4121606

Embed Size (px)

Citation preview

  • 8/13/2019 4121606

    1/23

    Low-Inflation Targeting and Long-Run UnemploymentAuthor(s): Per Lundborg and Hans SacklnSource: The Scandinavian Journal of Economics, Vol. 108, No. 3 (Sep., 2006), pp. 397-418Published by: Wileyon behalf of The Scandinavian Journal of EconomicsStable URL: http://www.jstor.org/stable/4121606.

    Accessed: 01/10/2013 03:27

    Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at.http://www.jstor.org/page/info/about/policies/terms.jsp

    .JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of

    content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

    of scholarship. For more information about JSTOR, please contact [email protected].

    .

    Wileyand The Scandinavian Journal of Economicsare collaborating with JSTOR to digitize, preserve and

    extend access to The Scandinavian Journal of Economics.

    http://www.jstor.org

    This content downloaded from 203.170.79.82 on Tue, 1 Oct 2013 03:27:36 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/action/showPublisher?publisherCode=blackhttp://www.jstor.org/action/showPublisher?publisherCode=sjehttp://www.jstor.org/stable/4121606?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/4121606?origin=JSTOR-pdfhttp://www.jstor.org/action/showPublisher?publisherCode=sjehttp://www.jstor.org/action/showPublisher?publisherCode=black
  • 8/13/2019 4121606

    2/23

    Scand.J of Economics108(3),397-418, 2006DOI: 10.1111/j.1467-9442.2006.00464.xLow-inflation Targeting and Long-runUnemployment*Per LundborgSwedishInstitute or SocialResearch,StockholmUniversity, E-10691 Stockholm,[email protected] SacklinSwedishInstitute or SocialResearch,StockholmUniversity,SE-10691 Stockholm,[email protected]

    AbstractThe model of Akerlof,Dickens and Perry (2000) (ADP) predictsthat low inflationmaycauseunemploymento persistathighlevels.Whenapplied o U.S.data, heirresults tronglyrejected he conventionalNAIRUmodel.Weapply he ADPmodelto Swedishdata.Thefactthatour Swedishdataalso reject heNAIRUmodelhas a numberof interestingmplicationsfor the Swedisheconomyand,potentially,or otherEuropean ountriesas well. The resultsindicate that raisingthe Swedish inflationtargetfrom 2 to 4 percentwould bring long-rununemploymentownby severalpercentage oints.Thepossibilityof ADP-type ong-runPhillipscurvesalso across heeurocountriesmayraise some concernabout he EMUproject.While detailedstudieson othercountriesare needed, here is nothing o suggestthatthesenon-verticalPhillipscurves would not differ considerably crossthe euro countries.Anysingle inflationlevel targetedby the ECB would then generateexcess unemploymentnindividualmember tates.Keywords: hillipscurve;efficiency wages; near-rationalityJELclassification:E24; E31;J41

    I. IntroductionFollowinga waveof heavyanti-inflationaryoliciesduring he lastdecades,inflation argetinghasnowbecomethe leadingprincipleof monetarypolicyin many European conomies.In the euro area,the UK and Sweden,thetargetsareset to 2 percentor less. In a historicalperspective uchinflation*We are grateful o two anonymous eferees for very detailedcommentsandto GeorgeA.Akerlof ordiscussionson an earlyversion.We have benefited rom commentsby LawrenceBall, MikaelApel andotherparticipants t the workshop ThePhillipsCurve:New TheoryandEvidence n Stockholm,May2002.We alsoacknowledge seful comments romseminarparticipants t the TradeUnion Institute or EconomicsResearch FIEF),the Institute orInternational conomicStudies, he Riksbank,he EconomicCouncil,StockholmUniversityandGrteborgUniversity.Theresearch asbeen financedby a grant rom he SwedishCouncilfor WorkingLife and SocialResearch.? Theeditorsof the Scandinavianournalof Economics 006.Published y BlackwellPublishing,600 Garsington oad,Oxford,OX42DQ,UK and 350 MainStreet,Malden,MA 02148,USA.

    This content downloaded from 203.170.79.82 on Tue, 1 Oct 2013 03:27:36 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 4121606

    3/23

    398 P Lundborgand H. Sacklenrates are strikingly ow; no OECDcountryhas experienced nflationbe-low 2 percentfor any extendedperiodafter WorldWarII and up to the1990s.Thewidespreadmplementationf low-inflationargets uggestsa strongcommitmento the idea of a vertical ong-runPhillipscurve.According othisview,inflationhas no long-runeffectson unemployment,mplying hatlow unemploymentn principlecould coexist with low rates of inflation.However,n sustainedperiodsof low unemployment,s in theperiod1955-1968, inflationhasbeenconsiderably igher han2 percent.'Moreover,heperiodof low inflation ince theearly1990shasbeencharacterizedy high,or veryhigh, jobless rates.Such observations aveprompted esearchersoquestionthe verticalPhillipscurve.The possibility of a long-run unemployment-inflationrade-offraisesseveralintriguingquestions.Rather hanattributing igh unemploymentoshiftsin the verticalPhillipscurve,a trade-offwouldimplythatheavyanti-inflationarypolicies might have caused unemploymento persistat highlevels. This,of course,shouldbe of specialconcernto countriesaimingatvery low inflation.If a trade-offexists, there shouldalso be some concernabout the viabilityof the EMU project.Since thereis nothingto suggestthatthe trade-off s identicalacrossmembercountries,how shouldan infla-tion targetfor the whole euro areabe determined?Giventhatmembershipimplies a common(and low) inflationrate, could a countryobtainmoreoutputandemployment utsidethe unionby optingfor a country-specificinflation arget?A numberof empirical tudieshavechallenged heconventionalwisdomthat the long-runPhillips curve is vertical. In particular, everal studieshave foundadverse ong-runeffects of low inflation.Forexample,Bullardand Keating (1995) reporteda negativelong-runresponseof outputto areductionn inflation n European ountrieswith low inflation.Studies onthe U.S. by King and Watson(1994) and Fair(2000) suggesta long-rununemployment-inflationrade-off.In addition,a growing heoreticaliterature as providedmicroeconomicrationales or a trade-offbetweenunemploymentndinflationat low infla-tion rates.Onecategoryof modelsbuildson the existenceof nominalwagerigidity.Using a wage-bargainingpproach,Akerlof, Dickens and Perry(1996) rely on the adverseeffects of low inflationon real-wage lexibilityas discussedby Schultze(1959), Samuelsonand Solow (1960) andTobin(1972). If the inflationrate falls from,say, 3 to 0 percent,some firms instochastic teadystatecannotreduce henominalwageandarethusexposed

    During heperiod1955-1968,averageunemploymentn the 19OECDcountrieswas around2 percentand average nflationaround4 percent, .e., substantially igherthanthe presenttargets.OThe editors of the Scandinavian Journal of Economics 2006.

    This content downloaded from 203.170.79.82 on Tue, 1 Oct 2013 03:27:36 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 4121606

    4/23

    Low-inflationtargetingand long-run unemployment 399to a real-wage hock thatcausesunemploymento rise. Holden(2004) con-structsa bargainingmodel based on legal requirementsn manyEuropeancountries hatnominalwage contracts an only be changedby mutualcon-sent,therebygivingworkersa strongerbargaining ositionat low inflation.Thekey implicationof both models is a long-run rade-offbetweenunem-ploymentandinflationat low inflationrates.Akerlofet al. (1996) appliedtheirmodelto U.S. data,and laterDjoudadand Sargent 1997) andFortin(2001) presented esultsbased on Canadiandata.Dickens(2001) includedthe samemechanism or a set of European ountries.All these studiesre-ject the traditional erticalPhillipscurve in favorof the idea that wagerigidityyields a negativelyslopedPhillipscurveat low inflation.Anotherapproach elies on near-rationality,.e., thatundercertaincondi-tions agents maydeviatefromprofit-or utility-maximizingehavior.Morethan30 yearsago, Eckstein and Brinner 1972) had alreadynotedthat inthe U.S., wage- andprice-setters artially gnored nflationduringyearsoflow inflation.Buildingon this idea,andreferring o extensivesociologicalandpsychologicalevidence,Akerlof,DickensandPerry 2000) (henceforthADP) set up an efficiency wage model in whichagents'behaviorchangesas the economyshifts betweenhigh- andlow-inflation egimes.If inflationis disregarded t low rates,the firm sets a lowerwage and a lowerpricerelativeto nominalaggregatedemand.As a result,unemploymentan besustainedat lowerlevels than if inflationwere fully accounted or. Whentested on U.S. data,the standardPhillipsrelation s rejected n favorof thenear-rationalityypothesis.2The Phillipscurveimpliedby the ADP modelhas several mportantm-plications.Above all, the model identifiesan inflationratethatminimizeslong-run unemployment.Departures rom this rate could potentiallygiverise to largecosts in terms of unemployment. his featuremakesthe modelparticularlynteresting or a countrylike Sweden,where inflation s nowtargetedat a low level (2 percent)and wherefull employment as tradition-ally been given highpriority.For this reason,we estimated he ADP modelon Swedishquarterlydatato examinewhethera different nflationtargetwouldreduce ong-rununemployment.Ourpreferred egressions basedonsurveydata on inflationexpectations) how that, by raisingthe inflationtarget rom2 to around4 percent, ong-rununemploymentwouldgo downfrom 4-5 percent o about 2-3 percent.In thenextsectionwe beginby considering gents'rationalesornottak-ing low inflation ully into account.We thenoutlinethe basicADP model.The econometric pecificationsandourdataare introducedn SectionIII,2 See also Fortin 2001) and Dickens(2001) for some mixedempiricalresultsfor Canadaanda numberof European ountries, espectively.

    ? The editors of the Scandinavian Journal of Economics 2006.

    This content downloaded from 203.170.79.82 on Tue, 1 Oct 2013 03:27:36 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 4121606

    5/23

    400 P Lundborgnd H. Sacklenwhich also contains the empiricalresults.In the final sectionwe discussthe policy implicationsof our findings.

    II. Near-rationalityand the Phillips CurveWhyInflation is Disregarded at Low RatesThe ADP model recognizesthat, duringhigh inflation,all firms behaverationallyby takinginflationfully into accountwhen they set wages andprices.However,a specialfeatureof the model is thatsome firms tend todisregardnflationwhen it is low and hence behavenear-rationally.

    Why would firms disregard ow inflation?ADP offer numerousargu-ments(see ADP,pp. 4-10, for a comprehensiveurvey),amongwhichtheso-callededitingarguments presumablyhe mostfundamental. irmsfacea myriadof everydaydecisionsconcerningproductdesign,whom to em-ploy, investments, ost reductions, tc., thatare all of some importanceoprofits. In such complexsituations, irms tend to ignorefactorsthat havelittleeffecton profits.ADP showconvincinglyhat osses fromdisregardinginflation n wage- and price-settingactuallybecomenegligibleat low in-flationrates.Hence, low-level inflationmaysimplydisappear s a variableof relevance o profits.3Is thereanyempirical upport orthe existenceof near-rationalehavior?ADP presenta rangeof such evidencefor the U.S. For instance,the im-portanceof inflationexpectations n determiningwages seems to dependon the inflationrate.ADP show that the magnitudeof the coefficientonexpectedinflationdependscruciallyon the level of inflation.Periodsofhigh inflationtend to yield coefficientsaroundone, while periodsof lowinflation end to yield estimatedcoefficientsbetween0 and0.5.4ADP develop heirargumentsn an efficiency-wage ramework,mplyingthatfirmssetwagesunilaterally. hismaybe anadequatedescription f theU.S. labormarket,butless so of Europeanmarketswherecollectivebargain-ing dominates.However,any economycontainselementsof bothunilateralwage-settingand wage bargaining.Whenapplying he ADP model to ourSwedishquarterly atawe implicitlyassume hatfirmshaveenough atitudefor efficiency wage-setting.Studiesby Agell and Lundborg1995, 2003)

    3 At some low level, inflationmaybe perceivedas pricestability.Or,as argued n Blinder,Canetti,Lebow and Rudd(1998), a prominentdefinitionof price stability s inflationsolow that it ceases to be a factor n influencingdecisions .4A consequenceof disregardingnflationwould be a morefrequentoccurrence f nominalpricecontracts.That nominalcontractsare more commonat low inflation s consistentwithexistingevidence(see ADP,fn. 9).? The editors of the Scandinavian Journal of Economics 2006.

    This content downloaded from 203.170.79.82 on Tue, 1 Oct 2013 03:27:36 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 4121606

    6/23

    Low-inflation argetingand long-run unemployment 401havealso shownthat even for a typicalbargaining conomyas in Sweden,firms' wage-setting s affectedby efficiencywage considerations.5The ADP Model6At anypointin time,firmsmaybehaveeitherrationallyindexedr) or near-rationally nr). Rationalfirms know their workers'inflationexpectationsand incorporatehese in wage-setting.For reasonsdiscussedabove,near-rationalfirms neglect to incorporatenflation when it is low. Whetherafirmbehavesrationally r near-rationallys not exogenouslygivenbutwillbe determinedby the rateof inflation.Firmsset efficiency wages that minimizelabor cost per expectedeffi-ciencyunit,wj/e

  • 8/13/2019 4121606

    7/23

    402 P LundborgndH. SacklknTheeconomyconsistsof n monopolisticallyompetitive irms thatdividetotal aggregatedemandamong them accordingto the relativeprices oftheirrespectivegoods.The first-order onditionof the profit-maximizationproblemimplies that the firm's price is determinedas a markupon theexpectedunitefficiencylaborcost. Thatis, pj= (//(0 - 1))(wj/ej), where/ is the priceelasticityof demand.It can easilybe shownthatnear-rationalityausesa relativedecrease nprofits. ADP demonstrate hat losses from disregardingow inflationarenegligiblebelow2 percentandpossiblyevenat inflationratesas high as 5percentor more.As inflation ncreases,however, he losses will eventuallybecome substantial; ee ADP,pp. 15-16. In line with our earlierreason-

    ing, near-rationalirms wouldthereforebe expected o becomeincreasinglyawareof forgone profitsas inflationincreases.Assumingthat this toler-ance towards osses differs across firms, it then follows that more andmorefirmswouldswitch to fully rationalbehavior.Wheninflationreachesa sufficientlyhigh level, all firms incorporatenflation n wage-andprice-setting,andwe are back in a worldof fully rationalagents.In the ADP model,the abovemechanisms specifiedin termsof thresh-old levels for losses inducedby near-rationalehaviorrelativeto profitsunder ull rationality.Once the relative oss reaches histhresholdevel, thefirm switches fromnear-rationalo fully rationalbehavior,or vice versa.Assuminga normaldistribution f these thresholds, he fractionof price-setterswho behavenear-rationally ill vary with the rateof inflation,7r.In the followingwe denote this fractionby 4(.), where 1 is the standardcumulativenormaldistribution.Calculating he averageprice level at time t as a weightedaverageofprices set by rationaland near-rationalirms, the short-runprice Phillipscurveis givenby

    ( A -Cu (1-)/ A - Cul [1 (4)(I + 7r,) = A t-)

    [I + (1)t(.)7 ], (4)A C u e B 1 ca twhereue_, is expectedunemploymentt the startof t- 1.8III. The Swedish Long-run Phillips CurveThe Lowest Sustainable UnemploymentRate of Inflation (LSURI)By takingthe logarithms f both sides of the short-runPhillipsrelation nequation 4), andmaking he sameapproximationss ADP (p. 16), we get8 The average price level is p, = l,(.)pp,, + [1 - ,(-)]Pn,,,t and 1 +7r, = Pt/Pt-I. To obtainexpressions orpr,t andPnr,t,we use (3), (2a) and (2b) in (1). The resultingexpression sthenused,alongwith (3), in the price equationspj,t= (01/( - 1))(wj,t/ee,t).? The editors of the Scandinavian Journal of Economics 2006.

    This content downloaded from 203.170.79.82 on Tue, 1 Oct 2013 03:27:36 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 4121606

    8/23

    Low-inflationtargetingand long-run unemployment 403

    7r = d - aut + Dt(?.)7r+7Aut. (5)

    The long-runsteady-statePhillipsrelation s characterizedy equalitybe-tween actual and expected inflationand by a constant(and known)un-employmentrate. Imposingthese restrictions, he Phillipsrelationin (5)reducestou = (d/a) - [1 - t(-)]Wr/a. (6)

    The standard,ully verticallong-runPhillipscurve,which impliesthatunemploymentlwaysequalsa natural ate,is derivedunder he assump-tion of full rationality.By allowingfor near-rationality,e maynote from(6) that this long-runPhillipscurvewill, in general,not be vertical.Hence,the conventional onceptof a natural ate of unemployments less rele-vantforthe ADP model;in the followingwe let the natural ate denotethelevel of unemploymenthat obtainswhen all firms behaverationally.Thisspecialcase of (6) occurs f inflation s sufficientlyhighsuchthat4(.)= 1.From(6), the natural ate is given by u = d/a.Since disregarding ero inflation is equivalent o fully rationalbehav-ior, the naturalrate also obtains at 7r=0. As inflationincreases abovezero, however,near-rationalirms opt for a wage that is lower than thewage set by fully rationalfirms. As this implies a lower averagewagecomparedto the wage in an economy where all firms always behaverationally, nemploymentwill be lowerthan the natural ate.Consequently,the long-runPhillipscurve has a negatively lopedsegmentat low inflationrates.As inflationrises, thereare two opposingforces at work.While near-rational irms disregardmoreinflation,which tends to furtherreduceun-employment, t also becomes increasinglycostly for firms to behavethisway.More and morefirmstherefore witchto full rationality,nd hencesethigherwages,which tends to increaseunemployment.Eventually, t someinflationrate,the share of rational irms has increased o such an extentthatunemploymentctuallystartsto increase.This processcontinuesuntilinflation has reachedthe level at which all firms have switchedto fullyrationalbehavior.At this level of inflation and above, unemploymentsagainat its natural ate.Thus, the model yields the hump-shapedong-runPhillips curve thatADP show in their Figure 1. This shapehas severalinteresting mplica-tions. Most importantly,here exists a lowest sustainableunemploymentrate of inflation(LSURI),yielding the lowest sustainableunemploymentrate (LSUR).Formally,he LSURI is the inflationratewhere u/o ir 0in (6).

    O The editors of the Scandinavian Journal of Economics 2006.

    This content downloaded from 203.170.79.82 on Tue, 1 Oct 2013 03:27:36 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 4121606

    9/23

    404 P Lundborgand H. SacklinEmpirical Specifications and DataIn order to proceedtowardsan empiricalspecification,we returnto theshort-run hillipsrelationn equation 5). In linewithADP,we approximatethe argumentn the standardnormalcumulativedistributionunctionbyD + Eir2, whereD and E areparameters.iL representshe effects of pastinflationon the likelihoodthatpeople act rationally oward nflation.ThePhillipscurve that we estimate henbecomes:

    rt= d + alu + a2ue_-1+(D + Eir.t)rte + kXt+ Et, (7)whered, D, E, a1, a2 andk areparameters, a vectorof dummyvariablesande an error erm.WeproxyI7Lby severaldifferent pecificationssuggestedby ADP.Oneis a geometricallydeclining weightedmovingaverageof past inflation nconsumerprices:

    =L,t (1 - 6)7fLt-1 + 6Ft-l, (8)in which 6 is estimated.An alternative pecification sEI=1 1 - iA)rti (9)L,t= I (1 iA)E = 1 iA

    where A is estimated, indexes quartersand I is set to 16 quarters. naddition to these formulations,we also apply the following weightingprocedure:I

    rL=,t gii-i (10)i=1in which the weights gi are estimated(0

  • 8/13/2019 4121606

    10/23

    Low-inflationtargetingand long-run unemployment 405

    domestically.For a largeeconomylike the U.S. (for whichADP estimatedtheirmodel),the consumerpriceindexmaybe a relevantndex.This,how-ever,wouldnot be the case fora smallopeneconomyas in Sweden.Instead,the averageconsumerpricelevel, p, is determined s a weightedaverageofthe prices set by domestic and foreign producers, p = (1 - m)pd + mpm,wherepd is the domesticconsumerprice, pm the price of importedcon-sumergoods and m the value of importedconsumergoods as a shareoftotalconsumption.By takingthe differencesof this index,holdingthe im-port shareconstant,we may derivethe relevantprice inflationfor goodsproducedandconsumeddomesticallyS= - mtA

    ,(11)

    Pt-1 - mtplwhich is our dependentvariable.9We used quarterly ata from 1963:1to 2000:2 which are annualizedbycalculating he percentagechangein the relevantprice indices duringthelast four quarters.10 igureAl in the Appendixdisplaysa scatterplotofopen unemploymentndconsumerprice inflation.Estimatingthe model for a small open economy implies taking intoconsideration he dependenceon external factors. As mentionedabove,our dependentvariable, nflationin productsproducedand consumedinSweden, s determined s the differencebetweenthe CPIandimportedn-flation.However,ourinflationseries for importedgoods includesnot onlyconsumptiongoods but also intermediary oods. This introducesa meas-urementerror into our domestic inflation series in periodswhen pricesof intermediate oods move differently hanpricesof consumption oods.To accountfor this, and to control for supply shocks, we introducedanumberof dummyvariables o captureoil price increasesin 1973-1974and 1979-1981, and decreasesin 1986. Dummiesalso cover price hikeson food inputsin the early 1970s, the Swedishtax reformin 1990-1991andthe extremewage increases n 1995-1996 that canbe traced o foreign9Let7rmdenotenflationn importedonsumeroods.For he ong-runteady-statehillipsrelationn(6)tobevalid,weassumehat xchangeates djustothat, xpressedn domesticcurrency, = 7rd= 7rholds orthelongrun.10Weareawarehat hiscould ntroduceomedependencen theerrorerms, .e.,autocor-relation. n the Swedish conomyhereareprice shocks n certainmonthsparticularlyJanuary ndJuly)due to changesthathit the whole economyat the sametime such as thegovernment'shangesntaxes,newwageagreements,tc.Seasonaldjustmentf datamustthereforeehandled ithgreat areand,bydefiningnflationnthisway,weeliminate llseasonaldependencen our data.How severe s theriskof autocorrelation?heDW statisticsare notreliablewhenwe have aggedvariablesn the RHS.However,hisis notthecasein ourpreferredegressions,.e.,thosebasedonsurveys f inflationxpectations.ere, heDWstatisticsare in the inconclusive angeandwe haveinterpretedhis as if we do not havea major utocorrelationroblem.

    C The editorsof the Scandinavianournalof Economics 006.

    This content downloaded from 203.170.79.82 on Tue, 1 Oct 2013 03:27:36 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 4121606

    11/23

    406 P Lundborgnd H. Sacklenincreases n pricesof pulp andpaper.All variablesare defined in greaterdetail in the Appendix.ResultsUsing maximum-likelihoodmethods, we estimated a total of 120specifications.Theregressionsdifferwithrespect o sampleperiods,meas-ures of unemployment nd the way in which inflationexpectationsareaccounted or. TableAl in the Appendixgives an overviewof the differ-ent specifications.In ourjudgment113 regressionscame out withoutanymajorproblems n terms of identification,meaningfulparameterstimatesetc.

    The followingresult section deals exclusivelywith the long-runPhillipscurve.12We first focus on the resultsusing estimated adaptive) nflationexpectations,not because we place moretrustin these results,but simplybecause this is the conventionalway of dealingwithexpectations.Wethenreportwhatwe consider o be a morereliableset of resultsbasedon surveysof inflationexpectations.Estimated ExpectationsLet us firstconsider he baselineregressions,.e., the regressions or 1963-2000 with unemployment efined as openunemployment.f the parameterE in (7) is zero,the coefficient on expectationswill not varywith inflation.The theorythat agents ignore inflationwhen it is low requires hatE begreater hanzero. In 23 of the 24 regressionswe find thatE is significantlypositive as predictedby theory.13The averageestimateof the LSUR is2.08 percent,associatedwith an average nflation,LSURI,of 2.61 percent.The vast majorityof LSURs rangebetween 1.6 and 2.5 percentand theassociatedLSURIsrangebetween2.0 and3.0 percent.11For the remaining ases we ran into problems imilar o those ADP hadin some of theirregressions or the U.S.,namelythat the estimatesof D andE tended o approachoo andooc, espectively. hat s, theoptimization rocedurewastrying o eliminate oefficientvaluesbetweenunityand the lower floor. Akerlofet al. did not interpretheir failed egressionsas resultssupportinga vertical long-runPhillipscurve, but ratheras the result of someunexplained onvergenceproblem. t shouldbe notedthatall sevenof ourregressionswithconvergence roblems all into the categoryof regressionswithestimated xpectations.Fourof the sevenregressions ccurwhen we use maleunemploymentn the regressions.Wehavenot been able to identifyany othercommoncharacteristics.We have chosen to disregardthese failedregressionsn this paper.12Specifically, he short-run urve in (7) implies the long-runrelationu = (d/Eat)- [1 -QD(D Er 2)]r/ a,, whereEat is the sumof the coefficientsfor laggedunemployment.13As discussed above, we use four alternative measures for ire, three measures for7ftL, andtwo alternativeag lengthsfor unemployment. his gives us 24 possiblecombinations. none regressionwe encounteredhe convergenceproblemmentionedn footnote11.OThe editors of the Scandinavian Journal of Economics 2006.

    This content downloaded from 203.170.79.82 on Tue, 1 Oct 2013 03:27:36 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 4121606

    12/23

    Low-inflationtargetingand long-run unemployment 407

    10-8- **w *

    6- ? *

    2- *0

    0 2 4 6Unemployment

    Fig. 1. Estimatedexpectations: he average Phillipscurve

    Figure1 summarizesomecrucial eaturesof the long-runPhillipscurvesimpliedby the regressions.Foreach regressionwe haveevaluatedhe un-employmentateassociatedwith0, 2, 4, 6 and 8 percent nflation.Toobtaina representative hillipscurve we then fitted a curvethrough he aver-age unemploymentateat the different evels of inflation.14 Forinstance,consider he effect on unemploymentf changing he rateof inflation romzero to the LSURI.Largeunemploymenteductionsakeplace as inflationis raisedfrom0 to 2 percentand at LSURIthe totalreductionamounts oalmost2.5 percentagepoints.By the time inflationhas reached4 percent,the resultssuggestthatunemployments again rising.In Table 1, columns (i) and (ii), we reportdetailed results for tworepresentativeegressions.For example,the table shows the size of thecoefficient on inflationaryexpectations, Q(.), when evaluatedat differ-ent rates of inflation. At zero inflation, the size is approximately .5.The coefficient then increases with inflation and is close to unity wheninflationreaches 6 percent.Recall that the theoreticalmodel interpretsthe coefficient 1(.) as the share of fully rationalfirms. This interpre-tation thus suggests that many firms take inflation fully into accountalreadyatpricestability,andalmostall firms behaverationally t 6 percentinflation.So far,the resultshaveindicateda greatdealof robustnesso variationsin the way expectationsare estimated.However,we are also interested ninvestigatinghow the model performswith respectto otherdefinitionsofunemployment nd to othersample periods.The resultsare discussedin14We capture he unemploymentminimumby fitting the curvethrough he averageLSURandLSURI.

    C The editors of the Scandinavian Journal of Economics 2006.

    This content downloaded from 203.170.79.82 on Tue, 1 Oct 2013 03:27:36 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 4121606

    13/23

    Table 1. Estimatedparameters or the long-run Phillips curveaIndependentvariables andcharacteristics (i) (ii) (iii) (iv)Constant 0.014 0.015 0.032 0.042

    (4.08) (3.73) (6.17) (8.48)ut-1 -1.089 -0.975 -0.595 -0.896(-2.36) (-1.85) (-1.40) (-1.66)ut-2 0.780 -0.002 0.169 -0.870(1.67) (-0.01) (0.39) (-0.85)D (constantin 0.141 0.047 -1.126 -0.906coefficient on (0.33) (0.10) (-1.79) (2.03)expectations)E (coeff. of rL 653.85 557.36 578.51 451.83in coefficient on (3.73) (2.28) (2.63) (3.46)expectations)

    ?0

    0?

    0000~00

    This content downloaded from 203.170.79.82 on Tue, 1 Oct 2013 03:27:36 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 4121606

    14/23

    Table 1. (Continued)Method for Geometrically Geometrically Geometrically 16-quarterMAconstructing7rL declining declining declining with differentweights, eq. (9) weights, eq. (9) weights, eq. (8) weights for eachyear, eq. (10)Method for 16-quarterag 16-quarterMA Survey data Survey dataconstructing7re with different with differentweights for each weights for eachquarter,eq. (10) year, eq. (10)No. of unempl. lags 2 12 2 12Sample period 1963:1-2000:2 1963:1-2000:2 1979:3-2000:2 1979:3-2000:2LSURI 0.0254 0.0281 0.0376 0.0401LSUR 0.0208 0.0194 0.0209 0.0296Natural rateu(7r= 0.0) 0.044 0.047 0.076 0.065u(ir= 2.0) 0.022 0.022 0.038 0.041u(7r= 4.0) 0.029 0.025 0.021 0.030u(7r= 6.0) 0.043 0.043 0.052 0.043

    (7r= 0.0) 0.556 0.519 0.130 0.1824(7r = 2.0) 0.655 0.606 0.185 0.234D(7r= 4.0) 0.882 0.826 0.420 0.427D(7r= 6.0) 0.994 0.980 0.831 0.764DW statistic 1.518 1.506 1.002 1.150R2 0.879 0.873 0.933 0.948aAsymptotic-valuesn parentheses.etailed esults orall regressionsreavailablen request.

    ?

    'a

    a

    This content downloaded from 203.170.79.82 on Tue, 1 Oct 2013 03:27:36 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 4121606

    15/23

    410 P Lundborgand H. Sacklenthe Appendix.The LSURIdoes not changein any significantway whenopen unemployments replacedby total unemployment r unemploymentamong prime-agedmales. In some regressionswe excludedthe turbulent1990s. This shifted the LSURs downwardswhile the LSURIswere notsignificantlyaffected.Direct Measures of Expected InflationIt is far fromobviousthatestimatedadaptiveexpectations apturehouse-holds' trueexpectationson inflation.To illustrate his point,Figure2 con-trasts he expectationsmpliedby regression i) in Table1 to surveydataonhouseholds'expectationswhichare available orthe period1979:3-2000:2.The estimatedseries follows the CPI quite closely. It is noteworthy hatthere are apparentdeviationsfrom the directlymeasured nflationexpec-tations. Giventhatthe surveydatabetterreflecthouseholds' rueinflationexpectations,one might suspectthat the resultsobtainedabove could bemisleading.The results for some of the regressionsusing our 1979-2000 surveydataarereportedn Table1, columns(iii) and (iv). A vital observationsthat the changeof data does not cause a rejectionof the overallresultsdiscussed above.The parameterE is still, withoutexception,significantlypositive. Hence, as predictedby theory,the reactioncoefficient of priceinflation to expectedinflation(D(.)) is significantlysmallerin low- thanin high-inflationperiods.Figure3 shows that the averageLSURI usingsurveydata(4.01 percent) s somewhathigherthanthe averageobtainedwith estimatedexpectations 2.61 percent).Similarly, he averageLSUR

    16 -12 -- - , -.-

    8 - --I

    0 I I

    1960 1970 1980 1990 2000....... CPIEstimatedxpectationsSurvey xpectations

    Fig. 2. CPIandexpectedinflation? The editors of the Scandinavian Journal of Economics 2006.

    This content downloaded from 203.170.79.82 on Tue, 1 Oct 2013 03:27:36 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 4121606

    16/23

    Low-inflationtargetingand long-run unemployment 411

    65-

    -4------------- A --------m3 ------------1-01I I0 1 2 3 4

    LSUR* Estimatedxp.1963-2000* Survey xp.1979-2000A Survey+ imputedxp.1963-2000

    Fig.3. Inflationndminimumnemploymentincreaseswhenwe substitutedurveydatafor estimatedexpectations,rom2.08 to 2.85.One disadvantage f using the relativelyshortsurveydataseries is thatwe areunableto utilize the full lengthof the dataon inflationandunem-ployment.Moreover,his raises the questionof whetherwe have too fewobservationsand to whatextentthe resultsof these regressionsare influ-enced by the particular ample period.To addressthese issues, we nextimputed urveydata for the period1963:1-1979:2 n order o obtaina fullsurveydataseriesfor 1963:1-2000:2.15nTable1,columns v) and(vi), weshowdetailedresults fromsome regressionsbasedon this extendedsurveydataset.The plot in Figure3 illustrates he most important esult,namelythatthe LSURIsareverysimilar o those obtainedusingthe originalsurveydata.Hence,the LSURIsappearnot to be affectedby the sampleperiod.To give an idea of what the individualPhillips curves look like,Figure4 displaystwo representativeurves from the regressionswherewe used surveydata on inflationexpectations.According o these curves,inflationneeds to be around4 percent n orderto be compatiblewith thelowestsustainable ateof unemployment. ote also that a monetarypolicyaimingat pricestabilityappears o be accompanied y largecosts in termsof highunemployment,round7 percent,which shouldbe comparedo the2-2.5 percentassociatedwith 4 percent nflation.

    Aretheseresultsconsistentwiththerecentmacroeconomic evelopment?By the end of 2001, the SwedishRiksbankhad succeededin stabilizing15 We fitted a regressionmodelthatdetermines ur surveydataas a non-linear unctionoflaggedCPIandpredicted series for the period1963:1-1979:2.

    OThe editors of the Scandinavian Journal of Economics 2006.

    This content downloaded from 203.170.79.82 on Tue, 1 Oct 2013 03:27:36 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 4121606

    17/23

    412 P Lundborgand H. Sacklen108

    c 6==42-

    0 2 4 6 8 10Unemployment- Survey xp. 1979-2000Survey+ imputed xp. 1963-2000

    Fig.4. Surveyata:wo epresentativehillipsurvesTable,columnsiv)andv))inflationat the 2 percent arget.16It is noteworthyhatunemploymenthenalso stabilizedandremained table for 24 consecutivemonthsat around4percent, .e., at the level predictedby our surveydataregressions.17Other Objectives than Minimum UnemploymentOne importantmplicationof the ADP model is thatproductivity hangeswith the rate of inflation.Consequently,here is no a priori reasonwhymaximumoutputand minimumunemploymentwould coincide at someinflationrate. Akerlof et al. largelyneglectedthis aspect of the model.Lundborg nd Sacklen(2003) discussthe implicationsof the ADP modelforthe relationbetween nflationand variablesotherthanunemployment-like output-that may be of importanceo welfare.Let outputbe definedas Q= (1 - u)e, where e is a weighted averageof effort in rationalandnear-rationalirms.18Lundborgand Sacklen (2003) then show that thetheoreticalmodeldoesnotrule outan effort-inflationrade-offat low infla-tion rates.Consequently,ince employmentandeffortmay go in different16An inflationtargetof 2 percent(with a lower and upperbound of 1 and 3 percent,respectively)was announced n 1993, and the SwedishRiksbankassessedthat the targetwould be met during he 12 monthsof 1995.17Unemploymentncreasedrom2003 to 2005 and was somewhere etween5 and6 percentin 2005. Inflation,on the other hand,droppedduringthe period 2003-2005, to around0.5 percent.1 Note that this rathercrude production unctionignores potentially mportant ong-runeffects of inflationon output hroughsavingsand investment,he qualityof price signals,etc.C The editors of the Scandinavian Journal of Economics 2006.

    This content downloaded from 203.170.79.82 on Tue, 1 Oct 2013 03:27:36 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 4121606

    18/23

    Low-inflation argetingand long-run unemployment 413directionswheninflation ncreases, hetheoreticalmodelcannotruleoutanoutput-inflationrade-off.In fact, minimumunemployment nd minimumoutputcan occurat the same rateof inflation. n this situation,governmentdecision-making bviouslybecomes more complicated hanjust directingmonetarypolicy toward he inflationratethatminimizesunemployment.What do our regressions or Swedenimply aboutthe relationbetweenoutputandinflation?Calculationsn Lundborg nd Sacklen 2003) suggestthat maximumoutputand minimumunemployment ccur at roughlythesamerate of inflation,slightlybelow4 percent.19IV. ConcludingDiscussionAkerlof et al. (2000) stronglyrejected he vertical ong-runPhillipscurvewhentheirmodel was applied o U.S. data.The fact thatour Swedishdataalso reject the idea of a vertical curve has a numberof interesting m-plicationsfor the Swedish economyand, potentially, or other Europeancountriesas well. A major consequence s that the level of the inflationtargetreally matters;almost all our regressionssuggest that an inflationrate below 2 percent s associatedwith largecosts in termsof unemploy-ment. There is more uncertainty oncerningthe inflationrate that min-imizes unemployment.Our estimatesvary across specifications,but thepoint estimatesof our preferred egressions20 suggestthat a doublingofinflation,from 2 to 4 percent,could restrictlong-rununemploymento1.5-3.5 percent.The SwedishRiksbank's percent nflation arget mpliesan unemploy-ment level above the lowest sustainable ate.The fact thatactualinflationsince the inceptionof the target n 1995 averages1.3percentsuggeststhatmonetarypolicy has been highlycontractive.

    19See Lundborg ndSackl6n 2003), Section2.3 andFigure7. The calculations xplore heestimates ortheregressionbasedon surveydata1963-2000that s shown n Figure4-andcolumn(v) in Table 1-of the presentpaper.Using the estimatesfor the U.S. in one ofADP's representative hillipscurve regressions, eported n the first columnof Table 2in Akerlof et al. (2000, p. 32), the calculationsn Lundborg nd Sackl6n 2003) suggestasimilarpattern or the U.S., namelythatthe output-maximizingateof inflationappears obe very close to the unemployment-minimizingate of inflation.It shouldbe stressedthatthese calculations ely on point estimates,and they have not been subjectto a sensitivityanalysiswhere the parameters'tandard rrorsare accounted or.20Werefer o the regressionsbased on surveydataon inflation xpectations s ourpreferredregressions.Thereasonfor this is simplythatsurveydatacan be expected o reflect house-holds'true nflation xpectationsn a betterwaythandifferentndirect stimation rocedures.Note also that our basic strategyof estimatinga wide rangeof specificationshas led us todisregardhe calculation,andpresentation, f detailedconfidenceintervals or the Phillipscurvesimpliedby our estimates.Hence,while our regressionsalmostuniformly ejecttheverticalcurve, here s still someuncertaintybout heexact locationandshapeof the curves.? The editors of the Scandinavian Journal of Economics 2006.

    This content downloaded from 203.170.79.82 on Tue, 1 Oct 2013 03:27:36 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 4121606

    19/23

    414 P Lundborgand H. SacklknFrom a policypointof view it could,of course,be argued hatthe ADPmodelis specifiedwithoutcoststhatwouldmakeanincreaseof the Swedishinflation arget rom2 percent o, say,4 percent ess desirable.Changingheinflationtargetmay involveshort-run djustment osts as it takestime toestablishcredibilityor a newtarget.Perhapsmoreimportantly,uranalysisis set forth in terms of minimumunemployment atherthan maximumoutputand,as such, it ignores potentially mportant ffects of inflationongrowth.21The possibilityof ADP-typelong-runPhillips curves across the eurocountriesas well may raise some concernaboutthe EMU project.Whiledetailed studies on other countriesare needed, there is nothingto sug-

    gest that these non-verticalPhillipscurves would not differ considerablyacross the euro countries.22Any single inflation level targetedby theECB would then generateexcess unemploymentn individualmemberstates.Could it be that the supportfor the ADP-typePhillips curve simplyreflects a misspecifiedmodel whereshifts of a verticallong-runPhillipscurve (due to institutional hanges)have not been fully accounted or?23Ourresultssuggestthat this is not the case. The robustnessests discussedin SectionIIIand theAppendix ndicate hata rightwardhiftof thePhillipscurvemay indeedhave occurredduringthe 1990s;see e.g. FigureA2 inthe Appendix.However, he basic hump hapeof the ADP-typePhillipscurveremains ntact.24 imilarly,heADP-typePhillipscurvegains supporteven if we dropthe 1960s and 1970s fromour data-see e.g. Figure3 andcolumns(iii)-(iv) in Table 1.21 Inthe literaturehere s a greatdeal of disagreementn possible sand r grease ffectsof inflation.No firm conclusionsaboutthe effects of inflationon growthcan be drawn.Some studiesindicatea large negativeassociationbetweeninflationand growth;see e.g.Cukierman,Kalaitzidakis, ummers nd Webb 1993), Fischer 1993) or BrunoandEasterly(1998). Otherstudiesraise doubtsaboutthe causalityand have argued hatthe effects ofinflationon growtharenot convincing; ee e.g. Romer 2001, p. 552). Changes n inflationare often the endogenouseffect of economicreformsthat could well be the reason for anegativessociationetweennflationndgrowth.22 f, forinstance,heincidencef editingdiffers cross irmsize,industrytructure,tc.,the trade-offwoulddiffer across countries.23 Inthe 1960s,Swedenexperiencedmoderatenflationandlow unemployment, hereas he1970s and 1980s witnessedhigh inflationand moderateunemploymentates.Followingamajormacroeconomichockntheearly1990s,unemploymentose ounprecedentedevels.Despite the severityof the shock, it has frequentlybeen arguedthat the unemploymentincreasealso could be the result of changesin labormarket nstitutions,.e., the result ofrightwardhiftsof a vertical ong-runPhillipscurve.24Excluding he 1990s, these particular egressionshave a limitednumberof observationson inflation n the 0-3 percent nterval.This, of course,makesit difficultto identifytheshapeof the Phillipscurvebelowthe LSURIwithoutrelyingon functionalormrestrictions.? The editors of the Scandinavian Journal of Economics 2006.

    This content downloaded from 203.170.79.82 on Tue, 1 Oct 2013 03:27:36 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 4121606

    20/23

    Low-inflation argetingand long-rununemployment 415

    Except for the NAIRUmodel with its verticalcurve, no other knownmodel emergesas a special case of the ADP model. This makes it dif-ficult to assess the risk thatdata are in fact generatedalong the lines ofsome other model (associatedwith a non-verticalong-runPhillipscurve),in whichcase oursupport or the ADP-typePhillipscurvewouldhave fol-lowedmainlyfromfunctionalmisspecification. n a preliminary ttempt oconstructa richermodel,Akerlofet al. (2000)presentregressionshatalsoinclude a term for nominalwage rigidity.These suggest a role for bothnear-rationalitynd nominalwage rigidity.25Hence, it shouldbe stressedthatthe rejectionof the NAIRU model in this paperin no way excludesthe relevanceof othermodelsthatcan yield non-verticalong-runPhillipscurves.A promisingarea for futureresearchwould be to stringentlyderivemodelsthatallow us to test the relativestrengthof various heories.AppendixDataCPI (ConsumerPrice Index): 1959:1-2000:2 from StatisticsSweden.Ourquarterlydataarecalculated s arithmeticverages f themonthlyigures.IPI (ImportPrice Index): 1963:1-2000:2 from Statistics Sweden. This index reflectsthepricesof goods mportedo Sweden.Ourquarterlyataarecalculateds arithmeticaverages f themonthlyigures.Importhares: 1963-2000 romStatistics weden.Foreachyearwe calculatehevalueof goodsandservices mportedo Sweden s a shareof GDPatmarket rices.Wethenassign he same mport hare o eachquarter.Surveydata on expected inflation: 1979:3-2000:2 from the National Instituteof Eco-nomic Research.Quarterly ataon households' xpectationsn CPI one yearahead,collected veryquarter.Unemployment:959:1-2000:2romStatistics weden LaborForceSurveys).Season-allyadjusted ataon openunemployments a shareof the labor orce(aged16-64).Totalunemployment:965:1-2000:2 rom StatisticsSweden(LaborForceSurveys).Seasonally djusted ataon open unemploymentlus workersn active abormarketprogramsaged 16-64).Male unemployment:1959:1-2000:2 from Statistics Sweden (Labor Force Surveys).Seasonally djusted ataon openunemploymentormenaged25-54.

    25 See Akerlof et al. (2000, pp. 30 and37). As noted in the introduction,therarguments,like nominalwage rigidity,can give rise to a negativelysloped long-runPhillipscurve.Forexample,the curves derivedfromnominalwage rigidityin Akerlofet al. (1996) andHolden 2004)have a negatively lopedsegmentat low inflationanda vertical egmentwheninflation s sufficientlyhigh (i.e., the shapeof aninvertedJ ).? The editors of the Scandinavian Journal of Economics 2006.

    This content downloaded from 203.170.79.82 on Tue, 1 Oct 2013 03:27:36 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 4121606

    21/23

    416 P Lundborgand H. SacklenDummy ariables:Dl = 1 for 1970:3-1970:4, zero otherwiseD2 = 1 for 1973:1-1974:1, zero otherwiseD3 = 1 for 1974:3, zero otherwiseD4 = 1 for 1975:3, zero otherwiseD5 = 1 for 1979:1-1980:1, zero otherwiseD6= 1 for 1980:2-1981:3, zero otherwiseD7 = 1 for 1981:4-1983:3, zero otherwiseD8 = 1 for 1986:1-1986:4, zero otherwiseD9= 1 for 1990:1-1991:2, zero otherwiseDIO= 1 for 1995:3-1996:2, zero otherwise

    Table Al. EstimatedspecificationsEstimated Surveydataoninflation xpectations inflation xpectationsUnemploymentmeasure 1963-2000 1963-1990 1979-2000 1963-2000

    Open 24a 24a 6b 6bTotal 24a 6bMale 24a 6baAs discussed in Section III,we use four alternativemeasuresfor 7re, three measuresfor 7rL,andtwo alternativelag lengths for unemployment.This gives us 24 possible combinations.bThreemeasures for L7rLnd two alternative ag lengths for unemploymentgive us six possible combinations.

    RobustnessUsing estimatedexpectations,we first exploredthe relationshipbetween inflation andtotal, rather than open, unemployment. Since government expenditures affect open

    1614-12-10-

    .9 8S6-4-

    2

    -2 2 4 6 8 10Unemployment

    Fig. Al. Swedish inflationandunemployment: uarterlydata, 1963-2000C The editors of the Scandinavian Journal of Economics 2006.

    This content downloaded from 203.170.79.82 on Tue, 1 Oct 2013 03:27:36 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 4121606

    22/23

    Low-inflationtargetingand long-run unemployment 417

    1963-1991 1963-2000108 *- *4

    --S4 - . .. .

    2-0

    0 2 4 6Unemployment

    Fig. A2. The Phillipscurve:1963-1991 and 1963-2000unemployment,and also change over the business cycle, it could be arguedthat totalunemploymentbettercapturesthe relevant labor marketsituation.We thereforewantedto see if the LSURI we obtained based on open unemploymentwould also yield alowest sustainable otal unemployment.Moreover,using total unemploymentmight sim-plify a comparisonof our results to those obtainedby ADP for the U.S. since openunemploymentwould be considerablyhigher in Sweden in the absence of labormarketprograms.Results from these regressionscan be found in Lundborgand Sacklen (2003). TheLSURI does not change in any significantway when open unemployments replacedbytotal unemployment.However,the average LSUR increases from 2.08 percentto 3.89percent.While this figure is considerablycloser to the LSURs estimatedby ADP,it stillfalls short of what appears o be the most common LSURs in their study.26Forsake ofcomparisonwith the resultsof ADP,we also ransome regressionsbased on prime-agedmales. The regressionsbased on open unemploymentamong males aged 25-54 yieldconsiderablylower LSURs, but the correspondinginflation rates do not differ muchfrom those obtainedpreviously.

    The turbulentSwedish labor market of the 1990s involved a majorincrease in un-employmentand a drasticdecrease in inflation. It thereforeseemed reasonableto testwhetherexclusion of the 1990s wouldyield very differentresults.In FigureA2 we showthe average Phillips curve for 1963:1-1991:2 and the Phillips curve for the period1963:1-2000:2 (see also Figure Al). Excluding the 1990s implies a leftwardshift ofthe curve:at just a slightly higher inflation ratethan for 1963-2000, the averagelowestsustainableunemploymentrate is now as low as 1.26 percent.This result suggests thatinclusion of the turbulent1990s in the data has some effect on LSUR, while the infla-tion rate remains stable. The observed change in Figure A2 could indicate that someparametershifts may have occurredduring the 1990s, which our model is unable tocaptureaccurately.26ADP neverreportan average stimateof theirLSURsto whichwe maycompareourestimates.However,he majorityof their estimatedLSURsapparentlyxceed4 percent.

    ? The editors of the Scandinavian Journal of Economics 2006.

    This content downloaded from 203.170.79.82 on Tue, 1 Oct 2013 03:27:36 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 8/13/2019 4121606

    23/23

    418 P Lundborgand H. SacklinReferencesAgell, J. and Lundborg,P. (1995), Theoriesof Pay and Unemployment: urveyEvidencefrom Swedish Manufacturing Firms, Scandinavian Journal of Economics 97, 295-307.Agell, J. and Lundborg,P. (2003), SurveyEvidenceon WageRigidityand Unemployment:Sweden in the 1990s, Scandinavian Journal of Economics 105, 15-29.Akerlof, G. A., Dickens, W T. and Perry,G. L. (1996), The Macroeconomics f LowInflation, Brookings Papers on Economic Activity 1, 1-59.Akerlof,G. A., Dickens,W.T andPerry,G. L. (2000),Near-rationalWageand PriceSettingand the Long Run Phillips Curve, Brookings Papers on Economic Activity 1, 1-44.Blinder,A. S., Canetti,E. D., Lebow,D. E. andRudd,J. B. (1998),AskingAboutPrices:ANew Approach to Understanding Price Stickiness, Russell Sage Foundation,New York.Bruno,M. andEasterly,W.(1998), InflationCrises andLong-runGrowth, ournalof Mon-

    etary Economics 41, 3-26.Bullard,J. and Keating,J. W. (1995), The Long Run RelationshipbetweenInflationandOutput in Post-war Economies, Journal of Monetary Economics 36, 477-496.Chen,P.andEdin,P.A. (2002), EfficiencyWagesandIndustryWageDifferentials:A Com-

    parison across Methods of Pay, Review of Economics and Statistics 84, 617-631.Cukierman,A., Kalaitzidakis,P., Summers,L. H. and Webb,S. B. (1993), CentralBankIndependence, Growth, Investment, and Real Rates, Carnegie-Rochester Conference Serieson Public Policy 39, 95-140.Dickens,W T. (2001), Commentson CharlesWyplosz's Do We KnowHow Low InflationShouldBe? , mimeo,Brookings nstitution,Washington,DC.Djoudad,R. andSargent,T.C. (1997), Does the ADP StoryProvidea BetterPhillipsCurvefor Canada?, npublishedworkingpaper,EconomicStudiesandPolicyAnalysisDivision,Department f Finance,Ottawa.Eckstein, 0. and Brinner, R. (1972), The Inflation Process in the United States, Joint Eco-nomic Committeeof the Congressof the United States, GovernmentPrintingOffice,Washington,DC.Fair,R. (2000), Testing he NAIRU Modelfor the UnitedStates,Reviewof EconomicsandStatistics 82, 64-71.Fischer,S. (1993), The Role of MacroeconomicFactors n Growth,Journalof MonetaryEconomics 32, 485-512.Fortin,P. (2001), InflationTargeting:The ThreePercentSolution,Policy Matters,Institutefor Research n PublicPolicy,Montreal.Holden, S. (2004), The Costs of Price Stability-DownwardNominal Wage Rigidity inEurope, Economica 71, 183-208.King, R. G. and Watson,M. W. (1994), The Post-warU.S. PhillipsCurve:A RevisionistEconomic History, Carnegie-Rochester Conference Series on Public Policy 41, 157-219.Lundborg, .andSacklen,H. (2003),Low-inflation argetingndUnemploymentersistence,WorkingPaperno. 188, FIEF,Stockholm.Romer,D. (2001),AdvancedMacroeconomics,nd edn., McGraw-Hill,New York.Samuelson,P.A. andSolow,R. M. (1960),Problems f AchievingandMaintaining StablePrice Level: AnalyticalAspects of Anti-inflationPolicy,AmericanEconomicReview,Papers and Proceedings 50, 177-194.Schultze,C. L. (1959), Recent Inflation n the UnitedStates,StudyPaperno. 1, JointEco-nomicCommittee f the 86thCongress,Government rintingOffice, Washington,DC.Tobin,J. (1972), InflationandUnemployment, mericanEconomicReview62, 1-18.

    Firstversion submittedMay2003;final versionreceivedDecember2005.

    ? The editors of the Scandinavian Journal of Economics 2006.