Wage-Price Dynamics and Deflation in Hong Kong

Transcription

1 HEI Working Paper No: 06/2004 Wage-Price Dynamics and Deflaion in Hong Kong Hans Genberg Graduae Insiue of Inernaional Sudies Lauren L. Pauwels Graduae Insiue of Inernaional Sudies Absrac This paper provides empirical evidence on he dynamics of prices and wages in Hong Kong. The resuls imply ha he deflaion in Hong Kong since 997 can be undersood using a convenional macroeconomic framework wherein foreign influences consiue he basic underlying shocks, and adjusmen processes in domesic wages and prices deermine he deails of he ransmission mechanism. Our resuls indicae ha he decline in local nominal prices owes much o declining prices of impored inermediae goods. The negaive oupu gap and he increase in unemploymen experienced during he deflaion period also have heir origin in foreign shocks, bu he domesic wage adjusmen process consiues an imporan conribuing facor. The Auhors. All righs reserved. No par of his paper may be reproduced wihou he permission of he auhors.

2 Wage-Price Dynamics and Deflaion in Hong Kong Hans Genberg* Lauren Pauwels* This Version: Augus 2004 Absrac This paper provides empirical evidence on he dynamics of prices and wages in Hong Kong. The resuls imply ha he deflaion in Hong Kong since 997 can be undersood using a convenional macroeconomic framework wherein foreign influences consiue he basic underlying shocks, and adjusmen processes in domesic wages and prices deermine he deails of he ransmission mechanism. Our resuls indicae ha he decline in local nominal prices owes much o declining prices of impored inermediae goods. The negaive oupu gap and he increase in unemploymen experienced during he deflaion period also have heir origin in foreign shocks, bu he domesic wage adjusmen process consiues an imporan conribuing facor. JEL classificaion: E3, F4, N5, C5 Paper prepared for he CEPR/HIEBS/HKIMR Conference on Deflaion, Pegs and Capial Flows, /2 December 2003, Hong Kong. Commens from our Guy Meredih, our discussan, oher paricipans a he conference as well as an anonymous referee are graefully acknowledged. The paper was prepared in par wihin he NCCR Finrisk projec Macro Risk, Sysemic Risk and Inernaional Finance. The NCCR is a research program suppored by he Swiss Naional Science Foundaion. *Graduae Insiue of Inernaional Sudies, 32 rue de Lausanne, 2 Geneva 2, Swizerland. genberg@hei.unige.ch and pauwels0@hei.unige.ch

3 I. Inroducion. During he six years ending in 2003 he consumer price index in Hong Kong fell by 5 percen. During he same period he unemploymen rae increased o 8.7% in he hird quarer of 2003 from 2.% in he hird quarer of 997. By any definiion his mus be wha we mean by deflaion. In his paper we aemp o explain he main feaures of he deflaion process by esimaing and inerpreing a se of equaion describing he inerrelaionships beween domesic inflaion, wage changes, he oupu gap, unemploymen, and heir muual dependence on foreign economic condiions. Our purpose is o see wheher he deflaion can be explained by a sandard macroeconomic approach involving foreign and domesic shocks ogeher wih a convenional propagaion mechanism. We conclude ha he deflaion process in Hong Kong can indeed be undersood in hese erms where he main sources of shocks are of foreign origin, and where he propagaion mechanism involves somewha sluggish wage adjusmens. We also find evidence suggesive of changes in he propagaion mechanism in he pos-997 period. The remainder of he paper is organized as follows. In he nex secion we presen some sylized facs abou he recen deflaionary period, including he evoluion of domesic and foreign price and business cycle indicaors. These sylized facs sugges a prima-facie case for a dominan role of exernal shocks in he deflaion process. This hypohesis is esed in subsequen secions. In secion III we use Granger-causaliy ess and a semi-srucural vecor auoregression model o documen he imporance of exernal price and oupu shocks for he evoluion of prices, wages and unemploymen in Hong Kong. We also find, however, ha domesic facors are imporan in he shor o medium run, especially for wages and unemploymen. Secion IV provides a srucural inerpreaion of hese findings. A brief concluding secion ends he paper. II. The Naure of he Curren Deflaion in Hong Kong. The curren deflaion began in 997/98. Measured by he implici GDP deflaor he price level peaked in he 4 h quarer of 997, and he consumer price index sared falling hree quarers laer. From heir respecive peaks he indices have fallen by 23% and 7% respecively (Figure II.). Oher imporan indicaors of nominal prices and coss have evolved in a similar manner. For example, he propery price index fell from a value of 04 a is peak in he 3 rd quarer of 997 o 35.5 in 2003:2. While nominal wages have no fallen during he price deflaion, heir growh rae has declined from an annual rae of abou 7% during he five years before he sar of he deflaion o essenially zero since (Figure II.2). Prices of goods impored ino Hong Kong and a weighed average of consumer prices in Hong Kong s main rading parners have also followed generally downward From 992:4 unil 997:4 nominal wages rose by 36.9%, and from 997:4 o 2003: he growh rae was -.%. 2

4 pahs in recen years, bu he sar of heir decline happened earlier. The uni value of impors peaked wo years before he domesic GDP deflaor, and has fallen by some 9% since hen. The weighed average of rading parners CPIs (expressed in erms of HKD) sopped rising in he middle of 995, fell close o 0% during he following hree years before recovering and again reaching he 995 level in he middle of (Figure II.3). These developmens should be inerpreed agains he background of he currency-board arrangemen in place in Hong Kong which implies a virual peg of he HKD o he USD a he rae of 7.8 HKD/USD. 2 Togeher wih he highly open naure of he economy, he fixed exchange rae implies ha exernal evens should play an imporan role for domesic price movemens. Indeed one of he hypoheses we will invesigae is he exen o which he curren price deflaion is compleely exernally deermined. An alernaive possibiliy is ha exernal developmens consiue he principal fundamenal shocks, bu ha domesic facors have an imporan role in he propagaion mechanism of hese exernal shocks. We will also invesigae a hird hypohesis which has frequenly been ariculaed namely ha he propery price decline, he sar of which precedes he decline in he general price level, is an imporan independen cause of he deflaion. Deflaion would be relaively innocuous if i simply mean declining nominal variables and if employmen and oupu were unaffeced. The problem of course is ha such a nea dichoomy may no be presen. In Hong Kong, he evoluion of indicaors of real economic aciviy suggess ha here migh be some reason for concern. The unemploymen rae has been rising a a seemingly acceleraing rae from is rough in 990 of abou % of he labor force o around 3% in 996 o close o 9% in mid (Figure II.4) A he same ime oupu growh has been slowing seadily from he 5-6% range in he lae 980s o he 2-3% range in he pas five years. Cyclical indicaors show sharp declines in economic aciviy in he immediae afermah of he Souh-Eas Asian Crisis and again in The exen o which he price deflaion in he pas six years has conribued o he slowing down of real economic aciviy will be invesigae laer in he paper. Here we simply noe ha a leas he cyclical swings in he Hong Kong economy follow quie closely, bu wih higher ampliude han hose in he main rading parners. (Figure II.5) 2 The value 7.8 represens a ceiling for he exchange rae and formally here is no floor. However, he flucuaions have been minimal from he poin of view of pass-hrough of foreign price changes. The lowes value of he Hong Kong dollar during he pas 0 years is less han one percenage poin from he maximum. 3

5 Figure II.: Price leveldevelopm ens 05 CPI(seas. A dj.) GDP deflaor (S eas. A dj.) '93 '933 '94 '943 '95 '953 '96 '963 '97 '973 '98 '983 '99 '993 '00 '003 '0 '03 '02 '023 '03 '033 0 Figure II.2: Nom inal w age and price developm ens G D P deflaor Nom inalw ages P ropery prices '93 '933 '94 '943 '95 '953 '96 '963 '97 '973 '98 '983 '99 '993 '00 '003 '0 '03 '02 '023 '03 '033 F ig u re II.3 : E x e rn a l p ric e d e v e lo p m e n s 0 W orld CPI(m easured in HKD) Uni va lue ofim p o rs '93 '933 '94 '943 '95 '953 '96 '963 '97 '973 '98 '983 '99 '993 '00 '003 '0 '03 '02 '023 '03 '033 4

6 Figure II.4: Unem ploym en ra e '84 '85 '86 '87 '88 '89 '90 '9 '92 '93 '94 '95 '96 '97 '98 '99 '00 '0 '02 '03 Figure II.5: O upu gaps (obained using HP file rs ) O upu gap H ong K ong O upu gap rading parners '84 '85 '86 '87 '88 '89 '90 '9 '92 '93 '94 '95 '96 '97 '98 '99 '00 '0 '02 '03 III. Inflaion dynamics: causaliy ess and evidence from semi-srucural models. In his secion we show ha a descripion of inflaion dynamics in Hong Kong requires firs and foremos careful aenion o inernaional influences in he form of world price, ineres rae and oupu movemens, reflecing exernal cos and demand facors. Bu our evidence also indicaes ha a full undersanding of he shor- o mediumerm dynamics of domesic prices requires aking accoun of he ineracion beween prices, wages and economic aciviy in he Hong Kong economy. Finally we argue ha propery price developmens do no appear o have had an imporan independen role in he inflaion process, bu have insead reaced endogenously o he exernal influences ha also drive inflaion more generally. 5

7 III. Granger-causaliy ess. Genberg and Pauwels (2003) [hereafer GP] presen exensive evidence from Granger-causaliy ess and bivariae vecor auoregressions (VAR) beween domesic prices (boh he CPI and he GDP deflaor), domesic nominal wages, propery prices and several measures of inernaional prices (CPI in he Unied Saes, a weighed average of CPIs in Hong Kong s main rading parners, and impor prices). For our purposes here, he resuls in ha paper can be summarized in hree poins. Firs, even hough he Hong Kong dollar is fixed in relaion o he US dollar, price developmens in he Unied Saes do no have a paricularly srong influence on Hong Kong prices and wages. Second, a weighed average of consumer prices in HK s main rading parners and a measure of HK impor prices do have srong influences on boh he local consumer price index and he GDP deflaor. Impor prices influence local wages and propery prices as well. Third, here is a muual inerdependence beween HK wages and he GDP deflaor. The Granger-causaliy ess in GP (2003a) were based mainly on bivariae regressions. This opens up he possibiliy ha he relaionships found are he consequence of common effecs of hird variables. To invesigae his possibiliy we conduc block exogeneiy es in a VAR sysem ha includes four domesic variables (he GDP deflaor, nominal wage raes, he nominal propery prices, and he unemploymen rae) and hree foreign variables (he US hree-monh T-Bill rae, uni value of Hong Kong impors, and a measure of he oupu gap in Hong Kong s main rading parners). We esimae he VAR sysem boh in levels and in quarerly raes of change. 3 The resuls are presened in Table III.. For each of he four domesic variables we conduced ess for he exclusion of he lags of he exogenous variables. Only hose variables whose lags are significan a he 0% level or lower are shown in he able. The resuls confirm and exend hose found in GP (2003a). Firs, foreign facors have significan effecs on each of he four domesic variables included in he sysem. Lagged impor prices have significan effecs on boh he GDP deflaor and nominal wages. They also influence he unemploymen rae in he level specificaion. The US ineres rae has an impac on he GDP deflaor in he level specificaion and on propery price inflaion in he growh rae specificaion, and he exernal oupu gap influences he domesic unemploymen rae in he sysem esimaed in growh raes. Second, he ineracion beween domesic variables remains afer we conrol for he effecs of foreign variables. The exac naure of he ineracion depends on wheher we consider he level or growh rae specificaions, bu some paerns can be discerned; 4 (i) he GDP deflaor influences and is influenced by nominal wages, (ii) he unemploymen rae has an impac on wages, and (iii) wages influence he unemploymen rae. 3 In each case he unemploymen rae, he T-Bill rae, and he foreign oupu gap are measured in percenages. 4 The significance level of he ineracion erms beween he domesic variables is also affeced by he lag lengh in he VARs. The imporance of foreign variables, however, is robus. 6

8 Table III.: Granger-causaliy ess Log-levels GDP deflaor p-values of a es for he exclusion of 2 lags of Nominal wages 0.09 Impor prices 0.00 US 3-M T-Bill rae 0.06 Nominal wages Unemploymen rae Impor prices Nominal propery prices GDP deflaor 0.03 Unemploymen rae Impor prices 0.0 Quarerly raes of change 2 p-values of a es for he exclusion of 4 lags of GDP deflaor Impor prices.0 Nominal wages Nominal propery prices Unemploymen rae GDP deflaor Unemploymen rae Impor prices GDP deflaor Nominal wages US 3-M T-Bill rae Nominal wages World oupu gap The unemploymen rae is measured in percenage poins. As exogenous variables, he esimaed equaions include US 3-monh Treasury Bill rae, he log of impor prices, and a measure of he world oupu gap. All equaions were esimaed wih wo lags of he independen variables. 2 The unemploymen rae is measured in percenage poins. As exogenous variables, he esimaed equaions include he US 3-monh Treasury Bill rae, he quarerly rae of change of impor prices, and a measure of he world oupu gap. All equaions were esimaed wih four lags of he independen variables. The imporance of exernal facors is no surprising in view of he size and openness of he Hong Kong economy, he currency board sysem, and he absence of conrols on capial movemens. A rigidly fixed exchange rae and free movemens of capial implies ha credi condiions in Hong Kong will be deermined predominanly in he inernaional marke as capured here by he US Treasury Bill rae. The inernaionally deermined credi condiions will in urn influence macroeconomic condiions hrough heir effecs iniially on asse prices (e.g. propery prices), consumpion and invesmen and ulimaely on he general level of aggregae demand, oupu, and he price level. Furhermore, given he fixed exchange rae, exernal price developmens will have an 7

9 impac boh on domesic consumer goods (and herefore on he CPI) and on impored inpus. As we will documen more in deail below, he variaions in he cos of impored inpus appear o be an imporan source of inflaionary, and deflaionary, pressures in Hong Kong. Finally, exernal business cycle condiions as measured by he aggregae oupu gap in he main rading parners will be imporan for aggregae demand for Hong Kong oupu. The exreme openness of he economy means ha variaions in expor demand is very imporan for he local economy. In addiion i is likely ha domesic economic agens will form expecaions abou fuure incomes on he basis of exernal developmens, and his in urn will influence domesic demand. The dependence of he domesic economy on exernal facors does no mean ha local developmens are unimporan, bu i suggess ha some of hese developmens may be deermined parly by exernal evens. For example, Peng, Cheung and Leung (200) have shown ha propery price developmens can have significan effecs on domesic aggregae demand and hence on inflaion hrough wealh and balance shee effecs. 5 This is no inconsisen wih our resuls which sugges ha propery price developmens do no Granger-cause inflaion or oupu in a mulivariae conex, because we would argue ha propery prices hemselves are in par endogenous. In he nex secion we propose a mehodology for measuring he relaive imporance of foreign and domesic facors for domesic macroeconomic flucuaions in general and for he curren deflaion in paricular. III.2 Evidence from a semi-srucural VAR. 6 To esimae he relaive imporance of domesic and foreign shocks in he Hong Kong economy we use he framework developed in Genberg, Salemi, and Swoboda (987) and applied o Hong Kong in Genberg (2003). We updae he esimaions carried ou here using daa unil he firs quarer of 2003, and we also exend he analysis by paying greaer aenion o he explanaion of he pos 997 period. The esimaion framework is a block-riangular VAR as in equaion () y x D( L) = 0 D D 2 22 ( L) y ( L) x + ε η () The vecors y and x conain respecively he domesic and foreign variables of ineres. The riangular form of he marix of coefficiens on he lagged variables embodies he small open economy hypohesis according o which domesic variables have no influence on foreign variables. The sysem () can be hough of as a reduced form of a general dynamic macroeconomic model of he home economy in which he 5 We are graeful o a referee for drawing our aenion o his sudy and ha of he IMF menioned in foonoe 8 below. 6 This secion draws heavily on Genberg (2003). 8

10 small economy assumpion is mainained. The reduced form errors ε and η will be funcions of he srucural errors, bu he small economy hypohesis ensures ha η will no conain any errors from he domesic srucural equaions. The block-riangular srucure herefore allows us a semi-srucural inerpreaion of (), namely ha η only conains foreign shocks. This will allow us o consruc a measure of he relaive imporance of foreign versus domesic shocks in he economy. The moving average represenaion of y and x is given in (2) y x D = I ( L) L 0 D D 2 22 ( L) L ( L) L ε η (2) From his we can wrie he forecas error a horizon h of each elemen of y as a linear combinaion of he VAR errors o he equaion for he domesic and foreign variables y D h F h i, + h E yi, + h ) = ξd, l ε d, + l + ψ f, l d = l = f = l = ( η f, + l (3) The variance of he forecas error is V D h F h D F h [ yi, + h E yi, + h )] = d, l σ ε + ψ f l + d, σ η f ( ξ d, lψ f, lσ ε d, η f d = l = f = l = d = f = l = ξ (4) In view of he small-economy assumpion ha says ha domesic shocks do no influence foreign variables, i mus be ha case ha he conemporaneous correlaion beween he VAR errors ε and η mus be due o he conemporaneous effec of x on y and no vice versa. Hence we can decompose he variance of he forecas errors ino a domesic par which is he firs erm of he righ hand side of (4) and a foreign par corresponding o he second and hird erms. A convenien way o achieve an equivalen decomposiion is o esimae he variance-covariance marix of he VAR errors and render i diagonal by a Choleski decomposiion in which he foreign variables come firs in he ordering. This effecively ensures ha he conemporaneous correlaion beween ε and η is due o he conemporaneous effec of x on y. Then we ge [ yi, + h E ( yi, + h )] = Di, h + Fi h V, where D h 2 2 D i, h = ξ d, l σ ε d d = l = F F h D F h 2 2 i, h = ψ f, lσ η + ξ f d, lψ f, lσ ε d, η f f = l = d = f = l = D i,h (F i,h ) sands for he domesic (foreign) conribuion o he forecas error variance of y i a horizon h. The relaive conribuion of foreign variables o he forecas error of each domesic variable is herefore simply 9

11 REL F i, h i, h = (5) D i, h F + F i, h To compue his we firs esimae he VAR as shown in (), i.e. imposing he block-riangular srucure beween domesic and foreign variables. We hen orhogonalize he errors using he Choleski facorizaion of he variance-covariance marix. Finally we calculae he forecas error variance decomposiion and compue REL F for each of he domesic variables. We use he same variables as in he Granger-causaliy ess, namely he GDP deflaor, nominal wage raes, nominal propery prices, and he unemploymen rae as domesic variables and he US 3-monh Treasury Bill rae, uni value of Hong Kong impors, and a measure of he oupu gap in Hong Kong s main rading parners as he foreign variables. 7 The esimaes of he relaive imporance of foreign shocks in he Hong Kong economy given in Table III.2 are consisen wih hose presened in Genberg (2003). Forecas errors of he GDP deflaor are predominanly due o foreign facors a horizons longer han 4 quarers. The same can be said for propery prices. Nominal wages are influenced by a larger proporion of domesic shocks a shor- o medium horizons, and he measure of he relaive imporance of foreign shocks only reaches 50% a he 3-year horizon. The unemploymen rae lies in beween he wo price indices and he wage rae in erms of is dependence on foreign variables. A he horizon of one year wo-hirds of he forecas error variance is accouned for by domesic shocks bu his proporion falls o less ha one hird a he 3 year horizon. 7 All variables excep he unemploymen rae, he US ineres rae, and he world oupu gap were expressed in log-levels. Five lags of each variable were included in each equaion. Genberg (2003) conains robusness checks of he main resuls using differen combinaions of domesic and foreign variables, and differen lag lenghs in he VAR. 0

12 Table III.2. Conribuion of foreign variables o he forecas error variance of domesic variables. (Percen) Horizon P GDP W U P PROP Sample period: 984:-2003:. The domesic variables are he log of he GDP deflaor (P GDP ), he log of nominal wages (W), he unemploymen rae in percen (U), and he log of he propery price index (P PROP ). The foreign variables are he US hree-monh Treasury bill rae, he log of he uni value of impors of Hong Kong, and a measure of he foreign oupu gap. Wha do he esimaes imply for he sources of he deflaion since 997? One way o answer his quesion is o make use of he fac ha in VARs i is possible o decompose he in-sample acual value of each variable ino one par ha is forecas on he basis he esimaed dynamics of he sysem and anoher par ha depends on shocks ha have occurred during a paricular ime period. To be specific, equaion (3) above can be rewrien as y = i, yi, h E h yi, yi, h ) D h d = l = d, h+ l F ( + ξ ε + ψ η (6) d, l h f = l = f, l f, h+ l The inerpreaion of his relaionship is as follows. The acual change in he variable y i from ime -h o ime can be decomposed ino (i) a forecas or expeced change (as of period -h) which is he firs erm on he righ hand side, (ii) he consequence of shocks o domesic variables in he ime inerval -h o (he second erm), and (iii) he consequence of shocks o foreign variables in he same ime inerval (he hird erm). The forecas, E -h (y i, - y i,-h ), represens he delayed effecs implici in he dynamics of he enire VAR sysem. In his secion his decomposiion is used o illusrae he relaive imporance of predicable facors, foreign shocks, and domesic shocks for he evoluion of oupu growh and inflaion during he deflaion since 997. We use he esimaed values of he domesic and foreign shocks o decompose he pah for hese wo variables from he las quarer of 996 (-h in he equaion) unil he end of he sample. In each case he pahs of wo counerfacual variables are calculaed and presened in chars ogeher wih he acual values: - No Shocks which is simply y ) E h ( i,

13 - Foreign shocks only which is defined by E h ( yi, ) + ψ f, l η F h f = l = f, h+ l We sar describing he resuls for he GDP deflaor and he nominal wage index ogeher wih he corresponding inflaion raes. Figures III. and III.2 show ha boh he price level and he wage level in 2003 are around he level hey would have been if only exernal shocks had influenced he Hong Kong economy since 997. This is in line wih he view ha over long enough horizons, developmens in he economy are dominaed by shocks originaing abroad. This long-erm view hides he fac ha domesic deflaionary shocks have been relaively imporan in he inervening period. From is peak in he fourh quarer of 997, he GDP deflaor fell during he subsequen hree years by some 4.7%. Had here been no shocks he model would have prediced a decline by.9%, and aking accoun he exernal shocks during , he model can explain a decline of 9.4%. In oher words, domesic facors accouned for slighly over five percenage poins of he oal decline. 8 For wages a similar paern is visible. Wih only foreign shocks, he model would have prediced an increase of wages of 5.4% whereas he acual increase was only.3%. Domesic shocks hus exered downward pressures on wages of 4.% over his period. During he period he paerns are reversed, which is bes seen in Figures III.a and III.2a ha conain he year-on-year inflaion raes calculaed from he daa underlying Figures III. and III.2. The growh raes of boh he price and wage series are now higher han prediced by he effecs of foreign shocks alone. In oher words, domesic facors were exering inflaionary pressures during his period. Turning o he behavior of he unemploymen rae (Figure III.3) we also noe a difference beween he period from he beginning of 998 o he end of 2000 and he period hereafer. In he firs period, domesic shocks keep he unemploymen rae higher han hey would have been in heir absence, whereas in he second period he reverse is rue. Noe, however, ha foreign facors are considerably more imporan han he domesic ones. For example from he firs quarer of 997 unil he fourh quarer of 999, he acual unemploymen rae increased from 2.3% o 6.3%, of which only one percenage poins can be explained by domesic facors. Similarly, he cumulaive effecs of ineria and foreign shocks would have prediced an unemploymen rae of jus over 8% in he beginning of 2003 when he acual rae was 7.5%. Domesic expansionary facors have had an effec of abou half a percenage poin. 8 I is possible ha price developmens in he propery secor are responsible for his as suggesed by Peng, Cheung and Leung (200) and IMF (2002). 2

14 Figure III.: GDP deflaor (percen deviaion fro m 997:)) 5 A cual N o shocks W ih foreign shocks :0 997:02 997:03 997:04 998:0 998:02 998:03 998:04 999:0 999:02 999:03 999: :0 2000: : :04 200:0 200:02 200:03 200: :0 2002: : : :0 Figure III.2: Nom inalwages (percen deviaion fro m 997:) 4 Acual N o shocks W ih foreign shocks :0 997:02 997:03 997:04 998:0 998:02 998:03 998:04 999:0 999:02 999:03 999: :0 2000: : :04 200:0 200:02 200:03 200: :0 2002: : : :0 Figure III.a: Inflaion (y -o -y % changes in GDP deflaor) Acual N o shocks W ih foreign shocks :0 998:02 998:03 998:04 999:0 999:02 999:03 999: :0 2000: : :04 200:0 200:02 200:03 200: :0 2002: : : :0 3

15 Figure III.2a: W age inflaion (y -o -y % changes in nom inal wages) Acual N o shocks W ih foreign shocks :0 998:02 998:03 998:04 999:0 999:02 999:03 999: :0 2000: : :04 200:0 200:02 200:03 200: :0 2002: : : :0 Figure III.3: Unem ploym en ra e (p e rc e n ) 9 Acual N o shocks W ih foreign shocks :0 997:02 997:03 997:04 998:0 998:02 998:03 998:04 999:0 999:02 999:03 999: :0 2000: : :04 200:0 200:02 200:03 200: :0 2002: : : :0 Taken ogeher he behavior of prices, wages, and unemploymen since 997 sugges an inernaional environmen which was decidedly deflaionary from 998 unil he second half of 999 and again from 200 unil he second half of Domesic facors conribued o he deflaion from 998 unil he end of 2000 afer which hey have been somewha inflaionary. I would be emping o use he esimaed VAR o separae he effecs of he domesic shocks ino heir individual componens. Unforunaely, in he absence of some idenifying resricions ha would allow us recuperae he srucural shocks from he reduced form VAR disurbances (he ε in equaion ()), we are no able o do so. Insead in he nex secion we shall search for srucural explanaions of he resuls obained here 4

16 by esimaing equaions explaining he wage and price inflaion as well as he unemploymen rae. IV. Srucural inerpreaions of he wage-price-unemploymen nexus. IV. The inflaion process. Much of he recen lieraure on inflaion dynamics has focused on heoreical developmens and empirical applicaions of he New Keynesian Phillips curve (NKPC). Influenial conribuors o his lieraure have been Galí and Gerler (999) [henceforh GG] who esimaed heir New Hybrid Keynesian Phillips curve for he USA, and Galí, Gerler and López-Salido (200) [henceforh GGLS] who applied he framework o he Euro area. The heoreical model is based on he Calvo (983) sicky price and Taylor (980) sicky wage framework and implies ha firms se prices as mark-ups over he marginal cos of producion. The empirical applicaions ypically focus on he behavior of wages and labor produciviy as componens of marginal cos. 9 The main findings of GG (999) and GGLS (200) can be summarized in four poins: () firms are found o be forward-looking, (2) he degree of price sickiness is subsanial, on average abou 5 quarers in he Euro area and 4 quarers in he US, 0 (3) he backward-looking behaviour is saisically bu no quaniaively imporan, (4) he New Keynesian Phillips curve provides a good descripion of inflaion dynamics The model used in GG (999) and GGLS (200), however, only feaures labor as he variable inpu in he cos funcion. Genberg and Pauwels (2003a, 2004) apply he NKPC paradigm o Hong Kong bu argue ha impored inermediae goods are an imporan facor of producion in he conex of open economies. Firms ha se prices in relaion o marginal cos herefore ake ino accoun no only labor cos bu also he cos of impored inermediae inpus. Our open economy model ness Galí and Gerler s (999) hybrid model as a special case when impored inpus are negligible. In his secion we summarize he basic srucure of our open economy version of he NKPC and repor resuls we have obained when we confron i wih dae from Hong Kong. The New Hybrid Keynesian Phillips curve The new hybrid Keynesian Phillips curve is based on Calvo (983) sicky price model emphasising forward-looking behaviour of firms. Each period, a fracion -θ of firms in he economy adjus heir prices, while he remainder are consrained o keeping hem fixed. The parameer θ is hus a measure of he degree of price sickiness. Since 9 The empirical applicaion of he GG (999) approach is no wihou conroversy. Boh he use of uni labour cos and raional expecaions in heir model have been criicised by Robers (200), Rudd and Whelan (200) and Lindé (200), who's resuls do no concur wih hose of Galí and Gerler (999). Galí, Gerler and López-Salido (2003) formulae a response showing ha heir model is robus o he criicism. Moreover, economeric problems associaed wih insrumenal variable esimaion end o affec he resuls obained for he NKPC (see Sock, Wrigh and Yogo (2002) for a weak insrumens and idenificaion survey). 0 I should be noed ha he degree of price sickiness is no precisely esimaed in hese sudies. For example, GGLS presen esimaes from 2 quarers o 7 quarers for he Euro Area. 5

17 firms are assumed idenical, he proporion of firms adjusing a ime, will choose he * same opimal price. The aggregae price level herefore follows p p ( ) = θp + θ p Gali and Gerler (999) assume ha a fracion ( κ) of he firms se prices opimally in a forward-looking fashion and while a fracion κ se prices using an average of opimally adjused prices from he previous period plus a correcion for pas inflaion (π ). Hence where p p * f f b ( ) p κp = κ + = µ + p π b = * + p k= 0 i n ( βθ ) ( βθ ) E [ mc ] + k f b where p is he price se by he forward looking agen and p is ha se by heir backward-looking counerpars. The implied reduced form of he hybrid NKPC can be wrien as h f b = δ mc + ω E π } + + ω π π { + ε (7) where β is he subjecive discoun facor and ^ mc he deviaion of he log of real marginal cos from is seady-sae value. The relaionship beween he srucural parameers and he reduced from coefficiens is defined by δ ω ω f b and ( κ)( θ)( βθ) βθψ κψ ψ θ + κ [ θ( β) ] ψ As done in he lieraure we assume ha β = implying haω f + ω b =. Micro-Foundaions The marginal cos measure presened in GG (999) is derived from basic principles of microeconomics. Generally, he cos minimizing problem for n inpus can n be wrien as: Min C = w X, wih as he i h inpu price valued a ime and X i= i, i, i, w i, is he i h inpu a ime, subjec o he producion funcion Y = f ( X ; α ). We assume a i, i X i, 6

18 i Cobb-Douglas producion funcion so ha Y = A X α i, wih αi =. The firs order condiions yield he following shadow price in real erms: n i= n i= λ real w X = i, i,, α PY i i where P is he oupu price. Afer aking he naural logarihm and deviaions from seady sae values, we can define a muli-inpu-marginal-cos gap measure by: so, ^ n i i i= * (, ) i, mimc = ξ s s ^ mimc n = ξ ( sˆ, ) i i i= wi, X where si, = ln PY share and ξ =. n i= i i, and where s i, is he seady-sae value of he ih inpu income Open-Economy Marginal Coss Galí and Gerler s model has capured inflaion dynamics in he USA and EU fairly well using solely he wage rae ha in urn influences marginal cos. This may no be he case in he conex of open economies, where i is believed ha exernal inflaion may have a more direc influence on domesic inflaion. Genberg and Pauwels (2003b) argue ha price seing in a highly open economy is likely o be influenced by foreign prices hrough inermediae inpus, as formulaed in Galí and López-Salido (2000) for heir analysis of Spanish inflaion. Gagnon and Khan (200) have aemped o modify he marginal cos measure hrough he use of differen ypes of producion funcions. Open economy consideraions also have been inroduced in srucural models of price seing by Galí and Monacelli (2000) and Balakrishnan and López-Salido (2002). In he wo inpus case using labour (L ) and inermediae impors (M ), he specificaion yields he following firs order condiion: λ = w L αy = P im M ( α ) Y As before we can define he following weighed average cos measure by: ^ L omc = ξsˆ + ( ξ) s ˆim 7

19 ^ where omc is he open economy measure of he deviaion of marginal cos from is seady sae value. The parameerξ is beween 0 and and needs o be deermined. When ξ is equal o, he model collapses o he specificaion by GG (999) feauring only labors coss. The final expression for he rae of inflaion is: h f b = δ omc + ω E { π } + + ω π π + ε (8) wih he relaionship beween he coefficiens and he srucural parameers as defined above following equaion (7). Applicaion o Hong Kong We esimaed he model for a sample exending from he firs quarer 984 unil he fourh quarer of 2002, as well as for a sub-sample spanning 984 Q 997 Q2 o check wheher excluding he deflaionary period of he sample has an impac on he esimaes. Inflaion ( π ) was measured as he log difference of he GDP deflaor, and he marginal cos variable was consruced using a seasonally adjused nominal wage index and a seasonally adjused quarerly uni value index of impors. To deal wih he { } ^ endogeneiy of E π + and omc, we used lagged values of he inflaion rae, curren and lagged values of CPI inflaion in Hong Kong s main rading parners, and curren lagged values of impor price inflaion as insrumens. A deailed discussion of he esimaion sraegy can be found in GP (2004). Here we simply highligh he main conclusions. Table IV. shows esimaed reduced form coefficiens as well as he implied srucural parameers from specificaions which yielded he mos significan coefficien (δ h ) on he marginal cos measure, as judged by he p-value. For he full sample, he esimaed weigh (ξ) on labour cos is ypically less han one half. In he preferred specificaion shown in he able, only values beween 0 and 0.3 give posiive and saisically significan values of δ h. This implies ha when firms adjus prices as mark-up over marginal cos, hey place more weigh on impored inpu coss han on wage coss. In Hong Kong he evidence clearly indicaes ha open economy consideraions mus be aken ino accoun in modelling inflaion along he line of he NKPC. The esimae of he parameer κ implies ha abou half of he firms se prices in a compleely forward-looking manner, while he esimae of θ suggess ha beween 46 and 49 percen of firms adjus prices each quarer. As a consequence, he mean lag of price adjusmens is esimaed o be beween 2.4 and 3.3 quarers. This is somewha 8

20 shorer han he esimaes found for he USA and he Euro Area indicaing ha price flexibiliy in Hong Kong is relaively high. Turning o he resuls for he shorer sample which excludes he deflaion since mid-997, we noe hree main differences. Firs, he imporance of labor in he marginal cos measure is greaer in he earlier period. This is consisen wih an evoluion of he Hong Kong economy away from labour-inensive producion. Second, he esimae of he effec of marginal cos on curren inflaion is larger in he earlier sample. This migh be he resul of greaer pricing power of firms when producion has shifed owards more high-value-added producs, bu i does seem o be inconsisen wih he process of price convergence wih he mainland. Finally, he esimae of κ, he proporion of price seers which are backward-looking, is smaller for he pre-crisis period. Table IV.: Esimaion of he Open Economy New Keynesian Phillips Curve. Sample 984: o 2002:4 Reduced form coefficiens ξ = 0 δ h (0.0) ω f 0.58 (0.38) Implied srucural parameers Price sickiness (in quarers) /(-θ) θ κ 0.5 ξ = 0.3 δ h (0.03) ω f 0.52 (0.4) θ κ 0.54 Noes: The insrumen se consiss of 5 lags of domesic inflaion and wo lags and he curren values of he CPI inflaion of main rading parners and impor price inflaion. Sample 984: o 997:2 Reduced form coefficiens ξ = 0.5 δ h 0.8 (0.034) ω f 0.89 (0.27) Implied srucural parameers Price sickiness (in quarers) /(-θ) θ κ 0.08 ξ = 0.55 δ h 0.02 (0.028) ω f 0.97 (0.098) θ κ 0.02 Noes: The insrumen se consiss of 5 lags of domesic inflaion and wo lags and he curren values of he CPI inflaion of main rading parners and impor price inflaion. 9

21 In spie of hese changes in he resuls, he degree of price sickiness does no appear o have changed over ime. Hong Kong has been and remains an economy wih relaively high degree of price flexibiliy. IV.2 Wage dynamics. In he model of inflaion ha we esimaed in he previous secion, he domesic wage rae is an imporan driving force, even if we have argued ha i does play second fiddle o exernal price impulses. To gain furher insighs ino he inflaion process we herefore specify and esimae an equaion describing domesic wage movemens. I is based on a framework suggesed by Blanchard and Kaz (999) in which rae of change in nominal wages is a funcion of expeced inflaion, produciviy growh, a measure of demand, and a ype of error-correcion erm according o which real wages in excess of hose warraned by produciviy leads o downward nominal wage adjusmens, i.e. w w e W = α 0 + α( p p ) + α 2 prod _ growh + α3d + α4 ln ω (9) P where w sands for he nominal wage rae, p for he price level, d for a measure of excess demand, and ω for he equilibrium real wage rae. Lower case leers refer o he naural logarihm of he corresponding capial leers. We expec α, α 2, and α 3 o be posiive and α 4 o be negaive. Daa. For he price level we used he consumer price index. In he empirical applicaion we replaced he expeced rae of price inflaion wih he curren rae and used insrumenal variables o projec his rae on lagged informaion. Labor produciviy was measured as real GDP per hour worked. The produciviy growh variable used in he regression is he quarerly growh rae in his measure smoohed using he Hodrick- Presco filer wih a smoohing parameer of 600. The excess real wage W erm, ln ω, real wage gap for shor, was obained as he residual from a P regression of he log of he real wage rae on a consan and he log of labor produciviy. Resuls. Table IV.2 conain coefficien esimaes obained when a measure of he oupu gap is used as he demand variable. 2 The firs column shows he resuls corresponding o he whole sample from 984 unil All coefficiens have he correc signs and all excep he one on he error-correcion erm are highly significan. The size of he See Table IV. for an explanaion of he esimaion echnique and he insrumens used. 2 The oupu gap was measured as he difference beween he log of real gdp and he HP filered (smoohing parameer = 600) of he same series. 20

22 coefficien on he oupu gap implies ha a gap of 5 percenage poins (he mean absolue value of he series is 2.2 percen and he maximum is 8.2 percen) will increase wage inflaion by 0.5 % in he curren quarer. The coefficien on he real wage gap implies ha if real wages are 5 percenage poins above heir equilibrium level (as measured here) he nominal wage rae will adjus downward by.47% in he curren quarer. 3 These esimaes imply a cerain sluggishness in he adjusmen of he wage rae o changes in prices and demand condiions. For example, an increase of 5 percen in he CPI would lead o an immediae increase in wages of abou 3 percen and hence a wage gap of minus wo percen. The wage gap would by iself increase wages abou 0.8 percen during he following four quarers. However, here would likely be oher facors influencing he wage process as well. As we shall see in he nex secion, he reducion in wages will have compeiiveness effecs increasing he oupu gap, which in urn will increase wages. In addiion, he inflaion model esimaed in he previous secion implies ha here will be feedback from he wage rae ono he price level. 4 Foreign variables do no ener he wage equaion direcly, bu hey do have indirec effecs. An increase in impor prices influences boh he GDP deflaor (see he previous secion) and he consumer price index (see foonoe 4). Hence hey will influence wages boh concurrenly and wih a lag as described above. Changes in exernal demand will also influence wages indirecly since, as we shall see in he nex secion, foreign oupu flucuaions have a direc influence on he domesic oupu gap. 3 The mean of he absolue values of he real wage gap is 2.6% and he maximum is 0.%. 4 Noe ha we assume ha i is he consumer price index ha is imporan for he wage process, whereas he inflaion model esimaed in he previous secion used he GDP deflaor a he dependen variable. The wo are clearly relaed. A simple OLS regression of he quarerly CPI inflaion rae on he corresponding rae of change of he GDP deflaor and he inflaion rae of impored goods gives π cpi = π gdp +.24π imp. 2

23 Table IV.2: Deerminans of wage inflaion () (2) (3) (4) (5) (6) (7) log(cpi ).6 (.06).59 (.05).52 (.04).52 (.04).58 (.05).66 (.05).53 (.04) y-gap.0 (.05).9 (.05).32 (.04).28 (.04).9 (.06).0 (.05).28 (.06) Prod_growh(HP).60 (.09).79 (.06).93 (.05).87 (.05).66 (.07).73 (.06).87 (.07) Wage-gap (.069) -.26 (.03) -.30 (.06) -.35 (.05) (.040) -.28 (.02) -.28 (.05) log(cpi )*Deflaion -.65 (.3) -.48 (.07) -.82 (.2) Wage-gap - *Deflaion.23 (.07).08 (.05).20 (.07) y-gap *Deflaion.28 (.2).4 (.07).07 (.08) Consan.03 (.002).006 (.002).005 (.002).008 (.00).03 (.002).007 (.003).007 (.007) Dependen variable: log(w ). All equaions are esimaed wih GMM. Insrumens are four lags of he dependen and he independen variables in each equaion. The sample period for all equaions 984: o 2003:2. Sandard errors are given in parenheses. The remaining columns in Table IV.2 conain esimaes of equaions inended o es wheher he wage adjusmen process is differen during he deflaion period since 997 compared o before. The esimaion sraegy is o creae a dummy variable, deflaion, which equals zero from 984: o 997:4 and one hereafer and o inerac his variable wih he CPI inflaion rae, he oupu gap, and he real wage gap. 5,6 When he reacion o he CPI inflaion rae is allowed o be differen in he deflaion period (column (2)) he coefficien falls o zero. 7 This implies a slower reacion of wages o consumer prices. I should be noed, however, ha he reacion o boh he real wage gap and he oupu gap is increased which parially, bu no fully, compensaes for he slower reacion o prices. The remaining columns essenially ell a similar sory of aenuaed reacions of wages o price shocks in he deflaion period. Column (3) shows 5 The produciviy growh rae is almos consan in he deflaion period so i was no possible o inerac his variable wih he deflaion dummy. 6 An alernaive sraegy would be o es wheher he adjusmen process is differen when he price level increases as compared o when i decreases. For our sample his is almos idenical o our sraegy because a dummy variable ha would disinguish beween increasing and decreasing CPI is he same as our deflaion dummy excep for hree observaions (ou of a oal of seveny six). The resuls wih he alernaive specificaion (available from he auhor on reques) are herefore very similar o he ones we repor in Tables IV.2 and IV.3. 7 The poin esimae is acually negaive, bu no significanly differen from zero. 22

24 ha he reacion o he real wage gap falls, and column (4) implies ha he reacion o boh prices and he real wage gap is lower. Column (6) shows ha when he reacion o boh price inflaion and he oupu gap are allowed o change, he former decreases, where he laer increases. The combined effec again implies a slower reacion of wages o price shocks. 8 Table IV.3 conain resuls for regressions ha are similar o hose in Table IV.2 excep ha he unemploymen rae replaces he oupu gap as he demand measure in equaion (9). The sign of he esimaes in col. () again correspond o wha he underlying model predics. In conras wih he previous able, however, he coefficien on he CPI inflaion rae is no significanly differen from zero implying ha ransmission of price shocks would occur only hrough he lagged effecs on he real wage gap and he unemploymen rae. When he ineracion erms are included in he regressions, he resuls in columns (4) and (6) show similar effecs as in he previous able. The response o prices is significan in he pre-deflaion period, bu fall o zero during he deflaion. The response o he real wage gap also falls during he deflaion period (col. 4), whereas he response o he unemploymen rae is essenially unchanged (col. 6). Table IV.3: Deerminans of wage inflaion () (2) (3) (4) (5) (6) (7) log(cpi ).02 (.0).44 (.09) -.02 (.09).35 (.06) -.03 (.05).57 (.07) (.034) Unemp -.02 (.003) -.05 (.003) (.004) -.02 (.002) (.002) -.00 (.002) (.002) Prod_growh(HP).2 (.08).42 (.08).6 (.05).62 (.05).25 (.07).57 (.07).57 (.3) Wage-gap (.06) -.28 (.08) -.23 (.08) -.35 (.05) -.27 (.06) -.37 (.06) -.42 (.0) log(cpi )*Deflaion -.8 (.22) -.47 (.06) -.64 (.05) Wage-gap - *Deflaion -.09 (.5).2 (.06).35 (.0) Unemp *Deflaion.003 (.00).0009 (.0008).006 (.00) Consan.2 (.02).065 (.03).3 (.02).057 (.008).3 (.0).036 (.009).2 (.0) Dependen variable: log(w ). All equaions are esimaed wih GMM. Insrumens are four lags of he dependen and he independen variables in each equaion. The sample period for all equaions is 984: o 8 The resuls presened in Tables IV.2 and IV.3 correspond o equaions where here is no dummy variable for he inercep. Adding such a dummy does alers some coefficien esimaes, especially in Table IV.3, bu does no modify our main conclusions. The corresponding resuls can be obained from he auhors on reques. I is worh noicing in his conex ha he deflaion dummy and log(cpi ) are highly collinear in our sample (he correlaion coefficien is -.85). 23

25 Taken ogeher he esimaes corroborae hose obained from he VAR model in secion III, where found ha he conribuion of foreign shock in he variance decomposiion of wages was relaively small a shor horizons. Our srucural inerpreaion in erms of he wage equaion esimaed here explains his by he limied direc effec of foreign prices in he wage equaion and by he significan lags involved in he ransmission via he real wage gap and he oupu gap/unemploymen rae. IV.3. The IS equaion. We have seen ha inflaion depends on wage developmens and ha wage developmens depend on domesic demand condiions. I remains o explore he deerminans of domesic demand. To his end we esimae a convenional IS-ype equaion of he form y f = 0 + β y + β 2 log( reer ) + β 3rr + β 4 y β (0) where y (y f ) sands for a measure of domesic (foreign) demand condiions, reer for he real exchange rae (measured such ha an increase represens a real appreciaion), and r for he real ineres rae. In he empirical applicaion he domesic and foreign demand condiions are measured by esimaed oupu gaps based on Hong Kong real GDP and real GDP of he main rading parners. Two ineres rae series were used, he US 3- monh T-Bill rae and he 3-monh Hibor rae. In view of he fixed exchange rae he wo follow each oher quie closely, bu a imes here have been significan differences, noably in he second half of 997 and in 998. To calculae real ineres raes, he year on year percenage change in he CPI was used as he inflaion measure. Esimaions were made using GMM wih four lags of he dependen and each of he independen variables as insrumens. As in he previous secions, we used quarerly spanning 984: o 2003:2. The resuls are presened in Table IV.4, column () when he US T-Bill rae is used and column (3) when i is replaced by he local 3-monh Hibor rae. No surprisingly, he exernal oupu gap has a srong influence on he Hong Kong business cycle. The coefficiens imply a close o one-for-one effec. The coefficien on he real exchange rae is negaive as expeced. A real appreciaion of en percenage poins will according o he poin esimaes reduce he oupu gap by abou four enh of one percenage poin. To pu his in perspecive, he oupu gap in Hong Kong declined abou 5 percenage poins beween he firs quarer of 997 and he firs quarer of 998. During he same ime he Hong Kong dollar appreciaed in real erms by 2%. Hence he esimaes in he able imply ha of he 5 percenage poin drop in he oupu gap only 0.5 can be explained by he real appreciaion. 9 9 These calculaions do no ake ino accoun he lagged effecs implied by he presence of he lagged oupu gap among he regressors. Taking his ino accoun, he long run effec of he real appreciaion would be. and.2 percenage poins respecively if we use only he IS equaion in he calculaions. However, he change in he oupu gap would elici changes in boh wages and prices as we have seen in he previous secions, so he full impac would have o be calculaed using hese equaions as well. 24

26 The effecs of ineres rae changes are negaive and significanly differen from zero. The esimaes using he HK rae are quie close o hose using he US rae as expeced given he high correlaion beween hem. The differences will concern mainly he inerpreaion of he downurn afer he 997 Asian crises ha we will commen on furher below. The size of he coefficiens imply ha an increase in he real ineres rae by one percenage poin will decrease he oupu gap by 0.32 percenage poins in he curren quarer (0.23 percenage poin using he Hong Kong ineres rae). The real Hibor rae increased by approximaely 3 percenage poins beween he hird quarer of 997 and he hird quarer of 998. The esimaes imply ha he direc impac of his was a decline in he oupu gap of abou 0.7 of a percenage poin. Columns (2) and (4) in he able are inroduced o es wheher he ransmission mechanism of foreign shocks has changed in he deflaion period following he Asia crisis. As in he previous secion we inerac one of he independen variables wih our deflaion dummy. We have limied he possible ineracion erms o he effec of he foreign oupu gap. The main consequence is a sriking increase in he impac of he exernal oupu gap in he pos 997 period. 20 For example, when he Hong Kong ineres rae is used (column 4) he response coefficien is.84 in he deflaion period as opposed o.50 before. Absen a heoreical explanaion of his change in response we simply offer his resul as an inriguing empirical regulariy a his sage. Table IV.4: The IS equaion () (2) (3) (4) y gap-world ().97 (.2).65 (.07).84 (.0).50 (.08) log[reer()] (.007) (.003) (.005) (.003) i US () (.0006) (.0004) i HK () (.0004) (.0002) y gap (-).52 (.04).62 (.03).6 (.03).64 (.02) y gap-world ()*deflaion().27 (.7).34 (.3) Consan.2 (.04).4 (.0).8 (.02).2 (.0) We end he discussion of he IS equaion by commening briefly on he residuals from he IS equaions in he sharp decline in economic aciviy during 998 and 999 and he pick-up in Figure IV. conains our measure of he oupu gap ogeher wih he residuals from hree of he equaions represened in Table IV.4 (he numbering of he shocks in he figure corresponds o he columns in he able). Consider he decline in he 20 Adding a dummy for he inercep as well does no aler his conclusion. 25