Labout Market Dynamics in Australia What Drives Unemployment

8/2/2019 Labout Market Dynamics in Australia What Drives Unemployment

1/26

Labour Market Dynamics in Australia: What DrivesUnemployment?*

MA RI K A K A RA N A S S OU

Department of Economics, Queen Mary,University of London, London, UK and Institute for the Study of Labor (IZA), Bonn, Germany

H E CT O R S A L A

Department of Applied Economics,Universitat Autonoma de Barcelona,

Bellaterra, Spain and Institute for theStudy of Labor (IZA), Bonn, Germany

The debate in Australia on the (constant-output) elasticity oflabour demand with respect to wages has wrongly sidelined the

role of capital stock as a determinant of employment (Webster,2003). As far back as 1991, Pissarides had argued that the influ-ence of capital stock on the performance of the labour market iscrucial but not well understood, a research area which is particu-larly relevant for Australia. This article attempts to fill this voidby estimating a multi-equation labour market model comprisinglabour demand, wage setting and labour supply equations. Themodel is used to examine the causes of the unemployment upturnin 19731983 and the subsequent decline in 19932006. Ourresults show that (i) the main determinants of the unemploymentrise in the 1970s and early 1980s were wage-push factors, the twooil price shocks and the increase in interest rates, and (ii) theacceleration in capital accumulation was the crucial driving force

of unemployment in the 1990s and 2000s. Furthermore, althoughthe recent boom in the terms of trade is equally important, itsdownward effect on unemployment was partially reversed by theresulting decrease in net foreign demand.

I IntroductionThe relationship between aggregate employ-

ment and real wages has attracted a lot of atten-tion in Australia. It is common in most studiesto estimate labour demand equations with out-put, rather than capital stock, as an explanatory

variable. However, Webster (2003, p. 141)argues that ...there is no theoretical reason whythe estimating function should include output asan explanatory variable. ...Little discussion isgiven in the literature of this issue and its prac-tice is questionable.

More than 15 years ago, Pissarides (1991)e st imat ed a t hr ee -e qu at io n mod el f or t he

* The bulk of this work was completed while Hec-tor Sala was visiting the School of Economics at theUniversity of New South Wales (Sydney, Australia).Their warm hospitality, as well as the insightful com-ments and suggestions received from Glenn Otto, aregratefully acknowledged. The authors are also gratefulto Jeffrey Sheen, two anonymous referees and David

Norman for insightful comments on an earlier versionof this article. Hector Sala acknowledges financialsupport from the Spanish Ministry of Science andTechnology through grant ECO2009-07636 and themobility programme Jose Castillejo.

JEL classifications: E22, E24, J21Correspondence: Hector Sala, Departament dEcon-

omia Aplicada, Edifici B Campus UAB, 08193Bellaterra, Spain. Email: [email protected]

THE ECONOMIC RECORD, VOL. 86, NO. 273, JUNE, 2010, 185209

185

2009 The Economic Society of Australiadoi: 10.1111/j.1475-4932.2009.00597.x


2/26

Australian labour market, where capital stockfeatured as a determinant of employment andthe wage rate. In his global appraisal of Aus-tralias unemployment experience, Pissarides(1991, p. 36) pointed out that

the links between capital, investment andemployment are crucial but they are not verywell understood. I make an attempt to discussthem and to bring out the role of investmentin the ups and downs of employment in the1980s. But much more work is needed in thisarea. In Australia even more than in othercountries, this area of research seems to holdmuch promise of important results.

In the light of the arguments by Webster(2003) on the estimation of employment equa-tions, and the largely disregarded points raisedby Pissarides (1991), the objective of this article

is twofold. First, we provide an updated estima-tion of labour demand using capital stock, ratherthan output, as a determinant of the Australianemployment. It should be noted that, as inLewis and MacDonald (2002), our estimationfollows the autoregressive distributed lag(ARDL) approach to co-integration analysis (butusing capital stock instead of output as a deter-minant of employment). Second, we assess thedriving forces of unemployment over the 19722006 period in the context of a labour marketmodel comprising, as in Pissarides (1991),employment, wage setting and labour force

equations.There are two stark differences between the

work of Pissarides and ours. First, in the Pissa-rides model capital stock does not influenceemployment in the steady state, as its directeffect on employment is exactly offset by itsindirect effect via wages (capital stock raisesthe wage rate which, in turn, decreases employ-ment). This is in the spirit of Layard et al.(1991) and the non-accelerating inflation rate ofunemployment (NAIRU) framework of analysis,where there is no room for growing variables toinfluence the unemployment rate. Second,

investment (i.e. the growth rate of capitalstock), through its positive effect on wages,ends up having a negative effect on employmentand a positive one on unemployment. Accordingto Pissarides (1991 pp. 4344),

this effect is probably due to new technologyembodied in the capital stock, which displaceslabour faster when it takes place at a higher

rate, so that it leads to a higher equilibriumunemployment rate. It may also be due to thepeculiarly narrow-based investment booms inAustralia and the wage spillovers that theycaused elsewhere in the economy.

In contrast to the aforementioned work, thisarticle follows the chain reaction theory (CRT)of unemployment, presented in Section III, andshows that capital stock affects employmentpositively in the long-run, whereas capital accu-mulation is inversely related with the ups anddowns of the unemployment rate. The contribu-tion of capital stock to the evolution of unem-p lo ymen t i s a mo ng t he s ev er al i mp or ta ntconsiderations which distinguish the CRT fromthe NAIRU framework of analysis.1 In fact,the significant influence of capital stock onunemployment has become one of the salient

empirical facts unveiled by the CRT in studiesfor the United Kingdom (Henry et al., 2000;Karanassou & Snower, 2004), the EU (Karanas-sou et al., 2003), the Nordic countries (Kara-nassou et al., 2 00 8a ), a nd S pa in ( Ban de &Karanassou, 2009; Karanassou & Sala 2009).2

This is in sharp contrast with the influentialnatural rate of unemployment (NRU) literature,which, on the basis of the observation that theunemployment rate is trendless, asserts thateither (i) policies that shift upward the time pathof capital stock have no long-run effect on theunemployment rate, or (ii) the unemploymentrate can be influenced by trendless transforma-tions of the capital stock (e.g. the unemploymentrate may depend on the capital labour ratio).However, as Karanassou and Snower (2004)argued, there is no reason to believe that thelabour market alone is responsible for ensuringthat the unemployment rate is trendless in thelong-run. In general, equilibrating mechanismsin the labour market and other markets are

1 See Karanassou et al. (2008a, 2009) for a compre-hensive exposition of the differences between theCRT and NAIRU.

2

Apart from the CRT approach, the impact of capi-tal stock on unemployment is becoming the focalpoint of analysis in other studies as well. Rowthorn(1999) and Kapadia (2005) examine production func-tions less restrictive than the standard CobbDouglasone, whereas Malley and Moutos (2001) augment theNAIRU framework with strategic trade policy models.Arestis et al. (2007) investigate the capitalunemploy-ment relation in the context of Robert Rowthorns1977 aspirations gap model.

186 ECONOMIC RECORD JUNE

2009 The Economic Society of Australia


3/26

jointly responsible for this phenomenon. Thus,restrictions on the relationships between thelong-run growth rates (as opposed to the levels)of capital stock and other growing exogenousvariables are sufficient for this purpose. As anexample, in the context of the illustrative labour

market model below (Eqns 5, 6), consider thefrictional growth stability restriction (8) and theanalysis in Section III(ii).

Whilst the exogeneity of capital stock in ourlabour market model can be deemed as a short-coming of our methodology, our view is that,unless one pins down an unquestionably univer-sally true model for economic reality, ourapproach is at par with other macrolabour theo-ries. Naturally, we can augment our CRT modelto include a capital accumulation equation (see,e .g . K ar an as so u & S al a, 2 00 8) b ut t hi s i sbeyond our scope in this study.

Here we aim at evaluating how capital stockand other supply- and demand-side factors haveshaped the unemployment rate trajectory. Therole of capital accumulation in the Australianeconomy is of special interest, as the unemploy-ment effect of capital stock has been largelysidelined in the relevant literature. Our workimplicitly responds to Gregory (2000, p. 124),who argues that

Perhaps we cannot proceed far with simpleanalysis as it has been applied here except toshow that common views and prescriptions donot accord well with the data. This suggeststhat more complex analysis capable of takinginto account many influences at the same timeis needed.

Our empirical three-equation model features avariety of feedback mechanisms and includes awide set of explanatory variables. The estimatedsystem tracks fairly well the actual evolution ofAustralian unemployment, and is thus used toassess the relative influence of the explanatoryvariables in shaping its trajectory during twodistinct eras: 19731983, when rising unemploy-ment attained a historical maximum; and 1993

2006, when falling unemployment reachedalmost full-employment levels.

Our results for the 1970s and early 1980s aresupportive of the mainstream view that shocks(oil prices and interest rates) and institutions(direct taxes, benefits and income policies) areresponsible for the increase in the unemploy-ment rate. However, this is not the whole story,as an external factor (e.g. terms of trade) and a

Keynesian policy in the late 1970s were alsoidentified among the driving forces of unem-ployment. In particular, unemployment wouldhave been even higher from 1976 to 1980 hadthe government not increased its expenditures.As we show in Section III, one advantage of the

CRT ap pr oa ch i s t ha t i t e na bl es u s t o t es twhether trended variables (such as capital stockor working-age population) affect unemploy-m ent. W e find that, although the overallcontribution of the growing variables over the19731983 period was minimal, the slow-downin capital accumulation in the mid-1970s hadprolonged and significant effects on unemploy-ment until the end of the decade.

One of our main findings is the crucial effectof capital accumulation on the decline in unem-ployment over the 19932006 period. We showthat it accounts for 80 per cent of the overall

fall in unemployment. Had capital stock growthstayed constant at its 1993 value, unemploymentwould have ended the period by approximately5 percentage points (pp) higher than it actuallydid. The external factors remain important,especially in the 2000s, but the continuing dete-rioration of net foreign demand partially negatesthe contribution of the terms of trade to thedeclining unemployment (which otherwisewould be as large as the contribution of capitalstock growth). In sharp contrast with the 19731983 period, institutional factors play no role inthe performance of the Australian labour market

during 19932006.3

The rest of the article is structured as follows.Section II outlines some crucial aspects of theliterature on the labour demand in Australia.Section III presents the CRT approach. SectionIV estimates a CRT multi-equation labour mar-ket model for Australia from 1972 to 2006. Sec-tion V discusses the economic and politicaldevelopments during our sample period. In thecontext of our empirical CRT model, Section VIevaluates the contributions of the exogenousvariables (capital stock among them) to theunemployment rate trajectory. Section VII

concludes.

3 Although our main interest lies in the 19731983and 19932006 periods, to check the robustness ofour findings we also examine the 19831989 and19891993 subsamples (the results, which are avail-able upon request, are summarised in Section VI).

2010 LABOUR MARKET DYNAMICS IN AUSTRALIA 187



4/26

II Some Key Features of the Literature onAustralia

The impact of wage changes on (un)employ-ment is a crucial issue, especially as mostlabour market policies influence wage setting. InAustralia, a tradition of centralised wage bar-

gaining and a long experience of income poli-cies implemented through the Accord (or Pricesand Incomes Accord) have given rise to aprolific literature on this issue.

Webster (2003), surveying the Australian andoverseas literature, documents a constant-outputelasticity of labour demand with respect towages in the range of)0.15 to )0.80. The con-stant-output adjective signifies that it is a par-tial elasticity, as it is estimated by holding theendogenous output constant, a practice criticisedby Webster (2003) as careless because it failsto capture the full effect of real wages on labour

demand. Instead, Webster (2003) argues that thetotal elasticity should be estimated by takinginto account the response of output to variationsin labour as a crucial production factor. Follow-ing the analysis of the archetypal aggregatesupply-side model, where the total cost of pro-duction is minimised subject to the productionfunction, the implication for the profit-maximis-ing firm is that its labour demand equationincludes capital stock as a determinant, ratherthan output. Thus, when output enters the esti-mated labour demand equation there is theproblem of potential endogeneity, which is a

serious concern for the results of most of thisliterature.4

Webster (2003), like Pissarides (1991), showsthat labour demand should depend on wages andreal interest rates (as the relevant prices of theproduction factors) and capital stock (as one ofthe main production factors). Of course, othervariables, such as technological change anddemand-side variables, can be considered asdeterminants of labour demand. Most of thestudies for Australia consider a standard Hicks-neutral, rather than Harrod-neutral, characterisa-tion of technical progress by including a linear

time trend in the employment equation (for alist, see Lewis & MacDonald, 2004). Although,as pointed out by Webster (2003), most of the

literature has tended to ignore any demand-sideinfluences, Pissarides (1991) argues for the pos-sibility of aggregate demand effects (see alsoLindbeck & Snower, 1994) and gives evidencefor demand-side influences in his estimatedemployment equation.

In terms of the estimation procedure, Lewisand MacDonald (2002) apply the same econo-metric technique used in this article (and gener-ally in the papers which estimate CRT models):5

the ARDL approach (also known as bounds test-ing approach) to co-integration analysis. TheARDL was developed by Pesaran (1997), Pesa-ran and Shin (1999) and Pesaran et al.(2001) asan alternative procedure to the standard co-inte-gration/error-correction analysis. One of themain advantages of the ARDL is that it avoidsthe pre-testing issues of identifying the I(1) or

I(0) properties of the variables, a problem that

accompanies the popular co-integration method-ologies (like the Johansen maximum likelihood).The voluminous literature on all the differenttypes of unit root tests developed since theinfluential paper by Dickey and Fuller (1981) isa clear manifestation of the problems involvedin testing whether a time series is integrated oforder zero or one. It can be shown that theARDL provides a robust econometric tool forestimating and testing the short- and long-runrelationships between the variables without hav-ing to classify them as I(1) or I(0). In addition,as Lewis and MacDonald (2002) explain, the

ARDL is capable of dealing with the endogene-ity issues that might arise when output is usedin the labour demand equation. We can furtherargue for the robustness of the empirical resultsby checking whether the long-run solution ofthe estimated equation via the ARDL methodol-ogy is in line with the co-integrating vectorobtained by the Johansen procedure (see Lewis& MacDonald, 2002; Karanassou & Snower,2004; Karanassou et al., 2008a).

Lewis and MacDonald (2002) use quarterlydata for Australia over the 19611998 periodand estimate dynamic versions of the labour

demand equation nt a0 + a1wt + a2yt + a3t + et(equation (6) in p. 20), where nt, wt and yt repre-sent the (log of) employment, real wage and realgross domestic product (GDP), respectively, t isa time trend, et denotes the error term and the as

4 Even the recent study on an employment equationfor Australia (Dixon et al., 2005) is in line with theconventional use of output as an exogenous determi-nant of labour demand.

5 See, for example, Karanassou and Snower (1998)and Karanassou et al. (2003, 2008a).




5/26

are constants. They find weak exogeneity for theright-hand side variables and estimate the fol-lowing long-run relationship (co-integratingvector):6

n w y t 1 0:46 1:05 0:003 :

We replicated the results of Lewis and Mac-Donald (2002) by using our annual dataset from1970 to 2006 (see Section IV(i) for data descrip-t io n) . I n p ar ti cu la r, f ol lo wi ng t he A RD Lapproach, our selected dynamic specification oftheir labour demand equation features twoannual lags (versus their seven quarterly lags)and has the following long-run solution:

n w y t 1 0:50 0:97 0:009 :

We confirmed the robustness of this long-runrelationship by finding that it is not statisticallydifferent from the co-integrating vector obtainedby the Johansen procedure (see Section IV(i) forthe details of the likelihood ratio or LR testused).

Despite using the same econometric techniqueas Lewis and MacDonald (2002) for the estima-tion of the labour demand equation, our empiri-cal model in Section IV differs from theirs (andfrom most of the existing literature) in two mainrespects. First, we take into account both Web-sters (2003) concerns about the use of output inthe labour demand equation, and Pissarides(1991) intuition that capital stock may yield

valuable insights in explaining labour marketo ut co me s i n A us tr al ia . S ec on d, w e u se adynamic multi-equation system with spillovereffects, that is, our model examines labourdemand in conjunction with wage setting andlabour supply, and, thus, takes into account allinteractions among them. A distinguishing char-acteristic of our empirical analysis is that itdoes not focus on the computation of the NRUbut, instead, provides an account of how thedriving forces of employment, wages and the

labour force contribute to the evolution of theunemployment rate. We should point out that,because our model derives from the CRT ofunemployment, it has radically different impli-cations for the impact of capital accumulationon employment from the Pissarides (1991)

model.Beyond the studies concerned with the aggre-

gate labour demand, the literature on the unem-pl oyment rat e is closely lin ked with theestimation of the equilibrium rate of unemploy-ment, commonly referred to as the NRU or theNAIRU. Several papers follow the work byLilien (1982) who estimated a microfoundedreduced-form unemployment equation for theUnited States. This group includes the study byGroenewold and Hagger (1998) who estimatedthe NRU for 19791993, the subsequent debateon the validity of such estimates (Groenewold &

Hagger, 1999; McDonald, 1999), and the morerecent work by Heaton and Oslington (2002).Other papers base their analysis on the esti-mation of the frictional unemployment rate(Bodman, 1999) or the Beveridge curve (e.g.Groenewold, 2003; Kennedy et al., 2008). How-ever, the main strand of this literature estimatesthe NAIRU within a Phillips Curve context (seeGruen et al., 1999, for a general appraisal). TheNAIRU is estimated by (i) applying the Kal-man filter technique (Debelle & Vickery, 1998),(ii) using Vector Autoregression (VAR) models(Crosby & Olekalns, 1998) and Structural VAR

(SVAR) models (Groenewold & Hagger, 2000)and (iii) by embedding the Phillips Curve into abig macroeconomic model such as the Treasurymacroeconomic (TRYM) model (Song & Freeb-airn, 2005). Finally, Lye et al. (2001), and Lyeand McDonald (2006) use the range model(which allows for a piece-wise linear short-runPhillips Curve), to evaluate the minimum equi-librium rate of unemployment, and contrast theirmodel to NRU specifications.

No matter how the equilibrium rate of unem-pl oyment is perceived, i t is now widelyacknowledged that it is hard to agree on its

value at any point in time.7

Hagger and Groene-wold (2003, p. 324) claim that it is time toditch the natural rate, as it has become a sourceo f g re at a nd g ro wi ng c on fu si on . Ly e a nd

6

Because this is obtained from a marginal produc-tivity condition, Lewis and MacDonald (2002) usethese values, together with a labour share of 0.6, tocompute (i) the implied constant-output elasticity,which they place below )0.20, and (ii) the correctelasticity of demand for labour with respect to realwages, which they find equal to )0.80. This interpre-tation prompted a hot debate (see, among others,Dowrick & Wells, 2004; Lewis & MacDonald, 2004),which is beyond the scope of our analysis.

7 See, for example, The Economist, 30 September2006, p. 108.




6/26

McDonald (2006, p. 227) conclude that basingmacroeconomic policy on the natural rate modelwould underpin the possibilities for economicwelfare in Australia. McDonald (2008, p. 283)remarks that the usefulness of the TV NAIRUis bedeviled by the lack of explicit economic

mechanisms to determine its movements. It letsthe data talk, but, uninstructed, the data can telllittle.

In the next section, we present the workingsof the CRT within a stylised framework of anal-ysis and explain how it differs from the conven-tional wisdom. We show that, under plausibleconditions, unemployment does not gravitatetowards its natural rate and, therefore, the cen-tral role of the NRU in policy-making is beingchallenged.8

III The CRT

The CRT is an interactive dynamics approachaimed at explaining the evolution of the unem-ployment rate within a framework of a dynamicmulti-equation model, which allows for spill-overs effects and growing variables.

In essence, the CRT postulates that the move-ments in unemployment are driven by the inter-play of shocks and lagged adjustment processes.Generally, the CRT refers to lags of the endo-genous variables as the lagged adjustment

processes. Shocks refer to changes in the exoge-nous variables, whereas lagged adjustmentprocesses refer to, among others, employment

adjustments, wage/price staggering, insidermembership effects, long-term unemploymenteffects and labour force adjustments. Spilloversarise when endogenous variables have explana-tory power in other equations of the system and,thus, generate interactive dynamics. The chainreaction adjective highlights the intertemporalresponses of the unemployment rate to changesin the exogenous variables (shocks) propagatedby a network of interacting lags.

Our arguments for differentiating the CRTmethodology from the traditional (dynamic)simultaneous equations models are as follows.

CRT models focus on dynamics and flourish in

a distributed-lag environment, placing emphasiso n t he r ol e o f i mp ul se r es po ns e f un ct io ns(IRFs).9 The reason that CRT models refer tothe inherent simultaneity issue as spilloversis to stress the plethora of feedback mecha-nisms, and flag the importance of the univariate

representation of unemployment for the investi-gation of its evolution (see Section III(iii)). Thisis used to derive the IRFs which are a focalpoint of the CRT and lead to the evaluation ofthe driving forces of unemployment, presentedin Section VI.

Another main feature of the CRT is that theunemployment trajectory is influenced not onlyby stationary variables (such as oil prices orinterest rates) but by trended variables (such ascapital stock and working-age population) aswell. Therefore, it has the capacity of account-ing for the unemployment effects of both tempo-

rary and permanent shocks to the labour market.This contradicts the conventional wisdom as itis expressed by the NRU theory. Furthermore,we should point out that the interplay of lagsand changes in growing variables gives rise to

frictional growth.In what follows, we give a brief outline of the

popular natural rate hypothesis, explain analyti-cally the phenomenon of frictional growth andhow it challenges the NRU, and present theCRT via a stylised labour market model.

(i) The Natural Rate Hypothesis

In macrolabour modelling the NRU (un

) isgenerally defined as the equilibrium value atwhich unemployment stabilises in the long-run(see, e.g. Ball & Mankiw, 2002).

Based on the observation that, on average, theunemployment rate has been trendless over sev-eral decades, the dominant view in the literatureargues that growing variables cannot influenceits value in the long-run (i.e. the NRU). Thesimplest illustration of this conventional wisdomis the following dynamic single-equation model(for expositional ease we consider an AR(1)process with one determinant, and ignore the

error term):

8 As our focus is on the real side of the economy,we do not consider nominal variables and, thus, werefer to the NRU rather than the NAIRU. Karanassouet al. (2008b) examine a CRT model consisting ofboth real and nominal variables. For a discussion ofNRU, NAIRU and CRT models, see Karanassou et al.(2009).

9 Although dynamics and IRFs are also focal pointsin the (structural) VAR framework, the crucial differ-ences between the two methodologies are beyond thescope of this article.




7/26

ut a ut1 bxt; 1

where ut denotes the unemployment rate, xt isan exogenous variable, the autoregressive (AR)parameter a is less than one in absolute valueand b is a constant.

Let us reparameterise this autoregression as:

ut b

1 axt

a

1 aDut; 2

where D denotes the difference operator. Inthe long-run, assuming that xt is not growing,the unemployment rate stabilises (i.e. ut ut)1 , Dut 0) and so Equation (2) reduces toits steady-state and gives the NRU:

un b

1 ax LR ; 3

where the superscript LR denotes the long-runvalue of the variable. In this case, the first termof equation (2) can be interpreted as trendunemployment and the second one as cyclicalunemployment.

(ii) The Phenomenon of Frictional GrowthLet us rewrite the above pedagogical AR(1) Equa-

tion (2) in terms of some endogenous variable yt:

yt b

1 axt|fflfflffl{zfflfflffl}

trend or steady-state

a

1 aDyt|fflfflfflfflffl{zfflfflfflfflffl}

cycle if long-run growth0

frictional growth otherwise

: 4

Note that the first term of Equation (4) capturesthe trend of yt, whereas the second termdescribes its cyclical variations if the growth of

xt is zero in the long-run. If however, the exoge-nous variable xt has a non-zero long-run growthrate, the second term of Equation (4) gives rise to

frictional growth. In other words, frictionalgrowth results from the interplay of lags andgrowth. It is also worth noting that the long-runelasticity of y with respect to x, a focal point ineconomics, is b(1 ) a) regardless of whetherfrictional growth is zero or not. It is this fact, we

believe, which has led economists to disregardthe role of frictional growth in macroeconometricmodels. As we demonstrate next, although fric-tional growth does not affect single-equationunemployment rate models (and generally NRUmodels), it has major implications for dynamicmulti-equation labour market models.

Without loss of generality, let us analyse oneof the simplest cases of multi-equation labour

market models where frictional growth canarise. Consider the following AR(1) labourdemand and static labour supply equations, eachfeaturing a single exogenous variable:

nt a1nt1 b1kt; 5

lt zt; 6

where nt denotes employment, lt is labour force,kt is capital stock, zt is working-age population,the AR parameter is 0 < a1 < 1 and b1 is a posi-tive constant.10 All variables are in logs and weignore the error terms for ease of exposition.

The unemployment rate (not in logs), can beapproximated by the difference between thelabour force and employment:

ut lt nt: 7

This definition implies that the unemploymentrate stabilises in the long-run, that is, DuLR 0,when DlLR DnLR. In other words, for unem-ployment stability in the long-run, the growthrate of employment should be equal to thegrowth rate of the labour force, say g.11 Thisrestriction can also be expressed in terms of thelong-run growth rates of the exogenous vari-ables:

DnLR DlLR g ,b1

1 a1DkLR

Dz LR g: 8

We refer to Equation (8) as the frictionalgrowth (FG) stability condition, as it ensuresthat the unemployment rate stabilises in thelong-run.

Next, reparameterise the labour demand (Eqn 5)as:

nt b1

1 a1kt

a1

1 a1Dnt; 9

and subtract it from the supply Equation (6) toobtain the following unemployment dynamics:

10 Observe that when all variables are I(1), thelabour demand and supply Equations (5) and (6)imply the co-integrating vectors (1, )b1/(1 ) a1)),(1, )1), respectively.

11 Note that the growth rate of log variables isproxied by their first differences, D(). We can plausi-bly assume that both capital stock and working-agepopulation are growing variables with growth ratesthat stabilise in the long-run.




8/26

ut zt b1

1 a1kt

|fflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflffl}

trend

a1

1 a1Dnt|fflfflfflfflfflffl{zfflfflfflfflfflffl}

cycle if long-run growth 0

frictional growth otherwise

:

10

Observe that, similar to the AR(1) Equation (4),the term in parentheses captures the trend ofthe unemployment rate and the second termdescribes its cyclical variations if employmentand labour force are not growing in the long-run. If, on the other hand, employment andlabour force have non-zero long-run growthrates (owing to the growing capital stock andworking-age population), the second term gener-ates frictional growth.

We should stress that the frictional growthstability condition (8) can be confirmed in the

light of Equation (10). Since the second termof Equation (10) is either zero or constant inthe long-run (depending on the value at whichthe long-run labour growth stabilises), theunemployment change is only associated withthe change in the first term of Equation (10).Therefore, in the long-run

Du Dz b1

1 a1Dk 0;

that is, unemployment stabilises in the long-runowing to the stability condition (8).12

An implication of the FG stability condition

(8), is that the long-run solution of Equation(10) is:

uLR z LR b1

1 a1kLR

|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}

NRU

a1b1

1 a12DkLR ;

|fflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflffl}frictional growth

11

where the first term in parentheses measures theNRU (the steady-state) and the second term cap-tures frictional growth, that is, the interplaybetween growth and the employment adjustmentprocess. Note that, although working-age popu-

lation and capital stock are growing variables,

the unemployment rate is trendless in the long-run owing to the FG stability condition (8).

The long-run value (uLR), towards which theunemployment rate converges, reduces to theNRU only when frictional growth is zero, that iseither when the exogenous variables do not

grow or when the labour demand and supplyequations have identical dynamic structures (inthis example, when both labour demand andsupply are static equations). Therefore, fric-tional growth implies that the NRU ceases toserve as an attractor for the unemployment rateand challenges its pivotal role in policy-making.Another implication of frictional growth is thatunemployment cannot be simply reduced intothe sum of trend and cyclical components.

Obviously, frictional growth does not featurein studies which (i) examine dynamic single-equation models (such as Nickell et al., 2005),

as there are no interactions, or (ii) use staticmodels (such as Nickell, 1997; or Blanchard &Wolfers, 2000), as labour market dynamics aresimply sidelined.

We should emphasise that one of the salientfeatures of the CRT models is that, in contrastto single-equation NRU models, they can alsoinclude growing exogenous variables. The onlyrequirement is that each equation is balanced(i.e. dynamically stable) so that each growingdependent variable is driven by the set of itsgrowing determinants. It can be shown thatequilibrating mechanisms in the labour market

and other markets jointly act to ensure that theunemployment rate is trendless in the long-run(Karanassou & Snower, 2004). In terms of theaforementioned analytical model, these mecha-nisms can be expressed in the form of the FGstability condition (8). In short, CRT modelsgenerate frictional growth which implies that, as

long-run NRUsteadystate

frictional growth,

1. unemployment may substantially deviatefrom its natural rate, and

2. it cannot be decomposed into trend and

cyclical components.

Owing to our limited knowledge of the long-run growth rates of the exogenous variables wecannot obtain reliable estimates of frictionalgrowth, and consequently of the long-run unem-ployment rate. Therefore, CRT models do notaim at evaluating the natural (or long-run)unemployment rate, but, instead, focus on the

12 This general result is analysed in Karanassou andSnower (2004), where it is explained why the assump-tion of a constant capital labour ratio (a1 1 ) b1) isunnecessarily restrictive. In other words, condition (8)guarantees the plausible property of unemploymentstability in the long-run.




9/26

contributions of the exogenous variables tothe evolution of the unemployment rate (see Sec-tion V).

(iii) Unemployment Determinants in a CRTModel

We illustrate the analytical workings of theCRT by adding a wage-setting equation to thesimple model of the previous section (Eqns 5,6), and by introducing spillover effects in thelabour demand and wage-setting equations:

nt a1nt1 b1kt c1wt; 12

wt a2wt1 b2bt c2ut; 13

lt zt; 14

where nt is employment, wt is real wage, lt is

labour force, ut is the unemployment rate, kt isreal capital stock, bt is real benefit and zt is theworking-age population; the bs and cs arepositive constants, and 0 < a1,a2 < 1. As in theprevious section, all variables except unemploy-ment are in logs and we ignore the error termsfor simplicity.

In the context of the aforementioned model, anumber of important remarks are to be made.

1. The AR parameters capture the employmentadjustment and wage/price staggering effects,respectively.

2. The cs generate spillover effects, as changes

in an exogenous variable in one equation,say working-age population in labour sup-ply, can also affect the real wage (by feed-ing through ut) and, in turn, labour demand(by feeding through wt). Observe that whenc1 and c2 are both non-zero, all labour mar-ket shocks generate spillover effects. If c2 0, that is, if unemployment does notp ut d own wa rd p re ss ur e o n w ag es , t he nlabour demand and supply shocks do notspillover to wages. Note that the main feed-back mechanism in this model is providedby the wage elasticity of labour demand. If

c1 0, that is, if labour demand is com-pletely inelastic with respect to wages, thenshocks to wage setting do not spillover toemployment and unemployment. In thiscase, benefits do not influence unemploy-ment, and the unemployment effects of theexogenous variables kt and zt can be ade-quately measured by individual analysis of

the labour demand and supply equations,respectively.

3. In the presence of spillover effects, the indi-vidual labour demand and supply equationscannot provide adequate measures of the sen-sitivities of unemployment to the exogenous

variables. We refer to the bs as the localshort-run elasticities (i.e. the elasticitiesobtained simply by eye inspection) to distin-guish them from the global ones, whichincorporate all the feedback mechanisms inthe labour market model. The global elastici-ties can be obtained by the univariate repre-sentation of unemployment, which expressesunemployment as a function of its own lagsand the exogenous variables in the system.

It is worth noting that the empirical CRTmodel in the next section is an expanded version

of the aforementioned stylised labour marketmodel characterised by a variety of laggedadjustment processes and feedback mechanisms.

We now show how to derive the univariaterepresentation of unemployment by rewritingthe labour demand and real wage Equations (12)and (13) as:

1 a1Bnt b1kt c1wt; 15

1 a2Bwt b2bt c2ut; 16

where B is the backshift operator. Substitutionof Equation (16) into Equation (15) gives

1 a1B1 a2Bnt c1c2ut 1 a2Bb1kt

c1b2bt: 17

Next, rewrite the labour supply (Eqn 14) as:

1 a1B1 a2Blt 1 a1B1 a2Bzt:

18

Finally, use definition (7) and subtract labourdemand (Eqn 17) from labour supply (Eqn 18) toobtain the univariate representation (or reduced

form) of the unemployment rate equation:

c1c2 1 a1B1 a2But

1 a2Bb1kt c1b2bt

1 a1B1 a2Bzt: 19

The term reduced form relates to the fact thatthe parameters of the equation are not estimateddirectly; instead, they are some non-linear func-tion of the parameters of the underlying labour




10/26

market system (12)(14).13 Alternatively, theunemployment rate Equation (19) can be writ-ten as:

ut /1ut1 /2ut2 hkkt hbbt hzzt

hka2kt1 /1zt1 /2zt2; 20

where /1 a1 a2

1c1c2; /2

a1a21c1c2

; hk b1

1c1c2; hb

c1b21c1c2

and hz 1

1c1c2.

The reduced form unemployment rate Equa-tion (20) displays the following characteristicsof the CRT. The AR coefficients /1 and /2embody the interactions of the employmentadjustment (a1) and wage-price staggering (a2)processes. The hs embody the feedback mecha-nisms built in the system, as they are a functionof the slope coefficients (bs) of individual Equa-tions (12)(14) and the spillovers (cs). Thus, the

hs describe the global short-run sensitivities ofunemployment with respect to the exogenousvariables. The interplay of dynamics acrossequations is further emphasised by the laggedstructure of the exogenous variables. Using timeseries jargon, we refer to the lagged exogenousvariables as moving-average terms.

Observe the difference between the localand global sensitivities of unemployment. Forexample, the local short-run semielasticity ofthe unemployment rate with respect to capitalstock is b1 (by Eqns 7 and 12), whereas itsglobal short-run elasticity is b1 (1 + c1c2) (by

Eqn 20). Generally, the plethora of spillovers inthe system may render the local elasticitiesunreliable, as they might affect both their signand size. The CRT approach takes this fact intoaccount and uses the global sensitivities of theunemployment rate, as opposed to the localones, to diagnose the economic plausibility ofthe labour market system.

As we already discussed in Section III(ii), akey element of the CRT is that capital stock, atrended variable, influences the time path of theunemployment rate, a stationary variable. In thecontext of our stylised labour market model we

can justify this result as follows. First, capitalstock enters the system as a determinant of

employment, a trended variable. Labour demand(Eqn 12) is a balanced equation as it is dynami-c al ly s ta bl e ( |a1| < 1 ). S ec on d, t he t re nd edlabour force is driven by working-age popula-tion (also a trended variable); and the staticlabour supply (Eqn 14) is itself a balanced equa-

tion. Third, the labour demand equation remainsbalanced once the wage (Eqn 13) has beensubstituted into it (see Eqn 17). As a result, thederived unemployment rate, Equations (19) or( 20 ), i s i ts el f d yn amic al ly s ta bl e, a s i t i sobtained by the difference of two dynamicallystable equations (i. e. labour supply anddemand).

The detailed interplay between dynamics andgrowth, and the interactions among the variouslagged adjustment processes portrayed in theCRT model cannot be captured by a single-equation NRU model. In other words, estimat-

ing a single unemployment rate equation is notequivalent to estimating a multi-equationlabour market system. Karanassou et al. (2003)show that if a single-equation NRU model, andeach equation of a CRT model, which com-prises labour supply and demand equations,have all identical regressors, then the two esti-mation procedures will yield identical results.However, in structural labour market systems,it is highly unlikely that each constituent equa-tion will have the same regressors, and so thesingle-equation model cannot be viewed as anunbiased summary of the CRT multi-equation

model.

IV Estimated EquationsIn the spirit of the aforegiven stylised CRT

model, we estimate a dynamic multi-equationlabour market system with spillovers comprisinglabour demand, labour supply and wage-settingequations. As already noted in Section II, ourestimation technique is the ARDL approach toco-integration analysis, which is also used inLewis and MacDonald (2002) and in most otherCRT studies.

We s ho ul d e mp ha si se t ha t t he s el ec te d

equations are dynamically stable and pass thestandard mis-specification tests (for linearity,structural stability, no serial correlation, homo-scedasticity and normality) at conventional sig-nificance levels. In addition, we estimate ourequations as a system using three-stages leastsquares (3SLS) to take into account the potentialendogeneity and cross-equation correlation. Inwhat follows, we present our results and provide

13 Equation (19) is dynamically stable (i.e. bal-anced) as (i) products of polynomials in B which sat-isfy the stability conditions are stable, and (ii) linearcombinations of dynamically stable polynomials in Bare also stable.




11/26

an overall evaluation of the empirical labourmarket model.

(i) DataWe use annual data running from 1970 to

2006. All time series were obtained from theOrganisation for Economic Cooperation andD ev el op me nt ( OE CD ) E co no mi c O ut lo ok except for the ASX 200 stock market index(source: Bloomberg), and oil prices (source:IMF International Financial Statistics). Table 1gi ves the definitions of the variables inthe selected specifications of the estimatedequations.

Most of our variables are quite standard, butsome deserve further clarification. Financial

wealth (fw), for example, is defined as the ASX200 stock index expressed in real terms and nor-malised by labour productivity (given by theratio of GDP over employment). This followsPhelps and Zoega (2001) who argue that theswings in economic activity are influenced bythe firms expectations about future productivity,and proxy the latter by using the financial wealthvariable. dwp is a wage-push dummy for thewage rise of 19741975, aiming at an equal payfor women across occupational groups. TheAccord dummy d

ac, in turn, captures the effectof; (i) the wage pause that took place in 1983,

and (ii) the Prices and Incomes Accord (i.e. thesystem of wage bargaining at a national level),which was implemented from the third trimesterof 1983 until the mid-1990s with the aim to pre-vent excessive wage growth. We use the wage-push and accord dummies in the wage-settingequation, following Pissarides (1991) whoargued that they capture the income policieseffects on wages (see Section V for details).

(ii) Labour DemandThe labour demand equation contains a large

set of determinants including supply- and

demand-side factors, and external influences (seeTable 2). The low persistence coefficient (mea-sured by the sum of the AR coefficients) of 0.19,indicates quick employment adjustments to mar-ket shocks. Note that our selected equationincludes the first two (annual) lags of employ-ment, which is consistent with the dynamic struc-ture of Lewis and MacDonald (2002) of sevenquarterly lags. This is in contrast to Pissarides

TABLE 1 Definitions of Variables

c, constantn, employment (log) tax, direct taxes on households (% GDP)l, labour force (log) gov, government expenditures (% total capital stock)u, unemployment rate l ) n fd, foreign demand exports imports (% GDP)w, real wage per employee (log) tot, terms of trade log(export prices/import prices)k, real total capital stock (log) r, real long-term interest rateb, social security benefits (% GDP)

fw, financial wealth logreal Asx 200 stock index

labour productivity

z, working-age population (log)

dwp, wage-push dummy 1 in 19741975, 0 otherwise

oil, real oil price (log) dac, accord dummy 1 in 19831995, 0 otherwise

Note: GDP, gross domestic product.Sources: OECD Economic Outlook, Bloomberg, IMF International Financial Statistics.

TABLE 2 Labour Demand Equation, 19722006. Dependent

Variable: nt; Estimation Methodology: ARDL

OLS 3SLS

Coefficient P-value Coefficient P-value

c 7.07 [0.000] 6.97 [0.000]nt)1 0.19 [0.038] 0.20 [0.004]Dnt)1 0.58 [0.000] 0.58 [0.000]wt )0.12 [0.064] )0.11 [0.037]Dwt 0.18 [0.016] 0.10 [0.106]rt )0.29 [0.017] )0.24 [0.015]Drt 0.27 [0.019] 0.26 [0.003]kt 0.45 [0.000] 0.44 [0.000]fwt 0.01 [0.042] 0.01 [0.031]oil t)1 )0.01 [0.025] )0.01 [0.001]

tot t 0.11 [0.003] 0.12 [0.000]fdt)1 0.39 [0.010] 0.44 [0.000]gov t 2.78 [0.000] 2.68 [0.000]Dgov t )1.51 [0.013] )1.58 [0.000]

Std. error 0.007 0.007R2 0.999 0.999

Notes: 3 SL S, t hr ee -s ta ge s l ea st s qu ar es ; A RD L, a ut o-

regressive distributed lag; OLS, Ordinary least squares.




12/26

(1991) finding of two quarterly lags and a muchhigher persistence coefficient of around 0.70.

The first subset of determinants reflects(mainly supply-side) fundamentals and includeswages, interest rates and capital stock (Pissa-rides, 1991; Webster, 2003). The long-run elas-

ticity of employment with respect to wages is)0.15, which is in the lower range of the esti-mates given by Webster (2003) for Australia(between )0.15 and )0.8). This is in contrast tothe )0.6 estimate given by Lewis and MacDon-ald (2002) who use output, rather than capitalstock, as an explanatory variable. We find thatthe long-run elasticity of employment withrespect to capital stock is 0.56. For Australia,the scarce evidence on this value is mixed.Pissarides (1991) finds it very close to 1 so thathe can safely restrict it to be unity, a valuewhich is consistent with an underlying Cobb

Douglas production function. In contrast toPissarides (1991), Rowthorn (1999) places theelasticity of substitution between capital andemployment at (i) 0.77 using Bean et al.s( 19 86 ) s tu dy , ( ii ) 0 .5 9 u si ng N ew el l a ndSimons (1985) estimates and (iii) 0.62 usingLayard et al.s (1991) results. The elasticity ofsubstitution predicted by our model is consistentwith the latter two.

The second subset of driving forces is alsostandard. Financial wealth, which Phelps andZoega (2001) argue is crucial for explaining thelong swings of the unemployment rate in the

OECD countries, enters our equation signifi-cantly and with the expected positive sign.Higher oil prices hurt labour demand and betterterms of trade increase employment.

Finally, in line with Pissarides (1991), weidentify the following two demand-side influ-ences, both with the expected sign: (i) net for-eign demand (as percentage of GDP) increasesemployment, and (ii) government expenditures(normalised by capital stock) have a positiveimpact on labour demand. Note that foreigndemand, oil prices and the terms of trade pro-vide a rich set of external factors affecting

the Australian labour market. Also, Pissarides(1991) argues that detrending governmentexpenditures by capital stock allows an evalua-tion in terms of business cycle effects, whereasour reasoning is somewhat different. Controllingfor capital stock, we interpret the coefficient ofgovernment expenditure as its effect on employ-ment in the absence of crowding out effects oninvestment.

(iii) Wage SettingThe wage-setting equation is pretty standard

real wages are determined by unemployment,capital deepening and a set of institutional vari-ables. The persistence coefficient is rather lowa t 0 .2 5 ( se e T ab le 3) , r efl ec ti ng t he q ui ck adjustments to shocks that characterise theAnglo-Saxon labour markets.

Unemployment reflects labour market demandconditions, and puts downward pressure on

wages through the change in, rather than the levelof, the unemployment rate (Dut). In turn, capitaldeepening (nt)kt), which commonly proxieslabour productivity, exerts the expected positiveinfluence. The institutional variables (socialsecurity benefits, bt, and direct taxes of house-holds, taxt) and the dummy d

wp appear as wage-push factors. In contrast, the Accord dummy dac,displays the expected negative sign.14

TABLE 3Wage-Setting Equation, 19722006. Dependent

Variable: wt; Estimation Methodology: ARDL

OLS 3SLS


c 5.00 [0.000] 5.07 [0.000]wt)1 0.25 [0.033] 0.24 [0.032]Dwt)1 0.34 [0.002] 0.31 [0.006]Dut )0. 80 [0.002] )0. 63 [0. 024]Dut)1 0.42 [0.237] 0.47 [0.068]kt)nt 0.25 [0.000] 0.26 [0.000]bt 1.09 [0.044] 1.03 [0.014]tax t 0.68 [0.040] 0.74 [0.019]dwp 0.06 [0.000] 0.06 [0.000]

dac

)0. 05 [0.000] )0. 05 [0. 000]

Std. error 0.016 0.016R2 0.978 0.978

Notes: 3 SL S, t hr ee -s ta ge s l ea st s qu ar es ; A RD L, a ut o-regressive distributed lag; OLS, ordinary least square.

14 Various attempts to include trade union densityas an additional wage-push factor were unsuccessful.

For the most part of our sample period real wagesexhibit alternate spells of stability and growth,whereas trade union density displays a downwardtrend. The contrasting trajectories of the two variablesmay be responsible for the clear lack of significanceof trade union density in the context of the otherwiserich specification of our wage equation. Of course,this does not negate the trade union influence on wagesetting evidenced in different contexts (see, e.g. thestudy by Waddoups, 2005).




13/26

(iv) Labour ForceAs indicated by the large persistence coeffi-

cient of 0.86 (see Table 4), labour force deci-sions are very persistent. These decisions aredetermined by economic and demographic fac-tors.

On the economic side, there is a discouraged-worker effect so that the higher the unemploy-ment rate the lower the incentives to participatein the labour market. We also find that the realwage has a positive (albeit mild) effect on laboursupply, which is slightly reinforced when itinteracts with the Accord dummy (wt d

ac). Inother words, the elasticity of labour force withrespect to the real wage increased during thePrices and Incomes Accord period. This resultis consistent with Lewis and Kirby (1988) whofind that the labour supply curve shifted out-wards with the Accord. However, it is in contrast

to Pissarides (1991) who argues that the incomeand substitution effects cancel, as there is nosignificant influence of wages on labour supply.

On the demographic side, labour supplydepends on the working-age population with along-run elasticity equal to unity.15 This impliesthat, for a given unemployment rate and realwage, all people within the working-age groupwill end up participating in the labour market(e.g. at some point, students will enter thelabour force).

(v) Model Diagnostics

To validate our estimated labour marketmodel we follow three main diagnostic avenues:

1. We plot the actual and fitted values of theunemployment rate and evidence the accu-racy of the latter.

2. We use the Johansen procedure and find thatthe long-run solutions implied by the ARDLapproach represent co-integrating vectors.

3. We compute the global (interactive) sensi-tivities of unemployment with respect to theexogenous variables and find that they dis-play the expected signs.

The fitted values of our estimated model areobtained by solving for the unemployment ratein the estimated labour demand, wage setting,labour supply equations (3SLS estimates inTables 24) and the unemployment definition(Eqn 7). Figure 1 plots the actual and fitted val-ues and shows that our estimation tracks thedata very well. We should emphasise that it ismore difficult to obtain a good fit with a multi-

equation labour market system than with a sin-gle unemployment rate equation. This is becauseof the numerous feedback mechanisms amongthe endogenous variables which are activatedwhen we solve the model for the unemploymentrate.

TABLE 4 Labour Force Equation, 19722006. Dependent

Variable: lt; Estimation Methodology: ARDL

OLS 3SLS


c )0.28 [0.020] )0.30 [0.007]lt)1 0.86 [0.000] 0.86 [0.000]Dlt)1 )0.18 [0.126] )0.14 [0.152]ut )0.32 [0.000] )0.31 [0.000]Dut)1 )0.37 [0.000] )0.36 [0.000]wt 0.05 [0.012] 0.05 [0.004]wt d

ac 0.003 [0.000] 0.003 [0.000]zt 0.14 (*) 0.14 (*)

Std. error 0.004 0.004R2 0.999 0.999

Notes: 3SLS, three-stages least squares; ARDL, autoregressive

distributed lag; OLS, ordinary least square; (*), restricted tounity.

0

2

4

6

8

10

12

1972

1974

1976

1978

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

Actual unemployment

Fitted unemployment

FIGURE 1Unemployment Rate: Actual and Fitted Values

15 The chi-squared Wald test on this restriction is3.38 [0.077], where the P-value is in square brackets.Non-rejection of the hypothesis indicates that theslope estimates in the restricted version are prettyclose to those in the unrestricted one, and implies thatt he parsi mony of t he l abour force equat ion i senhanced by imposing the restriction. (The unre-stricted estimates are available upon request.)




14/26

The second diagnostic deals with the long-runrelationships of labour demand, wage settingand labour supply, which we estimated byapplying the ARDL approach to co-integrationanalysis (second column in Table 5). In particu-lar, we test whether they are significantly differ-ent from the co-integrating vectors obtained by

the Johansen procedure (see also Lewis & Mac-Donald, 2002). Once the maximal eigenvalueand trace statistics confirm that the variablesinvolved in each equation are co-integrated, theJohansens co-integrating vectors (third columnin Table 5) are restricted to take the corre-sponding long-run values of our estimated equa-tions. The last column in Table 5 displays thecorresponding LR tests, which confirm that ourestimation procedure is consistent with that ofJohansen.16

TABLE 5Testing the Long-Run Relationships in the Johansen Framework

ARDL Johansen LR test

Labour demand ( N w k) ( N w k)OLS ( 1 0:15 0:56 ) ( 1 0:08 0:52 ) v2(2) 0.81 [0.666]3SLS ( 1 0:14 0:55 ) v2(2) 0.61 [0.617]

Wage setting ( w k n ) ( w k n )OLS ( 1 0:33 0:33 ) ( 1 0:27 0:22 ) v2(2) 0.23 [0.889]3SLS ( 1 0:34 0:34 ) v2(2) 0.21 [0.899]

Labour force ( L w z ) ( L w z )OLS (URTD) ( 1 0:05 1:19 ) ( 1 0:05 1:22 ) v2(2) 2.58 [0.275]3SLS RTD OLS ( 1 0:36 1:00 ) v2(2) 6.28 [0.043]

Notes: P-value in square brackets; 5% critical values: v2(2) 5.99. ARDL, autoregressive distributed lag; LR, likehood ratio;OLS, Ordinary least square; RTD, restricted; URTD, unrestricted.

TABLE 6Global Long-Run Unemployment Rate Sensitivities

k r b tax gov oil tot fd fw z

Long-run sensitivity )0.15 0.11 0.22 0.16 )1.23 0.006 )0.06 )0.20 )0.005 0.31

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Labour force

Employment

0.53

0.38

(a)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 150.45

0.40

0.35

0.30

0.25

0.20

0.15

0.10

0.15

(b)

FIGURE 2Responses to a Capital Stock Shock. (a) Employment

and Labour Force; (b) Unemployment Rate

16

It should be noted that the VAR model underly-ing the Johansen procedure contains all the variablesin our labour market model, both the I(0) and I(1)variables. Naturally, the co-integration tests only con-sider the I(1) variables in our models: nt, wt, lt, kt andzt. This implies that we test two restrictions in thelabour demand, wage-setting and labour supply equa-tions. To conserve space, we do not report the resultsof the underlying unit root and co-integration tests.These are available upon request.




15/26


16/26

Incomes Accord, known as the Accord, whichrun from 1983 to 1996 and consisted of a set ofwage-setting arrangements under which unionscommitted to restrain their wage claims inexchange for other compensatory social provi-sions (in the form, e.g. of selected tax cuts,

superannuation awards or improvements inessential social services). The scope of theAccord was to create a more flexible system(which would be more focused on productivityperformance), to modernise unions and to bringa broad change in industrial relations (whichwould result in fewer disputes). Other reformsentailed administrative changes to enhance theefficiency of public services, removal of tradebarriers (especially for the clothing, auto andwool industries) and also state-led reforms ofcentral economic sectors. In addition, the bank-ing sector opened to foreign companies.

F ro m 1 99 6 t o 2 00 7 t he g ov er nmen t wa sagain in the hands of the Liberal Party but thistime it was co-existing with several Labor stategovernments, which spend more than two-thirdsof total public revenues. This was a period ofcomprehensive structural reforms fostering eco-nomic liberalisation. A first pillar was alreadyset up in 1995, when Australias state govern-ments agreed on a so-called National Compe-tition Policy (NCP), which is considered as themost extensive economic reform programme inthe history of Australia. At the public level, theNCP aimed at preventing an anticompetitive

conduct of public enterprises, introducing taxreforms and, more recently, changing the fund-ing and delivery arrangements for various gov-ernment services at the public level. At theprivate level, the NCP aimed at further disman-tling trade barriers and enhance the deregula-tion of the financial system. A second pillarwas the removal of the Accord and the liberali-sation of the labour market, whereas a thirdpillar was the reduction of government reve-nues in conjunction with a fiscal consolidationprocess (indeed, quite often during th isperiod the government annual budget was in

surplus).20

The economic and policy developments ofthese years are pictured in Figure 3 with theplots of the 10 exogenous variables used inour model. Until the mid-1970s there was arise in social security benefits and direct taxes(together with a temporary increase in govern-

ment expenditures), which was followed by asudden reduction in the second half of the1970s (coinciding with the change of govern-ment in 1975). The period of 1970s and early1980s is characterised by the oil price shock,the increase in interest rates and the deteriora-tion of the terms of trade. In contrast, therewas no significant variation in trade surplus.Capital accumulation, financial wealth and thegrowth rate of working-age population experi-enced a sharp downturn in the beginning ofthe 1970s, but recovered afterwards and endedu p i n 1 98 3 wi th v al ue s s imil ar t o t ho se i n

1973.During the marathon boom of the 1990s and

2000s, there was not much action on the institu-tional side (benefits and direct taxes), but therewas a clear decline in government expendituresand a fall in interest rates. Financial wealth didnot change much, despite its brief increase inthe second half of the 1990s. Regarding thegrowth rate of working-age population, therewas a clear acceleration after its sharp declinein the previous decade. Like the 1970s, oilprices were increasing but, in clear contrast tothat period, there was a boom in the terms of

trade. This was accompanied by a sharp erosioni n t he t ra de b al an ce ( as p er ce nt of G DP ),which moved from a surplus close to 4 per centin the second half of the 1990s to a deficit of 4per cent in 2006. This dramatic performance ofthe external sector is one of the two salient fea-tures of the 19932006 period. The second oneis the profound acceleration of capital accumu-lation capital stock growth increased fromaround 2 per cent in 1993 to 4 per cent by theend of this period.

Apparently, the 19932006 boom in Australiawas not contaminated by the problems faced by

its traditional partners: the 1990s Japanese tur-moil and the US downturn of the early 2000s.We understand that two key factors acted as thesafeguards of the Australian performance.

First, as Parham (2004) argues, the accelera-tion in capital accumulation was important forthe productivity revival in the 1990s. AlthoughParham (2004, p. 239) acknowledges the linkbetween the stellar productivity performance of

20 ONeill and Fagan (2006) survey the last threedecades of economic reforms in Australia. The OECDEconomic Surveys for Australia (1998, 1999, 2000,2001, 2003, 2004, 2006) provide detailed accounts ofits developments with specific focus on the economicpolicy and the process of structural reforms.




17/26

1%

2%

3%

4%

5%

6%

1970 1974 1978 1982 1986 1990 1994 1998 2002 2006 1970 1974 1978 1982 1986 1990 1994 1998 2002 20068%

6%

4%

2%

0%

2%

4%

6%

8%

10%

1970 1974 1978 1982 1986 1990 1994 1998 2002 20063%

4%

5%

6%

7%

8%

9%

1970 1974 1978 1982 1986 1990 1994 1998 2002 20068%

9%

10%

11%

12%

13%

14%

15%

1970 1974 1978 1982 1986 1990 1994 1998 2002 200614.5%

15.0%

15.5%

16.0%

16.5%

17.0%

17.5%

18.0%

1970 1974 1978 1982 1986 1990 1994 1998 2002 20062.1

1.6

1.1

0.6

0.1

(a) (b)

(c) (d)

(e) (f)

(g) (h)

(i) (j)

0.4

0.3

0.2

0.1

0.0

0.1

0.2

1970 1974 1978 1982 1986 1990 1994 1998 2002 2006 1970 1974 1978 1982 1986 1990 1994 1998 2002 20064%

3%

2%

1%

0%

1%

2%

3%

4%

5%

1970 1974 1978 1982 1986 1990 1994 1998 2002 20064.0

4.5

5.0

5.5

6.0

6.5

1970 1974 1978 1982 1986 1990 1994 1998 2002 20060.8%

1.0%

1.2%

1.4%

1.6%

1.8%

2.0%

2.2%

FIGURE 3Exogenous Variables: Actual and Fixed Values in 1973 and 1993. (a) Capital Accumulation; (b) Real Interest

Rates; (c) Social Security Benefits; (d) Direct Taxes on Households; (e) Government Expenditures; (f) Oil Prices;(g) Terms of Trade; (h) Foreign Demand; (i) Financial Wealth; (j) Working-age Population (Growth)




18/26

Australia and its economic reforms (in line withthe OECD Economic Surveys), he states thatThe accumulation of physical and human capi-tal has laid a long-term foundation for produc-tivity growth. Over the 19501994 period,Madden and Savage (1998, p. 362) find open-

ness to trade and international competitivenessto be significant sources of Australian labourproductivity in the short-run, but in the long-run, fixed capital accumulation is the dominantsource of productivity improvement.

Second, the emergence of China in the globaltrading scene (by joining the WTO in 2001)boosted Australias economic activity. On theone hand, the strong Chinese demand pushedthe prices of Australian raw materials relent-lessly up (especially on iron ore, coal and alu-minium). On the other hand, China was a sourceof cheap imports (office supplies, appliances and

toys), which implied lower costs for Australianfactories and easily affordable consumer prod-ucts. However, we should point out that Austra-lias resilience over the 20002001 USrecession probably owes more to a favourablehousing market (driven by government incen-tives), than to China.21

VI Driving Forces of UnemploymentIn the context of the empirical CRT model of

Section IV, we assess the driving forces of thelabour market by evaluating the contribution ofthe exogenous variables to the evolution of

unemployment during two periods of interest the unemployment upturn in 19731983 and theunemployment decline in 19932006.

In each period we simulate the model (i.e.t he 3 SL S e qu at io ns i n T ab le s 2 4 a nd t heunemployment definition, Eqn 7) under thecounterfactual scenario that the exogenous vari-ables are fixed (one at a time) at their 1973 or1993 values. Figure 3 plots the actual seriesfor the whole sample (solid lines) and theirfixed values over the 19731983 and 19932006 periods (dotted lines). In turn, Figures 4and 5 plot the trajectories of actual unemploy-

ment (solid lines) and the simulated unemploy-men t r at e ( do tt ed l in es ) w he n e ac h o f t heexogenous variables is kept constant at its 1973and 1993 values, respectively. In this way, thedotted lines picture what would have been the

unemployment trajectory had the given exoge-nous variable remained at the initial value ofits time path, instead of having evolved as itactually did.

The dynamic contributions of the exogenousvariables to the evolution of unemployment are

then measured as the difference between itsactual and simulated values. Therefore, the con-tribution of a given exogenous variable over aperiod of time represents how much wouldthe unemployment rate have deviated from itsactual value in the absence of any change in thevariable.22

It is worth pointing out that the contributionof each exogenous variable reflects only itsdirect effect on unemployment, ceteris paribus,and does not capture any indirect effect throughits possible influences on other exogenous vari-ables in the model. The latter can be captured if

all other exogenous variables were to be endo-genised, which is beyond the scope of our work.We should also note that, although VAR modelshave the capacity of disentangling the effects ofall variables, as they are all endogenous, theyhave the disadvantage of examining the effectsof only generic one-off shocks instead of actualones. Thus, the CRT methodology owing tothe rich structure of endogenous and exogenousvariables, network of spillovers, overall modeldiagnosis and consideration of actual shocksand their unemployment contributions has acomparative advantage in the empirical assess-

ment of economic developments over VARsand traditional structural macroeconometricmodels.23

(i) 19731983The unemployment rate increased by 8.1

pp over this period, from 2.3 per cent in 1973 to10.4 per cent in 1983. Our analysis is picturedin Figure 4 and shows that a variety of factorsaccounted for this rise, some of them beingcommon in the macrolabour literature.

21 We a re gr at ef ul t o D avi d N or ma n f or th isinsight.

22 For the sake of brevity and clarity, the variableswith rather negligible contributions are not shown inFigures 4 and 5. These are foreign demand and work-ing age population in the first period (19731983),institutions and oil prices in the second period (19932006) and financial wealth in both.

23 See Karanassou and Sala (2008) for a discussionof the empirical methodologies of (structural) VARs,CRT and traditional structural macro models.




19/26

We find that our set of institutional variables,that is direct taxes, benefits and the (wage-pushand Accord) dummies, account for 2.2 pp of thesurge in unemployment (see Figure 4c). This isabout 27 per cent of the 8.1 pp unemployment

increase over this period. That is, in the absenceo f t he u pw ar d t re nd o f t he se v ar iab le s i nthe 1970s and early 1980s, documented inFigures 3c and 3d, the unemployment ratew ou ld h av e r ea ch ed 8.2 p er c en t i n 1 98 3,

2

3

4

5

6

7

8

9

10

11

1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983

Simulatedtrajectory2.3%

10.4%10.8%

1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983

Simulatedtrajectory

2

3

4

5

6

7

8

9

10

11

2.3%

10.4%

8.9%

1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983

Simulatedtrajectory

2

3

4

5

6

7

8

9

10

11

2.3%

10.4%

8.7%

1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983

Simulatedtrajectory

2

3

4

5

6

7

8

9

10

11

2.3%

10.4%

9.2%

1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983

Simulatedtrajectory

2

3

4

5

6

7

8

9

10

11

Actualtrajectory

2.3%

10.4%

8.2%

1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983

Simulatedtrajectory

2

3

4

5

6

7

8

9

10

11

Actualtrajectory

Actualtrajectory

Actualtrajectory

Actualtrajectory

Actualtrajectory

2.3%

10.4%

8.5%

(a) (b)

(c) (d)

(e) (f)

FIGURE 4Unemployment Contributions, 19731983. (a) Capital Accumulation; (b) Real Interest Rates; (c) Institutional

Variables; (d) Government Expenditures; (e) Terms of Trade; (f) Oil Prices

Note: Simulated trajectories result from fixing each exogenous variable at its 1973 value. Institutional variables

comprise benefits, direct taxes, and the income policy dummies.




20/26

instead of the actual 10.4 per cent.24 Note thatinstitutional (i.e. wage pressure) variables areconventionally blamed to be a key contributorto the rise in unemployment. For example,Gregory (2000, p. 123) asserts that

My v ie w i s t ha t l ar ge a ve ra ge r ea l w ag eincreases in Australia and most OECD coun-t ri es d ur in g t he mid 1 97 0s w er e l ar ge lyresponsible for the rising unemployment ratesover the following seven or eight years.

and Pissarides (1991) argues that about threequarters of the unemployment increase between19701973 and 19761979 was because of equi-librium factors, in particular the rise in the fiscalwedge and unemployment benefits.

Two other standard determinants in the litera-ture are commodity price and interest rates ho ck s. O ur mod el c ap tu re s t he s ho ck s i ncommodity prices through the fall in the termsof trade and the rise in oil prices. Figures 4eand 4f show that these account for a total of 2.9pp of the unemployment rise (1.7 pp owing tothe deterioration of the terms of trade, and 1.2pp owing to the increase in oil prices), which isabout one-third of the unemployment increaseduring this period. The unemployment impact ofthe terms of trade has already received someattention in Australia (Gruen, 2006). Regardinginterest rate shocks, we find that their contribu-tion was to increase unemployment by 1.5 pp(i.e. 18.5 per cent of its 8.1 pp rise). Had thereal long-term interest rate not been risingbetween 1973 and 1983, as shown in Figure 3b,the unemployment rate would have ended theperiod at 8.9 per cent rather than its actual 10.4per cent (see Fig. 4b).

These results show that changes in institu-tions, commodity prices and interest ratesaccount for 81.5 per cent (= (2.2 + 2.9 + 1.5)8.1) of the unemployment increase from 1973 to1983 and, thus, are in line with the conventionalwisdom that shocks and institutions (such asbenefits and taxes) are responsible for highunemployment rates.

However, our model also unveils the roleplayed by other factors, such as capital stockand government spending, which are quite oftensidelined in the literature. In particular, observethat the decrease in capital stock growth in thesecond half of the 1970s (Fig. 3a) was responsi-

ble for 1.1 pp, or 18 per cent, of the 6.0 ppincrease in unemployment from 1976 to 1980(Fig. 4a). The powerful influence of capitalaccumulation will become apparent in the nextsection, where we examine the 19932006 per-iod. In terms of government expenditures, ourfindings demonstrate that had they not beenreduced so dramatically in the early 1980s,unemployment would have been 1.9 pp lowerthan its 1983 actual value (Fig. 4d). Further-more, the relatively generous spending by thegovernment during 19761980 led to a reductionin the unemployment rate of around 2 pp, on

average. In view of these findings, we wonderwhether the fiscal consolidation of the early1980s came too soon, while other driving forceswere still pushing unemployment upwards.

Al th oug h fi na nc ia l w eal th a nd f or ei gndemand affect significantly labour demand, andworking-age population affects significantlylabour supply, they had a minimal contributionto the trajectory of the unemployment ratefrom 1973 to 1983. For example, our simula-tions find that had financial wealth remained atits relatively high 1973 level (Fig. 3i), unem-ployment would have ended up the period at

10.3 per cent instead of its actual 10.4 percent. In other words, financial wealth contrib-uted to approximately only 1 per cent of ther is e i n t he u ne mp lo ymen t r at e d ur in g t hi speriod. This is in contrast to the studies of thestructuralist theory advocates (e.g. Phelps &Zoega, 2001) who argue that financial wealth isthe crucial driving force of the unemploymentrate.

(ii) 19932006The unemployment rate declined by 6.3 pp

during this period, from 11.2 per cent in 1993

t o 4 .9 p er c en t i n 2 00 6. F ig ur e 5 d isp la ysthe contributions of the exogenous variables tothe fall in the unemployment rate. A centralresult of our analysis is that it distinguishesbetween the driving forces of the unemploymentrate during the 19731983 and 19932006periods.

First, capital accumulation contributed toreducing unemployment by 5.0 pp (Fig. 5a).

24 In particular, benefits account for a rise of 0.6pp, taxes for 0.5 pp, the wage-push dummy for 0.2 ppand the Accord dummy for 0.9 pp. As this dummywas implemented in the last year of the 19731983period, the value of 0.9 only refers to its effect in1983. That is, had the wage-pause and the Accord nottaken place, unemployment would have been higherby 0.9 pp in 1983.




21/26

Had capital stock growth stayed constant at its1993 value, unemployment would have been9.9 per cent in 2006 instead of its actual 4.9per cent. The progressive increase in capitalaccumulation since the early 1990s (Fig. 3a)was closely linked with much higher

investment in new technologies, especiallythose associated with the ICTs and the neweconomy.25 This sort of investment explainsmos t o f t he i nc rea se i n c ap it al d ee pe ni ng( Ba nk s, 2 00 1) a nd i s a ss oci at ed wi th t heunprecedented period of rapid productivitygrowth in Australia, very much like the one inUnited States (see Parham, 2004). Increasedinternational exposure also enhanced the per-formance of capital accumulation. In particular,the Australian Treasury (2006) reports that therecent terms of trade boom is being used bysome export-oriented companies to invest and

secure their advantageous international posi-tion.26

In addition, we find that the fall in interestrates (Fig. 3b), which accompanied the acceler-ation of capital stock since the early 1990s,accounted for 0.7 pp reduction in the unemploy-ment rate (Fig. 5b). Note that this effect is ontop of any potential influence of the interest rateon unemployment through capital accumulation(or another variable in the model).27

Second, although the impact of oil prices onunemployment in 19932006 was weaker thanin 19731983 (0.5 pp and 1.2 pp contributions,

respectively), the terms of trade factor becamecrucial for the Australian labour market. In par-ticular, Fig. 5e reveals that in the absence ofthe terms of trade boom, unemployment wouldhave remained high, reaching 9.0 per cent in2006, rather than its actual 4.9 per cent (i.e. theterms of trade were responsible for 4.16.3 = 65

per cent of the fall in unemployment). Theextraordinary evolution of the terms of tradesince 1999 (Fig. 3g) resulted from a combina-tion of cheap imports, mainly from China, andexports of raw materials (such as uranium, coal,nickel and other mineral resources) whose prices

were on the rise owing to sustained increases indemand, again mainly from China (AustralianTreasury, 2006).

Third, regarding the external-related drivingforces of Australian unemployment, foreigndemand played a noticeable role in the evolu-tion of unemployment only during the veryrecent years. In particular the fast deteriorationof the trade balance in the 2000s (from a sur-p lu s o f 4 p er c en t i n 2 00 1 t o a d efi ci t o f 4per cent in 2006) put substantial upward pres-sure on unemployment (see, respectively,Figs 3h and 5f), thus counteracting the terms

of tr ade influence. Had for ei gn d em andremained at its 1993 value, unemploymentw ou ld h av e b ee n l ow er b y 2 .7 p p i n 2 00 6.This deterioration in the trade balance reflects(i) an income boost, which allowed the pur-chase of a greater volume of imports, and (ii)supply constraints in Australias export sector(particularly resources).28

Fourth, benefits and taxes had no substantialimpact on unemployment during this period.This is in line with Giesecke (2008, p. 36) whoexamines the period 1996/19972001/2002 andconcludes that his results are not suggestive of

policy changes having exerted a large influenceon either GDP or consumption growth. Never-theless, we find that the abolition of the Accordcontributed to the fall in the unemployment rate(its abolition in 1995 led to a 2.3 pp reduction).

Note that the Accords objective was to put abrake to the large average real wage increasesof the mid-1970s, which were seen as responsi-ble for the rising unemployment rates of the fol-lowing several years. However, Gregory (2000)argues that, although average real wages did notincrease during the Accord period, the resultinglabour shift affected mainly women (as employ-

ment growth was directed primarily towardspart-time jobs and jobs traditionally held bywomen). In particular, The jobs did not reducethe unemployment rate but added to the partici-pation rate of women (p. 123). Therefore, our

25 Investment in Information and CommunicationTechnologies in Australia is part of a general restruc-turing process that has predominantly affected thewholesale, finance and insurance sectors.

26

The links between capital accumulation and pro-ductivity growth are clearly beyond the scope of thisarticle. In other papers we have dealt with the impactof productivity on the labour market (see Agnese &Sala, 2009, for Japan; Karanassou & Sala, 2008, forUnited States).

27 To disentangle the direct and indirect (via capitalstock) effects of the interest rate would require mod-elling capital stock as a function of the real interestrate, which is beyond the scope of this paper.

28 We thank David Norman for bringing up thisreflection.




22/26

finding (namely, that the absence of the incomespolicy led to a fall in unemployment) is in linewith Gregory (2000) under the provision thatthe labour force curve (mainly on the womensside) shifted inwards by more than the labourdemand curve did.

Furthermore, the 1990s was also a period of fis-cal consolidation (Fig. 3e). According to Songand Freebairn (2006) this consolidation had anoverall positive effect on the Australian econ-omy, mainly because it allowed the fall in interestrates. According to our simulation findings, the

4

5

6

7

8

9

10

11

12

1993 1995 1997 1999 2001 2003 2005

Simulated

trajectory

Actualtrajectory

Actualtrajectory

Actualtrajectory

Actualtrajectory

Actualtrajectory

Actualtrajectory

11.2%

4.9%

9.9%

1993 1995 1997 1999 2001 2003 20054

5

6

7

8

9

10

11

12

Simulated

trajectory

11.2%

4.9%

5.6%

1993 1995 1997 1999 2001 2003 20054

5

6

7

8

9

10

11

12

Simulated

trajectory

11.2%

4.9%

7.2%

1993 1995 1997 1999 2001 2003 20052

3

4

5

6

7

8

9

10

11

12

Simulated

trajectory

11.2%

4.9%

2.8%

1993 1995 1997 1999 2001 2003 20054

5

6

7

8

9

10

11

12

Simulated

trajectory

11.2%

4.9%

9.0%

1993 1995 1997 1999 2001 2003 20052

3

4

5

6

7

8

9

10

11

12

Simulated

trajectory

11.2%

4.9%

2.2%

(a) (b)

(c) (d)

(e) (f)

FIGURE 5Unemployment Contributions, 19932006. (a) Capital Accumulation; (b) Real Interest Rates; (c) Prices and

Incomes Accord; (d) Government Expenditures; (e) Terms of Trade; (f) Foreign Demand

Note: Simulated trajectories result from fixing each exogenous variable at its 1993 value.




23/26

reduction in government expenditures led to anunemployment increase of 2.1 pp (Fig. 5d). Thiswas the direct effect of the cutbacks in publicspending since 1993, which may have also hadcompensating macroeconomic effects throughlower interest rates or via other indirect routes (in

the form of changes in the various exogenousvariables of the model). Finally, as in the 19731983 period, the contribution of financial wealthto the unemployment trajectory is negligible.

To check the robustness of our results, wealso evaluated the contribution of the exogenousvariables to the unemployment trajectory duringthe 19831989 and 19891993 periods. Here isa brief summary of our findings.29 Capital accu-mulation substantially affects the evolution ofunemployment in all four periods, and its influ-ence becomes stronger as we move along thesub-samples. Although oil prices remain impor-

tant during 19831989 (as in the 19731983period), their impact significantly diminishesafter 1989. In contrast, the contribution of inter-est rates is significant throughout the sample:the rise in interest rates during 19731983 and19891993 (Fig. 3) pushes unemployment up(by 1.5 pp and 0.8 pp, respectively), whereastheir fall in 19831989 and 19932006 reducesunemployment (by 0.7 pp and 0.8 pp, respec-tively). Similar to the 19731983 period, therise in direct taxes over the years 19831989puts upward pressure on the unemployment rate.However, as taxes fluctuate within a narrow

range of around 2 pp during 19891993 (seeFig. 3d), their effects are negligible in this per-iod. Regarding the influence of the Prices andIncomes Accord, our results for the 19831989and 19891993 periods confirm our finding dur-ing the 19932006 years. Finally, social securitybenefits, the terms of trade and foreign demandvariables are also relevant over the 19831989and 19891993 periods.

VII ConclusionsIn this article, we examined the evolution of

Australian unemployment using the CRT method-

o lo gy . Th e CRT d if fe rs f ro m mai ns tr ea mapproaches in two key aspects. First, rather thanestimat

Documents

Labout Market Dynamics in Australia What Drives Unemployment