NAIRU: Is it Useful for Monetary Policy in Jamaica? Taffi Bryson 1 Research Department Research and Economic Programming Division Bank of Jamaica

Transcription

1 (DRAFT) Sepember 2008 NAIRU: Is i Useful for Moneary Policy in Jamaica? Taffi Bryson 1 Research Deparmen Research and Economic Programming Division Bank of Jamaica Absrac This paper esimaes he non-acceleraing inflaion rae of unemploymen (NAIRU) for Jamaica and examines is usefulness in he conduc of moneary policy. Using Kalman and HP filers, he sudy finds suppor for he hypohesis ha he unemploymen gap explains in par, he dynamics of inflaion in Jamaica. However, given he small coefficien derived from he model, he use of he NAIRU in policy decisions should be analyzed joinly wih oher indicaors. The produciviy-augmened NAIRU has a consisen negaive rend, which implies ha wage aspiraions end o exceed produciviy, wih a posiive impac on inflaion. Keywords: NAIRU; Unemploymen; Kalman filer; Phillips curve JEL Classificaion: C22, E31, C13 1 The auhor would like o hank all he saff of he Research Deparmen of he Bank of Jamaica for heir invaluable commens and conribuions o he developmen of his paper. All errors are, however, my own. The views expressed in his paper are hose of he auhor and do no necessarily reflec hose of he Bank of Jamaica

2 Table of Conens 1.0 Inroducion Concepual Issues NAIRU Limiaions of he use of he NAIRU for Moneary Policy Modeling he NAIRU Srucural Mehod Purely Saisical Mehod Reduced Form Mehods Recen Empirical Sudies Economeric Model The Model Incorporaing Produciviy Esimaion Issues Daa Resuls Conclusion References Appendix

3 1.0 Inroducion The non-acceleraing inflaion rae of unemploymen (NAIRU), beer known as he naural rae of unemploymen, is he rae of unemploymen ha is consisen wih he seady sae rae of inflaion in he absence of supply shocks (Saiger e al, 1997). In he shor run, he Phillips curve posis ha he unemploymen rae plays a role in he ransmission from unanicipaed changes in aggregae demand o inflaion. More specifically, models ha incorporae Philips curve analysis show ha increases in demand enhance real Gross Domesic Produc (GDP) relaive o is poenial level, hereby increasing he demand for labour and hence lowering he unemploymen rae relaive o he NAIRU. Empirical evidence has shown ha he NAIRU can be a good leading indicaor of fuure inflaion when he economy undergoes demand shocks. However, he NAIRU may give misleading policy signals when here is a supply shock, such as an increase in produciviy. An unexpeced increase in produciviy causes downward price pressures (as aggregae supply exceeds aggregae demand) as well as a reducion in he unemploymen rae (aribued o he increase in real GDP and is associaed increase in demand for labour). In his conex, policy makers may misinerpre he reducion in he unemploymen level relaive o he NAIRU as a reason o ighen policy in fear of higher inflaion This paper esimaes a ime varian NAIRU for Jamaica beween 1994 and 2007 using unobserved componens models. Sae space models are used as sincehe NAIRU is unobserved bu is assumed o be persisen. The analysis explicily accouns for changes in labour produciviy in he economy. The implicaions of he NAIRU in he assessmen of inflaionary pressures are also examined. This is moivaed he hypohesis ha increased produciviy growh in conjuncion wih ineria in real wage aspiraion on he par of workers allows unemploymen o decline below he non- acceleraing inflaion level wihou generaing inflaionary pressures (Ball & Moffi, 2001 and Gruber, 2003). Given ha moneary policy should be pre-empive o conrol for inflaion and ha he effecs of moneary policy have long lags (Esrella and Mishkin, 1998), he esimaion of he Jamaican NAIRU is an aemp o include he NAIRU as an addiional ool in making more informed policy decisions. 2

4 The paper is developed as follows: Secion 2 will review he heoreical underpinnings of he NAIRU. Secion 3 ses ou he modelling issues described in he lieraure and he daa properies. Secion 4 presens formulaion of he economeric model. Finally, secion 5 gives he resuls of he esimaions and secion 6 concludes wih a discussion of he implicaion of he resuls. 2.0 Concepual Issues 2.1 NAIRU In heory, whenever he unemploymen rae persiss below he NAIRU (refers o a igh labour marke), he rae of inflaion ends o rise and vice versa. The Phillips curve relaionship implies ha inflaion will sabilize a he permanenly higher level. Tigh labour markes induce employees o bid for higher wages, which given ha labour is generally he larges single componen of producion, implies ha persisen increases in wages should evenually resul in increased prices by firms and hence economy-wide price inflaion hrough he cos-push inflaion mechanism. (Espinosa-Vega and Russell, 1997). Given he large and susained changes in unemploymen, mos of he lieraure has focused on economeric models ha allow for variaion in he NAIRU overime. The NAIRU in he shor run is herefore more volaile and ofen affeced by srucural changes as well as supply shocks (Saiger, Sock, & Wason, 1995). The deerminans of he NAIRU are assumed o be influenced by iner alia demographics and echnological changes which give meri for he variable o be modelled as persisen (Laubach, 2001). Uncerainy abou he NAIRU does no render he Phillips curve useless for he conduc of moneary policy as posied by King and Wason (1994) who found ha unemploymen rae Granger-causes inflaion. 2.2 Limiaions of he use of he NAIRU for Moneary Policy While he applicaion of he NAIRU is deemed imporan for he conduc of moneary policy is use has o be seen wihin he conex of he uncerainy as idenified by Sagier, 3

5 Sock and Wason (1997). This uncerainy relaes o he specificaion of a proper model, he smoohness parameer and he margin of error associaed wih he esimaed parameers. The empirical lieraure generally finds large sandard error bands around he Kalman filer esimaes (Greenslade (2003), Slacelak (2005), Sagier e al., (1997)). In addiion, he shor-run rade-off beween unemploymen and inflaion may be unsable over ime and is sensiive o he way inflaion expecaions are formed. Hence any radeoff would end o disappear if policy makers aemped o exploi i sysemaically. Neverheless, Esrella and Mishkin (1998) conclude ha he above-menioned issues do no negae he use of he NAIRU bu raher reduce he magniude of he policy response o he indicaor. Esrella and Mishkin (1998) also noe ha he divergence of inflaion from is arge is of equal imporance as he NAIRU gap in developing a policy sance. However, hey warn agains he improper use of he NAIRU as a poenial arge given is shor run consruc. 3.0 Modelling he NAIRU The echniques used in he exising lieraure o esimae he NAIRU can be broadly classified ino hree caegories: srucural, saisical and reduced-form mehods. 3.1 Srucural Mehod The srucural mehod involves modelling aggregae wage and price seing behaviour in a sysem of equaions. The esimaion of he NAIRU is made under he assumpion ha markes are in full or someimes parial equilibrium (Richardson, 2000) 2. The model presupposes full adjusmen of firms and workers o all shocks. As such, he derived unemploymen corresponds o long run concep of he NAIRU which conrols for shocks raher han he shor run NAIRU which commonly appears in he reduced-form Phillips curve specificaion. Srucural models are useful as hey presen more informaion on he deerminans of he NAIRU given he heoreical relaionship beween macroeconomic shocks, policy insrumens and he long run equilibrium rae of unemploymen. However, heir 2 The equilibrium level of unemploymen is obained as a se of values for which inflaion is sable (Szeo and Guy, 2004). 4

6 shorcomings include dependence on he assumpions of he underlying behaviour of economic agens for which here is no general consensus. Furher, he inricacy of he esimaion prevens he imely calculaion of he esimaes and here exiss a number of economeric and measuremen issues associaed wih he mehod. 3.2 Purely Saisical Mehod Purely saisical mehods dichoomize he unemploymen rae ino is rend and cyclical componens, wihou accouning for he relaionship beween he unemploymen rae and inflaion. These models are inended o be predicive and rely on he assumpion ha equilibraing forces are swif and sufficien in bringing he unemploymen rae o is rend, so ha on average, he unemploymen rae will be close o he NAIRU. While purely saisical mehods can be esimaed in a imely and consisen way, hey are ofen dependen on arbirary assumpions. These assumpions relae o he modeling of he esimaed rend, is variance as well as he relaionship beween he rend and he cyclical componens. This is he case of he HP filer which idenifies unemploymen as a weighed average of acual unemploymen. However, given ha filers work as moving averages hey ofen respond slowly o apparen changes in unemploymen. 3.3 Reduced Form Mehods The reduced form mehod of esimaion of he NAIRU is grounded in he heoreical underpinnings of he expecaion-augmened Phillips curve, which allows for addiional facors oher han he inflaion/unemploymen relaionship. The reduced form approach has become he mos popular echnique in recen sudies. Noneheless, he reduced form approach is by no means perfec and has a number of disadvanages. These disadvanages include he heavy dependence of he NAIRU esimaes on he esimaion of inflaion expecaions as well as he mehod s aheoreical approach of esimaing he NAIRU via he reduced form mehod, which does no idenify he underlying srucural inflaion/unemploymen relaionship (Richardson e al, 2000). Also, he filers lack precision for end of sample esimaes, while he resuls are liable o be sensiive o arbirary choices of variance parameers such as he signal o noise raio. Sock and Wason s (1998) procedure is propose in order o obain median-unbiased esimaes of 5

7 he raio of he variance parameers (i.e. signal o noise raio). This paper follows he approach of Laubach (2001) in fixing he variance parameers. 3.4 Recen Empirical Sudies Richardson e al (2000) using he Kalman filer procedure, joinly esimae he Phillips curve and Time-Varying NAIRU for he 21 OECD member counries over he period 1980 o The shocks used in he sudy are real impor and oil prices. The resuls found suppor for he saisical significance of he NAIRU in explaining movemens in he inflaion. Laubach (2001) using a similar framework as Richardson e al (2000) includes a drif in he specificaion of he NAIRU and models he NAIRU as a persisen sochasic process. The shock variables used in his paper are he nominal exchange rae, commodiy prices and wo measures of price changes; he CPI and he GDP deflaor. Laubach (2001) examines he NAIRU beween 1971 and 1998 for he G7 member counries excluding Japan and Ausralia. The resuls show suppor for he significance of he NAIRU for he Unied Saes in predicing changes in infalion, however, for mos European counries he NAIRU specificaion does no explain he join progression of unemploymen and inflaion. Slacalek (2005) uses a similar framework o ha of Laubach (2001) and Richardson e al (2000) bu uses a produciviy-augmened Phillips curve. The approach adds o he augmened Phillips curve by modelling he NAIRU wih informaion conained in he rend of produciviy as opposed o modelling NAIRU as a purely persisen sochasic process. The main assumpion is ha produciviy has wo componens, namely, he capializaion effec and he creaive desrucion effec 3. Using quarerly daa beween 1960 and 2002 for he Unied Saes, Slacalek (2005) found he ne impac of he wo effecs o have a negaive correlaion beween produciviy growh rend and he NAIRU. 3 The capializaion effec represens he creaion of jobs due o increased labour produciviy. This effec saes ha labour produciviy is negaively relaed o unemploymen. The creaive desrucion effec represens he srucural change where increases in produciviy shorens employmen duraion and raises he NAIRU. The correlaion beween he produciviy growh and he NAIRU herefore depends on he relaive size of hese wo effecs. 6

8 Greenslade e al (2003) applies a reduced form framework of he Philip s curve o esimae he NAIRU for he Unied Kingdom from 1973 o The supply shock variables used are similar o Richardson e al (2000), real impor prices and oil prices. In addiion o examining he impac of he unemploymen gap o price inflaion, he also esimaes he relaionship beween he unemploymen gap and wage inflaion. The resuls found suppor for he use of he NAIRU (albei wih some level of uncerainy) as well as he imporance of he supply shocks in analyzing inflaionary pressure. 4.0 Economeric Model 4.1 The Model The NAIRU is inferred from a Phillips curve-ype regression of he following form: π π = β ( L)( π + ε (1) e e 1 π 1 ) + γ ( L)( U 1 U 1 ) + δ ( L) X 1 where β (L), γ (L) and δ (L) are he polynomial in he lag operaors, realized and expeced inflaion, respecively. π and e π denoe U denoes he NAIRU a ime, X is a vecor of variables capuring supply shocks and he random exogenous even, ε. Inuiively, he unemploymen gap is defined as ( U U ) and is negaively relaed o inflaion. 1 1 Given he aforemenioned resriciveness of he consan NAIRU assumpion as well as he pauciy of empirical evidence o subsaniae he assumpion, a ime varying parameer model (Kalman filer) is used o capure he srucural changes of he labour and commodiy markes. Specifically he model becomes: 2 Δπ = β ( L) Δπ L U U L X + ε ε ~ N(0, ) (2) 1 + γ ( )( 1 1 ) + δ ( ) 1 σ ε 2 U η + v v ~ N(0, σ ) = U 1 v 7

9 Here he random exogenous evens ε and v are assumed o be i.i.d. normal wih mean zero and variance 2 σ ε and σ, respecively, and cov( ε, ) = 0. 2 v v The sysem represens an expecaion-augmened Phillips curve consising of he Phillips curve (he firs equaion of (2)), which models he unexpeced inflaion as a funcion of pas deviaions from expeced inflaion, he unemploymen gap and shocks. Expecaions are assumed o be adapive and exhibi ineria represened by lags of inflaion, e i.e. π = π 1. The second equaion of (2) represens he ime varian NAIRU. The univariae NAIRU is modelled as an unobserved sochasic componen, assumed o follow a random walk. Given he absence of an upward rend in Jamaica s unemploymen daa i is plausible o assume ha he NAIRU can be specified as a random walk wihou drif (Laubach, 2001) 4. The assumpion of he size of he sandard error of v deermines how he NAIRU will move from quarer o quarer. Given he seady sae concep of he NAIRU, Gordon (1997) posulaes a smoohness prerequisie ha allows he NAIRU o move around wihou sharp quarer o quarer volailiy. Inuiively, he variaion of v is usually small. Similar o Laubach (2001), a bivariae specificaion is used wih he assumpion ha he unemploymen gap has a endency o rever o a zero mean over ime hereby imposing some srucure on he unemploymen gap. i.e., 1 U 1 = φ ( L)( U 1 U 1 ) k where ( 1 ) < 1 U + φ (3) The assumpion ha he unemploymen gap follows a saionary process yields addiional informaion abou he NAIRU (Laubach, 2001). This imposiion is in line wih Friedman s 1968 naural-rae hypohesis ha he unemploymen rae can be kep away from is naural rae only by ever acceleraing inflaion or deflaion. 4 Laubach (2001) modelled he U.S. he NAIRU as a random walk wihou drif given he mean reversion naure of is unemploymen daa. 8

10 The Kalman filer esimaes he reduced-form Phillips curve joinly wih he NAIRU 5. In he esimaion of he Kalman filer an ieraive procedure is used o idenify he NAIRU series and he coefficien on he unemploymen gap unil convergence is achieved. The paper draws on boh Laubach (2001) and Ball and Moffi (2001) who considers a specificaion of he NAIRU wihou drif. The repored NAIRUs are calculaed such ha he enire Phillips curve error is assigned o he variaion of he NAIRU (i.e. in each year he ime varying NAIRU is he value of he NAIRU ha would se he prediced value of inflaion equal o he acual value of inflaion). 4.2 Incorporaing Produciviy The mehodology of incorporaing produciviy in he Phillips curve emanaes from Ball and Moffi (2001), amongs ohers, who sough o explain he low unemploymen and low inflaion environmen in he Unied Saes in he 1990s. Ball and Moffi (2001) found ha he increased growh rae in produciviy ha occurred around he same ime was responsible for he change in he levels of he unemploymen and inflaion rade-off. The argumen proposed by he auhors is enrenched in he idea ha workers wage aspiraion adjus slowly o shifs in produciviy. The model assumes iniially ha wage adjusmens are largely based on pas wage increases and less on produciviy. Secondly, he model assumes ha wage inflaion depends negaively on unemploymen. The produciviy augmened Phillips curve is a combinaion of a price-seing and wageseing equaions. Assuming mark-ups are consan price seing in his model is given by π = ω θ + υ (4) 5 The reduced form approach was also he framework of choice in he recen sudies of Laubach (2001), Szeo e al (2004) and Gordon (1997) as opposed o he srucural models. The main reason for he selecion of he reduced form was ha he srucural model framework assumes full adjusmen o all shocks. 9

11 where inflaion, π, is deermined by he difference beween nominal wage growh, and he growh rae of produciviy, θ, and an error erm, υ. Mark-ups are assumed o be consan. Increases in wages above produciviy will ranslae ino upward movemens in prices. ω, The wage-seing process assumes ha in he seady sae, where here is flexibiliy, he equilibrium wage growh equal o he rae of inflaion plus he rae of produciviy growh. However, Ball and Moffi (2001) inroduce ineria ino he real wage adjusmen process and sugges ha workers, in addiion o examining conemporaneous inflaion and produciviy when seing heir wages, also look a pas levels of real wage growh which is capured in a wage aspiraions erm. These wage aspiraions, pas wage increases and are defined as: A, depend on A 1 β = β i= 1 i β ( ω π ) i i (5a) where β is he discoun facor which is he weigh placed on pas levels of real wage growh. Rewriing equaion 5a recursively gives: A ( β )( ω π ) = βa (5b) For he aspiraion erm, he iniial value of A is se equal o he saring value of he HPfilered real wage growh series. The discoun parameer is se equal o Theoreically, workers real wage aspiraions depend on a weighed average of curren produciviy growh and curren and pas produciviy wages encompassed in, as unemploymen, U. A, as well 6 Taken from Ball and Moffi (2001) and Gruber (2003). 10

12 Combining equaions (4) and (5a), A is he discouned sum of pas levels of produciviy growh and is a weighed average of pas increases wih exponenially declining weighs. The workers arge level of real wage growh becomes: ( ω E π ) = α γu + δθ + ( 1 δ ) A + 1 η (6) where E 1 is he expecaions operaor. Unemploymen is negaively relaed o he workers real wage growh arge, since as he number of job seekers increase, workers end o revise downwards heir wage expecaions. By subsiuing he wage-seing equaion (6) ino he price seing equaion (4), assuming adapive expecaions (i.e. Eπ = π 1) and allowing supply shocks, S, he produciviyaugmened Phillips curve is: ( U 1 U 1 ) + δ ( L X 1 f1( θ 1 A ) ε π ( + (7) = b L) π 1 + c1 ) 1 2 U η + v v ~ N(0, σ ) = U 1 v Equaion (7) implies ha inflaion is negaively associaed wih he excess produciviy growh over aspiraions of real wage growh. In he seady sae, changes in produciviy are direcly proporionae o changes in wage aspiraion ( θ = A ). In he shor run, however, movemen in produciviy θ is no mached immediaely by a shif in A which resuls in downward pressures on inflaion. In effec, movemens in θ A are reaed as persisen supply shocks ha shif he Phillips curve for a given NAIRU. In summary he specificaions esimaed in he paper are as follows: 1. Univariae NAIRU model (equaion (2)). 2. Univariae NAIRU model including produciviy (Equaion (7) and (9)). 11

13 3. Bivariae NAIRU model (equaions (3) and (2)). 4. Bivariae NAIRU model including produciviy (Equaion (3), (7) and (9)). 4.3 Esimaion issues The esimaion of he unobserved componens models, faces a pile up problem when here are non saionary sae variable. The Maximum Likelihood Esimaes of he signal o noise raio ( σ v ) has a poin mass of zero even if he rue value is greaer han zero σ τ (Sock, 1994). This paper adops he approach of Laubach (2001) in fixing he parameers σ andσ by using a diffuse prior. Secondly, wo alernaive mehods were used in his ε v paper o choose he iniial value for he sae, U 0. The firs approach chooses U = 0 U in compuing he iniial sae and is covariance marix for he bivariae specificaions. For he univariae specificaions he paper follows Laubach (2001) in defining a variable Z U U and re-wriing equaion (4) as 0 Δπ = c + β ( L) Δπ + ε i = 1... n (10) z + v = z γ ( L)( u 1 z 1 ) + δ ( L) X 1 The iniial value ( u 0 ) is given by c γ (L). In he specificaion he iniial value of z 0 is zero by definiion. 4.4 Daa The sudy uses quarerly daa over he period 1994:1 o 2007:4. The measure of inflaion used in he sudy is he change in he Consumer Price Index (CPI). Jamaica s official unemploymen rae is defined as hose persons of working age (14 years and older) who were (a) wihou work during he reference period; (b) currenly available for work; (c) seeking work, and also includes persons who are no seeking work 7. The unemploymen 7 This relaxed definiion follows he inernaional sandards on he basis ha in many developing counries he labour marke is largely unorganised, mos workers are self-employed or informally employed and ha 12

14 rae was no enered conemporaneously so as o avoid he simulaneiy issues. Expecaions are assumed o be adapive and exhibi ineria represened by lags of inflaion. Produciviy growh is he change in he log of oupu per worker. The shor erm shock variable, which is expeced o rever o zero afer a year, is impor prices measured by an index of impored raw maerials and food prices. Laubach (2001) in heir sudy found he coefficiens for he univariae model ranging beween 0.09 and The shock series was aken as deviaions from is mean, o ensure ha in he seady sae i does no impac he rae of inflaion. Wage aspiraions are measured as a weighed sum of discouned pas levels of real wage growh. Figure 1 shows ha boh he unemploymen rae and he inflaion series were relaive sable beween 1995 and Inflaion ranged beween 0.0 and 4.0 per cen while unemploymen ranged beween 16.1 o 17.4 per cen. Unemploymen declined sharply in 2003 wih a similar movemen in inflaion. The decline in he unemploymen rae could be aribued o he consrucion boom ha sared during he early 2000 s. The consrucion boom involved significan consrucion of highways, roads and hoels, which increased employmen in he respecive as well as suppor indusries. The remainder of he period under sudy was characerised by a declining unemploymen rae ranging beween 13.1 and 9.6 per cen while inflaion was more volaile and ranged beween -0.3 and 7.9 per cen. For he sample period, he unemploymen gap can be assumed o have averaged close o zero as he inflaion daa does no exhibi explosive behaviour (Laubach, 2001). here may be siuaions in which persons do no acively seek work because hey believe ha here will be no demand for heir skill. 13

15 Figure 1: Jamaica Unemploymen rae and Inflaion Inflaion Unemploymen Rae Inflaion Rae (%) Q1 1994Q4 1995Q3 1996Q2 1997Q1 1997Q4 1998Q3 1999Q2 2000Q1 2000Q4 2001Q3 2002Q2 2003Q1 2003Q4 2004Q3 2005Q2 2006Q1 2006Q4 2007Q3 Figure 2: Jamaica Labour Produciviy minus Wage Aspiraions 1997Q2 1997Q4 1998Q Q4 1999Q Q Q Q Q2 2001Q Q Q4 2003Q2 2003Q4 2004Q2 2004Q4 2005Q2 2005Q4 2006Q Q Q2 2007Q4 Excess Wage Aspiraion Exc ess Wage Aspiraion HP Filer Unemploymen Rae (%) 14

16 5.0 Resuls The difference beween θ and A is graphed wih is HP-filered rend, in figure 2. The consisen negaive rend implies ha wage aspiraions of real wage growh exceed produciviy, wih a consequen posiive impac on inflaion. The resuls illusrae ha he coefficien of he produciviy erm is negaive and significan and ha he supply shock is posiively correlaed wih inflaion. The resuls from he esimaion of each of he specificaion are presened in ables 1 hrough 4. They demonsrae a fairly srong negaive relaionship beween he unemploymen gap and inflaion wih he coefficien being significanly differen from zero a he 5% level. The esimaes of he coefficien on he unemploymen gap ranges beween and -0.15, and implies ha increases in employmen, above full employmen level, will lead o increased inflaion. Table 1: Univariae NAIRU Model Mehod: Maximum likelihood (Marquard) Sample: 1994Q1 2008Q1 Included observaions: 55 Coefficien Sd. Error Prob. ( u C u 1) Δπ X Log likelihood NAIRU Average SE Diffuse priors 5 15

17 Table 2: Bivariae NAIRU Model Mehod: Maximum likelihood (Marquard) Sample: 1994Q1 2008Q1 Included observaions: 55 Coefficien Sd. Error Prob. ( u 1 u 1) E π Δ 1 X E Log likelihood NAIRU Average SE Table 3: Univariae NAIRU Model wih Produciviy Mehod: Maximum likelihood (Marquard) Sample: 1994Q1 2008Q1 Included observaions: 44 Coefficien Sd. Error Prob. ( u C u 1) E Δπ E X ( ) θ A E Log likelihood NAIRU Average SE Diffuse priors 5 16

18 Table 4: Bivariae NAIRU Model wih Produciviy Mehod: Maximum likelihood (Marquard) Included observaions: 44 ( u Coefficien Sd. Error Prob. 1 u 1) E Δπ E X ( θ A ) Log likelihood NAIRU Average SE Examining he log likelihood saisics of he smoohed NAIRU esimaes, he resuls indicae ha he bivariae specificaions (ables 2 and 4) ouperform he univariae specificaion (ables 1 and 3). The resuls indicae ha he significance of he unemploymen gap increases when he produciviy erm is added o boh he univariae and bivariae NAIRU specificaions (see Tables1, 2, 3, and 4). The bivariae model wih produciviy is herefore he preferred model. A comparison of he resuls repored in ables 3 and 4 shows ha by adding an auoregressive srucure of he unemploymen gap o he Phillips equaion lead o an increase in size of he unemploymen gap. The Kalman filer NAIRU is esimaed as an auoregressive process ha mirrors he random walk specificaion, as seen in oher sudies. Inuiively, he auo-regressive form is consisen wih he NAIRU adjusing only slowly o enduring supply shocks. Furhermore, he auoregressive srucure is imporan in a shor erm forecasing conex as changes in he esimaed NAIRU over he recen pas may provide informaion relevan o is likely fuure profile (Richardson, 2000). The Kalman smooher was used o obain he poin esimaes of each specificaion (see figure 3, 4 and 5 in he appendix). Resuls from he granger causaliy ess rejec he hypohesis ha changes in he unemploymen gap does no Granger cause changes in inflaion for he bivariae model 17

19 wih produciviy (able 5) 8. Addiionally, granger causaliy ess rejec he hypohesis ha he unemploymen gap of he bivariae specificaion does no Granger cause wages. The hypohesis ha changes in wages does no Granger cause changes in inflaion was rejeced a he 10 percen level which maybe indicaive of feedback. 6.0 Conclusion The concep of he NAIRU can play an imporan role in he conduc of moneary policy. This paper esimaes and assesses several specificaion of he ime varying NAIRU for Jamaica and examines is usefulness for he conduc of moneary policy. Using a Kalman filer framework, he paper finds ha he NAIRU esimaes are considerably improved when produciviy is accouned for. The resuls demonsrae a negaive relaionship beween he unemploymen gap and inflaion. The produciviy-augmened NAIRU has a consisen negaive rend, which implies ha wage aspiraions of real wage growh end o exceed produciviy, which have a posiive impac on inflaion. Using he log likelihood es he bivariae specificaion wih produciviy produces more precise esimaes of he NAIRU. Granger causaliy es rejec he hypohesis ha he unemploymen gap does no granger cause changes in inflaion. The downward sloping excess wages shows ha coninuous wage increases above he produciviy is significan and causes increased inflaion. 8 This resul was found for all models wih he excepion of he univariae model wihou produciviy. 18

20 References Andrews, Donald and Werner Ploberger (1994), Opimal Tes When a Nuisance Parameer is Presen Only under he Alernaive, Economerica, Vol. 62, No. 6, pp Ball and Mankiw, (2002), The NAIRU in Theory and Pracice, The Journal of Economic Perspecives, Vol. 16, No. 4, pp Ball, L. and Moffi, (2001), Produciviy and he Phillips curve, John Hopkins Universiy. Espinosa-Vega, M. and Russell, S, (1997), Hisory and Theory of he NAIRU: A Criical Review, Economic Review, Federal Reserve Bank of Alana, Second Quarer, pp Esrella, A. and Mishkin, F.(1998), Rehinking he Role of NAIRU in Moneary Policy: Implicaions of Model Formulaion and Uncerainy Federal Reserve Bank of New York. Research Paper No Gordon, R. and Sock, J. (1998), Foundaions of he Goldilocks Economy: Supply Shocks and he Time-varying NAIRU, Brookings Papers on Ecnonmic Aciviy, Vol. 1998, No. 2, and pp Gordon, R. (1997), The Time-Varying NAIRU and is Implicaions for Economic Policy, The Journal of Economic Perspecives, Vol. 11, No. 4, pp Greenslade, J., Pierse, R. and Saleheen, J.(2003), A Kalman Filer Approach o Esimaing he UK NAIRU, Bank of England Working Paper No. 179 Gruber, J. (2003), Produciviy Growh and he Phillips Curve in Canada, Inernaional Finance Discussion Papers, Board of Governors of he Federal Reserve Sysem, No King, R. and Wason, M. (1994), The Pos-War U.S. Phillips Curve: A Revisionis Economeric Hisory, Carneige-Rocheser Conference Series on Public Policy Vol. 41, pp Laubach, T. (2001), Measuring he NAIRU: Evidence from Seven Economies, The Review and Saisics, Vol.82, No.2, pp Richardson, Pee, L. Boone, C. Giorno, M. Meacci, D. Rae and D. Turner (2000), The Concep, Policy Use and Measuremen Of Srucural Unemploymen: Esimaing A Time Varying NAIRU Across 21 Oecd Counries, Paris, Organisaion for 19

21 Economic Co-Operaion and Developmen, Economic Deparmen Working Papers No Slacalek, J. (2005), Produciviy and he Naural Rae of Unemploymen, Discussion Papers of DIW Berlin., DIW Berlin, German Insiue for Economic Research, No. 461 Sagier, D., Sock, J. and Wason, M. (1997), The NAIRU, Unemploymen And Moneary Policy, The Journal of Economic Perspecives, Vol. 11, No. 4, pp Sephanides, G. (2006), Measuring he NAIRU: Evidence from he European Union, USA and Japan, Inernaional Research Journal of Finance and Economics, Issue 1. Sock, W. and Wason (1998), Asympoically Median Unbiased Esimaion of Coefficien Variance in a Time Varying Parameer Model, NBER Technical Working Paper No. 201 Szeo, K. and Guy, M. (2004), Esimaing a New Zealand NAIRU, Wellingon, New Zealand Treasury, Working Paper No 04/10. 20

22 Appendix 21

23 Figure 3: Univariae NAIRU Model NAIRU Unemploymen Rae Figure 4: Univariae NAIRU Model Including Produciviy NAIRU Unemploymen Rae 22

24 Figure 5: Bivariae NAIRU Mmodel NAIRU Unemploymen Rae Figure 6: Bivariae NAIRU Model Including Produciviy Unemploymen Rae NAIRU 23

25 Table 5: Granger Causaliy of he Bivariae NAIRU Model wih Produciviy Pairwise Granger Causaliy Tess Sample: 1996Q1 2007Q4 Lags: 1 Null Hypohesis: Obs F-Saisic Probabiliy Unemploymen Gap does no Granger Cause WAGES WAGES does no Granger Cause Unemploymen Gap Inflaion changes does no Granger Cause WAGES WAGES does no Granger Cause Inflaion changes Inflaion changes does no Granger Cause Unemploymen Gap Unemploymen Gap does no Granger Cause Inflaion changes