interest and inflation rates through the lens of the...
TRANSCRIPT
– 5 –
sveriges riksbank economic review 2015:2
* WewouldliketothankClaesBerg,PaoloBonomolo,DavidKjellberg,HannaKöhler,EricM.Leeper,JesperLindé,MarianneSterner,DavidVestinandAndersVredinforvaluablediscussionsandsuggestions.Theopinionsexpressedinthisarticlerepresenttheauthors’personalopinionsandcannotberegardedasanexpressionoftheRiksbank’sview.
InterestandinflationratesthroughthelensofthetheoryofIrvingFisherMagnus Jonsson and andré reslow*MagnusJonssonhasaPhDinEconomicsandAndréReslowhasaMaster’sDegreeinEconomics.BothworkintheMonetaryPolicyDepartmentattheRiksbank.
The nominal interest rate and inflation are positively correlated with each other in the
short run and in the long run, in Sweden as well as in other countries. In this article, we
show that the positive correlation can be understood through Irving Fisher’s theory of the
relationship between the nominal interest rate, the inflation rate and the real interest rate.
In our analysis, the short run correlation can be explained by supply and demand shocks in
a standard macroeconomic model, where Fisher’s theory is a key factor. If we assume long
run monetary neutrality, Fisher’s theory implies that the correlation between the nominal
interest rate and inflation is positive also in the long run. Finally, we show that if the real
interest rate is equated to the GDP growth rate per capita, the long run implications of
Fisher’s theory fit the data in Sweden and several other countries. This provides empirical
support to the idea that a low policy rate over the long run could lead to low inflation, as
has been proposed by Narayana Kocherlakota, among others.
AlmostsevenyearshavepassedsincetheinvestmentbankLehmanBrothersdeclared
bankruptcyintheautumnof2008andtheglobalfinancialcrisisbrokeoutwithfullforce.The
crisisbroughtwithitthelargestfallinGDPsincetheGreatDepressionofthe1930s.Moreover,
therecoveryafterthecrisishasbeenunusuallyslowinmanycountries,seeFigure1(a).In
theeuroareaandJapan,GDPisataboutthesamelevelaswhenthecrisisbrokeout,while
inSweden,theUS,theUKandCanadathelevelisfivetotenpercenthigher.Additionally,
GDPpercapitaisatthepre-crisislevelinallthesecountries,seeFigure1(b).Swedenonlyjust
reachesthepre-crisislevelandboththeeuroareaandtheUKarebelowthatlevel.
Thecentralbanksreactedpromptlyandcutthepolicyratestonearzeropercentatthe
outsetofthefinancialcrisisinordertoavoidafurtherdeepeningofthecrisis,seeFigure2(a).
However,thepolicyrateshaveremainedattheselowlevels.InSweden,thepolicyrateis
currentlyslightlynegativeandinseveralothercountriesitisclosetozeropercent.
Inflationfellrapidlyatthebeginningofthefinancialcrisis,eventhoughitrecovered
relativelyquickly,seeFigure2(b).Forexample,inSwedeninflationwasabovetwopercent
around2011,butsincethenithasbeenfallingagainandremainedatlowlevels.Thelow
inflationratesintherecentyearsare,however,notjustaSwedishphenomenon,butare
beingexperiencedbyseveralothercountries.
– 6 –
sveriges riksbank economic review 2015:2
WHyISInFLATIonnoTRISInG?
Monetarypolicyhasthusbeencharacterisedbypolicyratesnearzeropercentsincethe
beginningofthecrisis.Inaddition,severalcentralbankshavepurchasedgovernmentbonds
andadoptedotherso-calledunconventionalmeasurestomakemonetarypolicyevenmore
expansionary.Atthesametime,inflationhasbeenlowandbelowtheinflationtargetin
80
85
90
95
100
105
110
115
00 02 04 06 08 10 12 14
(a) GDP
Sweden United States Euro area Japan United Kingdom Canada
80
85
90
95
100
105
110
115
00 02 04 06 08 10 12 14
(b) GDP per capita
Figure 1. GDP and GDP per capita in Sweden, the US, the euro area, Japan, the UK and Canada Index, 2007 = 100
Sources: Eurostat, Japanese Cabinet Office, Official Statistics of Japan, Statistics Sweden, Statistics Canada, UK Office for National Statistics, US Bureau of Economic Analysis and US Census Bureau
00 02 04 06 08 10 12 14
(b) Inflation rate
-4
-2
0
2
4
6
00 02 04 06 08 10 12 14
(a) Nominal interest rate
-2
0
2
4
6
8
Sweden United States Euro area Japan United Kingdom Canada
Figure 2. Nominal interest and inflation rates in Sweden, the US, the euro area, Japan, the UK and CanadaPer cent
Note. The nominal interest rates are measured by three-month treasury bills, except in the euro area where the nominal interest rate is measured by EONIA. Inflation is measured by CPI inflation in all countries, except in the US where it is measured by PCE inflation, and the euro area where it is measured by HICP inflation. Note that the high rates of inflation in Japan in 2014 were mostly, 2 percentage points, due to temporary effects from increased taxation on consumption (the calculations in the article is adjusted for this).
Sources: ECB, Eurostat, Macrobond, OECD, Statistics Sweden and the US Bureau of Economic Analysis
– 7 –
sveriges riksbank economic review 2015:2
manycountries.Whatiscausingtheselowinflationrates?Whyhasmonetarypolicy,which
hasbeenexceptionallyexpansionaryinahistoricalperspective,notledtohigherratesof
inflation?
Acommonexplanationisthatmonetarypolicyhasbeenconstrainedbythezerolower
bound,i.e.thepolicyratecannotbesubstantiallybelowzeropercent.Proponentsofthis
viewarguethatdemandisstillbelownormalandisthereforenotcreatinganyupward
pressureoninflation.Inotherwords,monetarypolicyhasnotbeenexpansionaryenough.1
Anotherexplanationthathasbeenputforwardisthatincreasingcompetition,entailed
byglobalisationanddigitalisation,hasmadeitmoredifficultforfirmstoraiseprices
withoutlosingcustomers.2
CAnLoWPoLICyRATESoVERTHELonGRUnLEADToLoWInFLATIon?
Perhapsamorechallengingexplanationisthatlowpolicyratesoverthelongruncouldlead
tolowinflation.Thismaysoundcontradictory,aspolicyratecutsaccordingtostandard
theoryshouldleadtorisinginflation.nevertheless,narayanaKocherlakota,presidentofthe
FederalReserveBankofMinneapolis,exploredthisideainaspeech2010,seeKocherlakota
(2010):
Long run monetary neutrality is an uncontroversial, simple, but nonetheless
profound proposition. In particular, it implies that if the FOMC maintains
the fed funds rate at its current level of 0-25 basis points for too long,
both anticipated and actual inflation have to become negative. Why? It’s
simple arithmetic. Let’s say that the real rate of return on safe investments
is 1 percent and we need to add an amount of anticipated inflation that will
result in a fed funds rate of 0.25 percent. The only way to get that is to add a
negative number, in this case, −0.75 percent. To sum up, over the long run, a
low fed funds rate must lead to consistent, but low, levels of deflation.
Kocherlakota’sclaimisbasedontwowell-knowneconomictheories.Thefirstisthetheory
oflongrunmonetaryneutrality,whichsaysthatachangeinthepolicyratedoesnotaffect
longrunvaluesofrealinterestratesandotherrealvariables.ThesecondisIrvingFisher’s
theoryofthelongrunrelationshipbetweenthenominalinterestrate,theinflationrate
andtherealinterestrate.3Fisher’stheoryissimpleandintuitive:inthelongrun,inflation
hastocorrespondtothedifferencebetweenthelongrunlevelsofthenominalandthe
realinterestrates.Takentogether,thesetwotheoriesimplythatalong-livedchangeinthe
nominalinterestratecorrespondstoanequalchangeininflation.
1 SeeHall(2014).2 SeeApeletal.(2014).SeealsoJonsson(2007)whoquantifiestheeffectsofincreasedcompetitiononinflationin
theRiksbank’smacroeconomicmodelRamses.3 SeeFisher(1977).
– 8 –
sveriges riksbank economic review 2015:2
Centralbanksusechangesinashort-termnominalinterestrate(i.e.thepolicyrate)to
influenceinflation.4Buthowthesechangesaffectinflationdependonseveraldifferent
factors.5Kocherlakotaemphasisedthatthelengthofthechangemaybeanimportant
factor.Wethereforededicatealargepartofthearticletoexplainwhyshort-livedchanges
mayaffectinflationdifferentlythanlong-livedchanges.
oUTLInEAnDSUMMARy
Theoutlineofthearticleisasfollows:ThenextsectiondescribesIrvingFisher’stheoryof
therelationshipbetweenthenominalinterestrate,theinflationrateandtherealinterest
rate.Thissectionalsodescribestheassumptionsunderwhichtherealinterestratecanbe
equatedtotheGDPgrowthratepercapita.
Thefollowingsectionshedslightontheshortrunrelationshipbetweenthenominal
interestrateandinflation.Wefirstshowthatthecorrelationbetweenthesetwovariablesis
positiveinSwedenandothercountries.Thismayappearcontradictory,asshort-livedcuts
inthepolicyratenormallyleadtorisinginflationandincreasestofallinginflation,which
shouldimplyanegativecorrelation.However,thisfactneednotnecessarilyimplythat
thecorrelationinthedataisalsonegative.Changesinthepolicyratearegenerallynotan
importantfactorbehindthebusinesscyclefluctuations.Theseareinsteadmainlydrivenby
supplyanddemandshocks.Wethereforeconcludethissectionbyexplainingwhythese
shocksgiverisetoapositivecorrelationbetweenthenominalinterestrateandinflation.
Wethen,inthenextsection,lookatthelongrunrelationship.Alsothelongrun
correlationbetweenthenominalinterestrateandinflationispositiveinthedata.We
demonstratethatinastandardmacroeconomicmodel,whereFisher’stheoryandlong
runmonetaryneutralityarekeyassumptions,along-livedchangeinthepolicyrateleads
toanequalchangeininflation,whichexplainsthepositivecorrelationandalsoconfirms
Kocherlakota’sclaim.
Inthesubsequentsection,wemakeuseofthefactthatundercertainassumptions
therealinterestratecanbeequatedtotheGDPgrowthratepercapita.Weshowthat
theaverageinflationinSwedenandothercountriescanbeexplainedbythedifference
betweentheaveragenominalinterestrateandtheGDPgrowthratepercapita.This
suggeststhatthelongrunimplicationsofFisher’stheoryfitthedatainseveralcountries.
Furthermore,itprovidesempiricalsupporttoKocherlakota’sclaim.
Japanwasbadlyhitbyafinancialcrisisinthebeginningofthe1990s.TheJapanese
experienceinthepost-crisisperiodisofparticularinterest,sinceJapan'seconomyhas
beencharacterisedbylownominalinterestratesandinflation.Weshowthattheaverage
4 Thetermsnominalinterestrateandpolicyrateareusedinterchangeablyinthearticle.Intheempiricalanalysis,thenominalinterestrateismeasuredbytheyieldonathree-monthtreasurybill.Thedifferencebetweenthisinterestrateandthepolicyrateisnormallymarginal.
5 MiltonFriedmansaidthatmonetarypolicyworkswithlongandvariablelags,bywhichhemeantthattheeffectsoninflationofchangesinmonetarypolicytakestimeandvariesovertime,seeFriedmanandSchwartz(1963).
– 9 –
sveriges riksbank economic review 2015:2
inflationinJapan,bothinthepre-crisisandpost-crisisperiods,canbeexplainedbythe
differencebetweentheaveragenominalinterestrateandtheGDPgrowthratepercapita.
Followingthis,thereisasectionthattakesamuch-debatedarticle,Bullard(2010),as
astartingpointforadiscussionoftheeffectsoflownominalinterestratesoveralonger
period.6Bullard’sarticleexplainswhyFisher’stheorytogetherwiththezerolowerbound
mayimplyasteadystatecharacterisedbynear-zeronominalinterestratesandlowor
negativeinflation.7ThedatasuggestthatitisstilltooearlytodeterminewhetherSweden
andothercountrieshavemovedintosuchasteadystatefollowingthefinancialcrisis.Seven
yearsofdataisnotenoughtodeterminethis.ontheotherhand,itappearsasifJapan,
followingthefinancialcrisisatthebeginningofthe1990s,hasmovedintoasteadystateof
lowinflation.Finally,wemakesomeconcludingremarks.
Therelationshipbetweenthenominalinterestrate,theinflationrateandtherealinterestrateaccordingtoIrvingFisher
IrvingFisher’stheoryoftherelationshipbetweenthenominalinterestrate,theinflation
rateandtherealinterestrate,theso-calledFisherrelation,isfundamentalinmonetary
theory.Thisrelationisalsoakeyfeatureinthemicro-foundedmacroeconomicmodelsthat
centralbanksuseintheirforecastingandpolicywork.TheRiksbank’smodelRamsesisan
exampleofsuchamodel.8
TheFisherrelationisformallyanarbitrageconditionbetweenarealandanominalasset
andcanbewritteninthefollowingway,
(1) 1 + Rt = (1 + Etπt + 1)(1 + rt),
whereR denotesthenominalinterestrate,π inflation,E expectations(Etπt + 1thusdenotes
expectedinflationintimet + 1thatanagenthasattimet),andr therealinterestrate.
AmathematicalderivationoftheFisherrelationcanbefoundinseveraltextbooks,see
forexampleWalsh(2003).Hereweexplaintheintuitionbehindtherelationthrougha
verbalreasoning.Supposethatwehavearealassetthatcostsoneappleinperiodt and
thatgivesthereturnof(1 + rt) applesoneperiodlater.Innominalterms,therealasset
costsPtinperiodt andPt + 1inthenextperiod.Thenominalpriceoftherealassetin
periodt + 1isthus(1 + rt)Pt + 1andthenominalreturnwillbe((1 + rt)Pt + 1 – Pt) / Pt.In
ordertoavoidarbitrageopportunities,thereturnontherealassetmustbeashighasona
nominalasset,Rt,i.e.Rt = ((1 + rt)Pt + 1 – Pt) / Pt.Thisexpressioncanbere-writteninterms
ofinflationrates,1 + Rt = (1 + πt + 1)(1 + rt). Ifwealsotakeintoaccountthatthefutureprice
levelisunknown,wegetequation(1).
6 SeeBullard(2010).7 Bullard’sarticlebuildsoninsightsfromBenhabibetal.(2001).8 SeeChristianoetal.(2011).
– 10 –
sveriges riksbank economic review 2015:2
Forsmallvaluesoftheinflationandrealinterestrates,equation(1)canbeapproximated
withthefollowingexpression,
(2) Rt = Etπt + 1 + rt.
Equation(2)isthe”shortrun”Fisherrelation,sinceitisvalidateachpointintimet .In
macroeconomicmodels,atimeperiodisusuallyaquarter.TheFisherrelationisalsovalidin
thelongrun,i.e.insteadystate,whereitiswrittenas,
(3) R‾ = π‾ + r‾ ,
whereabarindicatesthatitisalongrun(steadystate)valueofavariable.Longrunvalues
areusuallycalculatedastheaverageoveralongerperiod,usuallytenyearsormore.
THEREALInTERESTRATECAnBEEqUATEDToTHEGDPGRoWTHRATEPERCAPITAUnDER
CERTAInASSUMPTIonS
Thefollowingfactorscanbeshowntodeterminethelongrunrealinterestrateunder
certainassumptionsonthehouseholds’preferencesandthefirms’production
technologies,9
(4) 1 + r = 1β
(1 + γ )σ,
whereβ denotesthehouseholds’discountfactor,γ the(real)GDPgrowthratepercapita
(workingagepopulation)and1 / σthehouseholds’intertemporalelasticityofsubstitution.
Equation(4)showsthatthelongrunrealinterestratedependsontwofundamentalfactors.
Thefirstishowhouseholdsvalueconsumptiontodayversusconsumptiontomorrow,β,and
thesecondistheGDPgrowthratepercapita,γ .10
Explainingwhytherealinterestratedependsonβ isbestdonethroughanexample.
Assumethathouseholdsareimpatient,i.e.,theypreferconsumingtodayasopposedto
consumingtomorrow.Ifhouseholdsaretovalueconsumptiontomorrowashighlyas
consumptiontoday,theymustbecompensatedfordeferringtheirconsumption.11Thatisto
say,iftheysaveandpostponesomeoftoday’sconsumptionuntiltomorrow,theremustbe
apositiverealinterestrateonthissaving.
Inaddition,therealinterestratedependsonthegrowthrate.Themarginalutility
ofconsumptionisnormallydiminishing,i.e.asmallincreaseofconsumptionincreases
households’utilitybutatadiminishingrate.or,toputitdifferently,theutilityofeleven
applesisgreaterthanthatoften,butthemarginalutilityoftheeleventhappleisless
thanthatofthetenth.Inagrowingeconomy,theconsumptionleveltodayislowerthan
9 SeeJonsson(2002).10 ”TheGDPgrowthratepercapita”willhereafterbecalled”thegrowthrate”.11 Utilitymaximisationimpliesthathouseholdsareindifferentaboutconsumingtodayortomorrow.
– 11 –
sveriges riksbank economic review 2015:2
futurelevels,whichimpliesthatthemarginalutilityofconsumingtodayisgreaterthanthe
marginalutilityofconsuminginthefuture.Therealinterestratemustthereforebepositive
ifhouseholdsaretobeindifferentbetweenconsumingtodayandinthefuture.Howhigh
theinterestratehastobedependsontheintertemporalelasticityofsubstitutionofthe
households.Ifthewillingnessofsubstitutingconsumptionbetweenperiodsisrelativelylow,
i.e.ifσisrelativelyhigh,therealinterestrateneedstoberelativelyhigh.
notethataccordingtoequation(4),ifβ = σ = 1therealinterestrateequalsthegrowth
rate.Ifβ = 1thentoday’sgenerationwillvalueitsownconsumptionashighlyasfuture
generations’consumption,whichcouldbeareasonableassumptionfromtheperspective
ofjustice.Itmaybemoredifficulttojustifyacertainvalueforσ.Ithasalsoprovedtobe
difficulttoestimatethisparameterwithanycertainty.However,acommonvalueforthis
parameterinmanymacroeconomicmodelsis1,which,forexample,isthecaseinthe
Riksbank’smacroeconomicmodelRamses.Ifweassumethatβ = σ = 1thelongrunFisher
relationcanbedefinedasfollows,
(5) R‾ = π‾ + γ.
Thenominalinterestrateandinflationarepositivelycorrelatedintheshortrun
ThenominalinterestrateintheFisherrelationshouldaccordingtotheorybemeasuredby
ashort-termrisk-freenominalinterestrate.Giventhat,weusetheyieldonathree-month
treasurybillasameasureofthenominalinterestrateinSweden,theUS,Japan,theUKand
Canada.Fortheeuroarea,wheretherearenotreasurybillsissuedjointlybythemember
states,weuseEonIA(EuroovernightIndexAverage),whichisareferencerateforloans
betweenbankswithamaturityofonebankingday.Regardinginflation,thereareseveral
differentmeasurestochoosefrom.Althoughthereisnoindisputablemeasure,whichis
reflectedinthatcentralbanksusuallyreportseveralmeasuresintheirreports.Wefollow
thispracticeandreporttwomeasures.Thefirstisinflationaccordingtotheconsumerprice
index,i.e.CPIinflation.12Thesecondisameasureofso-calledcoreinflation.13
Figure3showsthecorrelationbetweenthenominalinterestrateandinflationinsix
countries:Sweden,theUS,theeuroarea,Japan,theUKandCanada.Thetimeperiodis
1961-2014forallcountries,excepttheeuroarea,whereitis1999-2014.Ascanbeseen,
thecorrelationispositivethroughout.ItisstrongestinJapan,withacorrelationcoefficient
of0.80,andweakestintheeuroarea,wherethecorrelationcoefficientis0.48.
12 However,fortheUS,weusePCEinflation(theconsumptiondeflator),whichisthepreferredmeasurebytheFederalReserve.
13 CoreinflationismeasuredbyCPIexcludingfoodandenergy,exceptinSweden,whereitismeasuredbyCPIFexcludingfoodandenergy,theUS,whereitismeasuredbyPCEinflationexcludingfoodandenergy,andtheeuroarea,whereitismeasuredbyHICPexcludingfood,energy,alcoholandtobacco.notethatmeasuresofcoreinflationarenotavailableforthe1960sand1970s.
– 12 –
sveriges riksbank economic review 2015:2
MonETARyPoLICySHoCKSRESULTInAnEGATIVECoRRELATIonBETWEEnTHE
noMInALInTERESTRATEAnDInFLATIon
Thepositivecorrelationbetweenthenominalinterestrateandinflationmayappear
contradictoryatfirstglance.Policyratecutsnormallyleadtorisinginflation,whileincreases
leadtofallinginflation,whichshouldimplyanegativecorrelation.However,thisdoesnot
necessarilyimplythatthecorrelationindataalsohastobenegative.Wecanexplainwhy
thisisthecasewiththehelpofasimplemacroeconomicmodel,inwhichtheFisherrelation
isakeyfeature.14
Thecentralbankfollowsasimple,linearruleàlaTaylor(1993)withthefollowing
parameterisation,15
(6) Rt = 0.8 Rt – 1 + (1 – 0.8) [R‾ + 1.5 (πt – π‾ ) + 0.1 (yt – y‾ )] + εt ,
whereRdenotesthepolicyrate,R‾ thepolicyrate’slongrunlevel,π inflation,π‾ inflation’s
longrunlevel(i.e.theinflationtarget),youtput,y‾ thelongrunoutputlevel,andεa
monetarypolicyshock.Thedifferences,πt – π‾andyt – y‾,arethusinflation’sdeviationfrom
theinflationtargetandtheoutputgap,respectively.
14 SeeMehandMoran(2010)andJonssonandMoran(2014)foradescriptionofthemodel.15 TheTaylorruleisasimplerecommendationforhowmonetarypolicyshouldbeconductedundernormal
circumstances.Therulehasalsobeenshowntoworkwellinmanydifferenttypesofmacroeconomicmodels,seePlosser(2008).
-2
0
2
4
6
8
10
12
14
16
-2 0 2 4 6 8 10 12 14 16
Nom
inal
inte
rest
rat
e
Inflation rate
Sweden, ρ = 0.67 United States, ρ = 0.71
Euro area, ρ = 0.48
-3
0
3
6
9
12
15
18
21
-3 0 3 6 9 12 15 18 21 24 27
Inflation rate
Japan, ρ = 0.80 United Kingdom, ρ = 0.60
Canada, ρ = 0.73
Figure 3. Correlation between the nominal interest rate and inflation in Sweden, the US, the euro area, Japan, the UK and CanadaPer cent
Note. ρ denotes the correlation coefficient. Quarterly data 1961-2014 for all countries, except the euro area, which is based on quarterly data 1999-2014. The nominal interest rates are measured by three-month treasury bills, except in the euro area where the nominal interest rate is measured by EONIA. Inflation is measured by CPI inflation in all countries, except in the US where it is measured by PCE inflation, and the euro area, where it is measured by HICP inflation.
Sources: ECB, Eurostat, Federal Reserve, Macrobond, OECD, Statistics Sweden and US Bureau of Economic Analysis
– 13 –
sveriges riksbank economic review 2015:2
Businesscyclefluctuationsareduetodifferenttypesofshockshittingtheeconomy,
towhichpricesandquantitiesadjust.Thisimpliesthatthecorrelationbetweendifferent
variableswilldependontheseshocks.Hence,inordertoexplain,forexample,the
correlationbetweenthenominalinterestrateandinflationitisnecessarytoknowwhich
shocksthathavecreatedthefluctuations.
ThemonetarypolicyshockintheTaylorrulemaybeinterpretedasachangeinthe
policyratethatisduetoneitherdeviationsofinflationfromthetargetnordeviationsof
outputfromitsnormallevel.Wecanthusillustratehowshort-livedchangesinthepolicy
rateaffectinflationbyallowingmonetarypolicyshockstodrivethebusinesscycle.Figure
4(a)showsthecorrelationbetweenthenominalinterestrateandinflationinthiscase.As
expected,thecorrelationisnegativeandthecorrelationcoefficientis–0.36.
0
2
4
6
8
0 1 2 3 4
(a) Monetary policy shocks, ρ = – 0.36
Nom
inal
inte
rest
rat
e
3.5
3.8
4.1
4.4
4.7
1.4 1.7 2.0 2.3 2.6
(d) Bank capital shocks, ρ = 0.68
3.0
3.5
4.0
4.5
5.0
1.0 1.5 2.0 2.5 3.0
Nom
inal
inte
rest
rat
e
(b) Supply shocks, ρ = 0.90
0
2
4
6
8
0 1 2 3 4
Inflation rate Inflation rate
(c) Demand shocks, ρ = 0.95
Figure 4. Correlation between the nominal interest rate and inflation after monetary policy shocks, supply shocks, demand shocks and bank capital shocksPer cent
Note. ρ denotes the correlation coefficient.
Source: Own calculations
– 14 –
sveriges riksbank economic review 2015:2
Animportantassumptioninthemodelisthatpricesarestickyintheshortrun,i.e.some
firmsarenotchangingtheirsellingpricesalongwithchangesindemand.Assumea
short-livedcutinthepolicyrateandthatthepricesaresostickythattheyinitiallyremain
unchanged.Thismeansthatthecut,Rt ↓,mustbematchedbyanequallylargefallinthe
realinterestrate,rt ↓,i.e.,
(7) Rt ↓= π‾ + rt ↓ ,
whereπ‾ indicatesthatinflationinitiallyremainsonthelongrunlevel.However,astime
passes,thefallingrealinterestratetendstoincreasehouseholds’consumptionandfirms’
investment.Thisboostsdemand,whichpushesupinflationwhenfirmseventuallystart
tochangetheirprices.Short-livedpolicyratecutsarethusassociatedwithrisinginflation
and,conversely,increasesareassociatedwithfallinginflation.Thisexplainsthenegative
correlationaftermonetarypolicyshocksinFigure4(a).
Buthowcanitthenbethatthenominalinterestrateandinflationarepositively
correlatedinthedatawhenmonetarypolicyshocksgiverisetoanegativecorrelation?This
issimplybecausemonetarypolicyshocksareingeneralnotanimportantcauseofbusiness
cyclefluctuations.Theseareusuallyduetoothershocks.
BUTSUPPLyAnDDEMAnDSHoCKSRESULTInAPoSITIVECoRRELATIon
Supplyanddemandshocksareoftenthemaindriversofthebusinesscycleand,hence,the
correlationsinthedata.Thesupplyshockismodelledasashocktothefirms’production
technology.Thisshockaffectstheproductionpossibilitiesofthefirmsandthereforehas
adirecteffectonthesupplyofgoodsandservices.Animprovementoftheproduction
technologymeansthatthefirms’productionpossibilitiesincrease,butitalsomeansthat
thecostofproductiondecreasesandthatthefirmscanlowertheirprices.Apositivesupply
shockthusleadstofallingprices.Thecentralbankreactstothisbycuttingthepolicyrate
tobringinflationbacktothetarget.Hence,thecorrelationbetweenthenominalinterest
rateandinflationbecomespositive.Thecorrelationcoefficientis0.90,seeFigure4(b).
Thedemandshockequalspublicconsumptioninthemodel,sincechangesinpublic
consumptionhavedirecteffectsonaggregatedemand.Anincreaseinpublicconsumption
pushesdemandupandthereforealsoincreasesinflationarypressures.Tokeepinflation
ontargetthecentralbankraisesthepolicyrate.Thisgivesrisetoapositivecorrelation
betweenthenominalinterestrateandinflation,seeFigure4(c),andacorrelation
coefficientof0.95.
Wehaveclaimedthatsupplyanddemandshocksarethemostcommonshocks,but
othershocksdoalsooccur.Thecreditcrunchassociatedwiththefinancialcrisisof2008is
anexampleofafinancialshock(aso-calledbankcapitalshock)wheretheeffectsspread
rapidlytootherpartsoftheeconomy.Thistypeofshockisunusualand,likethemonetary
policyshock,doesnotprovideanimportantexplanationofthecorrelationsindataover
– 15 –
sveriges riksbank economic review 2015:2
longerperiods.nevertheless,themacroeconomicmodelshowsthatthecorrelation
betweenthenominalinterestrateandinflationispositive,asinthedata,andmeasures
0.68whenthebanks’lendingisexposedtobankcapitalshocks,seeFigure4(d).
Thenominalinterestrateandinflationarepositivelycorrelatedalsointhelongrun
Figure5showsthelongruncorrelationbetweenthenominalinterestrateandinflation
infivecountries:Sweden,theUS,Japan,theUKandCanada.Thelongrunvalueshave
beencalculatedasten-yearaverages.Thetimeperiodcovers1965to2014,meaningfive
observationsforeachcountryandatotalof25observations.Aswecansee,thecorrelation
ispositivethroughoutandthecorrelationcoefficientis0.83.
ThebluelineinthefigureshowsthetheoreticalFisherrelation,inwhichwehave
calculatedthelongrunrealinterestrateastheaveragedifferencebetweenthenominal
interestrateandinflationforallcountriesovertheentireperiod.Thisgivesavalueof
1.56percent.TheredlineshowsanestimateoftheFisherrelation.Theestimatedslope
coefficientis0.86,whichisrelativelyclosetothetheoreticalvalueof1.Also,astatistical
testdoesnotrejectthenullhypothesisthattheslopeis1.16
0
2
4
6
8
10
12
14
16
-4 -2 0 2 4 6 8 10 12 14
Nom
inal
inte
rest
rat
e
Inflation rate
ρ = 0.83 Estimated Fisher relation Theoretical Fisher relation
Figure 5. Long run correlation between the nominal interest rate and inflation in Sweden, the US, Japan, the UK and Canada, ten-year averages, 1965-2014Per cent
Note. ρ denotes the correlation coefficient. The nominal interest rates are measured by three-month treasury bills. Inflation is measured by CPI inflation in all countries, except in the US where it is measured by PCE inflation. The blue line shows R–
t =1.56 + π– t. while the red line
shows R– t = 2.16 + 0.86π–
t.
Sources: Federal Reserve, Macrobond, OECD, Statistics Sweden, US Bureau of Economic Analysis and the Riksbank
16 AnF-testwiththenullhypothesisthattheslopeisoneintheestimatedFisherrelationgivesap-valueequalto0.06(F-value=3.79,covariance=0.01anddegreesoffreedom(1,23)).
– 16 –
sveriges riksbank economic review 2015:2
WecanseeinthefigurethattheestimatedFisherrelationliesrelativelyclosetothe
theoreticalrelation,butseveralisolatedpointsneverthelessliesomedistanceaway.This
canbeduetoseveralthings.Thelongrunrealinterestratemaynothavebeenthesame
inallcountriesovertheentireperiod.Itisalsolikelythatthelongrunrealinterestratehas
variedbetweentheten-yearperiodswithinthedifferentcountries.
ALonG-LIVEDPoLICyRATECUTIMPLIESTHATInFLATIonWILLBELoWERoVERTHE
LonGRUn
Wecanunderstandthelongruncorrelationbetweenthenominalinterestrateandinflation
throughtheimplicationsofthelongrunFisherrelation.Assume,likeKocherlakota,thatthe
realinterestrateisindependentofmonetarypolicyinthelongrun(monetaryneutrality)
andthatthecentralbankdeterminesthepolicyrate.Thissecondassumptionmeansthat
thecausalitybetweenthepolicyrateandinflationrunsfromthepolicyratetoinflation.
Letusassumethatthecentralbankcarriesoutalong-lived(inthisscenariopermanent)
cuttothepolicyrate,R‾ ↓,seeequation(8).Monetaryneutralityimpliesthatthereal
interestrateremainsunchangedatthelongrunlevel,r‾.Hence,thepolicyratecutonly
affectslongruninflation(i.e.theinflationtarget),whichisadjusteddownwards,π‾ ↓.Itis
adjusteddownwardsasmuchasthepolicyrateinordertoavoidarbitrageopportunities
betweennominalandrealassets.Along-livedpolicyratecutthusleadstolowerlongrun
inflationinaccordancewithKocherlakota’sclaim.Thisalsoexplainsthelongrunpositive
correlation,
(8) R‾ ↓ = π‾ ↓ + r‾.
notethedifferencetoashort-livedchangeinthepolicyrate.Aswehaveseen,inthiscase
thereisachangeintherealinterestrate,sincepricesinthemodelaresticky.Thismeans
thatthenominalinterestrateandinflation,intheshortrun,neednotnecessarilymovein
thesamedirection.
Themacroeconomicmodelcanbeusedtoillustratehowtherealinterestrateand
inflationmayadjustafteralong-livedpolicyratecut.Assumethatthelongrunlevelsofthe
policyandrealinterestratesareinitiallyfourandtwopercent,respectively.Inflationisthus
twopercent.Thecentralbankcutsthepolicyratesuccessivelytotwopercent,whereit
remains,seeFigure(6).Therealinterestrateinitiallyrisesslightly,butthenfallsbacktothe
longrunleveloftwopercent.Inflationfallssuccessivelyoverthewholeperiodtoitsnew
longrunvalue,whichaccordingtotheFisherrelationmustbezeropercent.
Anotablefeatureofthisscenarioisthatinflationsuccessivelyadjustsdownwardsto
itsnewlongrunlevel.Themainreasonforthisisthatagentsintheeconomyhavewhat
isknownasrationalexpectations.Thismeansthatthehouseholdsandthefirmsdonot
makeanysystematicerrorswhentheyformexpectationsoffuturemonetarypolicy.They
understandthatthepolicyratecutwillbelong-lived.Theyalsounderstandthatalong-
– 17 –
sveriges riksbank economic review 2015:2
livedcutisassociatedwithalowerinflationrateoverthelongrun.Thefirmsthereforestart
adjustingtheirpricesdownwardsatthemomentofthepolicyratecut.
-1
0
1
2
3
4
5
0 4 8 12 16 20 24 28
Quarter
Policy rate Inflation rate Real interest rate
Figure 6. The policy rate, the inflation rate and the real interest rate after a long-lived (in this scenario permanent) cut in the policy rate with two percentage pointsPer cent
Source: Own calculations
AccordingtothelongrunFisherrelationthereisonlyonelevelofthepolicyratethat
isconsistentwithaspecificinflationtarget,giventhelongrunrealinterestrate.Ifthe
policyratedeviatesfromthislevelduringalongerperiod,thereisariskthatagentsinthe
economymayinterpretthisasachangeintheinflationtarget.Theagentsmaytherefore
haveinterpretedthelong-livedcutinthepolicyrateasanintentiontolowertheinflation
targetfromtwotozeropercentinthisscenario.Inotherwords,ifthecentralbankhasan
inflationtargetoftwopercentandthelongrunrealinterestrateistwopercent,thepolicy
ratemustonaveragebefourpercentiftheinflationtargetistobeattained.Ifthepolicy
rateinsteadaveragestwopercent,theinflationrate,aswehaveseeninFigure(6),willon
averagebetwopercentagepointslower,i.e.zeropercent.
Thisscenarioalsoillustrateswhytheleveloflongrunrealinterestrateisimportant
tothecentralbank.Itmustbeawareofthislevelinordertosetapolicyratelevelthatis
consistentwiththeinflationtarget.Inthewakeofthefinancialcrisis,LarrySummersand
othereconomistshaveadvancedathesisofsecularstagnation.17Amongotherthings,they
arguethatthelongrunleveloftherealinterestratemayhavefallen.Ifthisiscorrect,the
longrunlevelofthepolicyratemustbeadjusteddownwardsinproportiontothefallinthe
realinterestrate,otherwiseinflationwillbetoohigh.
17 SeeSummers(2014).
– 18 –
sveriges riksbank economic review 2015:2
ThelongrunFisherrelationfitsthedatainSwedenandothercountries
Kocherlakota’sclaimthatlowpolicyratesoverthelongruncanleadtolowinflationrelies
onthelongrunFisherrelation.Inthissection,wecarryoutanempiricaltestofthisrelation.
Todoso,wecalculatetheinflationimpliedbythelongrunFisherrelationandcompareitto
theobservedaverageinflation.Inthesecalculationswemakeuseofthefactthatthereal
interestrateundercertainassumptionscanbeequatedtothegrowthrate,seeequation
(4).Hence,wecalculateinflationfromtheFisherrelationasthedifferencebetweenthe
averagelevelsofthenominalinterestrateandthegrowthrate.18Thisinflationrateiscalled
theFisherinflation,andisthusdefinedas,
(9) π‾ F = R‾ – γ.
TodeterminewhetherthelongrunFisherrelationfitsthedata,weshowthedifference
betweentheFisherinflationandtheobservedinflationfromdifferentcountriesinbar
chartsaswellasresultsfromastatisticaltest.notethattherealinterestrateinthis
test,whichisequatedtothegrowthrate,mayvaryfromcountrytocountry,unlikethe
calculationsinFigure5.
Totestlongrunrelationshipsthetimeperiodexaminedshouldbeaslongaspossible,at
thesametimeasthedatashouldnotdisplayanycleartrends.Itmayalsobeanadvantage
ifthemonetarypolicyregimeisthesameovertheentireperiod.Totakethisintoaccount,
wepresentresultsfromtwodifferentperiods.Thefirstlongerperiodislimitedbyaccessto
dataandstretchesfrom1961to2014.Forthisperiod,wehavedataforSweden,theUS,
Japan,theUKandCanada.
Thesecondshorterperiodis,totheextentpossible,characterisedbyastablemonetary
policyregimeandnocleartrendsinthedata.AlanGreenspanwasappointedChairmanof
theFederalReservein1987.TheperiodpriortoGreenspanwascharacterisedbybothhigh
nominalinterestratesandhighinflation.WethereforestartthesecondperiodfortheUSin
1987.FortheUK,CanadaandSweden,thesecondperiodstartsoneyearaftertheinflation
targetswereannouncedineachcountry,i.e.1992forCanada,1993fortheUKand1994
forSweden.Startingtheyearaftertheannouncementofinflationtargetingallowsusto
avoidincludingtheeffectsoftheeconomicandfinancialcrisisthathitthesecountriesat
thestartofthe1990s.Duringthe1960sand1970s,Japanunderwentatransitionfrom
relativelylowGDPlevelstolevelsinparitywiththoseofthedevelopedindustrialcountries.
TheshorterperiodforJapanthereforestartsin1981.Thisperiodalsoincludestheeuro
area.Theeurowasofficiallyintroducedon1January1999,whichiswhytheperiodstarts
in1999fortheeuroarea.
18 Hence,theempiricaltestdoesnotjustincludetheFisherrelationbutalsohowwellthelongrunrealinterestratecanbeapproximatedbythegrowthrate.Thereasonwemakethisapproximationisbecausethelongrunleveloftherealinterestrateisnotobservable.
– 19 –
sveriges riksbank economic review 2015:2
ThenominalinterestrateinSwedenaveragedjustoversixpercentbetween1961
and2014,whilethegrowthratewasjustovertwopercent,seeTable1.Thismeansthat
theFisherinflationwasaroundfourpercent,whichwasonlyslightlylowerthantheCPI
inflation,seeFigure7(a).FortheshorterperiodwecanseethattheFisherinflationwasin
linewiththeobservedinflation,i.e.boththeCPIandcoreinflation,seeFigure7(b).
IntheUS,thenominalinterestrateaveragedfivepercentandthegrowthrateaveraged
almosttwopercentovertheperiod1961-2014,seeTable1.TheFisherinflationwas
thereforeabout3.2percent.Thiswasjustafewtenthsofapercentagepointlowerthan
thePCEinflation,seeFigure7(a).TheFisherinflationalsocorrespondedwellwithboththe
PCEandcoreinflationfortheshorterperiod,seeFigure7(b).
IntheUK,theFisherinflationunderestimatedtheobservedinflationslightlyforthe
longerperiod,whileitoverestimatedtheCPIandcoreinflationslightlyfortheshorter
period,seeFigures7(a)and7(b).InCanada,theFisherinflationwasinlinewiththe
observedinflationinboththeshorterandthelongerperiods,seeFigures7(a)and7(b).
TheFisherinflationintheeuroareawaslowerthanboththeCPIandthecoreinflation,
althoughitwasnotthatfarfromthecoreinflation,seeFigure7(b).nevertheless,theresult
indicatesthattheFisherrelationseemstofitsomewhatlesswellintheeuroareathanin
theothercountries.Thiscouldbeduetoanumberoffactors.Thetimeperiodisrelatively
short,15years,andalmosthalfoftheseyearshavebeencharacterisedbythefinancial
crisis.Anotherproblemcouldbethatitdoesnotexisttreasurybondsissuedjointlybythe
memberstates,i.e.theEonIAratemayhavebeenapoorapproximationofashort-term
risk-freerateinseveralofthecountriesintheeuroarea.
Table 1. The nominal interest rate and the GDP growth rate per capita in different countries and time periodsPercentandannualpercentagechangerespectively,averages
CoUnTRy PERIoD THEnoMInALInTERESTRATE THEGDPGRoWTHRATEPERCAPITA
Sweden 1961-20141994-2014
6.33.3
2.22.0
UnitedStates 1961-20141987-2014
5.03.6
1.81.5
Euroarea 1999-2014 2.2 1.2
Japan 1961-19801981-19901981-20141991-2014
8.56.12.51.0
5.43.71.70.9
UnitedKingdom 1961-20141993-2014
7.04.1
2.01.7
Canada 1961-20141992-2014
6.03.4
1.81.5
Russia 1999-2013 9.4 4.9
India 1994-2013 7.4 4.8
China 1998-2013 3.4 8.3
Sources:ECB,Eurostat,FederalReserve,JapaneseCabinetoffice,Macrobond,oECD,ReserveBankofIndia,StatisticsCanada,StatisticsSweden,theWorldBank,UKofficefornationalStatisticsandUSBureauofEconomicAnalysis
– 20 –
SVERIGESRIKSBAnKEConoMICREVIEW 2015:2
THELonGRUnFISHERRELATIonDoESnoTFITTHEDATAInJAPAnInTHE1960sAnD
1970s–AnDFoRGooDREASonS
InJapan,theFisherinflationwasalmost1.5percentagepointslowerthantheobserved
inflationfortheperiod1961-2014,seeFigure7(a).ontheotherhand,fortheshorter
period,1981-2014,theFisherinflationwasentirelyinlinewiththeobservedinflation,
measuredbyboththeCPIandcoreinflation,seeFigure7(b).
InorderforthelongrunFisherrelationtofitthedatatheeconomyneedstobeina
steadystatewithnocleartrendsinthedata.TheJapaneseeconomygrewrapidlyfollowing
theendoftheSecondWorldWar,withgrowthratesofabout10percentperyear.These
highgrowthrateslasteduntilthebeginningofthe1970s,whentheeconomysuffered
fromtheeffectsofrisingoilprices.Hence,theJapaneseeconomyinthe1960sand1970s
wasinatransitionalphase,asitmovedfromrelativelylowGDPlevelstolevelsinparity
withthoseofthedevelopedindustrialcountries.Consequently,weshouldnotexpectthe
longrunFisherrelationtofitthedataduringthisperiod.
Japanesegrowthratesinthe1960sand1970saveraged5.5percentandthenominal
interestratewasjustover8.5percent,seeTable1.ThismeansthattheFisherinflationwas
around3percent.TheCPIinflationwasaround7.5percentduringthisperiod,seeFigure
8(a).Hence,theFisherinflationunderestimatedtheCPIinflationbymorethanfourper
cent.
0
2
4
6
8
Sweden1961-2014
the US1961-2014
Japan1961-2014
the UK1961-2014
Canada1961-2014
(a)
0
1
2
3
4 (b)
Sweden1994-2014
the US1987-2014
Euro area1999-2014
Japan1981-2014
the UK1993-2014
Canada1992-2014
Figure 7. Inflation according to the Fisher relation and observed inflation in Sweden, the US, the euro area, Japan, the UK and CanadaPer cent
Note. Core inflation is measured by CPI excluding food and energy, except in Sweden, where it is measured by CPIF excluding food and energy, the United States where it is measured by PCE inflation excluding food and energy, and the euro area where it is measured by HICP excluding food, energy, alcohol and tobacco.
Sources: ECB, Eurostat, Federal Reserve, Japanese Cabinet Office, Macrobond, OECD, Statistics Sweden, Statistics Canada, the World Bank, UK Office for National Statistics, US Bureau of Economic Analysis and the Riksbank
CPI-inflation Inflation according to the Fisher relationCore inflation
– 21 –
sveriges riksbank economic review 2015:2
Wecanobserveasimilarpattern,asinJapan’stransitionphase,incountriessuchasRussia,
IndiaandChina,wherethegrowthrateshavebeenhighinthelast15-20years.InRussia
andIndia,thegrowthrateshavebeenjustbelowfivepercentand,inChina,justover
eightpercent,seeTable1.Inotherwords,growthratesinparitywiththoseduringJapan’s
transitionphase.However,itisnotlikelythatthesegrowthrateswillbesustainableover
thelongrun–thoseofJapanwerenot.
Figure8(a)showsthattheFisherinflationinRussia,IndiaandChinawasclearlybelow
theobservedinflation,justastheFisherinflationwasinJapan’stransitionphase.Inline
withthetheory,thisconfirmsthatlongrunrelationsdonotapplytotransitionalphases.In
addition,andalsoinlinewiththetheory,itislikelythatthegrowthrateisapoormeasure
oftherealinterestrateintransitionalphaseswithhighgrowthrates.
ASTATISTICALTESTConFIRMSTHATTHELonGRUnFISHERRELATIonFITSTHEDATA
Byusingsimple”eyeballeconometrics”,wehaveshownthattheFisherinflationisinline
withtheobservedinflationinSwedenandseveralothercountries.Thisconclusioncanbe
confirmedbyat-testwiththenullhypothesisthattheFisherinflationcorrespondstothe
observedinflationineachcountry.TheaveragedifferencebetweentheFisherinflationand
thecoreinflationis−0.06,withap-valueof0.73,whilethedifferencebetweentheFisher
inflationandtheCPIinflationis−0.20,withap-valueof0.28,seeTable2.Forthelonger
period,1961-2014,wehaveonly4observations.Inthiscase,thedifferencebetweenthe
FisherinflationandtheCPIinflationis−0.34,withap-valueof0.17,seeTable2.
-1
0
1
2
3
4
Japan1981-1990
Japan1991-2014
Japan1961-1980
Russia1999-2013
India1994-2013
China1998-2013
-6
0
6
12
18
24 (a) (b)
Figure 8. Inflation according to the Fisher relation and observed inflation in Japan, Russia, India and ChinaPer cent
Note. Core inflation is measured by CPI excluding food and energy.
Sources: Japanese Cabinet Office, Macrobond, OECD, Reserve Bank of India and the World Bank
CPI-inflation Inflation according to the Fisher relationCore inflation
– 22 –
sveriges riksbank economic review 2015:2
Table 2. Matched t-test with the null hypothesis that the Fisher inflation equals the observed inflation
MEAnVALUE T-VALUESTAnDARDDEVIATIon
DEGREESoFFREEDoM P-VALUE
ΔKPI −0.34 −1.81 0.37 3 0.17ΔKPI −0.20 −1.21 0.41 5 0.28ΔUnd −0.06 −0.36 0.37 5 0.73
note.Intheempiricaltest,ΔKPI= π‾ F – π‾ KPI andΔUnd= π‾ F – π‾ Undarecalculated.At-testismadeoftheΔ-series,whichteststhenullhypothesisthatthethreeΔ-seriescomesfromanormaldistributionwithmeanzero.Row1referstoFigure7(a),excludingJapan.Rows2and3refertoFigure7(b).
DoESTHELoWPoLICyRATEInJAPAnExPLAInTHELoWInFLATIon?
ThefinancialmarketsinJapanwitnessedwidespreadderegulationinthe1980s,which
amongotherthingsledtoasteepriseinstockandpropertyprices.Thiscametoanabrupt
endatthebeginningofthe1990swhenboththestockandpropertypricesfellinthewake
ofafinancialcrisis.Sincethen,Japan’scentralbankhasheldthepolicyrateatlowlevels,
andatthesametimetheinflationhasbeenclosetozeroornegative.Thisperiod,which
beganatthebeginningofthe1990sandwhich,inallessentials,iscontinuingtoday,is
knownasJapan’s“lostdecades”.
ItmaybeofparticularinteresttoapplytheFisherrelationtoJapan'slostdecades,since
theywerecharacterisedbybothlowpolicyratesandlowinflation.Infact,wecanshow
thattheFisherinflationisentirelyinlinewiththeobservedinflationnotonlyduringthelost
decades,butalsointheprecedingperiod.
Inthepre-crisisperiod,1981-1990,theaveragenominalinterestratewasaboutsix
percentandthegrowthratealmostfourpercent,seeTable1.ThisimpliesthattheFisher
inflationwasjustovertwopercent,whichwasinlinewiththeobservedinflation,see
Figure8(b).Duringthelostdecades,boththenominalinterestrateandthegrowthratefell
toaboutonepercent.ThisimpliesaFisherinflationaroundzeropercent,whichwasalso
theleveloftheobservedinflation,seeFigure8(b).
DoesthismeanthatthelowinflationinJapan’slostdecadeswasduetothelowpolicy
rate?ourresultsareconsistentwithsuchaconclusion,butwecannotruleoutother
explanations.19nevertheless,itisworthnotingthattheaveragegrowthrateinJapan’slost
decadeswasjustbelowonepercent,whichcanbecomparedto1.4percentintheUS
and1.3percentinCanada.Inotherwords,Japanesegrowthrateswereaboutthesame
asthoseoftheUSandCanada–unlikeitsmonetarypolicy.Thenominalinterestratein
Japanwasaboutonepercent,whileintheUSitwas2.9percentandinCanadaitwas3.5
percent.Hence,theFisherinflationwas1.5percentintheUSand2.2percentinCanada,
whichcorrespondsrelativelywellwiththeactualoutcomeofabouttwopercentinboth
countries.InJapan,aswehaveseen,theinflationwasaboutzeropercentonaveragein
thedataandaccordingtotheFisherrelation.
19 Shirai(2012)discusses,forexample,theroleofdemographyineconomicdevelopmentsandshows,amongotherthings,estimatesoftheoutputgapwhichindicatethatithasbeennegativeoveralmosttheentireperiodsincethemid-1990s.However,healsonotesthatthereis,asofyet,noconsensusoverthefactorsthatcouldexplainsuchadevelopment.
– 23 –
sveriges riksbank economic review 2015:2
π‾ F
JamesBullardontheeffectsoflownominalinterestratesoveralongerperiod
Inamuch-debatedarticlefrom2010,JamesBullard,presidentoftheFederalReserveBank
ofSt.Louis,discussedtheeffectsoflownominalinterestratesoveralongerperiodoftime,
seeBullard(2010).ThepurposewastoshedlightontheriskthattheUScouldbemoving
toanewsteadystatewithdeflationarytendencieslikethoseinJapan.Hisanalysiswas
basedoninsightsfromearlierworkbyBenhabibetal.(2001),inwhichtheFisherrelation
andthezerolowerboundwerekeyingredients.20
MonetarypolicyisoftendescribedbyalinearTaylorruleinstandardmacroeconomic
models.21Asaconsequence,thereexistsnozerolowerboundforthepolicyrate.
Furthermore,thereisonlyonesteadystateinflationrate,whichcoincideswiththecentral
bank’sinflationtarget.However,ifthereexistsazerolowerbound,theremaybeasecond
steadystate,inwhichthepolicyrateisclosetozeropercentandinflationislowor
negative.
InFigure9weillustratethetwosteadystates.ThesolidlineshowstheFisherrelation,
givenalongrunrealinterestrateof0.5percent.Thedashedlineshowsanon-linear
Taylorrule,wherethenon-linearityisduetothezerolowerbound.Hence,therearetwo
pointswheretheFisherrelationandthenon-linearTaylorruleintersect,markingthetwo
steadystates.Theblueareamarksthesteadystatewherethenominalinterestrateis
wellabovethezerolowerboundandinflationisontarget.Thissteadystateisknownas
the”targeted”steadystate.Theothersteadystate,markedred,mayappearwhenthere
isalowerzerobound.Asmentioned,inthissteadystate,inflationmaybeloworeven
negative,whichmeansthatitwillbebelowthecentralbanksinflationtarget.Thissteady
stateisthereforeknownasthe“unintended”steadystate.
Thearrowsinthefigureillustrateapossibletransitionalpathfromthetargetedsteady
statetotheunintendedsteadystate.notethatmonetarypolicybecomespassiveoncethe
economygetsstuckintheunintendedsteadystate,sincethepolicyratewillnotreactto
changesininflation.Ifinflationweretofall,thecentralbankwouldbeunabletocutthe
policyrateduetothezerolowerbound.If,ontheotherhand,inflationweretorise,the
policyratecannotberaisedeither,sinceinflationisfarbelowtheinflationtarget.Inflation
expectationsthereforeremainatalevelconsistentwithanominalinterestrateofzeroper
centandarealinterestrateof0.5percent,i.e.anegativeinflationrateof−0.5percent.
Whentheeconomygetsstuckintheunintendedsteadystate,itcanbedifficultto
escapefromit,sincestandardmonetarypolicyactionsareineffective.Thecentralbank
mayinsuchasituationresorttounconventionalmeasures.Bullardarguesthatquantitative
easing(purchasinggovernmentbonds)isthemeasurethathasthebestchanceoftaking
theeconomybacktothetargetedsteadystate.
20 Thelevelofthelowerboundisnotimportanttotheanalysisinthissection.Theimportantpointisthattherede factoexistsalowerbound.Forthesakeofsimplicity,thislowerlimitisassumedtobezeropercent.
21 Themonetarypolicyruleinequation(6)isanexampleofthis.
– 24 –
SVERIGESRIKSBAnKEConoMICREVIEW 2015:2
-1
1
2
3
4
5
-1 1 2 3
Nom
inal
inte
rest
rat
e
Inflation rate
Unintendedsteady state
Targetedsteady state
Figure 9. The two steady states for the nominal interest rate and inflation when monetary policy is constrained by the zero lower boundPer cent
Fisher relation Non-linear Taylor rule
Source: Own illustration
JAPAnAPPEARSToHAVEGoTSTUCKInTHEUnInTEnDEDSTEADySTATE
Japanhasbeencharacterisedbylownominalinterestratesandinflationsincetheoutbreak
ofthefinancialcrisisinthebeginningofthe1990s.DoesthissuggestthatJapanhasgot
stuckintheunintendedsteadystate?Toinvestigatethis,welookatthemovementsofthe
nominalinterestrateandinflationbeforeandafterthecrisis.
Thenominalinterestrateinthepre-crisisperiodappearstohavevariedaroundalong
runlevelof6-8percentandinflationaroundalevelof2-3percent,seetheblueoutcomes
inFigures10(a)and10(b).outcomesinthepost-crisisperiodaremarkedred.Thenominal
interestrateappearstofluctuatearoundzeropercent.Theinflationratehasvaried,but
mostoftheobservationsarestillaroundzeropercent.ThisindicatesthatJapanmayhave
movedfromthetargetedsteadystatetotheunintendedsteadystate.22
SEVEnyEARSoFDATAISnoTEnoUGHToDETERMInEWHETHERSWEDEnAnDoTHER
CoUnTRIESHAVEMoVEDToTHEUnInTEnDEDSTEADySTATE
Swedenandseveralothercountrieshavesincetheoutbreakofthefinancialcrisisin2008
hadnominalinterestratesclosetozeroandlowinflation.Doesthissuggestthatthese
countriesareabouttomovetotheunintendedsteadystate?Sevenyearsofdataisnot
enoughtodeterminethis,butitmaybeenoughtodistinguishcertaintendencies.
22 ThisisalsoconfirmedbyAruobaetal.(2014).
– 25 –
sveriges riksbank economic review 2015:2
Figures10(c)and10(d)showthenominalinterestrateandinflationinSwedenbetween
theyears1994and2014.Inthepre-crisisperiodthenominalinterestratevariedaround
3-4percentandinflationaround1-2percent.Aftertheoutbreakofthecrisis,the
nominalinterestratehasbeenfallingtoaroundzeropercent,althoughitisnotpossible
todistinguishanewsteadystate.ThisisthecaseregardlessofwhetherwelookatCPI
inflationorcoreinflation.
IntheUS,thesteadystatepriortothefinancialcrisisappearstohavebeenabout4per
centforthenominalinterestrateandabout2percentforinflation,seeFigure11(a).Inthe
post-crisisperiod,thenominalinterestratehasbeenaboutzeropercent,butinflationhas
variedagooddealandmostoftheobservationsfluctuatearound1-2percent.
-2
0
2
4
6
8
10
-3 -2 -1 0 1 2 3 4 5 6 7
Nom
inal
inte
rest
rat
e
(a) Japan, CPI (b) Japan, core
-2
0
2
4
6
8
10
-3 -2 -1 0 1 2 3 4 5 6 7
-2
0
2
4
6
8
10
Nom
inal
inte
rest
rat
e
Inflation rate Inflation rate
(c) Sweden, CPI
-3 -2 -1 0 1 2 3 4 5 6 7-2
0
2
4
6
8
10
-3 -2 -1 0 1 2 3 4 5 6 7
(d) Sweden, core
Figure 10. Nominal interest rates and inflation in Japan and SwedenPer cent
Note. The nominal interest rates are measured by three-month treasury bills. Core inflation is measured by CPI excluding food and energy in Japan and by CPIF excluding food and energy in Sweden.
Sources: Macrobond, OECD, Statistics Sweden and the Riksbank
1981-1990 1991-2014
1994-2007 2008-2014 1994-2007 2008-2014
1981-1990 1991-2014
– 26 –
sveriges riksbank economic review 2015:2
Intheeuroareainflationhasbeenfluctuatingextensivelyinthepost-crisisperiod.At
lowest,itwas–0.3percentand,athighest,3percent,seeFigure11(b).Wenotethatthe
mostrecentobservationsindicatethatinflationisapproachingdeflationterritory.Butitis
hardtodistinguishanewsteadystateonthebasisofthis,eveniftheeuroareaappearsto
haveleftthesteadystateitwasinpriortothecrisis.neitherdothedevelopmentsinthe
UKandCanadainthepost-crisisperiodindicatethatthesecountrieshavebeenfallinginto
theunintendedsteadystate,seeFigures11(c)and11(d).However,liketheeuroarea,the
UKappearstohaveleftthesteadystateitwasinbeforethefinancialcrisis.
Figure 11. Nominal interest rates and inflation in the US, the euro area, the UK and CanadaPer cent
Note. The nominal interest rates are measured by three-month treasury bills, except in the euro area where the nominal interest rate is measured by EONIA.
Sources: ECB, Eurostat, Macrobond, OECD and the US Bureau of Economic Analysis
-2
0
2
4
6
8
10
-1 0 1 2 3 4 5
(a) United States, PCE
Nom
inal
inte
rest
rat
e
(b) Euro area, HICP
-2
0
2
4
6
8
10
-2
0
2
4
6
8
10
-2
0
2
4
6
8
10
-1 0 1 2 3 4 5
-1 0 1 2 3 4 5 -1 0 1 2 3 4 5
(c) United Kingdom, CPI (d) Canada, CPI
1987-2007 2008-2014 1999-2007 2008-2014
Nom
inal
inte
rest
rat
e
Inflation Inflation
1993-2007 2008-2014 1993-2007 2008-2014
– 27 –
sveriges riksbank economic review 2015:2
Concludingremarks
oneoftheobjectivesofthisarticlehasbeentoexplorenarayanaKocherlakota’sclaimthat
alowpolicyrateoverthelongruncanleadtolowinflation.Theideareliesonlongrun
monetarypolicyneutralityandthelongrunFisherrelation.Undertheseassumptions,it
followsthatalong-livedcutinthepolicyrateleadstoaproportionalfallininflation.
However,evenifeconomictheorypredictslowinflationifthepolicyrateisheldatalow
leveloveralongerperiod,thisdoesnotnecessarilyimplythatthisalsowillbethecasein
reality.Economictheoriesarebasedonvarioussimplifiedassumptionsandaretherefore,by
definition,inaccurateinonewayoranother.Itisthereforeimportanttoempiricallytestto
whichextenttheeconomictheoryisconsistentwithdata.Wehaveshownthat,ifthereal
interestrateisequatedtotheGDPgrowthratepercapita,theaverageinflationinSweden
andothercountriescanbeexplainedbythedifferencebetweentheaveragenominal
interestrateandthegrowthrate.Wefindthisinteresting,sinceitindicatesthat,inseveral
countries,thelongrunFisherrelationisconsistentwiththedata.Inaddition,itprovides
empiricalsupporttoKocherlakota’sclaim.
SeveralscientificarticleshaverecentlybeenpublishedwheretheFisherrelationisoneof
thekeyfactorsbehindtheresults.Schmitt-GrohéandUribe(2013)isoneexample.They
presentascenariowhichisintendedtoresembletheexperiencesoftheUSfollowingthe
outbreakofthefinancialcrisisin2008,i.e.alongperiodofnominalinterestratesclose
tozero,inflationexpectationsbelowtheinflationtargetandslowemploymentgrowth.
TheFisherrelationplaysakeyroleinhowhouseholdsandfirmsinterpretchangesinthe
policyrateintheirmodel.Apolicyrateincreaseisasignalofhigherfutureinflation.oneof
theirconclusionsisthatanincreaseinthepolicyratepushesupinflationexpectationsand
promotesemployment.
AnotherexampleisLeeperandLeith(2015).Theypresentamodelinwhichthefiscal
theoryofthepricelevelandtheFisherrelationaretwokeyfeaturesindetermining
inflation.23Thefiscaltheoryofthepricelevelisrelevantwhenexpansionsinnominaldebt
arenotexpectedtobefundedbyhighertaxesorlowerexpenditure,i.e.whenhouseholds
andfirmsdonotexpectfiscalpolicyexpansionstobefundedbyafuturesurplus.When
thisisthecase,raisingthepolicyrateraisesthenominalinterestratereceiptsofthebond
holders(i.e.thehouseholds).Sincetheydonotexpecthigherfuturetaxestofinancethe
increasedexpenditurestheirnominalwealthalsoincrease.Theincreaseofwealthpushes
upconsumptionanddemand,whicheventuallyalsopushesupinflation.
23 SeealsoLeeperandyun(2006)foradescriptionofthefiscaltheoryofthepricelevel.
– 28 –
sveriges riksbank economic review 2015:2
References
Apel,Mikael,EricFrohm,JyryHokkanen,ChristinanymanandStefanPalmqvist(2014),”Resultsfromasurveyoncompanypricing,”Economic Commentaryno.4,2014.
Aruoba,S.Boragan,PabloCuba-BordaandFrankSchorfheide(2014),”MacroeconomicDynamicsneartheZLB:ATaleofTwoCountries,”unpublishedpaper.
Benhabib,Jess,StephanieSchmitt-Grohé,andMartínUribe(2001),”ThePerilsofTaylorRules,”Journal of Economic Theory,96(1-2),pp.40-69.
Bullard,James(2010),”SevenFacesof’ThePeril’,”Federal Reserve Bank of St. Louis Review,92(5),pp.339-352.
Christiano,Lawrence,MathiasTrabandt,andKarlWalentin(2011),”IntroducingFinancialFrictionsandUnemploymentintoaSmallopenEconomyModel,”Journal of Economic Dynamics and Control35(12),pp.1999-2041.
Fisher,Irving(1977)[1930],The Theory of Interest,Philadelphia:PorcupinePress.
Friedman,Milton,andAnnaJ.Schwartz(1963),A Monetary History of the United States, 1867-1960.PrincetonUniversityPress.
Hall,RobertE.(2014),”TheRoutesintoandoutoftheZeroLowerBound,”FederalReserveBankofKansasCity,JacksonHoleSymposium2014.
Jonsson,Magnus(2002),”Therealinterestrateandmonetarypolicy,”Sveriges Riksbank Economic Review, 3,pp.45-64.
Jonsson,Magnus(2007),”Increasedcompetitionandinflation,”Sveriges Riksbank Economic Review, 2,pp.69-89.
Jonsson,MagnusandKevinMoran(2014),”Thelinkagesbetweenmonetaryandmacroprudentialpolicies,”Sveriges Riksbank Economic Review, 1,pp.6-26.
Kocherlakota,narayana(2010),”InsidetheFoMC,”speech,Marquette,Michigan.
Leeper,EricM.andTackyun(2006),”TheFiscalTheoryofthePriceLevel:BackgroundandBeyond,”International Tax and Public Finance13(4),pp.373-409.
Leeper,EricM.andCampbellLeith(2015),”InflationThroughtheLensoftheFiscalTheory,”HandbookofMacroeconomics,volume2,underpublication.
Meh,CésaireandKevinMoran(2010),”TheRoleofBankCapitalinthePropagationofShocks,”Journal of Economic Dynamics and Control34(3),pp.555-576.
Plosser,CharlesI.(2008),”TheBenefitsofSystematicMonetaryPolicy,”speech,FederalReserveBankofPhiladelphia.
Schmitt-Grohé,StephanieandMartínUribe(2013),”TheMakingofaGreatContractionwithaLiquidityTrapandaJoblessRecovery,”unpublishedpaper.
Shirai,Sayuri(2012),”HaveDemographicChangesAffectedJapan’sMacroeconomicPerformance?SomeImplicationsforMonetaryPolicy,”unpublishedpaper.
Summers,Larry(2014),”U.S.EconomicProspects:SecularStagnation,Hysteresis,andtheZeroLowerBound,”Business Economics49(2),pp.65-73.
Taylor,JohnB.(1993),”DiscretionversusPolicyRulesinPractice,”Carnegie-Rochester Conference Series on Public Policy39(1),pp.195-214.
Walsh,Carl(2003),Monetary Theory and Policy.2ndEdition,MITPress.