The Labour Market in Macroeconomic Models of the Australian Economy *

Transcription

1 The Labour Marke in Macroeconomic Models of he Ausralian Economy * Jim Thomson Melbourne Insiue of Applied Economic and Social Research The Universiy of Melbourne Melbourne Insiue Working Paper No. 8/00 ISSN ISBN Sepember 2000 *This paper is he resul of work being underaken as par of a collaboraive research program eniled Unemploymen: Economic Analysis and Policy Modelling. The projec is generously suppored by he Ausralian Research Council and he following collaboraive parners: Commonwealh Deparmen of Family and Communiy Services, Commonwealh Deparmen of Employmen, Workplace Relaions and Small Business and he Produciviy Commission. The views expressed in his paper represen hose of he auhors and are no necessarily he views of he collaboraive parners. Melbourne Insiue of Applied Economic and Social Research The Universiy of Melbourne Vicoria 300 Ausralia Telephone (03) Fax (03) melb.ins@iaesr.unimelb.edu.au WWW Address hp://

2 Absrac This paper examines he reamen of he labour marke in hree macroeconomeric models of he Ausralian economy: he Ausralian Treasury Macroeconomic (TRYM) model, he Access Economics Macro (AEM) model, and he Murphy model. In each of hese models, employmen and unemploymen are basically deermined a he aggregae level (hough in he Murphy model, labour demand is deermined a he indusry level). The unemploymen rae converges in he long run o an equilibrium level a which he average rae of real-wage inflaion across he economy is equal o he rae of produciviy growh in he economy. This rae, he non-acceleraing inflaion rae of unemploymen or NAIRU, is given by an expecaions-augmened Phillips curve. In each of he models, he NAIRU is reaed as exogenous and is value is esimaed as a parameer of he model. In he shor run, expeced wage inflaion depends on deviaions of he unemploymen rae from he NAIRU and on a number of oher variables including, in all he models, he change in he unemploymen rae (a speed-limi effec). This makes i possible o define a shor-run or flexible NAIRU as he unemploymen a which expeced real-wage inflaion equals he rae of produciviy growh, and his shor-run NAIRU depends on lagged unemploymen. 2

3 . Inroducion. General inroducion This paper examines he reamen of he labour marke in hree macroeconomeric models of he Ausralian economy: he Treasury Macroeconomic (TRYM) model of he Ausralian economy, he Access Economics Macro (AEM) model, and he Murphy model. There are srong similariies beween he models, in par because hey all descend (hough ofen wih subsanial modificaions) from he Naional Income Forecasing (NIF) models developed in he Commonwealh Treasury in he 970s and 980s. The similariy is especially srong wih respec o he labour marke, in paricular in relaion o unemploymen. There are hree key aspecs o his: Firs, in each of he models employmen and unemploymen are basically deermined a he aggregae level ha is, i is no buil up from indusry-specific, skill-specific, sae-specific, age-specific or gender-specific levels. The TRYM model works enirely a he aggregae level. The AEM model includes indusry and sae deail for oupu and employmen, bu his is a decomposiion of he corresponding aggregae variables. In he Murphy model, labour demand is deermined a he indusry level, bu he unemploymen rae is an aggregae variable. Second, in each of he models, he unemploymen rae converges in he long run o an equilibrium level a which he average rae of wage equaion across he economy is equal o he rae of produciviy growh in he economy. This rae, he nonacceleraing inflaion rae of unemploymen or NAIRU, is given by an expecaionsaugmened Phillips curve. In each of he models, he NAIRU is consan and is reaed as exogenous ha is, is value is esimaed as a parameer of he model an is no explicily relaed o oher aspecs of he economy. Third, expeced wage inflaion in he shor run depends on deviaions of he unemploymen rae from he NAIRU and also on several oher variables. In paricular, all models include he change in he unemploymen rae as a deerminan of expeced wage inflaion. This means ha, in he shor run, he unemploymen rae a which expeced wage inflaion equals he rae of produciviy growh is no equal 3

4 o he NAIRU unless he unemploymen rae is already equal o he NAIRU. This makes i possible o define a shor-run or flexible NAIRU as he unemploymen a which expeced wage inflaion equals he rae of produciviy growh, and his shor-run NAIRU will depend on lagged inflaion. The shor-run NAIRU is discussed on page 28 in subsecion for he TRYM model, on page 46 in subsecion 3.3. for he AEM model, and on page 5 in subsecion 4.3. for he Murphy model. The paper is divided ino four secions: an inroducory secion (Secion ) and secions given deails of he hree models examined Secion 2 on he TRYM model, Secion 3 on he AEM model, and Secion 4 on he Murphy model. The remainder of his secion discusses by way of background a number of elemens ha are common o all he models (subsecion.2) and gives a broad overview of he way he labour marke is handled in he models. Secions 2, 3 and 4 are each divided ino hree pars. Firs hey give some brief background on he model being deal wih in he secion; second hey give an overview of he way labour marke is reaed in he relevan model; and hird hey describe in deail he main labour marke equaions in he model. The equaions of he models as presened in he secions use more a exbook syle noaion han he documenaion for he models, which use more compuer-language syles. Appendixes D, E and F for he TRYM model, he AEM model, and he Murphy model, respecively give deails of how he noaion used here corresponds o ha used in he documenaion..2 Common elemens of macro models There are several elemens in common o mos of he macro models reviewed here. Firs, he TRYM, AEM and Murphy models all have equilibrium pahs o which, in he absence of shor-run shocks, he economy converges in he long run. Second, he longrun unemploymen level is deermined by a Phillips curve relaion ha links he growh rae of he expeced real wage o deviaions of he unemploymen rae from a naural rae or NAIRU. Third, he TRYM model includes unfilled vacancies as par of is measure of labour demand and models he Beveridge curve he relaionship beween he vacancy rae and he unemploymen rae. And fourh, he models inegrae he long- 4

5 run equilibrium pah of he economy wih shor-run dynamics in an error-correcion framework. This secion describes hese four elemens..2. Long-run equilibrium The macro models all feaure a long-run equilibrium growh pah o which he economy is coninually drawn. This is usually he pah he economy would follow if here were no shocks, prices were perfecly flexible, and agens had eiher raional expecaions or perfec foresigh. Shocks push he economy away from his equilibrium pah, bu here are forces in he economy ha draw he economy back. The speed a which he economy reurns depends on he size of physical adjusmen coss, such as hose in insalling new capial or maching workers wih job vacancies, and he speed wih which expecaions adjus o changes in he economy. The models assume a consan rae of labour-augmening echnological progress, λ say. Along he equilibrium growh pah, he real ineres rae is consan and equal o he world ineres rae, and he unemploymen rae is also consan, and equal o he NAIRU. All oupu relaed variables grow wih produciviy and labour inpus. Tha is, he real hourly wage grows a rae λ; GDP and is componens, measured per uni of labour inpu, grow a rae λ; and he capial-oupu raio is consan. This sor of long-run equilibrium is basically he one implied by he neo-classical growh model hough ha model does no usually include explici feaures ha would give rise o unemploymen. In mos respecs he macro models examined here are no consruced by adding shor-run adjusmen coss and sluggish expecaions o he neoclassical growh model raher hey are consruced in a more ad hoc manner, incorporaing feaures ha will cause he economy o converge o a pah wih he same properies as he neoclassical growh model..2.2 The Phillips curve and he NAIRU All he macro models considered here conain a wage-seing equaion based on an expecaions-augmened Phillips curve. The basic idea behind his goes back o Phillips (958) who found an inverse relaionship beween annual changes in nominal wages and he unemploymen rae in he U.K. over he period Subsequen sudies 5

6 found similar negaive relaionships beween nominal wage growh and unemploymen, and beween inflaion and unemploymen, for a large number of counries. These findings gave rise o he original Phillips curve heory he idea ha here is a negaive relaionship beween inflaion and unemploymen and ha here is a long-run rade off beween he wo. Aemps o explain or pu micro foundaions under he Phillips curve, in paricular by Friedman and Phelps, suggesed however ha rue relaionship is beween expeced real wage growh and he unemploymen rae. When he labour marke is igh (unemploymen is low) workers seek, and firms are willing o give, wage rises considerably above he expeced rae of inflaion; when here is high unemploymen, workers will accep a fall in he real wage. If we wrie he rae of growh of he nominal wage as he change in he log of wages in period +, ln(w + ), he expeced inflaion e rae in period + as π +, and he unemploymen rae in period as UR, hen here is a e relaionship of he form ( W ) = f ( UR ) ln π, where f < 0. There will be some + + unemploymen rae a which expeced real wage growh is equal o he long-run equilibrium rae of real wage growh, λ, given by he rae of echnological progress. If we call his unemploymen rae he naural rae and denoe i by UR na, hen he relaionship beween expeced real wage inflaion and he unemploymen rae has he ln, where < 0 e na form ( W ) = g( UR UR ) + π + λ g and ( 0 ) = 0 g. The heory of he expecaions-augmened Phillips curve has wo addiional elemens. Firs, wage rises end o feed ino price rises. Second, expecaions are backward looking so ha he higher recen inflaion has been, he higher fuure inflaion is expeced o be. Wih hese addiional assumpions, if unemploymen is kep below he naural rae in any period hen wage and price inflaion will no jus be relaively high, hey will be acceleraing. If he unemploymen rae is low in any period, nominal wage growh will be high relaive o he expeced inflaion rae and he pas inflaion raes on which expecaions are based. This fas increase in nominal wages will produce high inflaion and cause expecaions of inflaion o be revised upward. If he unemploymen rae says low in he nex period, he gap beween nominal wage growh and expeced This assumes ha wages for period + are se in period. 6

7 inflaion mus again be high, bu he expeced inflaion rae is higher han before so nominal wage growh mus be higher oo. This hen leads o higher inflaion, a furher upward revision of inflaion expecaions and even faser raes of wage and price inflaion in he following period. Tha is, permanenly low unemploymen leads o acceleraing inflaion. Similarly, permanenly high unemploymen an unemploymen rae above he naural rae will lead o a falling inflaion rae. The only unemploymen rae a which he inflaion rae can be consan (or no acceleraing) is he naural rae. Hence, his rae is ofen referred o as he non-acceleraing-inflaion rae of unemploymen or NAIRU. Under he heory of he expecaions-augmened Phillips curve, expeced inflaion is equal o acual inflaion along he equilibrium growh pah of he economy, he unemploymen rae is equal o he NAIRU, and real wages grow a he rae of echnological progress, λ. There are hree addiional poins o be made abou he expecaions-augmened Phillips curve and he way i is incorporaed ino he models reviewed here: Firs, in he descripion above, he rae of growh of expeced real wages depends only on he ighness of he labour marke, which in urn depends only on he unemploymen rae. The long-run equilibrium pahs of he economy in he macro models considered here are consisen wih hese assumpions. In he shor run, however, he models allow real wage growh o be affeced by facors oher han labour marke ighness (such as changes in wage-seing arrangemens) and do no measure labour marke ighness jus by he unemploymen rae. In paricular hey include changes in he unemploymen rae as well as he unemploymen rae iself. For a given unemploymen rae in period, he labour marke is igher and expeced real wage growh in period + is higher if he unemploymen rae had fallen beween period and period han if i had risen. Thus, he period- unemploymen rae a which period + expeced real wage growh is equal o he long-run equilibrium rae will always be beween he unemploymen rae in period and he NAIRU: i will be below he NAIRU if UR < NAIRU, i will be 7

8 above he NAIRU if UR = NAIRU. UR > NAIRU, and i will equal he NAIRU if Second, he value of he NAIRU in principle depends on he srucure of he economy; i migh change as he economy changes and be reaed as a funcion of variables relaed o he srucure of he economy. None of he models looked a here akes ha approach. They rea he NAIRU as consan and exogenous. Boh he AEM and Murphy models rea he NAIRU as consan and are esimaed over a period (from 976 on) for which ha is a plausible assumpion. The TRYM model allows he value of he NAIRU o differ beween he period before 974 and period from 974. I decomposes he change ino search-efficiency and wage seing componens, bu he wage-seing componen is basically a residual and he model effecively reas he value of he NAIRU as (consan and) exogenous in boh periods. Noe ha, while he long-run value of he NAIRU is exogenous in he models, shor-run facors can affec wage-seing for many periods and can give clues as o variables ha migh be included if he NAIRU were endogenised. Third, he value of he NAIRU migh depend on pas values of he unemploymen rae. In he exreme version of his, hyseresis, a period of high (or low) unemploymen can permanenly raise (or lower) he NAIRU. The pah of he unemploymen does no have permanen effecs on he NAIRU in any of he models looked a here, bu he inclusion of a change-in-unemploymen erm in he wageseing equaion could be inerpreed as implying a flexible NAIRU. Tha is, he long-run NAIRU is fixed, bu here is also a shor-run NAIRU ha in period is he value of he unemploymen rae UR for which he expeced growh in he real wage in period + is equal o he equilibrium growh rae, λ. This shor-run NAIRU is an increasing funcion of UR and equal o he long-run NAIRU only when UR = NAIRU..2.3 The Beveridge curve The AEM and Murphy models use acual employmen o measure labour demand and he size of he labour force o measure labour supply (he sum of employmen and unemploymen). The TRYM model, however, equaes labour demanded wih he sum of 8

9 employmen and unfilled vacancies and labour supplied wih he labour force. This helps wih idenificaion of he wo curves and is use in he model o decompose changes in he NAIRU ino hose associaed wih search efficiency and hose associaed wih wage seing. The inroducion of he exra variable, vacancies, requires an exra equaion and he TRYM model includes a negaive relaionship beween he rae of unemploymen and he vacancy rae (he raio of unfilled vacancies o he sum of employmen and vacancies). This relaionship is known as he Beveridge curve. The idea behind he Beveridge curve is ha when he labour marke i igh, unemploymen is low and firms find i difficul o find workers o fill vacan posiions. Thus he sock of vacancies is relaively high. On he oher hand, when he unemploymen rae is high, vacancies will fill quickly and he sock of unfilled vacancies a any ime will be low. The posiion of he Beveridge curve in vacancy rae/unemploymen rae space ha is, he level of he vacancy rae for any given unemploymen rae depends on how effecively he unemployed can be mached wih vacan jobs. This in urn depends on: How efficien he labour marke is a maching suiable workers wih suiable jobs maching unemployed facory workers wih vacan facory jobs and unemployed IT workers wih vacan IT posiions. The degree of mismach beween workers and vacancies maching will be slow if mos of he unemployed are facory workers bu mos vacancies are for IT workers. The level of general work skills of he unemployed for example, here may be a group of long-erm unemployed ha difficul o mach wih any vacancies. In he TRYM model here is an ouward shif in he Beveridge curve in 973 and 974, bu he relaionship is reaed as sable in he periods before and afer hose years. There are more deails in secion on he TRYM model below..2.4 Error-correcion models A common way of inegraing he shor-run and long-run relaionships beween he variables in models is hrough an error-correcion framework. For example, he equilibrium pah of a variable, y, migh depend on a vecor of variables, X, hrough a eq relaionship of he form, y = g X ). In he shor run, however, changes in he value ( 9

10 of y migh also depend on changes in variables in he vecor, Z (which migh include variables also in X and lagged values of y and of variables in X), hrough a funcion of he form, f ( Z ). An error-correcion model for his variable would have he form y = f ( Z ) a[ y g( X )], where he coefficien a is posiive. The firs erm on he righ-hand side represens he purely shor-run facors affecing y. The second erm, he error-correcion erm, represens a force pulling y oward is long-run equilibrium pah. The expression in brackes is he difference beween y and is equilibrium value in period. If y is above is equilibrium value, hen he erm begin subraced is posiive. In he absence of shor-run effecs on y, his makes y negaive; ha is, y falls beween period and period. Similarly, if y is below is equilibrium value, he error-correcion erm causes y o increase. Usually he value of he coefficien a is beween 0 and, in which case he correcion o y in period is a fracion of he error in period and he error-correcion erm ends o cause y o converge monoonically o is equilibrium pah. If a equalled one, however, hen convergence would occur wihin one period and any difference beween y and is equilibrium value would be due o Z. If a were beween and 2 hen he error-correcion erm would produce dampened oscillaions in y abou is equilibrium pah. (A negaive value of a, or a value above 2 would cause y o diverge.).3 General srucure of he models The hree models he TRYM model, he AEM model and he Murphy model each has an aggregae labour marke, hough he Murphy model builds up labour demand from secoral demands. There is one unemploymen rae in all models, however, and is long-run value is deermined by an aggregae wage-seing equaion. There are four componens of he labour marke in he AEM model and he Murphy model; he TRYM model has hese four componens and wo addiional ones. The componens are: Labour demand. In he AEM model and he Murphy model labour, demand is measured in persons and is idenified wih employmen; in he TRYM model i is measured in hours and includes boh he hours worked by hose employed and he hours ha would have been worked had all vacancies been filled. In all models 0

11 labour demand is he sum of he basically exogenously deermined labour demand by general governmen and governmen enerprises on he one hand, and of he endogenously deermined labour demand by he privae business secor on he oher. Labour supply by households. In he AEM model and he Murphy model, labour supply is measured in persons and is idenified wih he size of he labour force; in he TRYM model i is measured in hours and includes boh he hours worked by hose employed and he hours ha he unemployed would like o have worked In he TRYM model only, average hours worked by employees. This links labour demand and labour supply measured in hours o labour demand and labour supply in numbers of persons. Average hours are no required in he AEM model or he Murphy model, hough allowance is made for he changing proporion of par-ime work. Again in he TRYM model only, he relaionship beween unemploymen and vacancies or Beveridge curve. This provides he link beween labour demand and labour supply in persons on he one hand, and employmen on he oher hand. In all models, he relaionship beween expeced real wage inflaion and he unemploymen rae he Phillips curve. A low unemploymen rae ends o produce a large increase in he real wage. This ends o reduce labour demand and pushes he unemploymen rae up. Similarly, a high unemploymen rae ends o produce a small increase or a fall in he real wage, which ends o push he unemploymen rae down. This process gives a mechanism hrough which he unemploymen rae is pushed oward an equilibrium level. Price seing by privae secor firms. This ensures equilibrium in he goods marke o mach wha is happening in he labour marke. Of course, he models are all general equilibrium models so ha, in principle, evens in he labour marke canno be looked a in isolaion from hose in he oher pars of he economy. Wha happens in he labour marke affecs household consumpion decisions, firms price-seing and invesmen decisions, he exchange rae, ec. And changes o hese variables feedback o he labour marke, paricularly hrough he effec of oupu on demand for labour. In he shor-run, his ineracion beween he labour marke and he res of he economy is imporan for unemploymen and employmen levels. In he long run, however, employmen and unemploymen are deermined enirely from he

12 labour marke relaionships. The mos imporan is ha he long-run unemploymen rae is equal o an exogenously deermined NAIRU. This is common o all he models. The employmen relaionship differs across he models, however, and will be examined in he specific discussion for each model. 2. The Treasury Macroeconomic (TRYM) Model This secion examines he labour marke in he Commonwealh Treasury s TRYM model. Subsecion 2. gives some brief background on he TRYM model. Subsecion 2.2 oulines he way he labour marke is reaed in he TRYM model, parallelling and expanding on he more general bu briefer discussion in subsecion.3 above. Subsecion 2.3 examines he labour marke in he long run in more deail, and subsecion 2.4 discusses he long-run and shor-run versions of each of he main labourmarke equaions in he TRYM model. 2. Background o model The Treasury macroeconomic (TRYM) model is a highly aggregaed and relaively small quarerly model of he Ausralian economy. I has 29 esimaed equaions, 3 financial ideniies, 20 behavioural ideniies and 60 accouning ideniies. I follows on from he series of larger Naional Income Forecasing (NIF) models ha were developed by he Commonwealh Treasury in he 970s and 980s, he las of which was NIF88. Work on he TRYM model began in 990. I is used in Treasury for forecasing aggregae variables. The main documenaion for he TRYM model is: The Macroeconomics of he TRYM Model of he Ausralian Economy (Commonwealh Treasury, 996), which gives an overview of he model and oulines he general macroeconomic framework underlying he model; Documenaion of he Treasury Macroeconomic (TRYM) Model of he Ausralian Economy (Commonwealh Treasury, 996,999), 2 which gives deails of he equaions in he model; and The User s Guide How o Use he Treasury Macroeconomic (TRYM) Model of he Ausralian Economy wih TSP Sofware 2 In he version of his paper currenly on he Treasury web sie, he repored esimaion resuls for he labour marke use daa up o 999(2); he repored resuls for oher equaions use daa up o 995(3), in line wih he original version of he paper. 2

13 (Commonwealh Treasury, 996), which explains he use of he commercially available sofware for running he TRYM model. In addiion here are some 20 TRYM relaed papers dealing wih various aspecs of he TRYM model. Many of hese papers have been presened a conferences on TRYM or aspecs of he Ausralian economy. The ones of mos relevance o he labour marke are Johnson and Downes (994), Sacey and Downes (995), Downes and Sacey (996), and Downes and Bernie (999). All of his documenaion is available from he Treasury web sie hp:// In addiion, all daa series used in he esimaion he TRYM model, including series consruced especially for he model, are available a he ABS web sie (hp:// The discussion here is drawn mainly from Documenaion of he Treasury Macroeconomic (TRYM) Model of he Ausralian Economy, which is referred o as Documenaion, and Downes and Bernie (999). 2.2 Ouline of he labour marke in he TRYM model The labour marke in he TRYM model has he six componens referred o in he general inroducion (subsecion.3). A more deailed, bu sill brief, descripion of he way he six componens are reaed in he TRYM model is as follows: Labour demand. Labour demand in he TRYM model is measured in hours and includes boh employmen (filled jobs) and unfilled vacancies. Labour is demanded by hree secors in he economy he privae business secor, he governmen enerprise secor and he general governmen secor. Employmen in he general governmen secor was around 5 per cen of civilian employmen a he end of he 990s. I is reaed as being exogenous in he TRYM model. Employmen in he governmen enerprise secor as a proporion of civilian employmen has declined seadily over ime from around nine per cen in he middle of he 960s o around six per cen a he end of he 980s o wo o hree per cen of a he end of 990s. In he TRYM model, labour demand by he governmen enerprise secor (in hours) depends on invesmen by (and he capial sock of) he secor, which is reaed as exogenous. I does no depend on he real wage, even in he shor run. Employmen in he privae business secor accouned for over 80 per cen of civilian employmen a he end of he 990s (is lowes fracion over he pas 35 years occurred in he second half of he 970s when i was abou 75 per cen). In he long run, labour 3

14 demand by he privae business secor depends on oupu and he real producer wage hough as explained below, i depends ulimaely in equilibrium on labour supply. Labour supply. Labour supply, like labour demand, is measured in hours. I includes boh employmen and unemploymen. In he long run, labour supply in hours depends on hree facors: populaion, he employmen rae and an exogenous rend. Firs, labour supply is measured relaive o he working-age populaion (persons aged beween 5 and 64). This means ha labour supply grows wih populaion and allows for he effecs of he aging of he adul populaion due o falling moraliy raes. Second, he relaionship o employmen represens an encouraged/discouraged worker effec. I is allowed o vary wih he proporion of males o females in he labour force. For a given populaion size and level of average hours, an increase in employmen by en persons is esimaed o cause four addiional persons o ener he labour force, and hence o reduce he number of persons unemployed by six. Third, he exogenous rend is decomposed ino wo componens: one ha reflecs changes in he age composiion of he populaion (due o he baby boom) and he changes his produces because of he differences in paricipaion raes across age cohors; and a second componen ha picks up all oher rends in paricipaion. In he shor run, labour supply also reacs o flucuaions in labour demand, hough more srongly o changes in privae secor labour demand han o public secor labour demand. Average hours. In he long run, average hours follow an exogenous rend. This no doub reflecs facors such as he increase in par-ime work and in female relaive o male employmen, bu hese are no modelled explicily. In he shor run, average hours worked also respond o labour demand. Beveridge curve. The model posis a long-run negaive linear relaionship beween he log of he unemploymen rae and he log of he vacancy rae. This relaionship is allowed o shif ou over ime each vacancy rae is associaed wih a higher unemploymen rae. This shif is inerpreed as a reducion in search effeciveness. The shif is modelled by a logisical growh funcion wih he size, iming and speed of he shif being esimaed. The shif is esimaed o have occurred almos enirely beween 973(3) and 975(4) and o have added around half a percenage poin o he average unemploymen rae (NAIRU). This reflecs he change from he lae 4

15 960s and early 970s, when he vacancy rae was around per cen and he unemploymen rae was rarely much above wo per cen, o he period since he lae 970s, when he vacancy rae has mosly been below one per cen and he unemploymen rae has mosly been beween five and per cen. Phillips curve. In he TRYM model, he Phillips curve relaes changes in he expeced real consumpion wage o he unemploymen rae, changes in he unemploymen rae and changes in he degree of cenralisaion of wage deerminaion. Expecaions are backward looking and he expeced consumer-price inflaion rae used in calculaing expeced real wage inflaion is he average inflaion rae over he previous four quarers. This means ha changes in expeced inflaion lag behind changes in acual inflaion. For example, when he inflaion rae is increasing, expeced inflaion is less han acual inflaion. Changes in he nominal wage are based on expeced inflaion, so he increases in he nominal wage and he acual real wage are relaively small. This causes an increase in labour demand and pus downward pressure on he unemploymen rae. Tha is, in he shor run before inflaion expecaions cach up wih acual inflaion, increases in he inflaion rae end o lead o low unemploymen raes. There is an unemploymen rae, he NAIRU, a which here is no upward or downward pressure on wage inflaion (given no change in oher facors). The level of he NAIRU is modelled as he sum of wo componens a search-effeciveness componen and a wage-seing componen. The search-effeciveness componen comes from he Beveridge curve and he level of he unemploymen rae a which he unemploymen rae and he vacancy rae are equal. I hus increased as he Beveridge curve shifed ou from he end of 973 o he end of 974, from around 2¼ per cen o around 2¾ per cen. The wage-seing componen is modelled as a consan ha jumped in value beween 973(4) and 974(). The reason for his increase is no modelled. The combined effec of he increases in he wo componens was for he esimaed NAIRU o increase from around four per cen in he early 970s o abou 6½ per cen from he lae 970s onward. The NAIRU is effecively reaed as exogenous in he TRYM model excep during he brief period when Beveridge curve was shifing ouward. The decomposiion of he NAIRU ino search-effeciveness and wage-seing componens is endogenous, bu he wage-seing componen is, in effec, a residual. 5

16 In he shor-run, he growh rae of he expeced real wage also depends on he change in he unemploymen rae. For a given unemploymen rae, a decrease in he unemploymen rae pus addiional upward pressure on wage growh, and an increase in he unemploymen rae ends o reduce wage growh. The size of hese effecs is no symmeric he effec on wage growh of a fall in he unemploymen rae is esimaed o be almos wice as large as he effec of an increase in he unemploymen rae of he same magniude and he srengh of he effecs is made o depend on he fracion of he work force ha is unionised he effecs are sronger if a larger fracion of workers are union members. Changes in he degree of cenralisaion of wage deerminaion also affec wage inflaion. If in any quarer he degree of wage cenralisaion is greaer han i was in he corresponding quarer of he previous year, he rae of wage increase in he quarer will be differen from wha i would oherwise have been an increase in cenralisaion causes a emporary change in wage growh. 3 In addiion o he cenralisaion variable, here is a dummy for he meal rades wage decision in 974(3). 2.3 The labour marke in he long run In he long run virually all he labour-marke quaniy variables he unemploymen rae, he levels of employmen and unemploymen, average hours and vacancies are exogenous in he TRYM model. Average hours are explicily reaed as exogenous. The long-run equilibrium unemploymen rae is equal o he NAIRU, which is he sum of a consan (he wage-seing parameer) and he unemploymen rae a which he unemploymen rae and he vacancy rae are equal according o he Beveridge curve. Since he Beveridge curve is exogenous, boh componens of he NAIRU are 3 The direcion is a no clear from he TRYM model documenaion. In he wage-seing equaion in Downes and Bernie (999) and he main ex of Documenaion, he change in cenralisaion erm is a 3 (QCC, QCC 4 ) and he esimaed value of a 3,is posiive. This would imply ha an increase in cenralisaion emporarily reduces wage inflaion. In Appendix C o Documenaion, however, here is no negaive sign before he coefficien, a 3. If he value of a 3 is correcly repored in he ex, his would ha an increase in cenralisaion emporarily increases wage inflaion. The commens on he esimaion resuls in Downes and Bernie and Documenaion do no help clarify he issue hey merely noe ha [w]ages respond o changes in he insiuional environmen in he wage deerminaion sysem. 6

17 exogenous, so he NAIRU iself and he long-run unemploymen rae are exogenous. The esimae of he curren NAIRU in he TRYM model is 6.45 per cen. I requires only a small amoun of algebraic manipulaion o see ha he levels of employmen, unemploymen and vacancies are also exogenous. Since he unemploymen rae is he raio of he number of unemployed o he labour force, and he labour force is he sum of he number of persons employed and he number of persons unemployed, we can wrie he condiion ha he long-run unemploymen rae equals he NAIRU as E U + U = NAIRU, where E is he long-run equilibrium number of persons employed in quarer and U is he equilibrium number of persons unemployed. This implies ha he employmen rae (as a fracion of he labour force) is: E E + U = NAIRU. Taking logs gives: ln ( E + U ) = ln( E ) ln( NAIRU ) ln( E ) NAIRU. + The log of he labour force is approximaely equal o he sum of he log of unemploymen and he NAIRU. 4 The long-run labour supply equaion has he form: [ E + U ] ln 0, ln H LS LS E H = + a a, POP POP where H is desired average hours in period and POP is he working-age populaion in quarer. The numeraor in he expression on he lef-hand side of he equaion is labour supply in hours he produc of he number of persons waning o work (he sum 4 This resul uses he fac ha for small x, ln(+x) x, or ln( x) ) x. 7

18 of numbers employed and unemployed) and average hours. Thus he lef-hand side of he expression is he log of labour supply in hours relaive o he working populaion. Similarly, he second erm on he righ-hand side is he log of employmen in hours relaive o he working-age populaion. This erm reflecs he encouraged/discouraged worker effec. The esimaed value of he coefficien on he employmen erm is The oher coefficien, LS a 0,, reflecs paricipaion raes and demographic facors and can change over ime (hence he ime subscrip), bu is value is exogenous. If we expand he fracions in he labour-supply equaion and collec erms, we ge: LS LS LS LS ( E + U ) = a + a ln( E ) ( a ) ln( H ) + ( a ) ln( POP ). ln 0, Then subsiuing in he approximae expression for he log of labour force and solving for he log of employmen gives: LS a 0, NAIRU ln( E ) ln( H ) + ln( POP ). LS LS a a Since all he erms on he righ-hand side of his expression are exogenous in he TRYM model, he level of employmen is also exogenous. The expression ells us ha, for given populaion and average hours, a one percenage poin fall in he NAIRU is associaed wih a /( a LS ) per cen increase in employmen, or a one per cen increase in employmen is associaed wih a a LS percenage poin fall in he long-run unemploymen rae. Since he esimaed value of a LS is 0.400, his means ha a one percenage poin fall in he NAIRU is associaed wih a.7 per cen increase in employmen, or a one per cen increase in employmen is associaed wih a 0.6 percenage poin fall in he long-run unemploymen rae. Finally, he long-run Beveridge curve has he form: U ln ln, 0 BC BC V = + a a E + U E + U where V is he long-run level of vacancies and a BC 0 and a BC are parameers. The lefhand side of he expression is he log of he unemploymen rae and he final erm on he righ-hand side is he log of he vacancy rae. Since he long-run unemploymen rae is exogenously deermined, so is he vacancy rae. If we subsiue for he log of he unemploymen rae and he log of he labour force and rearrange, we ge: 8

19 ln BC a0 ( V ) + ln( NAIRU ) + ln( E ) NAIRU. + BC BC a a Subsiuing for he log of employmen gives: ln a LS BC LS 0, a0 a ( V ) ln( H ) + ln( POP ) + ln( NAIRU ) NAIRU. a LS a BC a BC a All he erms on he righ-hand side of he expression are exogenous, so he level of vacancies is also exogenous. 2.4 Deails of equaions in he TRYM model This secion provides furher deails of he main labour-marke equaions in he TRYM model. For each equaion, he long run and he shor run are examined separaely. The noaion used in he equaions in his secion differs somewha from he noaion in he TRYM documenaion. The relaionship beween he noaion here and ha in he TRYM model documenaion is given in Appendix A Unemploymen in he long run The wage-seing equaion, or Phillips curve, in he TRYM model makes he rae of growh of real wages depend, among oher hings, on he deviaion of he unemploymen rae from he NAIRU and on changes in he unemploymen rae. In he long run, he rae of growh of real wages is consan and equal o he rae of produciviy growh, and he unemploymen rae is equal o he NAIRU. The NAIRU, however, depends in par on he Beveridge curve relaionship. Specifically, he NAIRU is given by: NAIRU UR = UR ADJ ADJ + WSo + WS unil 973(4). from 974() The NAIRU in he TRYM model is no reaed as consan. Reasons for changes in he NAIRU are no modelled explicily bu hey are separaed ino wo ypes: LS 5 Appendices A,B and C o his working paper can be downloaded as a pdf file from he Melbourne Insiue websie 9

20 Those ha relae o changes in search efficiency as refleced in shifs in he Beveridge curve. These are refleced in he variable ADJ UR. Those ha relae o changes in wages seing. These are capured by he erms WSo and WS. The variable ADJ UR is he unemploymen rae adjused for search effeciveness ha is he unemploymen rae a which unemploymen and vacancies are equal in he Beveridge-curve relaionship. I also includes he residuals from he Beveridge curve equaion. If he residuals were no included hen, as explained below, ADJ UR would have been essenially consan and equal o 2.09 per cen unil 973(3), consan and equal o 2.64 per cen from 975(4), and increasing beween he wo levels beween 973(4) and 975(3). The iniial and final values of ADJ UR, he iming of he change and he rae of change are all esimaed. Noe ha he Beveridge curve was esimaed using daa for he period 967(3) o 999(2) so ha he 2.09 per cen value of can be aken as exending back ino he 960s. ADJ UR The second erm in he expression for he NAIRU is a consan ha akes he value WSo up o 973(4) and he value WS from 974(). The wage equaion is esimaed for he period 97() o 999(2). Had he sample period exended back ino he 960s, here would have been a leas one addiional change in he value of he consan. The iming of he break as 974() is imposed bu he values of WSo and WS are esimaed. The esimaes repored in Downes and Bernie (999) are WSo =. 83 and WS = Downes and Bernie give he corresponding values of he NAIRU, afer excluding he effecs of he residuals from he Beveridge curve, as 4.05 per cen in he early 970s and 6.45 per cen from he mid-970s. 6 (They noe ha he NAIRU was probably around wo per cen in he 960s.) 6 The esimaes for he NAIRU repored by Downes and Bernie (999) do no quie mach he esimaes hey repor for WSo and Ws and he values of UR ADJ, heir variable RNUST, implied by heir esimaed Beveridge curve and confirmed by he series for RNUST given in he ABS s TRYM model daabase. These imply values of he NAIRU of 3.92 per cen and 6.34 per cen, raher han 4.05 per cen and 6.45 per cen. 20

21 The TRYM model hus implies ha he NAIRU increased from around 2 per cen in he 960s o around 4 per cen in he early 970s, all he increase being due o wage-seing facors. The NAIRU increase by a furher 2½ per cen beween he early 970s and he lae 970s; around one-half of a percenage poin of his increase was due o a decrease in search efficiency and he remainder was due o wage-seing facors. Noe ha, apar from he fac ha he iming of he change in he wage-seing parameer (from WSo o WS) is imposed independenly of he Beveridge curve, he wage-seing componen of he change in he NAIRU is effecively a residual he par of he oal change no explained by he change in ADJ UR. The ime pah of he NAIRU in he TRYM model from 97() o 999(2) is shown in Figure. 7 The hinner line gives he value of he NAIRU including he effecs of he residuals from he Beveridge curve on ADJ UR ; he hicker, smooher line gives he NAIRU corresponding o zero values for he residuals. The NAIRU increases beween 973(4) and 975(3) wih a big jump beween 973(4) and 974() due mainly o he swich in he value of he wage-seing parameer from WSo o WS. Figure 2 shows he pah of he unemploymen rae for he period 959(3) o 999(2) and he TRYM model NAIRU, excluding he residuals from he Beveridge curve equaion, for he period 97() o 999(2). 8 Two feaures sand ou: The unemploymen rae is below he NAIRU hroughou he 970s and he increase in he NAIRU (mainly in 974) precedes he main increase in he unemploymen rae (which occurs in he second half of 975 and 976). In he 980s and 990s he NAIRU is considerably below he average unemploymen rae. Undersanding hese resuls requires knowledge of he shor run wage-seing equaion, so we ll examine his and hen reurn o possible explanaions for hese resuls. We ll 7 The NAIRU is calculaed as he sum of RNUST and he parameers, WSo and WS. The ime series for RNUST including residuals is aken from he ABS TRYM model daabase. The underlying series for RNUST is calculaed from he esimaed Beveridge curve. 8 The NAIRU is he same as in Figure. The unemploymen rae is aken from he ABS TRYM daabase. 2

22 also discuss some he reasons suggesed by Downes and Bernie (999) for he increase in he NAIRU. Figure. TRYM NAIRU 97() o 999(2) per cen Mar-7 Mar-73 Mar-75 Mar-77 Mar-79 Mar-8 Mar-83 Mar-85 Mar-87 Mar-89 Mar-9 Mar-93 Mar-95 Mar-97 Mar-99 Figure 2. Ausralian Unemploymen Rae and TRYM model NAIRU Unemploymen Rae NAIRU Sep-67 Sep-69 Sep-7 Sep-73 Sep-75 Sep-77 Sep-79 Sep-8 Sep-83 Sep-85 Sep-87 Sep-89 Sep-9 Sep-93 Sep-95 Sep-97 per cen The wage equaion in he shor run In he shor run in he TRYM model, he expeced rae of real-wage inflaion relaive o he rend rae of produciviy growh is modelled as depending on: The change in he degree of cenralised wage fixing. 22

23 Union membership as a proporion of oal employmen. The lagged change in he unemploymen rae. The lagged deviaion of he unemploymen rae from he NAIRU as a fracion of he unemploymen rae. In addiion here is a dummy for he meal rades wage decision in 974(3). The esimaed equaion has he form: π W π e C λ = a 2 + a Union 4 NAIRU UR On he lef-hand side of he equaion, + [ c UR + ( c ) UR ] UR π W + a 5 Q743. a 3 4 Cenral is he rae of nominal wage inflaion in quarer (he change in he log of he hourly wage rae beween quarer and quarer e ), π is he expeced rae of inflaion of consumpion good prices in period, and λ is C he rend rae of labour-augmening echnical progress. Thus he lef-hand side is he expeced rae of increase of he real wage rae in excess of produciviy growh (is rend rae of growh). Expeced consumer price inflaion is a weighed average of he change in he log of he deflaor of privae consumpion over he previous four quarers, wih he weighs declining geomerically. Tha is: 2 π C + aπ C + a 2 π C + a 3 π e C 4 π C =, a + a + a where π C is he acual rae of inflaion of consumpion good prices in period (he change in he log of he deflaor of privae consumpion beween quarer and quarer ). The esimaed value of a is so he weighs on he four lags of inflaion in he expression for expeced inflaion are 0.340, 0.27, 0.26 and On he righ-hand side of he wage equaion, Union is union membership as a proporion of oal employees in quarer, Cenral is a variable for he degree of cenralisaion of wage fixing, Q743 is a dummy variable ha equals one in 974(3) and zero in all oher quarers, and UR is he unemploymen rae in percenage poins. UR is he change in he unemploymen rae beween quarer and quarer. 3 UR is he posiive par of he change in he unemploymen rae i equals UR when UR is + posiive and equals zero when UR is negaive. Similarly, UR is he negaive par of 23

24 he change in he unemploymen rae i equals UR when UR is negaive and equals zero when UR is posiive. Thus, UR = UR + + UR. NAIRU is he value of he NAIRU in period, expressed in percenage poins, as explained in he previous subsecion. The firs erm on he righ-hand side of he wage equaion, + [ c UR + ( c ) UR ] a 2 Union, gives he effec of changes in he unemploymen rae on real-wage inflaion an increase in he unemploymen rae reduces he rae of wage growh. The srengh of his effec depends on he degree of unionisaion he greaer he proporion of workers who are union members, he greaer is he upward pressure on wages of a reducion in unemploymen. The degree of union membership rose from around 50 per cen a he sar of he 970s o around 55 per cen in he second half of he 970s hrough o he middle of he 980s. Since hen i has fallen seadily, down o 35 per cen a he end of he 990s. The effecs of posiive and negaive changes in he unemploymen rae are asymmeric. The esimaed value of c is 0.367, so c is and he upward effec on wage growh of a decrease in he unemploymen rae is 0.633/0.367 =.725 imes as grea as he downward effec on wage growh of an increase in he unemploymen rae. The esimaed value of he coefficien a 2 is The second erm on he righ-hand side of he wage equaion, 4 a Cenral, 3 represens he change in he degree of cenralisaion of wage seing in quarer compared o he corresponding quarer in he previous year. Cenral is a consruced variable. I ook he value 0.20 up o 975(), he 0.80 from 975(2) o 98(), 0.20 again from 98(2) o 982(4), and 0.80 from 983() o 985(3). Since hen i has been declining linearly, from 0.70 in 985(4) o 0.6 in 999(2). The esimaed value of he coefficien a 3 is The hird erm on he righ-hand side of he wage equaion, a 4 NAIRU UR UR he negaive of he deviaion of he lagged unemploymen rae from he NAIRU as a fracion of he unemploymen rae. Low unemploymen raes are associaed wih fas wage growh, and high unemploymen raes are associaed wih slow wage growh. The, is 24

25 esimaed value of he coefficien a 4 is 0.. If he NAIRU is 6.45 per cen, hen a consan unemploymen rae of 5.45 per cen in quarer (one percenage poin below he NAIRU) would add abou 0.20 per cen o he quarerly rae of wage inflaion in quarer, relaive o he scenario in which he unemploymen rae equalled he NAIRU; a consan unemploymen rae of 7.45 per cen in quarer (one percenage poin above he NAIRU) would subrac abou 0.5 per cen from he quarerly rae of wage inflaion in quarer. The final erm on he righ-hand side of he wage equaion, a Q743 5, is a dummy for he meal rades wage decision in he Sepember quarer of 974. The esimaed value of a 5 is Tha is, wages were 7.5 per cen higher in 974(3) han hey would have been wihou he wage decision. Figure 3 reproduced from Downes and Bernie (999), Figure 6 on page 32 shows he relaionship beween expeced real wage growh and he unemploymen rae from 970(3) o 996(). Also shown in he figure are he esimaed expecaions-augmened Phillips curves for he early 970s and for he period from he second half of he 970s esimaed using daa up o 996(). These earlier esimaes gave higher NAIRUs han he laes ones around 4½ per cen in he early 970s and around 7 per cen in he laer period. Downes and Bernie make he poin ha he Phillips curve for his laer period is exremely fla so ha he NAIRU canno be accuraely esimaed. Visual inspecion of he dae suggess ha he negaive slope of he curve is due almos enirely o he observaions for 974(2) and 974(3), 9 he wages boom and he wages freeze. Wihou hese observaions, expeced wage inflaion seems essenially unrelaed o he unemploymen rae. 9 The curve is esimaed for he period saring 974(2). The observaion for 974(3) corresponds o he meal rades wage decision and is dummied ou. 25

26 Figure 3: Esimaed Expecaions Augmened Phillip s Curves in TRYM Furher commens on he NAIRU in he TRYM model I was noed above ha in he 970s he NAIRU was above he unemploymen rae; he rend increase in he NAIRU preceded he main increase in he unemploymen rae in he mid-970s; and in he 980s and 990s he NAIRU was less han he average unemploymen rae. Le us consider each of hese in urn: The wage equaion predics ha expeced real wage growh will be high when he unemploymen rae is below he NAIRU. Tha is, he esimaed parameers in he wage equaion will end o be such ha any sub-period wih relaively high expeced real wage growh will end o be inerpreed as a boom a period wih an unemploymen ha is low relaive o he NAIRU. This means giving an esimae of he NAIRU ha is high relaive o observed unemploymen raes for he sub-period. This may provide a reason for he comparaively high NAIRU for he early 970s. Figure 6 in Downes and Bernie (999) (reproduced here as Figure 3) is consisen 26

27 wih his inerpreaion. There are hree queries ha migh be made in relaion o he observed rae of expeced real wage inflaion and is relaionship o he NAIRU: - Firs, he dependen variable is sricly he excess of real wage inflaion relaive o produciviy growh. In he wage equaion in he TRYM model he rae of produciviy growh is reaed as being consan over he sample period. There is considerable evidence however ha produciviy growh was higher in he early 970s ha in laer pars of he sample period (excep perhaps he second half of he 990s). Thus expeced real wage growh relaive o produciviy may be oversaed. - Second, expeced price inflaion is a weighed average of pas price inflaion. If agens are more forward-looking han he model assumes hen, in a period of rising price inflaion such as he early 970s, he measure of expeced price inflaion will undersae expeced price inflaion and measured expeced real wage inflaion will oversae acual expeced real wage inflaion. - Third, if here is a long-lasing increase in labour s share in oupu, as here was in he early sevenies, his will be associaed wih high realwage growh. I is no clear ha his would have he same effec on unemploymen as a faser han rend real-wage growh ha was expeced o be reversed in subsequen periods. On he iming of he increase in he NAIRU, he major par of he increase is due o he increase in he wage-seing parameer and he iming of his increase is imposed. I would be ineresing o know wheher he imposed quarer 974() would have been he one chosen if he iming of he break had been esimaed. On he low NAIRU in he 980s and 990s relaive o average unemploymen raes in hose decades, Downes and Bernie (999) explain his phenomenon in erms he asymmeric response of wages o unemploymen rae increases and decreases he upward effec on wage inflaion of a fall in he unemploymen rae is 70 per cen larger han he downward effec of an equal-size increase in he unemploymen rae. In addiion o his, he low NAIRU mus represen he flip side of some of he facors discussed above in relaion high NAIRU of he early 970s real wage growh has been lower han in he early 970s, he inflaion rae, he inflaion rae 27