http://www.nber.org/chapters/c10383.pdf

1 This PDF is a selection from an out-of-print volume from the National Bureauof Economic ResearchVolume Title: Business Cycles: Theory, History, Indicators, and ForecastingVolume Author/Editor: Victor ZarnowitzVolume Publisher: University of Chicago PressVolume ISBN: 0-226-97890-7 Volume URL: Date: n/aPublication Date: January 1992 Chapter Title: Major Macroeconomic Variables and Leading IndexesChapter Author: Victor ZarnowitzChapter URL: pages in book: (p. 357 - 382) ,institutions,andpolicies, wouldbesurprisingif thecharacterofthebusinesscycledidnotchan geinresponsetosuchfar-reachingdevelopmen tsastheGreatContractionofthe1930sandthep ost-Depressionreforms,theexpansionofgove rnmentandprivateserviceindustries,thedev elopmentoffiscalandotherbuilt-instabiliz ers, is consist-entwithourpriorsthatthedataforth eUnitedStatesandotherdevelopedmarket-ori entedcountriesgenerallysupportthehypothe sisthatbusinesscon-tractions werelessfrequent,shorter,andmilderafterW orldWarII thanbefore( ;seechapter3).

2 Althoughbusinesscycleshavemoderated,they retaina highdegreeofcon-tinuity, relativelycloseandstablerelationshipbetw eenpriorchangesinthisindexandchangesinma croeconomicactivity(VaccaraandZamowitz19 77;Auerbach1982;ZarnowitzandMoore1982;Di eboldandRudebusch1989).Yet,thecurrentlyp opular,smallreduced-formThischapterisco- authoredwithPhillIpBraunandisreprintedfr omAnalyzingModernBusi- , , (Armonk, ,1990). ,seeMoore1983, ; deeperlevel,thenotionthattheprivatemarke teconomyis similareffectinthatit , ,asimplevectorautoregressive(VAR)modelin logdifferencesofrealGNP,money, , ,thedominanteffectsonoutputcomefromthepa stmovementsoftheleadingindex,whiletherol esoftheothervariables(includingthelagged valuesofoutputgrowthitself)aregreatlyred uced( ).Ingeneral,omittingrelevantvariablesina VARwillcausethestandardexogeneitytests,i mpulseresponses,andvariancedecomposition stobebiased, ,includinginterestandinflationratesalong withoutput,monetary,andfiscalvariables, :(1)Thechangesintheleadingindexcanbeinte rpretedbroadlyasrep-resentingtheearlycol lectiveoutcomesofinvestmentandproduction (also,lessdirectly,consumption) , ,butthedemandeffectsmaybestrongerinthesh ortrun.

3 (2) numberofimportantfactorsthatwould ,eachtakenwithfourlags, ,it is notpossibleforsuchmodelstoaccom-modatemo revariablesandstillretaina " is,ofcourse,easytothinkofadditional,poss ibly ,buttheonlywaytoavoidit isbyassuminga fullstructuralmodel,whichcouldbestillmor e ,theimpliedsystem-widedynamics( ,impulseresponsefunctionsandvariancedeco mposi-tions) ,requirelongerseriesofconsistentdatathan areactuallyavailable,4andareprobablyofte nveryimpreciseevenforsmallersystems(seeZ ellner1985andRunkle1987withdiscussion).I ncon-trast,thereissufficientinformationt oestimatetheindividualequationswell, ,whicharelogicallypriortoinferencesonove ralldynamicsbutcarrynocommitmenttothepar ticularin-terpretationsofa ,it :1886-1914,1919-40, , , ,focusingonsimpleexogeneityandneutrality testsfora ,period, variablesandk=4 lags,km2= :2-1982:4cov-eredbyourpostwarseries, ,Symbols,andSourcesofDataNumberVariablea PeriodsbFormeSymboldSourceeNotesonDeriva tion!(1)(2)(3)(4)(5)(6)(7)RealGNP(19721, 2,3alnqB&GAnnualdatafromF&S1982andKuznet sdollars)1961andCommerceNIPA( ),interpolatedquarterlybyChow-Lin(1971)m ethodusingthePersons(1931)indexofindustr ialproductionandtrade2 Implicitpricedeflator1,2,3alnpB&G(nomina lGNP/realGNP)x1003 Monetarybase1,2,3alnbB&GBasedondatainF&S 1970and ,M 12,3alnmlB&G1919~6 ;1947-58:oldMl,FRB:1958-82:newMl,FRB5 Moneysupply,M21,2,3alnm2B&G1886-1907:F 1907-14and1919-80; ;1980-82:FRB6 Commercialpaper1,3 InlevelIB&G1886-89:inNewYorkCity,fromMac aulayrate1938;1890-1914,1919-80:4-6month prime, ;1981-82:6 month,fromFRB7 Commercialpaper2aiB& ,75leadingseriesAmplitude-adjusted(compo site)index,sixleadingseriesCompositeinde xofleadingindicators2323 Inlevel~lnInlevel~ln~ln~lnGXgxGfdcfdfFir estoneFirestoneBlanchardMooreShiskinComm erceBasedonDailyTreasuryStatementsoftheU nitedStates(seeFirestone1960, ,anddata,seasonallyadjusted, ) , ;basedondataforgovernmentspending,debt,a ndtaxesAnalyzedinMoore1961, , ;basedonspecific-cyclephases.

4 DatafromMoore1961, , , ;DatafromNBER filesThecompositeindexof12leadingindicat orsminusthreecomponents:M2inconstantdoll ars,changeinbusinessandconsumercreditout standing,andtheindexofstockprices, :1886-1914;period2:1919-40;period3 ~ln:firstdifferenceinnaturallogarithm;~ ; :BalkeandGordon1986, :Firestone,1960, ,Blanchard:BlanchardandWatson1986, , :BusinessConditionsDigest( ).362 ChapterTwelvesymbol, ,countingonepertimeperiod( and3). ,unitroottestsindicatethatinsomecases,le velsofa seriesoughttobeusedinoneperiodanddiffere ncesinanother.( ,whichcoversthestatisti-calframework.)As shownincolumn4, ,thefiscalindexin1949-82,andtheinterestr ateinbothoftheseperiodsarelevelseries; ;capitallettersforthosecastinlevels( , ).Theseriesrelatingtooutput(q),prices(p) ,the~ltemativemonetaryaggregates(b,ml'm2 ),andinterest(lori)arestapleingredientso fsmallreduced-form,orVAR, ,threerepresentfiscalvariables(8-10).For thepostwarperiod,thereisanindexcombining federalspending,debt,andtaxes(G).Forthei nterwarandprewarperiods,therearetwosegme ntsofthefederalexpendituresseries(gxandG X,respectively).

5 Finally,therearethreedifferentindexesofl eadingindicators(11-13).Theonlysuchserie spresentlyavailablefor1886-1914is a diffusionindexbasedonspecificcyclesinind ividualindicators(lde).Thecompositeindex esfor1919-40and,particularly,for1949-82( ldandl,respectively) "standard"historicalestimatesofGNPbefore WorldWarII,basedmainlyontheworkofKuznets ,Kendrick,andGallman, (1986b).5 Thesedataareconstructedfromthestan-dards eriesbymeansofquarterlyinterpolatorswhic hincludethePersons1931indexofindustrialp roductionandtradebefore1930,theFederalRe -serveBoard(FRB)industrialproductioninde xfor1930-40,constantterms, a sourceofunavoidableerroriftheunitperiodi s tobeshorterthana ,theimplicitpricedeflator(1982=100),andr ealGNPfor1869-1929arepresentedinBalkeand Gordon(1989). concludesthatrealGNPwasontheaverageabout asvolatileasthetraditionalKuznets-Kendri ckseriesindicate, varyingassortmentofweightedandsplicedser ies onbankclearingsoutsideNewYorkCity,produc tionofpigironandelectricpower,constructi oncontracts, ,butit is ,leadingtonewestimatesbyRomer(1986and198 7)andBalkeandGordon(1988).

6 Romer' periodonthepre-WorldWarI (Lebergott1986;Weir1986;BalkeandGordon19 88).It ismainlyforthisreasonthatwedonotuseRomer ' (monthlysinceMay1907,biennialearlier) ,too, (likethoseforincome,output,andprices) (Macaulay1938andFRB)isofgoodquality,atle astina designedtomeasure"theeffectoffiscalpol-i cyonaggregatedemandat giveninterestrates"(BlanchardandWatson19 86, ).Thisseriesmovescountercyclicallymostof thetime, ,nocomparablecomprehensiveindexis available,andweusetheseriesonfederalexpe ndituresfromFirestone('1960).FromtheComm erceDepartment'sleadingindexfor1949-82,w eexcluderealmoneybalances(M2deflatedbyth econsumerpriceindex) (Standard&Poor'sindexof500commonstocks)m aybestronglyaffectedbymonetaryandfiscald evelopments,andadoptinga con-servativebias, ,inventoryinvestment, :averageworkweek,newordersforrailroadcar loadingsandnetton-milesoffreight,indexes ofvolumeofmanufacturingandmin-ing, , (seechapter7). averageworkweek,manufacturing;averagewee klyinitialclaimsforunemploymentinsurance ; vendorperformance(Percentageofcompaniesr eceivingslowerdeliv-eries);changeinsensi tivematerialsprices;manufacturers'orders inconstantdollars,consumergoodsandmateri alsindustries;contractsandordersforplant andequipmentinconstantdollars;indexofnet businessformation;buildingpermitsfornewp rivatehousingunits; \velvedurable-goodsmanufacturers,nonfarm housingstarts,commercialandindus-trialco nstructioncontracts,newbusinessincorpora tions,andStandard&Poor' ,thoughnarrower,alsoresemblesthatofthepo stwarindex, ,whereasthepostwarindexis ,nocompositeindexofleadingindicatorsexis tsforthepre-WorldWarI.

7 ,wereport onsomeexperimentalworkwiththeonlyavailab leseriesthatsummarizestheearlycyclicalbe haviorofa ,financialmarkettransactionsandasset prices,bankclearings,loansanddeposits,se nsitivematerialsprices,inventoryinvestme nt,newordersforcapitalgoods,con-structio ncontracts, (1961)presentsa basedoncyclicalturnsineachofitscompo-nen ts:a seriesissimplycountedasrisingduringeachm onthofa specific-cycleexpansion(anddecliningothe rwise).Clearly,thetypeofsmoothingim-plic itinthisindexconstructionisill-suitedfor ourpurposesasit wasdesignedfora ,forlackofanyothermeasure, , (computedsoastoensuresymmetricaltreat-me ntofrisesanddeclines)arestandardized,tha tis,expressedasratiostotheirownlong-runm ean, ,andweightedaveragesbyCommerce,butthisma keslittledifferencesincetheCommerceweigh ts,basedonperformancescoresoftheselected indicators, ,theCommerceindexhasa trendadjustedtoequalthetrendintheindexof coincidentindicators(whichisclosetothetr endinrealGNP),whereastheShiskinindexhasn osuchadjustment(Shiskin1961a, ; , ).

8 ,whichareweighted4-monthmovingaver-ages, , "incredibleidentification"ofexistinglarg e-scalesimultaneousequationsystemsledtot herecentpopularityofvectorautoregression s,whichtreatallvariablesasendogenousands hununfoundeda (Sims1980a).Thischapterexaminestheintera ctionswithina ~,tasa genericvariablewithsdenotingthetimeserie s(q,p,b,..)andidenotingthetimeframe(prew ar,interwar,postwar).Foreachseries andtimeframe,weestimateordinaryleastsqua resregressionsoftheform(thesuperscriptii shenceforthomittedforsimplicity)(1)RJXs, t=as+~l+.L .L' ,t-j+et,r=Ij=IwhereR=3,..upto6 seriesandJ=4 t-statisticwhichteststhenullhypothesis(2 )JHo:.L'Ys,j=0j= anF-statisticwhichtests(3)againstthealte rnativethatnotall' (1982), (1988)andCampbellandMankiw(1987).Sims,St ock,andWatson(1986) , ,StockandWatson(1987) ,becauseofbothanexplicitinterestintheres ultsand alsothenecessityofhavingstationaryseries toemploystan-dardasymptotictheory,wecalc ulatea \velveJ(4) +etj=lagainstthegeneralalternative(5) + +etj= <1 and,dependingonthetest, unitrootandis stationaryeitherarounditsmean,whentVsisr estrictedtobe0,orarounda timetrend,whentVsis , , ,interwar,andprewarsampleperiods, testsfora singleunitrootforeachseries, containstestsfora , seriesis stationaryarounditsmean,weestimatetheDic key-FullerT~statisticandtheStock-Watsonq ' (eq.)

9 [5]) seriesisstationaryaroundalineartimetrend , (OLS)thefollowingtransformationofequatio n(5):(6) +tVl+< +L< + );j=2(7)andcalculatingtheadjustedt-stati sticfor< <PsI ~=SE(<p) (< )isthetypicallyreportedstan-darderrorof< (k,m)testforcommontrendsina (8)qj(k,m)=T[Re(~)-1],where~is thelargestreal(Re) ,m)statistictests thehypothesisofkversusmunitrootsforann-v ectortimeseries(m<k$;n). :QuarterlyData,1949:1-1982 'ffTq, +M( * + * * + + * **- ** * ** + + * *- ** *- **- **m, ** *- ** ** ** - ** * ** * *b- +- ** *- ** ** * * ** * ** * ** *- ** ** + **- - ** + ** * ** * *aOnthedefinitionsofthevariables, (inlogs,exceptforI)Lowercaselettersdenot efirstdifferences(inlogs,exceptfori). (1979)statisticcomputedusinga 'fis theStock-Watson(1986)statistic,alsofroma regressionwithfourlags."Tandq,areagain,r espectively,theDickey-FullerandStock-Wat sonstatisticscalculatedusinga denotedby**(exceptfortheDickey-Fullertes ts, );atthe1%levelby*;5%levelby+ ;and10% ,thenullhypothesis(eq.

10 [4])isoneunitroot,k=1,againstthealternat ive(eq.[5])ofnounitroot,m= 'fandq; constant,time, ( ),onlythefiscalindexisstationaryinlevels arounditsmeanaswellasarounda , :QuarterlyData,1919:1-1940:1 UnitRootTestStatisticsbt-StatisticsCSeri esaf~q' ( + +- + * * + * * * ** *- **- + \ - ** ** *- ** ** ** * ** ** * ** * ** + * ** *- ** * **- * ** *- ** *- ** , (inlogs,exceptforl)Lowercaselettersdenot efirstdifferences(inlogs,exceptfori).hf~ denotestheDickey-Fuller(1979)statisticco mputedusinga 'fis theStock-Watson(1986)statistic,alsofroma ,respectively,theDickey-FullerandStock-W atsonstatisticscalculatedusinga denotedby**(exceptfortheDickey-Fullertes ts, );atthe1%levelby*;5%levelby+ ;and10% isnecessarytotakefirstdifferencesofrealG NP,money, timetrend, built-innonstationarycompo-nentconstruct edfromthetrendofthecoincidentindex( , ).Becausethisnonstationarycomponentisimp licitlyrelatedtothetrendrateofgrowthofGN Pandwetakefirstdiffer-encesofGNP, ,Stock,andWatson(1986),thepresenceofsign ificanttrendsintheseriesofmoneyandpricec hangesmakesit necessarytoincludea ( ), :QuarterlyData,1886:1-1914:4 UnitRootTestStatisticsbt-StatisticscSeri esa7~ * + + * ** ** *- + * + * * *- ** * ** + * ** * ** * * **- *- ** * *- **- *- ** ** * ** * ** *- ** *- ** + * ** * ** *- ** * ** + , (inlogs,exceptforI)Lowercaselettersdenot efirstdifferences(inlogs,exceptfori).