Monetary Integration, Partisanship, and. Macroeconomic Policy

Transcription

1 Monetary Integraton, Partsanshp, and Macroeconomc Polcy Torben Iversen Department of Government Harvard Unversty and Davd Soskce Wssenschaftszentrum Berln Paper prepared for presentaton at the 95th Amercan Poltcal Assocaton Meetng at the Atlanta Hlton and the Marrott Marqus, September 2-5, 1999.

2 Abstract Most of the current lterature on macroeconomc polces focus on cyclcal phenomena such as poltcally nduced busness cycles, and how short-sghted polces can be overcome through proper nsttutonal desgn. Ths paper nstead focuses on the lastng effects of macroeconomc polces on such real economc varables as unemployment and dstrbuton, and we argue that tradtonal dstrbutve poltcs s mportant for understandng the choce over polces and nsttutons. We derve our hypotheses from a ratonal expectatons model of the choce over macroeconomc polces, wage barganng nsttutons, and unon wage strateges, and we present evdence from Europe that corroborates these hypotheses. The model helps us understand not only cross-natonal dfferences n macroeconomc polces, nsttutons and performance, but also the effects on these varables of nternatonal monetary ntegraton.

3 I. Introducton Durng the past two decades very lttle work n poltcal scence has been produced on the long-term effects of macroeconomc polces on ncome dstrbuton and unemployment, or the role of partsan poltcs n the choce over these polces. Presumably ths omsson s justfed by reference to the new classcal concept of polcy neutralty -- the thess that fscal and monetary polces cannot have any lastng effects on the real economy. Polcy neutralty s n turn thought to rest on the wdely accepted noton that people have ratonal expectatons. In ths paper we show that both vews are ncorrect, and that there are strong reasons to expect macroeconomc polces, and the nsttutons that support them, to be shaped by partsan conflcts over the real economy. Understandng the effects of macroeconomc polcy-makng on the real economy was once an ntegral part of comparatve poltcal economy. Most notably, Hbbs semnal 1977 artcle on poltcal partes and macroeconomc polces generated countless studes on the role of partsanshp n explanng macroeconomc polces and outcomes. Hbbs argued that conflctng dstrbutve nterests of the core consttuences of left and rght partes would lead these to choose dfferent ponts on the Phllps curve trade-off between unemployment and nflaton. But whle Hbbs argument about the dstrbutve nterests of partes survved, the noton of an explotable long-term Phllps curve dd not. As shown by Fredman (1968), usng an adaptve expectatons framework, and by Lucas (1972), usng a ratonal expectatons framework, the Phllps curve was not consstent wth ratonal behavor. And wthout ths trade-off, Hbbs poltcal explanaton for endurng cross-natonal dfferences n polces and outcomes had no foundaton n economcs. The only remnants of Hbbs argument are n Alesna s ratonal partsan busness cycle model, whch hypotheszes that left and rght governments create dfferent post-electon polcy cycles, whle leavng the real economy unaffected (Alesna 1988; Alesna, Cohen and Roubn 1992). Consstent wth Alesna s approach, as well as wth non-partsan poltcal busness cycle models, much of the lterature turned to the queston of how to desgn macroeconomc 1

4 nsttutons that would overcome short-term ncentves by governments to nflate (the tme nconsstency problem). For example, a key queston became how to wrte a central bank contract that would bnd the government to a non-nflatonary polcy rule wthout smultaneously undermnng ts ablty flexbly to adapt to the busness cycle (see Rogoff 1985 and Lohmann 1992). Ths paper nstead revsts Hbbs orgnal nsght that the choce over macroeconomc regmes s motvated by partsan preferences and have lastng effects on real outcomes. Contrary to conventonal wsdom, we show that the polcy-neutralty thess does not follow from ratonal expectatons, but nstead from the neo-classcal assumpton that labor markets are atomstc. Once we allow for the possblty of large prce and wage setters, or coordnated wage barganng, fscal and monetary rules affect unons real wage demands as well as employers support for centralzed and redstrbutve wage barganng. Because coordnated barganng s pervasve n Europe, and plays a role n most other regons of the world (see OECD 1997), ths s an mportant nsght for poltcal economy. The key to our argument s that the employment costs of mltant unon behavor, and hence the ncentve to behave restraned, depends on the extent to whch macroeconomc polces are accommodatng. If macroeconomc authortes keep nomnal spendng and the money supply constant, the wage demands of large ndvdual unons wll have an effect of real demand and hence employment, thereby furnshng unons wth an ncentve to behave restraned. By contrast, f the government seeks to keep real spendng and the real money supply constant, no ndvdual unon can affect demand or employment and wll consequently have lttle ncentve to exercse restrant. In ths stuaton, the only way for employers to control wage costs may be to centralze wage barganng to the peak level, although such centralzaton typcally comes at the expense of severe constrants on frms ablty to shape the wage structure. Our model also helps us understand the consequences of nternatonal monetary ntegraton. In a natonal economy wth an autonomous central bank, large ndvdual unons can 2

5 have sgnfcantly affect monetary polcy f the central bank rases nterest rates n response to nflatonary wage clams. Ratonally antcpatng ths, unons have an ncentve to nternalze the macroeconomc costs of mltancy. But when nterest rates are set for a whole regon by a sngle central bank, the wage demands of any partcular natonal unon wll not have much of an effect on these rates. The costs of mltancy can therefore be externalzed, and wage restrant and employment wll fall. Ths logc helps us understand why European labor markets are no longer functonng as well as they once dd, and why equlbrum unemployment n Europe has been rsng. But contrary to economsts who predct that monetary unon wll force governments to deregulate labor markets, we argue that the sngle currency wll restore fscal polcy autonomy and possbly lead to a prolferaton of socal pacts between governments and unons. Gven the contnued strength of unons, t s also poltcally more realstc to expect the formaton of new forms of cross-natonal wage coordnaton than wholesale labor market deregulaton. The paper s organzed nto three man sectons. Secton II consders the nteracton of monetary, fscal, and wage polces n a before captal moblty world where governments enjoy fscal and monetary polcy autonomy. As we show theoretcally, and llustrate emprcally, ths s a world where partsan poltcs s mportant and where countres cluster nto a small set of dstnct natonal models. Secton III dscusses what happens when economes ntegrate nto a system of fxed-but-adjustable exchange rates wth complete captal moblty. We explan why such ntegraton have deleterous effects on employment and undermnes natonal varaton n nsttutons and polces. Fnally, secton IV consders the lkely effects on macroeconomc coordnaton n a monetary unon. Whle problematc for wage restrant and employment n the short run, such a unon could regnte dstnct natonal varetes of coordnaton and/or produce new forms of nternatonal coordnaton. Throughout we llustrate the key mplcatons of the theoretcal model wth data from Europe. 3

6 II. Natonal Varetes n a Before Captal Moblty World In Mundell-Flemng s well-known open-economy model t s only possble for governments to combne at most two of the followng three objectves: monetary autonomy, captal moblty and fxed exchange rates. For our purposes, the model provdes a useful typology of nternatonal monetary regmes, wth two deal types and one hybrd. The frst deal type refer to a stuaton wth closed natonal captal markets, or what Freden (1991) calls a before captal moblty (BCM) world. It does not matter for our argument whether exchange rates are fxed or flexble; the key s that the government enjoys monetary polcy autonomy. In the European context ths stuaton characterzed the perod from the Second World War untl about the late 1970s or early 1980s. The second deal type s an after captal moblty (ACM) world wth rrevocably fxed exchange rates. In fact, there are two sub-speces of ths regme type: one where monetary polcy s determned by a natonal central bank (CB) targetng domestc prces, and one where a transnatonal CB targets prces n the whole currency area. The European Economc and Monetary Unon (EMU), and the European Central Bank (ECB), are an example of the latter. The former stuaton s better exemplfed by the Exchange Rate Mechansm (ERM) nsde the European Monetary System, wth the German CB settng monetary polcy for all members. But lke vrtually all real-lfe examples of a fxed exchange rate systems, the ERM was n fact a system of captal moblty wth fxed-but-adjustable exchange rates. If exchange rates are not rrevocably fxed t affects the analyss of fscal polcy as we dscuss. Consequently, we consder ths hybrd as a separate type of nternatonal monetary system. Wth these dstnctons n mnd, the extensve form game n Fgure 1 sets out the structure of our argument for the BCM world. As we swtch the analyss to the other two regme types, we wll show how the equlbrum of the game changes. [Fgure 1 about here] 4

7 The game has three prncpal actors -- the government, employers, and unons and we focus on two dfferent outcomes: unemployment and wage equalty. 1 Both outcomes are central to comparatve poltcal economy, and of obvous nterest to governments and labor market actors alke. In the frst stage of the game, the government commts to a partcular monetary and fscal rule that s then mplemented (by a central bank n the case of monetary polcy). The monetary rule s gven by: (1) M = P 1 β where M s the nomnal money supply, P s the aggregate prce level, and â = [0,1] s a parameter measurng the degree to whch the CB follows a non-accommodatng polcy rule. Note that f the CB rule s completely accommodatng, â = 0, the CB fxes the real money supply by settng M equal to the prce level, whereas f the CB s completely non-accommodatng t fxes the nomnal money supply and sets M equal to unty. 2 The fscal rule s defned n an analogous fashon: (2) G = P 1 β F, where G s nomnal government spendng, and â F = [0,1] s the fscal rule. If â F =0 t means that the government holds real expendtures constant, correspondng to an accommodatng rule; f â F =1 t means that the government fxes nomnal spendng, correspondng to a nonaccommodatng rule. For presentatonal ease Fgure 1 assumes that governments pursue consstent monetary and fscal polces. Although there s nothng nherent to our model that requres monetary and fscal polces to move n tandem, ths s n fact what we tend to observe n a BCM world. There are two prncpal reasons for ths. If the government controls both monetary and fscal polcy t has an ncentve to pursue consstent polces n order to maxmze ther effect. On the other hand, f the mplementaton of a non-accommodatng polcy rule s delegated to an ndependent central 5

8 bank, as s typcally the case, then the CB can undermne accommodatng fscal polces by rasng nterest rates (Scharpf 1991; Tabelln 1990). Ths gves the government an ncentve not to engage n accommodatng polces. Once the government s commtted to a partcular set of fscal and monetary rules, unons and employers bargan wages. We dstngush between two types of barganng systems: coordnated and uncoordnated. In uncoordnated systems wages are set at the plant or frm level, and there s no or lttle capacty for coordnaton of wages across frms or ndustres. In coordnated systems, by contrast, wages are barganed collectvely at the ndustry- or peak-level. Furthermore, we assume that wage barganers n coordnated systems, unlke barganers n uncoordnated ones, have the capacty credbly to commt to centralzed barganng provded that employers (or ther organzatonal representatves) are wllng to go along. If not, barganng wll occur at the ndustry (or sectoral) level. The reason that we need to nclude a separate note for employers choce over centralzaton s that centralzed barganng may or may not be n ther nterest. The benefts of centralzed barganng are descrbed well n the neo-corporatst lterature, and concerns the capacty of employers to control average wage costs (e.g., Cameron 1984; Lange 1984; Crouch 1985). But s t s now also wdely acknowledged that there may be costs to employers of centralzaton. In partcular, centralzaton can nterfere wth employers ablty to desgn a wage structure that maxmze shop-floor cooperaton and employee nvestment n sklls. As more and more wages are subjected to the same collectve agreement, low wage unons are lkely to gan n barganng power and call for a more egaltaran dstrbuton of wages (Iversen 1996; Pontusson and Swenson 1996). Ths makes t harder for employers to create a desred wage dfferentaton between sklled and unsklled workers. 3 For ths reason, whether employers wll consent to peaklevel barganng depends on whether ther ncreased capacty to control wage costs wll outwegh ther reduced capacty to control the wage structure. 6

9 At the last stage of the game monopoly unons choose a wage strategy that can be ether mltant (m) or restraned (r). Strctly speakng, the game does not end here because once wages have been barganed, frms must set prces, and the macroeconomc authortes must then choose a partcular fscal and monetary polcy. Two smplfyng assumptons, however, allows us to effectvely treat the choce of unons as the last stage n the game. The frst assumpton s that frms behave as Bertrand compettors and apply a constant markup to the money wage. The second assumpton s that the government, or ts agent (the CB), s credbly commtted to mplementng ts polcy rules. The frst assumpton s techncal and for convenence only. Other prcng rules could be assumed wthout affectng our results. The second assumpton s substantve and mples that how the government can credble commt to a partcular rule s exogenous to the model. Whle the topc of commtment s mportant, and has attracted much attenton n the central bank ndependence lterature, t s not one that we address n ths paper. What matters to our argument s that governments can and do adhere dfferent polcy rules, not how these rules are nsttutonalzed. In determnng unons choce of wage strategy we assume that the economy has a fxed number of dentcal sectors,, and that wages are barganed smultaneously and ndependently by one monopoly unon n each sector. 4 The th unon s concerned wth real wages and employment among ts members and maxmzes the followng utlty functon: (3) U = w α e, where w a s the average real wage rate for unon s members, and e s the employment rate among unon s members. The margnal utlty of hgher wages s assumed to be declnng n w (0<á<1). Fnally, we defne the followng aggregate demand functon for sector : (4) q = e = m / N η p = m / N η w, 7

10 where m s the aggregate real money supply, and p = P /P (P s the prce n the th sector, and P s the aggregate prce level for the economy). Snce W = P wth Bertrand prcng, and normalzng labor supply to unty, ths expresson mples the standard trade-off between the employment rate, e, and the real wage, w. 5 Wth these assumptons n mnd, the game can now be solved through backward nducton. To derve the equlbrum real wage and employment rate we fnd the maxmum of unon s utlty functon subject to the demand functon n (4). Substtutng (4) nto (3), dfferentatng by w, and rearrangng gves (2) α m m [ m N η w ] + = η w. / N m w w The bracketed expresson s equal to e, and m/n can be expressed n terms of e and w usng (4): (3) α e + ln m ) ln w ( e + η w = η w. Ths gves the followng expresson for w : (4) w ln m α + α ln w = ln m η 1 ln w e. ln m The term s crucal, and we need to evaluate t: ln w (8) ln m ln w = ln m ln P ln ln P P ln P lnw lnw ln P ln W 1. 8

11 = 1/ N P Gven the monetary rule defned n (1), and gven a geometrc prce ndex, equaton (8) can be reformulated as P (9) ln ln m w β = N 1 1 N 1 β =. N 1 Substtutng (9) nto (7) yelds the followng expresson for the real wage chosen by unon : (10) w β α = N 1 β η 1 + N 1 e. We can now derve the equlbrum level of employment snce the real wage must be equal to unty n equlbrum: 6 (11) β η * N e =. β α N 1 Equatons (10) and (11) gve the basc results for our model of monetary polcy. 7 Two obvous and standard results follow from equaton (11): The less the mportance attached by the unon to ncreased real wages (.e., the lower s á), the greater wll be the equlbrum employment rate; smlarly, the more elastc s product market demand (ç) the greater wll be the equlbrum employment rate. The theoretcally nterestng results, however, are those related to â (how nonaccommodatng the monetary rule s) and N (the number of unons). There are two specal cases. If N=1 the real wage s fxed at whatever money wage the encompassng unon chooses. The 9

12 unon then decdes on a money wage, say W* (and hence prce level) to secure the real money supply (hence aggregate demand) whch produces the unon s desred rate of employment, say e e *. 8 If N s very large, e*=ç/á, wth hgher unemployment than n the encompassng case. Although all unons know that f they were to coordnate ther barganng perfectly, they would choose the same money wage as the encompassng unon and get an employment rate of e e *, they are n a Prsoner s Dlemma: Each unon faces a trade-off between the real wage and employment, snce f the sectoral money wage, W, s ncreased, gven money wages n the other sectors, the sectoral real wage rses and the unon moves up the sectoral demand curve for labor. Ths makes t mpossble to sustan e e * n equlbrum a central result that s reflected n vrtually all neocorporatst nterpretatons of collectve wage barganng (see Lange 1984). Nether of these specal cases leaves any role for monetary polcy, whch s consstent wth both neo-corporatst and new classcal analyses. For N=1 and N=4 monetary rules are neutral wth respect to unemployment. But ths s not the case when N s fnte or small. Specfcally, as we go from a stuaton wth many unons to a stuaton wth few, the postve effect on employment s greater when the monetary rule s non-accommodatng than when t s accommodatng. The reason for ths central result s that when there s a small number of unons, whether each of these unons wll have an apprecable effect on the real money supply, and hence real demand and employment, depends on the monetary rule. The more non-accommodatng ths rule, the greater the effect of each unon s wage demands on the real money supply. And the more each unon s behavor affects employment, the greater the ncentve to be restraned. Exactly the same logc apples n the case of fscal polcy. To see ths we need to reformulate the aggregate demand functon so that t ncludes government spendng: (12) d q = a / N + g / N + ( c / N) q + η ηw, 10

13 where a s real autonomous demand (e.g., nvestment) n the economy as a whole (wth a/n demanded n each sector); g s real government spendng on goods and servces (wth g/n spent n each sector), and q d s dsposable ncome, out of whch a proporton c (propensty to consume) s consumed (wth 1/N spent n each sector). Assume now that g = τ where ô s real tax recepts. In other words, we assume that budgets are balanced (although the model can easly be modfed to allow for surpluses and defcts). Assume further that government revenues are rased from a poll tax to avod the complcatons of dstortonary taxaton. Now sum (12) over the N sectors and assume that w = N (an assumptons that s always true n equlbrum snce w = 1, as noted above). Gven that q d = q τ we can wrte the aggregate demand functon as (13) q = a + g + c( q g) = g + a / ( 1 c). ca a Notng that a / N + =, and rearrangng, we have the followng N ( 1 c) N ( 1 c) expresson for the sectoral demand functon: (14) e = q = a + g / N + η η w. N ( 1 c) As before, unon maxmzes ts utlty functon (3) subject to the sectoral demand functon, whch produces: (15) a α + g / N N (1 c) + η η w + g N w g g w = η w (Smlar to (5)), or, n terms of w : 11

14 (16) w α = + ( g / Nq ) η ln g ln w e. Usng an dentcal logc as n (7) and (8), we can now establsh that (17) ln g ln w F β =. N 1 g / N Substtutng (17) nto (16) and defnng σ g, --.e., the share of government q expendture n the output of sector we get (18) w α = σ g, η F β N 1 e where the degree of restrant depends nversely on the number of sectors and postvely on the share of government expendture n the sector. Fnally, the equlbrum employment rate s (19) η e* =. F β α σ g N 1 As n the case of monetary polcy, employment s rsng n â F and fallng n N, and the reason s exactly the same as before: Indvdual unons have a greater effect on real sectoral demand, and hence a greater ncentve to behave restraned, when the fscal rule s nonaccommodatng than when t s accommodatng. If we allowed for unbalanced budgets by holdng ô constant, there would be an even bgger ncentve for unons to be restraned snce the effect on q of a non-accommodatng polcy rule would then also operate through q d. 9 These predctons are borne out by data from Europe n the perod (see Table 1). Brtan, France are here treated as examples of economes wth uncoordnated wage barganng n 12

15 ths perod, whle those countres wdely regarded as havng coordnated barganng was dvded nto one group wth predomnantly ndustry-level barganng (Austra, Belgum, Germany, Netherlands, and Swtzerland) and one group wth predomnantly peak-level barganng (Fnland, Denmark, Norway, and Sweden). Fnally we splt the sample nto one half wth below-medan, and one half wth above-medan, scores on two varables desgned to measure fscal and monetary non-accommodaton (dscussed below). Although ths dvson s really only relevant for the coordnated cases, we retaned t across the cases for consstency. Note that unemployment s sgnfcantly lower n the coordnated than n the uncoordnated European economes, and that wage equalty ncreases wth centralzaton. The smple correlaton between centralzaton and OECD s measure of earnngs equalty (d1/d9 ratos) s 0.72 for all 12 countres and 0.90 f we exclude Austra (an oft-noted outler). Ths pattern s echoed n emprcal studes usng more sophstcated statstcal technques (see Rowthorn 1992, Wallersten 1999, and Rueda and Pontusson 1998). The relatonshp between coordnaton and unemployment also receves strong support n the emprcal lterature (see Layard et al. 1991; Hall and Franzese 1998; and Iversen 1999). [Table 1 about here] An mportant mplcaton of these results s that employers can elct the same degree of wage restrant from unons at a lower level of centralzaton when monetary and fscal polcy rules are non-accommodatng than when they are accommodatng. All else beng equal, employers therefore have less ncentve to agree to peak-level barganng when polces are nonaccommodatng than when they are accommodatng. 10 In the game tree (Fgure 1) we have llustrated ths by employers consentng to peak-level barganng only when rules are accommodatng; otherwse they wll prefer ndustry-level barganng. 13

16 Obvously, f we conceve of both centralzaton (1/N) and non-accommodaton (â) as contnuous varables, the predcton s smply that centralzaton wll be greater the more accommodatng the macroeconomc rules. Ths s precsely what we observe n Europe for those countres that are wdely consdered to be characterzed by coordnated barganng systems (see Fgure 2). 11 Monetary rules are proxed here by an ndex (standardzed to vary between 0 and 1) trackng relatve exchange rate movements (apprecaton equals non-accommodaton), but the pattern s the same f we nstead use an ndex of central bank ndependence (wdely consdered an ndcator for the conservatveness or non-accommodaton of the central bank). Centralzaton s an ndex of the predomnant level of barganng and the concentraton of unon membershp at that level (see Iversen 1999, ch. 3). Both varables are averages for the perod. The smple correlaton between the two varables s [Fgure 2 about here] The relatonshp s the same, and equally strong, f we consder fscal rules (r=-0.86). Fscal rules are here proxed by the nverse of government consumpton, and to see that ths s a sensble operatonalzaton of the theoretcal varable, note that â F n the formal model s gven by (20) β F α = η / e * 1 σ N 1 g 1 If we assume that á and ç do not vary across countres, and f we defne centralzaton of wage barganng as 1/N-1, then a good operatonal measure of fscal non-accommodaton would be: Fscal non-accommodaton = e/n -1 *ó g = (1-unemployment rate)/(centralzaton * spendng). 14

17 Snce numbers are readly avalable for these varables, t s straghtforward to measure fscal nonaccommodaton emprcally. In practce, however, t turns out that ths measure s hghly correlated wth government consumpton (0.94) (ó g ). Ths s fortunate because f we want to examne the relatonshp between fscal non-accommodaton and centralzaton, we need a measure of the former that does not contan the latter as a component. We have therefore smply used the followng measure for fscal non-accommodaton based on a standardzed government consumpton rates: 1 σ σ max g max g σ σ g mn g The relatonshp between non-accommodaton and centralzaton becomes even stronger f we use the mean of fscal and monetary non-accommodaton as a proxy for the overall orentaton of the macroeconomc regme. The only slghtly devant case s Norway, whch was able to rapdly expand government spendng wthout sgnfcantly ncreasng the share of spendng n GDP because of huge ol revenues. The smple correlaton s ncludng Norway, and excludng Norway. For coordnated market economes n Europe there s thus no queston that a close emprcal assocaton exsts between macroeconomc accommodaton and centralzaton. Havng consdered the choces of unons and employers, and the outcomes assocated wth these choces, we are now at the frst stage n the game: The choce of a monetary and fscal polcy rule. To explan ths choce we propose a very smple hypothess: It depends on the hstorcal strength of the poltcal left. If left partes are concerned wth equalty and the nterests of low-pad workers -- a core consttuency for the tradtonal left -- these partes should favor accommodatng polces. Because accommodatng monetary polces nduce employers to consent to more centralzed nsttutons, and because such nsttutons are assocated wth wage compresson, t s n the nterest left partes concerned wth the welfare of ther core consttuences to pursue such polces. In addton, government spendng on goods and servces whch are 15

18 closely lnked to our concepton of accommodatng fscal polces have drect redstrbutve effects because ctzens are gven equal access to free or low-prced servces. Such access wll be more valuable to low-ncome people than to hgh-ncome people. Agan, ths makes accommodatng fscal polces n the nterest of tradtonal left partes. By contrast, snce rght partes cater less to the nterests of low-pad workers, and snce they are presumably more concerned wth the nterests of employers, these partes are expected to pursue more nonaccommodatng monetary and fscal polces. To test ths smple conjecture we looked at the assocaton between the left government partsanshp and our measures for fscal and monetary non-accommodaton. Snce partsanshp s only lkely to have an apprecable effect on rules through a slow process of nsttuton- and reputaton-buldng, we have here used a hstorcal average ( ) of the partsan complexon of the government. 12 As expected there s a strong assocaton between partsanshp and nonaccommodatng polces. For fscal polcy the smple correlaton s For monetary polcy t s 0.86 The former result s confrmed by a number of detaled emprcal studes showng left governments to spend consderably more than rght ones on goods and servces (see Cusack 1997; Box 1998; Iversen and Wren 1998). The relatonshp for monetary polcy often acknowledged, but rarely explaned. From a new classcal perspectve, snce accommodatng rules leads to unnecessary nflaton only short-termsm could explan the adherence to such rules. 13 But left governments are scarcely more short-sghted than rght governments. [Fgure 3 about here] A systematc attempt to explan the role of partsan poltcs for macroeconomc polces s Hbbs semnal 1977 artcle whch argues that because of the poltcal left s greater concern for equalty and full employment, left governments wll pursue more accommodatng macroeconomc polces. However, Hbbs conclusons were derved under the assumpton of an explotable and 16

19 long-term trade-off between nflaton and unemployment (the Phllps curve); an assumpton that has been dscarded n modern economcs because t s ncompatble wth ratonal expectatons. Our model restores Hbbs basc nsght that partsanshp s mportant for explanng macroeconomc polcy-makng and outcomes, but does so under the assumpton of ratonal expectatons. The new classcal polcy neutralty thess s not mpled by the ratonal expectatons assumpton, as wdely beleved, but by the assumpton that labor markets are atomstc. Once we allow for strategc players n the labor market, whch s essental for any understandng of European economes, macroeconomc polcy makng has lastng effects on the real economy that are bound to dvde partsan poltcans. III. The perls of monetary ntegraton Consder now a system of open economes operatng nsde a system of fxed-butadjustable (FBA) exchange rates wth complete captal moblty (see Fgure 4). Ths approxmates the stuaton for members of the Exchange Rate mechansm of the European Monetary System from the md-1980s untl the launch of the euro n It s close to the experence of most other European economes from the early 1990s when most governments pursued fxed exchange rate polces. [Fgure 4 about here] As s well understood, most countres n a FBA exchange rate system wll enjoy only lmted monetary and fscal polcy autonomy (see Freden 1991; Gross and Thygesen 1992; and Andrews 1994). In the case of monetary polcy the reason s straghtforward. If exchange rates are truly fxed, any nterest rate above the level n the system wll cause captal nflows untl the dfferental s elmnated; f nterest rates n an economy are below the level n the system, captal wll leave untl domestc rates are brought back to the common level. 17

20 When exchange rates are not rrevocably fxed, governments could n prncple pursue autonomous monetary polces by allowng occasonal exchange rate adjustments. But such polces are unlkely to be very effectve. The reason s that currency traders wll demand an nterest rate premum to hold a currency that s lkely to devalue n the future as a result of nflatonary polces. Such rsk premums obvously defeat the objectve of an expansonary polcy, and makes t less attractve to engage n such a polcy. Ths argument also apples to fscal polcy. Thus, any fscal polcy that s perceved to be nflatonary wll lead to captal outflows and hence the need for the CB to rase nterest rates. Antcpatng ths, the government wll have less of an ncentve to engage n expansonary polces. So long as governments algn ther fscal and monetary polces to the requrements of a fxed exchange rate, monetary polcy wll therefore be the same for all countres nsde the fxed exchange rate system. Of course, someone n the system must set nterest rates for the whole system, and ths prvlege usually befalls a large country commtted to prce stablty. The reason s that f there s even the slghtest uncertanty about future exchange rates, currency traders would prefer to hold the currency wthn the system that s least nflatonary (as a hedge aganst devaluatons). Ths makes t very dffcult for any country to devate much from the polcy of the most deflatonary country, whle mposng no constrans on the leadng country to pursue deflatonary polces (Gross and Thygesen 1992). Although never desgned to produce such an outcome, the ERM created a sngle domnant monetary polcy-setter: Germany. The development of monetary and fscal polces n European countres snce the 1980s clearly bear the marks of ntegraton nto a common deflatonary monetary regme. Along wth growng captal moblty and the transton to a more stable exchange rate system, especally wthn the ERM, came a sgnfcant reducton n the varablty of nflaton and nterest rates, as well as a slowdown n the growth of government consumpton coupled wth fscal consoldaton (see Table 2). Comparng the perod to the perod, average nflaton rates were cut n half from 9.8 to 4.7 and so was the standard devaton n these rates. In the same perod, 18

21 real nterest rates rose from 0.3 to 5.0 percent whle the standard devaton n these rates dropped from 2.2 to 1.2. In terms of government spendng on goods and servces, t went up by an average of 3 percent durng the frst perod, but only by 0.8 percent n the second perod. And whle government defcts ncreased everywhere n the frst perod, most countres experenced retrenchment n the second perod. [Table 2 about here] In terms of our game n Fgure 2, the accommodatng polcy branch was effectvely blocked off by monetary ntegraton. But ths s not the only, or even the most mportant, consequence of movng from a BCM to an ACM world. It also had the effect of elmnatng coordnated barganng for all but the domnant country wthn the currency area. The reason for ths mportant effect s that unons n these countres can no longer exert any nfluence on the real money supply wthn the hard currency area. Because the number of unons s defned relatve to the sze of the currency area, the number of unons n effect ncreases from a few natonal unons to a large number of regonal ones. Assumng that N s very hgh wthn the regon, real wages and equlbrum employment are therefore gven by the followng modfed versons of equatons (10) and (11): (21) w = η α e, and (22) α e* =. η 19

22 Everythng else beng equal, both real wages and unemployment wll thus be hgher n a FBA exchange rate system compared to a BCM world wth monetary polcy autonomy. An alternatve way to explan ths effect of ntegraton s to thnk of the aggregate demand effects of wage ncreases as a cost that s ncreasngly externalzed when movng from a system of several natonal CBs to a system wth only one CB. In the former case, f a large unon rases wages, a non-accommodatng CB would respond by ncreasng nterest rates (whch s n practce how the CB affects the money supply). Because ths rases unemployment, the unon has an ncentve to nternalze ths cost. In the latter case, by contrast, unless the polcy-settng CB s the natonal CB (a case we consder below), the same unon would have no effect on nterest rates because ts wages do not have much mpact on the prces that the foregn CB cares about. The aggregate demand cost of mltancy would therefore be externalzed, undermnng the ncentve for unons to exercse restrant. 14 Hence, when we move from an autonomous fnancal system to an ntegrated one, we expect equlbrum unemployment to rse. Ths s ndeed the pattern we observe among the countres partcpatng n the ERM and ts predecessor, the European currency snake (launched n 1973), and t helps account for the rse n equlbrum unemployment n other European countres snce the early 1990s (see Fgure 5). The fgure shows the evoluton of standardzed unemployment rates n Germany compared to two sets of countres: the regular EMS/snake countres (mnus Germany), and the regular non-ems/snake countres. Notce that untl the early 1990s the rse n unemployment among the EMS countres far exceeded the rse among the non- EMS countres. By the early 1990s, however, most non-ems countres had abandoned any attempts to pursue autonomous monetary polces. Sweden, one of the holdouts for monetary polcy ndependence, for example, made a clear break wth past polces n 1991 when the krona was pegged to the ecu (a peg that soon had to be abandoned, but wthout a return to past polces). 20

23 [Fgure 5 about here] We have kept the data for Germany separate n the fgure because t was the domnant country n the ERM, and ths makes an mportant dfference to the analyss. Because the monetary authorty n the domnant country wll contnue to target the domestc prce level, whle smultaneously settng polces for other countres, unons n the domnant country wll exert the same effect on the real money supply (through the nterest rate) as they dd before the exchange rate system was put n place. Thus, under ERM the trend-settng metalworker s unon n Germany had to antcpate the reacton of the Bundesbank n exactly the same manner as before the ERM; that s, t had to consder the mpact of ts own wages on the German real money supply (m=m(w G )). As explaned n the prevous secton, the German real wage barganng schedule would therefore be: (23) w β α = N 1 β η 1 + N 1 G e G, where w G and e G are the German real wage and employment rates, respectvely. Ths mples more restrant than n the rest of the ERM member countres (compare to equaton 21). Germany s stll negatvely effected by the economc slowdown caused by unemployment n ts man tradng partners, whch were prmarly ERM members, but because macroeconomc coordnaton contnues to functon well wthn Germany tself, we should expect unemployment performance to be better n Germany than n the ERM. Ths s by and large also what we observe. Once the Germany adapts to the 1979 ol shock, unemployment starts to revert to ts orgnal level. Not surprsngly, unemployment shoots up agan after unfcaton n 1991, whch reflected not only the collapse of neffcent east German frms, but also very deflatonary polces by a CB respondng to defct-fnanced government transfers to the east. Yet, once the man economc 21

24 actors n the German economy had tme to adapt to the new stuaton, wage and fscal polcy restrant returned just as t dd after the second ol shock. Indeed, several barganng rounds snce 1994 were on average very moderate. For the centralzed coordnated economes n Europe, monetary ntegraton was not only bad for unemployment, but nvolved broader nsttutonal changes. As we saw n the prevous secton, Northern Europe n the early 1980s could be dvded nto a Scandnavan cluster wth centralzed egaltaran barganng systems and accommodatng full employment polces, and a Germanc cluster n whch there was nether compresson of dfferentals, nor full employment guarantees, nor centralzed barganng. A decade later most of the former (mnus Norway) had moved consderably closer to the latter (Iversen and Pontusson 2000). Ths poses the queston of why some governments that had adhered to accommodatng polces n the past took part n ths ntegraton process. Answerng ths queston requres us to endogenze the choce over nternatonal monetary system a task that s well beyond the scope of ths paper. However, we can pont to two forces of change that almost certanly played a role -- one s nternatonal, the other technologcal. As descrbed by Scharpf (1991), Hellener (1994), and Soskce (1999), once fnancal market lberalzaton got underway n mportant countres lke the UK and the Unted States, and once the U.S. Federal Reserve embarked on a radcally deflatonary strategy, the German Bundesbank was compelled to follow sut. Ths pulled those countres n Europe that dd not have very elaborate captal controls down a deflatonary monetary path. In turn, we know from the analyss n the prevous secton that when macroeconomc polces turn non-accommodatng, there s less of a wage-cost ratonale for havng hghly centralzed wage barganng systems. In part, therefore, the abandonment of encompassng peak-level barganng n countres such as Belgum, the Netherlands, Denmark and Sweden can be seen as a natural response to the reduced capacty of governments to engage n accommodatng natonal fscal and monetary polces. 22

25 In part, however, we must also see the move towards more restrctve macroeconomc regmes as a response to growng pressures for decentralzaton. These pressures could not be effcently addressed wthn the confnes of accommodatng polcy regmes for reasons have been elaborated n Pontusson and Swenson (1996, 1999) and Iversen (1996). Wth the transton to more skll-ntensve producton technologes based on a commtted workforce wth extensve frm and ndustry specfc sklls, narrow wage dfferentals became a barrer to compettveness. Frms were mpared n ther ablty to reward employees for nvestng n specfc sklls and for takng on more responsbltes at the shop floor, and full employment exacerbated these problems because t ncreased the barganng power of the unons representng the lower pad relatve to that of sklled workers unons. Optng out of centralzed barganng, however, requred alternatve means to control wage pressures, and non-accommodatng macroeconomc polces were the soluton. Although the mmedate effect of optng out of centralzed barganng was to ncrease nflatonary pressures, whch undermned compettveness, the threat of massve captal flght made t nearly mpossble for governments to mantan an accommodatng polcy stance. IV. The promses of monetary unon: Restoraton of polcy autonomy? Now consder the case where countres abandon ther natonal currences and become part of a currency unon wth a sngle transnatonal monetary authorty. Ths s the EMU scenaro llustrated n Fgure 6. Snce there are a large number of unons wthn the common currency area, and assumng that these bargan wages smultaneously and ndependently, the effect of any ndvdual money wage ncrease on the European prce level wll be small. In ths case the stuaton for all economes n the unon are smlar to the non-domnant economes n the fxed exchange rate system descrbed n the prevous secton. Hence all economes wll now be descrbed by the followng employment equaton (24) α e* =. η 23

26 As far a monetary polces are concerned, therefore, the what s new compared to the fxed exchange rate system s bad news: Wage restrant and employment wll be fallng n the domnant economy. In fact, t wll also be fallng n other member countres, but by a smaller amount. [Fgure 6 about here] To see ths, consder Fgure 7, whch summarzes the results untl now. The bottom upwards-slopng lne s the wage equaton (10), whch shows the barganed real wage schedule when countres have ther own ndependent monetary systems. Wth coordnated barganng, unons have an ncentve to take nto account ther own effect on the real money supply, and therefore to act wth greater restrant than f barganng was uncoordnated. The employment rate n ths case s determned by the ntersecton of the barganed real wage schedule and the feasble or prce-determned real wage whch s unty n equlbrum (the horzontal w=1 lne). The labor demand schedule e(m,w) wll also pass through ths pont. Ths s the scenaro that was dscussed n secton II. [Fgure 7 about here] Wth the move to a fxed exchange rate system lke the ERM, unons n the non-domnant economes can no longer exert any affect the real money supply, and they have no ncentve to take nto account ther effect on the aggregate prce level. The real wage schedule for these countres s gven by equaton (21), and represented by the steeply upwards-slopng lne n Fgure 7. However, the domnant country n the fxed exchange rate system -- Germany n the case of the ERM wll stll be characterzed by the old barganng schedule. Snce m erm s the same for all ERM members, the wage-employment combnaton (w, e) of the ERM members excludng Germany must be on the same labor demand schedule, e(m erm, w) as the German wage 24

27 employment combnaton (w G, e G ). To satsfy the condton that the geometrc average of the real wages of the ERM members s unty, the German real wage must be below and the real wage of the other members above unty. As can be seen from Fgure 7, that mples that employment both n Germany and n the other member states s lower than n the case where countres have ndependent monetary systems, but the fall n employment s greater n the non-german ERM countres. Ths s the scenaro consdered n secton III. Fnally, n the case of a monetary unon (EMU), unons n all countres f they bargan ndependently and the monetary authorty (the ECB) targets the average prce level wll be characterzed by the steeply upwards-slopng barganng schedule, wth a correspondngly lower rate of employment. As can be seen from the fgure, all countres wll experence a reducton n employment, but the drop wll be greater n Germany where unons no longer face a central bank targetng German nlfaton. The mltant demands of IG Metall, the pace-settng unon n Germany, n the past couple of wage barganng rounds, wth the Bundesbank on ts death bed or actually dead (as a polcy-makng entty), appear to bear ths out. For the frst tme snce European monetary ntegraton was ntated, German unemployment s now hgher than n the former regular ERM countres. The monetary unon case also dffer from the fxed-but-adjustable exchange rate case n terms of fscal polcy. But whereas monetary unon mples the loss of natonal monetary polcy autonomy, the exact opposte s the case for fscal polces. Wthout exchange rate rsks and wth fully moble captal, governments are no longer lmted to pursung non-accommodatng polces. Of course, an nterest premum wll stll be assessed aganst proflgate governments as a hedge aganst default, but that s also the case n the BCM world. Furthermore, unlke ndependent natonal CBs a transnatonal central bank such as the ECB cannot punsh ndvdual countres for expansonary polces no more than t can punsh ndvdual natonal unons for mltant behavor. In terms of fscal polcy, therefore, we are returned to the analyss of autonomous monetary systems n secton II, and all the conclusons about fscal polcy from ths secton now 25

28 apples agan. Indeed t takes on added mportance because monetary polcy s now neffectve as a deterrence aganst both wage mltancy and expansonary fscal polces. Hence, there s a real possblty that the macroeconomc regme could once agan turn accommodatng n some countres. Ths rases the mportant possblty that the varaton n macroeconomc coordnaton that we observed across countres n the BCM world, whch vanshed durng the fxed exchange rate phase, wll reemerge. Recall that accommodatng fscal polces wll, ceters parbus, lead to more centralzed barganng. If monetary polces are no longer effectve, some form of centralzed barganng nvolvng fscal polces may therefore (re-)surface n some countres. It s possble, for example, that left governments, as part of a broader fscal strategy, wll bargan wage restrant wth the unons n return for fscal actvsm. Employers may have few optons but to go along wth such a strategy. However, t s unlkely that we wll see a return to Swedsh-style soldarstc and centralzed barganng. There are several reasons for ths. Frst, as we already noted, the barganng poston of employers has probably ncreased wth the transton to more knowledge-ntensve forms of producton. Employers are wllng to ncur a hgher cost to avod hghly ntrusve centralzed regulaton of local employment and wage condtons. Secondly, the tradtonal core consttuency of left partes sem-sklled blue-collar workers has been n declne and are no longer essental for electoral success (Ktschelt 1994). To the extent ths s the case, left governments wll be less nsstng on redstrbutve wage polces, but no less concerned wth full employment. Indeed, rather than beng smple poltcal vehcles for redstrbuton, t seems qute plausble that socal contracts nvolvng fscal polcy wll also have provsons for ncreases n the flexblty of work and wages, as well as tax reforms and subsdes to encourage the employment or educaton of low-sklled workers. The Dutch Wassenaar accord s perhaps the best example of a negotated reform of labor markets and socal polces, and s wdely credted for the mprovements n Dutch labor market performance snce the md-1980s (see Vsser and Hemerjk 26

29 1997). Smlar experments have seen the lght of day n countres as dverse as the Netherlands, Italy, Span, Portugal and Ireland (see Regn 1997 and Rhodes 1998). The neo-lberal alternatve to more coordnaton s wholesale deregulaton of labor markets. Advocates of such a soluton can appeal to a strong sense among many polcy-makers that labor markets n Europe are no longer workng as well as they once dd. In the neo lberal nterpretaton ths s because unons have become too strong, and markets too regulated, to produce full employment. But strong unons and labor market regulaton have always characterzed the European economes. If our argument s correct, at least part of the explanaton for the lack of and wage restrant s that monetary ntegraton has caused a breakdown of coordnated barganng. If ths dagnoss s correct then restoraton of coordnaton, not deregulaton, may not be answer. In part such coordnaton may come at the natonal level n the form of socal pacts. But coordnated barganng may also re-emerge n varous forms at the European level. For example German unons could emerge as wage leaders for the rest of Europe, especally f the ECB targets German rather than European nflaton and thereby provdes German unons wth a greater ncentve to exercse restrant. More ambtous plans of explct coordnaton between natonal unons and employers assocatons, even European-wde barganng, are clearly also beng contemplated by unons. 15 But whether deregulaton of labor markets or new forms of macroeconomc coordnaton ultmately prevals s not our man concern here. The mportant pont s that after a phase of convergence n macroeconomc polces and nsttutons, there s once agan scope for real poltcal contestaton over macroeconomc polces, and such contestaton could lead to renewed natonal dvergence and nsttutonal nnovaton at the European level. Theoretcally, and such contestaton analyss shows that although the transton from a BCM world wth separate currences to an ACM world wth sngle currency can be seen as a process leadng to ever hgher levels of monetary ntegraton, the effects of ths process are non-lnear and can only be 27

30 understood n the ntersecton of macroeconomc polcy-makng and strategc choces of organzed nterests n the labor market. V. Concluson Durng the past two decades, the unemployment rate n European countres has doubled, trpled, or even quadrupled. At the same tme unemployment performance contnue to vary across countres, and dvergence has been on the rse recently wth some countres, such as the Netherlands and Denmark, dong better, and others, such as Germany and Sweden, dong worse. Smlar stores could be told for other regons of the world. Curously enough whle these major, and poltcally hghly senstve, changes unfolded, poltcal economsts ncreasngly turned ther attenton to explanng short-term fluctuatons n nomnal varables. Ths paper s ntended to redrect some of the energes n poltcal economy to understand changes n the real economc varables, and to revve the study of macroeconomc polces and nsttutons. Contrary to wde held belefs, economc theory does not mply that macroeconomc polces and nsttutons are unmportant for varables lke unemployment and earnngs equalty. Once we ntroduce large organzed players nto the economy, such as ndustry unons and employers assocatons, macroeconomc polces do affect real economc behavor and help us understand both cross natonal dfferences and changes over tme. For example, the theory we have presented suggests that monetary ntegraton n Europe has been a man cause of the rse n European unemployment. Because monetary ntegraton causes the macroeconomc costs of mltant unon behavor to be ncreasngly externalzed, real wages and unemployment wll rse. Although many beleve that monetary ntegraton has such effects, the mechansm s rarely dentfed. Fnally, our model revves Hbbs thess that long-terms dfferences n macroeconomc polces, dstrbuton, and unemployment are due n part to partsan poltcs. Thus, pror to the early 1980s dfferences n the hstorcal strength of the left has been a major cause behnd the 28