COMPARING PAST AND PRESENT INFLATION 現(xiàn)在與過(guò)去的通脹比較

NBERWORKINGPAPERSERIES

COMPARINGPASTANDPRESENTINFLATION

MarijnA.Bolhuis

JuddN.L.Cramer

LawrenceH.Summers

WorkingPaper30116

/papers/w30116

NATIONALBUREAUOFECONOMICRESEARCH

1050MassachusettsAvenue

Cambridge,MA02138

June2022

TheviewsexpressedhereinarethoseoftheauthorsanddonotnecessarilyreflecttheviewsoftheNationalBureauofEconomicResearch,IMF,itsExecutiveBoard,orIMFmanagement.

NBERworkingpapersarecirculatedfordiscussionandcommentpurposes.Theyhavenotbeenpeer-reviewedorbeensubjecttothereviewbytheNBERBoardofDirectorsthataccompaniesofficialNBERpublications.

?2022byMarijnA.Bolhuis,JuddN.L.Cramer,andLawrenceH.Summers.Allrightsreserved.Shortsectionsoftext,nottoexceedtwoparagraphs,maybequotedwithoutexplicitpermissionprovidedthatfullcredit,including?notice,isgiventothesource.

ComparingPastandPresentInflation

MarijnA.Bolhuis,JuddN.L.Cramer,andLawrenceH.Summers

NBERWorkingPaperNo.30116

June2022

JELNo.C43,E21,E31,E37

ABSTRACT

TherehavebeenimportantmethodologicalchangesintheConsumerPriceIndex(CPI)overtime.Thesedistortcomparisonsofinflationfromdifferentperiods,whichhavebecomemoreprevalentasinflationhasrisento40-yearhighs.Tobettercontextualizethecurrentrun-upininflation,thispaperconstructsnewhistoricalseriesforCPIheadlineandcoreinflationthataremoreconsistentwithcurrentpracticesandexpendituresharesforthepost-warperiod.Usingtheseseries,wefindthatcurrentinflationlevelsaremuchclosertopastinflationpeaksthantheofficialserieswouldsuggest.Inparticular,therateofcoreCPIdisinflationcausedbyVolcker-erapoliciesissignificantlylowerwhenmeasuredusingtoday’streatmentofhousing:only5percentagepointsofdeclineinsteadof11percentagepointsintheofficialCPIstatistics.Toreturnto2percentcoreCPIinflationtodaywillthusrequirenearlythesameamountofdisinflationasachievedunder

ChairmanVolcker.

MarijnA.Bolhuis

InternationalMonetaryFund

mbolhuis@

JuddN.L.Cramer

judd.cramer@

LawrenceH.Summers

HarvardKennedySchoolofGovernment

79JFKStreet

Cambridge,MA02138

andNBER

lhs@

FullseriesofheadlineandcoreinflationwithOERcorrectionandconstantweightsfordifferentyearsisavailableat:

/category/inflation/

2

1.Introduction

AsconcernsaboutUSinflationhavegrown,theConsumerPriceIndex(CPI)hascomeundercloserscrutiny.2TheCPIgrew8.3percentinthetwelvemonthsendinginApril,downslightlyfromthepreviousmonthbutstillwellaboveanyotherperiodsince1981(Figure1).Whileaworryingfigure,thisremainsfarbelowtheofficialMarch1980peakof14.8percent.Thattheheadlinenumberhadalreadyfallento2.5percentbyJuly1983,followingthepolicydecisionsofFederalReserveBoardChairmanPaulVolcker,hasservedastheexemplumofthepowerofhawkishmonetarypolicy(GoodfriendandKing2005).Sincemuchlessofadeclineisneededtoreturntotrendtoday,somecommentershavesuggestedthatpolicymakersmightbeabletodecreaseinflationtowardsdesiredlevelswithoutlargemacroeconomicconsequences(DeLong2022;Krugman2022).Yet,methodologicalchangesintheCPIovertimemakedrawingconclusionsfromthesetypesofintertemporalcomparisonsfraught.

ThispapershowsthatusingthemovementofthepublishedCPIduringpastdisinflationaryperiodstoexplorethecurrentsituationcanleadresearchersastray.Forexample,inarguingagainstpolicymakersfalling“behindthecurve”inthefaceofrisinginflation,Blanchard(2022)showedthattoday’sgapbetweencoreinflation—whichremovesvolatilefoodandenergyprices—andrealinterestratesisapproachingabout70percentofthe1975gap.WeargueagainstusingofficialCPIinflationtoassessthisgap.Mostimportantly,priorto1983measurementofshelterinflationwasmechanicallyresponsivetoFederalReserveinterestratepolicyviamortgagerates.Thismethodmadepre-1983peakCPIinflationmeasures,especially

2ThroughoutthispaperweuseCPIasashorthandforCPI-U,theConsumerPriceIndexforAllUrbanConsumers.

3

duringtheVolcker-era,artificiallyhighatthebeginningofthetighteningcycle,anddeclineslookartificiallyfast.

Figure1:HeadlineCPIinflation,1946-present

Source:BureauofLaborStatistics.

Note:Percentchangefrom12monthsearlier.

Tobettercontextualizethecurrentrun-upininflation,thispaperconstructsnewhistoricalseriesforCPIheadlineandcoreinflationthataremoreconsistentwithcurrentpracticesandexpendituresharesfortheentiretyofthepost-warperiod.UsingpubliclyavailableBureauofLaborStatistics(BLS)dataforthepost-warperiod,wedevelopnewestimatesofCPIheadlineandcoreinflationthatcanbebettercomparedacrosstime.Thefullseriesisavailableonourwebsiteat/category/inflation/.Ouranalysisrevealsthatcurrent

inflation,especiallycoreinflation,isconsiderablyclosertopreviouspeaksthanintheofficialseries.OfficialcoreCPIinflationpeakedat13.6percentinJune1980,whereasweestimatethatcoreinflationwas9.1percentinthatsamemonthwhenadjustingforthetreatmentofshelter

4

inflation.OurestimatesalsosuggestthatthelocaltroughofcoreCPIinflationin1983wasconsiderablyhigherthanoriginallyreported.Overall,theseestimatesimplythattherateofcoreCPIdisinflationcausedbyVolcker-erapoliciesissignificantlylowerwhenmeasuredusingthecurrenttreatmentofhousing:only5percentagepointsofdeclineinsteadof11percentagepointsintheofficialCPIstatistics.Toreturnto2percentcoreCPItoday,wethusneeddisinflationofasimilarmagnitudeasChairmanVolckerachieved.

Similarissuesaffectconclusionsdrawnfromcomparisonsofcurrentinflationwithotherperiodsofelevatedinflation.RecentworksuggeststhattheyearsfollowingWorldWarIIhavestrongsimilaritiestothecurrentinflationenvironment(e.g.,Rouseetal.,2021;DeLong,2022).Weshowthatduetothegreaterweightoftransitorygoodscomponents—especiallyfoodandapparel—intheindexofthe1940sand1950s,pastinflationspikeswerehigherandmoreshort-livedthantoday’s.Whenusingcurrentweights,weestimatethatthepeakofcoreCPIinflationinJune1951fallsfrom7.2to5percent,andthepeakofheadlineCPIinflationfallsfrom9.4tojust3.3percent.Thesetwopointsserveasacautionagainstoverlyoptimisticforecastsofaninexpensivedisinflationinthecurrentcycle—thedisinflationthatneedstobeachievednowislargebyhistoricalstandards.

Therestofthispaperisstructuredasfollows.Section2reviewsthemethodologyusedbytheBureauofLaborStatistics(BLS)informinghomeownershipcostsbeforeandafterthechangesintroducedin1983andsuggeststhatitisnaturaltoexpectthatpastinflationwouldbelowerusingcurrentmethods.Section3describesthedataweemployanddifferentstatisticalmodelsweuseforcreatingmoreconsistentCPImeasuresbothtodealwiththemeasurementofhousingandtherelativelymore“sticky”natureoftheindextoday.Section4presentsourestimatesfor

5

thepost-warinflationsanddisinflationsusingourmeasuresandshowsthecurrentsituationisofasimilarmagnitudewithpastepisodes.Section5offerssomeconcludingobservations.

6

2.MeasuringHousingInflation

Housingisbothaninvestmentandaconsumptiongood.Ontheonehand,itisthelargestfixedinvestmentthatmostAmericansmakeintheirlives.3Ontheotherhand,itprovidesaservice,shelter,thatisconsumeddaily.Between1953and1983,theBureauofLaborStatistics(BLS)valuedhomeownershipcostsfortheCPIwithoutdisentanglingthesetwoqualities.Itproducedameasurethatbroadlycapturedchangesintheexpensesofhomeowners,takinghouseprices,mortgageinterestrates,propertytaxesandinsurance,andmaintenancecostsasinputs.Morespecifically,thehome-purchaseexpenditureweightwasthenetpurchaseofowner-occupiedhousesinthesurveyperiod,andthemortgage-interestexpenditureweightwasthetotalinterest(undiscounted)thatwouldbepaidoverhalfthetermonallmortgagesincurredduringthesurveyperiod(Dugganetal.1997).Sheltercostswerethusdirectlyaffectedbymonetarypolicy,duetotheeffectofthefederalfundsrateonmortgagerates.ThismethodologywaswithoutconceptualfoundationanditsuseresultedinasubstantialupwardbiasintheCPI(Gillingham1980;Gillingham1983).Theapproachproducedavolatileshelterseriespre-1983,asshowninFigure2(a),thatmovedalmostinstepwiththefederalfundsrateuntil1983.Muchmoresothanrents,theseestimateswereresponsivetochangesininterestratesasseeninthepre-1983periodofFigure2(c).Duringthetighteningcyclesof1967-1969,1972-1974,and1977-1981,shelterinflationincreasedsharply,onlytofallprecipitouslywhenthepaceoftighteningsloweddown.4

3ThishasbeenthecasethroughoutAmericanhistory.See,forexample,Shelton(1966).

4Besidesthemechanicalrelationshipbetweenmonetarypolicyandhomeownershipcosts,theoldapproachfacedseveralothercriticisms.Otherformsofinterestpayments,suchascreditcardspending,werenotpartoftheconsumerbasket.Moreover,consumerseitherpaidforthehomepurchasepriceorthemortgagepayments,butnotboth.Includingbothledtoalargershareofhousingintheconsumptionbasket.Furthermore,theoldmethodologyassumedallhomeownershadthesame30-yearfixed-ratemortgage.

7

Figure2:ShelterCPIinflationandtheFederalFundsRate,1954-present

(a)

(b)(c)

Source:BureauofLaborStatistics.

Notes:Percentchangefrom12monthsearlier.ChangeinOERinitiatedJanuary1983.

In1983,aftertenyearsofstudy,theBLSexchangedhomeownershipcostsforowners’equivalentrent(OER)(GillinghamandLane1982).Byestimatingwhatahomeownerwouldreceivefortheirhomeontherentalmarket,theBLSstrippedawaytheinvestmentaspectofhousingtoisolateowner-occupiers’consumptionofresidentialservices.Sincethis1983shift,

8

shelterCPIhasbeenmuchlessvolatileandmuchmorecorrelatedwithrentCPI(Figure2(b)).Ofcourse,thisshadowpriceofowner-occupiedshelterisnotobserved.TheBLSusesstatisticaltechniquestoinferOERusingrentalpricesforsimilarunitsinthearea.5Inparticular,theestimatedaverageOERvalueisdeterminedbyalinearregressionofimputedrentsonpropertyvalue,income,andnumberofroomsfromtheConsumerExpenditureSurvey.6Thisprocedureisdoneforallunits,notjustnewlysignedleasesornewlypurchasedhouses,withhousingunitsinthesamplequeriedeverysixmonths.EstimatedOERwillthereforelagspotpricesandismechanicallycorrelatedwithrents.7

The1983housingchangeshadlargeeffectsonbothheadlineandcoremeasuresofinflation.InDecember1982,“homeownership”received26.1percentoftheweightoftheoverallCPIindex.ForthecoreCPI,itwasafull36.1percent.Initslastmonth,homeownershipcostswerereportedtohavedeclinedby1.7percentfromNovembertoDecemberasthefinancing,taxes,andinsurancecomponent,whichmostlyvariedwithmortgagerates,declined3.7percent.InJanuary1983,owners’equivalentrentofresidence(OER)accountedforonly13.5percentoftheweightfortheoverallCPIinitsfirstmonthofexistence.Itwasreportedtohavegrown0.7percentfromDecember1982.TheBLSwasawarethattherewouldbealargediscontinuityinthepublished

5ForarecentdiscussionofmethodsseeGindelskyetal.(2019).

6Forcompletedetails,seeChapter17oftheBLShandbookofmethods.Thelinearregressioncoefficientsarethenappliedtodecennialcensusvaluesforthesameindependentvariablestoestimatetheaverageowners’equivalentrentforeachsegmentviathenonlinearregression

0ER=Fo+(F1×pTopval)+(F2×pTopval2)+(F3×income)+(F4×Tooms),

whereOERisthepredictedvaluethatthehomewouldrentfor,propvalisthemarketvalueofthehome,incomeistheincomeoftheconsumerunit,androomsisthenumberofroomsinthehouse.TheBLSrepeatsthisprocedureacrossdifferentgeographicareas.Afterthemodellingprocess,CPIweightsarethendeterminedbysurveysofhomeowners.

7Forfurtherdiscussionofthislagstructure,seeBolhuisetal.(2022)

9

weightsandmeasureswiththechangeinmethodology.Toallowthepublicandresearcherstobetterprepareforthetransition,beginningin1978,theBLSstartedpublishingtheCPI-U-X1serieswhichusedtheOERconceptstocreateanalternative,overlappingseries.ThisallowedforcomparisonoftheOEReffectpriortotheredesign,withinvestigationsshowingalowerpeakforthealternateseries(DeLong1997).FollowingthechangeoveroftheofficialCPI-UtoitscurrentOERapproachin1983,theCPIhasundergonenumerousotherchanges,fromqualityadjustments

forusedcars,tousinggeometricmeanstocalculatedpricechangesforsubcomponents,andmore.8

Tokeeptrackofallthesechanges,andtheirpossibleeffects,theCPIpublishesaCPIforallUrbanConsumersResearchSeries(CPI-U-RS)whichseekstoanswerthequestion:“whatwouldhavebeenthemeasuredrateofinflationfrom1978forwardhadthemethodscurrentlyusedincalculatingtheCPI-Ubeeninusesince1978.”9Asweexploreourownmeasures,whichseektoanswerslightlydifferentquestions,wealsoinvestigatetheCPI-U-RSandshowthatitleadsustosimilarthoughslightlymoreconservativeconclusions.BecausetheCPI-U-RSisonlypublishedfrom1978forward,wemust“backcast”ourownmeasurestostudythepre-1978inflationcycles.

8Forfulllistsee:/cpi/research-series/r-cpi-u-rs-changes.htm.

9Theseseriesareavailableat/cpi/research-series/r-cpi-u-rs-home.htm.

10

3.NewEstimatesofHistoricalCPIInflation

BasicCPIMethodology

IntheCPI,theurbanareasoftheUnitedStatesaredividedinto32geographicareas,calledindexareas.10Thesetofallgoodsandservicespurchasedbyconsumersisdividedinto211categoriescalleditemstrata:209CommoditiesandServicesitemstrata,plusthe2housingitemstratathatarethemainfocusofthispaper.Thisresultsinover7,000item-areacombinations.

Weabstractawayfromthislevelofdetailandfocusonthenationwidegroupsthataremostconsistentfrom1946topresent.ItistheevolutionoftheweightsassignedtodifferentcomponentsovertimeandtheirvariedperformancesonwhichwefocusforoursecondmajoradjustmenttopastCPImeasures.

AggregationandReweighting

Oncesamplingandanalysisareusedtoproduceameasuredpriceincreaseforeachofthe7,000plusindexes,amodifiedLaspeyrespriceindexisusedtoaggregatebasicindexesintothepublishedCPI-U.TheLaspeyresindexusesestimatedquantitiesfromthepredeterminedexpenditurereferenceperiodtoweighteachbasicitem-areaindex.Thesequantityweightscurrentlyremainfixedforatwo-yearperiod,andarereplacedinJanuaryofeacheven-numberedyearwhentheaggregationweightsareupdated.InaLaspeyresaggregation,consumersubstitutionbetweenitemsisassumedtobezero.TheCPIisnotapureLaspeyresindexasthegoodsbasket

10ThisdiscussionborrowslargelyfromtheBLS’sHandbookofMethods,lastupdatedinNovemberof2020.

11

hasevolvedsignificantlyovertimewithnewinventionsandhigherincomelevels.WedescribethefullreweightingscheduleintheAppendix.

Data

WeusepublicdatafromtheBLSovertimetoexplorethechangeinthenatureofinflationduringthepost-warperiod.Ourdatasetcontains32componentsthatcoveraround90percentoftheoverallCPIsince1946,aslistedinAppendixTable1.FigureA.1intheAppendixplottheinflationrateofcomponentsovertime.

WealsocollectdataonquantityweightsandrelativeimportanceofCPIcomponentsovertime.11InAppendixFigureA.2,weplottheevolutionofrelativeimportanceratiosofthecomponentsovertime.Sincethe1940s,consumerspendinghasshiftedfromgoodstoservices.Asaresult,theweightsandrelativeimportanceofgoodscomponentshasfallenovertime,mainlydrivenbyadeclineintheimportanceoffoodandapparel.Themirrorimageofthistrendisthattheweightsandrelativeimportanceofhousing,medicalcare,education,andpersonalcarehaveincreasedsincethestartofthepost-warperiod.

Usingourdataoninflationratesandweightsofthe32components,weconstructtwonewmeasuresofCPIinflation.First,wereplicatetheofficialheadlineandcoreCPIinflationrate.Toensureourbottom-upestimateofofficialheadlineCPIequalsthepublishedseries,weadda

11Therelativeimportanceofacomponentisitsexpenditureorvalueweight.Whenthequantityweightsarecollected,theyrepresentaverageannualexpenditures.Foryearsotherthanthebaseyear,relativeimportanceratiosrepresentanestimateofhowconsumerswoulddistributetheirexpendituresaspriceschangeovertime,butconsumersdonotchangetheirrealconsumptionpatterns.Notethatweightsareestimatedusingmicrodataonexpenditures,whereasrelativeimportanceratiosarenotdirectlyobserved.

12

residualCPIcomponent.Thisresidualcomponentmainlycoversrecreationandinformation.WethenadjustCPIinflationpre-1983byestimatingOERusingtheCPIrentseries.WebackcastwhatwethinkthatOERinflationwouldhavebeenpre-1983hadthepost-1983methodbeenused.WedothisbyregressingOERonrentalinflationpost-1983.For1979to1983,wecheckourestimatesagainsttheretroactiveCPI-U-RSseriesoftheBLSthatmeasuresCPIinflationconsistentlyusingcurrentmethods.Finally,toassesstheimportanceofdifferencesinthevolatilityofCPIcomponentsforoverallinflation,wecreateasecondversionoftheestimatedCPIseriesthatuses2022quantityweightsovertheentireperiodwhilealsoadjustingthepre-1983datawithestimatedOER.Ouronlinedatasetalsoincludesseriesthatuseconstantweightsfromothertimeperiods.

13

4.Findings

OurestimatessuggestthatthecurrentinflationrateisclosertothepeakofothercyclesthantheofficialCPIdatasuggest.Figure3showsthatthepeakoftheVolcker-erainflation(March1980),currentlyunderstoodtohavebeenat14.8percent,isonly11.4percentwhenadjustedfortheswitchfromhomeownershipcoststoOER.ThegrowthincoreCPIatitspeakinJune1980fallsfrom13.6percentto9.1percentwhenmeasuredusingtheOERmethodasseeninFigure4.ThelargedifferencesbetweentheofficialandadjustedseriesreflectboththesubstantialweightofOERintheindex,especiallyincoreCPI,andthelowerpeaksofestimatedOERrelativetohomeownershipcosts.Fromalowof14.5percentweightin1983,asAmericanshaveshiftedmoreoftheirconsumptiontowardshousing,OERhasrisentorepresent24.3percentofoverallCPIand30.6percentofcoreCPIin2022.Whilepastinflationpeaksarelowerusingtheconsistentmethodology,theaverageinflationratepre-1983wouldalsohavebeenlower.Ourestimatesshowthatthemeanheadlineinflationratebetween1949and1983is0.4percentagepointslowerwhenaccountingfortheshifttoOER.

14

Figure3:OfficialandEstimatedHeadlineCPI

Source:BureauofLaborStatistics,authors’calculations

Notes:Percentchangefrom12monthsearlier.Left-handside:HomeownershipcostsarereplacedwithestimatedOERpre-1983.Right-

handside:HomeownershipcostsarereplacedwithestimatedOERpre-1983andquantityweightsarefixedat2022levels.

Figure4:OfficialandEstimatedCoreCPI

Source:BureauofLaborStatistics,authors’calculations

Notes:Percentchangefrom12monthsearlier.Left-handside:HomeownershipcostsarereplacedwithestimatedOERpre-1983.Right-

handside:HomeownershipcostsarereplacedwithestimatedOERpre-1983andquantityweightsarefixedat2022levels.

15

Table1:PastInflationCyclesandToday

HeadlineCPI

CoreCPI

Official

Today’sBasis

Today’sOfficial

Basisand

weights

Today’sBasis

Today’sBasisandweights

1949-541

Start

Peak

EndReflation

Disinflation

-2.9

9.4

-0.7

12.3

10.1

-2.9

9.4

-0.7

12.3

10.1

0.1

3.3

-0.5

3.2

3.8

0.4

7.2

-0.4

6.8

7.6

-0.2

6.9

-0.4

7.1

7.3

0.3

5.0

-0.6

4.7

5.6

1972-762

Start

Peak

EndReflation

Disinflation

2.7

12.3

4.9

9.6

7.4

2.4

11.0

5.1

8.6

5.9

2.2

9.1

5.9

6.9

3.2

2.8

11.7

6.1

8.9

5.6

2.2

9.5

6.3

7.3

3.2

2.1

7.8

6.1

5.7

1.7

1978-833

Start

Peak

EndReflation

Disinflation

6.5

14.8

2.5

8.3

12.3

5.7

11.6

3.4

5.9

8.2

5.6

11.4

3.4

5.8

8.0

6.5

13.6

3.0

6.1

10.6

5.0

9.1

4.3

4.1

4.8

4.9

9.8

3.6

4.9

6.2

Today4

Start

Peak

Reflation

0.1

8.5

8.4

0.1

8.5

8.4

0.1

8.5

8.4

1.2

6.5

5.3

1.2

6.5

5.3

1.2

6.5

5.3

Disinflation

(to2%target)

Sources:BureauofLaborStatistics,Authors’calculations

Notes:Wedefine‘Start’and‘End’asthelocalminimaofofficialannualheadlineCPIgrowthatthestartandendofeachcycle.Wedefine‘Peak’asthelocalmaximumofofficialannualheadlineCPIandcoreCPIgrowthforeachcycle.ForofficialcoreCPI,weusetheCPIlessfoodseriesfortheperiodbefore1958.Today’sbasis:HomeownershipcostsarereplacedwithestimatedOERpre-1983.Today’sbasisand

weights:HomeownershipcostsarereplacedwithestimatedOERpre-1983andquantityweightsarefixedat2022levels.

1Start:July1949.Peak:February1951forheadlineandJune1951forcore.End:October1954.

2Start:June1972.Peak:December1974forheadlineandFebruary1975forcore.End:December1976.

3Start:April1978.Peak:March1980forheadlineandJune1980forcore.End:July1983.

4Start:May2020.Peak:March2022forbothheadlineandcore.

16

Morebroadlyspeaking,pastinflationarycycleswouldhavebeenlessvolatileusingtheconsistentmethodologythatusesOER.Thepaceofreflationduringthecycleupswings,andthepaceofdisinflationduringthecycledownswingsarelowerundertoday’smethodology,assummarizedinTable1.ThesedifferencesimplythattheresponsivenessoftheCPItomonetarypolicywasconsiderablylowerduringthe1960sand1970sthantheofficialCPIstatisticssuggest.WestressthatanyconsequencesofthedifferenceinmeasurementarelargerforcoreCPIthanforheadlineCPI,duetotheconsiderablylargerweightofOERincoreCPI.

OurestimatesalignwiththeretroactiveCPI-U-RSseriesoftheBLSthatmeasuresCPIinflationconsistentlyusingcurrentmethods.FigureA.3intheAppendixplotstheCPI-U-RSseriesforheadlineandcoreinflation,whichoverlapwithourestimatesfortheperiod1979-1983.WhereastheCPI-U-RSseriesestimatesthepeakofheadlineinflationduringtheVolcker-erainMarch1980at11.8percent,weestimateourpeakinthesamemonthat11.6percent.Forcoreinflation,bothouradjustedseriesandtheCPI-U-RSpeakat9.9percentinDecember1980,wellbelowtheofficiallyreportednumberof12.2percentforthatmonth.

AnalternativeapproachtomakingCPImorecomparableistoattempttoapplypre-1983techniquestothismoment.ThishasbeenpursuedLeeandBarton(2022),whichextendsworkofHazelletal.(2020).Afterperformingthecomplementaryprocess,theyalsofindthatpastandpresentinflationlookmoresimilarthanoriginallyreported,withcurrentlevelshigherthanofficialseriesduetotherecentincreaseinmortgagerates.

Ourestimatesalsoindicatethatpastinflationcycleslookmorevolatilethantoday’sduetothegreaterweightoftransitorygoodscomponentsinpastmeasurements.Intheearly1950s,forexample,foodandapparelaccountedforcloseto50percentoftheheadlineCPIindex(Figure5).Afterthegrowthrateofthesecomponentsshotabove10percentduetoKoreanwar-induced

17

shortages,headlineinflationfellfrom9to2percentwithinayear.Today,however,foodandapparelonlyreceive17percentoftheweightofheadlineCPI.WeestimatethatthepeakofheadlineCPIinflationin1951wouldhavebeen3.3percent,insteadof9.4percent,whenmeasuringCPIusingtoday’sweights.OuradjustedpeakofcoreCPIinflationin1951is5percent,comparedtoanofficialpeakof7.2percent.

Duringtheinflationcycleoftheearly1970s,boththetreatmentofshelterinflationandthegreaterweightofvolatilegoodscomponentspushedupofficialCPIinflationcomparedtocurrentmethods.WhenadjustingforthetreatmentofOER,ourestimateofthepeakofcoreCPIinflationinFebruary1975fallsfrom11.7to9.5percent.WeestimatethatthepeakofcoreCPIinflationin1974wouldhavebeenonly7.8percentwhenusingtoday’sweightscoupledwiththeOERcorrection.Thepre-Volckeradjustedtroughwasstill5.9percent,suggestingthelessonsfromtheearly1970sfordecreasingcurrentlyelevatedinflationtoaround2percentarelimited.

Theindexbearswitnesstothisshiftfromtransitorygoodscomponentstolessvolatileservices,withstickyindustriesgainingweightintheCPIacrosstheboard(BryanandMeyer,2010).Thecurrentapproachrendershousinginflationparticularlysticky(Bolhuisetal.,2022).SuchchangesinmeasurementmethodscomplicatecomparisonsbetweenCPIratesfromdifferentperiods.ByconstructingCPIseriesusingconstantweights,weaimtoimprovethecomparisonsofCPIinflationovertime.

18

Figure5:AverageFoodandApparelWeightinHeadlineCPIbyRevision

100%90%80%70%60%50%40%30%20%10%0%

1946-19531954-19631964-19771978-19871988-1997After1997

FoodApparel

Source:BureauofLaborStatistics,authors’calculations

Note:Arithmetic