Labor market regimes and the effects of monetary policy

Transcription

1 Labor market regmes and the effects of monetary polcy CEFOS Workng Paper Ncola Acocella & Govann D Bartolomeo Department of Economcs, Unversty of Rome La Sapenza Douglas A. Hbbs, Jr.* CEFOS Abstract In ths paper we use a standard mult-unon, monopolstc competton model to nvestgate the qualtatve and quanttatve responses of nflaton and unemployment to monetary polcy actvsm under dfferent nsttutonal arrangements n the labor market, whch are defned by the rgdty of nomnal wages. We show that the effects of monetary polcy on the real economy depend crtcally on the wage formaton regme, and on the ways n whch the restrctveness of polcy nteracts wth product prce competton, wth unon centralzaton, and wth the weght placed on real wage premums as compared to unemployment n unons optmal programs. Our nterpretaton of the results emphaszes how the posture of monetary polcy toward nflaton nfluences the strategc calculatons that drve unon wage settng behavor n dfferent nsttutonal settngs. * douglas@douglas-hbbs.com Douglas.Hbbs@cefos.gu.se Centrum för forsknng om offentlg sektor Center for Publc Sector Research

2 CEFOS Workng Paper 2 Labor market regmes and the effects of monotary polcy Ncola Acocella, Govann D Bartolomeo & Douglas A. Hbbs, Jr., 2005 ISSN: CEFOS Centrum för forsknng om offentlg sektor Göteborgs unverstet Box Göteborg offce@cefos.gu.se Tel CEFOS Center for Publc Sector Research Göteborg Unversty P.O. Box 720 SE Göteborg, Sweden offce@cefos.gu.se Tel

3 Fnal Revson Labor market regmes and the effects of monetary polcy Ncola Acocella *, Govann D Bartolomeo *, and Douglas A. Hbbs, Jr. ** forthcomng n Journal of Macreconomcs Abstract In ths paper we use a standard mult-unon, monopolstc competton model to evaluate analytcally and numercally the effects of monetary polcy on nflaton and unemployment under dfferent nsttutonal arrangements n the labor market that are defned by the rgdty of nomnal wages. We show that the effects of monetary polcy on the real economy depend crtcally on the wage formaton regme, and on the ways n whch the restrctveness of polcy nteracts wth product prce competton, wage settng centralzaton and the utlty weght unons place on real wage premums as compared to unemployment. Our nterpretaton of the results emphaszes how the posture of monetary polcy toward nflaton nfluences the strategc calculatons drvng unons wage settng behavor n dfferent nsttutonal envronments. JEL Classfcaton: E52, E58, J51. Keywords: Polcy games, monetary polcy neutralty, trade unons, monopolstc competton, labor markets. * Department of Economcs, Unversty of Rome La Sapenza. ** Correspondng author: Douglas Hbbs, CEFOS, Göteborg Unversty, Box 720, Göteborg 40530, Sweden; douglas@douglas-hbbs.com Acocella and D Bartolomeo gratefully acknowledge fnancal support from the Unversty of Rome La Sapenza; Hbbs s grateful for support from the Göteborg Unversty Donatonsfonder. An earler verson of ths paper appeared as Workng Paper No 58 of the Publc Economcs Department, Unversty of Rome La Sapenza. 1

4 1. Monetary polcy, wage settng nsttutons and macroeconomc performance Monetary polcy neutralty means that monetary nstruments are unable to affect real varables, such as output and employment. 1 The Barro-Gordon (1983) model and ts many varants, nspred by the semnal paper of Kydland and Prescott (1977), are the man templates for modern analyss of monetary polcy ssues. In terms of the framework used n ths paper and lad out below, the bare-bones Barro-Gordon setup corresponds to a game n whch the central bank sets the money supply n order to mnmze nflaton and unemployment after unons set nomnal wages so as to optmally trade off expected real wages and unemployment of ther members. Although unons are Stackelberg leaders, the central bank s objectves and optmal polcy reactons are common knowledge and unon wage polces are condtoned on ratonal expectatons of the money supply. Nomnal wages are therefore adjusted to crowd-out the postve effects that monetary expansons otherwse would have on output and employment by moderatng real wages. The result s a Stackelberg equlbrum characterzed by monetary neutralty and excess nflaton. 2 A more favorable mplcaton of ths lne of research s that a conservatve central bank pursung a strngent, non-accommodatng polcy s able to contan nflaton wthout real costs n the form of systematcally hgher unemployment and depressed output a vew developed theoretcally by Rogoff (1985) and supported to varous degrees emprcally by evdence n Grll et al. (1991), Alesna and Summers (1993), Bleaney (1996) and Ejffnger et al. (1998)) and others, all of whch took ratngs of central bank ndependence as good proxes for polcy conservatsm. Non-neutralty was shown to arse, however, f wages only partally adjust to monetary changes because of the exstence of mult-perod overlappng contracts as n Fscher (1977) or because the polcy authorty has an nformaton advantage over wage and prce setters as n Canzoner (1985) or because unons have a pure dstaste for nflaton as n Gylfason and Lndbeck (1994). 1 The classcal defnton of monetary neutralty mples that autonomous changes n the money supply have no nfluence on the level of real output (Patnkn (1956)). In the polcy games lterature whch came a generaton later, money s typcally endogenous. A defnton better suted to modern frameworks of analyss would be that monetary polcy s neutral (non-neutral) when equlbrum output and employment do not (do) depend on the preferences of the monetary polcymaker (Acocella and D Bartolomeo (2004)). 2 Unon power over nomnal wage settng, however, mght nfluence the monetary authorty s objectves and constrants. Fscher and Summers (1989), for example, argued that other thngs beng equal ndexaton lowers the cost of nflaton, whch by tself gves the authortes an ncentve to pursue more nflatonary polces. The mplcaton s that unon coordnaton effectvely ndexng wages would tend to ncrease nflaton. On the other hand, Waller and VanHoose (1992) ponted out that ndexaton steepens the aggregate supply curve, reducng the output and employment gans from (unantcpated) nflaton. The ncentve to pursue an nflatonary polcy n the frst nstance s therefore dmnshed, mplyng that through ths channel coordnated unon acton to ndex wages mght reduce an nflaton bas. 2

5 More recent contrbutons to the polcy game lterature stress new channels of monetary nonneutralty that do not depend on stcky wages, nformaton asymmetres and drect unon averson to nflaton, but nstead operate through the nteracton of central bank polces wth wage and prce settng nsttutons. Theoretcal demonstratons by Bratoss and Martn (1999), Soskce and Iversen (2000), Corcell et al. (2004, 2005), among others, mpled that when there s a multplcty of wage settng unons and product markets are monopolstcally compettve, a Barro-Gordon framework may delver polcy non-neutralty. In partcular, restrctve monetary polcy rules that are nternalzed by wage and/or prce setters can lead to favorable equlbrum outcomes for output and unemployment as well as nflaton. Data reported n Soskce and Iversen (2000) and more extended emprcal analyses by Hall (1994), Hall and Franzese (1998), Iversen (1999 ch.3), Cukerman and Lpp (1999) and Franzese (2001a, 2002 ch.4), whch tested explctly for the contngency of real monetary polcy effects on the nsttutonal framework of wage/prce formaton, yelded evdence broadly supportng ths lne of theory, although n hs masterful revew of these and related studes Franzese (2001b) conceded that dsagreement remans over the precse nature of these nteractve effects (p.457). The am of ths paper s help clarfy the sources and potental numercal magntudes of nsttutonally contngent monetary non-neutralty. Specfcally, we apply a workhorse multunon, monopolstc competton model to pn down analytcally and evaluate numercally the nomnal and real macroeconomc effects of monetary polcy n two dstnct wage formaton regmes: () rgd wage labor markets n whch unons contract wages n advance for at least one perod (commtment) and cannot alter ther clams after the monetary authorty acts, and () flexble wage markets n whch unons nteract smultaneously wth the central bank and may adjust wages freely n any perod. The remander of the paper s organzed as follows. In the next secton we ntroduce the workhorse mult-unon, monopolstc competton model. Solutons to the optmal programs of the central bank and labor unons are gven n Secton 3. In Secton 4 we derve analytc solutons for wages, prces and unemployment levels n flexble and rgd wage labor markets, and we show the qualtatve responses of these varables to varatons n monetary polcy conservatveness, centralzaton-coordnaton of wage settng, and prce competton n product markets. In Secton 5 we assgn plausble ranges of numercal values to model parameters, and we calculate the quanttatve effects of monetary polcy conservatveness on macroeconomc outcomes under dfferent assumptons about wage settng centralzaton, 3

6 prce competton, and the relatve weght placed upon real wages vs. unemployment by unons operatng n flexble and rgd wage nsttutonal regmes. We gve ntutve nterpretatons of the results, placng specal emphass on how the posture of polcy n dfferent nsttutonal settngs affects the strategc calculatons of unons n settng nomnal wage levels. 2. The economc setup We use a smple, workhorse model 3 of an mperfectly compettve economy composed of several unons and monopolstcally compettve frms. Several unons represent the entre labor force and exert monopoly power over nomnal wage levels. 4 We ndex unons by [ 1, n] and for smplcty assume they are equal n sze, each representng a share of the labor force equal to σ = 1/n. An array of frms producng dfferentated goods and servces set product prces. One unon represents the labor force and sets the nomnal wage of workers n each frm. Frms assocated wth unon are ndexed by j. Frms maxmze a one perod proft functon under demand and producton technology constrants. Frm-level demand s (1) Y d j P j = P η M P where d Y j s demand for a frm s output, and P j and P are frm-product prces and the general prce level, respectvely. The latter s convenently assumed to equal the geometrc average of j P. η ( 1, ) s the elastcty of product demand wth respect to relatve prce. Frm-level demand s also affected by aggregate demand, whch equals the aggregate real money supply M P. The central bank controls aggregate demand perfectly by settng the money supply at any gven prce level. Each frm uses a producton technology defned by decreasng returns to labor nputs (2) Yj L α j = α ( 0,1) 3 The Appendx gves fuller dervaton of the economc setup whch draws upon Bratoss and Martn (1999), Soskce and Iversen (2000), and especally Corcell et al. (2005) and Cukerman (2004). 4 The model therefore s not applcable to economes n whch wages are set manly by ndvdual bargans, as n the contemporary US, or to (notonal) economes n whch wages are manly set by just one all-encompassng unon. Ths pont s made explct n analyses of the model presented ahead. 4

7 where L j s the labor nput of frm j. Note that technologcal progress (secular productvty) s mplctly fxed at 1.0. Equatons (1) and (2) mply that frm j s condtons for proft maxmzaton under monopolstc competton can be wrtten as log-lnear equatons for relatve product prce and labor demand (3) p p = θ ( 1 α)( m p) + α( w p) j d (4) l = θ m p η( w p) j where lower case varables denote logs of the correspondng upper case varables, log nomnal wage level of workers n unon and θ ( 0,1) 1 =. α + η α ( 1 ) w s the As shown n the appendx, wthout loss of generalty we set log labor supply parametrcally to zero, from whch t follows that the unemployment rate among unon s members mpled by equaton (4) s (5) U = θ η( w p) ( m p). Takng averages of equatons (3) and (4), one obtans (as shown n the Appendx) the followng aggregate reduced forms for the log prce level and the rate of unemployment n the macroeconomy: (6) = α + ( 1 α) p w m (7) U = w m. By usng equaton (6) we can rewrte equaton (5) as (8) θ η α( η 1) U = w w m. Equaton (8) mples that the unemployment rate of unon s members s postvely related to unon s log nomnal wage clams, w, and negatvely related to the economy-wde, average log nomnal wage level, w. The unon-specfc wage always domnates the economy-wde wage because η α( η 1) >. Unemployment s also decreasng n the nomnal money supply. The economy-wde log nomnal wage level s equal to ( 1 ) w= σw + σ w, where recall 1 σ = s the relatve sze of unon s membershp (whch s equvalent to the degree of n 5

8 centralzaton or coordnaton of nomnal wage settng), and wage set by other unons. It follows that equaton (8) can be expressed {( 1 ) ( 1 )( 1) } (9) U = θ η ασ η w α σ η w m. w s the average log nomnal Equaton (9) s the reduced form unemployment rate for unon. When settng the nomnal wage, unon faces two opposte effects. Frst, rasng w decreases employment demand for the unon s members due to the hgher labor costs mposed on the frm. Second, rasng contrbutes to ncreasng the economy-wde wage, whch by tself makes frm j more compettve. The former effect naturally domnates the latter whenever wages are set by more than one unon. w 3. Optmal polces 3.1 The central bank s problem We assume that the central bank sets the log money supply m to maxmze the followng quadratc objectve functon, subject to (6) and (7): (10) CB β U π 2 2 = 2 2 where π = ( p p ) s the nflaton rate and β ( 0, ) 1 defnes the central bank s averson to nflaton (or deflaton) relatve to unemployment. For a one perod optmal polcy, the lagged prce level, p 1, s gven parametrcally, and so wthout loss of generalty t can be set to zero, allowng us to treat current prces and nflaton rates nterchangeably; π = p. Solvng the central bank s problem yelds the wage-contngent optmal monetary polcy rule (11) m= φw. The reacton parameter ( ) ( α) 2 α 1 α β 1 φ = 1 β + 1 represents the central bank s wllngness to accommodate nomnal wage settlements, whch we shall call monetary polcy 1 conservatveness. When β < and thereforeφ < 0, wage ncreases are at least α 1 α partly accommodated, and the central bank s lberal or populst. As β goes to ts lower bound of zero, φ approaches the lower bound of ts doman, -1, and the central bank s ultra- 6

9 lberal. Its objectve functon s targeted solely on mnmzng the unemployment rate and monetary polcy accommodates fully nomnal wage ncreases. The reverse s true when β > α 1 ( 1 α ). In ths case φ > 0 and the central bank s conservatve, settng the money supply so as to offset at least partly wage-nduced ncreases to the prce level. As the bank s nflaton averson parameter β goes to ts upper lmt, φ approaches ts upper bound, α /(1 α), and we observe the ultra-conservatve central bank that ams only to mnmze 1 nflaton. At β =, the reacton coeffcent φ = 0. The central bank s passve and α 1 α monetary polcy s nether accommodatng nor non-accommodatng The labor unons problem Each unon seeks to maxmze a lnear-quadratc preference functon, membershp s log real wage ( w p) and unemployment rate (12) TU = γ ( w p) { 1, 2,... n} 2 U 2 where γ > 0 s the relatve utlty weght of the log real wage premum. 6 The frst order condton for each unon s (13) γ p U 1 U = 0 w w U as arguments: TU, wth the where recall that the current perod log prce level, p, defnes the nflaton rate, π. The frst order condton (13) represents the senstvty of unon wage polces to nflaton and unemployment, respectvely. As wll become clear n the next secton, after substtuton for the endogenous varables π and terms of parameters and gven varables. U we obtan the correspondng optmalty condton n 4. Labor market regmes 5 Hence, at a productvty parameter of 23 α = (the typcal value for advanced economes), polcy would be passve when the Central Bank s utlty weght on nflaton relatve to unemployment, β, s equal to 4.5. Values greater or less than 4.5 therefore defne what we call conservatve and lberal polces, respectvely. 6 The real wage s the net-of-productvty premum because the producton functon n (2) excludes technologcal progress, and ths mples that under perfect competton the equlbrum values of w, p, m and U are normalzed to zero. These features of the model should be kept n mnd when nterpretng the settngs of γ and other parameters n secton

10 4.1 Regmes and nformaton Equlbrum outcomes of the game are obtaned by solvng equatons (11) and (13) under the unons nformaton constrant. Dfferent equlbrum concepts are assocated wth dfferent nformaton settngs and assocated specfcatons of expectatons and, therefore, they vary by labor market regme. Followng Ljungqvst and Sargent s (2000) dscusson, a Stackelberg game mples that unons maxmze ther objectves by nternalzng the central bank s reacton to ther wage polces, and they cannot change nomnal wages thereafter (because, for example, wages are bound by non-renegotable contracts extendng more that one perod). Consequently, ths regme s assocated wth pre-determned wages set one perod n advance. By contrast, n a non-cooperatve Nash gamng envronment unons and the monetary authorty nteract smultaneously, takng realzatons of m and w, respectvely, as gven. What we desgnate a Nash game therefore mples flexble wages that can be adjusted every perod along wth the money supply. In both flexble and rgd wage labor markets monopolstc frms set prces after the realzatons of wages and the money supply. The sequences of moves can be depcted Unons Central Bank ( Stackelberg ) Frms Unons & Central Bank ( Nash ) Frms. Both wage settng regmes are plausble. The Stackelberg leader-follower setup s consstent wth the fact that n unonzed economes wages are normally set one or more years n advance, whereas the central bank n prncple can adjust monetary nstruments wth great frequency, gvng t consderable flexblty vs-à-vs wage outcomes. On the other hand, monetary polcy effects may be sluggsh, whch motvates the vew that Nash soluton concepts yeld reasonable one-shot approxmatons to a repeated game between unons and the monetary authorty. The most approprate characterzaton of the labor market regme may therefore depend on the lag structure of polcy effects, whch for reasons outsde the scope of ths paper could well vary across tme and space. 4.2 Flexble labor markets In the Nash flexble wage case, each unon fnds the nomnal wage that maxmzes ts objectves by dfferentatng the aggregate prce (nflaton) constrant n (6) and the unemployment functon n (9) wth respect to nomnal wages of other unons, w, takng both the money supply, m, and the w, as gven, and substtutng these dervatves nto the 8

11 optmalty condton n (13). Each unon s optmalty condton n flexble wage markets s then (14) γ ( ασ) θ η ασ( η ) 1 1 = 0 U whch mples the wage settng functon(s) (15) η α η 1 α 1 σ η 1 η α η 1 w = m+ w + γ ( 1 ασ) > 0 η ασ η 1 η ασ η 1 η ασ η 1 for { 1, 2,... n}. Unon always reacts to monetary expanson by rasng ts wage clams less than proportonally, snce the postve coeffcent of m n (15) s always less than one, except when all workers are represented by a sngle unon ( σ = 1), n whch case the reacton coeffcent on m s unty. Moreover, unon always responds to ncreases n the average wage of other unons by rasng ts nomnal wage, no matter how conservatve the central bank s. The Nash equlbrum s found by solvng the system of n + 1 equatons mpled by equatons (11) and (15). Imposng the symmetry condton w= w = w yelds the economy-wde, equlbrum log nomnal wage level as 2 (16) ( 1) 1 η α η w = 1 ασ γ > φ η ασ η 1 After substtutng the optmal value of w gven by (16) and the optmal value of m gven by (11) nto equatons (6) and (7), equlbrum nflaton and unemployment n the macroeconomy are (17) (18) ( 1 ασ )( η α ( η 1) )( α φ + αφ ) ( η ασ( η 1) )( 1+ φ) π = γ ( 1 ασ ) η α ( η 1) η ασ( η 1) U = γ > 0. Snce the central bank s preferred nflaton rate was normed to zero, equaton (17) shows that an nflaton bas exsts, whch represents an neffcent outcome of the non-cooperatve Nash equlbrum that ncreases the central bank s dsutlty wthout affectng a unon s felcty. 0 9

12 The bas declnes wth ncreases to central bank conservatveness, 7 wth ncreases to wage settng centralzaton, and wth ncreases to product market competton: ( αη ( 1) η)( ασ 1) 2 ( 1 ) ( 1) π = γ < 0 φ + φ ασ η η ( ( 1 )) ( 1) ( 1 ) ( 1 ) 2 π α φ α α η η = αγ < 0 σ + φ η+ ασ η ( α φ( 1 α) )( 1 σ)( ασ 1) ( 1 ) ( 1 ) π = αγ < 0. η + φ αη η + η 2 Snce the conservatveness parameter, φ, s the only term dependent on the central bank s preferences, t s clear from (18) that systematc monetary polcy cannot affect the real economy, represented here by the unemployment rate (U). Hence, n the flexble wage Nash regme we obtan the standard Barro-Gordon results of nflaton bas and real-sde polcy neutralty, notwthstandng the complcatons to the economc envronment ntroduced by the presence of a multplcty of unons n the labor market and monopolstc prce competton n product markets. However, unemployment decreases as both centralzaton (coordnaton) of wage settng and product market competton ncrease: (( 1) ) ( 1 ) U α η η = αγ < 0 σ αη η + η 2 ( 1 )( 1) ( 1 ) U σ ασ = αγ < 0. η αη η + η 2 At σ = 1 or η =, unemployment falls to ts sngle unon compettve market mnmum, ( 1 ) U = α γ, whch for gven productvty of labor, α, ncreases lnearly wth the weght γ attached to the real wage premum relatve to unemployment. Note that U ( 1 ) = α γ s not the perfect competton outcome. Unemployment goes to the compettve equlbrum U = 0 only when unons wth monopoly wage power place no weght on achevng a real wage above that warranted by productvty, that s, only when γ = Rgd labor markets 7 Note that the bas asymptotes to zero when the central bank becomes ultra-conservatve, that s, when the bank s utlty weght on nflaton approaches t upper lmt β = and, therefore, the conservatveness polcy φ = α /1 α. parameter approaches ts upper lmt 10

13 The rgd labor market or Stackelberg equlbrum s found n the usual way by solvng the two-stage game by backward nducton. 8 In the second stage the central bank solves ts problem takng nomnal wages set by unons as gven, whch yelds the optmal wagecontngent monetary polcy rule of equaton (11). In the frst stage each unon solves ts problem by nternalzng the central bank s reacton functon (11), n the lght of equatons (6), (9) and the gven wage clams of other unons. After substtuton of the approprate wage dervatves nto (13), we derve the followng realzaton of unon s optmal wage condton: (19) π U γ 1 ( κ1w κ2w ) = 0 w w κ γ ( w p) w U w 2 w = w + κ1 κ1 for { 1, 2,... n} wth κ = η ασ( η ) θ + σφ, 1 1 κ 2 = α 1 σ η 1 θ 1 σ φ. Equaton (19) represents a system of n equatons. By mposng the symmetry condton w w w (20) = = and solvng, we obtan the economy-wde equlbrum log nomnal wage as ( 1 ( α φ+ αφ) σ) ( α + ( 1 α) η) ( η( 1 σ( α φ+ αφ) ) + ασ( 1+ φ) )( 1+ φ) w = γ > Inflaton and unemployment n the macroeconomy may be derved drectly from equaton (20) gven equatons (6), (7), and (11). We have (21) (22) ( 1 ( α φ + αφ) σ)( α φ + αφ) ( α + ( 1 α) η) ( η( 1 σ ( α φ + αφ) ) + ασ ( 1+ φ) )( 1+ φ) 0. π = γ ( 1 ( + ) ) + ( 1 ) 1 ( + ) + 1+ α φ αφ σ α α η U = γ > η α φ αφ σ ασ φ As n the flexble wage regme, equaton (21) shows that n rgd wage labor markets optmal polcy agan creates an nflaton bas that declnes wth central bank conservatveness, wage settng centralzaton, and product market competton: 0. 2 n( σ( α + φ( α 1) ) 1) + ( αη ( 1) η) 2 ( 1 ) ( 1 ) 1 ( 1 ) π ασ 1+ φ 2σ φ 1 α α φ 1 α + 1+ α = γ < 0 φ + φ ασ + φ + η σ α φ α ( ) 2 ( ) { } 8 Strctly speakng we have a three stage game wth frms movng last to set fully flexble prces that are perfectly antcpated and nternalzed by unons and the monetary authorty. Our focus s on unon-central bank nteractons and nothng s lost by abstractng from the prce settng stage. 0 11

14 ( α φ( 1 α) )( α( η 1) η) { ( 1 ) ( 1 ( 1 ) )} 2 π = αγ < σ ασ + φ + η + σφ α ασ ( α φ( 1 α) )( 1 σ) ασ 1+ σφ( α 1) ( 1 ) ( 1 ) 1 ( 1 ) { } 2 π = αγ < η + φ ασ + φ + η + σφ α ασ 0 0. Equaton (21) also mples that the nflaton bas agan goes to zero when the central bank becomes ultra-conservatve, that s as β and φ α /1 ( α). Unlke the stuaton n the flexble wage regme, however, equaton (22) shows that unemployment n the rgd wage regme s a negatve functon of central bank conservatveness φ ( 1 ) ( ( 1) ) { ( 1 ) ( 1 ( 1 ) )} 2 U σ σ α η η == αγ < φ ασ + φ + η σφ α ασ Hence, monetary polcy s non-neutral, and conservatve monetary polcy s able to acheve both lower nflaton and lower unemployment when wages are pre-commtted vs-à-vs monetary polcy. Unemployment also falls, as n the flexble wage case, wth ncreases to unon centralzaton and product prce competton: ( ) ( 1+ ) ( 1) { ( 1 ) 1 ( 1 ) } 2 U φα η η = αγ < σ σ + φ + η σ α φ α ( ) ( 1 ) ( 1 ) 1 { ( 1 ) 1 ( 1 ) } 2 U σ σ α φ α = αγ < η ασ + φ + η σ α φ α Summary and nterpretaton of results The general pattern of macroeconomc outcomes assocated wth flexble and rgd wage labor market regmes, along wth those assocated wth the lmtng cases of a completely centralzed wage settng and perfectly compettve product markets, are summarzed n Table 1. 9 Note however that ths effect goes to zero as wage settng centralzaton σ goes to 1 (all wages are set by a sngle unon) or to 0 (all wages are set ndvdually). We dscuss ths below. 12

15 Table 1. Parameter effects on nflaton and unemployment n varous settngs * Parameter effects (a) Flexble wage regme ( Nash ) (b) Rgd wage regme ( Stackelberg ) (c) Sngle unon ( σ = 1) (d) Perfect competton (η = ) π U π U π U π U Monetary polcy conservatveness, φ Product market prce competton, η N N N N N Unon wage centralzaton, σ N N Unon real wage weght, γ * The table reports the sgn of the effect of ncreases n the parameters n the frst column on nflaton (π) and unemployment (U). N ndcates no effects. The outcomes assocated wth cases (c) and (d) are ndependent of the wage regmes. The frst row of Table 1 ndcates that the degree of central bank conservatveness always affects nflaton, but has capacty to affect unemployment (the real economy) only f a multplcty of uncoordnated unons pre-commt wages n an economy wth less than perfectly compettve product markets. The other rows show the sgns of the effects of product market competton, wage settng centralzaton, and the weght unons place upon real wage premums as compared to unemployment none of whch depend on the operatve wage regme. Rows 2 and 3 mply that ncreasng ether product market competton or wage settng coordnaton reduces both nflaton and unemployment n the both flexble and rgd wage regmes. Note that the sze of unons (the degree of centralzaton of wage settng) affects macroeconomc outcomes only f frms exert market power over product prces, that s when η <. Smlarly, the degree of market competton affects outcomes only f there s a multplcty of unons (less than full coordnaton of wage settng), σ < 1. Absent ether a multplcty of unons or mperfect product competton, monetary polcy s neutral. Row 4 reflects the usual consequences of unon emphass on real wages under a dmnshng returns producton technology and a downward slopng demand for labor. The greater the weght placed on the (net of productvty) real wage relatve to unemployment, the hgher are both unemployment and nflaton n all types of labor and product markets. The overall 13

16 pattern of results delvers qute standard nferences: A low unemployment and low nflaton macroeconomy are promoted by conservatve, ant-nflaton central bankers, hghly compettve product markets, and hghly coordnated unons that place relatvely small weght on real wage premums as compared to unemployment. The more llumnatng ssue of whether rgd or flexble wage labor markets delver better macroeconomc performance depends on the central bank s nflaton averson. It s clear from prevous analyses that comparatve statcs of the parametrc results are cumbersome, and so the topc s more nformatvely addressed by graphcal analyss of smulated outcomes. In order to evaluate the quanttatve mpact of monetary polcy under dfferent wage formaton regmes we need to pn down values of some parameters, and to lmt the range of others. 5.1 Baselne parameter settngs Labor s share of ncome n developed economes generally les n the vcnty of two-thrds of GDP, so we set the productvty parameter to α = 23n all smulaton experments. 10 Prce elastctes of demand, even for sectors producng relatvely elastc goods and servces, are n general unlkely to exceed 1.5, and only nfrequently to exceed 2.0. We use a baselne settng for η = 1.3 n some smulatons, and nvestgate the effects of values spannng 1.1 to 2.0 n another. 11 Wage formaton n developed European market economes s typcally domnated by 2 to 3 large unons, 12 whch mples a representatve value of σ = 0.4 for the wage settng centralzaton parameter. 13 A labor force organzed by one all-encompassng unon (or a degree of nter-unon coordnaton that amounts to the same thng) has never exsted n a democratc captalst economy, and lkely never wll. Even durng the heydays of soldarty barganng n Scandnava there were two prncpal peak unons whose wage polces were only partly synchronzed. In smulatons nvestgatng the quanttatve mpact of wage centralzaton we analyze values of σ spannng 0.20 to 0.8; the hghest settng s taken to 10 Note however that α can be equated to labor s share only under perfect competton. When frms have market power, the share tends to understate the output elastcty of labor nput. For our purposes the baselne value adopted suffces however. 11 As shown n the Appendx (and as n any case s well known), the optmal markup of prce over margnal cost 1 1 among frms wth market power s 1. Calbrated analyses by Gord (1995) produced markups n the η range 2.2 to 4.5, whch correspond to values of η spannng 1.3 to 1.8, whch are covered by our settngs. 12 Golden, Lange and Wallersten (2002). 13 However see the remarks ahead about effectve centralzaton under a monetary unon. 14

17 represent the effectve degree of wage settng coordnaton n a labor market wth two, less than fully coordnated peak unons. Recent econometrc estmates of average equlbrum unemployment rates n European economes under the polces, labor market nsttutons and product market condtons prevalng n the late 1990 s fall n a farly broad range between 5% and 8.5%. 14 For that range of emprcal estmates, calculatons based on our parameterzatons of equlbrum unemployment mply that the weght placed by unons on the net-of-productvty real wage premum lkely falls n the nterval γ [ 0.1, 0.15], gven a labor nput elastcty n the vcnty of α = 23. We study the effects of a somewhat broader range γ [ 0.075, 0.175], but we use γ = (that s, a targeted real wage premum of 12.5%) as the baselne settng n smulatons fxng the value of ths parameter. Central bank behavor, on the other hand, appears emprcally to have spanned far more of ts theoretcal doman than other varable parameters n the model. At varous tmes durng pre- EMU perod n Europe, the monetary polces of Banca d Itala probably came closer than the polces of other monetary authortes n developed economes to ultra-lberal, near full accommodaton of wage nflaton, represented n the model by β 0 and φ At the opposte end of the spectrum, the Deutsche Bundesbank generally pursued very restrctve polces, as has the European Central Bank snce the monetary unon. In these cases, the posture of polces was probably not been too far from the ultra-conservatve lmt β =, ( 1 ) φ = α α. 5.2 Model smulatons Fgure 1 shows graphs of nflaton and unemployment outcomes n flexble and rgd wage labor market regmes n relaton to the full range of monetary polcy conservatveness when other parameters are set to baselne values. The model smulatons ndcate that moderately conservatve monetary polces yeld low sngle-dgt nflaton rates and unemployment rates n the 6-8% range, as opposed low double-dgt nflaton and unemployment rates of 8-10% under moderately lberal monetary polces. The smulatons delver quanttatve results broadly consstent wth experence n EU economes, whch mples that the baselne parameter settngs are credble. 14 See, for example, Douven (2002) and Logeay and Tober (2003) 15 The lberal lmts proper are ruled out because at β = 0, φ = 1 and nflaton goes to nfnty. See equatons (17) and (21). 15

18 Fgure 1. Inflaton and Unemployment n Flexble and Rgd Wage Regmes as Monetary Polcy Conservatveness Increases * (1a) * Percentage nflaton rates and percentage pont rates of unemployment are log approxmatons and were generated by parameter settngs α = 23, γ = 0.125, σ = 0.4 and η = 1.3 as φ goes from ts lower bound, 1, to ts upper lmt, α ( 1 α) = 2. The ultra-lberal lower bound of conservatveness s truncated n 1(a) because π as φ 1. (1b) We know from equatons (17) and (21) that n both flexble and rgd wage regmes nflaton declnes wth central bank conservatveness φ. Analyss of those equatons shows that when monetary polcy s lberal or populst ( φ < 0 ), nflaton s lower n flexble wage than n a rgd wage regmes. However, f polcy s conservatve ( φ > 0), the opposte s true and nflaton s lower n rgd regmes than n flexble ones. But the quanttatve dfferences are small to vanshng for plausble parameter settngs, and they cannot be detected at all from the nflaton graph lnes n Fgure 1(a) where polcy s n the moderately lberal to ultraconservatve range ( φ >.5 ). When polcy s passve and therefore does not respond to nomnal wage developments ( φ = 0 ), nflaton naturally converges across wage formaton regmes. Inflaton outcomes across regmes also converge to zero when polcy approaches the ultra-conservatve maxmum φ α ( 1 α) =. Equaton (18) showed that unemployment s unresponsve to systematc monetary polcy n flexble wage regmes, whereas eq. (22) mples that n rgd wage regmes unemployment s a decreasng convex functon of central bank conservatveness. Analyss readly demonstrates that f monetary polcy s lberal, unemployment equlbrums are always hgher n Stackelberg, rgd wage envronments than n Nash, flexble wage ones. When monetary s 16

19 conservatve, the pattern s reversed and rgd wage systems yeld superor unemployment (employment) outcomes. Moreover, by contrast to the nflaton outcomes, the dfferences become large as monetary polcy devates sgnfcantly n ether a lberal or conservatve drecton from the passve polcy posture of φ = 0. Fgure 1(b) depcts the pattern of results scaled to baselne values of model parameters. The explanatons for the patterns n Fgure 1 can be traced to the monetary transmsson mechansms and assocated ncentves of unons to pursue ambtous wage polces n dfferent nsttutonal settngs. Under monopolstc prce competton and a multplcty of atomstc unons, the contrbuton of unon-specfc nomnal wages, w, to the general prce level, p, s small by comparson to ts mpact on frm-specfc product prces, p j. Each uncoordnated unon ratonally explots the wedge between log real consumpton wage receved by ts members, ( w p) ( w pj), and the log real product wage faced by ndvdual frms,, by ncreasng nomnal wages n order to acheve hgher real wages, dscountng the mpact of ts behavor on the general prce level and, therefore, on the real wages of other unons. Snce all unons behave n lke fashon, the ensung ncrease to the general prce level (eq. 6) reduces aggregate demand va the negatve effect on real money balances (eqs. 1, A.11), whch n turn pushes up unemployment (eq. 5). In a flexble wage Nash regme, unons nteract smultaneously wth the central bank, and take the money supply, along wth the wages of other unons, as gven. Unon nomnal wage polces are therefore unaffected by systematc monetary polcy reactons (eq. 15). Even a Draconan ant-nflaton polcy s unable to overcome any of the unemployment costs created by the wage behavor of uncoordnated ndvdual trade unons (eq. 18) pursung best-reply Nash strateges. By contrast n a rgd wage Stackelberg regme, unons nternalze reactons of the central bank (eq. 11), and ths affects ther strategc calculatons. The unemployment effects of unon wage polces are ether aggravated or mtgated, dependng on the posture of polcy. Each unon n a rgd wage labor market takes account of the fact that a lberal central bank reacts to nomnal wage rses wth an accommodatng expanson of the money supply ( φ < 0) that yelds hgher nflaton for gven a wage ncrease (eq. 21). Unons nternalze the antcpated hgher prce level, whch rases the optmal nomnal wage consstent wth ther real wage-unemployment objectves (eqs. 13, 19-20), thereby magnfyng both the nflaton bas and aggregate unemployment costs assocated wth decentralzed wage formaton (eqs. 17

20 21-22). Under lberal monetary polces both economy-wde nflaton (eq. 6) and unemployment (eq. 7) are therefore hgher when nomnal wages are rgd as compared to beng flexble, as depcted n Fgure 1. Analogously, unons n rgd wage regmes nternalze the fact that a conservatve central bank reacts to nomnal wage ncreases wth a non-accommodatng contracton of the money supply ( φ > 0). Monetary contractons negatvely affect nflaton (the general prce level) and ths rases the real wage premum and the ensung unemployment nduced by a gven ncrease to the nomnal wage. As a result the real wage satsfyng every unon s optmalty condton can be acheved wth smaller nomnal wage ncreases (eq. 19). Hence by assurng some degree of prce stablty (a low nflaton bas), the central bank partly allevates the negatve employment externalty arsng from decentralzed wage formaton. Consequently, both nflaton and unemployment are lower than when wages are set ndependently of antcpated polcy reactons. Intuton s deepened by evaluaton of each unon s optmal condton for nomnal wage settng gven by equaton (13). The frst order condton γ p U 1 U = 0 mples that w w unemployment of unon s members s proportonal to the mplct margnal rate of substtuton between the real wage premum and unemployment embedded n the unon s p U preference functon: U γ 1. In flexble wage regmes, optmal wage settng w w and, consequently, unemployment among unon s members s polcy-ndependent (ndependent of φ ) (24) ( 1 p w ) ( 1 ) = U w γ ασ γ w flex ( η ασ η 1 ) ( α η ( 1 α )) +. However n rgd wage regmes wage settng t s polcy-dependent (25) γ ( ασ σφ α ) γ ( ( )) ( 1 p w ) 1 + ( 1 ) = U w ( 1) 1 η ασ η α + η α + σφ w rgd and so the unemployment rate each unon falls as the restrctveness of monetary polcy ncreases γ ( 1 p w ) U w w rgd φ < 0. Moreover, n a rgd wage labor market each unon s 18

21 margnal rate of substtuton s (1) greater than ts margnal rate of substtuton would be n a flexble wage market when φ < 0, (2) equal to t when φ = 0, and (3) less than ts margnal rate of substtuton would be n a flexble wage market when φ > 0. Snce all unons behave symmetrcally, we obtan the macroeconomc outcomes assocated wth rgd and flexble wage regmes descrbed above and graphed n Fgure 1. Fgures 2-4 graph unemployment outcomes n rgd and flexble wage regmes n relaton to the full range of central bank conservatveness and constraned varaton n the remanng model parameters σ, η and γ. We focus on unemployment outcomes alone because the responses of nflaton to plausble varatons of all model parameters do not yeld szeable dfferences across wage settng regmes. As n the baselne results graphed n Fgure 1, t s on the real sde of the economy represented n our model by the unemployment rate where bg dfferences n monetary effects emerge n rgd as compared to flexble wage labor markets. The results graphed n Fgure 1(b), as well as the results of the prevous comparatve statcs, mpled that the nfluence of monetary polcy on relatve unemployment outcomes orgnates wth the way that systematc central bank reactons to wage changes affect unons strategc calculatons n rgd wage labor markets. It s clear from Fgures 2-4 that under almost all reasonable varatons of wage settng centralzaton, product prce competton, and unon real wage orentaton, the effects actvst monetary polces on unemployment performance n rgd wage regmes are several percentage ponts n magntude, even when the monetary polcy reacton functon m= φw s evaluated wthn a truncated range that lkely corresponds to nearly all emprcal experence, φ [ 0.5, 1.5]. 16 Fgure 2 shows the response of unemployment rates to monetary polcy as unon centralzaton vares over emprcally relevant values. When unons are bg and therefore set wages for a large fracton of the labor force, they understand that ther wage polces have great mpact on the general prce level and, consequently, that the wedge between product real wages and consumpton real wages opened up by nomnal wage ncreases s correspondngly small. Hence, n both rgd and flexble wage envronments, large unons correctly perceve that the real wage return to aggressve nomnal wage polces s comparatvely weak, especally n relaton to the utlty cost of wage-nduced ncreases to 16 At α = 23, the correspondng range of the central bank s utlty weght on nflaton as compared to unemployment s β [ 1.8, 45]. 19

22 unemployment among unon members (eqs. 8-9). Unon sze (centralzaton or coordnaton of wage settng) therefore tempers optmal wage aspratons, yeldng lower frm-level prce rses and lower unon-level unemployment as centralzaton of wage settng ncreases the rght-sdes of equatons (24) and (25) both declne as σ rses. Consequently, as shown analytcally n secton 4 by the comparatve statcs for equatons (17)-(18) and (21)-(22), economy-wde rates of nflaton and unemployment fall n both wage settng regmes wth greater centralzaton of nomnal wage formaton. In rgd wage systems, however, the effects of unon centralzaton nteract wth the restrctveness of monetary polcy, and ths can create large dfferences n unemployment outcomes across wage settng regmes. At any gven degree of centralzaton, the nternalzaton of lberal polces by unons that pre-commt nomnal wages, ratonally leads to more aggressve wage polces as monetary polcy becomes more accommodatng, and ths yelds rates of nflaton and unemployment that always exceed the correspondng rates n flexble wage regmes where the money supply s taken as gven. The otherwse benefcal effects of wage settng centralzaton are eroded completely as polcy approaches the ultralberal lmt φ = 1, whch s depcted by the convergng graph lnes n Fgure 2 for unemployment outcomes n rgd wage labor markets at large negatve values of φ. Fgure 2. The Effects of Monetary Polcy Conservatveness on Unemployment by Varaton n Centralzaton of Wage Settng * * Percentage pont rates of unemployment are log approxmatons and were generated by parameter settngsα = 23, γ = and η = 1.3 as φ goes from ts lower bound, 1, to ts upper lmt, α ( 1 α) = 2 for varous degrees of wage settng centralzaton, σ. Rgd wage outcomes are depcted by the convex graph lnes; flexble wage outcomes by the parallel lnes. 20

23 The reverse s true n rgd wage labor markets when polcy s conservatve and the central bank contracts the money supply n proporton to nomnal wage ncreases. The bgger (or more coordnated) are unons, the greater s the nternalzaton of ant-nflaton monetary polces, and the lower are rates of unemployment (and nflaton) by comparson to outcomes n flexble wage labor markets, as Fgure 2 shows. However, central bank conservatveness s only a second best way of solvng coordnaton problems among atomstc unons. The lowest rate of unemployment attanable by an ultra-conservatve central bank operatng n a rgd ( 1 ) wage regme s U η α + = α ( 1 α ) γ, whch exceeds the sngle unon-perfect η 1 α + ασ competton mnmum U ( 1 ) = α γ by a factor that declnes as wage settng becomes more centralzed. At the lmt, when all wages are set by one all-encompassng unon that s, as σ 1 or, equvalently, as the wage behavor of notonally ndependent unons becomes perfectly coordnated whch amounts to the same thng monetary polcy no longer affects unon wage polces because a sngle unon fully nternalzes on ts own the macroeconomc consequences of ts nomnal wage behavor. Consequently monetary polcy s neutral. 17 (See row 1, column c of Table 1) However, as mentoned before, complete centralzaton of wage settng has never been observed n a market economy. Note that the establshment of a monetary unon notably the EMU effectvely decreases wage settng centralzaton because the nomnal wage rses obtaned by even the largest natonal unons have comparatvely small effect on the unon-wde wage and prce levels and, consequently, have correspondngly small nfluence on unon-wde monetary polcy. Ceters parbus, monetary unon therefore tends to rase equlbrum unemployment, 18 unless there are offsettng changes elsewhere n the macro poltcal economy. One offsettng change could be and, n fact, has been n the case of the EMU a shft to more conservatve monetary polcy (bgger φ ) facng unons of the typcal member naton. Another s a parallel ncrease 17 Another way to thnk about ths, whch orgnates wth an nsght of an anonymous revewer, s that the 1 presence of a multplcty ( n ) of uncoordnated unons defned by σ = < 1 amounts also to sayng that n each unon behaves as f t were the only Stackelberg leader. Alternatvely, f each unon were to nternalze the behavor of the others n ts wage settng behavor, polcy would agan be neutral. As noted n the man text, such behavor defnes perfect wage coordnaton and s functonally equvalent to σ =1 (n=1). Note too that polcy becomes neutral as σ 0 ( n L ), that s, when the wage settng s completely atomzed so that the mpact of any ndvdual wage on economy-wde aggregates s neglgble. 18 Ths potental effect of monetary unon appears to have been dentfed frst by Hall (1994), and was subsequently dscussed by Hall and Franzese (1998), Soskce and Iversen (1998) and Cukerman and Lpp (2001). 21

24 n market competton (bgger η ), whch also has been an ntegral feature of the deepenng of the European Communty project. 19 Fgure 3 shows that the drect mpact of product market prce competton on unemployment 20 and on the effcacy of restrctve polcy n drvng down unemployment are very bg at plausble settngs of model parameters. In rgd wage labor markets, where monetary polcy has capacty to nfluence unon wage clams, unemployment s agan exacerbated f polcy s populst and s mproved f polcy s conservatve. However note that polcy exerts more potent effects on unemployment as product markets become less compettve the convex graph lnes get steeper as η decreases. The reason s that relatvely small values of η (less market competton) yeld relatvely hgh monopoly rents to frms and, therefore, hgher wages to unonzed workers, whch n turn rases unemployment. 21 Fgure 3. The Effects of Monetary Polcy Conservatveness on Unemployment by Varaton n Product Market Competton * * Percentage pont rates of unemployment are log approxmatons and were generated by parameter settngsα = 23, γ = and σ = 0.4 as φ goes from ts lower bound, 1, to ts upper lmt, α ( 1 α) = 2 for varous elastctes of prce competton, η. Rgd wage outcomes are depcted by the convex graph lnes; flexble wage outcomes by the parallel lnes. Polcy looses potency as markets become more compettve. Although we consder a relatve prce elastcty of product demand of 2.0 to be a reasonable upper lmt, f competton were to become much fercer, for example because of government polces that dramatcally 19 Holden (2005) analyzes the ncentves that monetary unon may create for natonal unons to ncrease wage coordnaton, whch n our model means rasng the effectve magntude of σ. 20 Raw emprcal data reported n OECD (2003) show large, monotonc responses of employment to product market lberalzaton between 1978 and 1998 n 21 countres. 21 For extended analyss see Nckell (1999). 22

25 reduced barrers to trade and market entry, 22 monetary polcy would become rrelevant to unemployment performance n rgd wage regmes, where otherwse t exerts great effects. Frms chargng non-compettve prces would be drven to bankruptcy and labor would be reallocated to survvng producers, no matter what the wage nflaton posture of the central bank. The ensung unformty of product prces would elmnate the wage wedge, thereby nullfyng the ncentve of unons to pursue wage polces explotng a gap between the real consumpton wage and the real product wage. At the perfect competton lmt η =, unemployment falls to U ( 1 ) = α γ at all values of φ. Fgure 4 graphs smulatons of the unemployment effects of varatons n the weght unons attach to real wage premums. γ s a parameter that dstorts compettve outcomes n labor markets, n that t represents the wllngness of unons to mpose wage levels exceedng the underlyng productvty of the (gven) labor force, thus drvng unemployment above the perfect competton equlbrum U = An mportant message of Fgure 4 s that unemployment outcomes are very senstve to unon real wage objectves. Across plausble settngs for γ, whch n our smulatons mply unon real wage goals that range from 7.5 to 17.5 percentage ponts above the market clearng level, equlbrum unemployment rates shft by as much as 7 percentage ponts. γ s drectly analogous to the status of η n product markets, and t nteracts wth monetary polcy conservatveness n much the same fashon as η. 24 Hence monetary polcy exerts greatest effects when unon goals are most dstortng that s, at the larger values of γ. As γ gets small, unon wage behavor corresponds more on ts own to compettve behavor and the condtonng effects of polcy dsspate, just as s the case when the product prce competton parameter η gets large. In prncple γ should be seen as a behavoral parameter, rather than a deep structural parameter lke α. Hence the relatve weght unons attach to real wages as opposed to unemployment n the determnaton of ther nomnal wage polces s subject to planned adjustment, although we make no attempt n ths paper to model the nternal dynamcs of unons that mght help explan revsons of ther utlty programs. 22 See, for example, the analyss of Blanchard and Gavazz (2003). 23 Recall that productvty growth and the equlbrum rate of unemployment under pure competton are both normalzed to zero n our model. 24 Blanchard and Phlppon (2003) analyze a model n whch the unon utlty weght on the real wage level declnes wth unon belefs about the degree of product market competton. In our settng that would make γ proportonal to η, rather than beng an ndependent parameter wth separable effects. 23