Wespac $ Insiuional Bank April 2002 Explaining he NZ-Ausralian exchange rae occasional paper Paul Conway and Richard Franulovich Wespac Insiuional Bank PO Box 691 Wellingon New Zealand Phone (644) 381-1414 paul_conway@wespac.co.nz A division of Wespac Banking Corporaion ABN 33 007 457 141
Wespac $ Insiuional Bank Explaining he NZ-Ausralian exchange rae, April 2002 Execuive summary In his paper we invesigae wheher 'economic fundamenals' have a significan influence on he value of he New Zealand dollar relaive o he Ausralian dollar (NZD/AUD). To do his we esimae an economic model ha capures he hisorical relaionship beween he real (i.e., he inflaion adjused) NZD/AUD exchange rae and he following relaive economic variables: Produciviy growh House prices (o capure relaive domesic inflaion pressures) Ineres rae differenials Commodiy prices Curren accoun balances (o capure differen savings/invesmen balances) We find ha he relaionship beween hese fundamenals and he real NZD/AUD exchange rae is consisen wih he predicions of economic heory and significan in a saisical sense. On he basis of his relaionship we calculae a measure of he 'equilibrium' (or fair value) NZD/AUD exchange rae. This racks he acual exchange rae reasonably closely, indicaing ha he economic fundamenals in our model successfully explain he broad rends in he currency over he las fifeen or so years. From he perspecive of he currency union debae, his finding indicaes ha he NZD/AUD exchange rae acs as an effecive macroeconomic shock absorber, raher han a source of shock. If he NZD/AUD exchange rae were superseded by some form of currency union hen is shock absorber properies would be los. To compensae, real resources (i.e., labour and capial) would need o move more beween he wo counries in response o economic shocks. There would also be more variabiliy in relaive inflaion raes across he wo economies. Our esimaed model also explains reasonably shor-erm exchange rae developmens. On a quarerly basis he effec of changes in he economic fundamenals on he exchange rae are generally as we would expec. I follows ha if we have a good undersanding of how he New Zealand and Ausralian economies are likely o evolve, we sand a good chance of correcly forecasing he fuure rend in he exchange rae. We es he abiliy of our model o forecas he exchange rae and find ha i comprehensively ouperforms an alernaive naïve forecasing echnique ha doesn' accoun for economic fundamenals (a 'random walk' model). 1
Wespac $ Insiuional Bank Explaining he NZ-Ausralian exchange rae, April 2002 1) Inroducion In his paper we develop a model of he real (inflaion adjused) NZD/AUD exchange rae. Our moivaion is wofold. Firs, we wan o es wheher heories of exchange rae deerminaion provide a useful framework for explaining he pah of he NZD/AUD exchange rae since boh currencies have been floaing. If economic fundamenals do play a role, our model will capure he relaionships involved. We will hen be able o see how imporan he fundamenals are in comparison o speculaive behaviour in currency marke as drivers of he NZD/AUD exchange rae. If he fundamenals do explain he NZD/AUD, his suggess i acs more as a macroeconomic shock absorber, raher han source of shock. This is useful informaion in undersanding he cos of a currency union, whereby macro flexibiliy would be reduced wih a single currency. Our second and relaed moivaion for building an exchange rae model is o develop a forecasing ool. Because i quanifies he relaionships involved, an exchange rae model can serve as a useful bridge beween our macro forecass for New Zealand and Ausralia, and he NZD/AUD exchange rae. This will give us an idea of he exen of any over or under valuaion in he currency and he mos likely direcion of fuure movemens. The paper is srucured as follows. In secion 2 we ouline a heoreical framework for hinking abou he fundamenal deerminans of a currency's value. We use his framework o jusify he choice of economic variables ha we include in our model. In secion 3 we give a brief and inuiive explanaion of he modelling echnique we use. The resuls of esimaing our model are given in secion 4. We es he abiliy of our model o forecas he NZD/AUD exchange rae in secion 5 and offer some concluding commens in secion 6. 2) Exchange rae drivers The New Zealand dollar was floaed in March 1985. The floa has been 'clean' in ha he RBNZ has no direcly inervened in currency markes. However, he exchange rae has played an imporan role in New Zealand's moneary policy. Over he early 1990s he RBNZ relied exensively on he direc effec of he exchange rae on impor prices o influence inflaion. This implied ha ineres raes were ofen se wih he objecive of influencing he near-erm level of he exchange rae. In more recen imes he RBNZ has been less concerned wih he direc influence of he exchange rae on inflaion. However, he indirec effec on inflaion via inernaional demand for New Zealand's goods and services coninues o be an imporan consideraion in policy decisions 1. The Ausralian dollar was floaed in December 1983. The RBA have, from ime o ime, inervened direcly in currency markes ha is, brough and sold Ausralian dollars so as o influence is value. In principle, he moivaion for inervenion has been o calm markes ha are hreaening o become disorderly. In pracise, inervenion has been used o help reverse perceived overshoos in he currency and, on some occasions, o give moneary policy greaer room for manoeuvre. Alhough cenral-bank inervenion may affec exchange raes a he margin, economic fundamenals are likely o be a far more significan influence over he medium erm. In his secion we ouline he range of economic facors ha are likely o affec he value of he currency and he naure of he relaionships involved. We use his o jusify he choice of explanaory variables ha we include in our model of he NZD/AUD exchange rae. 1 For a discussion on he role of he exchange rae in moneary policy in New Zealand see Orr, Sco, and Whie (1998). 2
Wespac $ Insiuional Bank Explaining he NZ-Ausralian exchange rae, April 2002 Exchange rae deerminaion in he long run Over he long run he nominal exchange rae should adjus so ha he prices of goods and services are he same across counries. The basic idea is ha he purchasing power of a currency, in erms of he goods ha can be bough wih i, should be he same regardless of where i is spen. This heory of exchange rae deerminaion is known as purchasing power pariy (PPP). The heoreical raionale behind PPP is basically one of arbirage. If he same good is priced differenly in differen counries hen inernaional demand will swich o he cheaper good. This will end o increase is domesic price and appreciae he currency of he counry in which i is sold. Over ime, he forces of compeiion if allowed will equalise prices across counries. Arbirage in capial invesmen will also end o resul in PPP-ype behaviour. For example, if a counry's exchange rae is such ha he local cos of labour in foreign dollars is low relaive o oher counries, hen producion will be araced o his base. Producion capaciy will end o move from he 'overvalued' o he 'undervalued' economy. This effec will ac o bring he exchange rae back o is PPP equilibrium ha is he level ha equaes prices across counries. The empirical evidence on PPP has been mixed. A number of researchers have found ha nominal exchange raes do end o exhibi PPP-like behaviour over very long periods of ime. This conclusion is based on research ha uses long runs of hisorical ime-series daa or panel daa ses for a number of differen counries. The nominal NZD/AUD exchange rae and he level of consumer prices in Ausralia relaive o ha in New Zealand are graphed in figure 1a. On he face of i, he nominal exchange rae does appear o display PPP-ype behaviour over he long run. From 1960 unil he mid-1980s he exchange rae seadily depreciaed as inflaion in NZ was persisenly higher han in Ausralia. This mean ha he reduced purchasing power of he kiwi dollar in New Zealand was broadly mached by a fall in is purchasing power when convered ino Ausralian dollars. Since he lae-1980s consumer price inflaion in NZ has been more similar o ha in Ausralia and so he nominal TWI has cycled around a more consan mean. The real exchange rae is calculaed by simply subracing he (log) of relaive prices from he (log) of he nominal exchange rae ha is, he difference beween he wo lines in figure 1a. Because he real exchange rae accouns for price differenials across counries, i is a measure of an economy's compeiiveness in inernaional markes 2. The real NZD/AUD exchange rae has cycled around a consan long-run average a leas since he 1960s (figure 1b). Again, his suggess ha he heory of PPP provides a reasonable descripion of long-run movemens in he NZD/AUD exchange rae. Over shorer ime horizons he exchange rae can deviae from is PPP equilibrium. This can be seen from he deviaion of he real exchange rae from is long-run average in figure 1b. According o a review aricle by Ken Rogoff (1996), he consensus amongs economiss is ha deviaions of exchange raes from heir PPP level damp ou a a rae of roughly 15% per year. Pu anoher way, hese deviaions have a 'half life' of hree o five years. This is painfully slow. There are a number of reasons why PPP doesn' hold over shorer spans of ime. Alhough inernaional markes are becoming more inegraed, hey are sill relaively segmened and have large rading fricions across a broad range of goods. These fricions are due o facors such as ransporaion coss, ariffs, non-ariff barriers, and pricing o marke behaviour by firms. A he exreme, a large proporion of goods and services are no raded on inernaional markes. These fricions make i difficul o arbirage away inernaional price differenials and mean ha exchange raes are only loosely ied o heir PPP equilibrium. Oher macroeconomic fundamenals ha reflec business cycle pressures can cause he real exchange rae o deviae from is PPP equilibrium. Fads and fashions in currency markes will also cause exchange rae o deviae from some noion of equilibrium. The key o explaining movemens in he exchange rae lies in undersanding hese forces. 2 A depreciaion (appreciaion) of he real exchange rae is synonymous wih an increase (decrease) in inernaional compeiiveness. 3
Wespac $ Insiuional Bank Explaining he NZ-Ausralian exchange rae, April 2002 Figure 1a: Nominal NZD/AUD exchange raes relaive prices 140 130 120 AU/NZ price level NZD/AUD exchange rae 110 100 90 80 70 60 1960 1964 1968 1972 1976 1980 1984 1988 1992 1996 2000 Figure 1b: Real Exchange Rae 130 120 110 100 90 80 70 60 1960 1964 1968 1972 1976 1980 1984 1988 1992 1996 2000 Deviaions from PPP A differen rae of produciviy growh across counries is one reason why he real exchange raes can move away from is PPP equilibrium. PPP relaes o arbirage in goods ha are raded on inernaional markes. Bu he real exchange rae is measured using price indices ha include non-raded goods. As such, he real exchange rae can move away from is PPP equilibrium because differen raes of produciviy growh affec prices for non-raded goods across counries. In shor, economies wih high produciviy growh will end o have appreciaing real exchange raes 3. While differen raes of produciviy growh are likely o explain medium o long-run movemens in he real exchange rae, shorer-run cyclical movemens are more likely o reflec he exen of relaive disequilibrium in he economy. In oher words, he real exchange rae may ac as a ype of 'safey valve' and move o offse cyclical pressures in he economy. The real exchange rae can play his role because of he effec i has on an economy's ne inernaional rade. 3 This is known as he Balassa-Samulson effec. For a good discussion of he Balassa-Samulson effec in he New Zealand conex see Brook and Hargreaves (2001). 4
Wespac $ Insiuional Bank Explaining he NZ-Ausralian exchange rae, April 2002 For example, if domesic demand increases relaive o anoher counry for some reason, hen he real exchange rae should appreciae. This would cause ne expors o fall, offseing he increase in domesic demand and resoring he balance beween aggregae supply and demand in he economy. Accordingly, if an economy is experiencing a business cycle upswing relaive o is rading parners, he real exchange rae should end o appreciae. 'Porfolio balance' ype behaviour in financial markes may lead o a posiive relaionship beween real exchange rae changes and ineres rae differenials across counries. The heory of PPP is concerned wih arbirage in he inernaional markes for goods and services. However, he movemen of financial capial from counry o counry in search of he highes reurn is a far larger source of foreign exchange ransacions. For example, if higher ineres raes increase he expeced reurn from invesing in New Zealand hen financial capial will be araced o New Zealand, puing upward pressure on he real exchange rae 4. Furher, in his age of inflaion-argeing cenral banks ineres raes end o reflec he sae of he business cycle and are anoher indicaor of relaive disequilibrium in he economy. Because New Zealand and Ausralia are boh small open economies ha are highly dependen on exporing commodiies, commodiy prices are also likely o effec he real value of he currency. If prices (expressed in world price erms) for New Zealand's commodiies increase relaive o Ausralia's, hen he real NZD/AUD exchange rae may end o appreciae. This would reconcile he forces of supply and demand in boh economies and resore inernal balance. Also, higher prices for New Zealand's commodiies increases he reurn on New Zealand's asses and herefore heir value. This should also lead o an appreciaion in he real value of he currency. As well as faciliaing inernal balance, he real exchange rae can also be (equivalenly) hough of as a mechanism for ensuring exernal balance beween he domesic economy and he res of he world. For example, if he level of savings in an economy is insufficien o finance is invesmens needs, hen he sock of ne foreign liabiliies wha a counry owes he res of he world will seadily increase. To finance his increase in foreign deb ne expors will need o increase, hence he need for a currency depreciaion. The ne balance of savings and invesmen flows wihin an economy is refleced by he balance on he curren accoun. Consequenly, we would expec a posiive relaionship beween a counry's (relaive) curren accoun and is real exchange rae. In summary, economic heory suggess he following fundamenals may explain deviaions of he real exchange rae from is PPP equilibrium: Differen raes of produciviy growh across counries Relaive disequilibrium (business cycle pressures) in he economy Ineres rae differenials Relaive commodiy prices Differen savings/invesmen balances (curren accoun balance). The daa We model he real NZD/AUD exchange rae, which we calculae using producer price indices. By choosing o model he real exchange rae we are effecively imposing PPP and assuming insananeous adjusmen o nominal shocks. However, his isn' really a major issue given ha New Zealand and Ausralia have experienced broadly similar raes of inflaion over he sample period of our sudy. As can be seen from figure 2a, he real and nominal NZD/AUD exchange raes have followed similar pahs since he mid-1980s. 4 Noe ha exbook definiions of uncovered ineres pariy (UIP) conjecure a negaive relaionship beween he expeced exchange rae and ineres rae differenial. According o UIP if ineres raes are higher in New Zealand han Ausralia hen invesors mus be expecing he NZD/AUD exchange rae o depreciae so as o equalise reurns across he wo counries. We are modelling he acual exchange rae, no he expeced exchange rae, and herefore posulae a posiive relaionship consisen wih a capial flow noion of exchange rae dynamics. 5
Wespac $ Insiuional Bank Explaining he NZ-Ausralian exchange rae, April 2002 We use oupu per person employed in New Zealand relaive o Ausralia o capure produciviy differences beween he wo economies (figure 2b). As well as capuring he impac of longer-run produciviy differenials on he real exchange rae his variable will also reflec o an exen he relaive business cycle posiions of New Zealand and Ausralia. Our primary measure of disequilibrium or business cycle pressures across he wo economies is relaive house prices. House prices are seen as a good predicor of he real exchange rae over he business cycle. Grimes e al (2000) noe a close cyclical relaionship beween New Zealand's real exchange rae and house prices. In a recen paper Bowden (2001) oulines heoreical reasons as o why his migh be he case. In essence, changes in asse prices are driven be he same forces ha aler he balance beween non-raded and raded goods prices, and hence he real exchange rae. I follows ha asse prices ac as a signal for curren or prospecive changes in he real exchange rae. Figure 2c shows he igh relaionship beween he real NZD/AUD exchange rae and relaive house prices. In choosing our measure of relaive disequilibrium in he wo economies we ried a range of alernaive indicaors such as relaive GDP growh and oupu gaps. However, none of hese alernaives could explain cyclical swings in he real exchange rae as well as relaive house prices. Furher, house price daa is available on a more imely basis han GDP daa and isn' prone o significan revisions. We use he spread on real 10-year bond yields in New Zealand and Ausralia as our measure of he ineres rae differenial (figure 2d). Real raes are calculaed as he nominal rae minus acual producer price inflaion. By using real 10-year bond yields we also capure medium-erm facors, such as he risk premium, in our model. Commodiy price effecs are capured by he raio of New Zealand o Ausralian commodiy prices (figure 2e). The relaive savings/invesmen balance beween he wo counries is capured by he relaive curren accoun balance (figure 2f). Furher deails on he daa sources can be found in he appendix. 0.95 Figure 2a: Real and Nominal NZD/AUD Exchange Rae 10.0 Figure 2b: Relaive Produciviy NZ-Aus 0.00 0.90 0.85 0.80 0.75 0.70 Nominal exchange rae 0.65 Real exchange rae 0.60 Mar-85 Mar-87 Mar-89 Mar-91 Mar-93 Mar-95 Mar-97 Mar-99 Mar-01 9.5-0.05 9.0 8.5-0.10 8.0 7.5-0.15 7.0-0.20 Real exchange rae (LHS) 6.5 Relaive produciviy (RHS) 6.0-0.25 Mar-85 Mar-87 Mar-89 Mar-91 Mar-93 Mar-95 Mar-97 Mar-99 Mar-01 10.0 9.5 9.0 8.5 8.0 7.5 Figure 2c: Relaive House Prices NZ-Aus 7.0 2.1 6.5 Real exchange rae (LHS) 2.0 NZ/AU house prices (RHS) 6.0 1.9 Mar-85 Mar-87 Mar-89 Mar-91 Mar-93 Mar-95 Mar-97 Mar-99 Mar-01 2.7 2.6 2.5 2.4 2.3 2.2 10.0 9.5 9.0 8.5 8.0 7.5 7.0 6.5 Figure 2d: 10-year Bond Yields NZ-Aus Real exchange rae (LHS) 10-year bond spread (RHS) 6.0-4 Mar-85 Mar-87 Mar-89 Mar-91 Mar-93 Mar-95 Mar-97 Mar-99 Mar-01 10 8 6 4 % 2 0-2 6
Wespac $ Insiuional Bank Explaining he NZ-Ausralian exchange rae, April 2002 10.0 9.5 9.0 8.5 8.0 7.5 7.0 6.5 Figure 2e: Relaive Curren Accoun NZ-Aus Real exchange rae (LHS) Relaive curren accoun (RHS) 6.0-8 Mar-85 Mar-87 Mar-89 Mar-91 Mar-93 Mar-95 Mar-97 Mar-99 Mar-01 8 6 4 2 0-2 -4-6 % GDP 10.0 9.5 9.0 8.5 8.0 7.5 Figure 2f: Relaive Commodiy Prices NZ-Aus 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 7.0-0.05 Real exchange rae (LHS) -0.10 6.5 Relaive commodiy prices (RHS) -0.15 6.0-0.20 Mar-85 Mar-87 Mar-89 Mar-91 Mar-93 Mar-95 Mar-97 Mar-99 Mar-01 3) The modelling echnique To capure he dynamics of he real NZD/AUD exchange rae we use a 'vecor error correcion' (VECM) model. This echnique was developed by he Swedish economerician Soren Johansen, and has been used by a number of economiss o model exchange raes 5. There are wo appealing aspecs of his echnique. Firs, i allows us o model he exchange rae as par of a sysem. Wihin ha sysem each variable has is own separae equaion. In his way he model capures all he inerdependencies beween he variables. For example, an increase in domesic demand in NZ will no only influence he exchange rae direcly, bu will also affec prices, ineres raes, and he curren accoun. In urn, changes in hese variables will have an addiional indirec impac on he exchange rae. The VECM echnique allows us o capure all of hese influences boh he direc and indirec effecs. In conras, single equaion models only capure he direc effecs and can miss poenially imporan dynamics. The oher appealing aspec of he VECM approach is ha i accouns for boh long-run and shor-run influences in he same model. Long-run influences are capured by looking for equilibrium relaionships beween he variables of he model. Broadly speaking, an equilibrium relaionship exiss if he variables influence each oher over a long period of ime. If such a relaionships exiss, he variables are said o be 'coinegraed'. As an analogy for coinegraion consider he case of a drunk couple walking home afer a nigh on he own. Despie he appearance of randomness, he movemen of each individual is influenced by he oher. If hey sagger oo far apar, an overriding force brings hem back ogeher and each individual arrives home a approximaely he same ime. If he variables in our VECM model share a long-run equilibrium relaionship hen we can use his o help explain heir shor-run dynamics. In much he same way as he drunk couple end o move ogeher hrough he srees, he variables of he model will move in such a way so as o resore equilibrium in he long run. For example, if he exchange rae has recenly deviaed from he equilibrium pah hen we can expec i o gradually drif back owards equilibrium over ime. To capure he shor-run dynamics in he daa we include lags of he firs difference of each variable in he equaions of he sysem. In his way quarer on quarer change in each of he variables is influenced by recen changes in iself and he oher variables of he model. In erms of he drunk couple, hese effecs are analogous wih a quick sleep in a bush, a rip o he burger bar, or any oher emporary disracion on he way home. A mahemaical descripion of he model and he ess for coinegraion are given in he appendix. 5 Examples include: MacDonald (1997), MacDonald and Taylor (1993, 1994) and MacDonald and Marsh (1996). Auhors who have modelled he NZ exchange rae using his echnique include: Lee (2001), and Sephens and Choy (2001). For one of he original papers on coinegraion see Johansen (1998). 7
Wespac $ Insiuional Bank Explaining he NZ-Ausralian exchange rae, April 2002 4) Esimaion resuls The long-run equilibrium exchange rae We esimae he model over he sample period 1986Q2 o 2001Q3. Tess on he model indicae srong evidence of coinegraion. In oher words, he variables of he model do appear o share a long-run equilibrium relaionship. So over long periods of ime he variables do no move independenly of each oher. Insead, he long-run rends are linked. We assess he naure of his linkage by looking a he coefficiens in he equilibrium relaionship. If he equilibrium relaionship is meaningful, hen we would expec he linkages beween he variables o be consisen wih he economic heory oulined in secion 2 and saisically significan. Esimaion of he model reveals ha all he coefficiens in he long-run relaionship have he expeced posiive sign. However, ess of significance reveal ha he coefficien on relaive ineres raes is no saisically differen from zero. This indicaes ha he ineres rae differenial does no play a significan role and can be excluded from he long-run equilibrium relaionship. This leaves us wih a coinegraing relaionship ha can be wrien as 6 : e =µ+0.339(cp nz -cp au )+0.556(dd nz -dd au )+0.514(ca nz - ca au (0.047) (0.028) (0.182) (0.110) )+0.435(p nz - p au )+ε The variable e sands for he real NZD/AUD exchange rae, cp is commodiy prices, dd is domesic demand, ca is he curren accoun, and p is labour produciviy. All of he variables are in logs. Superscrips denoe he relevan counry. µ is a consan and ε is he equilibrium error erm. The numbers below he coefficiens are sandard errors. All of he coefficiens are saisically significan and a plausible magniude. So no only is here srong evidence of a long-run relaionship beween he variables of he model, bu he naure of his relaionship is consisen wih economic heory. Because he rends of he variables are linked (coinegraed), he dynamic pahs of he variables will be influenced by he curren deviaion from he equilibrium relaionship. In oher words, here is a connecion beween he change in he variables and he deviaion from equilibrium. The srengh of his connecion is capured by he 'speed of adjusmen' parameer. In he exchange rae equaion, he speed of adjusmen parameer is -0.6. This has he expeced negaive sign and is highly significan in a saisical sense wih a -saisic of 3.6. This means ha he equilibrium relaionship is an imporan deerminan of quarerly changes in he real exchange rae. If he exchange rae wanders away from equilibrium, i will end o drif back over ime. In fac, he speed of adjusmen parameer is reasonably large and indicaes ha deviaion of he exchange rae from equilibrium are relaively shor lived, lasing only a few quarers on average. In figure 3 we plo our measure of he equilibrium NZD/AUD exchange rae, derived from our model, and he acual exchange rae. For ease of inerpreaion we have convered he real exchange rae ha is used in our model ino he nominal value. The firs poin o noe is ha he equilibrium exchange rae racks he acual exchange rae reasonably closely and maches he broad swings ha have occurred over he sample period. This indicaes ha he economic fundamenals included in our model explain rend movemens in he NZD/AUD exchange rae over he las fifeen years of so reasonably successfully. 6 This equaion could be wrien wih any of he variables on he lef-hand side. However, because we are ineresed in he exchange rae, we make his he subjec of he equaion. 8
Wespac $ Insiuional Bank Explaining he NZ-Ausralian exchange rae, April 2002 I also raises he issue of wha exacly do we mean by he erm 'equilibrium exchange rae'. This is a erm ha mus always be inerpreed in conex. For example, a one exreme i could be argued ha because he exchange rae is coninually deermined by he forces of supply and demand, i is always a an 'equilibrium' value. A he oher exreme, he exchange rae can move away from PPP-based measures of 'equilibrium' for years on end. The equilibrium measure of he NZD/AUD exchange rae derived from our model and shown in figure 3 is clearly in beween hese wo exremes. Our equilibrium measure can be hough of as he exchange rae ha would prevail if deermined solely by he economic fundamenals included in he model. These fundamenals reflec he effec of he business cycle and oher shor-run economic perurbaions. As such, our measure of he equilibrium NZD/AUD exchange rae also reflecs his ype of volailiy. This makes i a reasonably shor-erm, emporary equilibrium, raher han one ha is consisen wih he noion of a 'seady sae' in he economy. Because our measure of equilibrium reflecs he economic fundamenals, any deviaion of he acual exchange rae from equilibrium can be hough of as largely he resul of asse bubbles and oher speculaive behaviour (and any fundamenals omied from he model). According o our model, he peak of he sharp appreciaion in he exchange rae in 1988 was speculaive in naure. Also, he appreciaion of he currency in 1993 was premaure on he basis of he fundamenals and pre-emped he ensuing increase in he equilibrium exchange rae. Since he lae 1990s he depreciaion of he currency has been largely jusifiable on he basis of he economic fundamenals capured in our model. Alhough speculaive facors can move he exchange rae in he shor run, our model indicaes ha rend movemens in he NZD/AUD exchange rae are ulimaely deermined by economic fundamenals. This means ha he NZD/AUD exchange rae has moved o offse relaive cyclical pressures in he wo economies. From he perspecive of he currency union debae his highlighs a poenial cos, in erms of a loss of an effecive shock absorber, ha would resul from forming a currency union wih Ausralia. The resuls from esimaing our model also suggess ha if we have a good undersanding of he relaive economic fundamenals in NZ and Ausralia, we should be reasonably successful a predicing rend movemens in he NZD/AUD exchange rae. 0.95 Figure 3: Acual and Equilibrium Exchange Rae 0.90 0.85 0.80 0.75 0.70 0.65 Nominal exchange rae Equilibrium nominal exchange rae 0.60 Mar-85 Mar-87 Mar-89 Mar-91 Mar-93 Mar-95 Mar-97 Mar-99 Mar-01 9
Wespac $ Insiuional Bank Explaining he NZ-Ausralian exchange rae, April 2002 Shor-run dynamics The las secion was concerned wih how well he model capures he long-run rends in he NZD/AUD exchange rae. In his secion we assess he shor-run dynamics of he model. In oher words, we wan o see how well he model capures he quarer on quarer relaionships beween he exchange rae and he oher variables of he model. We use wo differen echniques o do his. Firs, we assess shor-run dynamics by calculaing 'impulse response funcions' (IRFs). IRFs are calculaed by subjecing a variable of he model o an unexpeced 'shock' and seeing how he oher variables reac. Using his echnique we can see, for example, how he real exchange rae responds o an unexpeced increase in New Zealand's commodiy prices. If he reacions are broadly consisen wih he predicions of economic heory we can be conclude ha he model has sensible shor-run properies. The response of he real NZD/AUD exchange rae o unexpeced increases in he oher variables of he model is graphed in figure 4. As we would expec, he response is generally posiive in all cases. Shocks o relaive house prices and commodiy prices in paricular have a srong influence on he real NZD/AUD exchange rae. Shocks o he ineres rae differenial exer a minimal effec. The second way of assessing he shor-run properies of he model involves looking a how well he predicions of he model 'fi' he daa. In oher words, we compare he model's predicions for quarerly changes in he exchange rae wih he acual exchange rae daa. Figure 5 plos he acual change in he real NZD/AUD exchange rae hrough ime and he fied value derived from our model. The model does a reasonably good job of maching he daa and correcly picks he direcion of quarerly changes in he exchange rae 80% of he ime. All in all, he shor-run dynamics of he model appear o be sensible. 2.5% Figure 4: Shor-run response of he exchange rae o various shocks 2.0% 1.5% Ineres rae shock Curren accoun shock Produciviy shock Domesic demand shock Commodiy price shock % change 1.0% 0.5% 0.0% -0.5% 1 2 3 4 5 6 7 8 9 10 10
Wespac $ Insiuional Bank Explaining he NZ-Ausralian exchange rae, April 2002 Figure 5: Acual and fied values from full model 0.08 0.06 0.04 0.02 0.00-0.02-0.04-0.06-0.08-0.10 Fied Value Acual -0.12 Mar-87 Mar-89 Mar-91 Mar-93 Mar-95 Mar-97 Mar-99 Mar-01 5) Forecasing abiliy Fiing he daa is an 'in sample' es in ha i uses all available informaion o forecas pas movemens in he exchange rae. A more difficul es for our model is o see how well i forecass 'ou of sample'. To conduc an ou of sample es we esimae he model over a 'rolling' sample period. Specifically, we begin by esimaing he model over he period 1987Q2 o 1997Q1. We hen use his version of he model o forecas he exchange rae over he nex eigh quarers ha is, 1997Q2 hrough o 1999Q1. We hen re-esimae he model using an addiional quarer of daa and calculae anoher se of exchange rae forecass. This process is repeaed unil he end of he sample period, 2001Q2. We compare his se of exchange rae forecass wih he acual exchange rae daa and calculae measures of forecas error a each ime horizon 7. These give us a good idea of he forecasing abiliy of he model. To evaluae he model we compare is forecasing abiliy wih ha of an alernaive mehod of predicing exchange rae movemens. The alernaive ha we use is a 'random walk' exchange rae forecas. The inuiion behind his forecasing mehod is ha in he nex quarer he exchange rae will equal wha i is oday plus or minus a random componen. The random walk model is a good parable for how currency raders pick exchange rae movemens and a common meric by which exchange rae model are assessed 8. The raio of forecas errors from our model o forecas errors from he random walk model is given in able 2. Over all ime horizons he raio is less han one indicaing ha he forecas errors from our model are smaller han hose of he random walk model. In oher words, our model is clearly more proficien han a random walk a forecasing he NZD/AUD exchange rae. 7 Specifically, we calculae roo mean squared error saisics. 8 This approach was used in a famous paper by Meese and Rogoff (1883). MacDonald (1997) also uses his es. 11
Wespac $ Insiuional Bank Explaining he NZ-Ausralian exchange rae, April 2002 Table 1: Raio of forecas errors our model/random walk Quarers ahead Raio 1 0.564227 2 0.672784 3 0.496917 4 0.746577 5 0.600053 6 0.495325 7 0.547338 8 0.698036 6) Conclusions The overriding conclusion ha we draw from his work is ha economic fundamenals give good insigh ino he behaviour of he NZD/AUD exchange rae. In paricular, our model does a credible job of explaining he broad rends in he real NZD/AUD exchange rae over he las fifeen or so years. This is clearly useful when i comes o forecasing he NZD/AUD exchange rae. If we have a good undersanding of how he fundamenals of Ausralia and New Zealand are likely o unfold, we sand a good chance of correcly picking he fuure rend in he exchange rae. I is imporan o noe ha his is no always he case wih exchange rae models. Recen aemps a modelling he NZD/USD and AUD/USD exchange raes find ha he New Zealand and Ausralian dollars are currenly significanly undervalued on he basis of economic fundamenals 9. We surmise ha our model is reasonably successful a explaining rends in he NZD/AUD exchange rae because srong marke senimen owards he US dollar is no an issue. Because his disoring influence washes ou, economic fundamenals play a larger role and he behaviour of he exchange rae becomes more predicable. Unforunaely, he old adage sill applies: "All models are wrong, bu some are useful". The only hing ha is cerain abou our model, or any economic model, is ha is quarerly exchange rae forecass will be wrong o some degree. However, he model we presen in his paper provides a framework for considering inerdependencies beween macroeconomic fundamenals and he NZD/AUD exchange rae. 9 See, for example, Sephens and Choy (2001). 12
Wespac $ Insiuional Bank Explaining he NZ-Ausralian exchange rae, April 2002 References Bjorksen, N. (2001), "The curren sae of New Zealand moneary union research", Reserve Bank of New Zealand Bullein, December 2002. Brook, A. and Hargreaves, D. (2001), "PPP-based analysis of New Zealand's equilibrium exchange rae", Reserve Bank of New Zealand Discussion Paper, DP2001/01. Enders, W. (1995), Applied Economeric Time Series, John Wiley and Sons, inc. New York. Grimes, A. and Holmes, F. wih Bowden (2000), An ANZAC dollar? Currency union and business developmen, Insiue for Policy Sudies, Vicoria Universiy of Wellingon. Johansen, S. (1988), "Saisical analysis of coinegraing vecors", Journal of Economic Dynamics and Conrol, vol 12, 231-254. Johansen, S. and Juselius, K. (1990), "Maximum likelihood esimaion and inference on coinegraion wih applicaion o he demand for money." Oxford Bullein of Economics and Saisics, 52, 169-209. Lee, M. (2001), "Real exchange rae deerminaion and generalised purchasing power pariy: New Zealand and Ausralia. Paper presened a he New Zealand Economis's Associaion annual conference. 2001. MacDonald, R. (1997), "Wha deermines real exchange raes? The long and shor of i." IMF Working Paper, WP/97/21. MacDonald, R. and Marsh, W. (1996), "On casselian PPP, coinegraion and exchange rae forecasing". Review of Economics and Saisics. MacDonald, R. and Taylor, M. (1994), "The moneary model of he exchange rae: long-run relaionships, shor-run dynamics, and how o bea a random walk". Journal of Inernaional Money and Finance, vol 13, 276-90. MacDonald, R. and Taylor, M. (1993), "The moneary approach o he exchange rae: Raional expecaions, long-run equilibrium and forecasing". IMF Saff Papers, vol 40, 89-107. Meese, R. and Rogoff, K. (1988), "Was i real? The exchange rae-ineres differenial relaion over he modern floaing-rae period". The Journal of Finance, vol XLIII, no4. Orr, A., Sco, A., and Whie, B. (1998), "The exchange rae and inflaion argeing", Reserve Bank of New Zealand Bullein, Vol 61, No 3. Rogoff, K. (1996), "The purchasing power pariy puzzle", Journal of Economic Lieraure, vol XXXIV, 647-668. Sephens, D. and Choy, W.K. (2001), "Wha is a Kiwi worh? Valuing he New Zealand dollar according o purchasing power pariy and uncovered ineres pariy". Paper presened a he New Zealand Economis's Associaion annual conference. 2001. 13
Wespac $ Insiuional Bank Explaining he NZ-Ausralian exchange rae, April 2002 Daa Appendix Variable Descripion Source Nominal NZD/AUD exchange rae Average for quarer Bloomberg NZ 10-year bond yield Average for quarer Daasream: NZI61 AU 10-year bond yield Average for quarer Daasream: AUI61 NZ producer prices Daasream: NZPP.F AU producer prices Daasream: AUOCPPMFF NZ curren accoun % of GDP Daasream: NZCURBALB AU curren accoun % of GDP Daasream: AUOCCBALB NZ house prices Quoable Value NZ AU house prices Ausralian Bureau of Saisics NZ produciviy GDP per person employed Saisics NZ AU produciviy GDP per person employed Ausralian Bureau of Saisics NZ commodiy prices 3 monh average ANZ AU commodiy prices 3 monh average Reserve Bank of Ausralia Technical Appendix 13 Our exchange rae model can be wrien as: z = A 1 z -1 + + A k z -k + µ + ε where z is a (n 1) vecor of he real exchange rae and he oher variables of he model. The vecor µ conains consans and ε is a vecor of whie noise disurbances wih a zero mean. On he basis of convenional lag lengh ess we include 4 lags in his equaion. Tha is, k=4. This equaion can be rewrien in vecor error correcion (VECM) form as: z = Γ 1 z -1 + + Γ k -1 z -k+1 + Πz -1 + µ + ε where he Γ's are (n n) marices of shor-run coefficiens and Π conains he long-run coefficiens. symbolises 'change in'. The rank of Π deermines he number of coinegraing vecors in he sysem. If Π is of reduced rank hen here exis marices α and β such ha Π = αβ' The columns of β are he linearly independen coinegraing vecors and he α marix is he adjusmen marix and indicaes he speed wih which he sysem responds o las period's deviaions from he coinegraing vecor. To es for coinegraion we use he race es saisic developed by Johansen. This is a es ha here are a mos r independen coinegraing vecors. The es saisic is calculaed as: N λ race = T i= r+ 1 ln(1-λˆ i ) 13 In his appendix we provide a very brief explanaion of he modelling echnique we use. For a more comprehensive discussion Enders (1995) is a very readable reference. 14
Wespac $ Insiuional Bank Explaining he NZ-Ausralian exchange rae, April 2002 where is he esimaed eigenvalues of he Π marix and T in he number of observaions. The criical values for λˆi his saisic are calculaed by simulaion sudies and can be found in Johansen and Juselius (1990). The resuls of his es for coinegraion in he variables of our model are given in able A1. As discussed in he ex, hese resuls sugges srong evidence of coinegraion. To calculae he IRF's we use a Choleski facorisaion on he residuals of he reduced form model. We assume he following causal ordering for his facorisaion: e, (i nz -i au ), (dd nz -dd au ), (cp nz - cs au ), (p nz -p au ). Table A1: resuls of coinegraion ess Null Hypohesis Alernaive Hypohesis λ race values r=0 r>0 135.99*** r 1 r>1 88.42 r 2 r>2 47.79 r 3 r>3 26.15 r 4 r>4 8.70 r 5 r>5 1.57 *** Significan a he 1% level The ime-series properies of he daa are an imporan consideraion. In paricular, weaher hey conain a uni roo has a bearing on he mos appropriae modelling sraegy. The resuls of uni roo ess on he variables in our model are repored in able A2. The numonics are he same as in he main ex. The resuls indicae ha all he variables have a uni roo. The resuls for he relaive curren accoun and ineres raes are boarder line. However, he modelling echnique ha we use is able o cope wih a mix of I(1) and I(0) variables so borderline non-saionariy is no a problem. Table A1: ADF uni roo ess: levels daa 3 2 1 Φ3 Φ1 Conclusion e -2.3-2.4-1.1 3.0 I(1), zero drif (p nz -p au ) -2.7-1.5-0.7 3.8 1.4 I(1), zero drif (i nz -i au ) -2.0-1.8 2.4 1.7 I(1), zero drif (cp nz -cp au ) -3.2-1.5-0.8 5.6 1.2 I(1), zero drif (dd nz -dd au ) -2.3-2.1-0.8 2.8 2.5 I(1), zero drif (ps nz - ps au ) -2.2-2.8-1.8 2.5 2.5 I(1), zero drif **5% sig, *10% sig. 15