OUTPUT ADJUSTMENT IN DEVELOPING COUNTRIES: A STRUCTURAL VAR APPROACH

Transcription

1 OUTPUT ADJUSTMENT IN DEVELOPING COUNTRIES: A STRUCTURAL VAR APPROACH Seven Morling School of Economics The Universiy of Queensland Brisbane Qld 4072 Ausralia fax: May 2002 Discussion Paper No 307 ISSN Morling, 2002 This discussion paper should no be quoed or reproduced in whole or in par wihou he wrien consen of he auhor.

2 OUTPUT ADJUSTMENT IN DEVELOPING COUNTRIES: A STRUCTURAL VAR APPROACH Seven Morling Absrac: This paper examines wheher emporary flucuaions in oupu around poenial in developing counries are induced primarily by aggregae demand shocks or emporary aggregae supply shocks. Srucural vecor auoregression mehodology using long-run resricions is used o idenify emporary oupu shocks for a large sample of developing counries. Impulse response funcions are used o examine wheher he emporary shocks behave like demand or supply shocks. The permanen/ransiory decomposiion appears o spli he shocks ino permanen supply shocks, and emporary demand or supply shocks depending on which influence dominaes in a paricular counry. In a lile over half of he counries, emporary shocks behave like emporary aggregae supply shocks; in a lile under half of he counries he emporary shocks behave like aggregae demand shocks.

3 1. INTRODUCTION The oupu gap he gap beween acual and poenial oupu plays an imporan role in macroeconomic modeling and policy formulaion (Nickell, 1988; Chadha e al., 1992; De Masi, 1997). Flucuaions around poenial oupu represen an accumulaion of aggregae demand shocks and emporary aggregae supply shocks. Demand shocks are generally hough o dominae hese processes, a leas in developed counries. In developing counries, however, shor-erm flucuaions in oupu around poenial are also likely o be influenced by he effecs of emporary economy-wide supply shocks such as oil price shocks, erms of rade shocks, and oher disrupions such as droughs, floods, cyclones and social and poliical upheavals. Hoffmaiser and Roldos (1996, 1997), for example, find ha supply shocks are he dominan influence on shor-erm oupu flucuaions in Asian and Lain American counries; Hoffmaiser e al. (1998) find similar resuls for Sub-Saharan African counries. Given he heerogeneiy of developing counries i is no clear herefore wheher convenional measures of he oupu gap are picking up demand- or supply- induced flucuaions in oupu for individual counries. No surprisingly, his disincion has imporan implicaions for how policy-makers inerpre hese measures when hinking abou inflaion and making policy judgemens. To examine he quesion of wheher oupu gaps in developing counries are dominaed by he effecs of aggregae demand shocks or emporary aggregae supply shocks, emporary shocks are idenified for a large sample of developing counries using srucural vecor auoregression mehodology. Long-run idenifying resricions in he spiri of Blanchard and Quah (1989) and Shapiro and Wason (1988) are used o idenify srucural shocks and o decompose shifs in oupu ino permanen and emporary componens. Impulse response funcions are used o examine wheher emporary shocks o oupu behave like demand or supply shocks in hese developing counries. 2. THEORETICAL BACKGROUND A useful framework for hinking abou he effecs of various shocks on oupu is a basic bare bones AS-AD model skeched below in equaions (1) o (5): e IS as y = α + α 1 r + ε + ε (1) md m p = α 2 y α 3i + ε (2) ε ms (3) m = IS md ms ( )[ + ] + θ ( ) ε e as p = δ p + (1 δ ) p + φ L ε ε + 1 ε L (4) * [ e ] ( ) + = i 1 i ψ E + (5)

4 2 where y is he log of oupu, r e is he expeced real ineres rae, i is he nominal ineres rae, p is he log of he price level, p e is he expeced rae of inflaion in period +1 (condiional on informaion available a ), m is he log of he money supply and e is he log of he nominal exchange rae. The erms ε IS, ε as, ε md and ε ms are exogenous shocks o domesic absorpion, aggregae supply, money demand and money supply respecively. IS shocks are inerpreed as exogenous shifs in fiscal policy. The supply shock represens echnology and labour supply disurbances. 1 I also includes emporary aggregae supply shocks. Srucural reforms are also likely o be a significan source of domesic supply shocks in developing economies. 2 The money shocks are inerpreed as domesic money shifs in money demand and exogenous money supply changes. ψ is a wedge o allow for he possibiliy of capial conrols. L is a lag operaor. Equaions (1) and (2) are relaively sandard IS and LM equaions describing equilibrium condiions in goods and asse markes. 3 Equaion (3) is a money supply equaion and (4) is a loosely-specified price adjusmen equaion. Equaion (5) is he usual open-economy uncovered ineres rae pariy condiion, modified o allow for he possibiliy of risk premia and/or capial conrols. In he model skeched above he long-run aggregae supply curve is verical, while he aggregae demand curve is downward sloping. Permanen movemens in oupu are due o supply shocks; demand shocks do no have permanen effecs on oupu. 4 The shor-run aggregae supply curve is upward sloping, because he prices of some inpus are assumed o be deermined under aucion-like condiions, because of markup pricing and/or because of assumed informaional asymmeries. This allows for he possibiliy ha oupu can be varied in he shor-run in response o shifs in aggregae demand. In pracice, he slope of he shor-run aggregae supply curve is expeced o be relaively seep, so ha a subsanial porion of he adjusmen o demand shocks is likely o be borne by shifs in prices, or emporarily, hrough excess money holdings. Some of he shor-erm shifs in oupu are likely o be due o emporary supply shocks. As discussed below, his has imporan implicaions for observed price and oupu dynamics. The price adjusmen equaion is loosely specified. The firs wo erms reflec he role of inflaion expecaions and he degree of price flexibiliy in he price adjusmen process. The lagged inflaion erm allows for he possibiliy of backward-looking expecaions (adapive expecaions) and/or wage and price ineria. The hird erm in he price adjusmen equaion is a vecor of exogenous demand shocks. While he incorporaion of he expeced inflaion erms allows for inflaion expecaions o adjus insananeously o expeced or anicipaed changes in 1 Since developing counries are ofen relaively open, supply shocks may also include changes in world oil prices, erms of rade changes and shifs in world real ineres raes. 2 These reforms include: public secor reforms, including ax reforms and resrucuring and privaisaion of public enerprises, he removal of rade and capial conrols, relaxaion and/or removal of price conrols, and labour marke reforms. 3 The ineres rae is reained in equaions (1) and (2) for exposiional sake alhough is role is likely o be severely limied by he relaively undeveloped naure of financial markes in developing counries and he conrols or ceilings on ineres raes ofen imposed by moneary auhoriies. The opporuniy cos of holding financial asses in erms of goods expeced inflaion is he componen of he nominal ineres rae ha is likely o be mos relevan for many developing counries. 4 The simplificaion ha demand shocks do no have permanen effecs on oupu is widely used in he lieraure, hough no unconroversial (for example, see Gali, 1992). For he purposes of his sudy he requiremen is only ha he effecs are relaively small, compared wih supply shocks (Blanchard and Quah, 1989)

5 3 endogenous variables (such as changes in he money supply, oupu or he exchange rae) he model allows for he possibiliy ha unexpeced demand shocks may affec prices indirecly hrough emporary disequilibrium in goods and/or money markes. A relaively seep supply curve, for example, would see demand shocks direcly affec prices, while a relaively shallow curve (for example in he presence of employee mispercepions in he new classical model or sicky prices/wages in new Keynesian models) would see much of he iniial effec fall on oupu or, in he absence of well-funcioning financial markes, on excess holdings of money. The fourh erm in he price adjusmen equaion is exogenous supply shocks. These exogenous shocks include permanen shocks, and emporary shocks, which may emporarily drive a wedge beween acual and poenial oupu. The sandard can be used o show he effecs on oupu of a demand shock and a emporary supply shock under he usual new Keynesian assumpions of sicky prices and wages (Figure 1). The demand case is sraighforward. An decrease in demand from D 0 o D 1, o he exen ha prices are fixed, falls iniially on oupu, which emporarily falls from Q 0 o Q 2. If some prices are flexible he decrease in demand will be moderaed by an iniial decline in prices. Over ime, he shor-run supply curve will shif down when new price and wage agreemens are negoiaed. The iniial negaive disequilibrium (he oupu gap) will be associaed wih a fall in prices (from P 0 o P 1 ). In he case of a supply shock, however, an iniial negaive disequilibrium is associaed wih a rise in prices. In he case of a emporary adverse supply shock he producion funcion and he labour demand curve shif down. If wages and prices are sicky, real wages will no iniially adjus o clear he labour marke and he economy will emporarily operae a a level of oupu below he naural rae (such as Q 2 ). Over ime, prices will rise as price and wage conracs are renegoiaed. If moneary auhoriies accommodae he emporary shock, oupu will iniially fall and prices will rise (o P 1 ); if auhoriies adop a neural policy he aggregae demand curve would shif o he righ o inersec he new aggregae supply curve a he (permanen) naural level of oupu (a E 2 ), puing furher upward pressures on wages and prices. Over ime, he oupu effecs will be reversed as he effecs of he emporary supply shock abae. The permanency of he effecs on he price level depends on he sance of he moneary auhoriies. In boh he demand and he emporary supply shock cases here is he familiar emporary defici capaciy in he goods and facor markes. In he former case, however, he negaive gap is associaed wih a fall in prices; in he laer case he negaive gap is associaed wih a rise in prices.

6 4 Figure 1: Demand Shocks and Temporary Supply Shocks DEMAND SHOCK D 1 D 0 S 0 Price Level P 0 P 1 E 0 E 1 Q 2 Q 0 Real GDP TEMPORARY SUPPLY SHOCK S 1 Price Level P 2 P 1 P 0 D 0 E 1 E 2 E 0 S 0 Q 2 Q 1 Q 0 Real GDP

7 5 3. ESTIMATING THE OUTPUT GAP USING SVAR METHODOLOGY Since poenial oupu and he oupu gap are no observable various economeric echniques are commonly used o consruc hem. These echniques include univariae and mulivariae filers and srucural models. The various mehods have been reviewed in several recen papers including De Masi (1997), Dupasquier, Guay and S-Aman (1997) and S-Aman and Van Norden (1997). An alernaive echnique o he mechanical filers and he more sophisicaed srucural models discussed above is he srucural VAR echnique. This mehod uses minimum heoreical resricions o idenify he major shocks o he sysem and o decompose movemens in oupu ino permanen and ransiory componens. Those shocks can hen be used o consruc he measures of he oupu gap. A paricular varian of ha approach based on long-run resricions on oupu proposed by Blanchard and Quah (1989) and Shapiro and Wason (1988) is used in his paper. Wihin he srucural VAR approach, here are several echniques ha can be used o recover he informaion required. The reduced-form VAR provides reduced-form errors from which he srucural shocks can be recovered. Blanchard and Quah (1989) show ha he recovery of he srucural shocks from he reduced-form errors requires he idenificaion of he elemens of he marix of conemporaneous coefficiens ha relaes he srucural shocks o he reduced-form errors. One way o do his is o use a Choleski decomposiion, which resrics he marix of conemporaneous coefficiens o be a lower riangular marix. The effec of his is ha one variable is assumed no o have a conemporaneous effec on he oher(s). Boh shocks may affec he conemporaneous value of one variable, bu only one shock is allowed o affec he oher. Unforunaely his mechanical procedure is very sensiive o he ordering of he VAR variables, paricularly when he correlaion beween he reduced-form errors is high. Bernanke (1986) and Sims (1986) use direc resricions on he conemporaneous ineracions amongs he variables based on economic heory. An alernaive approach he one ha is used in his paper follows Blanchard and Quah (1989) and Shapiro and Wason (1988) and uses resricions on he long-run dynamic effecs of he shocks on paricular variables in he sysem o idenify he srucural shocks. An advanage of he Blanchard and Quah approach is ha i does no impose any resricions on he conemporaneous ineracion amongs he variables in he sysem and leaves he dynamics of he sysem unconsrained. This is paricularly useful for developing counries where shor-run srucural relaionships are less clearly defined. The idenifying resricions used in his approach are he orhogonaliy resricion, he resricions embedded in he variance covariance marix of he reduced-form innovaions and he long-run resricion. The basic AS-AD model skeched above provides he heoreical underpinnings for he long-run resricions used here. The AS-AD model assumes ha aggregae supply shocks have permanen effecs on he level of oupu while aggregae demand shocks, and emporary aggregae supply shocks, have only emporary effecs. A posiive permanen supply shock shifs he long-run supply curve ou, resuling in a permanen increase in oupu and a permanen fall in prices. A posiive demand shock shifs he aggregae demand curve ou, which increases oupu

8 6 (emporarily) and prices. Since he long-run aggregae supply curve remains unchanged, prices will rise furher over ime and oupu will evenually reurn o he long-run equilibrium level accordingly, only he price level is affeced in he long run. A posiive emporary supply shock shifs he aggregae supply curve ou which increases oupu (emporarily) and reduces prices. These idenifying resricions are relaively unconroversial he shor-run long-run dichoomy and he noion of long-run neuraliy of money are consisen wih mos models including Mundell-Flemming ype models and real business cycle models. A bivariae vecor auoregression sysem using oupu and price daa is used o esimae permanen and ransiory oupu shocks. This specificaion follows Bayoumi (1992) and Bayoumi and Eichengreen (1994), who exend Blanchard and Quah by using oupu and prices raher han oupu and unemploymen. Bergman (1996) and Keaing and Nye (1998) also use a bivariae VAR using oupu and price daa. The choice of his sysem reflecs boh heoreical and pracical consideraions. The approach is relaively sraighforward and he underlying assumpions of his approach are consisen wih he AS-AD framework discussed earlier. I has he advanage ha i has limied daa requiremens only oupu and price daa are required for each counry. This is an imporan consideraion he pauciy of developing counry daa, and heir poor qualiy, is well-recognised (Heson, 1994; Srinvasan, 1994) so using a limied number of he beer-qualiy series has pracical advanages. While here may be some gains o be had by including more informaion o idenify he srucural componens, here is also a cos in erms of less precise esimaes. There is also a cos in imposing addiional idenifying resricions. The idenificaion procedure follows Blanchard and Quah (1989), using he simplificaion proposed by Lasrapes (1992), as followed recenly, for example, by Hoffmaiser and Roldos (1997), as oulined below. The model is expressed as an infinie moving average represenaion of he variables such ha: x A0 ε + A1ε = Ai ε i A( L)ε (6) = = i = 0 y where x = ε p ε = ε P T I is assumed ha he change in oupu, y, and he change in prices, p, are saionary, and ha he permanen and ransiory srucural errors, ε P, and ε T respecively, are uncorrelaed whie noise disurbances. The variance of he srucural shocks is normalised so ha E(ε ε ) = I, he ideniy marix. The moving-average represenaion of he reduced-form of he model is: x e + C1 e = C i e i C( L)e (7) = = i = 0

9 7 where e is a vecor of esimaed reduced-form residuals wih variance E(e e ) = Ω and he marices C i represen he impulse response funcions of shocks o y and p. From (6) and (7) i follows ha he srucural innovaions are a linear ransformaion of he reduced-form innovaions. The reduced form residuals are relaed o he srucural residuals by: e = A( 0)ε (8) where A(0) is a marix of he conemporaneous effecs of he srucural innovaions. I follows ha: E ' ( ) A( 0) E ( ) A( 0) e e = ε ε (9) and since E(ε ε ) = I: A ( 0) A(0) '= Ω (10) To recover he srucural innovaions i is necessary o provide sufficien resricions o idenify he elemens of he marix A(0). From equaions (6) and (7) noe ha C(0) = 0 and hence: A ( 0)ε = e (11) Lagging (6) gives: and herefore: A( j) ε = C( j e (12) j ) j A ( j) = C( j) A(0) (13) Using he simplificaion proposed by Lasrapes (1992), from (13) for j=1: and using (10): A ( 1) = C(1) A(0) (14) A( 1) A(1)' = C(1) ΩC(1)' (15) Where A(1) is lower riangular (as in he case where he cumulaive effec of an ε T shock on he y sequence is equal o zero), i can be calculaed as he lower Choleski decomposiion of C(1)Ω C(1)'. The marix A(0) is hen calculaed as:

10 8 A 1 ( 0) C( 1) A( 1) = (16) This allows he rerieval of he srucural shocks using he residuals from (7). The exogenous shocks idenified from he bivariae srucural models (6-16) and he A marix of coefficiens can be used o consruc measures of he oupu gap. Following S-Aman and van Norden (1997) poenial oupu is consruced as he level o which oupu revers as he effecs of demand disurbances, or emporary supply disurbances dissipae ha is, he level of aggregae oupu achievable over ime wihou placing undue pressures on resources. I is assumed ha shifs in poenial oupu reflec permanen labour supply and produciviy shocks ha is, hose ypes of disurbances ha are likely o have permanen effecs on aggregae oupu. Oher disurbances, are reaed as demand disurbances or emporary supply disurbances, ha is he cyclical (or in he case of supply disurbances, irregular) componen of oupu. From equaion (6) poenial oupu, y*, is given by he sum of he projeced deerminisic rend in oupu and he cumulaive effecs of pas permanen supply shocks: P = µ + A ( L )ε (17) * y 1 where µ is he projeced deerminisic rend in oupu and ε P are permanen aggregae supply shocks. The oupu gap is given by: * ( y y ) = y µ + A ( ) 1 L P ( ) ε (18) The inerpreaion of he oupu gap recovered in (18) depends on he naure of ε T, he emporary shocks inroduced above. If demand shocks dominae movemens in acual oupu around poenial, he permanen/ransiory decomposiion will separae flucuaions in oupu ino supply and demand shocks; if however, shor-erm flucuaions in oupu are dominaed by emporary supply shocks he permanen/ransiory decomposiion will separae flucuaions in oupu ino hese wo componens. 5 The impulse response funcions, repored below, provide an indicaion of he naure of hese shocks. 5 Keaing and Nye (1998, p. 245) make a similar observaion. In referring o nineeenh cenury (hen developing) counries hey noe Suppose here are only wo kinds of shocks in hese hree nineeen cenury economies: Supply shocks ha have emporary oupu effecs and supply shocks ha have permanen effecs on oupu. Under his assumpion, a permanen-ransiory decomposiion would effecively separae oupu ino hese wo supply shocks.

11 9 5. ESTIMATION The srucural VAR model is esimaed using annual daa over he period 1973 o The sample period was chosen o cover a maximum esimaion period while covering a criical mass of counries. 6 The saring poin coincides wih he sar of he pos Breon-Woods era. 7 Fifyeigh developing counries produce he required daa over ha period. The gaps are esimaed separaely for each of he counries in he sample. For analysis, counries are grouped ino five regional groups Africa, Asia Middle Eas, Wesern Hemisphere and Wesern Hemisphere (high inflaion). Regional groups are more likely o share similar srucural and insiuional characerisics and be affeced by similar disurbances han he broader grouping. Daa are from he Inernaional Moneary Fund, Inernaional Financial Saisics. The price measure is he gross domesic produc deflaor (99b/99bp). The gross domesic produc deflaor is used, since his measure is generally regarded as he mos comprehensive measure of a counry s price level (Brajer, 1992). The oupu measure is gross domesic produc a consan prices (99bp). Prior o esimaion, he saisical properies of he daa were examined. The saisical properies of he daa play an imporan role in he ype of empirical analysis conduced here. There are imporan differences, for example, in daa ha have a uni roo process and hose ha have a rend saionary process. Shocks o difference saionary processes permanenly shif he rend; shocks o rend saionary processes have a ransiory effec as he effecs dissipae over ime. Accordingly, idenificaion of permanen and ransiory shocks o series requires differen echniques depending on he naure of he daa generaing process. Similarly, he ypes of ransformaions ha are necessary o induce saionariy will vary depending on he saisical properies of he daa.. The srucural VAR approach requires ha he growh rae of oupu follows a saionary sochasic process. The order of inegraion of prices in he VAR can vary. An I(1) or I(2) variable can be differenced accordingly and a rend-saionary variable can be used by aking he residuals from he deerminisic rend. Three ess are used o examine he order of inegraion of he series: he Augmened Dickey- Fuller 8 (ADF) es, he Phillips-Perron (1988) (PP) es and he Kwiaowski, Phillips, Schmid & Shin (1992) (KPSS) es. 9 The resuls sugges ha oupu is generally I(1) alhough for several counries he evidence suggess ha i is rend saionary. Prices generally appear o be I(1), bu for several counries he resuls sugges ha he series may be beer described by I(2) processes and for a couple of counries hey appear o be rend saionary Mos developing counries publish only annual oupu daa. Since oher sudies sugges ha adjusmen o disequilibrium condiions appears o be relaively slow in developing counries, his may no be oo limiing. 7 The mid 1970s was also a period of subsanial shifs in exchange rae regimes in developing counries (Aghevli e al., 1991). 8 See Dickey and Fuller (1979), Dickey and Panula (1987) and Said and Dickey (1984). 9 The resuls are no repored here bu are available on reques. 10 Alhough he order of inegraion of prices remains a conroversial area, Phillips (1995) shows ha he SVAR echnique performs badly if eiher non-saionary or quasi-non-saionary series are used. Accordingly, he change in

12 10 The firs sep in he process was o esimae reduced-form VARs for each counry in he sample using he bivariae sysems as described above. 11 Oupu was firs differenced and prices were differenced once or wice, depending on he order of inegraion, or aken as a deviaion from rend in he case of rend saionary series. 12 The lag srucure of he VARs was based on a general o specific approach, wih a sequenial likelihood raio saisic used o deermine he appropriae lag lengh. Generally he ess indicaed a lag lengh of one, alhough in a few cases wo lags were necessary o ensure he null of no auocorrelaion was rejeced. However, DeSerres and Guay (1995) argue ha he sandard informaion crieria ofen sugges he inclusion of an insufficien number of lags. A one-period lag lengh may be inadequae o properly capure he dynamic responses of he sysem o economic disurbances (Ahmed and Park, 1994). Accordingly, a lag lengh of wo was used for all counries. 13 The srucural shocks were recovered as described above. The hisorical decomposiion of he variable was consruced by seing all he ransiory shocks o zero and using he permanen shocks o obain he permanen changes in he variable. The various disequilibrium erms were consruced as deviaions from he accumulaion of hese permanen shocks (and he rend). Before proceeding o he resuls, some imporan caveas should be raised. A a general level he SVAR approach suffers from much he same criicisms as leveled agains ordinary VARs he absence of a complee specificaion derived from firs principles including preferences, echnologies and explici equilibrium condiions. The parsimonious naure of bivariae SVARs, while commonly used in economeric research, could arac analogous criicisms a a macroeconomic level. More specifically, inappropriae shock aggregaion and ime aggregaion may become an issue wih SVAR mehodology (Faus and Leeper, 1997). Small dimension SVARs allow he idenificaion of only a limied number of shocks. In pracice, however, here are likely o be many differen ypes of shocks affecing a sysem so ha he shocks idenified by he VAR mehodology are likely o be an aggregaion of shocks. If he shocks are dissimilar han he aggregaion may provide a poor represenaion of he underlying relaionships and see he aggregae shocks poorly idenified. 14 inflaion is included in he vecor auoregession for hose counries where he uni roos ess could no rejec he hypohesis ha prices are I(2). 11 Esimaion was conduced using RATS sofware. 12 Where oupu was I(2) or rend saionary, he SVAR procedure was no appropriae and hose counries were dropped from he sample. 13 As par of he early screening he residuals from he reduced form VAR were examined for he presence of ouliers. While ouliers can conain useful informaion a imes, hey may also reflec measuremen problems, breaks in series and one-off shocks ha canno be adequaely covered in a parsimonious model specificaion. These disorions are likely o be paricularly imporan for developing counries. Ouliers were omied from he reduced form VAR if hey exceeded hree sandard errors. 14 Higher dimension SVARs, however, require addiional idenificaion resricions, which may be difficul o suppor. There is also a loss of precision wih he large dimension VARs, paricularly where daa is limied. The assumpion made in his sudy is ha he shocks are dominaed by wo main underlying shocks; supporing informaion from impulse response funcions is used o provide some confirmaion of he heoreical priors.

13 11 The issue of inappropriae ime aggregaion is a common problem in all empirical ime series work in ha he causal relaionships beween variables may occur wihin he periodiciy of he daa used. This is paricularly likely when annual daa is used, and where he focus is on disequilibrium dynamics, as in his sudy. Rossana and Seaer (1995) argue ha annual daa disor he dynamic effecs of he shocks, relaive o higher periodiciy daa. Some comfor is provided, however, by widespread evidence of rigidiies and ineria in developing counry macro aggregaes and by he findings of oher sudies, which confirm he slow pace of adjusmen o disurbances in many of hese counries. A final cavea is he usual, bu imporan qualificaion abou he qualiy of developing counry daa. The lack of daa in developing counries, and is poor qualiy is widely recognised. This paper uses a limied se of wha is regarded as he more imporan, and hence more widely available and beer qualiy macro daa. Sreamlined heoreical and empirical srucures are used o keep he daa needs wihin he consrains of he available daa. Neverheless, he daa are clearly of poor qualiy and his is an imporan cavea o he resuls. 6. RESULTS 6.1 Impulse Responses The adjusmen of oupu o shocks was examined using impulse response funcions for each counry. Impulse response funcions show he response of he sysem o a one-sandarddeviaion shock o one of he variables. As expeced, impulse response funcions suggesed ha permanen shocks o oupu behave like aggregae supply shocks, wih posiive supply shocks boosing oupu and permanenly lowering prices. 15 This resul is consisen across each of he subsamples, suggesing ha, consisen wih he findings of oher sudies (Hoffmaiser and Roldos, 1996, 1997; Hoffmaiser e al.,1998), supply shocks are he dominan source or permanen movemens in oupu in developing counries. The mos ineresing resul, however, is he effec on oupu of a emporary shock. In he case of a (posiive) nominal shock, oupu and price would be expeced o move upwards ogeher. In he case of a emporary supply shock, however, a rise in prices would be expeced o be associaed wih a emporary fall in oupu. Resuls of he impulse response funcions for each of he five subsample groups, are shown in Figure 2. In a lile more han half of he counries a emporary shock (o he orhogonalised error in he price equaion) leads o a emporary fall in oupu and a rise in prices. This suggess ha emporary shocks o oupu in many developing counries are dominaed by supply shocks wih he effecs dissipaing over a horizon of wo or hree years and in some cases longer. In slighly 15 The sance of moneary policy is a major deerminan of he permanency of price changes. This suggess ha he naure of he exchange rae regime is an imporan consideraion. Mos counries in he sudy had flexible exchange rae arrangemens over he period; mos of hose ha had fixed arrangemens were unable o hold he level of he peg seady for susained periods. Only wo of he 58 counries in he sample mainained a fixed exchange rae a he same level of he full period, and he resuls for hese wo counries were broadly consisen wih he AS-AD framework.

14 12 Figure 2: Oupu Response o a Temporary Shock Africa 4% 2% Oupu 0% -2% -4% Years Asia 2% 1% 0% Oupu -1% -2% Years Middle Eas 8% 4% 0% Oupu -4% -8% Years

15 13 Figure 2: Oupu Response o a Temporary Shock Wesern Hemisphere 2% 1% Oupu 0% -1% -2% Years Wesern Hemisphere (High Inflaion) 4% 2% Oupu 0% -2% -4% Years less han half of he developing counries examined, he shocks behave like more radiional demand-side shocks, leading o a emporary rise in oupu and higher prices. A simple check on he resuls is o examine he correlaion beween inflaion and he esimaed cyclical componen of oupu (Chadha and Prasad, 1993). If emporary movemens in oupu are primarily due o demand shocks, he correlaion beween inflaion and he cyclical movemens in oupu would be expeced o be posiive; if emporary movemens are mainly due o supply shocks i would be expeced o be negaive. The resuls confirm he findings from he impulse response funcions. 6.2 Variance Decomposiion Variance decomposiions for he full sample and for each of he subsamples are repored in Table 1. The variance decomposiion shows he proporion of he variance of he forecas error ha can be aribued o each of he endogenous variables. Each panel shows he percenage of he oupu

16 14 and price variance ha can be aribued o shocks o oupu and price respecively; forecas horizons of one, hree and five years are repored. For he full sample, he resuls sugges ha permanen supply shocks explain a large par of movemens in oupu in developing counries. Supply shocks explain around 88 percen over a one-year forecas horizon, rising o around 94 percen over a five-year horizon; nominal shocks accoun for around 12 percen over a one-year horizon. For prices, permanen real shocks explain only 14 percen over a one-year horizon, rising o around 22 percen over a five yearhorizon; nominal shocks accoun for a large par of he forecas variance in prices. These resuls are broadly similar across he five subsample groups. Table 1: Variance Decomposiion Developing Counries Oupu Prices Years Africa Oupu Prices Years Asia Oupu Prices Years Middle Eas Oupu Prices Years

17 15 Table 1: Variance Decomposiion Wesern Hemisphere Oupu Prices Years Wesern Hemisphere (High) Oupu Prices Years CONCLUSION The resuls sugges ha he decomposiion of oupu ino permanen and ransiory shocks is no sufficien o properly idenify demand and supply shocks for many developing counries. Raher he echnique appears o spli he shocks ino permanen supply shocks, and emporary demand or supply shocks depending on which influence dominaes in a paricular counry. While he permanen/emporary decomposiion is sill useful, clear caveas apply o he inerpreaion of he emporary componen, a leas for developing counries. Since oupu gaps are effecively jus an accumulaion of emporary oupu shocks, he same caveas apply. The resuls sugges ha measures of he oupu gap consruced for developing counries are likely o capure a mix of supply- and demand-induced deviaions in acual oupu from poenial oupu. The dominan influence may, or may no be, demand shocks. Accordingly, he resuls sugges ha oupu gap measures are a less useful guide for assessing inflaionary pressures and for conducing moneary policy in developing counries. Given he ambiguiy of hese measures, oher measures of excess or deficien demand such as money velociy gaps or exernal price gaps may be more useful guides for policy making in developing counries For money gaps see for example Hendry (1995) and (Fung and Kasumovich (1998). For exernal price gaps see Kool and Taom (1994) and Garcia-Herrero and Pradha (1998).

18 16 REFERENCES Ahmed, S. and Park, J.H. (1994) Sources of macroeconomic flucuaions in small open economies, Journal of Macroeconomics, 16(1):1-26. Aghevli, B.B., Khan, M.S. and Moniel, P.J. (1991) Exchange rae policy in developing counries: Some Analyical Issues, Inernaional Moneary Fund Occasional Paper 78. Bayoumi, T. (1992) The effecs of he ERM on paricipaing economies, IMF Saff Papers, 39: Bayoumi, T. and Eichengreen, B. (1994) Macroeconomic adjusmen under Breon Woods and he pos-breon Woods floa: an impulse-response analysis, The Economic Journal, 104: Bergman, M. (1996) Inernaional evidence on he sources of macroeconomic flucuaions, European Economic Review, 40: Bernanke, B. (1986) Alernaive explanaions of he money-income correlaion, Carnegie- Rocheser Conference Series on Public Policy, 25: Blanchard, O.J. and Quah, D. (1989) The dynamic effec of aggregae demand and supply disurbances, American Economic Review, 79: Brajer, V. (1992) Empirical inflaion models in developing counries: sensiiviy o he definiion of inflaion, The Journal of Developmen Sudies, 28(4): Buler, L. (1996) A semi-srucural mehod o esimae poenial oupu: combining economic heory wih a ime-series filer, The Bank of Canada s New Quarerly Projecion Model, Par 4, Technical Repor 77, Bank of Canada, Oawa. Chadha, B., Masson, P.R. and Meredih, G. (1992) Models of inflaion and he coss of disinflaion, Inernaional Moneary Fund Saff Papers, 39(2): Chadha, B. and Prasad, E. (1993) Inerpreing he cyclical behaviour of prices, IMF Saff Papers, 40(2): Cogley, T. and Nason, J. (1995) The effecs of he Hodrick-Presco filer on rend and difference saionary ime series: implicaions for business cycle research, Journal of Economic Dynamics and Conrol, 19: Conway, P. and Hun, B. (1997) Esimaing poenial oupu: a semi-srucural approach, Reserve Bank of New Zealand Discussion Paper G97/9. de Brouwer, G. (1998) Esimaing oupu gaps, Reserve Bank of Ausralia Research Discussion Paper 9809.

19 17 De Masi, P. R. (1997) IMF esimaes of poenial oupu: heory and pracice, IMF Working Paper 97/177, Inernaional Moneary Fund, Washingon. DeSerres, A. and Guay, A. (1995) The selecion of he runcaion lag in srucural vars (or vecms) wih long-run resricions, Bank of Canada Working Paper DeSerres, A., Guay, A. and S-Aman, P. (1995) Esimaing and projecing poenial oupu using srucural VAR mehodology: he case of he Mexican economy, Bank of Canada Working Paper Dickey, D. and Fuller, W.A. (1979) Disribuion of he esimaes for auoregressive ime series wih a uni roo, Journal of he American Saisical Associaion, 74: Dickey, D. and Panula, S. (1987) Deermining he order of differencing in auoregressive processes, Journal of Business and Economic Saisics, 15: Dupasquier, C., Guay, A. and S-Aman, P. (1997) A comparison of alernaive mehods of esimaing poenial oupu and he oupu gap, Bank of Canada Working Paper Gali, J. (1992) How well does he IS-LM model fi poswar U.S. daa?, The Quarerly Journal of Economics, 107(2): Garcia-Herrero, A. and Pradhan, M.J. (1998) The domesic and foreign price gaps in he P-sar model, IMF Working Paper WP/98/64, Inernaional Moneary Fund, Washingon. Giorno, C.P., Richardson, P., Roseveare, D. and van den Noord. P. (1995) Poenial oupu, oupu gaps and srucural budge balances, OECD Economic Sudies, 24: Guay, A. and S-Aman, P. (1996) Do mechanical filers provide a good approximaion of business cycles? Bank of Canada Technical Repor 78. Faus, J. and Leeper, E. (1997) When do idenifying resricions give reliable resuls, Journal of Business and Economic Saisics, 15: Fung, B.S.C. and Kasumovich, M. (1998) Moneary shocks in he G-6 counries: is here a puzzle? Journal of Moneary Economics. Hendry, S. (1995) Long-run demand for M1, Bank of Canada Working Paper Heson, A. (1994) A brief review of some problems in using naional accouns daa in level of oupu comparisons and growh sudies, Journal of Developmen Economics, 44(1): Hoffmaiser, A.W. and Roldos, J.E. (1996) The sources of macroeconomic flucuaions in developing counries: Brazil and Korea, IMF Working Paper 96/20, Inernaional Moneary Fund, Washingon.

20 18 Hoffmaiser, A.W. and Roldos, J.E. (1997) Are business cycles differen in Asia and Lain America? IMF Working Paper 97/9, Inernaional Moneary Fund, Washingon. Hoffmaiser, A.W., Roldos J.E. and Wickham, P. (1996) Macroeconomic flucuaions in Sub- Saharan Africa, Inernaional Moneary Fund Saff Papers, 45(1): Inernaional Moneary Fund (2000) Inernaional Financial Saisics, Washingon, D.C., January. Keaing, J.W. and Nye, J.V. (1998) Permanen and ransiory shocks in real oupu: esimaes from nineeenh-cenury and poswar economies, Journal of Money, Credi and Banking, 30(2): Kool, C. and Taom, J.A. (1994) The P-sar model in five small economies, Federal Reserve Bank of S. Louis, May/June. Kwiakowski, D., Phillips P., Schmid, P. and Shin, Y. (1992) Tesing he null hypohesis of saionariy agains he alernaive of a uni roo, Journal of Economerics, 54: Lasrapes, W.D. (1992) Sources of flucuaions in real and nominal exchange raes, Review of Economics and Saisics, 20: Nickell, S. (1988) The supply side and macroeconomic modeling, in R.C. Bryan e al., Empirical Macroeconomics for Inerdependen Economies, Washingon, D.C., Brookings Insiuion, Phillips, P.C.B. (1995) Impulse response and forecas error variance asympoics in nonsaionary VARs, Working Paper 1102, Cowles Foundaion. Phillips, P.C.B. and Perron, P. (1988) Tesing for a uni roo in a ime series regression, Biomerica, 75: Rossana, R.J. and Seaer, J.J. (1995) Temporal aggregaion and economic ime series, Journal of Business and Economic Saisics,13(4): Said, S.E. and Dickey, D.A. (1984) Tesing for uni roos in auoregressive-moving average series of unknown order, Biomerica, 71: Shapiro, M.D. and Wason, M.W. (1988) Sources of business cycle flucuaions, NBER Working Paper Sims, C. (1986) Are forecasing models usable for policy analysis? Federal Reserve Bank of Minneapolis Quarerly Review, Winer, S-Aman, P. and van Norden, S. (1997) Measuremen of he oupu gap: a discussion of recen research a he Bank of Canada, in Moneary Policy and he Inflaion Process, BIS Conference Papers, 4, Basle, 1-38.