overvaluaion Source: BOU s cashflow, he BOP, and governmen s 2003/04 and 2004/05 budge cashflow. 38 Since he bulk of he expor secor uses a very small proporion of impored inpus (e.g. fish, coffee, gold, ea, flowers), he losses are esimaed on he basis of expor earnings muliplied by he difference beween exchange rae a June 2003 and curren period exchange rae. For BOU, since he Bank has minimum foreign liabiliies, he esimaed loss is simply he sock of foreign reserve holdings (ne of IMF liabiliies) earnings muliplied by he difference beween exchange rae a June 2003 and curren period exchange rae. 82

85 Exchange rae overvaluaion is very cosly o he overall economy. In 2004/05, oal losses (BOU, Gov, and Exporers) are esimaed a Shs billion, up from he losses of Shs billion incurred in he previous fiscal year. Exchange rae overvaluaion is cosly o governmen, paricularly when i is a ne saver of foreign exchange. For example, in 2004/05 governmen is esimaed o incur implici losses amouning Shs bn had he exchange rae no o appreciae a leas on nominal erms. There is a general view ha if BOU sold more forex o serilize liquidiy, hen domesic ineres coss would have been lower. Summing he presumed lower ineres coss already incurred plus he implici losses from exchange rae appreciaion suggess ha oal coss o governmen increased from Shs bn in 2003/04 o Shs bn in 2004/05. Governmen s case ha ineres coss o governmen of Shs bn incurred in 2004/05 is a leas lower han Shs bn incurred in 2003/04 misses he poin ha he oal coss o he budge are much higher in an environmen of an appreciaing exchange rae when he governmen is a ne saver of foreign exchange. BOU has indirecly suffered from an appreciaion and overvaluaion of he shilling. The implici oal coss are esimaed o amoun o Shs bn, up from Shs bn in 2003/04. In general, using foreign exchange serilizaion as a ool of liquidiy managemen a a ime when he exchange rae is appreciaing and he main agens are ne savers of foreign exchange is more cosly o he economy han using domesic OMO operaions. Heavy reliance on foreign exchange sales for liquidiy serilizaion and less use of domesic insrumens is very cosly o BOU, governmen and exporers. Given he coss of exchange rae overvaluaion o he economy, BOU s appropriae policy mix should be one ha avoids nominal and real exchange rae overvaluaion is suppor of exernal macroeconomic sabiliy, while domesic policy aims a ensuring a low and sable inflaion is suppor of inernal macroeconomic sabiliy. Managemen of domesic liquidiy should be suppored largely hrough he use of domesic insrumens, and o a lesser exen serilizaion of liquidiy using foreign exchange. Alhough i is anicipaed ha his will cause a moderae rise in nominal and real ineres raes, his will be beneficial as i will miigae some of he MacKnown-Shaw hypohesis effecs of financial repression, by supporing more savings which may in urn suppor more domesic invesmens. I is inuiive ha higher domesic savings would be consisen wih governmen s effors o reduce donor dependence. 83

86 V CONCLUDING REMARKS AND POLICY IMPLICATIONS This sudy has empirically esimaed Uganda s equilibrium real exchange rae (EREER) during 1993M1 o 2004M12 using ARDL approach o coinegraion. Overall, we find ha a long-run relaionship exiss beween Uganda s REER and is deerminans, driven largely by rade balance, openness, fiscal deficis, and capaciy uilizaion. Armed wih hese resuls, we hen esimaed a long-run relaionship of he EREER model. On he basis he resuls obained, we hen compared he acual REER and he EREER and obained resuls from which we concluded ha Uganda s REER is overvalued over he recen period, The sudy hen examined he macroeconomic financial coss of his overvaluaion. We found coss o be quie high for he Ugandan economy. Policy-wise, i can be concluded ha when an economy is a ne saver of foreign exchange, hen policy acions of governmen and he cenral bank mus aim a avoiding exchange rae overvaluaion in suppor of exernal macroeconomic sabiliy. On he basis of he analysis above, one can conclude ha BOU s use of foreign exchange sales as policy insrumen for liquidiy managemen be revisied in he conex of he exchange rae and is effecs on exernal macroeconomic sabiliy objecive. In general, BOU s policy effor on liquidiy managemen aimed a keeping inflaion low need o be largely suppored by issuance of domesic insrumens, and oher non-marke relaed insrumens, raher han heavy reliance on forex sales since his risks causing exchange rae misalignmen. Susained overvaluaion of he exchange rae risks one o reign in a Duch Disease unless acion is aken o reverse he exchange rae rends. In he case of Uganda, i is difficul o susain an overvalued exchange rae on he basis of aid inflows, unless his is suppored by a susained high growh and produciviy in he expor secor, augmened by large capial inflows. Policywise, allowing boh he nominal and real exchange raes o depreciae moderaely may provide incenives for he economy o build reserves and miigae some of he losses in Uganda s expors secor. The consequences of his migh be liquidiy injecions, which can be parly absorbed hrough issuance of domesic insrumens, and/or increased governmen savings wih he banking secor. The view ha real domesic ineres raes migh increase, hereby risking crowding-ou he privae secor remains an area for fuure research. One could es he proposiion ha ineres raes go up as governmen coninues an expansionary fiscal policy sance, while he cenral bank remains commied o is low-inflaion objecive. I would also be a useful, fuure research o invesigae he consequences of exchange rae misalignmen from he viewpoin of an LDC economy ha is a ne user of foreign exchange. 84

87 References Abbo, A., Darnell, A.C. and Evans, L The influence of exchange rae variabiliy on U.K. expors. Applied Economic Leers. 8 (2001), pp Adam, C.S., and Bevan, D Uganda: Aid, Public expendiure, and Duch disease. Universiy of Oxford. U.K. DFID-EA (Draf). Aingi-Ego, M and Sebbude, R.K Uganda s equilibrium real exchange rae and is implicaions for expor performance, Bank of Uganda Saff Papers, 2 (No.1). Edwards, S Real exchange rae, devaluaion, and adjusmen: Exchange rae policy in Developing Counries. MIT Press. Cambridge, MA. Kihangire D.A The impac of exchange rae variabiliy on Uganda s expors ( ). Ph.D. Thesis. Universiy of Bradford. Kihangire D., Pos D J, and Cameron S Has exchange rae variabiliy affeced Uganda s Coffee expors ( )? Discussion Paper 5, Universiy of Bradford MacDonald, R Modeling he equilibrium Real ERffecive Exchange Rae of he Uganda Shilling. Universiy of Srahclide. (Unpublished). Moniel, P.J The Long-run equilibrium Real Exchange Rae: Concepual issues and empirical research. In Hinkle E.L. and Moniel, P.J. (eds) (1999). Exchange rae misalignmen: Conceps and measuremen for Developing Counries, p Moniel, P.J Deerminans of he long-run Equilibrium Real Exchange Rae: An analyical model In Hinkle E.L. and Moniel, P.J. (eds) Exchange rae misalignmen: Conceps and measuremen for Developing Counries.p Musinguzi, P. and Muambi, B A monhly model of Uganda s Equilibrium Real Exchange Rae (ERER) Pah, inflaion, oupu gap and expors: An Auoregressive Disribued Lag (ARDL) approach o coinegraion. Pesaran, M.H. and Pesaran B Working wih Microfi 4.0: Ineracive economeric analysis. Oxford. Oxford Universiy Press. Pesaran, M.H., Shin, Y. and Smih, R Bounds Tesing Approaches o he analysis of level relaionships. Journal of Applied Economerics. 16, pp

88 Richaud, C, Varoudakis, A. and Veganzones, M Real Exchange Rae and openness in emerging economies: Argenina in he long-run. Applied Economics, 2000, (32), pp

89 The Bank of Uganda Saff Papers Journal Vol. 1, No Bank of Uganda IS INFLATION ALWAYS AND EVERYWHERE A NON- MONETARY PHENOMENON: EVIDENCE FROM UGANDA David Kihangire 39 and Alber Mugyenyi Absrac The paper examines he deerminans of inflaion in Uganda during 1994M7-2005M6. We es he hypohesis ha Uganda s inflaion is always and everywhere a non-moneary phenomenon. A heoreical background relaing inflaion o moneary and oher non-moneary facors is explored followed by a deailed empirical analysis of he facors. We apply he auoregressive disribued lag (ARDL) approach o coinegraion due o lack of a pure se of I(0) or I(1) daa for all he series. The resuls sugges insufficien evidence o accep he research hypohesis, as Uganda s inflaion rae is significanly correlaed wih boh moneary and non-moneary facors. Key Words: Uganda, Inflaion, exchange rae, exchange rae misalignmen, exchange rae variabiliy, erms of rade, rainfall, capaciy uilizaion, and ineres raes. 39 Auhors acknowledge commens from Dr. Louis Kasekende, Dr. Polycarp Musinguzi, and Dr. Michael Aingi-Ego and saff of he Research Deparmen, Bank of Uganda. The views expressed in his paper are hose of he auhors and do no necessarily represen he posiion of he Bank of Uganda. The auhors remain responsible for all oher errors and omissions. For correspondence on his paper use dkihangire@bou.or.ug. 87

90 I INTRODUCTION Inflaion is defined as he rae of increase in he general price level. I is commonly measured as he percenage change in he consumer price index (CPI) over a given period, usually one year. The remi of he cenral bank of Uganda, he Bank of Uganda (BoU), is o promoe and mainain he sabiliy of he value of he currency of Uganda; and regulae he currency sysem in he ineres of he economic progress of Uganda. The BoU is charged wih he formulaion and implemenaion of moneary policy direced fosering price sabiliy. Price sabiliy of necessiy requires moneary sabiliy. Exogenous facors, ha is, cosly energy and drough, have played a major role in he recen evoluion of inflaion. Indeed in he debae of recen inflaionary developmens, he exogenous facors appear o have gained prominence over he moneary aspecs. This would swing he pendulum away from he Friedmanie view ha inflaion is always and everywhere a moneary phenomenon. I can be said ha he view ha here is limied room for moneary policy o address he prevailing inflaion siuaion has gained currency, which invies he proposiion ha: Uganda s inflaion is always and everywhere a non-moneary phenomenon. This paper conribues o he ongoing debae on inflaion developmens in Uganda and explores he role of moneary policy in such a price environmen. II THEORETICAL PERSPECTIVE There are differing views regarding facors explaining inflaion, boh a he heoreical and empirical levels. Two heoreical perspecives are paricularly relevan o Uganda s curren inflaionary siuaion. Friedman (1953) posied ha inflaion is always and everywhere a moneary phenomenon. Accordingly, moneariss end o emphasize he imporance of moneary policy and he view ha inflaion is essenially a domesic phenomena semming from excessive money supply relaive o he growh and supply of goods and services. In his heoreical framework, i can be hypohesized ha inflaion varies posiively wih he rae of change in money supply, and negaively wih he rae of change in real income. On he oher hand, ohers (Durevall and Njuguna, (2001); De Grauwe and Polan, (2001); Nachega, (2001); Aa, Jefferis and Mannahoko, (1996); Chhibber, (1992)) argue ha inflaion is boh a moneary, as well as cospush-driven and hence posulae inflaion in he conex of purchasing power pariy (PPP) framework. In his framework, inflaion is as a posiive funcion of prices of raded goods (P T ); prices of non-raded goods ( P N ); and conrolled or regulaed prices, for example, uiliy ariffs (P C ). Inflaion is deermined according o absolue PPP, in which prices for non-raded goods are approximaed by excess supply of real money balances, prices of raded goods are approximaed by prices of impored goods, and oher cos-push measures are approximaed by uni-labour coss (wages). However, he PPP perspecive has a noed drawback as highlighed by some empirical sudies 88

91 (Isard, 1995), ha is, he pass-hrough from foreign prices o domesic prices is no very obvious, paricularly when he exchange rae appreciaes. This is because agens end o rea a reducion in impor coss (arising from exchange rae appreciaion) as increased profis, raher han domesic price reducions. Prices are indeed sicky downwards. An example of he PPP-inflaion nexus relaes o he ransmission mechanism of foreign prices o domesic prices, arising from he increases in oil prices on he world marke. The moneary policy response o increased oil produc prices is no clear-cu because such increases pose boh demand and supply shocks o an oil imporing economy 40. Higher oil prices push up inflaion (calling for a reducion in money sup ply or higher ineres raes), bu dampen gro wh (requiring he spurring of economic aciviy hrough easing money supply or cuing ineres rae s). The aggregae demand and aggregae s upply in he figure 5.1 below helps illusrae he effec of he double shock. F igure 5.1: Effecs of a double shock using agg regae demand an d aggregae supply Source: The Economis In he lef hand side of figure 5.1, he economy is in equilibrium a poins P1 and Q1. A higher oil price affecs an oil imporing economy wo-fold. I could lead o increases in firms producion coss and reduce profis and hereby reduce supply a any given price, shifing he aggregae supply curve lef o S2. I could also ransfer income from he oil-imporing counry o oil producers (he exen may be reduced hrough higher expors, where he richer oil producers increase demand for impors). Income and spending migh conrac in he oil-imporing counry, shifing he demand curve lef; o D2 (see lef hand panel of figure 5.1). A modes number of empirical sudies have invesed he deerminans of inflaion in Uganda. Mikkelsen and Peiris (2005), find ha: broad moneary aggregaes have he larges impac on prices, wih a 1.0 percen increase in money causing prices o increase by nearly 0.5 percen; changes in ineres raes have no effec on prices; and own innovaions (inflaion expecaions) 40 The Economis, The Crude Ar of Policymaking, Economics Focus, June 10 h

92 have a significan bu low persisen effec on fuure prices. The sudy also suggess ha changes in exchange rae have an impac on price levels, alhough he feed hrough effec is less han uniy. Nachega (2001) reveals ha inflaion in Uganda is a moneary phenomenon, consisen wih he findings of a cross-secion sudy by De Grauwe and Polan, (2001). III LINKING THE THEORETICAL PERSPECTIVES WITH THE RECENT INFLATION TRENDS IN UGANDA 3.1 Moneary growh and inflaion IS inflaion a moneary phenomenon? Table 5.1 below shows ha here is a close relaionship beween moneary expansion and all caegories of inflaion in Uganda. The correlaion beween inflaion and real money balances was negaive, and significan, consisen wih he heory relaing high inflaion wih reduced real money demand. This indicaes ha he view ha inflaion is a moneary phenomenon canno be ignored. Table 5.1: Correlaions beween Inflaion and Nominal and Real Money Growh Raes RBASMORM3 RM2 BASMO M3 M2 CPIH CPIU CPIF RBAS MO RM RM ** BASM O 0.946** M ** 0.878** 0.631** * M ** 0.911** ** CPIH ** ** ** ** CPIU ** * * * * CPIF ** ** ** ** 0.190* * Source: Auhors Figure 5.2: Base money versus desired levels (Shs. millions) Base Money 1,200,000 1,000, , , , , Jun Jul Aug Sep Oc Nov Dec Jan Feb Mar Apr May ,000 80,000 60,000 40,000 20, ,000 40,000 60,000 Policy Sance +(Tighen) (Ease) Toal Moneary Base Desired Moneary Base Moneary Policy Sance +(Tighen) -(Ease) Source: Bank of Uganda 90

93 The recen rends in figure 5.2 depicing base moneary growh vis-à-vis he arge levels reveals ha base money has on average been above desired levels, excep for he monhs in which he performance of base money managemen is judged (usually a he end of each quarer). The lumpiness of governmen expendiures parly accouns for such a rend. Uneven expendiure flows pose challenges o liquidiy managemen wihou exering undue money marke pressures and volailiy in ineres raes and he exchange rae. The above nowihsanding, he BoU has esablished a rack record of meeing he base money arge on a period average basis. Therefore, he inermien rend of base money above arge, may no necessarily ranslae ino underlying inflaion. However, persisence of base money above rend over a long period, could pose inflaionary pressures. 3.2 Nominal exchange rae and inflaion Is inflaion exchange rae driven? From he PPP heory, despie is drawbacks, exchange rae appreciaion should resul in lower inflaion raes, paricularly for raded goods. To invesigae his phenomenon, a simple plo was made beween underlying inflaion (comprising largely raded goods) vis-à-vis he exchange rae. The parial correlaions sugges a weak relaionship beween inflaion and exchange rae depreciaion. The correlaion beween inflaion and percenage changes in nominal exchange raes revealed weak correlaions of 0.1 each for boh headline and underlying inflaion. Figure 5.3: Inflaion V s changes in nominal exchange raes Inflaion % 5% 0% 5% 10% Jan 1994 Nov 1994 Sep 1995 Jul 1996 May 1997 Mar 1998 Jan 1999 % Change in Nov 1999 Sep 2000 Exchange rae Jul 2001 May 2002 Mar2003 Jan2004 Nov2004 Source: Bank of Uganda Underlying Headline % change in Exchange raes 3.3 PPP, foreign prices and inflaion Secion II indicaed he likely ransmission of foreign prices, such as increases in he world prices of peroleum producs, ino domesic inflaion. However, he impac on underlying inflaion (where core inflaion excludes energy prices) is no obvious rising or falling depending on he shapes of he demand and supply curves and he relaive sizes of heir lefward shifs. Headline inflaion (and underlying inflaion in Uganda s case where core inflaion includes energy prices) would rise. The imporan issue, herefore, is 91

94 he second round effecs, if high energy coss ransmi ino general prices and oher coss (such as higher coss of producion) across he whole economy. Therefore, he ulimae effec of higher oil prices would depend on how moneary policy reacs, hereby influencing he posiion of he demand curve. If moneary policy eases, in a bid o suppor oupu a Q1, increasing money supply (or cuing ineres raes), he economy risks ending up wih much higher inflaion a P3 in he righ hand side of figure 5.1, and possibly negaive real ineres raes. Bringing inflaion down henceforward would require aggressive conracion, which if excessively done, could cause a recession. Increases in exogenous facors such as adminisered prices (uiliy ariffs, among ohers) and oil prices, may yield one-ime increases in inflaion, which he cenral bank canno preven, in he firs round. However, he cenral bank he cenral bank mus srive o curb he spreading of he increases o general prices and coss. To assess he srengh of he poenial for ransmiing one-ime price increases economy wide, one needs o consider how flexible pricing power and labour markes are in Uganda. Pricing power may have spill-over effecs o many secors, (hough less so in Uganda, where mos wage conracs may no be indexed o inflaion and here is a considerable srucural slack in he labour marke). Neverheless, failure of moneary policy o respond cauiously would engender higher inflaionary expecaions, risking he spillage of exogenous upward price pressures economy wide, which would compromise macroeconomic sabiliy. 3.4 Ine res raes and inflaion The ineres pariy heory posis ha low domesic ineres raes exer inflaionary pressures hroug h expecaions of exchange rae depreciaion. Figu re 5.4 and Table 5.2 ( parial correlaions) are used o invesigae he relaionship beween inflaion and inere s raes, which boh reveal an inverse relaionshi p, beween ineres raes and inflaion. This may parly be explained by Uganda s low level of financial deepening o relaive o ha of Eas Africa. Broad money a s a percenage of gross domesic produc is 20.5 percen while he Eas Africa average is 25.6 percen World Developmen Indicaors, (2005). 92

95 Figure 5.4: Treasury bill raes and inflaion raes Treasury Bill Raes (Deflaed by Underlying Inflaion) and Inflaion Raes s % Jan 01 Apr 01 Jul 01 Oc 01 Jan Rae 02 Apr 02 Jul 02 Oc 02 Jan 03 Apr 03 Jul 03 Oc 03 Jan 04 Apr 04 Jul 04 Oc 04 Underlying Inflaion Headline Inflaion Real 91 day TB Raes ( Source: Bank of Uganda The link beween inflaion and ineres raes is ambiguous. In spie of he fac ha he BoU arges underlying inflaion we observed weak correlaion beween he underlying inflaion and real ineres raes. On he oher hand, consisen wih sudies ha used he headline inflaion rae (Mikkelson and Peiris, (2005); and Nachega, (2001)); one observes a posiive and significan correlaion beween inflaion and real ineres raes 42. Table 5.2 shows he correlaions beween he various measures of inflaion and real ineres raes over he period January 2001-Decemb er Table 2: Correlaions beween inflaion and real ineres raes CPIF CPIF CPIU CPIH R91H R91U CPIU CPIH ** ** R91H ** ** R91F R91U ** R91F ** * * ** ** R182F * * ** ** ** ** Source: Auhors The correlaions above sugges: R182F A srong posiive correlaion (0.9) beween food price inflaion and headline inflaion, bu a wea k one (0.2) wih underlying rae of inflaion. A srong posiive correlaion (0.5) beween headline inflaion and underlying rae of inflaion. This suggess ha headline inflaion migh affec underlying inflaion, mos likely hrough inflaion expecaions. 42 I is difficul o reconcile he observed posiive relaionship beween inflaion and Tbill rae, paricularly given he fac ha Tbills are used for moneary policy liquidiy managemen raher han for fiscal usage. 93

96 The underlying rae of inflaion is very weakly correlaed wih real ineres raes. This suggess ha measures o increase or reduce real ineres raes migh no have a significan impac on inflaion. The real rae of ineres (deflaed by underlying CPI) is weakly and negaively correlaed wih food inflaion. Secions wo o four examined he heoreical and empirical background o inflaion in Uganda. The following secions build on he background by exploring he salien driving facors, wih a view o obaining policy recommendaions. IV METHODOLOGY AND THE MODEL We adop he inflaion model employed by Durevall and Njuguna-N dungu (2001), incorporaing addiional variables such exchange rae variabiliy, exchange rae misalignmen, rainfall, and erms of rade, and a growh equaion. Consequenly, we esimaed an inflaion model of he form given by: ( LBASMOp ) = γ 0 + γ LCUIID + γ2r + γ3lneero + γ LVNEERO + γ5mislnmt + ε (5.1) 1 4 p MISLNMT + v = 0 + β1 LCUIID + β2lneero + β3ltotu + β4lvneero + β5ldefgdpt + β6 β (5.2) LCUIID RAIN+ v = 0 + η1 LNEERO + η2ltotu + η3ldefgdpt + η4mislnmt + η5 η (5.3) The reduced-form inflaion equaion can be wrien as:- LCPI = α 0 + α1 LBASMO + α2lcuiid + α3 LR91 + α4ltotu + α5 LNEER α LVNEERO+ α LDEFGDPT+ α MISLNMT+ α LRAIN+ ε (5.4) Where: LCPIH = log of he headline consumer price index LBASMO = log of he base money LCUIID = log of he capaciy uilizaion measured by index of indusrial producion as a raio of rend oupu LR91 = log of 91-day real Treasury bill rae LTOTU = log of erms of rade LNEER = log of nominal effecive exchange rae LVNEERO = log of nominal effecive exchange rae variabiliy LDEFGDPT = log of fiscal defici as a raio of GDP MISLMNT = he model-based exchange rae misalignmen (Kihangire e al (2005)) LRAIN = log of average monhly rainfall 94

97 4.1 Time series properies of he daa To invesigae he order of inegraion of he ime series daa used here, we used he Augmened Dickey-Fuller (ADF) es. For a paricular ime series variable, he ADF model akes he form:- x = 0 + βt Where, α + n λ x 1 + i i= 1 α x i + ε ( 5.5) x = he variable x, wh ose uni roo es is being invesigaed for he order of inegraion T = Time rend, and ε = Error erm. Using he sandard abulaed ADF saisics, we es he null hypohesis: H : λ = 0 (5.6) 0 Agains he alernaive hypohesis ha: H : λ 0 (5.7) A The resuls of he uni-roo analysis (Table 5.3) sugges ha here is lack of a pure series of I(0) or I(1) for all he variables of he equaion model. The series are found o be of mixed orders of inegraion I(0), I(1) or I(2). Consequenly, we employed he auoregressive disribued lag (ARDL) approach o coinegraion. The resuls of he inflaion equaion (9) based on ARDL approach presened in Table 5.4 and

98 Table 5.3: Descripion of Variables Variable Variable Name ADF(X) ADF( X) ADF( X) Order of Inegraion log of he headline LCPIH I(2) consumer price index Log of he base money LBASMO I(2) Log of he Money, M2 LM I(2) Log of he capaciy LCUIID I(2) uilizaion measured by index of indusrial producion as a raio of rend oupu Log of 91-Day real Tbill LR I(0), I(2) rae Log of Uganda s Terms LTOTU I(2) of Trade Log of Nominal LNEER I(0), I(2) Effecive Exchange Rae Log of Nominal LVNEERO I(0), I(2) Effecive Exchange Rae variabiliy Log of fiscal defici as a raio of GDP LDEFGDPT I(0), I(2) Exchange Rae MISLMNT I(1) Misalignmen Log of Uganda s LRAIN I(0), I(2) average monhly rainfall Sources: au hors Given ha here are mixed orders of inegraion, we esimaed he corresponding ARDL-ECM of equaion (9) of he form given b y:- n n n n n n 0 + β 1 i LCPI 1 + LBASMO + β LCUIID i + LR91 i + LTOTU i + β i LNEER+ i= 1 β2 i 3i β4i β5 i 6 i i= 0 i= 1 i= 0 i= 0 i= 0 LCPI = β n n... + β7 8 + i LVNEERO 1 + β i LDEFGDPT i β9i MISLNMT i + β10i LRAIN i i= 0 i= 0 i= 0 i= 0 + γ 1 LCPI γ 2LBASMO 1 + γ 3LCUIID 1 + γ 4LR γ 5LTOTU 1 + γ 6LNEER n n γ 7 LVNEERO 1 + γ 8LDEFGDPTγ 1 + 9MISLNM 1 T+ γ10lrain 1 + ε (5.8) 96

99 Table 5.4: General o Specific ARDL-ECM CPIH inflaion esimaion. Base money Monhly lags and leads C DLCPIH (1.7764) DLBASMO ( ) ( ) ( ) (-3.138) ( ) DLCUIID ( ) (-2.798) (-3.896) DLR ( ) (2.5886) ( ) DLTOTU (-1.356) DLNEER (-2.913) (2.4107) ( ) ( ) DLVNEERO (4.2010) ( ) DLDEFGDPT ( ) ( ) ( ) (-2.952) (-2.748) DMISLMNT ( ) (-3.483) DLRAIN (2.9858) (-3.192) DLCPIH (2.8120) DLBASMO ( ) DLCUIID (4.6338) DLR ( ) DLTOTU (2.7005) DLNEER ( ) DLVNEERO ( ) DLDEFGDPT ( ) DMISLMNT (2.3877) DLRAIN ( ) R 2 -Bar DW Sa RSSR Serial F(12.65) =1.131 Correlaion Funcional F(1.76)= form Normaliy CHSQ(2)= (0.9) Heeroskadasi CHSQ(1)= (0.19) ciy Source: Auhors 97

100 Table 5.5: General o Specific ARDL-ECM CPIH inflaion esimaion. M2 Monhly lags and leads C (7.2099) DLCP IH (1.3152) ( ) DM ( ) ( ) DLCUIID ( ) ( ) ( ) ( ) ( ) DLR DLTOTU ( ) ( ) ( ) (2.1053) ( ) ( ) DLNEER DLVNEERO (1.4384) (4.3301) (2.5055) (3.3460) (1.2297) (1.8989) (2.6015) ( ) ( ) (2.1741) ( ) (1.5742) (2.0722) ( ) DLDEFGDP T ( ) ( ) ( ) ( ) ( ) DMISLMNT ( ) ( ) ( ) ( ) DLRAIN DLCPIH (4.7920) (3.9856) (3.2311) (3.1980) ( ) DLBA SMO (5.5688) DLCUIID ( ) DLR (1.5281) DLTOTU ( ) DLNEER DLVNEERO ( ) ( ) DLDEFGDP T ( ) DMISLMNT ( ) DLRA IN R 2 -Bar DW Sa RSSR Serial Correlaion Funcional form Normaliy Heeroskadas iciy Source: Auhors ( ) F(12.65)=1.131 F(1.76)= CHSQ(2)= (0.7 50) CHSQ(1)=5.6840(0.01 7) ( ) 98

101 V RESULTS AND DISCUSSION The paper ess he cenral hypohesis ha inflaion is no a moneary phenomenon. The resuls of he analysis focusing on headline inflaion, and base money, B0, and M2 lead o he following conclusions: a) Is here a long-run relaionship beween headline inflaion and is major deerminans? The resuls of he regression were affirmaive. The coefficien of he lagged levels of CPIH is negaive and significan consisen wih he F-bounds es saisic. b) Does money growh maer for inflaion? The resuls were affirmaive, conrary o he proposiion and he cenral hypohesis. A 1.0 percen increase in broad money (M2) leads o a 0.2 percen increase in headline inflaion. Likewise, a 1.0 percen increase in base money leads o a 0.1 percen increase in inflaion. The resuls sugges ha regulaing he growh of money has a significan conribuion o regulaing inflaion. c) Does real aciviy maer for inflaion? The resuls sugges ha alhough real aciviy maers, he measured effecs are insignifican and very inelasic. d) Do real ineres rae changes maer for inflaion? In boh equaions, he resuls sugges ha i maers. A rise in real ineres raes migh lead o inflaion, alhough he measured effecs are inelasic. e) Do changes in erms of rade maer for inflaion? The empirical evidence suggess ha hey maer. An improvemen in he erms of rade helps o reduce inflaion and he measured effecs are significan alhough inelasic. f) Does he exchange rae maer for inflaion? The empirical evidence suggess ha i is he second mos imporan and significan facor in explaining inflaion. The elasiciy of response is 0.2 suggesing ha a 100 percen increase in he nominal effecive exchange rae is associaed wih a 20 percen reducion in inflaion. g) Does exchange rae variabiliy maer for inflaion? The resuls sugges ha alhough exchange rae variabiliy is associaed wih an increase in inflaion, he measured effecs are inelasic and insignifican. h) Does he fiscal defici/gdp maer for inflaion? The resuls sugges ha his is he single mos imporan facor in explaining Uganda s inflaion rae. The elasiciy of response is high (0.7) suggesing ha a 100 percen increase in he fiscal defici is associaed wih 65.0 percen increase in inflaion. i) Does exchange rae misalignmen maer for inflaion? The resuls sugges ha alhough exchange rae misalignmen is negaively and significanly associaed wih inflaion, he measured effecs are negaive and inelasic. j) Finally, one migh ask wheher rainfall maers for inflaion in Uganda. The resuls sugges he affirmaive; an increase in rainfall is associaed wih a decline in inflaion alhough he measured effecs are inelasic and a imes insignifican. In he inflaion equaion 99

102 involving M2, he elasiciy of response is only -0.02, while he corresponding equaion involving base money, he elasiciy of response is only The analysis above reveals ha here are significan shor-run and long-run relaionships beween inflaion and money, fiscal deficis, changes in he exchange rae, among ohers. However, he lag-lenghs of he effecs vary among facors. Wha are he moneary policy implicaions of hese resuls? The following secion examines his issue in more deails. VI POLICY IMPLICATIONS Given he above exposiion, here are several quesions facing moneary policy makers, hus: a) Wha would a cenral bank do if boh he headline and underlying rae of inflaion were negaive? b) Wha would a cenral bank do if he headline inflaion rae is highly (double-digi) posiive bu he underlying rae is negaive? c) Wha would a cenral bank do if he underlying inflaion rae is highly posiive, bu he headline is negaive? d) Wha would a cenral bank do if boh he headline and underlying inflaion raes are rising, and he headline is in double digis? The answers o cases (a) and (d) are sraighforward, while hose for (b) and (c) are complex. In he IS-LM-Bexchange rae regime era, moneary policy can be eased in response o case heoreical framework, and in he floaing (a), paricularly if he corresponding fiscal policy response is no oo loose. In he case of (d), moneary policy could be ighened paricularly if fiscal policy is unchanged or loose. Cases (b) and (c) are complex and require moneary policy resrain. In case (b), one would ighen moneary policy relaive o wha one had iniially programmed. This is because of he significan correlaions beween headline inflaion and he underlying inflaion as observed above, which could resul in headline inflaion feeding ino he underlying inflaion. In he case of (c), one could consider easing moneary policy, since he pursuance of moneary policy is for economy- wide concerns. The response o increased oil produc prices is no clear-cu because such increases pose boh demand and supply shocks o an oil imporing economy 43. Higher oil prices push up inflaion and dampen growh. However, given he remi of he cenral bank, he ideal response o an oil price shock such as he presen one would be o slow down money supply growh over ime. 43 The Economis, The Crude Ar of Policymaking, Economics Focus, June 10 h

103 Also, while he increases in uiliy ariffs and oil prices may yield sep-wise hikes o inflaion; he appropriae moneary policy response mus aim a curbing he ensuing second round effecs, o anchor inflaion expecaions. Failure of moneary policy o respond cauiously would soke general inflaion. In assessing he possible policy response by he cenral bank, i is imporan o reflec on he recen sudy by he African Deparmen 44 of he IMF, which showed ha inflaion in Uganda is driven by is own innovaions, as well as by moneary and exchange rae shocks. I has been shown ha broad moneary aggregaes have an idenifiable ransmission mechanism on consumer prices while ineres raes do no. A 1.0 percenage poin increase in M2 leads o a 0.2 percen rise in core inflaion in hree monhs. On he basis of he money muliplier, i can be inferred ha reducing he growh in he moneary base migh assis in lowering inflaion. However, i mus be emphasized ha moneary policy, alone, migh no address oher exogenous srucural shocks ha pose upward inflaionary pressures. Moneary policy can anchor inflaion expecaions, by ensuring ha prices in non-food iems slow down, relaive o oher prices. Exchange rae shocks have a srong effec on boh core and headline inflaion, persising for abou hree o six periods. Therefore, he BoU mus ensure exchange rae sabiliy. Likewise, pruden fiscal policy sance would ease inflaionary pressures. VII CONCLUSION This sudy aimed a invesigaing he proposiion ha Uganda s inflaion is always and everywhere a non-moneary phenomenon. The resuls of his sudy sugges ha here is insufficien evidence o accep he research proposiion. On he conrary, moneary developmens are a very imporan and significan facor in explaining inflaion in Uganda--- reducing moneary growh helps o reduce he rae of headline inflaion. However, here are also oher imporan facors ha help o explain Uganda s inflaion, he mos imporan ones being he fiscal defici and he exchange rae. Reducing he fiscal defici and ensuring exchange rae sabiliy help o reduce inflaion. From he foregoing exposiion, we conclude ha Bank of Uganda s moneary and exchange rae policies have a role o play in sabilizing he rae of inflaion. Tighening of moneary policy would help o anchor inflaionary expecaions downwards, when faced wih srong inflaionary pressures arising from exogenous facors. Policy accommodaion of exogenous shocks in a bid o suppor growh would prove ime inconsisen and couner producive. The pruden moneary policy response o exogenous inflaion shocks is one ha hals he ensuing second round effecs from spilling o general prices. 44 IMF (2005), Uganda: Seleced Issues and Saisical Appendix, January

104 References Aa, J.K, Jefferis, K.R. and Mannahoko, I Small counry experiences wih exchange raes and inflaion: Boswana. Journal of African Economies. 5(2), pp Chhibber, A Exchange reforms, supply response and inflaion in Africa. In I. Goldin and L.A. Winers (eds). Open Economies: Srucural Adjusmen and Agriculure. Cambridge Universiy Press. De Grauwe P. and Polan, M Is inflaion always and moneary phenomenon? CEPR and Universiy of Leuven. everywhere a Durevall, D. and Njuguna S. Ndung u A Dynamic model of inflaion for Kenya, Journal of African Economies. 10(1), pp The Economis. The crude ar of policymaking, Economics Focus, June 10 h Friedman, M Essays in posiive hinking. Universiy of Chicago. IMF Uganda: Seleced issues and saisical appendix, Inernaional Moneary Fund, washing D.C. Isard, P Exchange rae economics. Cambridge. Cambridge Universiy Press. Kihangire, D.A., Abuka, C. and Musinguzi, P Exchange rae misalignmen and financial liberalisaion: Empirical evidence and macroeconomic implicaions for Uganda, Bank of Uganda Saff Papers (forhcoming) Mikkelsen, J. and Peiris J. S Uganda: Seleced issues and saisical appendix. IMF, Washingon D.C. Nachega, J-C Financial liberalizaion, money demand, and inflaion in Uganda. IMF Working Paper Number WP/01/

105 The Bank of Uganda Saff Papers Journal Vol. 1, No Bank of Uganda MEASURING EFFICIENCY OF A MARKET IN TRANSITION: THE UGANDAN FOREIGN EXCHANGE MARKET 45 Michael Aingi-Ego and Rachael Ssebudde Absrac Since is inroducion in 1993, he Ugandan iner-bank foreign exchange marke has grown in size and deph. However, boh economeric and survey resuls sugges ha he marke is sill inefficien and marke dealers do no seem o ake advanage of he exising profiable rading opporuniies offered by he inefficiency, he seasonaliy of foreign exchange flows and rading paerns during he week. Alhough exisence of a few big marke players, who lead he marke, is one plausible explanaion for his, delays and even lack of informaion assimilaion is even more imporan. Marke dealers indicae ha wih excepion of BOU inervenion, he bulk of economic informaion from he auhoriies has minimal impac and akes long o be assimilaed ino expecaions. While i may drive he exchange rae away from is fundamenal value, BOU inervenion reduces marke volailiy. Such a role is crucial and criical especially for a marke characerized by seasonal movemens o eliminae unwarraned volailiy and speculaive endencies. The main policy challenges ha arise from his sudy include he developmen of forward foreign exchange markes and insrumens o hedge agains volailiy in he exchange rae, increasing he sensiiviy of exchange rae o BOU policy raes hrough srenghening he link beween he domesic money and foreign exchange marke. The oher policy dilemma is he effec of coninued serilizaion of liquidiy injecion arising ou of povery reducing donor inflows in he foreign exchange marke, which reduces incenives for forward marke developmen and could discourage porfolio flows. Key words: Marke efficiency, foreign exchange marke, marke expecaions, cenral bank inervenion. 45 The auhors would like o express heir graiude for he consrucive commens, echnical and financial suppor hey received from he EPRC. In addiion, hey grealy acknowledge he conribuions from heir colleagues a he Bank of Uganda for he invaluable suppor received in form of echnical discussions and informaion sharing. The views expressed herein are purely hose of he auhors and are mean for generaing policy debaes and any commens are highly welcome. Auhors Addresses: Ego@bou.or.ug and RSebudde@bou.or.ug 103

106 I INTRODUCTION In order o improve he efficiency in he allocaion of foreign exchange resources and he exernal balance posiion of he counry, Uganda in November 1993 liberalized is foreign exchange marke and allowed he exchange rae o be marke deermined. To dae, he number of acors in Ugan da s foreign exchange marke has seadily grown and he spreads beween he buy and sell raes have also gradually narrowed possibly reflecing increased compeiion. Our expecaion is ha as markes evolve, so does he behavior of he marke paricipans. The quesion ha arises is wheher his evoluion has been followed by some degree of efficiency. The Efficien Marke Hypohesis (EMH) posulaes ha a marke is efficien if here are a large number of buyers and sellers, wih no single agen on is own able o influence he pricing in he marke. Under such condiions, he large number of produc derivaives, free enry and exi, and flow of informaion o all agens allows he prices prevailing in he marke o reflec full informaion, and hence raders earn only normal profis. Depending on how much informaion is used o form expecaions abou fuure prices, differen forms of efficiency can be deduced - weak form, semi-srong form and srong form efficiency (Fama, 1970). When marke agens exrapolae exchange raes using he mos recen end, he marke would be characerized by bandwagon or desabilizing expecaions. Conversely, regressive expecaions occur when marke paricipans expec he exchange rae o reurn o some equilibrium level especially following an earlier overshoo in he exchange rae. On he oher hand, absence of insrumens and markes o hedge agains risk exposure and uncerainy associaed wih he exchange rae movemens cause a herding behavior in agens. I resuls ino self- fulfilling expecaions abou he evoluion of he exchange rae and hence insabiliy in foreign exchange marke (see Kasekende and Aingi-Ego (1996)). The disincion beween views relaing o he way expecaions are formed inhibis an imporan policy debae on exchange rae managemen especially for developing counries, including Uganda. Differen views have been posulaed on he effec of speculaion on markes. In conras o Friedman s (1953) who posulaes ha speculaors aciviies ensure ha exchange raes reflec he fundamenal value of a currency, Froo and Thaler (1990) argue ha speculaion in foreign exchange can be desabilizing and ha excess volailiy imposes large coss on producers, who in urn make less efficien allocaion decisions. One such poenial cos likely o subsis in small one-commodiy-expor and impor-dependen economies is he possibiliy of adoping a wai and see aiude, when invesors are confroned wih uncerain fuure exchange rae changes. Given he wide flucuaion in he exchange rae ha could resul from a marke-deermined exchange rae, Uganda, like many oher auhoriies adoped a managed floa exchange regime o miigae excessive insabiliy in he exchange raes by inervening in he foreign exchange markes. The success of hese acions in influencing he behavior of marke paricipans is also an empirical quesion whose answer may vary from one marke o anoher, probably 104

107 depending on he form of efficiency of hese markes. In ligh of he above argumens, he main objecives of he sudy are fourfold: (i) To deermine he form of efficiency of he Ugandan foreign exchange marke and he way expecaions on exchange rae are formed, drawing on he Fama (1970) definiions of efficiency; (ii) Depending on he level of efficiency derived in (i) above, o empirically es for exisence of profiable rading opporuniies in he Ugandan foreign exchange marke and he presence of speculaive aciviies; (iii) Invesigae he impac of inervenion in he foreign exchange marke by he Cenral Bank; and (iv) Formulae policy proposals for he foreign exchange managemen based on he resuls derived above. The res of he paper is organized in six secions. Secion II summarizes he evoluion of he foreign exchange marke in Uganda. In Secion III, a review of he relaed lieraure is summarized while he mehodology; model formulaion and daa requiremens for esing efficiency and exisence of profiable opporuniies are presened in Secion IV. The resuls, boh from he empirical sudy and he survey of he foreign exchange marke, are presened in Secion V, closely followed by he conclusions and policy recommendaions in he final secion. II EVOLUTION OF THE FOREIGN EXCHANGE MARKET IN UGANDA Among he key facors ha have influenced he evoluion of he foreign exchange markes in Uganda are differen economic and poliical regimes. In he immediae pos 1960 period, for insance, he srong expor performance ensured ha Uganda s impor needs were largely me from expor receips wihou serious implicaions on he curren accoun balance. This helped o susain he posiive growh raes ha were posed during his period. In spie of he fixed exchange rae regimes pursued hen, supply and demand equilibraed and no parallel markes developed for foreign exchange. The decade of he 1970s did no only face huge poliical and economic mismanagemen bu also winessed he firs oil crisis and declining inernaional prices for expor commodiies ha led o worsening erms of rade. To he exen ha he auhoriies mainained he fixed exchange rae regime and used i o allocae he foreign currency a raes ha could no equilibrae supply and demand, parallel markes in foreign exchange emerged o saisfy he unme demand from he official channels. By 1981, impor requiremens o suppor and susain economic growh could no be me. The overvaluaion of he exchange rae led o he deerioraion of he exernal balance and he worsening inernal balance exacerbaed his. A his ime, he price of foreign currency in he parallel marke was over 30 imes higher han he official exchange rae of UShs 7.14 per US Dollar. 105

108 Effors o resore macro-economic sabiliy in early 1980s included exensive innovaions in he area of exchange rae managemen, which were iniiaed wih a managed floa regime in he period Following his was he dual exchange rae regime, which esablished wo windows a he BOU in Weekly aucions were laer adoped wih he merging of he wo windows in 1984, and finally o dual sysem again in 1986 a raes of UShs 14 per US Dollar and UShs 50 per US Dollar for essenial and non-essenial ransacions respecively. Toward he end of he year, hese raes were merged ino a single rae of UShs 14 per US Dollar. While his was happening a he official BOU channel, he parallel marke ha had emerged in he lae 1970 s inensified and consequenly, he premium beween he official exchange rae and he parallel marke rae widened. Following he poliical change in 1986, an Economic Recovery Program was launched in The reform programme ha benefied from he subsanial financial and echnical suppor from he IMF, World Bank and he inernaional communiy was largely direced a macro-economic sabilizaion and kick-saring producion in he economy. Among he major reforms implemened were he rade and paymen arrangemens, he financial secor and markeing liberalizaion. Economic Recovery Credi (ERC) received in May 1987, suppored a currency reform ha saw one hundred shilling exchange for one new shilling and his was accompanied by a devaluaion of 77 percen o address he imbalances in he exernal secor. A he ime, he premium beween he parallel and official marke also declined from 1000 per cen o 50 percen. The reforms were gradualis in approach and his largely refleced he fac ha here was no general consensus among he key players on which direcion he exchange rae reforms should follow (Holmegren e al (2000)). To suppor impor dependen indusries, policies and programs were pu in place during his period including he Open General License (OGL, 1988), Special Impor Programs (SIP) 46 and Dual Licensing schemes for exporers wishing o impor. Even wih hese programs in place, arbiraging could no be eliminaed, paricularly when muliple foreign exchange markes exised. All his was clear esimony ha supply and demand could no be equilibraed a he prevailing official exchange rae. The infrequen bu large discree devaluaions effeced by governmen were no sufficien o remove he disequilibria exising in he marke. Effecive July 1989, governmen adoped a more acive sance when i moved o a crawling peg exchange rae regime. The official exchange rae now adjused in line wih he inflaion differenials beween Uganda and is main rading parners, Kenya, UK and US on a monhly basis. One year laer, in July 1990, Governmen ook a bold sep in he foreign exchange managemen policy o legalize he parallel marke and commission he Foreign exchange bureaux. This was laer followed by he inroducion of he weekly Duch aucion for foreign exchange a he official channel. The 2 SIP I was launched in 1988, while SIP II and SIP111 were respecively inroduced in 1989 and SIP

109 premiums beween he muliple foreign exchange markes exising hen (direc sales from BOU s official cash flow, he Duch aucion for donor funds and he forex bureaux) sared declining. To faciliae convergence of he marke, Governmen inroduced he iner-bank marke for foreign exchange in November 1993 and his compleed he liberalizaion of he exchange and paymens sysem. The marke deermined exchange rae regime was expeced o provide a more efficien and reliable mechanism for he allocaion of foreign exchange resources. Under his regime, commercial banks and forex bureaux were expeced o underake all forex ransacions wih he privae secor a he marke deermined buy and sale raes. BOU adoped he weighed exchange rae derived from he raes se by he inerbank marke as he official rae. Inervenion by BOU is mainly o smoohen wide flucuaions in he exchange rae in order o sabilize marke condiions 473. While bureaux are allowed o concenrae on reail spo ransacions in insrumens such as cash, cheques and bank drafs, banks ha are licensed as auhorized dealer banks can operae wholesale, reail and forward ransacions. In rying o undersand he evoluion of he Ugandan foreign exchange marke, Abuka and Sajjabi, (1996) invesigaed he facors responsible for he appreciaing rend of he foreign exchange rae over he period 1993 o In addiion, Aingi-Ego and Sebudde (1999) esimaed he equilibrium pah for he exchange rae and also esimaed levels of misalignmens for he period Specifically hey invesigaed he impac of he misalignmens in he shilling on performance of he non- radiional expors. None of he sudies have direcly ackled he issue of marke efficiency in he Ugandan foreign exchange marke, speculaion and insabiliy and he impac of he cenral bank s inervenion. III LITERATURE REVIEW There is a wealh of lieraure devoed o esing he hypohesis of efficiency of foreign exchange markes bu i has produced mixed resuls. Fama (1965) described an efficien marke as consising of a large number of compeiive 47 Effecive April 1, 2002, BOU changed he modaliies for enering he IFEM in order o reflec he disincion beween is objecives. Whereas prior o 2002, is presence in he IFEM was on accoun of inervenion o sabilize he marke, he increase in donor financed governmen expendiures necessiaed he use of forex sales o serilize he liquidiy. Consequenly, effecive April 2002, a framework of foreign exchange sales comprises of boh serilizaion and inervenion. Serilizaion is mean o remove liquidiy generaed by povery reducing donor flows in a manner ha exers minimum pressures on he exchange rae. I does his in a rading like manner where i ransacs wih an auhorized dealer bank posing he mos compeiive rae. The variaions in amouns raded are dependen on he liquidiy ha BOU is desirous of removing from he sysem. BOU does no announce is presence in he marke when i is serilizing. On he oher hand, he moivaion of BOU inervenion acions is o influence he exchange rae. I does so by announcing is presence in he marke bu no he amouns. Once again, ransacions are sruck wih he auhorized dealer banks posing he mos compeiive raes on he Uganda Invesmen Services (UIS). The period under which he changes in modaliies were effeced is ouside our sudy period, which has used daa from 1998 o

110 profi maximizers ineracing in a marke and uilizing all available informaion in a raional manner. In an efficien marke, prices should fully reflec all he relevan and available informaion; hence, no profi opporuniies are lef unexploied. If currency markes are efficien, he spo or forward exchange raes should embody all relevan informaion, and hey should no be able o forecas he spo or forward exchange rae as a funcion of anoher. Also, he curren forward rae should be an unbiased predicor of he fuure spo rae if we assume risk neuraliy and a covariance saionary risk premium (i.e he curren forward exchange rae should forecas he fuure spo rae if he markes are efficien). This is a long-sanding issue in he lieraure on foreign exchange markes. As Hodrick (1987) argues, here is very srong evidence agains he hypohesis ha forward exchange raes of any mauriy from one day, o one week o one or hree monhs are unbiased predicors of fuure spo raes. When he unbiasedness hypohesis does no hold, here is evidence of a risk premium, marke inefficiency or boh. While Baillie and Bollerslev (1989), MacDonald and Taylor (1989), Diebold e al (1994) and Rapp and Sharma (1999) argued ha coinegraion among exchange raes in differen currencies implies failure of marke efficiency, Dwyer and Wallace (1992), Crowder (1994, 1996) and Engel (1996) have demonsraed ha here is no connecion beween co-inegraion of spo raes and marke inefficiency. Levin and Lin (1992), Im e al. (1995) and Wu and Chen (1998) showed ha he improved saisical power of uni roo ess derived from using grouped or pooled cross-secional daa oher han individual series srongly suppors he hypohesis of marke efficiency. Alexakis and Apergis (1996) also prove he presence of he efficien foreign exchange marke hypohesis by modeling condiional heeroskedasiciy hrough ARCH models. As is depiced by he survey by Froo and Thaler (1990), he convenional es for efficiency assumes ha he forward exchange rae is an unbiased esimae of he ex pos spo rae. A broad range of lieraure has also proposed he use of expecaions survey daa o improve furher he efficiency es, jus in case here are any biases ha can be observed in his efficiency ess. Consequenly, Ellio and Io (1995) use micro survey daa o examine he efficiency of he forward yen/dollar marke and find ha he survey daa is an imporan source of supplemenary informaion on he behavior of he markes. The idea ha speculaion sabilizes exchange rae markes goes back o Friedman s (1953) conjecure. This view is, however, inconsisen wih he McKinnon (1976) argumen ha idenifies excessive exchange rae volailiy wih speculaion. A survey by Cheung e al (1999) explains he idea ha volailiy, sabilizaion (in he sense of moving owards fundamenals) and speculaion go hand in hand. Their findings from a survey of Unied Saes foreign exchange raders showed ha 84% of raders agreed wih he proposiion ha speculaion increases volailiy. Ineresingly, raders also 108

111 view speculaion as pushing exchange raes oward heir fundamenal values. According o heir resuls, speculaion enhanced marke liquidiy by 81%, and improved marke efficiency by 74%. An ineresing oucome of hese responses is ha speculaion is viewed as an inegral aspec of he foreign exchange marke, and ha volailiy is no inimical o working markes. In relaion o markes in ransiion, Dockery and Taylor (1998) invesigaed he long run dynamics of parallel and official exchange raes in four Easern European counries and heir resuls depic a varying speed of adjusmen ha is dependen on he economic mauriy of he counry and he presence of weak form informaion efficiency in he parallel markes. Sarwar (1998) invesigaed he weak form efficiency in he Souh Eas Asian parallel markes and concludes ha here exiss arbirage opporuniies in hese markes due o lack of efficiency. Likewise, Ogiogio (1994) who regresses he exchange rae on he raio of inflows o ouflow of foreign currency and Ayogu (1995, 1997) who uses VAR analysis and nonlinear algorihm of Hamilon (1989), have addressed he issue of marke efficiency and speculaive aciviies in he parallel marke in Nigeria alhough hey come up wih wo divergen views abou he exisence of noise rading in he marke. The marke efficiency hypohesis formulaion has wo major problems: i) he inerpreaion associaed wih he rejecion of marke efficiency and ii) he ambiguiy in consrucing alernaives o he null of efficiency. These problems arise since failure o find evidence in favor of he null may imply eiher a rejecion of he informaion se (probably on he assumpion ha agens are raional bu have he wrong model) or ha he informaion se has all he relevan informaion bu agens are no using he available informaion and hence irraional. Ayogu (1997) adops he hypohesis independen of hese implicaions for raional expecaions and hence he erm speculaive efficiency o sress he focus on opporuniies for risk-adjused excess reurns. In ligh of hese conflics, here is need o search for alernaive formulaions of marke efficiency. Exensive lieraure has been developed sudying speculaive efficiency and in paricular esing for exisence of profiable rading opporuniies. Bilson (1981), Dooley and Shafer (1982), Bilson and Hsieh (1983) and Hodrick and Srivasava (1984) are some of he examples. Ayogu (1997) measures he risk-reurn profile implici in he parallel foreign exchange marke by using a random walk funcion for he spo exchange rae and minimizing he level of diversifiable risk in he porfolio of currencies given a level of reurn o derive he efficien fronier for analyzing he opimal risk/reurn rade-offs implici in he parallel marke. He goes furher o explore he issue of speculaive dynamics (noise raders in he marke), an approach ha has been used o sudy gradual swings of prices away from fundamenals. The noise-rader approach makes wo major assumpions: i) he exisence of some invesors who are no fully raional and whose demand for risky asses is affeced by beliefs and senimens no fully jusified by fundamenal news, on one hand, and ii) exisence of raional invesors who 109

112 limi arbirage as his is deemed risky. (Shleifer and Summers, 1990). These wo assumpions imply ha when invesor senimens change, hey are no fully counered by arbirageurs and hence would affec reurns on asses. De Long e al. (1989, 1990a) sugges ha invesors who rade on noise fare worse han if heir expecaions had been raional. However, Shleifer and Summers (1990) argue ha opimal rading by raional agens would compleely reverse he effecs of irraional raders only if he raional agens are no consrained by wealh and are risk neural. Noise raders can also affec he welfare of he communiy by subjecing hem o wide savings in he exchange rae. On he role of cenral bank inervenion in foreign exchange marke, a large body of research has also been generaed. Typically, researchers conclude ha foreign exchange inervenion has lile effec on he firs momen (see Obsfeld, 1990). Tha nowihsanding, Edison (1993), Dominguex and Frankel (1993) and Kaminsky and Lewis (1996) have argued for a channel for inervenion hrough he signaling of fuure moneary policy. The resuls of a survey by Cheung e al (1999) are consisen wih he view of lile effec, wih opinion evenly spli beween inervenion pushing currencies away and oward heir fundamenal value. Similarly here is a spli opinion on wheher such inervenion is successful where he crierion of success is deermined by he responden s inerpreaion of he cenral bank s goal. There is a slighly more posiive response on he appropriaeness of he iming of cenral bank inervenion (60 percen), while 61 percen of he survey respondens view cenral bank inervenion as exacerbaing volailiy. Alhough hese las wo responses appear conradicory, he view of marke paricipans is ha increases in volailiy may go hand in hand wih marke efficiency. In fac, hese resuls are no inconsisen wih hose obained by Edison (1998) in her case sudies of cenral bank inervenion. She finds ha US inervenion in recen years has been infrequen, and sizable by hisoric sandards. Moreover, he Unied Saes markes have effeced changes, albei shor-lived, in he rend of exchange raes. This experience may explain why he US raders have a relaively posiive view of cenral bank inervenion. The fac ha news causes he grea bulk of movemens in exchange raes has a venerable hisory, going back o as far as Frenkel (1977). However, empirical aemps o link exchange rae movemens o specific announcemens of macroeconomic variables have been hampered by he difficuly in exracing he unexpeced componen in hese announcemens, as well as he fac ha he sudies are ofen conduced wih relaively low frequency daa. In paricular, i may be ha announcemen effecs have dissipaed by he ime he exchange rae daa are sampled, even when he daa frequency is daily or even hourly. In heir survey, Cheung e al (1999) found ha in fac he response of he exchange raes o news is exremely rapid on he order of minues for mos variables. In such cases even minue-by-minue daa migh no cach his news effec. 110

113 IV METHODOLOGY OF THE STUDY Expecaions surveys are an imporan source of supplemenary informaion in undersanding he behavior of markes and also provide a mode for assessing he effecs of some key elemens of he marke like reacion o news, for which economeric sudy is hampered by one facor or anoher. This sudy herefore adoped he approach by Ellio and Io (1995) and will herefore, underake a survey of he foreign exchange marke and use economerics ess for efficiency, exisence of profiable opporuniies in he marke and he impac of cenral bank inervenion. 4.1 Survey The survey of he foreign exchange marke was argeed a boh he commercial banks and he forex bureax. As per aached quesionnaire (Appendix 2), he survey inended o undersanding he evoluion of key characerisics of he foreign exchange marke in Uganda and also o provide informaion needed in explaining he economeric resuls. 4.2 Tesing for marke efficiency Fama (1970) deduced d ifferen forms of efficiency of a marke (i.e. weak form, semi-srong form and srong form efficiency) depend ing on how much informaion is used in forming expecaions abou he fuure price. In his sudy, we es for wo forms, he weak form and he srong form efficiency form, by he mehodology oulined below Weak form efficiency Tess for weak form marke efficiency have normally been based on he predicive power of he forward rae for he fuure spo rae (Swarna, 1994). The es o deermine wheher he forward rae is an unbiased predicor of he fuure spo rae in he Ugandan foreign exchange marke was herefore specified as follows: = α + β ln F + u (6.1) Where; S +1 is he hree-monh fuure spo rae is he calculaed forward rae a ime for delivery a ime (+1) β is he +1 relaionship beween S +1 and F +1 While he forward marke is no very acive in he Ugandan foreign exchange marke, he few agens ha underake hese ransacions base hem on he 111

114 ineres rae differenials beween Ugandan and foreign ineres raes 48 and hose of is foreign parners. The forward is herefore compued as: F =S x1+i h 1+i 0 Where i h is he ineres rae in Uganda I 0 is he foreign ineres rae. S is he spo rae a ime. (6.2) The es for efficiency in Equaion 6.1 is comprised in esing he null hypohesis α= 0 and β = Srong form efficiency For srong form efficiency of he marke, he lieraure suggess ha here should be coinegraion beween fuure spo and forward rae series. Like Swarna (1994), he Engle-Granger (1987) bivariae wo-sep coinegraion regression procedure is applied. The reverse regression from Equaion 1 above is specified as: +1 = α + β ln S +1 + u 1 (6.3) The direc regression specified in Equaion 6.1 and he reverse regression in Equai on 6.3 are conduced o deermine wheher he Ugandan foreign exchange marke is characerized by srong form efficiency. 4.3 Speculaion efficiency - exisence of profiable rading opporuniies Efficien marke hypohesis onl y holds as an exreme case of perfec risk free a rbirage ha has no pracical relevancy. Nowihsanding he argumens abou wheher foreign exchange markes are affeced by invesor senimens, we adop a model based on unpredicabiliy or randomness of changes in invesor senimens o assess noise rading in he Ugandan foreign exchange marke. The inuiion behind he model is ha if here exiss noise raders in he marke, hey would cause he exchange rae o overreac and hus yield excess reurn. Equilibrium reurns based on marke fundamenals reward only sysemaic risks Tesing for speculaive efficiency 48 For his sudy, he 91-day T-bill rae is used, while a hree monh US T-bill rae is used o proxy he foreign ineres raes. 112

115 We adoped he Ayogu (1997) approach ha esed for speculaive efficiency under a weak form efficien marke of Nigeria s parallel marke. Under his approach, we adop he efficiency hypohesis independen of is implicaions for raional expecaions. To parameerize he hypohesis, we have o model he process ha deermines evoluion of exchange rae in he IFEM. Based on he available lieraure, he choice of he model of nominal exchange raes o achieve his can be cumbersome. For he forecasing equaion, we choose o use a general specificaion ha allows for a random walk, wih or wihou a drif and for saionariy around a rend. Hence he nex period s exchange rae under he null hypohesis ha he marke is in weak efficien form can be specified as: i+1 = β i0 + βi1t + βi2s i + ε i+1 (6.4) Where S i = spo rae for Uganda n shillings per uni of foreign currency i for period i+1 = he random error erm T= he rend componen The null hypohesis ha one canno bea he marke assers ha Ho:( β i0 β i1 β i2 ) = (0,0,1) for all i s. The hypohesis, as specified, canno be esed for is disribuion because i is unknown and does no converge o a Gaussian process asympoically. We hus reparameerize he null hypohesis, following Dickey and Fuller (1981), ino a dynamic model of he form: i S = α + βt + θs i-1 + γ i i S -1 + ε (6.5) he unresriced model, and i i = γ S i S -1 + ε (6.6) he resriced model where θ = (β 2-1); i denoes he ih difference of S and e ~ iid (0,δ 2 ). The null of efficiency hus assers, Ho : (α,β,θ) = (0,0,0) and γ i = 0 (i=1,...k). This imposes he weaker maringale propery ha expeced changes in he ime series of he spo rae S are zero. We employ he likelihood raio es o deermine which model fis Uganda s foreign exchange marke. The es 113

116 saisics φ 2 (Dickey and Fuller, 1981) is he likelihood raio es of (β β β ) i0 i1 i2 = (0,0,1) agains he alernaive (β i0 β i1 β i2 ) (0,0,1). The saisics would be compued as he sandard F es of his disribuion and significance of g s assessed on he assumpion ha he asympoic disribuions of saisics are sandardized normal. The resul of he specificaions ess would guide us in he choice of forecasing model (i.e. resriced or unresriced). The forecasing equaion would hus be specified as: i+1 = β 0i + S i + γ ij i S i-j + ε i +1, i=1,2,...n (6.7) he unresriced model, or i = S + γ i+1 i ij S i-j + ε i +1, i=1,2,...n (6.8) he resriced model The error erm ε i in (6.7) or (6.8) above follows a complicaed process ha capures marke characerisics and hence can be used as a proxy for he risk premium for any invesor in ha currency marke. For a baske of currencies wihin a marke, here would be a combinaion of currencies ha minimizes risk and maximizes reurn Compuing reurns from noise rading For marke where he weak form efficiency (unbiasedness of forward raes in predicing fuure spo raes) hypohesis does no hold, i is argued ha here mus be evidence of eiher a risk premium, marke inefficiency or boh. Irrespecive of wha prevens he hypohesis from holding, agens can exploi such rading opporuniies. The mehodology used o compue for noise rading is developed on he basis of exisence of risk premium or excess reurns in he foreign exchange marke aribuable o changes in invesor senimens. Suppose, he expeced reurn on a paricular currency i a ime is ER i and he realized reurn on he currency wihin he same period is R i, hen we can define he excess reurn or premium, i as: ) = R ER = i i i S (6.9) i S i where S i = S i -S i-1 and informaion a ime. ) S i = ER I he expeced reurn using all available 114

117 The expeced reurn ) S would be calculaed posulaing ha if i arbirageurs exis in he marke, hey would adop heir pricing eiher according o he exising fundamenals in he marke or according o he forecasing equaion based on heir models. The expeced reurn can be spec ified as: ) S i f = S where S i f i -S f i-1 (6.10) = esimaed exchange rae a ime. Alernaively, a simple nonsaionary AR(1) process could be used based on he specificaion of he forecasing equaion above such ha ) S i i = α i + βs i-1 + ci S i-j + v i, i=1,2 (6.11) The proxied expeced reurn would hen be compared o he acual reurns from he marke, which may or may no include he speculaive aciviies. The null for no noise raders can be parameerized as: Ho: I = 0 And esed wih he Wald es of he exclusionary resricions Serial correlaion and seasonaliy While exisence of profiable opporuniies could also hinge on seasonaliy paerns, i is normally useful o augmen he above ess of efficiency wih ess for serial correlaion and seasonaliy o deermine wheher he expecaions of agens in he marke are regressive or heavily influenced by seasonaliy ha may lead o herding behavior characerisic of a small, one commodiy economy, like Uganda. In his es we wan o idenify a paern in foreign exchange rae movemens using Fama (1965) serial correlaion es. In he serial correlaion es, we run he following regression: R = α + βr -k (6.12) Where R is he reurn a ime Is he reurn a ime -k -k k is he lag in he number of ime periods, k=1,2,3,. β is he relaionship beween R and R -k The es for seasonaliy is done on boh he monhs of he year and days of he week. 115

118 4.4 Assessing he impac of cenral bank inervenion The behavior of he foreign exchange marke in reacion o cenral bank inervenion can be capured hrough he assessing volailiy around is rades. This manifess ino a es for abnormal volailiy around he cenral banks rades. Suppose reurns in he marke are specified as: S) 2 2 ( = R = α+ β D. (6.13) BOU Where: R = S = he reurn a ime = he dummy variable a ime BOU β p = he relaionship beween R i and cenral bank s presence in he marke, D BOU The dummy variable is mean o capure he cenral bank s presence in he marke akes a value of D BOU = 1, when he cenral bank rades and D BOU = 0, when i is no in he marke. Allowing for he effec of size of inervenion requires ha reurns are specified as: S = R i = α + βa i + e i (6.14) Where R i = he reurn a ime A i = he Bank of Uganda inervenion amoun a ime βs = he relaionship beween R and A i i Cenral bank sales or buys in he foreign exchange marke are expeced o have varian effecs. To ake direcion of rades ino accoun, reurns are specified as: S = R i = α + β 1 D1 + β 2 D2 + e i (6.15) Where R i = he reurn a ime D1 i = he dummy for Bank of Uganda sells a ime D2 i = he dummy for Bank of Uganda buys a ime β 1 = he relaionship beween R i and D1 i β 2 = he relaionship beween R i and D2 i Tess for he impac would be esing he significance of he coefficien on he dummy variables in he hree equaions relaing reurns o cenral bank inervenion; under he null Ho ha cenral bank inervenion has no effec on reurns specified as: 116

119 Ho: β p = β s = β 1 = β 2 = 0 And esed wih he Wald s es of coefficien resricions. V EMPIRICAL AND SURVEY FINDINGS This sudy followed he approach of Ell io and Io (1995) and used he field survey resuls o complemen he econom eric s udy in assessing he efficiency of he Ugandan foreign exch ange marke. The su rvey, inende d o collec some micro-informaion from he forex dealers, covere d 17 commercial banks, and 48 of h e 72 regisered forex bureaus in Kamp ala, represening 98.6 percen of he cumulaive forex ransacions over he period July - November The preceding analysis of field survey resuls was based on he responses from 14 b anks a nd 38 f orex bureaus. Informaion from he survey provided a good saring po in o our economeric sudy of efficiency of he Ugandan foreign exchange marke, bu also incorporaed sever al elem ens ha we re crucial in explaining some of he findings from he sudy. In he following subsecions, herefore, a synopsis of he characerisics of he foreign exchange marke as envisioned by he survey respondens is presened before he resuls of he economeric sudy are presened. 5.1 Survey response on characerisics of he foreign exchange rade Our survey resuls sugges ha considerable evoluion has been realized in he Ugandan foreign exchange marke since is liberalizaion in The changes have no only been winessed in he medium of rade, bu also in he producs and caegories of paricipans in he marke. Trade in he Uganda s foreign exchange marke ha was originally concenraed in cash ransacions has evolved and was, by 2001, almos equally disribued beween cash and elegraphic ransfers (TTs). As shown in Table 6.1, banks repor ha in 1994, up o an average of 29 percen of heir oal rade was ransaced hrough cash, bu his mean rose o 37.2percen in 2001, while ha for TT regisered an average of 17.8 percen in 1994 and has now risen o 46.7 percen. Noable reducion in he use of Traveler s cheques, drafs and leers of credi was repored over his period. Table 6.1: Average conribuion of differen forms of rading in he foreign exchange marke Trading Form Min Max Mean Min Max Mean Min Max Mean 117

120 Banks Cash Traveler s cheques Drafs Leer of Credi Telegraphic Transfers Bureaux Cash Traveler s cheques Drafs Leer of Credi Telegraphic Transfers Source: Auhors On he oher hand, he regulaions require Bureaux limi heir rade o over he couner ransacions, hence he bulk of heir ransacions are in cash. This nowihsanding only 21 ou of 38 bureaus surveyed indicaed ha 100 percen of heir ransacions were in cash by end of Up o 4 bureaus repored ha over 20 percen of heir ransacions are on accoun of raveler s cheques while 5 repored ha over 20 percen of heir ransacions are elegraphic ransfers. The average conribuion of cash and elegraphic ransfers increased from 70.6 percen and 1.3 percen in 1994 o 84.6 percen and 8.5 percen respecively by For caegories of counerpars in he foreign exchange business, in spie of he considerable increase in he ransacions volume in he iner-bank and iner-bureau business, he conribuion of hese caegories o oal foreign exchange marke has declined from a mean of 28.5 percen and 30 percen for iner-bank and iner-bureau respecively in 1994, o 18.7 percen and 16.7 percen respecively in This loss has been on accoun of he increase in cusomer business ha rose from 59.0 percen o 66.2 percen over he same period. Likewise, he bureaus cusomer ransacions ha accouned for 70 percen in 1994 peaked o an average of percen by end of Boh banks and bureaux repored heir inra-ransacion o have increased. Table 6.2 summaries hese developmens. Table 6.2: Caegorizaion of foreign exchange ransacions as repored by banks Min Max Mean Min Max Mean Min Max Mean Banks 118

121 Iner-bank business Bureau bank business Cusomer business Oher nonbank/bure aux Bureau x Inerbureux business Bureau bank business Cusomer business Oher nonbank/bureaux Source: Auhors On he c urrencies of rade i n his m arke, while he bulk of foreign exchange ransacions were conduced in Unied Saes dollars during h e period , here has been widening currency rade. The conribu ion of he US dollar declined from 76.7 percen o percen beween 1994 and 1997 before slighly recovering o 73.6 percen in As indicaed in Table 3 below, he pound serling followed a similar rend over he period being recorded a respecive levels of 18.3 percen, 14.8 percen before rising o 15.5 percen in The overall decline over he period for hese wo currencies was a he benefi of he Kenya shilling, which has increased in imporance as a regional currency, and he Deuchmark ha rose from 1.7 percen o 7.1 percen beween 1994 and Like banks, he bureaus mainly rade in US dollars. By 2001, hey repor ha US dollar rade accouned for on average 79.1 percen of oal rade, while some even depic maximum conribuion of 98.1 percen. Oher currencies, he UK pound serling, he Deuchmark and he Kenya shilling sood a averages of percen, 6.0 percen and percen respecively. Table 6.3: Common currency of foreign exchange business in Uganda Currency Min Max Mean Min Max Mean Min Max Mean Banks 119

122 US Dollar UK pound serling Deuch mark/euro Kenya shilling Bureaux US Dollar UK pound serling Deuch mark/euro Kenya shilling Source: Auhors One imporan facor o deermine how he m arke operae s is informaion abou he way agens deermine heir ransacions. A larger number of boh banks and bureaux descri be heir spo ransacions as being driven by cusomer orders. Bu also, echnical rading rules is indicaed as anoher imporan deerminan of heir ransacions. As shown in Table 6.4, in 2001 for insance, 7 banks ou of 14 indicae ha heir spo ransacions are based on echnical rading rules, compared o 23 ou of 38 for bureaux. However, cusomer orders ake a voing of 10 banks ou of 14 and 27 bureaux ou of 38 over he same period. Table 6. 4: Basis for he spo ransacions Basis Banks Bureaux Technical rading rules 3/14 6/14 7/14 4/38 7/38 23/38 Fundamenal analysis 1/14 1/14 1/14 1/38 3/38 10/38 Driven by cusomer orders 3/14 6/14 10/14 1/38 6/38 27/38 Source: Auhors From he survey resuls he spreads beween he bid and ask raes are higher in banks han in bureaux on boh he US dollar and he UK pound serling marke. In he US $ marke, he average spared beween he buy and ask rae sood a Shs Compared o he spread beween he US $ raes in he bureaux, of Shs , his is much higher. The repored average of spreads for oher currencies sands a Shs for he UK pound serling, Shs for he Deuchmark and Shs 1.67 for he Kenyan shilling for banks. As seen in Table 6.5, he comparable averages sood a Shs , Shs and Shs for he bureaux. Table 6.5: Average spreads in he foreign exchange marke (Ushs.) Banks Bureaux Min Max Mean Min Max Mean US Dollar UK Pound Serling

123 Deuch Mark/Euro Kenya Shs Source: Auhors Wih a rich background on he characerisics of he Ugandan foreign exchange marke, we proceeded wih he empirical sudy o deermine efficiency or inefficiency of he Uganda foreign exchange marke and exisence of profiable reurns in he marke. Table 6.6: Average spreads in he foreign exchange marke (Ushs.) Banks Bureaux Min Max Mean Min Max Mean US Dollar UK Pound Serling Deuch Mark/Euro Kenya Shs Source: Auhors Wih a rich background on he characerisics of he Ugandan foreign exchange marke, we proceeded wih he empirical sudy o deermine efficiency or inefficiency of he Uganda foreign exchange marke and exisence of profiable reurns in he marke. 5.2 Forms of marke efficiency Some economeric and survey hins Though he marke rades in several oher currencies, complee series on he nominal exchange raes 49 were only available for he Uganda shilling per US dollar rae. Moreover, mos of he oher currencies are priced hrough cross raes, which do no necessarily reflec marke characerisics wihin he Ugandan marke. The economeric sudy herefore concenraes on he US $ currency. Daily observaions of foreign exchang e weighed-average midraes for he period January 1995 o December 2001 were used. The exchange raes are quoed as he Uganda shilling value or price of one US Dollar (normally referred o as a direc quoe). As indicaed in Secion above, he es for weak form efficiency manifess in esing for he significance of he coefficiens in Equaion (6.1) under he null hypohesis ha α =0 and β= 1. The forward rae F was derived from he ineres raes differenials beween he Unied Saes and Uganda. As per resuls in Appendix Table 1, based on monhly daa for he period May 1998 o December 2001, Equaion (6.1) is represened as: ln S+3 = * ln F+3 (R1) (18.35) 49 The Research Deparmen of Bank of Uganda compiles he daily exchange raes beween he Uganda shilling and he US $, he mos common currency. This was he source of he daa for he empirical analysis. 121

124 (4.49) Values in brackes are -values of he regressors. A 95 percen level of confidence, we rejeced he null hypohesis of he forward rae being an unbiased predicor of he fuure spo rae since β 0. Tha he unbiasedness hypohesis does no hold in he Uganda foreign exchange marke is eviden of eiher exisence of a risk premium and/or ha he marke is no efficien. On srong form efficiency, he ess performed were geared owards esablishing wheher here is co-inegraion beween he spo and forward raes in Uganda (in line wih he wo-sep bivariae mehodology proposed in Secion 4.2.2). As per resuls summarized in Appendix, Table 6.2, he search for a co-inegraing vecor yields sugges ha he long run elasicies can be represened as: ln S +3 = * ln F +3 (R2) The Engle Granger wo-sep ransformaion of he resuls of equaion R2 above ino he forma suggesed in Equaion 3, i.e. ln F = a + bln F is presened in Appendix 2, bu can be summarized as: ln F+3 = * ln S *ln S-2 (00.79) (0.19) (-0.02) 0.85* lnf * lnf * ecm (R 3) (-9.30) (-0. 84) (14.06) 2 where value in brackes are -values, R = 0.86, D.W saisic = 0.77, serial correlaion F(12,20) = 5.075, funcional form F(1, 31) = and Heeroskedasciy F(1,36)= The resul in R3 suggess ha no only are changes in he spo exchange raes insignifican in explaining changes in he forward exchange rae, bu also ha he model is characerized by error exacerbaion as shown by he sign on he adjusmen co-efficien aached o he error correcion erm (ecm) 50. This could herefore mean ha he esablishmen of he co-inegraing vecor limis he relaionship of he fuure spo and forward exchange rae o one ha is weakly exogenous and evidence of srong exogeneiy is no suppored given he failure of he dynamic equaion o rejec he insignificance of he changes in he spo exchange raes in explaining forward raes. Even if he forma of Equaion 6.1 were used for esing he srong form of he marke, he resuls in Appendix Table 6.2 do no suppor exisence coinegraion beween he spo rae and he forward rae, given ha boh he 50 This is also confirmed by he failure of he serial correlaion es a 5% significance level 122

125 error erm and he changes in he forward rae are insignifican in explaining he changes in he fuure spo exchange rae. The implied inefficiency of he Ugandan foreign exchange marke is furher confirmed by exisence of a few big players dominaing he marke, also re vealed by our survey resuls. By he percenage of respondens who hough ha a few big players dominae he marke, his phenomenon increased dramaically beween 1994 an d Marke paricipans revealed ha his phenomenon is mos predominan in he US dollar marke - wih 12 ou of 14 banks indicaed a few big players do minae his marke. For UK pound serling marke, only 8 ou of 14 banks indicaed h a here i s his dominance and only 1 ou of 14 banks indicaed he same for he Kenya shilling marke. The same message is brough ou from he responses by bureaux, for which 29 ou of 38 bureaux, 8 ou of 38 bureaux and 5 ou of 38 bureaux indicae ha he hree currency markes are dominaed by a few big players (see Table 6.7) Table 6.7: Exisence of a few big players dominaing he marke - average posiion on ranking of 1-10 Banks Bureaux US Dollar 4/14 6/14 12/14 6/38 12/38 29/38 UK Pound Serling 2/14 3/14 8/14 4/38 4/38 8/38 Kenya Shilling 1/14 1/14 1/14 2/38 2/38 5/38 Source: Auhors The inerbank marke herefore is characerized by concenr aion in he few big banks ha have he compeiive edge in foreign exchan ge business. These accoun for abou 60 per cen of he IFEM ransaci ons and have a significan influence on he movemen of he exchange rae. Furher sill, while business sense would dicae ha he bid-ask rae ake cos as he mos imporan facor, many respondens are comforable following he marke leaders. Indeed, when asked abou facors deermining he bid-ask raes, only a few banks/bureaux ha indicaed ha he poenial cos of a quoe is imporan for deermining he spreads as opposed o following he marke convenion 51 or generally acceped levels of spreads (Table 6.8). The reason for choosing o use he generally acceped sandard is o secure a good 51 In his paper, we use he erm marke convenion o refer o generally acceped levels/sandards hough no necessarily agreed upon like i is normally defined when referring o more developed and organized markes. regulaion and bureaux policy. Securing a good marke image indeed comes in wih he wors rank of 3.71 h. 123

126 marke image. As is shown in Table 6.9, he facor secure a good marke image was indeed on he bes score of 1.6 h rank, on a scale of 6 facors proposed o he marke as possible reasons for he acions hey underake in he foreign exchange marke. Table 6.8: Facors deermining he Bid-Ask rae Banks Bureaux Marke convenion 5/14 7/14 11/14 13/38 12/38 37/38 Poenial cos of making quoe 1/14 1/14 1/14 2/38 3/38 2/38 Source: Auhors No surprising, oher imporan facors for conforming o marke convenion include rade regulaion and mainenance of an equiable and reciprocal rading relaionship wih ohers. The rankings are summarized in Table 8. For he bureaus, while marke convenion is voed highly han poenial cos of making he quoe, hey on average, indicae ha heir mos imporan consideraions in sicking o marke convenion are following rade. Table 6.9: Mos imporan facors for conforming o marke convenion (Average posiion on ranking of 1-6) Facor Banks Bureaux Bank/bureau policy 3.50h 2.11h Mainaining an equiable and reciprocal rading relaionship 2.25h 2.44h Securing a good marke image for he bank 1.60h 3.71h Maximizing rading profis 2.33h 3.42h Following he pracice of major raders 3.00h 3.47h Following rading regulaion 2.20h 2.00h Oher(specify) Source: Auhors The above resuls sugges ha in spie of he increased deph and produc diversificaion in he Ugandan foreign exchange marke, he marke has no ye achieved a significan degree of efficiency. Marke paricipans have always suggesed he poor flow of informaion as he main hindrance o efficien operaion of he marke. Recen developmens, however, poin o srenghened links beween he domesic securiies marke and foreign exchange marke. Raes of reurn on domesic securiies and he expeced exchange rae depreciaion now o a large exen deermine he allocaion of savings beween shilling asses and foreign currency asses. In a bid o promoe informaion flow, a leas regarding he raes of dealing, in July 2001, BOU direced all auhorized dealer banks o begin using he Reuer Sysem Uganda Invesmen Services (UIS), o pos heir indicaive buy and sell raes. I is hese raes, which BOU now uses in is ineracions wih he marke. Since he inroducion of he posing of indicaive raes on he UIS screen, informaion flow has improved. Ranges wihin he indicaive ask and offer raes have narrowed while he spreads have also come down dramaically. The spreads only end o increase when here are cases of insabiliy in he 124

127 marke. In general, here now appears o be a quicker response o he cenral bank s signals when i inervenes in he marke as opposed o wha was he case five years a go. 5.3 Speculaion efficiency exisence of profiable rading opporuniies Given he inefficiency of he marke indicaed above, he quesion ha arises now is wheher any profiable rading opporuni ies are available in he marke for agens o make abnormal profis, eiher hrough speculaion or noise rading Speculaive efficiency As sipulaed in Secion 4.2.1, he ess for he null of one canno bea he marke (Ho: (α,β,θ) = (0,0,0) and γ i = 0 (i=1,...k)), was examined under boh he resriced and unresriced models (see Appendix Table 3 and 4). The exchange rae equaion is summarized as: i S = e-06*T 0.006*S i * S * and is i 49* S S = 0.170* he resriced model he unresriced model, i The Wald es resricions on he coefficiens of he above wo models are summarized in Table 6.10, below. i -1 S -2 i (R4) (R5) Table 6.10 Resuls of he Wald Tes on resricions of Coefficiens in model R4 ad R5 Unresriced model (R4) Resriced model (R5) Null Hypohesis: Null Hypohesis: α = 0 125

128 β = 0 γ 1 = 0 θ = 0 γ 2 = 0 F-Saisic = Probabiliy = F-Saisic = Probabiliy = Chi-Square = Probabiliy = Chi-Square = Probabiliy = Conclusion: rejec null hypohesis Conclusion: rejec null hypohesis Source: Auhors As suggesed by he resuls of Wald resricion on he coefficiens α=0, β=0, θ=0, he null of one canno bea he marke was rejeced a 5 % level of significance. This indeed suggess ha here are profiable opporuniies in he Ugandan foreign exchange marke because changes in he exchange rae are no expeced o be zero. To es wheher agens in he marke ake advanage of hese profiable rading opporuniies, we aemped o deermine a wha reurn agens minimize risk or wha risk agens maximize profis. Due o daa limiaions on oher currency invesmens, his analysis was limied o he US$ currency. The reurns were compued as: R ( s = e ) * 100 (6.16) log s 1 Where R is he reurn a ime S is he exchange rae a ime S -1 is he exchange rae a ime -1 This assumes ha an invesor holding Unied Saes dollars makes a posiive reurn if he Uganda shilling depreciaes agains he dollar. In such a case, he invesor would sell one dollar for more Uganda Shillings han he price a which he bough i. Boh hese models depic a heavy concenraion of risk a zero, for which he bes fi of a relaionship is a horizonal sraigh line. This implies ha higher risk would no be rewarded wih higher reurn in his marke. On inerpreaion of his is ha, in spie of he abnormal profi opporuniies, he marke has in general been very risk averse. Hence, while he null of one canno bea he marke is rejeced, our analysis suggess ha agens do no make abnormal profis hrough speculaive rading. This again is probably on accoun of he phenomenon of having a few big players who lead he marke, while oher smaller players follow. Resricing our analysis o he backward looking model above would however ignore he behavior of hose agens in his marke who use he forward premium o price foreign exchange forward. Indeed, wih he forward-rae-forecasing model, he risk/reurn fronier depics some degree of elasiciy of reurn o risk, as depiced in figure 6.1 below. The char shows 126

129 ha he agens would demand for beer compensaion (higher reurn) if hey have o ake on any addiional risk in heir invesmens. Figure 6.1 Risk versus reurn (forward rae based model) Source: Auhors Reurns from noise rading The es for exisence of reurns from noise rading (secion above) ma nifess in comparison of reurns from he forecased exchan ge rae wih he acual reurn in he mark e. As summ arized in Table 6.11 below, he null hypohesis of non-exise nce of noise raders reurns was no rejeced. This implies ha in spie of he exisence of profiable opporuniies in he marke, rading has no realized any reurns even from noise rading. The deailed resuls for boh he Unresric ed and Resriced mo dels are presened in Appendix Tables 6.8 and 6.9. Given he characerisics of he Ugandan IFEM, adapive expecaions based on hisorical paerns and seasonaliy should be an imporan facor in undersanding rading behavior. Table : Tess for reurns in Noise Trading he Ugandan IFEM Equaion esed Null Decision Hypohesis Unresriced model Resriced model i S^ = S + Ho: = 0 Ho: = 0 i S^ = S + Noe: Decision based on confidence inerval of mean of residuals. Did no rejec null => no reurns Did no rejec null => no noise Noise rade reurns could herefore be manifesed as a correlaion of previous day s reurns or as a seasonal facor. In line wih Secion IV.II.III above, (i.e. he auo-regressive process AR(1, 2.., n)) we invesigaed wheher any rading opporuniies could be earned using daily, weekly and even monhly cyclical/seasonal paerns observed in he paern of he exchange rae. 127