MPRA Munich Personal RePEc Archive The real exchange rae of he rand and compeiiveness of Souh Africa s rade Elvis Monga Universiy of Cape Town 15. December 26 Online a hp://mpra.ub.uni-muenchen.de/1192/ MPRA Paper No. 1192, posed 16. December 26
The rreall exchange rrae off he rrand and compeiiiiveness off Souh Affrriica s rrade Elvis Monga 1 School of Economics, Universiy of Cape Town 15 Augus 26 Absrac I n he las 1 years since Souh Africa ransformed ino a democracy, he rand has seen an increase in volailiy of is real exchange rae. These flucuaions in he rand s real exchange rae have raised quesions as o wheher hey signify significan misalignmen of he currency and hereby undermine compeiiveness of Souh Africa s expors abroad. This is a perinen quesion in he Souh African conex because foreign rade has been criical o he growh of he economy. Effors o address curren high levels of unemploymen and widespread povery among he majoriy of he populaion have depended upon his growh. This sudy invesigaes he exen o which flucuaions in he rand s real exchange rae have impaced on he compeiiveness of Souh African rade flows by deermining wheher, a some poin, he rand had been misaligned, and he likely consequences of such a misalignmen. Using daa from 1972 o 23, and an equilibrium correcion model of he rand s real exchange rae drawn on exising lieraure, he sudy finds ha, from 1994 o 1996, and also in 1998, he rand s real exchange rae became undervalued by an average 1%. By early 22, he exen of overshooing had reached 2%. However, he srong recovery of he rand a he sar of 22 reversed his overshooing and insead pushed he real exchange rae above is equilibrium by an average 16 o 17% a he end of 23. This suggess significan loss of rade compeiiveness during 23 and needed a nominal depreciaion o correc he imbalance. 1 I hank very sincerely Prof. Melvin Ayogu for very useful and consrucive commens received on an earlier draf of his paper. Financial assisance from he Economic Policy Research Insiue (EPRI), Cape Town, is also graefully acknowledged.
Ta b l e o f c o n e n s T a b l e o f c o n e n s... i Lis of figures... ii Lis of Tables... iii 1 Inroducion... 1 2 Mehods for assessing real exchange rae misalignmens... 5 2.1 The purchasing power pariy approach... 6 2.2 The macroeconomic balance approach... 8 2.3 The behavioural equilibrium exchange rae (BEER) approach... 1 3 Modelling he rand s equilibrium real exchange rae and assessing he exen of misalignmen... 11 3.1 Facors ha deermine real exchange raes... 14 3.2 Empirical Applicaion... 18 3.2.1 Specificaion of he empirical model... 18 3.2.2 Tesing mehodology... 19 3.2.3 Daa definiions and sources... 26 3.2.4 Empirical resuls... 31 4 Conclusion... 5 References... 52 i
Lis of figures Figure 1-1: Real Effecive Exchange Rae for he rand: 197 o 25 2 Figure 3-1: Souh Africa: fundamenal deerminans of he real exchange rae 3 Figure 3-2: Real exchange raes of he rand (model 1B) 44 Figure 3-3: Real exchange raes of he rand (model 2A): 1972-23 45 Figure 3-4: Real exchange raes of he rand (model 4A): 1985-23 46 Figure 3-5: Real exchange raes of he rand (model 4C): 1985-23 47 Figure 3-6: Real exchange raes of he rand (model 5B): 1985-23 48 Figure 3-7: Real exchange rae of he rand (model 5C): 1985-23 49 ii
Lis of Tables Table 3-1: Resuls of Dickey-Fuller s uni roo ess for saionariy properies of he daa...32 Table 3-2: Chi-squared saisics for join VEC lag exclusion Wald es...36 Table 3-3: Chi-square saisics for join ess of VEC model diagnosics...36 Table 3-4: Resuls of Johansen s coinegraing es...37 Table 3-5: Resuls of esimaing co-inegraing relaionships among he variables (1972q1-23q4 and 1985q1-23q4)...41 iii
1 Inroducion M anaging he exchange rae has remained a he cenre of Souh Africa s effors o achieving and mainaining boh macroeconomic sabiliy and a susained increase in he growh rae of he economy, ever since he counry ransformed o democracy in 1994. In his period, he problem of high and rising unemploymen, combined wih widespread povery, emerged as he mos pressing challenge facing he economy, mainly because of he counry s hisory of segregaion under Aparheid. Economic policy responses o his unemploymen challenge have aimed a aaining a susainable growh rae of he economy, which is seen by he governmen as a basis for job creaion. However, he growh performance of he Souh African economy has proved insufficien o address he unemploymen problem, wih he counry experiencing a conracion in oupu for he mos par of he early 199 s, going by he growh in real gross domesic produc per capia, which accouns for populaion growh in measuring he growh rae an economy. Though a rebound in economic growh has occurred since he mid-199s, his has been oo low, peaking jus above 2% only in 1996, 23, and 24. As he Souh African economy is highly dependen on he global economy, foreign rade has been criical o growh of he economy 2. Sabiliy of he real exchange rae is hus seen by he Governmen as key o aainmen of is growh objecives, since his no only impacs on global and domesic demand for Souh Africa s producion bu on domesic prices also. Being a relaive price of domesically produced goods relaive o foreign goods, he real exchange rae provides a measure of how much, on average, he cos of Souh African produced goods and services will cos relaive o a comparable baske of foreign produced goods and services, and herefore of heir compeiiveness overime. More so ha expors reflec foreign demand of Souh Africa s producion, heir increase means ha all oupu, employmen, and incomes associaed wih heir producion may increase overime and hereby conribue o he growh rae of he 2 Foreign rade flows now accoun for more han 3 percen of Gross Domesic Produc 1
economy. Unwarraned flucuaions in he real exchange rae, hus, end o be derimenal o he compeiiveness of expors. As such, he Governmen s growh and employmen objecives may be undermined. However, he rand s real exchange rae has experienced considerable volailiy. As is eviden in figure 1-1, which depics he evoluion of he rand s real effecive exchange rae from 197 hrough o 25, flucuaions in he real exchange rae have, a imes, been large and persisen for several monhs. Afer appreciaing srongly in he 197s by 5 percen from 1972 o1983, he rand collapsed in he early 198s by 47percen over a space of jus wo years beween 1983 and 1985. Subsequenly, he rand rebounded by 44 percen beween 1985 and 1994, bu weakened again by 41 percen beween 1994 and 21. By December 23, he level of he real effecive exchange rae index represened an appreciaion of more han 5% over is December 21 level. Figure 1-1: Real Effecive Exchange Rae for he rand: 197 o 25 18 Real effecive exchange rae of he rand (Index: 1995=1) 18 16 16 14 14 12 12 1 1 8 8 6 7 72 74 76 78 8 82 84 86 88 9 92 94 96 98 2 4 Source: I-Ne Bridge 6 These flucuaions in he rand s real exchange rae have raised quesions as o wheher hey indicae misalignmen of he currency. The real exchange rae is, 2
convenionally, regarded as misaligned when i s observed values exhibi persisen deparures from is long run equilibrium rend, which is he value consisen wih long run rends in economic fundamenals (IMF, 1998). In consequence, a misaligned real exchange rae is labelled overvalued when is value is higher (appreciaes more) han is long run equilibrium value, and undervalued once i falls (depreciaes) below he equilibrium value. Thus, o make an assessmen of real exchange rae misalignmen, a measure of he equilibrium real exchange rae is ofen required as his forms a benchmark agains which he acual or observed real exchange rae can be gauged (Maeso-Fernandez, e al, 21). This quesion of wheher or oherwise he rand s real exchange has, a any poin in ime, been misaligned is perinen o Souh Africa given, as already noed, he imporance o he economy of foreign rade and capial flows. In such an environmen of greaer openness of he economy, misaligned real exchange raes are known o have he poenial o impose large coss o he economy, in he sense of having an adverse effec on resource allocaion in he economy. When a counry s real exchange is overvalued, he implicaion is ha domesic goods become more expensive relaive o foreign goods, since i reflecs he price of domesic goods relaive o foreign goods. This has he poenial of increasing he demand for foreign goods a he expense of domesically produced goods, and as a resul, here will be considerable scope for impors o increase, while he producion of domesically produced impor subsiues is more likely o decline. Because of he increase in he cos of producing hem, and he associaed loss of profiabiliy, expors will be harmed. Thus, overime, a counry s exernal posiion will end o deeriorae, and rising unemploymen may accompany his. On he oher hand, a real undervaluaion means ha domesic goods are becoming cheaper relaive o foreign goods, hereby making hem more compeiive. As a resul of he likely shif in demand in favour of domesic goods, here will be opporuniies o increase he producion of domesic impor subsiues, hereby giving rise o a siuaion of a faser growh of expors. The growh of impors will also end o slow down and, 3
overime, here will be endency for counry s balance of paymens o improve, as well as he employmen siuaion. The adverse effecs of real exchange rae misalignmen are, hus, well recognized in he lieraure. However, he significance of hese ofen differen and complex effecs of changes in he real exchange rae will, in many insances, depend of he economic condiions prevailing in he counry, such as wheher he economy is in recession or boom (De Kock Commission of Enquiry, 1985). Much also depends on he prevailing sance of oher policies, such as fiscal and moneary policies, which may srenghen or weaken he ransmission of hese exchange rae effecs o res of he economy. While his may seem o sugges ha he adverse impacs of real exchange rae misalignmen are no auomaic, an undersanding of he exen of real misalignmen is noneheless imporan, as if such misalignmens were o be large, his may serve o sugges he likely coss ha he economy is likely o face in he fuure, if no adjusmens o domesic policies are made. Thus, we invesigae, in his sudy, he exen o which pas movemens of he real exchange rae of he rand have impaced on he compeiiveness of he Souh African rade flows by deermining wheher a some poin he rand had been misaligned or oherwise, and he likely consequences of such a misalignmen. Because misalignmen is wih reference o some benchmark equilibrium measure, we idenify he variables ha are fundamenal in valuing he rand, which we hen use o esimae a long run model of he rand s real exchange rae. Based on he findings of he long run real exchange rae model, we derive a long run equilibrium real exchange rae value for he rand, which we compare wih he observed real exchange rae so as o make judgmen on wheher or no he rand has been misaligned. There already exis oher sudies ha have empirically examined he deerminans of he rand s real exchange raes and heir implied misalignmens. Aron, Elbadawi and Khan (1997, 2) esed for long run deerminans of he rand s real exchange using daa for he period 197 o 1995 and found he erms of rade, he gold price, official reserves, governmen expendiure, long erm capial flows, and degree of rade 4
resricions o have been he main deerminans. However, he sudy did no probe he issue misalignmen in he real exchange rae. Using daa from 197 o 22 and a somewha smaller se of explanaory variables, MacDonald and Ricci (23) also esimaed an equilibrium real exchange rae model for he rand, which hey used o measure he degree of misalignmen. Their analysis showed ha, in early 22, he rand had oversho is equilibrium value by a leas 25%. Similarly, Samson e al (23a, b) found ha he rand s real exchange rae had become significanly undervalued in 22, following a prolonged period of considerable depreciaions during 21. Previous sudies are hus confirmaory of he imporan of economic fundamenals in providing useful informaion o undersanding movemens in he rand s real exchange raes. Our sudy adds o his lieraure in wo ways. Firsly, we es for he significance of addiional variables and employ a somewha differen specificaion. Secondly, we use a large sample and es our model across wo differen sample periods. In erms of organisaion, he res of he discussion has been srucured as follows. Secion 2 examines he lieraure on he analyical and mehodological framework conneced wih esimaion of equilibrium real exchange raes and assessmen of real exchange rae misalignmen. Nex, in secion 3, we discuss he empirical approach adoped by he sudy, beginning firs wih a discussion of he heoreical model, and hen he economeric mehodology used o analyze he relaionship beween he real exchange rae and a se of economic fundamenals considered and he empirical resuls. Finally, we conclude he discussion, in secion 4, wih a presenaion of a summary of main findings. 2 Mehods for assessing real exchange rae misalignmens E ven hough, concepually, real exchange raes can be precisely defined as prices of home goods relaive o foreign goods, heir empirical examinaion poses challenges. The basic problem is ha hese inernaional pricing relaionships are pracically 5
unobservable and have o be calculaed based on an analyical framework 3. This poses he difficul ha, in he lieraure, one finds muliple analyical frameworks, which give rise o differen empirical values for he real exchange rae o choose from. On one hand, economiss have focused on he so called exernal real exchange rae, which is derived as a raio of he home o he foreign value of a broad based price index, where boh prices are denoed in he same currency (Hinkle & Moniel, 1999). However, maers are complicaed by he fac ha here exis muliple price indices o choose from, wih he implicaion ha here are as many empirical real exchange raes as here are price indices (Mark, 21). On he oher hand, he real exchange rae has been defined as a relaive price of radable goods in erms of non-radable goods, and his is labelled he inernal real exchange rae (Hinkle & Moniel, 1999). Even wih his approach, mulipliciy of concepual definiions give rise o several esimaes of he real exchange rae, hus raising he quesion as o when i migh be appropriae o use one raher han he oher. Noneheless, subjec o hese concepual and empirical ambiguiies, hree alernaive mehods have been widely used o assess misalignmens in real exchange raes. 2.1 The purchasing power pariy approach The purchasing power pariy (PPP) approach is he basic mehodological framework ofen used for he empirical examinaion of real exchange misalignmens. This approach is based on he law of one price, which holds ha if markes are compeiive and here is free rade of goods beween hem, he same baske of goods and services should command he same price across all markes when expressed in a common currency. To he exen ha he law of one price held rue, he implicaion is ha he nominal exchange rae ha ensures convergence of prices of a common baske of goods and services beween wo counries consiues an equilibrium value for he nominal exchange rae, which is labelled he PPP-exchange rae (Clark, e al, 1994). 3 See Hinkle and Moniel (1999) for a more deailed discussion of his 6
If for some reason he marke exchange rae significanly differed from he PPPexchange rae, hen profis can be made by buying he common baske of goods in he low price marke and selling i in he high price marke, a siuaion ha will evenually end o push he exchange rae back o is PPP equilibrium value overime. In consequence, PPP predics ha he real exchange rae remains unchanged overime, saring from a posiion of equilibrium in a base year when a counry s balance of paymens posiion was judged o be susainable (Inernaional Moneary Fund, 1998). Assessing real exchange rae misalignmen under he purchasing power pariy approach hen involves a comparison of prices of a baske of goods produced by he home counry wih ha of a comparable baske abroad and calculaing he exchange rae ha would equalize hem, relaive o a base year in which he balance of paymens was judged o be susainable (Wren-Lewis, 23). If is curren rae is higher or lower relaive o is level in ha base year, hen he real exchange rae is misaligned. However, as is also eviden in figure 1-1, he volailiy in real exchange raes has cas doubs over he validiy of he purchasing power pariy approach. More ofen researchers have empirically observed large and persisen flucuaions in real exchange raes since he adven of free floaing exchange rae regimes in 1973, for example, Isard (1995). Oher evidence agains he purchasing power pariy approach has come from a large empirical lieraure esing is validiy, which finds he approach an empirical failure and ha deviaions of real exchange raes from heir implied PPP values ake an average 4 years o be reduced by half (see Froo & Rogoff, 1995). These empirical observaions of large and persisen deviaions in real exchange raes have been aken as a suggesion ha oher facors may be a play, paricularly in he shor and medium run periods, which he approach ignores, as aenion is focused only on price and inflaion differenials. Consequenly, consensus has emerged in he economic lieraure ha he purchasing power pariy approach is no adequae o measure and esimae equilibrium real exchange raes. A bes, he approach provides only a parial approximaion of equilibrium real exchange raes, and herefore, of he exen of heir misalignmens. 7
2.2 The macroeconomic balance approach The empirical criicisms of he purchasing power pariy approach discussed above have led Williamson (1985, 1994a, b) and Isard and Faruqee (1998) o advocae he macroeconomic balance approach o assessing misalignmens in real exchange raes 4. This enails calculaing he equilibrium real exchange rae as he rae ha simulaneously yields he economy s inernal and exernal equilibrium. Exernal equilibrium is aken o mean achievemen of a susainable curren accoun defici, meaning one ha can be financed wihou undue foreign borrowing or an unnecessary loss of foreign reserves (Inernaional Moneary Fund, 1998). Inernal balance, on he oher hand, is defined as he level of oupu corresponding o full employmen and low inflaion or macroeconomic sabiliy. As he equilibrium real exchange rae depends on inernal and exernal balance, i will change once his posiion changes. The equilibrium exchange rae is, herefore, hough of as a range of equilibrium exchange raes raher han a single number and changes overime consisen wih flucuaion of is fundamenal deerminans (Inernaional Moneary Fund Insiue, 1998). In addiion, he approach incorporaes explicily he ime horizon in is noion of equilibrium. This implies ha he real exchange rae should be considered o be in equilibrium only in he conex of economic fundamenals deermining inernal and exernal equilibrium over a given ime frame (Hinkle & Moniel, 1999). This follows because economic fundamenals ha may be relevan for achievemen of inernal and exernal equilibrium a say a shor run period may differ from hose ones wih he mos influence over a much longer ime horizon. Thus, in pracical applicaions, researchers ofen selec a differen se of economic fundamenals defining boh inernal and exernal equilibrium, and herefore a differen value for he exchange rae, depending on he ime horizon under invesigaion (Clark e al, 1994). In his respec, he macroeconomic balance represens a major advance over he purchasing power pariy approach. 4 For an exensive discussion of his approach, see Clark e al (1994) and Isard and Faruqee (1998) 8
However, i has is drawbacks also. One is ha he approach is a mehod for calculaing equilibrium real exchange raes, raher han a heory of exchange rae deerminaion and does no herefore embody heoreically esable predicions. The oher relaes o he fac ha hough he macroeconomic balance provides a precise descripion of inernal and exernal equilibrium, and herefore of he equilibrium real exchange rae, his is difficul o deermine in pracice. When exacly an economy s macroeconomic equilibrium prevails in pracice can no be easily discerned. For his, one would require a macroeconomic model ha accuraely capures relaionships defining curren accoun balance, full employmen level of oupu, and inflaion. Bu, he pracicaliy of consrucing such a model is ofen hampered by resource consrains. Thus, in empirical applicaions, researchers have aemped o overcome he laer problem by employing heoreical models designed o seek a se of economic relaionships describing an economy s oupu and balance of paymens posiion, and where he value of underlying deerminans have been se a heir full employmen or susained levels (Clark & MacDonald, 1998). The real exchange rae calculaed from such an exercise is labelled he fundamenal equilibrium exchange rae (FEER), reflecing he emphasis on fac ha only hose facors judged o be imporan over he medium o long erm period should be considered in is derivaion, absracing from ransiory facors. Applying his approach, some researchers have used large economeric models 5, for example, Williamson (1994b), who used several macroeconomic models o esimae equilibrium real exchange raes for currencies of he major indusrial counries in he 199s. His analysis revealed significan deviaions of real exchange raes from heir implied equilibrium values. So oo did Bayoumi e al (1994), Faruqee e al (1996), Wren- Lewis and Driver (1997), Wren-Lewis (23), and Rosenberg (23). Oher researchers have applied relaively small heoreical models, using empirical esimaes of underlying parameers o esablish he relaionship among oupu, he balance of paymens posiion, 5 Clark e al (1994) and Clark and McDonald (1998) survey many of hese sudies 9
and he real exchange rae. An example is Sein (1995) who calculaed equilibrium real exchange raes by esimaing regression models of he real exchange rae wih several macroeconomic variables. Infac, mos sudies employing he macroeconomic balance approach have focused on indusrial counries. 2.3 The behavioural equilibrium exchange rae (BEER) approach Finally, a more recen mehod is he behavioural equilibrium exchange rae (BEER) approach. Pioneered in Clark and McDonald (1998, 2), he BEER approach is a modelling sraegy designed o seek a long run relaionship beween observed real exchanges and a se of fundamenal deerminans derived from a heoreical real exchange rae model. Is has he aracive feaure ha he real exchange rae is required o be in equilibrium only in erms is value given by he appropriae se of explanaory variables over a specific sample period. This allows represenaion of he equilibrium real exchange rae in erms of he dynamic srucure ha generaes he daa on he real exchange rae and is fundamenal deerminans (Hinkle & Moniel, 1999), even hough he variables hemselves are derived from a long run srucural model. Applicaion of he BEER approach requires, as a firs sep, specificaion of a behavioural relaionship of he real exchange rae and a se of economic fundamenals drawn from a reduced-form heoreical model. This relaionship is hen esimaed o esablish he long run impac of fundamenal deerminans on he real exchange rae. However, since curren values of fundamenals may include he influence of ransiory facors, he approach also requires judging wheher he economic fundamenals deerminans hemselves are a heir long run or susainable values (Clark & McDonald, 1998). On his, a common pracice in he lieraure is o calibrae values of explanaory variables using a derending filer such as Hodrick and Presco s (1997) H-P filer or Beveridge and Nelson (1981) decomposiion, boh of which decompose a ime series variable ino is long run rend and shor run cyclical value. The variables long run rend values obained from such decomposiion are hen used o approximae heir long run equilibrium values and subsiued ino he esimaed 1
real exchange relaionship o esablish he long run equilibrium real exchange rae. Finally, an assessmen of misalignmen is made by comparing he value for he real exchange generaed by his relaionship conaining de-rended values for he economic fundamenals wih he acual value for he real exchange rae. However, because of differences in mehodological approaches o idenifying fundamenal deerminans, and hose of esimaing hese relaionships, he BEER approach ends o yield differen values for he equilibrium real exchange raes and he exen of heir misalignmens. All he same, here is somewha exensive applicaion of he BEER approach o analysis of currencies of developing and middle-income economies. An influenial sudy is Edwards (1989) who applied he mehodology o 12 developing and middle income counries from 1962 hrough o 1984, and found significan misalignmens in heir currencies. So oo did Elbadawi s (1994) sudy of currencies for Chile, Ghana, and India, and Baffes e al (1999) for Coe d Ivoire. More recenly, Manhisen (23) used a heoreical model drawn on Edwards (1989) o invesigae real exchange rae misalignmen in Malawi and found several episodes during which he Malawan Kwacha had experienced several misalignmens. However, Cady (23) found ha he Malagasy franc did no seem o have experienced any major misalignmens for he mos of he 199s. There is herefore confirmaion in he lieraure ha real exchange raes are predicable based on heir underlying deerminaes, despie variaions in mehodologies used o esimae hem. 3 Modelling he rand s equilibrium real exchange rae and assessing he exen of misalignmen We apply he Behavioural Equilibrium Exchange Rae (BEER) mehodology, and base our examinaion of he relaionship beween he real exchange rae of he rand and is fundamenal deerminans relying on reduced-form relaionships implied by he heoreical model developed in Moniel (1999). 11
In Moniel s model 6, he real exchange rae is defined as he relaive price of nonraded goods in erms of raded goods, and he long run equilibrium exchange rae is ha rae ha is consisen wih achievemen of inernal and exernal equilibrium, when underlying deerminans are a heir susainable values. The model develops on he basis of a small open economy wih a fixed exchange rae sysem, financially open o inernaional financial markes, and flexible domesic wages and prices. The producion side of he economy consiss of producers who produce raded and non-raded goods, and are price akers in he world marke. Oupu in each secor is produced wih a fixed, secor-specific, homogenous, and freely mobile labour inpu, which is also subjec o diminishing marginal reurns. In boh secors, producers make heir allocaion decisions on he basis of profi maximizaion objec by seing he marginal produc of labour o he wage rae. The demand side of he economy is modelled in erms of he behaviour of households and he public secor. The behaviour of he represenaive household reflecs is acions o maximize curren and fuure discouned consumpion of boh he raded and he non-raded good, subjec o a wealh consrain, which also deermines he pah of consumpion expendiure. A each period, he acions of he represenaive household are such ha i decides o allocae is ne worh beween ne foreign bonds (financial asses) and domesic money. Bonds can be held eiher in domesic currency, which pays a nominal ineres rae r, or foreign currency, which pays a nominal ineres rae, r, and he wo ineres raes are relaed hrough an arbirage relaionship described by he uncovered ineres rae pariy condiion. The holding of domesic money, on he oher hand, is moivaed by he desire o reduce ransacion coss relaed o consumpion. This allocaion decision defines he represenaive household s wealh consrain. The public secor, on he oher hand, is a consolidaed public secor, made up of he governmen and he cenral bank. The cenral bank s funcions are o manage he exchange rae, which i achieves by exchanging domesic currency for foreign currency 6 For a similar approach, see also Feyzioglu (1997) 12
upon demand, and providing credi o governmen. In addiion o receiving credi form he cenral bank, he governmen also receives lump-sum axes from he privae secor, which i uses o consume raded and non-raded goods. Wih regard o he exernal posiion, he counry can borrow from and lend o he res of he world, since he economy is financially open o he res of he world. However, is financial liabiliies are deemed imperfec subsiues for hose of he res of he world. Given his, he rae a which is residens can borrow from or lend o he res of he world reflecs a risk premium, which reflecs he counry s abiliy o borrow or is inernaional indebedness. This is defined by he uncovered ineres rae pariy arbirage condiion, adjused for a risk premium. Given he acions of households, he governmen, and he counry s relaionship wih he res of he world, he model, when solved, implies an exernal and exernal equilibrium condiion for he economy. Exernal equilibrium is aained when he level of consumpion and he real exchange yield a susainable curren accoun balance. This equilibrium condiion is represened by he following relaionship: π a = y ( e) + r a ( τ + θ ) c (1) T g T Where π is he world inflaion rae, a is he ne foreign asses of he counry, y T (e) is raded goods oupu, inversely relaed wih he real exchange rae, e is he real exchange rae [measured as he relaive price of non-raded goods in erms of raded goods], he foreign nominal ineres rae, a r is ineres receips on ne foreign asses, τ is he ransacion coss associaed wih consumpion, θ is he share of raded goods in oal consumpion, and g T is governmen consumpion of raded goods. r is In he above expression, he righ hand side denoes he curren accoun balance, and equals real naional saving (i.e. i shows he surplus over household and governmen consumpion of raded goods). The Lef hand side, on he oher hand, represens he susainable value for he counry s capial accoun, which is he value of ne foreign 13
asses adjused for inflaion. These wo componens ogeher underscore he idea ha, in he long run, he real curren accoun mus equal he real value of a counry s claims on he res of he world. Because raded goods oupu is inversely dependan on he real exchange rae, a fall of is consumpion appreciaes he real exchange rae. Wih regard o inernal equilibrium, his holds when he marke for labour and he non-raded good clears. This is defined by he following relaionship: ( 1 ) c g N y N ( e) = c N + g N = θ e + (2) Where y N denoes oal producion of non-raded goods, which is posiively relaed o he real exchange rae, c is oal privae consumpion of he non-raded good, measured in erms of he raded good, g N is governmen consumpion of he non-raded good. Given he posiive relaionship beween non-raded goods producion and he real exchange rae, increased household and governmen consumpion of he non-raded good will increase is producion, and hence appreciae he real exchange rae. 3.1 Facors ha deermine real exchange raes Since he equilibrium exchange rae is given as ha rae which is consisen wih simulaneous achievemen of exernal and inernal equilibrium, facors deermining his macroeconomic balance condiion consiue he real exchange rae s fundamenal deerminans. In he framework of Moniel s model 7, hese are idenified as domesic supply side facors, he sance of fiscal policy, changes in he inernaional financial environmen, and he sance of rade and commercial policy. Domesic supply-side facors A candidae variable for changes in domesic supply relaed facors is he relaive produciviy growh rae beween a counry and is rading parners, firs suggesed by Balassa (1964) and Samuelson (1964) and is ofen referred o as he Balassa-Samuelson 7 For a deailed discussion, see Moniel (1999b) 14
effec. Balassa and Samuelson s argumen is ha if a counry were o experience a produciviy shock ha increases producion in is radable goods secor relaive o is non-radable goods secor, his would have he effec of appreciaing he equilibrium real exchange rae. This would follow because increased oupu of radable goods would permi an increase in demand for labour in favour of his secor, and hence he real wage. Resuling from his, a shif in labour from he non-radable goods secor would be expeced, which, a he given real exchange rae, would mean an expansion of he radable secor while he radable secor would conrac. This would have he effec of creaing excess demand in he non-radable goods secor; hereby give rise o a higher relaive price non-radable. By increasing producion of radables relaive o nonradables, he produciviy shock also gives rise o a rade surplus, which means ha he real exchange rae would have o appreciae o mainain eernal equilibrium. Accordingly, counries ha experience higher raes of produciviy growh in heir radable goods secors relaive o heir rading parners are expeced o experience appreciaions of heir real exchange raes (Aghevli e al, 1991). Sance of fiscal policy Fiscal policy has an ambiguous effec on he equilibrium real exchange rae: he direcion of is quaniaive influence depends on he secoral composiion of he change in governmen expendiure. If governmen increased is expendiure on radable goods, his would have he effec of increasing demand for radable goods hrough an increase in impor consumpion, hereby creaing a rade defici. To mainain exernal equilibrium, a depreciaion of he real exchange rae would be required. By conras, increased governmen expendiure on non-radable goods will end o creae excess demand in he non-radable goods marke. To mainain goods marke equilibrium, an increase in he relaive price of non-raded goods would be required, hereby appreciaing he real exchange rae. Thus, changes in he sance of fiscal policy can eiher depreciae or appreciae he real exchange rae. Facors relaed o changes in he counry s inernaional economic environmen 15
Aspecs of changes in a counry s inernaional economic environmen considered in he lieraure include changes in he erms of rade. Here an improvemen in a counry s erms of rade, which refers o an increase in he price of is exporable goods relaive o imporable goods, is expeced o appreciae he real exchange rae. This follows because increases in he price of expors permi an expansion of he expor secor oupu, and hus gives rise o excess supply of exporable goods and a rade surplus. This increase in expor secor oupu would allow an increase in he real wage in he raded goods secor relaive o ha of he non-raded goods secor. As a resul of his, a redisribuion of labor from he non-raded o he raded goods secor would be expeced. A he same ime, oupu of non-raded goods would be expeced o conrac, hereby creaing excess demand in he non-raded goods marke, and hus a higher relaive price. Boh he improvemen of he rade balance and he rise in he price on non-radable goods require an appreciaion of he real exchange rae if equilibrium is o be mainained. However, MacDonald and Ricci (23) poin ou ha in pracice few sudies find he erms of rade variable o be significan largely on accoun of he arbirariness involved in calculaing his variable (p. 4). Thus an alernaive approach used has been o proxy he erms of rade variable wih real commodiy price movemens of major commodiy expors, which is no only an accurae measure paricularly for major commodiy exporers, bu has also received empirical suppor (MacDonald & Ricci, 23). Anoher facor conneced o he change in a counry s inernaional financial environmen is he flow of inernaional capial and ransfers. Increases in he flow of capial imply a higher ne foreign asse posiion of he counry, which increases he counry s level of real income. This hen permis a larger expendiure on domesic (nonradable) goods (MacDonald & Ricci, 23), creaing excess demand and a rise in he price of non-radable goods. As a resul, an appreciaion of he real exchange rae would be expeced. Changes in a counry s real ineres rae differenial relaive o is rading parner counries will also end o appreciae he equilibrium real exchange rae. For an economy ha is highly open o inernaional capial markes, a higher domesic real ineres relaive 16
o ha of is rading parners, represening say aggregae demand condiions, produciviy levels, or persisen ighening of moneary policy (MacDonald & Ricci, 23), creaes opporuniies for capial o flow o ha counry. As a resul, an improvemen of he curren accoun would be expeced, hereby appreciaing he real exchange rae. By conras, changes in he world inflaion rae will end o cause a depreciaion of he equilibrium real exchange rae. The ransmission mechanism for his is hrough he effecs of inflaion on ransacion coss associaed wih changes in he real demand for money. The direcion of such ransacions coss on he equilibrium real exchange rae will end o depend on wheher hey are incurred hrough consumpion of raded or non-raded goods, which in boh cases, reduces heir supply. In he case of raded goods, his would require a real exchange rae appreciaion. For non-raded goods, he implicaion is ha he real exchange rae should depreciae. To he exen domesic prices reflec he relaive price of non-raded goods, a higher domesic inflaion for a counry relaive o ha of is rading parners would imply a depreciaion of is real exchange rae. Sance of rade and commercial policy Finally, greaer openness of he rade regime by way of rade liberalizaion is expeced o cause he equilibrium real exchange rae o depreciae. The effec of rade liberalizaion works by channelling resources from he radable secor ino he nonradable secor. For insance, a ariff reducion would have he effec of reducing he relaive price of imporables and also creaing excess supply boh exporables and nonraded goods (Aghevli, e al, 1991). Here he emergence of excess supply in he nonradable goods secor permis a depreciaion of he equilibrium real exchange rae. 17
3.2 Empirical Applicaion 3.2.1 Specificaion of he empirical model On he basis of heoreical moivaions discussed above, we expec he following facors o influence real exchange rae movemens of he rand, he real ineres rae differenial vis-à-vis Souh Africa s main rading parners, he inflaion differenial vis-à-vis Souh Africa s main rading parners, variaions in he erms of rade for Souh Africa, he level and composiion of governmen expendiure [sance of fiscal policy], relaive echnological or produciviy progress [proxy for Balassa-Samuelson effec] vis-à-vis Souh Africa s main rading parners, inernaional capial flows, openness of he rade regime [proxy for he sance of commercial and rade policy], he sock of foreign currency reserves, and he sock of foreign deb. Accordingly, our empirical model has he following specificaion: q = α + β + ε (3) F Where q denoes he rand s real exchange rae, and F is a vecor of fundamenal variables menioned above, β is a vecor of coefficiens, and ε is a random error erm. Specifically, F = [( r r ), ( π π ), o, gxp, prod, ka, open, fx, fxb ] where r r is he real ineres rae differenial for Souh Africa relaive o is main rading parners, π π is he inflaion differenial, o is he erms of rade, gxp is governmen 18
consumpion (expendiure), prod is relaive produciviy growh (proxy for Balassa- Samuelson effec), ka is inernaional capial flows, open a measure for openness of he rade regime, fx is foreign currency reserves, fxb is foreign deb, and he aserisk ( ) denoes foreign variables. The real ineres rae differenial, erms of rade, produciviy growh, capial flows, and foreign reserves all ener he model wih a posiive sign, since hese variables are expeced o appreciae he real exchange rae. Conversely, he inflaion rae differenial, rade openness, and foreign deb variables depreciae he real exchange rae, and are hus specified wih a negaive sign, whereas as he expeced impac of governmen expendiure is ambiguous. 3.2.2 Tesing mehodology In he specificaion discussed above, he objec of ineres is one of esing wheher a long run relaionship exiss beween he rand s real exchange rae wih his se of fundamenal deerminans by esimaing β, he vecor of parameers. Our choice of he mehodology o accomplish his was dicaed by properies of our daa. Togeher wih oher macroeconomic variables, exchange raes appear o possess non saionary sochasic rends ha allow variables o grow overime wih no endency o rever o heir mean values. The presence in he daa of such non saionary sochasic rends is known o invalidae sandard saisical inferences. Thus, we employed coinegraion analysis, which addresses his non saionary problem in he daa. Coinegraion ess have he benefi of allowing one o es for and esimae equilibrium relaionships among variables and he adjusmen process o such equilibrium, even hough he processes ha generae he daa are hemselves non saionary sochasic rends. In he lieraure on coinegraion analysis, various approaches have been suggesed, bordering on boh single equaion and muliple equaions approaches (see Pesaran & Pesaran 1997). For our purposes, we applied Johansen s (1991, 1995) mulivariae coinegraion ess, which examines he problem of esing for coinegraing relaionships in he conex of a sysem of joinly deermined non-saionary variables ha are generaed 19
as a vecor auoregressive (VAR) process, and models he relaionships among he variables using a vecor equilibrium correcion specificaion (Cady, 23, p.4). This sars wih examining a VAR model of he following form: p Z = a + a1 + Ψi Z i + Bw + U (4) i = 1 Where Z is an 1 nx vecor of joinly deermined non-saionary ( 1) I variables (here n is he number of variables in Z), w is a vecor of exogenous saionary[ I() ]variables, is a deerminisic rend variable, α is a vecor of inercep coefficiens, Ψ is an i nxn marix of coefficiens, Β is nx1 vecor of coefficiens on exogenous saionary variables, and U is o an nx1 vecor of whie noise error erms. When coinegraing relaionships are presen among he joinly deermined variables, he above vecor auoregressive process gives rise o a dynamic vecor equilibriumcorrecion [error-correcion] model (VECM), which is essenially a resriced vecor auoregressive process ha has co inegraion resricions buil ino he specificaion (Eviews3. Users Guide, p.54). This specificaion follows from Engle and Granger s (1987) represenaion heorem which says ha if a group of non-saionary daa processes are co inegraed, hey can be regarded as having been generaed by an equilibrium correcion process ha resrics he long run behavior of joinly deermined variables o heir long run values while allowing for a series of parial adjusmens over he shor run (Maddala & Kim, 1998). Hence we may rewrie he VAR process ino is VECM form as follows; p Δ = a + a + ΠZ + 1 1 1 Γi ΔZ i + B i = 1 Z w + U (5) p Where Π = ( Ι Ψ j ) and Γi = ( Ι Ψi ). i = 1 j = 1 p 1 Here Π denoes a marix of long run coefficiens ha defines long run relaionships among he variables, while adjusmens among he variables (Pesaran & Pesaran, 1997). Γ i is a coefficien marix capuring shor run dynamic 2
When specified in his VECM form, a vecor auoregressive process has he advanage of providing informaion on boh he long run relaionships among he variables and heir shor run adjusmen o such long run equilibrium relaionships. To begin wih, since he marix Π conains he long run coefficiens ha underlie long run relaionships among he variables, he relaion ΠZ represens he long run equilibrium relaionships among he variables. The relaions conained in he lagged erm ΠZ 1 serve as equilibrium correcion devices ha ensure ha he shor run adjusmens of variables are ied o heir long run values overime. Secondly, he erms Δ represen he shor run dynamics or deviaions of he variables from heir long run equilibrium, whose impac is represened by. Since coinegraion is assumed in his formulaion, is saionary, as is Δ considering ha Z is iself a non-saionary I(1) process. As a resul, he vecor error-correcion model is a saionary process, which allows applicaion of sandard inference saisics even hough he elemens of he sysem are hemselves non-saionary. The vecor error-correcion model is hus appropriae o model nonsaionary daa processes ha are coinegraed. Z i Γi ΠZ 1 Z i i Because he marix Π conains long run coefficiens, and hus provides informaion on abou long run relaionships among he variables, he presence of co inegraion among he variables is indicaed by he rank of his marix, which requires finding r, he number of linearly independen columns ha are presen in he marix. Three cases are disinguishable in his respec: Case (I). If he marix Π is of full rank, corresponding o number of variables in he sysem), hen here exis columns in he marix Π. In his case, he variables in r = n (where n equals he r = n linearly independen are saionary, and no co-inegraion exiss among he variables. A VAR model wih variables specified in heir levels is hus he appropriae model. Case (II). On he oher hand, if he rank of he marix Π is zero, when he elemens in Z are non-saionary, hen no coinegraion exiss among he variables. A VAR Z 21
model wih variables specified in heir firs differences in he appropriae model in his case. Case (III). Finally, if he marix Π of reduced rank when he elemens in Z are nonsaionary, hen here exis a mos r n 1linearly independen columns in he marix Π, which indicaes he presence of r co-inegraing relaions in Z. In his case, he VECM specificaion is he appropriae framework for modelling he relaionships among he variables. Accordingly, he hypohesis of he exisence of co inegraing relaionships among he variables can be formulaed in erms of a hypohesis of he reduced rank of marix Π. This can be saed as follows: H ( r) : rank( Π ) = r n 1. Under his hypohesis of co inegraion (i.e. case iii above), he pie marix is equivalen o Π = αβ, which flows from Engle and Granger s (1987) represenaion heorem. Here α is a full column rank n r marix represening he speed of adjusmen of variables in he model, whereas he marix β is also a full column rank n r marix of long run coefficiens, wih he propery ha β Ζ is saionary. In he ligh of his decomposiion of he pie marix, he vecor error-correcion model can be wrien as follows: p 1 a + 1 + αβ Z 1 + ΓiΔZ i + B i 1 Z = a Δ w + U (6) = In his vecor error correcion specificaion of he VAR model, he saionary coinegraed relaionships among he variables are now defined byξ = β Ζ 1, and his characerizes he long run relaionships among he variables. As a resul, he hypohesis on he rank of he marix Π is now equivalen o formulaing a hypohesis abou he rank of he marix β, which has he implicaion ha each column of β is he coinegraing vecor (Eviews5. Users Guide, p. 723). We may resae his as follows: 22
H ( r ) : rank( Π) = rank( β ) = r n Tesing for coinegraion herefore requires an esimae of he pie marix Π = αβ. 1 In he Johansen s approach, his is achieved by esimaing he pie marix Π = αβ from he unresriced VAR model, and hen esing wheher he resricions imposed by he reduced rank of marix Π can be rejeced by he daa (Eviews5. Users Guide, p. 724). This is esed using wo ypes of ess advocaed by Johansen. The firs is he race saisic, which allows one o perform a log likelihood raio (LR) es for he null hypohesis ha here exis r co inegraing vecors agains he alernaive of inegraing vecors (where saisic is compued as follows: n co n is he number of endogenous variables in he VAR). This LR race n = 2log( Q) = T log(1 ˆ λ ) ( r =,1, K, n 2, n 1) i = r + 1 i The second is he maximum eigenvalue ( λ max ) saisic. This is a log likelihood es ha ess he null hypohesis ha here is r co inegraing vecors agains he alernaive ha here is r + 1 co inegraing vecors. This saisic is compued as follows: LR ˆ max = T log( 1 λ r + 1 ) ( r =,1, K, n 2, n 1) Asympoic criical values for conducing hese wo ess have been provided by Johansen (1988), and also Oserwald-Lenum (1992). However, more ofen, he daa series under invesigaion will no only exhibi sochasic rends, bu may also exhibi non-zero means and or deerminisic rends. In his even, he co inegraing relaions among hese daa series will also exhibi non-zero means and or deerminisic rends, and his would have o be aken ino accoun during he course of invesigaion. Bu hese log likelihood raio saisics for he esing for he presence of co inegraion discussed above do no poses he usual chi-square saisics, and have been found dependan upon he rend assumpions underlying he daa under 23
invesigaion (Eviews5. User s Guide, p. 724). Because of his, he underlying daa rends would have o be ascerained prior o implemening he ess. On his, five rend assumpions on wheher he daa series conain inerceps and or ime rends, and wheher he inerceps and or ime rends are resriced (i.e. form par of he co inegraing relaions) are disinguishable (Pesaran & Pesaran, 1997) 8 : Case (I). Case (II). Case (III). Case (IV). The level series daa have no rends and he co inegraing relaions have no inerceps (i.e. a = a ). 1 2 = The level of he daa have no deerminisic rends and he co inegraing relaions have inerceps (i.e. resriced inerceps and no rends in co inegraing relaions, a = 1 and a = Πρ where ρ denoes an inercep coefficien). The level daa have linear rends bu he co inegraing relaions have only inerceps (i.e. unresriced inercep and no rends, a 1 = and a ). Boh he level daa and he co inegraing relaions have linear rends (i.e. unresriced inercep and resriced rend, a and a1 = Πγ where γ denoes a rend coefficien). Case (V). The level daa have quadraic rends and he co inegraing relaions have linear rends (i.e. unresriced inercep and rend, a and a 1 ). In pracical applicaion, deerminaion of he number of co inegraing relaions using he Johansen s mehod is a sequenial process condiional on he rend case assumpion discussed above, esing firs he hypohesis har =, and hen proceeding o r = n 1 unil he es fails o rejec he hypohesis. Afer deermining he number co inegraing relaionships among he variables, he nex sep is o esimae of he vecor equilibriumcorrecion model in order o examine he long run predicabiliy and shor run forecasing abiliy of he model. 8 For furher discussion of he five rend case assumpion, see also Eviews5. Users Guide (pp.724-727) and Pesaran and Pesaran (1997, pp. 132-136) 24
Tess for he order of inegraion of variables However, esing for coinegraion is valid only when he variables are generaed by a non-saionary process. Asceraining he order of inegraion of he daa under invesigaion is herefore a key preliminary sep o is applicaion. One widely used mehodology for esing he order of inegraion of variables, which his sudy employs, is Dickey and Fuller s (1979, 1981) DF and ADF ess. This ess wheher ρ = 1, which is a es for exisence of a uni roo, assuming ha he ime series daa under invesigaion is generaed by a firs order auoregressive process of he following form: s = α + α1 + ρs 1 + ε (7) Where α is a consan, is a deerminisic rend, andε, is a whie noise sochasic process. Alernaively, he above auoregressive process can be wrien as follows. Δs = α + α + γ + ε 1 s 1 (8) Whereγ = ρ 1. The ADF es is performed on his re-parameerized form of he auoregressive process. In paricular, he hypohesis for exisence of a uni roo is formulaed as H : γ, under which case he process s } is said o be a non-saionary I ~ (1), = agains he alernaive H : γ <, which has he implicaion ha he process is saionary. a However, in is pracical applicaions, he ADF es requires esimaing a more general auoregressive process ha includes lagged differences of he variable under invesigaion, which are included in order o conrol for auocorrelaion in he sochasic error erm. In erms of his consideraion, he es involves esimaing he following regression model; { 25