HOUSING AND THE BUSINESS CYCLE: DATA APPENDIX

Transcription

1 HOUSING AND THE BUSINESS CYCLE: DATA APPENDIX Morrs A. Davs 1 Federal Reserve Board of Governors 2 h and C S. NW Washngon DC 2551 U.S.A. mdavs@frb.gov Jonahan Heahcoe Duke Unversy Deparmen of Economcs 21B Socal Scences Buldng Durham NC heahcoe@econ.duke.edu Sepember 2 Absrac Ths paper serves as an emprcal companon pece for "Housng and he Busness Cycle" by Davs and Heahcoe (2). A large par of he paper s devoed o documenng he growh varably and co-movemen of major macroeconomc varables. We pay parcular aenon o he busness cycle facs relang o resdenal nvesmen and house prces. We descrbe a mehod for usng he NIPA Inpu-Oupu ables o allocae value added n fnal goods across hree nermedae goods secors: consrucon manufacurng and servces. We apply hs mehod o esmae he 1992 shares of he hree nermedae secor npus n consumpon resdenal nvesmen busness nvesmen and GDP. We consruc me seres for Solow resduals n our hree nermedae secors usng secor-specfc esmaes of capal s share and annual daa on secor oupus capal socks and hours. Fnally we use a GMM approach o conssenly esmae quarerly AR(1) producvy processes gven hese annual resduals. JEL Classfcaon: E E1 E2 1 The opnons expressed here are hose of he auhors and do no necessarly reflec he vews of he Federal Reserve Sysem of s saff. 1

2 I Fnal Goods Oupu To show he rends n he composon of GDP by broad caegory of fnal demand (prvae consumpon oal fxed nvesmen and governmen consumpon) n fgure 1 we graph he shares of nomnal GDP of hese caegores snce The op wo panels of hs fgure show ha he prvae consumpon share of GDP has been rsng and he governmen consumpon share of GDP has been declnng. I appears hese movemens largely offse; no obvous upward or downward rend s apparen n he oal fxed nvesmen o GDP rao shown n he boom panel. In able 1 we repor he cyclcal volaly of quarerly real GDP and s componens; we also repor he cross-correlaon of GDP wh fve leads and lags of he componens of GDP. Even hough he NIPA daa have been revsed a number of mes 2 In wha follows oal fxed nvesmen refers o gross fxed prvae and governmen nvesmen ncludng resdenal nvesmen and excludng governmen defense nvesmen; governmen consumpon refers o governmen consumpon and defense nvesmen expendures; and prvae consumpon refers o NIPA personal consumpon expendures. 3 Nomnal and Real Gross Domesc Produc by Fnal Demand are avalable n Tables 1.1 and 1.2 respecvely of Seleced NIPA Tables publshed by he US Deparmen of Commerce n he Survey of Curren Busness. Nomnal and real personal consumpon expendures by Major Type of Produc are avalable n NIPA ables 2.2 and 2.3. Nomnal and real governmen consumpon expendures and gross nvesmen by ype are avalable n NIPA ables 3.7 and 3.8. Nomnal and real gross prvae fxed nvesmen by ype are avalable n NIPA ables 5.4 and 5.5. The Chan-Type Quany ndexes used o conver nomnal quanes o real $1996 quanes are avalable n Tables ; hese ables also conan Chan-Type Prce Indexes we use o compue relave prce movemens of he componens of GDP. 2

3 snce Cooley and Presco (1995) and more daa are a our dsposal he (comparable) numbers n hs able are que close o hose n her Table 1.1. I.1 Consumpon NIPA consumpon ncludes spendng on nondurables goods and servces as well as spendng on durable goods and he (largely mpued) consumpon of housng servces. 4 As shown n fgure 2 durable goods purchases accoun for approxmaely 12 percen of NIPA consumpon expendures; he consumpon of housng servces accouns for anoher 15 percen. Durable goods purchases are a small fracon of oal consumpon expendures so her ncluson does no subsanally change he consumpon prce ndex (shown n he op panel of fgure 3) even hough he prce of durable goods relave o oher consumpon goods has been fallng. The ncluson of durable goods ncreases he busness cycle volaly of consumpon ( ncreases he percen sandard devaon from.87 o 1.33) bu does no affec cyclcal consumpon movemens; see able 1 and he boom panel of fgure 3 for deals. I.2 Toal Fxed Invesmen We dvde oal fxed nvesmen no resdenal fxed nvesmen governmen nvesmen (excludng defense nvesmen) and busness fxed nvesmen. Shown n he op panel of fgure 4 he resdenal and governmen shares of oal fxed nvesmen have 4 In calculang he boh he renal value of enan-occuped and owner-occuped dwellngs NIPA consumpon of housng servces s se equal o he produc of a ypcal ren per dwellng mpued n he case of owner-occuped dwellngs mes he number of dwellngs. See pp. 21 and 6 62 of Personal Consumpon Expendures (199) publshed by he US Deparmen of Commerce for deals. 3

4 remaned relavely sable a approxmaely 25 and 15 percen respecvely mplyng he busness fxed nvesmen share of nvesmen has also remaned sable. The nonresdenal srucures share of nvesmen has fallen que a b snce he early 198s (boom panel of fgure 4) ndcang ha he equpmen and sofware nvesmen share (he oher componen of busness fxed nvesmen) has rsen. The busness-cycle properes and rend prces of he dfferen fxed nvesmen seres are que dfferen. Table 1 shows ha resdenal nvesmen s more han wce as volale as oal nonresdenal fxed nvesmen defned as he sum of busness fxed nvesmen and governmen non-defense nvesmen; resdenal nvesmen also clearly leads GDP whle oal nonresdenal fxed nvesmen lags GDP. As for prces he op panel of fgure 5 confrms ha oal nonresdenal fxed nvesmen prces have rsen less rapdly han resdenal fxed nvesmen prces; shown n he boom panel he dfference s accouned for by he prce ndex of equpmen and sofware (he do lne): he nonresdenal srucures (dash lne) and governmen non-defense nvesmen (sold lne) prce ndexes are nearly dencal. I.3 Governmen Governmen consumpon (as we have defned ) s slghly more volale han GDP (able 1) does no appear o be cyclcally correlaed wh GDP (able 1) and has approxmaely he same rae of prce nflaon as prvae consumpon expendures (fgure 6). 4

5 II Fnal Goods Prces: Housng In he model prvae consumpon oal nvesmen and governmen consumpon share a common prce and he prce of housng s expressed relave o he prce of hs compose good. As we have seen he governmen consumpon and prvae consumpon prce ndexes are nearly dencal bu fgure 7 shows ha oal nonresdenal fxed nvesmen prces (do lne) have been fallng relave o consumpon (and hus governmen) prces he sold lne. Toal nonresdenal fxed nvesmen accouns for less han 15 percen of nomnal GDP excludng resdenal nvesmen so he appropraely calculaed prce ndex for GDP excludng resdenal nvesmen (dash lne) has almos he same me-seres pah as he prce ndex for consumpon also shown n fgure 7. Because hese wo prces ndexes are nearly dencal n he work ha follows we use he prce ndex for consumpon o compue all real relave prces. To consruc he real relave prce of housng we need o denfy a prce ndex for one un of undeprecaed housng sock. We know of wo such prce ndexes: he prce ndex for new resdenal nvesmen whch comes drecly from NIPA able 7.6 and he Chan-Type Annual-Weghed Prce Index (Fsher Ideal) of New One-Famly Houses Sold Includng Value of Lo a seres ha s publshed by he Bureau of he Census. 5 The wo prce ndexes dffer along a number of dmensons. Frs he NIPA prce ndex racks ncreases o he cos of npus whle he Census drecly measures 5 See able 7 of he Aprl 2 ssue of he Curren Consrucon Repors publshed by he US Deparmen of Commerce. Noe ha a smlar Laspeyres ndex s avalable as able 7a. 5

6 changes o new house prces afer conrollng for movemens n house prces caused by changes o house arbues. 6 Second he NIPA prce ndex measures changes o he prce of npus used o buld all resdenal srucures (sngle famly mul-famly and oher ) whle he Census prce ndex only apples o new one famly houses. In 2:Q1 resdenal nvesmen n sngle famly homes consued approxmaely 55 percen of oal resdenal nvesmen. Fnally changes o he Census prce ndex nclude changes o he prce of land whle he BEA measure does no. Over he avalable range of daa (he Census prce seres sars n 1979:Q1) he wo house prce ndexes have very smlar rends as shown n he op panel of fgure 8. The boom panel of fgure 8 shows he real relave prce of new houses for boh seres. The relave prce seres are calculaed by dvdng he housng prce ndexes by he prce ndex for consumpon. The resulng rao s self a prce ndex; hs ndex does no yeld he relave prce of housng a a gven me (n he graph he relave prce of housng s normalzed o 1. n 1979:Q1 for boh seres) bu changes o he ndex are reflecve of changes o he relave prce of housng. Ths graph shows ha he real prce of a new house has ncreased abou 15 percen snce If old houses are good subsues for new appears ha houses do no apprecae quckly f a all. Table 2 repors he busness cycle relaonshp of real relave house prces measured usng he NIPA seres GDP and housng nvesmen whle able 3 shows hese 6 In he Census prce ndex characerscs of houses such as he square-fooage locaon number of bedrooms and bahrooms ec. are held consan based on he knds of houses sold n For a descrpon of he hedonc regresson mehodology used o consruc hs prce ndex see Appendx A of he March 1997 ssue of Curren Consrucon Repors. 6

7 busness cycle relaonshps wh he Census real relave house prce seres. The boom row of able 3 shows ha he wo prce seres are hghly correlaed (he conemporaneous correlaon coeffcen equals.85) bu he Census seres s 1.4 mes more volale han he NIPA seres. 7 These ables also yeld four busness-cycle facs ha are robus o he choce of he real house prce seres. Frs real house prces are mldly procyclcal. Second resdenal nvesmen and real house prces are mldly conemporaneously correlaed a cyclcal frequences. Thrd resdenal nvesmen leads house prces bu house prces negavely lead housng nvesmen shown n he RES row of boh ables. Fnally new house prces measured by he NIPA seres are less volale han GDP; accordng o he Census seres hey are equally as volale as GDP. Ths las pon s mporan because s ofen saed ha house prces a busness cycle frequences are much more volale han GDP. We beleve such clams mus be relaed o he busness-cycle volaly of he average or medan sale prce of exsng homes. The average prce seres has he advanage n ha represens he ypcal sale prce of exsng homes and no jus he prce of new resdenal nvesmen. The dsadvanage s ha does accoun for any year-o-year dfferences n he ypcal qualy of homes ha are sold; hgh year-o-year varably may be due o measuremen of boh prce and qualy varaon. 8 7 The conemporaneous correlaon marx of oupu consumpon nvesmen governmen consumpon and he Census new house prce seres s avalable n able 4. 8 When we say qualy varaon we are referrng o varaon n house arbues. One such arbue s house sze so n hs sense qualy varaon s dencal o quany varaon. 7

8 To aemp o uncover he varaon n he real average sale prce seres due o qualy varaon we compare he Average sales prce of knds of houses sold n 1992 (esmaed from prce ndex) wh he Average sales prce of houses acually sold; boh daa seres are avalable from able 8 of Aprl 2 ssue of Curren Consrucon Repors. The frs seres ( Adjused ) repors he average sellng prce of a new house f house characerscs were held consan a he ypcal house bul n 1992 whereas he second seres ( Unadjused ) smply repors he average sellng prce of a new house. 9 The op panel of fgure 9 plos he me-seres of he wo prce seres over her avalable range of daa. 1 Ths panel ndcaes ha house qualy has been rsng over me. The boom panel shows he cyclcal varaon n he wo prce seres afer boh have been deflaed usng he consumpon prce ndex. New house qualy appears o vary over he busness cycle. When measured from 1982:Q1 o 1997:Q4 he percen sandard devaon of unadjused new house prces s 3.4% nearly double ha of he adjused seres (1.8%). 9 The adjused seres s consruced usng he Laspeyres (no Fsher) prce ndex. See he March 1997 ssue of Curren Consrucon Repors page A-4. 1 The unadjused seres sars n 1963:Q1 whle he adjused seres sars n 1977:Q1. 8

9 III Aggregae Socks of Capal and Housng To calbrae he oal producve capal sock of he model (he model varable k) we add NIPA esmaes of he nonresdenal fxed prvae capal sock o he NIPA esmaes of he federal governmen non-defense and he sae and local governmen capal socks. For he housng sock h we use NIPA esmaes of resdenal fxed prvae capal. 11 We exclude he value of he sock of durable goods owned by consumers from he producve capal sock and housng sock. NIPA esmaes year-end capal sock valuaons.e. he repored 1994 sock of prvae busness capal refers o he esmaed sock of hs capal on December In conras he NIPA oupu nvesmen and consumpon daa refer o he flow of hese varables over a gven me perod. For calbraon we defne he year capal sock as he geomerc mean of he NIPA repored capal sock n year and -1. Ths mddle-of-year capal sock measure more closely algns oupu (n a gven year) wh he capal sock used o produce ha oupu; more closely algns he argumens of household uly n a gven year (consumpon durng he year hours worked durng he year and he housng sock); and allows us for more accurae calculaon of deprecaon raes shown laer n hs secon. The op panel of fgure 1 plos he rao of he nomnal sock of producve capal o he nomnal housng sock. Snce 1955 he oal producve sock has been on average 1.53 mes larger han he housng sock. As fgure 11 shows governmen nondefense capal accouns for slghly more han hry percen of oal producve capal 11 See Fxed Reproducble Tangble Wealh n he Uned Saes (1999) publshed by he US Deparmen of Commerce for deals. 9

10 (op panel) and sae and local capal accouns for almos nney percen of hs nondefense governmen capal (boom panel). Fgure 12 plos he rao of he annual nomnal producve sock of capal o annual nomnal GDP. From 1955 hrough 1998 he average rao has been 1.53 shown by he doed lne. The fac ha he rao of he capal sock o he housng sock has averaged he same as he rao of he housng sock o GDP mples on average he sze of he housng sock equals annual GDP. The wo panels of fgure 13 graph our esmaed annual deprecaon raes (n percen) of he producve sock of capal and he housng sock. We calculae annual deprecaon raes for year by dvdng he NIPA esmae of he value of nomnal deprecaon 12 n year by our mddle-of-year measure of he year nomnal sock. The op panel of fgure 13 shows ha he housng sock deprecaes a approxmaely 1.6 percen per year. Ths s much lower han he deprecaon rae of producve capal 5.3 percen per year shown n he boom panel. Ths dfference s arbuable o he hgh deprecaon rae (6.6 percen per year) of busness fxed capal shown n he op panel of fgure 14; a 2.3 percen per year he deprecaon rae on governmen capal (boom panel) s no much hgher han ha of he housng sock. 12 Nomnal deprecaon ables are ncluded as par of he NIPA supplemenary capal sock ables avalable on he web a hp:// 1

11 IV Inermedae Goods To decompose oupu no value added by nermedae ndusry we use he annual NIPA Gross Produc by Indusry (GPO) ables. 13 These ables parse Gross Domesc Income (GDI) equal o GDP mnus he sascal dscrepancy no value added orgnang from 1 dfferen ndusres. These NIPA ndusres are Agrculure foresry and fshng (AFF); Mnng; Consrucon; Manufacurng; Transporaon and publc ules; Wholesale rade; Real rade; Fnance nsurance and real esae (FIRE); Servces; and Governmen. 14 Fgure 15 shows he fracon of nomnal prvae value added defned as GDI excludng governmen value added arbued by NIPA o he varous goods-producng ndusres (op panel) and servces-producng ndusres (boom panel) as classfed n able 3 of Lum and Yuskavage (1997). The op panel shows ha from he share of nomnal value-added from manufacurng has seadly declned whle he boom shows he shares of value added from servces (do-do-dash lne) and FIRE (do-dash lne) have ncreased. In 1947 goods-producng ndusres accouned for 5 percen of prvae value added; n 1998 hese ndusres accouned for only 26-1/2 percen. We om governmen value added from our calculaons (approxmaely 14 percen of GDP) 15 because we do no wan o arbrarly assgn governmen value added o goods- or servces- producng ndusres Recen GPO daa are avalable n he June 2 ssue of Survey of Curren Busness. GPO ndusry classfcao ns follow he 1987 Sandard Indusral Classfcaon sysem. 15 From 1955 hrough 1998 governmen value added has accouned for an average 13.8 percen of GDP (wh a sandard devaon of.77 percen). 11

12 IV.1 Capal Shares Fgure 16 graphs he capal share by NIPA ndusry for goods-producng (op panel) and servces-producng (boom panel) ndusres. For each ndusry we calculae he capal share n year θ as COMP θ = 1. (A.1) VA IBT PRO where COMP s nomnal compensaon of he employees n he ndusry n year VA s he nomnal value added of he ndusry n year IBT are he ndusry s nomnal ndrec busness ax and nonax lables and PRO s ha ndusry s nomnal propreor s ncome n year. 16 As n Cooley and Presco (1995) p. 19 we derve equaon (A.1) by assumng he share of IBT and PRO gong o labor n year s (1-θ ). The op panel of Fgure 16 demonsraes ha he capal share of he consrucon ndusry (dash lne) s he lowes of all goods-producng ndusres. The boom panel shows ha he capal shares of wholesale rade (do lne) real rade (dash lne) and servces (do-do-dash lne) are que smlar; n conras he capal share of ransporaon and publc ules (sold lne) s a b hgher and rsng over me whle he capal share of FIRE (do-dash lne) s much hgher han all of he oher capal shares. Ren pad on housng s ncluded n FIRE oupu explanng why he capal share n FIRE s remarkably large. Lne 58 of able 7 of he Gross Produc by Indusry ables (see foonoe 16) lss he value added of Nonfarm housng servces a subcaegory of 16 These daa are avalable n Lum Moyer and Yuskavage (2) and Naonal Income and Produc Accouns Tables n he December 1999 ssue of Survey of Curren Busness. 12

13 FIRE; hese servces approxmaely equal 8 percen of he nomnal consumpon of housng servces (lsed n Table 1.1 of he NIPA). From he NIPA GPO ables arbued an average of 5 percen of FIRE value o nonfarm housng servces. 17 When hese servces are purged from FIRE value added he average FIRE capal share drops o.28 slghly hgher han he average capal share of wholesale rade (.24). The me-seres varably of hs modfed FIRE capal share seres s que hgh rangng from almos.45 n 1948 o.11 n 1968 and hen.38 n 1986 and beyond; see fgure 17. As menoned he NIPA gross produc by ndusry ables nclude daa for nne prvae ndusres bu n he model we only have hree nermedae goods secors consrucon manufacurng and servces. To calbrae he model s consrucon secor we use daa from he NIPA consrucon ndusry. For manufacurng we group ogeher all NIPA goods-producng ndusres excep for consrucon: AFF Mnng and Manufacurng. To calbrae he servces secor of he model we aggregae all of he NIPA servces-producng ndusres excludng FIRE: ransporaon and publc ules wholesale rade real rade and servces. We exclude he modfed FIRE seres because he varably of s capal share looks suspcous The percenage has been seadly declnng snce 1963 when was 63 percen. In 1998 he percenage was 42-1/2. 18 Ths excluson also eases he compuaon of he echnology shocks o he model s servce secor: Incluson of he modfed FIRE seres requres excludng an arbrary fracon of FIRE capal sock n order o calculae a Solow resdual. 13

14 We calculae he capal share of he model s hree nermedae goods secors 19 {bms} a dae (θ ) as j θ = 1. (A.2) { VA j IBT PRO } j j j COMP where j ncludes he AFF mnng and manufacurng ndusres f =m and he ransporaon and publc ules wholesale rade real rade and servces ndusres f =s. 2 The hree panels of fgure 18 plo he me-seres pah of he capal share of he hree nermedae goods secors n he model as we have defned hem. The op panel shows he capal share of he consrucon ndusry; s average value from s.13. The mddle panel graphs he capal share for he manufacurng secor of he model ( Model Secor ) along wh he capal share of he NIPA manufacurng ndusry Manufacurng Only. The average capal shares over he perod are.31 and.26 respecvely. The boom panel shows he capal share for varous servce measures. The average capal share of he NIPA servces ndusry n he pos-1948 perod s only.13 shown by he dashed lne n he boom panel. The ncluson of he wholesale and real rade ndusres rases he average capal share o.18 he doed lne; he ransporaon and publc ules ndusres ncreases he average capal share o.24 he sold lne n he boom panel. For calbraon and he compuaon of echnology shocks by nermedae ndusry we se j 19 Conssen wh he model secon of he paper b sands for he consrucon secor m manufacurng and s servces. 2 j equals he consrucon ndusry f =b. 14

15 θ b (consrucon) θ m (manufacurng) θ s (servces) IV.2 Annual Technology Shocks We assume ha he hree nermedae goods are produced usng Cobb-Douglas echnology wh homogenous capal and labor npus: 1 θ ( z n ) { b m s} θ x = k (A.3) In hs equaon x s he real oupu of nermedae good secor a year k s he real capal sock of secor a year n s he supply of labor o secor a year and z s a labor-augmenng supply or echnology shock o secor a dae. Takng he naural logarhm of (A.3) and solvng for he log of he secor specfc echnology shock yelds log 1 = (A.4) ( ) [ log( x ) θ log( k ) ( 1 θ ) ( n )] z log 1 θ for {bms}. Calculaon he me-seres pah of he log of he consrucon manufacurng and servce secor echnology shocks herefore requres daa on real oupu real capal sock and hours worked by ndusry. Table 9 of he Gross Produc by Indusry ables (see foonoe 16) lss he Quany Indexes for Gross Domesc Produc by Indusry. For each ndusry mulplyng hs quany ndex by 1996 nomnal ndusry oupu yelds real oupu n chan-weghed 1996 dollars. Alhough hese calculaons drecly yeld real oupu of he consrucon secor of he model o consruc real oupu n 1996 dollars for he manufacurng secor we need o chan-wegh (correcly add) he real oupu of he AFF mnng and manufacurng ndusres. Smlarly o consruc real oupu of he 15

16 servce secor we chan-wegh real oupu of he ransporaon and publc ules wholesale rade real rade and servces ndusres. To consruc he real capal sock by model secor we perform analogous calculaons usng ables 5 and 6 of he NIPA Fxed Reproducble Tangble Wealh; 21 we creae our mddle-of-perod capal sock measure by calculang he geomerc mean of he resulng year and -1 chan-weghed real capal sock seres. Annual hours worked n consrucon are drecly observable from able 6.9c of he annual NIPA ables Hours Worked by Full-Tme and Par-Tme Employees by Indusry Group. 22 To creae annual hours worked n he manufacurng and servce secors we add ogeher he hours of he approprae consuen ndusres all of whch are also locaed n hs able. Table 5 shows he busness cycle volaly and cross-correlaon wh annual GDP of real annual oupu hours worked and real capal for he consrucon manufacurng and servces secor for To show he nfluence of a dfferen sample range and dfferen flerng parameer on he repored busness-cycle sascs of able 1 hs able also repors he busness cycle volaly and cross-correlaon wh GDP of he 21 Table 5 of hese ables repors he nomnal value of ndusry-specfc capal socks and able 6 lss he quany ndexes. In hese ables ndusres are classfed accordng o he 1987 Sandard Indusral Classfcaon sysem (SIC). See Fxed Reproducble Tangble Wealh n he Uned Saes (1999) for deals. 22 See Naonal Income and Produc Accouns Tables n he December 1999 ssue of he Survey of Curren Busness. 23 The real oupu by ndusry daa s avalable from 1977 o he curren. The BEA has no plans o release pre-1977 real oupu by ndusry daa. For he conemporaneous correlaon marx of annual oupu consumpon nvesmen governmen spendng house prces and nermedae secor oupu hours worked and capal see able 6. 16

17 componens of fnal demand (consumpon oal nonresdenal nvesmen resdenal nvesmen and governmen consumpon). A few facs emerge from hs able. Consrucon oupu s more han wce as volale as manufacurng oupu (whch self s nearly wce as volale as servces oupu). Consrucon and manufacurng hours and capal are more volale han he respecve servces seres. Also he capal socks of he hree nermedae secors are more volale han aggregae capal and he hours worked of consrucon and manufacurng are more volale han aggregae hours. Gven our esmaes of he secor-specfc capal shares we calculae he meseres pah of he annual logarhm of he echnology shocks of he hree secors from 1977 (see foonoe 23) hrough Fgure 19 shows he me-seres pah of he log shocks log(z ). 24 To esmae he rae of growh of he echnology shock we regress he log shocks on a consan and a me rend. As depced n hs fgure he echnology shock has grown a 3.67 percen per year (.91 percen per quarer) n he manufacurng secor 1.19 percen per year (.3 percen per quarer) n he servce secor and. percen per year n he consrucon ndusry. 25 In calbraon of he model we se g zb (consrucon) g zm (manufacurng) g zs (servces) The 1977 value has been normalzed o 1. n hs fgure. The growh raes shown n fgure 19 are he focus of conroversy among economss: some economss fnd hard o beleve ha here has been no sgnfcan ncrease n consrucon mul-facor producvy for example. See Jorgenson and Sroh (2) Gullckson and Harper (1999) Corrado and Slfman (1999) and Peper (199). 17

18 Fgure 2 shows he lnearly derended logarhm of he annual secor specfc echnology shocks denoed log ( z~ ) n he model secon of he paper. The derended log echnology shock o servces (dash lne) s clearly less volale han ha o consrucon (sold lne) and manufacurng (do lne). Sascs verfy he nuon mpared by hs graph: he sandard devaon of he derended logarhm of he servces echnology shock (.17) s one-hrd he sze of ha of consrucon (.54) and manufacurng (.44). IV.3 Quarerly Technology Shocks The lnk beween he quarerly logged derended resduals (ha we do no observe) and he logged derended annual resduals (ha we observe) s sraghforward bu algebracally complcaed. For secor of year n quarer q denoe he Solow resdual as z q. The quarerly analog of (A.4) s log 1 ( ) [ log( x ) θ log( k ) ( 1 θ ) ( n )] z q q q log q 1 θ = (A.5) where x q s quarerly real oupu of secor n quarer q of year (expressed a an annual rae) n q are quarerly hours worked n secor n quarer q of year also expressed a an annual rae and k q s he capal sock of secor n quarer q of year. Averagng boh sdes of hs equaon over he quarers n a year and explong he properes of logarhms yelds log 4 1 = 1 θ 4 q= 1 z q log 4 4 q= 1 x q θ log 4 4 q= 1 k q ( 1 θ ) log 4 4 q= 1 n q (A.6) 18

19 19 (A.6) shows ha f annual oupu capal and hours were equal o he geomerc mean of he quarerly numbers hen he annual logarhm of he Solow resdual equals he logarhm of he geomerc mean of he quarerly Solow resduals. Our mddle-ofyear capal sock measure (he geomerc mean of he year-end annual capal socks) should be farly close o he geomerc mean of quarerly capal socks. In he NIPA however annual oupu and hours are he arhmec average of quarerly oupu and hours (expressed a annual raes). Our nuon however s ha he geomerc means of quarerly oupu and hours approxmaely equal he arhmec means. We assume ha he quarerly logged derended resduals n each secor follow a frs-order auoregressve process wh auoregressve coeffcen a.e. for q>1 ( ) ( ) q q q e z a z 1 ~ log ~ log + = (A.7) and ( ) ( ) ~ log ~ log e z a z + = (A.8) for he frs quarer. Gven hs process we know ( ) ( ) q q q e z a z 1 4 ˆ ~ log ~ log + = (A.9) wh ˆ ˆ ˆ ˆ e e a e a e a e e e a e a e a e e e a e a e a e e e a e a e a e = = = = (A.1) Takng he average of boh sdes of equaon (A.9) for q=1 4 n each year produces. ˆ 4 1 ~ log ~ log = = = + = q q q q q q e z a z (A.11)

20 The varable on he lef-hand sde of (A.11) s he logarhm of he annual derended resdual and he varables on he rgh hand sde are he logarhm of he prevous year s annual derended resdual and a mean zero error. 26 (A.11) reveals ha f he quarerly logged derended Solow resduals follow a frs-order auoregressve process hen he annual resduals follow a frs-order auoregressve process as well. For convenence rewre (A.11) n erms of he annual derended resdual as ( ~ z ) α log( ~ z ) + ε log = 1 (A.12) where 4 4 α = a and ε = ˆ e q 4 q= 1 1. Ordnary leas squares of equaon (A.12) does no produce an unbased esmae of α because he error erm ε s correlaed wh he regressor ~ z Usng he annual logged derended Solow resdual daa from we esmae α of (A.12) for each of he hree nermedae ndusres usng GMM. We assume ha e q s ndependenly drawn over me and wh hs assumpon any varable daed year -2 or earler s a vald nsrumen as long as he varable s correlaed wh boh ~ and ~ z 1. Ths yelds an nfne number of possble nsrumens. We have z found ha dfferen nsrumens and ses of nsrumens yeld dfferen (unbased and conssen) esmaes of α for {bms}; our esmaes of α m and α s are especally sensve o he choce of nsrumens. For he consrucon and manufacurng ndusres (=bm) we use he year -2 value of he annual log derended consrucon resdual as he 26 We assume E[ e ] q =. q 27 e -1q s a componen of he error erm (see equaon (A.1)) and e -1q s correlaed wh z -1q by (A.7). 2

21 nsrumen; for he servces ndusry (=s) we use he annual log derended manufacurng resdual as he nsrumen. We use hs parcular se nsrumens because hey yeld he larges esmaes of α for each ndusry. Our esmaes for α and (and hus a = (α ) 1/4 ) are 28 Consrucon Manufacurng Servces α (annual) a (quarerly) Gven hese esmaes we se he sandard devaons and correlaons of he annual nnovaons ε as Consrucon Manufacurng Servces Sd. Dev. (ε ) Cor(ε b ε m ) Cor(ε b ε s ) Cor(ε s ε m ) To compue he varance and covarances of he quarerly nnovaons gven he above annual esmaes noe ha e ˆ q equals ( / 4)[ A e~ 4 ] q= 1 1 ι where ι 4 s a 1x4 vecor of he elemen 1 ~ e = [ e e e e e e e ] and a 1 a a 3 2 a a a 1 A =. (A.13) 3 2 a 1 a a 3 2 a a a 1 28 The based OLS esmaes of α b α m and α s are and.446 respecvely. 21

22 As noed we assume ha he nnovaons o he quarerly log derended resdual { e e } b q mq e sq are ndependenly dsrbued over me bu may be conemporaneously correlaed. Usng he seral ndependence we derve he followng expresson from (A.13) E 1 [ ε ε ] [ ι A A ι ] E[ e e ]. = 4 j 4 q j (A.14) 16 j q (A.14) relaes he varances ( = j) and covarances ( j) of he quarerly nnovaons e q o he varances and correlaons of he error n (A.12) ε. Based on hs relaonshp we calbrae Consrucon Manufacurng Servces Sd. Dev. (e q ) Cor(e bq e mq ) Cor(e bq e sq ) Cor(e mq e sq )

23 V Fnal Good Technology We assume ha frms ha produce fnal goods aggregae nermedae goods accordng o he followng echnology B j M j S j y = b m s (A.15) j j where S j equals 1 - B j - M j. In hs equaon y c s he oupu of he consumponnvesmen fnal good y h s he oupu of he resdenal nvesmen fnal good {b c m c s c } are he quanes of nermedae goods (consrucon manufacurng and servces) used n producon of he consumpon-nvesmen good {B c M c S c } are he shares of consrucon manufacurng and servces n producon of he consumpon-nvesmen good and {b h m h s h B h M h S h } are he quanes and shares of he nermedae goods n he producon of resdenal nvesmen. Denoe he prce of nermedae goods n uns of he consumpon-nvesmen good as p for {bms}. Derved n he model secon of he paper he frs order condons of he fnal goods frms mply ha j j and p b p m B = b c m c s c c = M c = S c (A.16) yc yc y c p s p b p m B = b h m h s h h = M h = S c (A.17) ph yh ph y h p h yh p s where p h s he prce of one un of resdenal nvesmen oupu expressed n uns of he consumpon-nvesmen good. Prof maxmzaon hus mples ha he share parameers {B c M c S c } and {B h M h S h } equal he rao of he value of purchased nermedae goods o he value of fnal goods ha are produced. 23

24 To calculae he value added from he consrucon manufacurng and servce secors no PCE busness fxed nvesmen and resdenal nvesmen we use he Use able of he 1992 NIPA Inpu-Oupu Accouns (IO) for he U.S. Economy. Ths IO Use able has wo complemenary sub-ables. In he frs he oal sales of an ndusry (for all nermedae ndusres) are allocaed o value-added of ha ndusry and sales from oher ndusres. In he second fnal sales of each ndusry comprsng he componens of fnal demand (PCE gross prvae fxed nvesmen ec.) are lsed. 29 Taken ogeher hese wo ables allow a decomposon of PCE prvae nvesmen ec. no value added by nermedae ndusry. 3 In row 6 of box 1 on he nex page we ls he 1992 fnal sales 31 of he consrucon manufacurng servce and oher (governmen) ndusres from hs subable of he IO Use able. 32 Fnal sales conss of sales purchased from oher ndusres rows 1 hrough 4 and value added row 5. For example he 68 bllon of sales orgnang from he consrucon ndusry (row 6 of column 1) consss of 212 bllon of sales purchased from manufacurng ndusres (row 2 column 1) 153 bllon of sales purchased from servces ndusres (row 3 column 1) 772 mllon of sales purchased 29 For more deals on he IO accouns see Lawson (1997) and Benchmark Inpu- Oupu Accouns of he Uned Saes 1992 (1998). 3 Noe ha he IO ables do no use he 1987 SIC o classfy nermedae ndusres a pon ha o whch we reurn laer. 31 These sales exclude he value of sales purchased from non-comparable mpors accounng for less han one percen of oal sales for all ndusres. 32 We group ogeher he AFF Mnng and Manufacurng ndusres as he Manufacurng ndusry and he Transporaon Trade Servces and FIRE ndusres as he Servces ndusry. 24

25 Box 1: IO-Use Table (1992) Decomposon of fnal sales (n mllons of curren dollars) by ndusry Consrucon (1) Fnal Sales from Indusry Manufacurng (2) Servces (3) Oher (4) (1) Consrucon (2) Manufacurng (3) Servces (4) Oher (5) Value added (6) Toal sales from oher (row 4 column 1) 594 mllon purchased from whn he consrucon ndusry (row 1 column 1) and 313 bllon of value added (row 5 column 1). In box 1 we subrac mpued renal ncome from housng from he oal sales of he servces ndusry (column 3 row 6) and from he value added of servces (column 3 row 5). 33 The sum of he value added of he nermedae ndusres ncludng he mpued renal ncome from housng equals nomnal 1992 GDP. Ineresngly enough he I/O esmaes of value added by nermedae ndusry row 5 do no equal he 1992 GPO esmaes of value added by ndusry (whch sum o nomnal 1992 GDI). 34 Ths dscrepancy resuls because he NIPA GPO accouns and I/O accouns use dfferen 33 We defne he mpued renal ncome from housng as I/O code 711 (FIRE ndusry); n 1992 hs value was $45725 mllon. We assume ha NIPA allocaes all mpued renal ncome o FIRE sales mpued renal ncome s enrely FIRE value added and mpued renal ncome s no sold o any of he oher ndusres. 34 For example he 1992 value added of he consrucon ndusry measured by he GPO accouns s 234 bllon (curren) dollars. 25

26 ndusry classfcaons and dfferen daa. 35 We do no correc for hese dfferences n our calbraon procedure. From he nformaon n box 1 we employ an nfne recurson o decompose sales arbued o one of he nermedae ndusres no value added from all four nermedae ndusres. To undersand why we use an nfne recurson consder column 1: fnal sales arbued o he consrucon ndusry nclude some value added from consrucon (row 5 column 1) and some sales purchased from he manufacurng ndusry (row 2 column 1). Bu sales arbued o manufacurng ndusres nclude sales from purchased from he consrucon ndusry (row 1 column 2) and hese sales n urn nclude consrucon value added. To help undersand he recurson we employ denoe he sales arbued o he consrucon manufacurng servces and oher ndusres as S B S M S S and S O respecvely. Furhermore denoe he value added of he consrucon manufacurng servces and oher ndusres as V B V M V S and V O respecvely. We know from column 1 ha S B = (.9)S B + (.3122)S M + (.2256)S S +(.11)S O +(.462)V B and we can wre smlar equaons for he oher hree columns correspondng o he sales from he oher hree ndusres. Usng hese equaons we derve he followng expresson lnkng sales arbued o dfferen nermedae ndusres o value added of he four nermedae ndusres See Parker (1997) for a dscusson of hs ssue. The columns should sum o

27 V B 1 V M VS VO = S B S M S S SO (A.18) As noed he second sub-able of he IO Use able lss componens of demand by sales arbued o nermedae ndusres. Usng he same ndusry classfcaons as before n box 2 we repor an abrdged poron of hs sub-able from he 1992 IO Use able. Columns 1 hrough 3 are drecly coped from he I/O Use able. The I/O use able does no have a resdenal nvesmen column so n column 4 we assume ha all bllon dollars of resdenal nvesmen n 1992 are arbued o sales from he consrucon ndusry. 37 In addon we subrac sales he mpued renal ncome from housng servces from he sales of he servces secor o PCE. Inerpreng hs able n 1992 PCE excludng he mpued ncome from housng conssed of $ of fnal sales from he consrucon ndusry $869.3 bllon of fnal sales from manufacurng ndusres 37 The I/O ables do ls fnal sales arbued o I/O consrucon ndusres 1111 (New resdenal 1 un srucures nonfarm) 1112 (New resdenal 2-4 un srucures nonfarm) 1115 (New resdenal addons and aleraons nonfarm) 1118 (New resdenal garden and hgh-rse aparmens consrucon) and 1211 (Manenance and repar of farm and nonfarm resdenal srucures). In 1992 he sum of sales arbued o hese ndusres equaled $ mllon almos he same as he $225.5 bllon recorded for resdenal nvesmen. If hese represen sales o consumers and no sales o oher nermedae ndusres our procedures are somewha valdaed. 27

28 Box 2: IO-Use Table Decomposon of Fnal Demand no Fnal Sales From Indusres PCE (1) Prvae Invesmen (2) Governmen Expendures a (3) Resdenal Invesmen (4) Consrucon Manufacurng Servc es Oher Toal a. Includes governmen consumpon and governmen nvesmen expendures. $2859. bllon of fnal sales from servces ndusres and -$9.8 bllon dollars of sales from oher (governmen) 38 addng o $3718 bllon dollars. 39 Usng equaon (A.18) wh he fnal sales by spendng caegory gven n box 2 we map fnal sales by ndusry no value added by ndusry for he four dfferen ndusres for PCE oal prvae fxed nvesmen resdenal nvesmen busness fxed nvesmen and he sum of PCE BFI and governmen non-defense nvesmen (GOVI); see box 3. For example o calculae he value added by nermedae ndusry for PCE we se S B = S M = S S = and S O = (aken from column 1 of box 2) and apply he formula gven n (A.18). To compue he consrucon manufacurng and servces share of PCE we smply dvde he value added of hose ndusres (gven n he rows of column 1) by oal We unforunaely can no provde an nerpreaon for negave sales nomnal PCE equals $421 bllon dollars; he consumpon of housng servces and sales arbued o non-comparable mpors accoun for he dfference. 28

29 Box 3: Decomposon of Fnal Demand no Value Added From Indusres PCE (1) Prvae Invesmen (2) Resdenal Invesmen (3) BFI (4) PCE+BFI +GOVI (5) Consrucon Manufacurng Servces Oher Toal PCE mnus value added from oher. Ths yelds shares of and percen of PCE for he consrucon manufacurng and servces ndusres. For resdenal nvesmen (column 3) we calculae consrucon manufacurng and servces shares of and percen respecvely. For Busness Fxed Invesmen (column 4 = column 2 mnus column 3) we calculae shares of and percen. In he model he consumpon-nvesmen good s he sum of PCE BFI and governmen nvesmen. We assume ha he composon of governmen nvesmen by nermedae ndusry value added s he same as busness fxed nvesmen. In 1992 nomnal governmen nvesmen was equal o 25.8 percen of busness fxed nvesmen. Mulplyng he rows of he BFI column by and hen addng he rows of he PCE column yelds he value added by ndusry of he model s consumpon-nvesmen good he PCE+BFI+GOVI column. The value-added shares of hs good are 3.7 percen for consrucon percen for manufacurng and percen for servces. Therefore for calbraon of he model we se 29

30 B c M c S c B h M h S h

31 Bblography Cooley Thomas and Edward Presco (1995). Economc Growh and Busness Cycles n Thomas Cooley (ed.) Froners of Busness Cycle Research Prnceon: Prnceon Unversy Press. Corrado Carol and Lawrence Slfman (1999). Decomposon of Producvy and Un Coss. Amercan Economc Revew 89(2) Gullckson Wllam and Mchael J. Harper (1999). Possble Measuremen Bas n Aggregae Producvy Growh. Monhly Labor Revew February Jorgenson Dale and Kevn Sroh (2). Rasng he Speed Lm: U.S. Economc Growh n he Informaon Age. Brookngs Papers on Economc Acvy forhcomng. Lawson Ann M (1997). Benchmark Inpu-Oupu Accouns for he U.S. Economy 1992: Make Use and Supplemenary Tables Survey of Curren Busness November Lum Sherlene Moyer Bran C. and Rober E. Yuskavage (2). Improved Esmaes of Gross Produc by Indusry for Survey of Curren Busness June Lum Sherlene and Rober E. Yuskavage (1997). Gross Produc by Indusry Survey of Curren Busness November Parker Rober (1997). Noe on Alernave Measures of Gross Produc by Indusry. Survey of Curren Busness November

32 Peper Paul E (199). The Measuremen of Consrucon Prces: Rerospec and Prospec n Erns R. Bernd and Jack E. Trple (ed.) Ffy Years of Economc Measuremen: The Jublee of he Conference on Research n Income and Wealh Chcago: Unversy of Chcago Press. U.S. Deparmen of Commerce Bureau of he Census (1997). New One Famly Houses Sold. Curren Consrucon Repors March A1-A8. U.S. Deparmen of Commerce Bureau of he Census (2). New One Famly Houses Sold. Curren Consrucon Repors Aprl 3-9. U.S. Deparmen of Commerce Bureau of Economc Analyss (199). Personal Consumpon Expendures. Mehodology Paper Seres MP-6. Washngon DC: U.S. Governmen Prnng Offce. U.S. Deparmen of Commerce Bureau of Economc Analyss (1998). Benchmark Inpu-Oupu Accouns of he Uned Saes Washngon DC: U.S. Governmen Prnng Offce. U.S. Deparmen of Commerce Bureau of Economc Analyss (1999). Fxed Reproducble Tangble Wealh n he Uned Saes Washngon DC: U.S. Governmen Prnng Offce. U.S. Deparmen of Commerce Bureau of Economc Analyss (1999). Naonal Income and Produc Accouns Tables. Survey of Curren Busness December U.S. Deparmen of Commerce Bureau of Economc Analyss (2). Seleced NIPA Tables. Survey of Curren Busness June D2-D29. 32

33 TABLE 1 BUSINESS-CYCLE VOLATILITY OF GDP AND KEY COMPONENTS 1955:Q1 1997:Q4 Cross-Correlaon of GDP wh: Varable SD% x(-5) x(-4) x(-3) x(-2) x(-1) x x(+1) x(+2) x(+3) x(+4) x(+5) GDP Consumpon expendures CONS CNDS CD Toal fxed nvesmen INVT Toal nonresdenal nvesmen INVTXH GOVI BFI Resdenal nvesmen RES Governmen consumpon and defense nvesmen GOVC Noes: All varables are real chan-weghed 1996$ and have all been logged and HP flered from 1947:Q1 o 2:Q1 wh λ = 16. GDP sands for GDP; CONS personal consumpon expendure; CNDS consumpon of nondurables and servces; CD consumpon of durables; INVT gross fxed nvesmen; INVXH gross fxed nvesmen excludng resdenal nvesmen; GOVI governmen non-defense nvesmen; BFI gross busness fxed nvesmen; RES gross resdenal fxed nvesmen; GOVC governmen consumpon and defense nvesmen purchases. 33

34 TABLE 2 BUSINESS-CYCLE VOLATILITY OF RELATIVE NEW HOUSE PRICES (NIPA) AND OTHER SERIES 1955:Q1 1997:Q4 Cross-Correlaon of Relave New House Prces (NIPA) wh: Varable SD% x(-5) x(-4) x(-3) x(-2) x(-1) x x(+1) x(+2) x(+3) x(+4) x(+5) New House Prces (NIPA) GDP RES Noes: All varables have been logged and HP flered from 1947:Q1 o 2:Q1 wh λ = 16. The New House Prces (NIPA) seres equals he chan-ype prce-ndex for NIPA resdenal nvesmen dvded by he chan-ype prce ndex for NIPA personal consumpon expendures. GDP sands for real $1996 chan-weghed GDP and RES sands for real chan-weghed $1996 gross resdenal fxed nvesmen. TABLE 3 BUSINESS-CYCLE VOLATILITY OF RELATVE NEW HOUSE PRICES (CENSUS) AND OTHER SERIES 1982:Q1 1997:Q4 Cross-Correlaon of Relave New House Prces (Census) wh: Varable SD% x(-5) x(-4) x(-3) x(-2) x(-1) x x(+1) x(+2) x(+3) x(+4) x(+5) New House Prces (Census) GDP RES New House Prces (NIPA) Noes: All varables excep for New House Prces (Census) have been logged and HP flered from 1947:Q1 o 2:Q1 wh λ = 16. The New House Prces (Census) seres equals he Chan-Type Annual-Weghed Prce Index (Fsher Ideal) of New One-Famly Houses Sold Includng Value of Lo dvded by he chan-ype prce ndex for NIPA personal consumpon expendures; he resulng rao s logged and HP flered from 1979:Q1 o 2:Q1. For oher noes see above. 34

35 TABLE 4 CONTEMPORANEOUS CORRELATIONS COMPUTED WITH QUARTERLY DATA 1982:1 1997: GDP CONS INVTXH RES GOVC P H 1 GDP CONS INVTXH RES GOVC P H 1 Noes: All varables have all been logged and HP flered wh λ = 16. All spendng varables are n real chan-weghed 1996$. GDP sands for GDP; CONS personal consumpon expendure; INVTXH gross fxed nvesmen ncludng governmen non-defense nvesmen and excludng resdenal nvesmen; RES gross resdenal fxed nvesmen; GOVC governmen consumpon and defense nvesmen purchases; P H s he Census new house chan-ype prce ndex dvded by he chan-ype prce ndex for NIPA personal consumpon expendures. 35

36 TABLE 5 BUSINESS-CYCLE VOLATILITIES COMPUTED WITH ANNUAL DATA Cross-Correlaon of Oupu wh: Varable SD% x(-1) x x(+1) GDP Fnal Goods CONS INVTXH RES GOVC Inermedae Goods Oupu CONSTR MANUF SERVICES Hours ALL CONSTR MANUF SERVICES Prvae Fxed Capal ALL CONSTR MANUF SERVICES Noes: All varables have been logged and HP flered usng all avalable daa wh λ = 1. GDP CONS INVTXH RES and GOVC are defned he same as n able 1. CONSTR MANUF and SERVICES sand for he consrucon manufacurng and servces secor of he model; manufacurng ncorporaes agrculure foresry and fshng mnng and manufacurng; servces ncorporaes ransporaon and publc ules wholesale rade real rade and servces. ALL prvae hours equals he sum of he hours worked n all ndusres excep governmen; ALL prvae fxed capal equals he sum of he socks of nonresdenal prvae fxed capal and governmen non-defense capal. 36

37 TABLE 6 CONTEMPORANEOUS CORRELATIONS COMPUTED WITH ANNUAL DATA GDP CONS INVTXH RES GOVC P H X b X m X s N all N b N m N s K all K b K m K s 1 GDP CONS INVTXH RES GOVC P H X b X m X s N all N b N m N s K all K b K m K s 1 Noes: All varables have all been logged and HP flered wh λ = 1. All oupu and capal varables are n real chan-weghed 1996$. All correlaons of (xp H ) are esmaed from 1979 o The manufacurng secor ncludes he AFF mnng and manufacurng ndusres. The servces secor ncludes he ransporaon and publc ules wholesale rade real rade and servces ndusres. GDP sands for GDP; CONS personal consumpon expendure; INVTXH gross fxed nvesmen ncludng governmen non-defense nvesmen and excludng resdenal nvesmen; RES gross resdenal fxed nvesmen; GOVC governmen consumpon and defense nvesmen purchases; P H s he Census new house chan-ype prce ndex dvded by he chan-ype prce ndex for NIPA personal consumpon expendures; X b s oupu of he consrucon secor; X m s oupu of he manufacurng secor; X s s oupu of he servces secor; N all are he hours worked n all prvae ndusres; N b s he hours worked n he consrucon secor; N m s he hours worked n he manufacurng secor; N s s he hours worked n he servces secor; K all s he oal producve capal sock n he economy (busness fxed capal and governmen non-defense capal); K b s he capal sock of he consrucon secor; K m s he capal sock of he manufacurng secor; K s s he capal sock of he servces secor. 37

38 FIGURE 1 PRIVATE CONSUMPTION GOVERNMENT CONSUMTION AND TOTAL FIXED INVESTMENT SHARE OF NOMINAL GDP 1955:Q1 2:Q1 7 Prvae Consumpon Governmen Consumpon Toal Fxed Invesmen

39 FIGURE 2 DURABLE GOODS AND HOUSING SERVICES SHARE OF NOMINAL NIPA CONSUMPTION 1955:Q1 2:Q1 16 Share of Nomnal NIPA Consumpon Consumer Durables Expendures Consumpon of Housng Servces 39

40 FIGURE 3 PRICE INDEX AND BUSINESS CYCLE DYNAMICS OF VARIOUS MEASURES OF CONSUMPTION 1955:Q1 2:Q1 12 Consumpon Prce Indexes = NIPA xcl Durables Durables.4 Derended (Log HP-Flered) Consumpon.3 NIPA xcl Durables NIPA xcl Durables 4

41 FIGURE 4 NOMINAL SHARES OF GROSS PRIVATE FIXED INVESTMENT 1955:Q1 TO 2:Q1 4 Resdenal Invesmen Governmen Invesmen Nonresdenal Srucures

42 FIGURE 5 PRICE INDEXES FOR VARIOUS FIXED INVESTMENT SERIES 1955:Q1 2:Q = Toal Nonresdenal Resdenal = Governmen Nondefense Prvae Equpmen and Sofware Prvae Nonres. Srucures 42