sign in sign up
<<
Home , Capacity utilization, William Nordhaus

LAWRENCE R. KLEIN University of Pennsylvania assisted by VIRGINIA LONG WhartonEconometric Forecasting Associates

Capacity Utilization: Concept, Measuremnent, Iand Recent Estimates

ON SEVERAL OCCASIONS in the past fifteen years, my colleagues and I have tried to explain disparities among alternative measures of capacity utiliza- tion and to justify our own approach to the measurementproblem.1 George Perry, in his contribution to this issue of Brookings Papers, has indicated many of the important issues, and I would like to amplify his points or restate them from another viewpoint in the of clarification. Briefly, the Federal Reserve estimates that as of mid-October 1973, a fair amount of spare capacity existed in the American economy. The esti- mated operating rate was only 83.4 percent for 1973:3. By contrast, the Wharton index for was as high as 96.7 in the same period, and was rising faster than the Federal Reserve index. The McGraw-Hill index was at an intermediate level of 86.5 percent in September. Similar divergences among the indexes had been apparent for several months. These messages are so different that they suggest the need for a close look into the whole subject. For some years the Wharton and Federal Reserve indexes of utilization followed similar paths. They began to diverge in 1965, when the Wharton index rose sharply, as if signaling the onset of the

1. LawrenceR. Klein and Robert Summers,The WhartonIndex of CapacityUtiliza- tion(University of Pennsylvania,Wharton School of Financeand Commerce, ResearchUnit, 1966). 743 744 Brookings Papers on Economic Activity, 3:1973 stronginflationary pressures that accompaniedthe VietnamWar.2 Perry's Table3 bringsout sharplythis breakin the relationship.For the two sub- periods1954-65 and 1966-72the seriesall havelarge pair-wise correlations, but the Whartonindex correlatespoorly with the others for the whole period 1954-72. After 1965, the short-runmovements in the Wharton seriesare like those in the others,but its level is much higherin termsof a utilizationrate. I feel that the Whartonindex gave the correctsignals on inflationin 1965 and again at the beginningof 1973.

The Concept

Full capacityhas been variouslydefined as a minimumpoint on a cost function,a full input point on an aggregateproduction function, and a bottleneckpoint in a generalequilibrium system. Full capacityshould be definedas an attainablelevel of outputthat can be reachedunder normal inputconditions-without lengthening accepted working weeks, and allow- ing for usual vacationsand for normalmaintenance. Preoccupation with measuresfor individualindustries, considered separately from others at the same time, tends to overstatecapacity for the system as a whole. The standardWharton measure of trendlines throughpeaks of individualpro- ductionseries provides a systemof attainablepoints because many or most industriespeak approximatelytogether. The Whartonapproach satisfies the conditionof feasibilityunder normal conditions; it runsthe risk, how- ever,of callinga local maximumpoint one of full capacitywhen it may be only a partialrecovery point. This is the problemof the so-called"weak peak." I believethat 1959was the last time that the economypeaked out at a local maximumwith less than full recovery. Some adjustmentfor the Whartonindex around 1959 became necessary. Preston and I estimatedin- dustryproduction functions and insertedthe full employmentlabor force andcapital stockfor each industry. The valueof productioncomputed from theseinputs at full use yieldedestimates of capacityoutput that could then be used to adjustweak peaksupward.3

2. See WhartonQuarterly, Vol. 6 (Winter 1971), p. 16. 3. L. R. Klein and R. S. Preston, "Some New Results in the Measurementof Ca- pacityUtilization," American Economic Review, Vol. 57 (March 1967), pp. 34-58. Lawrence R. Klein and Virginia Long 745 By distributingthe available(full employment)labor force over indus- tries as it would be, historically,at full employmentpoints, we implicitly take accountof the feasibility,in a generalequilibrium sense, of attaining the outputlevel that we designateas full capacity.This approach is different from one relyingon a purelymacroeconomic production function or from one that estimatesoutput at full capacityin each sectorwithout taking ac- count of relationshipsto other sectors.Subsequent to 1959, the economy consistentlypeaked out at full capacitypoints. The overshootingof inter- mediate peaks like those in 1959 when the trend lines are established throughpeaks has providedus fairly reliabletrends on capacitythat we now use in the Whartonindex. As the economyapproaches a turn,however, before the maximumpoint has been fullyreached-as at the end of 1973-the capacityvalues in a few cases keep exceedingthose in the previousquarter so that the capacity trendsmust be movedslightly upward. This is a fault of our method,but I thinkthat the resultingupward bias in capacityutilization is less than2 full percentagepoints in presentcircumstances. Accordingto presentforecasts, it is likely that the economy will slow down in coming months; thereforesubsequent quarters should not push economicperformance in separateindustries on to new, higher,capacity points that would lead us to reviseupward our capacitytrend lines. The figuresin Perry'sTable 9 areinteresting to me in thatthey indicatea muchsmaller upward bias for the Whartonindex over the FederalReserve or McGraw-Hillmeasures than would be suggestedby the comparisonof the usual aggregateindexes. The FederalReserve estimates in Table 9 for major materialsindustries are more firmlybased on direct information fromengineering and tradesources. They tend to be no morethan about5 points below the correspondingWharton indexes. Generallyspeaking, I like productionfunction estimates of capacityout- put, but for quicknessof estimationand simplicityof concept,I preferto workon a largescale with our presentmethod of trendlines through peaks, usingproduction function estimates only for adjustmentand extrapolation.4

4. Interestingmeasures of productionfunctions are containedin a researchpaper by Robert M. Coen and Bert G. Hickman, "AggregateUtilization Measuresof Economic Performance,"Research Memorandum 140 (StanfordUniversity, Center for Researchin , February1973; processed). 746 Brookings Papers on Economic Activity, 3:1973 Yet anotherapproach draws on linear programmingcalculations. In studies of chemicalproduction and petroleumrefining, Malenbaum and Griffinmeasure capacity as the "bottleneck"point in expansionalong a given ray correspondingto a fixed productmix.5 When one producthits such a bottleneck,all othersdependent on it for intermediateinput are re- strictedat less thanfull capacityutilization. This provides a maximumout- put point while preservinga given productmix. The Malenbaumand Griffinestimates for petroleumand chemicalswere checkedagainst the Whartonestimates of utilizationfor the same indus- tries. In general,little agreementwas found betweenthe resultsfrom the engineering-typeestimates and the standardWharton procedures, espe- cially for short-runmovements. But in most cases, the Whartonestimates of utilizationfor petroleumrefining are higherby about 4-5 percentage pointsthan those based on the linearprogramming model. For the chemical industry,the resultsare closer,but in a few isolatedyears the discrepancy exceeds5 percentagepoints. Systematicanalysis and coverage of all major sectors by production functionestimation, input- methods, and linearprogramming calcu- lations promisethe greatestreturn in precisionof measurement,but that combinedapproach seems to be a long way off, exceptfor intensive research in separateindustries. In this connection,I disagreewith Perry'sview that limitationson pro- ductionresulting from of intermediateproducts (like today'san- ticipatedshortages of fuel)should not affectthe conceptof capacity.Capac- ity is a generalequilibrium concept, which should be alteredin the light of bottleneckswhose effectscan be tracedthrough an input-outputanalysis. That is the whole point in usingcapacity utilization measures as signalsof inflationarypressure, and accounts for my view that other measures stronglyoverstate the amount of spare capacityavailable by not taking accountof interrelationshipsamong industries.6

5. Helen Malenbaum,"Capacity Balance in the Chemical Industry,"in Lawrence R. Klein (ed.), Essays in IndustrialEconometrics, Vol. 2 (University of Pennsylvania, Economics Research Unit, 1969); James Griffin, CapacityMeasurement in Petroleum Refining:A Process Analysis Approachto the Joint Product Case (Heath, 1971). 6. A techniquefor estimatingcapacity utilizationrates in an input-outputmodel is given in L. R. Klein, "Some Theoretical Issues in the Measurementof Capacity," Econometrica,Vol. 28 (April 1960), especiallypp. 280-86. Lawrence R. Klein and Virginia Long 747

CapacityUtilization as an ExplanatoryVariable

Indirectuse of capacitymeasures is importantin the constructionof econometricmodels and servesas a validationtest for the seriesactually beingconsidered. The indirectuses arein equationsfor (a) priceformation; (b) capitalformation; (c) .Capacity utilization is one of the most stra- tegicvariables in the Whartonmodel, and showsup in severalplaces. In the basic equationfor formation-the manufacturingdeflator-a non- lineartransformation of capacityis one of the most significantvariables:

loglg1- I CPMF' where CPMF is capacityutilization in manufacturingconstrained to lie between0.87 and 0.99. In this nonlineartransformation, stronger upward pressureon pricesdevelops as CPMF comes closer to its limitingvalue, 0.99. In equationsfor capitalformation, a one-quarterlag effectis introduced by the linear term (CPMM)-1, where CPMM is capacity utilizationin manufacturingand mining. This is a strong near-termeffect. The other variableshave effectsspread out over more quarters.Perhaps a nonlinear transformationwould be appropriatehere too, but it has not been tried. In some studiesof exportequations, capacity limitations on exportper- formancehave provensignificant. In the trademodel of the Organisation for Economic Co-operationand Development,exports for the United Statesare made to dependon "pressureof ,"which is quiteclose to the Whartonindex measure.7 In new estimatesof U.S. exportfunctions for use in the Whartonmodel, I havefound that Europeancapacity utilization, measuredanalogously with the Whartonmethods, is a highly significant variable. These are the main avenuesby which the capacityutilization variables enterthe entiremodel. They are strongly significant, from a statisticalview- point, in all these cases. But a properassessment of the marginalcoeffi-

7. F. G. Adams, H. Eguchi, and F. Meyer-zu-Schlochtern,An EconometricAnalysis of InternationalTrade (Paris: Organisationfor Economic Co-operationand Develop- ment, 1969). 748 Brookings Papers on Economic Activity, 3:1973 cientsof capacityvariables depends on good specificationof the restof each equationand on use of nonlineartransformations where appropriate. Best econometricpractice should be followedin specifyinginvestment functionswith lag distributionsof outputand capitalstock; capitalrental variablesshould be appropriatelyintroduced. Unless these steps are taken, the marginaleffects of capacityutilization on investmentmay not be cor- rectlyassessed. In the sameway, the priceequations with capacityutiliza- tion variablesshould have unit labor costs, insteadof rateswithout appropriateproductivity corrections as in Perry'sestimates. His estimates and assessmentsof the role of capacityin modelsleave me doubtfulthat he has capturedthe rightmarginal effect, because some of his othervariables are not optimallyspecified and becausehe does not use a nonlineartrans- formationof capacitywhere needed, in the priceequation. This is particu- larlyimportant when rates are high. Using capacitymeasures as strategicregressors depends on the statistical compilationof historicalseries-measured as trendsthrough peaks of pro- ductioncurves in the Whartoncase. But forecastingapplication requires a procedurefor extrapolatingcapacity utilization in the sameway that other endogenousvariables are extrapolated.This is accomplishedthrough the use of productionfunctions. These equations are firstestimated from his- toricalsample data with observedinputs of actualemployment in the form of a short-rundynamic adaptation to a long-runproduction function. The short-runmodel is A log L = X(log L* - log L-1), where

L* =- -logA - t- -log K+ - log X + e, and

L = employment K = stock X = output e = error. The parameterratio A/a is estimatedfrom long-runstatistics on average factorshares. Capacity output is estimatedfrom this pair of equationsun- dertwo conditions:A log L = 0; and L = LF, the laborforce. A separateestimate of a full capacityproduction function is not made; Lawrence R. Klein and Virginia Long 749 full capacityoutput is definedas a point on the long-runproduction func- tion, determinedby using full employmentinputs.

ContemporaryFindings

The precedingdiscussion describes the analyticaleconometric applica- tion of capacitymeasures. What signalshave the actual data been giving recently?Since the last low point,in the fourthquarter of 1970,the Federal Reserveindex has gained9.2 points,rising from 74.2 to 83.4. The Wharton index for manufacturinghas gained 13 points in the same period,moving from 83.7 to 96.7 (see Table 1). More significant,96.7 is very near the his- torical high for the index. After the index surpassed95, in the first quarterof 1973,the stronginflationary pressures that had been suspected earlier, when the index exceeded 90 in 1972:3, became unmistakable. Aggregatecapacity utilizationrates for the United Kingdom and five WesternEuropean countries are shown in Table 2. It has been evidentfor some time that one strategicindustry after an- otherwas passingcritical values for utilizationrates each quarter,and that the slack areaswere specializedsectors where definite cutbacks had been ordered-aerospaceis a typicalexample. Table 3 showsestimates for indi- vidualindustry components of the Whartonindex. Throughout1973, the sectors of high capacity utilizationare primaryand fabricatedmetals; nonelectricaland electricalmachinery; motor vehicles and parts; instru- ments;clay, glass, and stone products;lumber and furniture;miscellane- ous manufactures;textiles; apparel; paper; printing and publishing;chem- icals; petroleum;rubber and plastics; food; crude oil and natural gas extraction;electric power; and gas .With a line-upof high (Whar- ton) rates for a group of industrieslike these, severecapacity limitations quite evidentlywere building up. Sincebusiness cycle peaks (both specificand reference)have been strong and well-definedsince 1959,we have felt quiteconfident that our measures usingtrend lines through peaks provide adequate estimates of capacity,but until a new turnoccurs, we mustalways be in doubtat the end of a cyclical phaseabout the exact position of the trendlines. Thereis only one refer- encepoint and not two, as is the case for moredistant peaks. Yet, in the last half of 1973,we foundlittle occasionto reviseour series.The overallindex in the firstquarter for manufacturing,mining, and utilitieswas estimated -- w- N r- N 0 tl o zt zt tnt ) CN m n oo ,C r

ON CIACN \ CN O\000O0 0000000 00) C)C CNC NO

cl: U6C t- 6 6C,r,

' S. OCOC0000 00 000 0 00 0 0O OC 0 0

1.it t .8 t~~o o m r r- n r- o tl oo co v x Cd O cnO ~\000 000000r- 00 000 00 ~ceq\ ON 0 N m ON ON ON 0 N C, Cl ON ON ON CI Cm Ct s o~~~~~,*

W~~~~~.4eq t i N S 0> F9c en (ON en 00 0 F n cO ov

X ...... Q [ ) _~~~~b c~~~~. \,w 'w -w - ) O" 4 m , ,w \c t 'w 00 w'

'w Po Y~~~~~~eeq 'w Fn r ) o ol It o N CI r- C5 o eq vO . ;~~~~~~~, 08 'w 'w~ F1 t tr e1 cs o o~ tr ^6 1n'w'6c

~~~~~~~0 .D~~~~~OP~~~~~~~~~~~~~~~~ ON aNw (O 00 O mt 00 00 0 00 W0 00 0 Ur)O CN CONC

Cd 0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.~~~~~~~~~~~C) C) N It IOtwwOn N ON 00 r- Fn

n C(3D 3~~~~~~~1l, 6ciC ,3 enc\~.t)~j~C -~4- C1e )'f'/ 0*-w0s *~~~~~~C0~~~~~~~~~~~~~~~~~~NO CIN ON ON ONO N O 01 NO ON 4-4 S 0

t UWb t-qen . cl 00000C 00000\ o ~ ~ ~ 000000~~~N 0 16 i

r- X X Q o o r o o N Ii 00 CI r- 00 0

P= 0~~~~~~~~~~~~~0

O000000 005 00o 00 0000 00 00 00

00 O000 0lc CIN C 00

ri~~~~~~~~~~~~~~~r

cB o XXXXX .0 0*.

ON00 0000 O00~

I b ooo oc- c-.o

Q .u 01 ' O'Oe cs 00 c N O cd X CC :s cfi b C5 csfi nn ? t o o O mO O .: ef to o m nN

0 0 0~~~~~~~~~~~~~~~ 752 Brookings Papers on Economic Activity, 3:1973 Table 3. U.S. Capacity Utilization Rates and Weights, by Industry,First Quarter, 1973, Estimated in April and October 1973 Utilizationrate, first quarter1973 (percent) April October 1973 1973 Industry estimate estimate Weight Primarymetals 100.0 97.8 0.0188 Fabricatedmetal products 93.6 96.3 0.0194 Nonelectricalmachinery 89.3 93.6 0.0316 Electricalmachinery 95.1 97.6 0.0281 Motor vehicles and parts 100.0 99.9 0.0220 Aircraftand miscellaneoustransportation equipment 65.4 65.6 0.0162 Instruments 91.0 93.6 0.0084 Clay, stone, and glass products 97.9 98.0 0.0087 Lumberand products 99.5 100.0 0.0067 Furnitureand fixtures 99.1 95.6 0.0050 Miscellaneousmanufactures 100.0 100.0 0.0045 Textile mill products 94.5 99.0 0.0086 Apparel products 92.2 95.2 0.0107 Leatherand products 77.9 76.9 0.0026 Paper and products 100.0 100.0 0.0102 Printingand publishing 90.9 90.7 0.0154 Chemicalsand products 100.0 98.4 0.0213 Petroleumproducts 100.0 98.0 0.0067 Rubber and plastic products 100.0 98.5 0.0081 Food 99.2 100.0 0.0214 Tobacco and products 84.6 87.3 0.0018 Coal 84.5 85.9 0.0010 Oil and naturalgas extraction 91.8 92.0 0.0031 Metal mining 94.2 88.3 0.0009 Stone and earth minerals 77.8 85.0 0.0015 Electricutilities 100.0 99.0 0.0144 Gas utilities 96.1 92.8 0.0043 Source: Same as Table 1. at 94.2 in April 1973.In October,this same value, with hindsightand re- vised data, was estimatedto be 95.0.8The changeswere usually not more than one or two points,and rarelyas largeas five or six. They are detailed by industrygroups in Table4. As in 1965,the overallindicator of average utilizationhas servedas a fairlygood earlywarning signal.

8. This figureis more comprehensivethan the manufacturingseries, cited earlier.If constructionand serviceindustries are included,the averagerate is estimatedat 94.4 for 1973:1. -St C, 0 00 00 0

0 en O o CN N It (O 00 O' Nn 0 C, F o 00 en c, 0 r r O Noom ooooo oo oww omo N 0oC7

.k .cq SX a(ON n o W) ? w tn (N M en o x E K~~~~~e 00o moN en? t?W tn t

a ~~~~~~~~~~'Oo1 \' NC N0 oC -r 110 o'IOI' ~o\'O\

, C^ o N onX CN r- (N 0 afi WI, O oNvO oo

w w r- www w w w (O c\ 0 X~~~~~~~~~~OE ow (N o n (ON

o O~~~~~~~~~~~~~~CO> oCN C;

t 3 s ;: to mct cs~~~~~~n \ eWn ooen\- Mm ay N \,t- o o E~~~~~~~~~~~~OO oo o eon Nene in oo I a ~ ~~~~~~~~~~~~~~~~~~~~~~IU ,o Iom mmm wmm ,oC pw Cs (ON Cr CN00 00 0 00 00 ON(O(N

2;~~~~~~~~~~0asi o ON o 0C% C o t qQ t o o o o~~~~~~~~~~~~~~Co .N~~~~~~~~~~~~0\oO xX'm en t eon enN en en \c

Q~~~~~~~~~~Oxs c. cN (o,o oooo o o c C(OcNo

1d *;- |C =; C

( SU > XNIt 1 N m N (N - oo \0 O -f 5g> o 00

C; C;

m W) C m w t-: 6 C'i(: Q c; '-d:(,i (0 ,6 0 W C c;a-, oo oo oo oow w w w t- tQ wwwt- t- t- t-

en (ON t- -z O

c; C; C;

N M C tn "o C 't w m tn 110ON N tn 110 -tu 4 e e e (: 0 r-: tr;%6 %6 rz C _; en W)en -t c c) aN aN oo oo oo oo oo oo oo oo oo oo (F aN (oN c c

00 00 t4 C> C) C;

tn t- O tr) en t-- 00 CrN N M "t 't CINw It O t3 (: (: C;0 C C %6 t-: 0 W), 0 O' CN' CN' C-, CN 0 (3 CN C w w w w CN C% w C (31, CN ON CN

16W

00 %C tn

10 42 t3 CN 0", 0 oo r- t-, W; (0 : z C (ON0 (01 OIN(01 C C CN CN C C (3 CN CN ON 0 (01 0

PC 0 0tr) C)In atri 0tr) tri pat3 tu 0 0

WI en r- N o tn w --4 ON N \0 N 0 PC In W) N W) 6 -,; 6 W; 6 6 C44 W;0 CN C% ON ON 00 00 00 00 00 00 00 00 00 (ONON ON ON CN

00

C; C;

0 en ON w In t, N -4 t- tl- t- in C> 0 0 00 %O tn tn a, C C F a all ON ON ON CN CN ON CN (DIN0% CN (ON

00 cN (ON oo 00 00 O

C V3 0 . W). t- . -t . N . \0 . "o. . C,,. a . tn . N . ON. \0 . t- . w . 0 . 0 . tn . o oo tr) I- -4 C t- t-- r- -4 oo oo O -4 en In oo 0 CA (34 C (0-, w w w w (F w w (ONF C (3-% c a c

N en 't N en It C,4 en t C14 en It C4 en 0 00 > 3oo oo o Oo o

Uo . Ce 00 r- ootc e ooN

Ux t o 0 0 o o

o. C0. I o o oo en oo 0n o I o 0 o CbV6 rF "6C? oo08 oo: C? oo:0 oo0oo 08o- CO o Ch Ch

S S o0 0 0 ~~O~~~0 ~~00 0 000

. oo b ooooOoo o ?o No o "D t N oo tn e 4 S s bcz~~tn O mNCh tm Ch oo \W Ch W; 4 ?? Ch CO 00 0o oo N~~~~~~~~~~0oo0 00m Cmm mONON Co o 0 0

eQ0000 0000XO 00 ID'00 oo- e r 0O O CWCS00000 a0a 00000000C tr ) O \ O 00 00 0 CO

00 0N 00 t 0 0

4

<~~~~~~~~~OON S~C Ch o~ F Ct ^ CN o~CN CN ON ooCo sO 0 0 0 0

0 o 0 0 0

Q4 Q4 ON 00 0-W 0 n" 0DC ~ o O O O OC o 0 g cjIo 0I0tD 0 0 o 0 0

0 ON O0N 0N c ^ ? c0 O

A bE _ N ^ t _ON _NO O_N _NO _ _ t . 0) ~~~~~~~~~~.~C

0 ON5 0N mO ONCt 0 m cz

s~~~~~~~~0,.It -4en t e $O $ s: o' 'M..~~~~~~~ . I ?l o?6 ?o'C; r:0~C zC t t N \-O0O 0 Q Oma -.-'J. 0WOeC~ C'0cm0

C 0 0 \ O ? o o 8 aSo nN0 o 0 S 0

a:a~ooo~ \0ooo- 00'Oen oInor-N 0o

000 , 00oo 00o 00 o t-\0\CC\0 o-0O C X o 0 X~~~5, o o 0? 0? 0?

* 0 o~ooo~ O~0ooo oFo~o~ Oo~O~O oo0o

1 M ~~c-Or-a ~ICteC,c- Mtn%OOr--"V v~00 M

400 odC 00b _-6^* 0en_ 6et _ Ne* _ >^ _ t: c-iCO C C0C me C

0o 00 0 0~~~~~~~~~~~C0~W NC 0 0 0 0 ~~~~~~ 0 0 0~~~~~~~~~~~~~~~~~~~~~~~~~c 5 5 0 0~~~~~~~~~~~~~~~~~~~~~~~~~~~1

Z - r 0lr a OC Om %o m ON~--oe ON~CeC e- Commentsand Discussion

Alan Greenspan:I will be brieferthan usual, since the Klein-Longpaper has anticipatedmany of my comments.The total operatingrate sought hereis not an arithmeticweighting of operatingrates in individualindus- tries,but ratherthe rate obtainedfrom some type of input-outputor linear programmingsystem. Since interrelationshipsexist amongmaterials flows fromindustry to industry,the weightingsystem needed to obtainan overall operatingrate is one that rests on these flows. The measureof the econ- omy'sposition in termsof an overalloperating rate dependson the degree of shortagein partsof the system.Today shortagesexist and they are re- strictingproduction. Particular industries, such as coal mining,may exhibit excesscapacity, but this "slack"has only a limitedimpact on the adjust- ment of the overalleconomy to inflationarypressures. Unfortunately,it is not easy to use an input-outputsystem to get the kind of answerthat is required.The economyis faced with specificengi- neeringconstraints-that, for example,so muchelectric power is neededin the electrolyticprocess to produceso much aluminumingot; but thereis also a considerablerange of substitutabilityamong materialswhich is basicallya functionof price.Therefore, the analysisrequires not only some judgmentsabout engineeringinterrelationships, but also the price vector that determineswhich materials or processeswill be used. Withoutsome- thing of that sort, the conceptof capacityis meaningless. The real test of the concept is about to be thrustupon us, becausethe economyfaces a classicalcase of deprivationof a basicmaterial input-oil. Whatwill that do to the level of industrialactivity? Some preliminary work has had resultsthat I don't fully know how to interpret.The analysissug- geststhat as soon as the price of energybegan to rise in 1972,energy per unit of outputdropped rapidly, clearly implying some priceelasticity in the relationshipof energyinput to outputof . The infrastructureof pro- ductionin this countryis obviouslybased upon the low levels of energy 757 758 Brookings Papers on Economic Activity, 3:1973 pricesthat have persistedfor many decades.What is uncertain,unfortu- nately,is whatthe structureof productionwould be withenergy two to threetimes what they have been. The input-outputcoefficients system would be markedlydifferent. What is visibleto date is dramaticevidence of the short-termelasticity of energyinput with respect to price.It wouldbe nice to believethat all the nationhas to do is sit tight, allow pricesto rise, andlet the wholeadjustment process take place with zero effect on capacity. But thereis no way to be sure, becauseof uncertaintyabout engineering constraintsthat complicatethe economicconstraints. In principle,"capacity" has meaning.In the currentperiod of lengthen- ing lags in deliveries,and of all sorts of reportedshortages and difficulties in obtaininggoods (regardless of price),the economymust be producingat capacityin any meaningfulsense. If the numbersdo not indicatethat the economy is currentlyat capacity, then I suggest that the numbersare wrong,and that the correctconcept of capacitymust reflectthe situation. Next, I wouldlike to raisea coupleof issuesabout the Whartonmethod of measuringcapacity. As I understandit, outputof two-digitSIC manu- facturingindustries is used in constructingthe estimates.One problemis that the two-digitdata imply a lowerlevel of capacitythan would four-digit data since all four-digitindustries within a two-digitgrouping would not reachoperating rates of 100 at the same time. Since this is such a simple computationalproblem, I don't understandwhy the classificationsof the industrialproduction index are not used in theirfull detail.If they were,I suspectthat the aggregateoperating rate would automaticallylose a point or two. Furthermore,seasonality is a very tough problemfor periods of peak output.I gatherthe unadjusteddata were not used becauseoccasionally theyproduce some peculiarresults. Seasonally adjusted data may be just as misleading.In July, for example,seasonally adjusted operating rates in- creasedsharply, merely reflecting the fact that unadjustedproduction fell less than usual duringnormal vacation periods becausedemand was so strong.I thereforehave considerabledifficulty interpreting seasonally ad- justed operatingrates when the economyis close to peak output. I have a furtherquestion about how decliningindustries are handled. For example,when military equipment capacity is beingdismantled or the aerospaceindustry is declining,a gap developsbetween the McGraw-Hill operatingrate and the Wharton operatingrate. The McGraw-Hillrate George L. Perry; and Lawrence R. Klein and Virginia Long 759 tends to be much higherin the decliningindustries. The Whartonratchet techniqueworks predictably on the up side, but how does it work on the down side? Finally,just a note on the use of capitalstock figures.They are notori- ouslypoor on retirements,but thereare also problemson additions.A con- siderableportion of gross fixed investmentrepresents plants under con- struction,a categorythat is highly variable.In a periodlike the present, when there are long lags betweenconstruction starts and completions,I think(though I haveno datato verifythis) that a disproportionateamount of the increasein grossfixed assets, both as currentlyreported and as calcu- lated using survivalcurves, results from the increasein plants undercon- struction.The FederalReserve's use of the capitalstock with a 50 percent weightmay thus introducea significantupward bias to theircapacity esti- mates.I do hope that Perry'spaper will lead to a thoroughgoingrevision in this index. I find Perry'sanalysis of the cyclicalbias in the McGraw-Hillindex of operatingrates exceptionally interesting. It explainsa lot of whathas been observedbut could not be accountedfor in those series.I thinkthat Doug Greenwaldcould improvethe McGraw-Hillnumbers through an effortto adjustfor these effects.

DouglasGreenwald: I wouldlike to clarifya few pointsabout the McGraw- Hill indexrelevant to GeorgePerry's paper. First of all, McGraw-Hillgets all its answerson capacityutilization rates and investmentplans from one surveyin the spring.It is a packagesurvey which we thinkdraws relatively consistentresponses from companies. But it does havesome problems. One involvesdiversification by firms,which may make industryclassifications misleading.For example,the rubberindustry may reportthat it is adding 10 percentto capacity,when in realityall of this additionis devoted to chemicals.I don't know how to correctfor this, but it is an important qualificationto capacitynumbers for individualindustries. Another prob- lem is determiningthe amountof investmentdevoted to expansion,mod- ernization,pollution control, and employeehealth and safety. We have beenconcerned about the firstthree of these for some time, and questions designedto get at expendituresfor healthand safetyequipment were added to our surveylast spring. Withregard to data on operatingrates in specificindustries, we do have 760 Brookings Papers on Economic Activity, 3:1973 on our worksheets data on all two-digit manufacturing industries, and I will make these data available. We also now have unpublished data on thirty-eight three- and four-digit industries. Finally, I think the differences in operating rates among firms can be il- luminating. In a survey last fall, we asked companies where they were op- erating with respect to their preferredrate, and their reasons for operating above or below this level. The responses included some very odd reasons for operating above or below their preferred rate, but I do think their is significant. While 47 percent of the companies reported that they were operating under their preferred rate, 30 percent said that they were operating over the preferred rate and 23 percent said that they were operating either at the preferred rate or very close to it.

Nathan Edmonson: I agree with most of what Alan Greenspan said, es- pecially on the question of bottlenecks. A high rate of capacity utilization in one industry may not tell much about the economy as a whole, given con- siderable ability to substitute materials. Copper strikes, for example, have brought out almost unlimited possibilities for substitution for copper as an input. Our intent in forming the major materials utilization series was to assemble information on an admittedly small, but nevertheless strategic, group of materials industries. If the average utilization rate in these indus- tries were high, and the variance of the rates for individual industries around this average were small, then substitution would become a moot point-firms would have to substitute one scarce good for another. The al- ternative to this approach would presumably be to use industry capacities and input-output coefficients to build a matrix of constraint relations. Ca- pacity would then be defined as the maximum of , or some other measure of production, subject to these constraints. The weakness of this approach is that it ignores the possibilities of substitution. That is why we prefer to use information on the major materials industries. My second comment is in response to a specific point made in Perry's paper: that the capacity index obtained by dividing the industrial produc- tion index by the McGraw-Hill utilization rate shows positive correlation with the output series (Table 1). Perry suggested that one possible expla- nation of this cyclical movement is that industries find capacity when things get tight and forget about it when things get slack. This is especially likely in an industry with a very heterogeneous product mix and a fairly nebulous notion of capacity. Another possible explanation may relate to the value- George L. Perry; and Lawrence R. Klein and Virginia Long 761 addedweighting of the industrialproduction index. The benchmarkingof the industrialproduction index to the Censusof Manufacturersis done by seven-digitproduct classes, which are then added togetherusing value- addedweights. When capacity gets tight, I think it is safe to assume,pro- ducerstend to favortheir more profitableproduct lines. Assuminga posi- tive correlationbetween value added and profitability,the value-added weightingsystem would tend to exaggeratethe level of the cyclicalpeaks in the productionindex, as the productmix wouldbe shiftedtoward higher- value productswhen utilizationis high. The implied capacity measure is constructedby dividingindependent estimates of utilizationcollected fromfirms into the relevantcomponents of the industrialproduction index. If the product-mixeffect is in the productionindex (the numerator)but not in the reportedutilization rate (the denominator),the resultingcapacity measurewould exhibit the sameupward movement at cyclicalpeaks as the productionindex. The workwe have done with industrypeople leads us to believethat these product-mixeffects may not be pickedup in the reports on utilization.In materialsindustries especially, capacity is a fairlysimple and well-establishedconcept, often based on engineeringestimates that do not takeaccount of the productmix. In the paperindustry, capacity utiliza- tion is expressedas the ratio of actual tons of paperto potentialtons of paper,and the tonnagemeasure is not sensitiveto productmix. Much the same can be said for estimatesof utilizationrates in the steel industryand in petroleumrefining. To the extent that productmix may not affect the reportedutilization rates as it does the industrialproduction index, a cycli- cal bias would emergein this estimateof capacityas it alone among the capacity indexes studied is directly benchmarkedto industrialproduc- tion.

GeorgePerry: I would like to commenton the suggestion,made by Klein and Long and by Greenspan,that an input-outputapproach be taken to measuring aggregate capacity. Certainly, a lot of important information is lost in simply averaginga lot of disparateoperating rates for individual industries.And a more sophisticatedconcept of aggregatecapacity, which took accountof bottlenecksand shortages,would be niceto have.But, even if the many difficultiesin constructingsuch a measurecould be overcome, it mightnot be the best one for most uses. A materialshortage or a bottle- neck in an industrysupplying intermediate products can restrainoutput and raiseprices in a userindustry, just as a shortageof physicalcapacity in 762 Brookings Papers on Economic Activity, 3:1973 that industry would. But the two sources of restraint differ in many re- spects. It would be helpful to know that, even in the presence of under- utilized facilities, auto output could not expand because steel was unavail- able; but, for an equation, this concept might give the mislead- ing idea that the auto industry was operating at "capacity" in this situation. Quite apart from this, none of the present indexes clearly meets the criterion of a more sophisticated aggregation better than the others. If output in 1973 was constrained by bottlenecks and material shortages to a greater extent than usual, that in itself does not recommend the index that registers the highest operating rates for the year. Rather, with the present state of the art of measuring capacity, it seems better to note separately the level of operat- ing rates and conditions in the labor and materials markets and analyze from there.

General Discussion

William Branson questioned the procedure of choosing among utiliza- tion measures on the basis of their performance in explaining prices, invest- ment, trade flows, and the like. He directed this criticism at both Perry's paper and Klein's defense of the Wharton measure. Lawrence Krause, on the other hand, approved a methodology by which one picks a utilization measure on the basis of the questions one wishes it to answer. Krause agreed that a linear programming approach might produce a useful mea- sure of potential production for the entire economy, but noted that such an approach would have to make specific allowance for the possibilities of sub- stituting foreign for domestic materials. William Nordhaus suggested that labor constraints as well as capital constraints be considered in formulating a capacity concept. There was some discussion of the merits of the Wharton measure in par- ticular. Klein answered Greenspan's question about declining industries: in such cases, judgments are made as to how much of the retired capacity could be reactivated at short notice, and the downward trend in capacity is "leveled off" accordingly. Branson pointed out that the effectiveness of other utilization variables in explaining trade flows in other models weakens Klein's case for the Wharton measure. Murray Foss reported the success of the Wharton measure, as compared to the FRB measure, in his own regres- sions explaining investment between 1964 and 1973. was disturbed by the inherent inability of the Wharton measure to determine George L. Perry; and Lawrence R. Klein and Virginia Long 763 differences in capacity pressures at different cyclical peaks, except to the extent that the aggregate measure would vary due to a differenttime disper- sion of individual industry peaks. In particular, he noted that its failure to distinguish between inflationary and noninflationary peaks would hinder the Wharton measure's performance in price equations. Charles Holt and thought it would be useful to sort out the effects of preferredand actual utilization rates by including both as independent variables in Perry's regressions. Stephen Goldfeld added that this procedure would be particularly important if the response of the de- pendent variable to utilization was nonlinear as the preferredrate was ap- proached. Perry agreed that in principle one expected these nonlinearities to be important; but the limited attempts to find nonlinear effects reported in the text did not turn up many significant cases. Modigliani and Nord- haus objected to the use of utilization levels to explain rates of price in- crease. They noted that since an increase in utilization will raise prices by raising the desired markup, the proper explanatory variable for the rate of change of prices is the rate of change of the utilization rate. Otherwise, a utilization rate that simply remained high could create an ever-widening gap between prices and . Perry replied that he had reported on results from both specifications because he agreed with these doubts conceptually, but found that the empirical results did not confirm the doubts. Arthur Okun suggested that the price equation might be viewed as a reduced form of a larger model in which utilization worked through investment to accel- erate the growth of capacity. In this case, an unusually high rate of utiliza- tion could be sustained only by the continued frustration of plans to expand capacity to meet demand. Thus prices could reasonably be expected to rise, even though utilization remained stable at a high level. Foss cited independent evidence on the tightness of capacity in 1973. Two series related to capacity utilization published in Business Conditions Digest-percent of firms reporting slower deliveries, and percent of firms reporting long-term commitments to buy production materials-exceeded previous postwar peaks during 1973. Also, as of June 1973, 48 percent of the firms (weighted by gross fixed assets) responding to a BEA survey re- ported that they were short of capacity, as compared with 48 percent at the peak in 1969 and 51 percent at the 1966 peak. Foss also reported that the chief reason given by firms responding to the McGraw-Hill survey for operating below their preferred rates was a supply constraint; only 24 percent of those operating below preferred rates cited insufficient demand.