capacity in manufac

8/12/2019 Capacity in Manufac

1/42

GEORGE L. PERRY

Brookings nstitution

apacity n

anufacturing

INFORMATION ABOUT THE AMOUNT of available ndustrial apacity andabout he degree of its utilization re employed n many orms of economicanalysis. Data on utilization-operating rates-have proven useful n ex-plaining price movements nd in forming projections f the future course

of business capital nvestment. Less directly, operating ates can help ex-plain the cyclical behavior of productivity nd, through hat, changes nprofits and income shares. In some industries, operating ates could con-tribute o explanations f the size of order backlogs and, at times, couldoffer clues to real limitations on the expansion of output.

Three series of indexes on operating rates are regularly available andwidelyused. They are published by the Federal Reserve Board, McGraw-Hill, and the Wharton School. Together with data on output, each of theseseries mpliesan index of capacity. n addition, a separately stimated ndexof capacity, published by McGraw-Hill, an be used with output data toprovide a fourth measure of operating ates. All of the four measures reavailable or the manufacturing ector as a whole. The Federal ReserveBoard ndex s also disaggregated nto a two-way classification f primary

Note: I am grateful o Ellen Hahn for her able research assistance n the preparationof this paper, and to Richard Benson of Harvard University for his assistance n esti-mating some of the price equations used in the paper and for providing price and wagedata from the Data Resources, Inc., data bank. And I want to thank Nathan Edmonsonof the Division of Research and Statistics of the Federal Reserve Board for data and forinformation about capacity measures.

701


2/42

702 Brookings Papers on Economic Activity, 3:1973

processing and advanced processing ndustries, while the three other in-dexesare available or finer ndustry breakdowns, enerally t the standardindustrial lassification wo-digit ndustry evel. The main purpose of thispaper s to examine hese four measures.

For a limited number of industries, rade associations ompile furtherinformation n capacity. Reoently such data have been combined by theFederal Reserve nto an index of capacity utilization n major materialsindustries. ' The index s a potentially useful addition o the available n-formation about capacity and deserves crutiny by the profession. How-ever, t covers only a small raction of all manufacturing nd s not publiclyavailable or the individual ndustries rom which t is derived.Thus, t can-not be easily evaluated or used to check the more comprehensive ndexesjust described, lthough ome casual comparisons re reported below.

Finally, numerous ttempts avebeenmade o build up properly weightedserieson capital tocks o serve n much he same way as capacity measuresin economic analysis.2 n fact, capital stocks are employed n one part ofthe procedure or developing he Federal Reserve Board capacity ndex.However, no measure f capacity or operating ates generated ntirely rom

capital tock estimates has come nto widespread seand none s evaluatedhere.

Recent Developments

An unusual amount of attention has centered on operating ates duringthe expansion of 1971-73, particularly ince the outburst of inflation n

1973. In the six quarters ollowing he introduction of the new economicprogram n 1971, real gross national product grew at an average annualrate of 7.2 percent. Real growth exceeded n 8 percent ate n the last quar-ter of 1972and the first quarter f 1973.This extremely apid expansion edmany observers o argue hat the economy was stretching ts available pro-

1. Nathan Edmonson, Capacity Utilization n Major Materials ndustries, FederalReserve Bulletin, Vol. 59 (August 1973), pp. 564-66. The Federal Reserve had maintainedsuch an index in the past, but had not done so for many years before Edmonson recon-

structed t.2. A recent attempt s described n Robert M. Coen and Bert G. Hickman, Aggre-gate Utilization Measures of Economic Performance, Memorandum 140 (StanfordUniversity, Center for Research n Economic Growth, February 1973; processed).


3/42

George L. Perry 703

ductive capacity and that capacity imits would restrain subsequent x-pansion and create strong nflationary ressures n the industrial ector ofthe economy. In the second and third quarters f 1973, real GNP growthdid slow substantially rom ts hectic earlier pace, to an average nnual rateof 3.0 percent, t the same ime that wholesaleprices or industrial roductsrose rapidly.

The capacity-limits ypothesis s a tempting explanation or these de-velopments. However, other forces were at work pushing up prices n thisperiod: the price control program had been substantially weakened, hedollar was devalued against other currencies, nd worldwide ommodityprices weresoaring. Against his background, he price acceleration annotbe blamed implistically n excessively igh operating ates. Nor is it pos-sible to identify readily he role, if any, of capacity imitations n slowingthe growth of real output after the first quarter of the year. Consumptionspending alone accounts ully for the deceleration n GNP growth n thesecondquarter, nd consumer emand s extremely olatile and presumablywasrestrained ythe sharp rise n prices during he period. More revealing,virtually he entire lowdown n real GNP between he first and subsequent

quarters f 1973 s traceable o output n two sectors: agriculture nd auto-mobiles.Aside from hese sectors, ealGNP grewat a 6 percent ate n boththe first and third quarters nd at a 41/2 ercent ate n the second. Omittingthe GNP produced by other minor sectors hat have nothing to do withU.S. capacity imits-general government, ouseholds nd nstitutions, ndthe rest of the world-the annual GNP growth rate n the remainder f theeconomywas6.4 percent, .2 percent, nd 6.5 percent n the first hreequar-ters of 1973,respectively. These statistics offer ittle support or a supply-

limit hypothesis of a slowdown. On the other hand, the rate of inventoryaccumulation n the second and third quarters was lower than most fore-casters had predicted and could constitute evidence of supply constraintsfor some products.

Operating ates have also been the focus of attention or forecasters fplant and equipment pending n 1973,and continue to be an importantclue to many projections f business nvestment or 1974.High operatingrates have generally produced high levels of business nvestment outlays,and the extent of such an investment boom is one of the important deter-minants of the length and strength of any economic expansion.

Operating ates, then, promise significant lues to the most importantstabilization uestionsof the day. Unfortunately, he published ndexes ell


4/42


quite different tories. For the second quarter f 1973, he Federal ReserveBoard reported a rate of 83.4, McGraw-Hill's tilization rate index was86.9, Wharton eported 6.4, and calculations based on the McGraw-Hillcapacity urvey ndicated an operating ate of 81.4.3Such a range corre-sponds to the difference o be expected n manufacturing perating atesbetween he trough and peak of a mild business cycle. There s some am-biguity about whether he levels of utilization rates n these measures aredirectly omparable. Historical vidence presented elow suggests hat theyare. But even adjusting he FRB and McGraw-Hill measures y preferredoperating ates -the largest adjustment or comparability hat anyone has

suggested-still leaves a sizable discrepancy among the measures. Forsecond quarter 973, his adjustment uts the FRB index at 89.7, the indexfrom the McGraw-Hill tilization urvey at 93.4, and calculations ased onthe McGraw-Hill apacity urveyat 87.5,compared with 96.4for Wharton.By these measures, he Federal Reserve ndex and the McGraw-Hill apac-ity survey ndicated mple pare capacity, but the Wharton ndex suggestedthat a wide range of manufacturing ndustries were pushing against capac-ity limits. The answer o the crucial question of how much unused capacity

exists n American ndustry depends o an altogether unacceptable egreeon which of the widelyused measures one looks at.

AlternativeMeasures f Capacity

No one concept of capacity has general acceptance. Very oosely, capac-ity is meant o measure he output hat can be produced with the availablestock of plant and equipment. This evel of output depends on the amountof other nputs used with the capital and on changing echnical relation-ships that definehow the inputs are combined and how much output heywill produce n various combinations. f the mix of output varies, and ifcapital s specialized n its uses, the relation s complicated urther. Finally,a givenphysical acility can be utilized more or less fullyby operating moreor fewer shifts and longer or shorter workweeks.

None of the indexes attempts o narrow his range of ambiguity by de-

3. The McGraw-Hill measures analyzed here are available as end-of-year data. Esti-mates for periods within the year are made by the author by interpolating capacitygrowth between ts yearly end-points. For 1973 quarters, apacity is estimated for thispurpose to grow at the same rate as it did in 1972.


5/42


6/42


them directly omparable with government tatistics or each industry. To

obtain more aggregated stimates, ndividual ndustries re combined usingweights rom the index of industrial production.6McGraw-Hill oes not define capacity or operating ates, and firms re-

spond according o their own definitions. Firms also indicate what theirpreferred perating ates are, and again McGraw-Hill oes not attempt

to define he concept. The capacity urvey asks firmsabout both their plansfor additions o capacity n the current year and their actual additions nthe previous ear. The analysis n this paper s based on the actual additionsto capacity hat firms have reported.

The McGraw-Hill eries on additions o capacity and on operating atesare separate nd ndependent rom one another. The level of capacity ndi-cated by the one cannot be divided nto a measure f output for the indus-try to obtain the operating ate provided by the other. The survey on op-erating rates refers o December of each year and in the present analysis streated as an average operating ate for the entire month. Thus, when di-vided into seasonally adjusted output for December as measured by theFederal Reserve Board's ndex of industrial production, his series yieldsan estimate of capacity or that month.

The McGraw-Hill apacity urvey s not benchmarked o a level of utili-zation rates. Capacity n each industry s an index number qual o 100 n1967.Dividing his index nto output yields a utilization ndex. In the pres-ent analysis, his utilization ndex for each ndustry was then scaled so thatits average over the entire data period equaled he average rate from theMcGraw-Hill tilization urvey or the corresponding ndustry.

Both of the McGraw-Hill measures re subject o the normal echnical

problems f survey ampling. By the nature of what hey measure, hey canalso be suspected of having certain distinctive trengths and weaknesses.Estimates rom the capacity urvey uffer rom having no periodic bench-mark, so that any systematic error in the annual estimates of capacitygrowth will cumulate. By contrast, errors n estimates of capacity growthfrom he operating atesurveyare unlikely o cumulate ince new operatingrate benchmarks re provided annually. This survey, however, ould sufferfrom a cyclical bias if respondents reated marginal acilities differently t

6. Aggregation should be made using capacity weights rather than output weights.However, all the measures reviewed have used some form of output or value-addedweighting and the error n doing so is probably small.


7/42

George L. Perry 707

different tages of the cycle, say, by ignoring some idle facilities n esti-mating operating ates during lack periods but counting hem when theywere put back into use.

FEDERAL RESERVE BOARD INDEX

The Federal ReserveBoard ndex of manufacturing apacity s the mosteclecticof the indexes n construction, elying on three distinct sources ofinformation. The data from these sources are combined n a way that aimsto utilize the best features of each set and minimize ts weaknesses. The

FRB methodology relies on the McGraw-Hill utilization rate survey tobenchmark ts capacity ndex over the longer run. However, o estimateyear-to-year hanges in capacity, it uses two indicators of short-termcapacity growth: year-to-year hanges n the McGraw-Hill apacity urveyand estimates of the size of the capital stock. The FRB index merges hesethree kinds of information on capacity growth by estimating he historicrelationship etween he two short-run ndicators nd the utilization ratesurvey.

The historic drift in capacity between he two McGraw-Hill urveys sestimated by time trends. The time trend or the latest nterval s then ap-plied to the estimates rom the capacity urvey o provide one estimate ofyearly capacity growth. The same procedure s used to establish he timetrend of the drift between he capital stock and capacity calculated romthe utilization urvey. This time trend s then applied o the annual growthof the capital stock to provide a second estimate of capacity growth yearby year. These wo estimates fyearly apacity rowth-one from he drift-adjusted apacity urvey and one from he drift-adjusted apital tock-arethen averaged o provide he final FRB capacity ndex. Quarterly stimatesare nterpolated rom yearly estimates; and quarterly stimates of capacityutilization re derived by dividing apacity nto the FRB industrial roduc-tion index.7

A serious weakness of the FRB index is that the benchmarking o theutilization urvey s based on historic tatistical elationships hat are sim-ple at best and that may change substantially. n particular, stimates orrecent years are based on simple ime trend estimates of the drift that are

7. The most thorough published description of the index and its construction s givenin Frank de Leeuw, A Revised Index of Manufacturing Capacity, Federal ReserveBulletin, Vol. 52 (November 1966), pp. 1605-15.


8/42


heavily weighted with historical nformation. The estimates are not cur-

rently updated;8 and even if they were, they would still not adequatelyreflect any abrupt ecentchanges n the relation of investment nd capitalstock to capacity or in the bias in the McGraw-Hill apacity eries.

THE WHARTON INDEX

The Wharton measure of capacity s produced by an extremely impleprocedure. Seasonally adjusted quarterly data on output for each of thetwo-digitmanufacturing ndustries re recorded o determine eak quartersof output, and output at the peaks are aken as measures f capacity n eachindustry. Between uccessivepeaks, capacity s assumed o grow along astraight-line ath connecting hem. For the period after the most recentpeak, capacity s assumed o grow along the same straight ine that it fol-lowed before hat peak. If output subsequently oes above this line, a newcapacity estimate s definedby that level of output, and a final estimate sestablished when output eventually urns down. Thus at no time doesutilization xceed 100, and it reaches 100 at every cyclicalpeak. Some ex-ceptional ases are dealt with separately, uch as a peak followedby a briefdecline and a return of output o new highs, or a declining ndustry whoseoutput achieves ocal peaks that lie below previous peaks. In arriving t acapacitymeasure or all manufacturing, ndividual ndustries re aggregatedusing value-added weights.

On one occasion, Klein and Preston checked he estimates of capacitybased on the basic Wharton methodology by comparing hem with esti-mates rom a production unction or several ndividual ndustries.9 n light

of evidence hat some of the basic Wharton eries were drifting away fromthe production unction estimates, he Wharton ndex was adjusted up-ward hrough 1960.Wharton apacity estimates or later years have beenmade with the basic methodology.'0

The obvious drawback o the Wharton methodology s its treatment f

8. The most recent estimates of the drift use data through 1970.9. L. R. Klein and R. S. Preston, Some New Results n the Measurement f Capac-

ity Utilization, American Economic Review, Vol. 57 (March 1967), pp. 34-58.

10. The basic methodology s not considered acred and apparently ome of the esti-mates have occasionally been amended. However, the index is constructed essentiallyas described here.


9/42

George L. Perry 709

everypeak n output as a point of full utilization. The criticism hat the

1959-60output peak clearly should not have been regarded s a point offull utilization purred he adjustment o the series ust described. And intheir analysis, Klein and Preston concluded hat the adjustment was re-quired because he output peaks n the early 1950s represented verutiliza-tion for purposes of the Wharton ndex as well as because the 1959-60peaks were periods of less than full utilization.1' Since, for individual n-dustries, perating ates at peaks are defined o be 100, the basic Whartonmethodology annot distinguish differences n the intensity of utilizationfrom one peak to another. The peak operating ate for all manufacturingor some other aggregation will be less than 100 since all industries do notpeak n the same quarter. Thus peak operating ates or manufacturing andiffer rom cycle to cycle because of differences n the distribution f in-dividual ndustry peaks n time, but not because he intensity of utilizationat those industry peaks s measured o be different.

Because he most recent estimate of capacity n the Wharton methodol-ogy is provisional until a peak in output s reached, another drawback fthe Wharton methodology s that current estimates of capacity and op-erating ates are always ubject o revisiondepending n the course of out-put. If output exceeds he capacity ine extrapolated rom the most recentpeak, capacity will be defined o coincide with output until output slowsand a new peak is established o define capacity for the present cycle.Retrospectively, perating rates initially reported as 100 may be reviseddownward ubstantially. Conversely, f output expands weakly and peaksbefore reaching he capacity ine that had been extrapolated rom the mostrecent peak, nitially eported perating ateswillbe revisedupward. Quar-

ters in which spare capacity was initially reported o have been ample willhistorically be shown to have been periods of full utilization. Thus theWharton ndexcan tell different tories o the researcher sing t historicallyand to the decision maker using t currently.

The simplicity of the Wharton methodology s also its great strength.The technique s easily applied and yields prompt estimates of capacityutilization over a wide range of industries. n recent years, t has been ap-plied to data on industrial production n many countries other than the

United States o produce historic and current stimates f utilization. While11. Some New Results, pp. 54-55.


10/42


its drawbacks re unmistakable, he Wharton ndex may serve very wellwhere a measure of changes n utilization rates over relatively hort timeintervals s useful.

Characteristics f the Measures

Several haracteristics f these four measures f manufacturing apacityand utilization an be examined by comparing heir past behavior. Someof the findings rom this examination re suggested by the descriptions fhow the several measures re constructed nd how they have been utilizedin the past. But there are also a few surprises.

CYCLICAL BIAS

Table 1 shows regression stimates summarizing he relation betweencapacity as estimated by the four alternative measures nd two other vari-

ables, output and the capital stock. All variables are in logarithmic ormand thus summarize he relation between hanges n output and n the cap-ital stock and changes in the capacity measure. Separate estimates areshown or the 1954-65 and 1966-72 periods. A variety of evidence, nclud-ing the regression stimates f Table 1themselves, oint to a change aroundthe mid-sixties n the relation among he four measures f capacity and inthe relation between wo of the measures and investment or the capitalstock. The form of the equation does not represent ny structural ypoth-

esis, but rather offers a preliminary way to view the characteristics f theseveral apacity measures.The principal esult of interest s the significant ositive relationship e-

tween output changesand changes n capacity s measured y the McGraw-Hill utilization survey. The estimated effect is virtually dentical n bothregression eriods. There s no important elation between current outputand current capacity or any of the other measures. Since output enterswithout a lag, and since variation n the capital tock should capture muchof the true variation n capacity, t is extremely doubtful hat this relationbetween output and capacity epresents genuine case of rising output n-ducing capacity growth. Rather, capacity as measured rom the utilization


11/42

George L. Perry 711

Table 1. Relation betweenManufacturing apacity, Output, and

Capital Stock, Measured y Four Indexes, 1954-65 and 1966-72aCoefficient stimatesb Standard Durbin-

error of WatsonIndex Period a b c estimate R2 statistic

McGraw-Hill Cc 1954-65 -22.2 0.08 1.38 0.015 0.992 1.0(-12.3) (1.1) (12.6)

1966-72 -21.5 0.12 1.34 0.016 0.980 0.9(-12.4) (0.8) (12.2)

McGraw-Hill Uc 1954-65 -16.2 0.23 1.03 0.009 0.996 2.9(-13.9) (4.9) (14.5)

1966-72 -11.6 0.23 0.80 0.004 0.997 2.0(-24.7) (5.7) (27.0)

Wharton 1954-65 -26.7 0.05 1.38 0.015 0.990 0.7(-14.1) (0.7) (11.9)

1966-72 -17.2 0.02 0.89 0.008 0.988 1.0(-19.8) (0.3) (16.3)

Federal Reserve 1954-65 -21.7 0.05 1.37 0.007 0.998 0.7

Board (-26.1) (1.4) (26.9)1966-72 -20.8 0.01 1.32 0.001 0.996 1.3

(-29.0) (0.1) (29.4)

Sources: Capacity indexes are from McGraw-Hill Publications Company, Economics Department,Annual Survey of U.S. Business' Plans for New Plants and Equipment, April 1973 and preceding annual

issues; Wharton School of Finance and Commerce, University of Pennsylvania, Economic Research Unit,United States Aggregate Industrial Capacity Utilitization Rates (July and September 1973; computer

printouts); Board of Governors of the Federal Reserve System (computer cards). Industrial production andcapital stock data were provided by the Federal Reserve System.

a. The data are observations for the fourth quarter of each year. The numbers in parentheses aret-statistics.

b. The estimating equation is

In C = a + b In Q + c In K,whereC = capacity measure, last quarter of each yearQ = Federal Reserve index of industrial production in manufacturingK = capital stock.

c. Here and in the following tables McGraw-Hill-C and McGraw-Hill-U are indexes based onMcGraw-Hill capacity and utilization surveys, respectively.

surveydoes seem o have a cyclicalbias. It appears hat respondents findcapacity when output rises sharply, and lose t when output slackens.

It is not clear whether he results reflect imply a bias of respondents o

the survey, or the thinking of management bout how much capacity sactually available. f they reflect he thinking of decision makers, hen areported tilization ate hat subsequently roved o be too high when out-


12/42


put expanded and new capacity was discovered ould still produce he

economic onsequences f tight capacity. The ndexwith cyclicalbias couldpredict well.Indeed, ince most economicdevelopments ne would orecastusing utilization ates are ust as cyclical as the bias in the index s, predic-tions could be expected o be little affectedby the bias, whatever ts source.

The estimated lasticity between output and capacity measured by theutilization urvey s about one-quarter. Thus, assuming any given actualgrowth n capacity, f output were to grow 8 percent ather han zero in agiven year, the utilization urvey would indicate a 2 percent difference ncapacity rowth and a corresponding wo-point narrower pread n utiliza-tion rates han actuallywouldcharacterize he two alternative utput paths.For individual ears, there is some evidence hat the bias may have beennoticeably arger: n 1966, the survey ndicated a growth in capacity ofmore than 10 percent, and in 1970, a growth of only 0.3 percent. If, asseems reasonable, apacity s interpreted s the quantities hat firms indthey can produce when actually put to the test, the utilization urvey esti-mates are most reliable at high levelsof utilization, and comparisons withother ndexes are best made for such periods.

This findingalso validates he use of the utilization urvey o benchmarkinformation bout capacity evels n the FRB index while other measuresare used to estimate year-to-year hanges n the index. However, he FRBmethodology of estimating ime trends among he different measures maynot be optimal, ince or relatively hort ime ntervals, trend estimate anbe too much nfluenced y a few observations. t might be better o bench-mark o estimates rom the utilization urveyat its latest peak. Even bettermight be adjustment f the capacity mplied by the utilization urvey with

an equation uch as that n Table 1and application f the adjusted stimateto correct he drift n the other nputs of the FRB index.

The other result of interest n Table 1is the change n the relation of cap-ital stock growth o capacity growth between he two periods or the util-ization survey ndex and the Wharton ndex. The relation s little changedbetween he two periods or the other capacity ndexes. Sincesome drift sexpected between measures of capital stock growth and capacity, and thetiming between he two is not known accurately, o great significance s

attached o the differences n coefficients mong he measures. But the de-cline n the estimated lasticitybetweenperiods n two of the measures oescorrespond o the hypothesis hat pollution control efforts, and possiblyother developments, ave reduced he annual ncrement o capacity hat


13/42

George L. Perry 713

goes along with a given evel of investment pending. The size of the effectis substantial. According o the estimates or the utilization urvey ndex,to achieve a given percentage ncrease n capacity now requires 5 percentfaster growth n the capital stock-roughly equivalent o 10 percent moregross investment-than it used to. The effect estimated or the Whartonindex is even greater.

GROWTH IN CAPACITY

In Table 2, the growth n manufacturing apacity, as measured by the

four alternative ndexes, is shown for different ntervals since the mid-1950s.The capacity measures are for the end of each year, and the endpoints of the intervals hown correspond pproximately o cyclicalpeaks noutput. Average apacity growth rates were not far apart n the four mea-sures for the decade rom the end of 1956to the end of 1966.But in the1966-72 period, they diverge substantially. The FRB index and theMcGraw-Hill apacity survey record a speedup n capacity growth whilethe McGraw-Hill tilization urvey and Wharton record lowdowns.This

divergence oincides with the changes n the relation of the capital stockto the various measures f capacity mplied n Table 1.The growth rates for the smaller ubintervals hown n the second part

of the table are more erratic, particularly or the McGraw-Hill tilizationmeasure during he first decade. Allowing or the cyclicalbias that has beenidentified n this index smooths he picture onsiderably. Over he 1956-60interval, output grew at an average annual rate of only 0.6 percent; taccelerated o an 8.1 percent growth rate over the 1960-66 interval. The

Table 2. Growth Rates n Manufacturing apacity, s Measured y FourIndexes, 1956-66, 1966-72, and SubintervalsAverage percent per year

Intervalsa Subintervalsa

Index 1956-66 1966-72 1956-60 1960-66 1966-69 1969-72

Federal Reserve Board 4.6 5.0 4.3 4.8 6.0 4.0McGraw-Hill-C 4.7 5.4 4.7 4.7 5.6 5.1McGraw-Hill-U 4.7 3.7 2.7 6.0 4.8 2.6Wharton 4.3 3.5 4.8 4.0 3.8 3.2

Sources: Same as Table 1.a. Capacity was estimated for the fourth quarter of each year, except for the McGraw-Hill measures,

which are for December of the years shown,


14/42

CON

.4 ~ ~ ~ ~ ~ ~ ~ ~ ~~~ . '~~~~~~~~~~~-4

00


15/42

George L. Perry 715

indicated regression adjustment to the capacity growth estimate wouldnarrow the speedup of 3.3 percentage points shown between the two inter-vals to about 1.6 points. On an adjusted basis, the capacity growth rates inthis index for the two periods would be about 3.7 percent and 5.3 percent.

The growth rate in the Wharton index slows successively in each of thesubintervals; the deceleration during the first decade seems particularlyimplausible. Investment ncentives were introduced in 1962 and contributedto a spectacular rise in investment spending and a marked acceleration inthe growth of the capital stock. While a translation of this development intoa capacity estimate cannot be made with any precision, it seems unlikely

that capacity growth would have slowed in this period.The adjustment to the Wharton index that was made through 1960 put

its growth rate up to that point in line with those of the other measures.But as a result of lagging the others noticeably after that time, Whartonutilization rates were substantially higher at the 1966 and 1969 peaks thanthose recorded by the other measures:12

Index 1966 1969

Wharton 96.1 96.2McGraw-Hill-U 90.0 85.8McGraw-Hill-C 91.5 87.3Federal Reserve Board 92.3 87.1

The divergences starting in the mid-1960s can be seen in Figure 1, whichcharts historical utilization rates for all four measures. Curiously, a dis-proportionate amount of the departure of Wharton from the other indexesoccurs in one year, 1966.

PREFERRED OPERATING RATES

The usefulness of any index of operating rates can be judged from itsperformance as an economic time series. The absolute level of the seriesneed not have any well-defined meaning. But for purposes of comparingone index with another, it is useful to know how their levels are expectedto be related. The Federal Reserve index was explicitly benchmarked to theMcGraw-Hill utilization survey when it was constructed, so no adjustment

12. Peaks did not occur in exactly the same quarters or all measures. The operatingrates given are for the second quarters of 1966 and 1969.


16/42


is needed n comparing heir operating ates. The McGraw-Hill apacity

survey yields only an index number of capacity evels that, for the presentstudy, has been benchmarked o average he same as the utilization urvey.Thus, of the four indexes under review, only the McGraw-Hill tilizationsurvey and the Wharton ndex produce ndependent stimates of the levelof operating ates. It is useful to know how they should be related.

Respondents o the McGraw-Hill utilization survey indicate a pre-ferred evel of operating ate as well as an actual level in each period.Analysts have often suggested hat actual rates should be adjusted by pre-ferred ates or the purpose of comparing he estimates rom the McGraw-Hill survey with those from Wharton13-in other words, hat the level ofthe Wharton ndex should be comparable with the level of the utilizationsurvey as a fraction of the preferred operating rate the survey reports.However ntuitively appealing uch a simple adjustment may be, the his-tory of the two measures does not support ts application.

The data below show the ratio of operating ates in the McGraw-Hillsurvey o those estimated by Wharton or periods starting with the firstyear for which the survey was available-1954-and ending with everyyear since 1965.As of the mid-1960s, he average evel of operating atesreported y the two series was virtually dentical. The ratio between he twodeclines very year as the period s extended rom 1965 o the present. Thisoccurs because he Wharton ndex rose substantially bove the McGraw-Hill in 1966 and stayed above t thereafter. As a result, or the sevenyearsfrom 1966to 1972, he McGraw-Hill ndex (for December) averaged only90.8 percent of the Wharton ndex for the fourth quarter):

McGraw-Hill peratingrate as a percentage

Period of Wharton ate

1954 o:1965 100.41966 99.61967 99.01968 98.31969 97.61970 97.41971 97.01972 96.7

13. See, for instance, Klein's comment n Measures of Productive Capacity, pp. 56-57.


17/42

George L. Perry 717

The average relation between the two measures through the mid-1960s isappropriate for comparing the indexes. The Wharton index was examinedand revised at about that time.14 Furthermore, the recent and current evi-dence on operating rates is best examined using evidence about the relationbetween the two measures established from previous years. However, evenif the ratio of the two measures up to 1972-96.7-were used to make themcomparable, such a procedure would still imply a much smaller adjustmentthan adjusting by the preferred rate reported by McGraw-Hill, which was93 in 1972 for all manufacturing.

An examination of individual industries reported in both surveys sup-

ports the conclusion that no level adjustment is required between the two.By 1968, when the average ratio over the 1954-68 period was still 98.3 per-cent in all manufacturing, the 1954-68 average McGraw-Hill operatingrate exceeded the average Wharton operating rate over the same period infive of the eleven individual industries. Thus there is simply no evidencethat over most of the history of the two surveys the levels of operating ratesthey report should not be compared directly, either at the all-manufacturinglevel or at some level of disaggregation. Adjusting the McGraw-Hill index

by its preferred operating rate would bring operating rates in the twomeasures closer together for the most recent years, but only at the expenseof moving them further apart in earlier periods.

OPERATING RATE CORRELATIONS

Despite the differences among the four measures already cited, theirmeasures of operating rates are highly correlated, even over the whole1954-72 period. The correlation of

the Wharton index with the others isclearly the weakest. But as the marked divergence of the Wharton indexfrom the others occurs rather abruptly in the mid-1960s, the Wharton cor-relations are much higher if the whole period is divided into two parts.When the period is separated into intervals covering fourth quarter 1954to fourth quarter 1965 and first quarter 1966 to fourth quarter 1972, theWharton correlations rise substantially and become virtually indistinguish-able from the correlations among the other measures. These correlationsare presented in Table 3.

Correlating operating rates is a weak test of the similarity of the indexes

14. Klein and Preston, Some New Results.


18/42


Table 3. Correlation Matrix of Manufacturing perating RatesamongFour Indexes, 1954-72, and Subperiods

Index

McGraw- McGraw- FederalIndex Hill-U Hill-C Wharton Reserve Board

1954:4 to 1972:4McGraw-Hill-U 1.00McGraw-Hill-C 0.93 1.00Wharton 0.73 0.71 1.00Federal Reserve Board 0.96 0.99 0.75 1.00



Sources: Same as Table 1.

sinceall four measures ave ndustrial roduction s a common numerator.The correlations o show that despite heir differences, ll four are likelyto be useful or many purposes. At a minimum, hey must all gauge cyclicalvariations with some success, measuring ifferences etween years of highand ow utilization hat followone another airly losely. Since he measureshave been shown o drift substantially part over time, however, hey can-not be comparably uccessful n answering harder questions, such aswhether operating ates n 1972were already near peak levels.

Recent Hazards or Estimating Capacity

Much of the preceding nalysis points up the fact that alternative apac-ity estimates have been agreeing ess with one another ince the mid-1960sthan they used o. Coinciding with this development reseveraloccurrencesin the economy hat may have posed special difficulties or capacity mea-surements. Two of these, about which only a little can be said here, are theacceleration f wages and the striking hanges n competitiveness etweenthe United States and other nations.


19/42

George L. Perry 719

The acceleration n wages of the late 1960s and 1970s is sometimesthought o have spurred more capital-intensive roduction echniques ndhence o have altered he relation between he capital stock and capacity.If the prices of capital goods reflect actual wage levels-which they prob-ably do-and if costs of financial apital reflect expectedwage ncreases-which s much less certain-no shifts in production echniques would bepredicted imply rom he onset of inflation. t is beyond he present naly-sis to judge whether n the real world of recent years, inflation n factaffected production echniques nd capacity growth.

Changes n international ompetitiveness ave been striking and couldwell have affected economic capacity n U.S. industries. A concentratedoligopolistic ndustry uch as steel has been subjected o competitive res-sure rom mports produced y technically more advanced acilities broad.Modernization f the U.S. facilities lso may have been profitable or sometime; but they were not treated as obsolete until foreign competition n-truded. More generally, he inroads of foreign competition n particularlines can make U.S. capacity obsolete even in competitive ectors. But aspervasive s such a developmentmay appear o be, if the newest echnology

is available o U.S. producers, he presence of foreign competition tself isnot the key to accelerated bsolescence. The key is a slower rise in unitlabor costs abroad han here because of greater moderation f wages rela-tive to productivity rowth. The importance of this factor cannot be an-alyzed further here, and it is hard even to speculate about which of theavailable measures f capacity might most successfully etect obsolescencefrom this source.

THE ANTIPOLLUTION DRIVE

Somewhat more can be said about a third development n the economyover this recent period that may have posed special difficulties or someways of measuring apacity: he intensified rive o reduce environmentalpollution rom industrial ources.

In recent years, a great deal of public attention has been focused on theproblems f polluted air and water. ndustry as been dentified s a majorsource of pollution and has been the object of intensified antipollutionefforts. These efforts have the effect of altering historically stimated mea-sures of the capital stock and industrial apacity at both ends of the pro-ductive ife of fixed capital: To the extent that some portion of current


20/42


investment xpenditures re made for the purpose of cleaning up produc-

tion processes, nvestment will add less to capacity than it has addedhistorically, and less to the capital stock conceived as a means of aug-menting production. And to the extent hat some facilities are abandonedahead of schedulebecause hey cannot economically e altered o conformto new environmental tandards, he capital tock and capacity are reducedby retirement more quickly han historical xperiencewould predict.

Since 1967,McGraw-Hill as surveyed irms o determine he amount ofinvestment xpenditures eing devoted o pollution abatement. Accordingto these surveys, uch expenditures ave risen from $785 million n 1967to $2.6 billion in 1972 for the manufacturing ector as a whole, and rep-resented bout 8 percent of total plant and equipment utlays by manufac-turing irms n the later year.15

A good deal of ambiguity urrounds he McGraw-Hill stimates. Somefirms may report new facilities hat meet pollution standards s pollutioncontrol measures ven hough hey are also additions o the firm's apacity.I know of no way to make an allowance or this possibility. But on theassumption hat the outlays reported by McGraw-Hill re exclusively orpollution control and do not add to capacity, an estimate of the bias in acapital stock measure uch as that used in constructing he FRB capacityindex can be made. I have done this for the two subgroups f manufactur-ing for which separate ndexes are presented, dvanced and primary pro-cessing industries. The McGraw-Hill stimates of outlays for pollutioncontrol were calculated or these two industry roups and subtracted romtotal investment outlays in each group for each year. New estimates ofcapital stock were then generated y deflating his reduced evel of invest-

ment spending, and these new estimates were translated nto capacityfigures by means of the FRB formulas.

By the end of 1972,according o these calculations, he portion of theFRB capacity measure hat is generated sing nvestment nd capital stockestimates was 3.3 percent oo high n the primary rocessing ndustries nd1.5 percent oo high in the advanced processing ndustries. Since the FRBcapacity estimates use the McGraw-Hill apacity survey together withthese capital stock estimates, he actual FRB capacity ndex would be off

by only half these amounts on the basis of this one adjustment. And ifsome part of the outlays dentified with pollution abatement lso add to

15. Business' Plans for Plant and Equipment, 1972-75.


21/42

George L. Perry 721

capacity, he importance f this adjustment or the aggregates hat the FRBuses must be regarded s small. For individual ndustries, he effect couldbe considerably arger. But if the pollution control movement has had animportant ffect on industrial apacity over a range of industries, t musthave come by the other route-by forcing early retirement of existingfacilities.

There are no direct measures f the retirement r obsolescence f capitalfacilities. Historical estimates are available on average ives of varioustypes of capital and these are used in empirical nvestigations uch as theFRB capacity ndex. But they offer no warning of changes n typical his-torical patterns. n principle, ome of the capacity stimates hat are underreview would be capable of detecting uch changes. If output peaks werealways caused by capacity imitations, he Wharton methodology woulddetect them along with other influences n capacity growth. The half ofthe FRB capacity measure hat rests on a historical elation between cap-ital stock and capacity ould not detect such a change. And one can onlyhypothesize hat respondents o the McGraw-Hill urveys take properaccount of this source of change n their capacity. t could be that the ca-

pacity surveydetects he capacity nlargement rising rom new nvestmentmore accurately han it does the subtractions ue to retirements r obso-lescence.The former nvolvesmoney-capital budgeting, ontracting, ndspending-while the latter does not. There s no comparable eason o be-lievethat the operating ate survey s biased n its allowance or retirementsand obsolescence ince t represents, deallyat least, a fresh assessment achyear of the utilization of available apacity.

The capacity growth rates displayed n Table 2 are consistent with these

hypotheses about retirements nd obsolescence: hat they were unusuallyheavy during his period and that the utilization urvey detected his factwhile he capacity urvey ailed o do so. The utilization urvey ecordedsharp lowdown n capacity growth after he mid-1960s,while he capacitysurvey eflected nlya slight slowdown nd recorded capacity rowth atetwice as large as the utilization urvey during he 1969-72 nterval.

DisaggregatingMeasures f Capacity

As the overall evel of business activity varies, capacity pressures re noteven among ndividual ndustries. While some industries how more pro-


22/42


nounced yclicalvariation n output han others, heir endency o do so is

not strong nough o ensure egularity n capacitypressure cross ndustriesfrom one cycle o another. Aggregatemeasures f capacity utilization, uchas the widely reported measures of operating ates in all manufacturing,thus conceal a great deal of irregularity n the position of individual n-dustries. Yet it may make quite a lot of difference f the average operatingrate for all manufacturing s a couple of points below full utilization be-cause most manufacturing ndustries are in that position rather han be-cause some industries are operating at the limits of their capacity whileothers suffer an overhang f idle facilities. n the 1973 economy, nforma-tion from a variety of sources ndicates hat capacity imited he expansionof output n several ndustries uch as paper, petroleum, teel, and somelines of chemicals, while many others exhibited evidence of ample sparecapacity.

The significance f measures of capacity and of capacity utilization willvary from ndustry o industry. Prices an be expected o be more sensitiveto the degree of utilization n some ndustries han n others. n sufficientlyconcentrated ndustries, where irms aim for a target rate of return, pricesmay even have a negative relation o utilization, rising when utilization,productivity, nd profits are low. Similarly, he significance f utilizationrates or explaining nvestment willdiffer among ndustries. n some ndus-tries, capacity represents true physical imit to production. Those em-ploying continuous process operations, such as petroleum and paper,typically use facilities as intensively as demand permits, running themnearly wenty-four ours a day, seven days a week. When all facilities arerunning at these rates, a meaningful physical imit to capacity s reached.

Average ost curves may be flat or declining ight up to this point in suchoperations. By contrast, n others, such as the automobile ndustry, pro-duction s geared o a typical workweek ut is easilyexpanded by runningproduction ines more days or longer hours each day. The average aborcost of doing so is higher, at least after a point, because of overtime pay.But other costs are spread more widely. It may be profitable o expandoutput very substantially eyond the normal operating evel, with costcurves lat or decliningwellpast what scustomarily egarded s 100percentof capacity. The implications, oth for new investment nd for

price pres-sures, are thus widelydifferent or petroleum efiningand for automobileproduction.

Disaggregating he available measures of manufacturing apacity per-


23/42

George L. Perry 723

mits a look at how utilization affects he different ndustries and, at thesame time, how the alternative measures of utilization are as forecastingvariables. Such analysis can be helpful n several ways in evaluating hemerits of alternative measures nd n assessing he current tate of capacityutilization. t can identify he measures hat perform he best, and the in-dustries or which correct measures f utilization are mportant. And it candetermine he industries n which some agreement xists among he alter-native measures.

Unfortunately, he measures f capacity and utilization available or allmanufacturing re not all available n the same disaggregated orm. The

Federal Reserve Board ndex s disaggregated nly into advanced nd pri-mary processing ndustries. The Wharton ndex is thoroughly disaggre-gated, basically at the two-digit ndustry evel. The two McGraw-Hill ur-vey measures re available or somewhat ewer ndustries. n the followinganalysis omparisons re imited o the Wharton and the two McGraw-Hillmeasures, nd among hese, to industries or which data were available orat least two of the indexes.

Predicting Capacity Growth

High operating ates should, other things equal, nduce irms to add totheir capacity. A natural est of the indexesunder eview, herefore, s theirability to predict their own capacity growth rates from their own pastutilization ates. How the three measures under review ared n such a testis reported n Table 4. The percentage ncrease n capacity or each measurewas explained y past valuesof its own operating atesand past ncreases noutput. Output change s included as a way of capturing he effect of ex-pected uture changes n output on capacity decisions. The exact form ofthe equation used is shown in Table 4. The table presents he t-ratios ofthe operating ate variable n the equation explaining apacity growth oreach of fourteen ndustries s well as for all manufacturing.

Both McGraw-Hill measures xplain hemselveswell. The capacity ur-vey measure egistersa wrong sign in only one of the fourteen ndustries,food, and has a t-ratio ower han 2.0 in three others. The utilization urveymeasure oesabout as well.It has wrong igns n two industries nd t-ratioslower han 2.0 in only one other. Both of these measures lso perform wellin explaining heir own estimates of capacity growth n all manufacturing.


24/42


Table 4. Capacity Growth of Selected Industries as Explained byOwn Operating Rates of Three Indexes

t-ratio of operating ate variablein industry apacity equationa

Standardindustrial McGraw- McGraw-

classification Hill-C Hill-U Whartoncode Industry index index index

20 Food -5.0 3.2 -6.522 Textiles 1.3 4.8b 4. 9b26 Paper 4.4 -0.5b - 3.0b28 Chemicals 3.2 5.9 -0.429 Petroleum 3.6 3. lb 1.130 Rubber 4.0 2.9 0.6b32 Stone, clay, and glass 7.7 4. lb -1.333 Primary metals 4.3b ... _3.4

333-36, 339 Nonferrous metals 5.Ob -1.2b34, 38 Fabricated metals and instruments 0.6 5. 6b 1 .Ob

35 Machinery 6.9 0.7 -4.036 Electrical machinery 3.Ob 2.7 -8.4

371 Motor vehicles 6.9b 3.2 -0.4b372-75, 379 Other transportation equipment 1.0 2.8 -1.1

All manufacturing 5.1 5.1 0.2Sources: Same as Table 1.a. The capacity equation used was

Ct -Ct-4 = a + b(Ut-3 + Ut-4 + Ut-5 + Ut-6 + c Qt3 -Qt-7)Ct-4 4 1 Qt-7' '

whereCt = quarterly capacityUt = quarterly operating rateQt = quarterly output.

The period of estimation is 1956:2 to 1972:4.b. The output term had the wrong sign in the basic capacity equation. The t-ratio shown is for the operat-

ing rate term in the equation with the output term omitted.

The Wharton ndex fails in almost every ndustry. Only n textiles does itsown estimate of operating ates succeed n explaining ts estimate of capac-ity growth. Its operating rate variable has the wrong sign in nine otherindustries nd t-ratios essthan 2.0 in the remaining hree ndustries. t alsohas a negligible oefficient n its equation or all manufacturing.

Predicting nvestment

For GNP forecasting, he analyst s primarily nterested n predictinginvestment pending rather than capacity growth. While the expected


25/42

George L. Perry 725

causal relation between operating ates and investment s less precise hanthe relation between operating rates and capacity growth that was justanalyzed, nvestment s, in most instances, he primary action that firmscan take to expand capacity. In practice, operating rates are commonlyused by forecasters o help explain nvestment, nd the ability of such anindex to do so is an important est of its general usefulness.

How the three ndexes ared n explaining nvestment s shown n Table 5,which displays -ratios or their operating ates n an investment quationfor twelve industries as well as for all manufacturing. he equation ex-plained the ratio of deflated nvestment o output by past values of op-erating rates. Capital stocks for individual ndustries were not available,so it was not possible o try to explain nvestment s a fraction of the capi-tal stock. The exact form of the estimating quation s given in the table.

Table 5. Performance of Alternative Measures of Operating Rates inExplaining Investment

t-ratio of operating ate variablein industry nvestment quationa

Standardindustrial McGraw- McGraw-

classification Hill-C Hill-U Whartoncode Industry index index index

20 Food -2.3 1.2 1.422 Textiles 3.2 3.1 6.126 Paper 3.8 4.8 0.228 Chemicals -2.4 3.7 0.129 Petroleum 0.0 2.9 4.130 Rubber 3.9 2.2 1.032 Stone, clay, and glass 5.0 7.5 2.2

33 Primary metals 3.0 n.a. 2.534,38 Fabricated metals and instruments 2.0 3.9 4.7

35 Machinery 6.4 4.1 4.136 Electrical machinery 2.8 2.7 3.237 Transportation equipment 5.7 2.3 2.4

All manufacturing 7.3 5.3 3.5

Sources: Same as Table 1.a. The equation used is

It = a + b U1(t) + U4(t-1) + U3(t-1) + U2(t-l)Qt 4

where

It = investment in year tQt output in year tUT(t) = utilization in the T quarter of year t.

The equations were estimated for the period 1956-71.n.a. Not available.


26/42

Table 6. Prediction f Price Changes n Selected ndustries romOperating Rates of Alternative ndexes

Coefficient and t-ratio (in parentheses)of operating ate variable nindustry rice equationa

Standard McGraw- McGraw-industrial Hill-C Hill- U Wharton

classification ode Industry index index index

20 Foodb -0.048 -0.033 0.090(-2.5) (-0.6) (1.9)

22 Textiles 0.1160 0.1270 0.057c, d(3.4) (3.3) (2.4)

26 Paper 0.202 0.236 0.044(5.2) (5.8) (1.7)28 Chemicals 0.010 0.027 0.025

(1.0) (1.1) (1.2)29 Petroleum 0.031f 0.050f 0.390

(0.5) (0.6) (2.3)30 Rubber (0.031)0 0.0980 0.120e

(1.7) (2.6) (4.3)32 Stone, clay, and glass 0.005 0.015 -0.019

(0.3) (1.0) (-0.9)331,332 Iron and steelg 0.073 ... 0.064h

(4.4) (3.2)333-36, 339 Nonferrous metalsg 0.198d 0.196d 0.134d,h

(3.6) (3.3) (2.8)35 Machinery 0.031 0.035 0.021

(3.6) (2.7) (2.6)36 Electrical machinery 0.005 0.017 0.007

(0.3) (1.0) (0.5)371 Motor vehicles -0.034 -0.049 -0.038

(-3.6) (-3.8) (-3.6)

Sources: Price data were assembled by Richard Benson of Harvard University from wholesale price in-dexes. Wage data are from Employment and Earnings, various issues. The sources of the indexes are the

same as in Table 1.a. The equation used isPt =a + bU laggedt + c W + d Pt

p laggedt-i w laggedt-1 P laggedg-l

where the time interval is one quarter andPt = the wholesale price of industry outputUt = capacity utilization, scaled as a decimal (for example, 0.90)wg = straight-time hourly earnings in the industryPt = the price of material inputs to the industry

p laggedt = O.4pt + 0.3pt-1 + 0.2pt_2 + 0.1pt_3, and w lagged, U lagged, and P lagged are definedanalogously.

The period of estimation is 1955:3 to 1971:2.b. Data on wages were not available.c. The input price variable entered the equation with the wrong sign. The coefficient shown is for the

equation with input prices dropped.d. The wage variable entered the equation with the wrong sign and was dropped.e. Data on input prices were not available.f. The utilization rate variable is for petroleum refining only (SIC industries 291 and 299).g. The wage variable is for primary metals (SIC 33).h. The utilization rate for primary metals (SIC 33) is used.


27/42

George L. Perry 727

The Wharton ndex of operating ates explains nvestment much moresuccessfully han it explains ts own estimate of capacity growth. None ofthe indexes works well in the food industry, and Wharton displays a t-statistic greater han 2.0 in eight of the remaining leven ndustries s wellas in all manufacturing. t still does not perform quite as well as the twoMcGraw-Hill measures. The utilization eries has a significant oefficientin every ndustry ut food, while he capacity eries ails n two others. Theyalso have noticeably higher -ratios n the all-manufacturing quation. Butthe Wharton ndex explains nvestment well, and, in the present orm ofthe equation, only slightly ess well than its competitors.

Predicting rice Changes

All three measures of operating rates prove useful in predicting pricechanges. Over all, the Wharton measure does as well as the two McGraw-Hill surveys, nd no one of the three s clearly uperior n predicting rices.A comparison f their performances s provided n Table 6, where he co-

efficients and t-statistics or the operating rate terms are compared nequations or twelve separate ndustries.

All the price equations were estimated or periods ending n the secondquarter f 1971.Data were available o start the estimation period, withthree-quarter ags, in the third quarter of 1955. Ending the estimationperiod n mid-1971avoided he Phase I, II, and III price control episodes.Since here s no way to know the effect of controls on prices n individualindustries during this era of abruptly changing wage-price policies, esti-

mates of normal effects would be distorted. A utilization ndex that wastoo low during he initial reeze and Phase II would predict prices betterthan t should n that period. An index hat was too high might do betterthan t should during he Phase III stage of suddenly bsent controls. Con-fining the estimation o the years before Phase I avoids making specialallowances n the equations or all these changes. During the wage-priceguidepost eriod of 1962-68, restraints were much milder and, on the evi-denceas I interpret t, acted mainly by moderating wage ncreases nd keep-ing prices n step with wages. Since cost changes are accounted or in theprice equations stimated here, little room remains or a separate guide-post effect. n a fewindustries ome effectmay have been felt, particularlyin 1968,but no attempt was made to allow for it.


28/42


For each industry, basic price equations were estimated explaining hechange n the price of the industry's utput by the level of capacity utiliza-tion, the change n the average wage n the industry measured y straight-time hourly earnings of production workers), nd the change n the priceof the industry's material nputs. In two cases, data on wages or materialsprices were not available. And in some others, one or both took on thewrong sign and the equation was reestimated without them. The threechange variables had the form,

xt/(0.4xt-, + 0.3xt-2 + 0.2xt-3 + O.1Xt-4).

When the variable s changing at a steady rate, this can be thought of asapproximately + 1/2 (annual ate of increase). Thus he coefficient n theutilization rate in this equation can be thought of as approximately ne-half the elasticity of the variable with respect o the utilization rate, al-though the precise ag structure s rather complicated.

In addition o the industries or which statistics are reported n Table 6,it was possible o estimate price equations or six other wo-digit ndustries-tobacco, apparel, umber, furniture, eather, and fabricated metals-

using ust Wharton utilization data. Judged by t-statistics reater han 2.0,utilization ates were successful ariables n all cases but apparel.Besides he equation et forth n Table 6, two alternative orms of price

equations were estimated or each of the industries. Since the level ofutilization rates s used in the Table 6 equations, he estimates mply thatin a steady state in which the utilization rate was unchanged, he rate ofprice increase would be unchanged. Since the rates of change of inputprices and of wages n the industry are measured eparately y the other

explanatory ariables n the equation, his implication s implausible. Oneshould not expect margins o keep expanding or contracting ndefinitely.Of course, one would not expect such a steady state to prevail. Expandingmargins would be expected to induce firms to expand capacity faster,thereby reducing operating ates. The equations reported n Table 6 aretoo simple o capture all such effects. One might want to look for an in-dependent ffecton prices rom changes n utilization ather han from tslevel. Quite apart rom this argument, n industries whose pricing s char-acterized by market-clearing ehavior, one would expect the change inutilization, ather han its level, to explain price movements.

Equations ncluding he change in utilization were estimated or eachindustry and with each measure of utilization. Of the industries hown n


29/42

George L. Perry 729

Table 6, this variation was successful in textiles and petroleum; it also suc-ceeded in the lumber and leather industries, for which only Wharton utili-zation data were available.

Alternative equations were also estimated using a nonlinear form of theutilization rate, 1/(1.2 - U), where the U term was a distributed lag asbefore. In the denominator of this expression, 1.2 is used to avoid too muchnonlinearity and the explosion of the term to infinity as utilization ratesoccasionally reached 1.0. It seems likely that utilization effects on pricesare nonlinear, but these equations were virtually indistinguishable fromthose reported in Table 6.

The coefficients for two industries deserve special comment. Utilizationhad a significant negative effect on prices for motor vehicles, using all threemeasures, and for tobacco, for which only a Wharton estimate is available.It seems sensible to interpret this result as evidence of pricing based on atarget rate of return in these industries.16 It is much less likely that thesignificant negative coefficient on food estimated using the McGraw-Hillcapacity survey can be interpreted in this way, since this industry is not asconcentrated as autos and tobacco. While the equation reported in Table 6

cannot be considered an optimal pricing equation, it does seem to capturethe importance of utilization rates once costs have been accounted for, anddoes provide a fairly straightforward comparison of the utilization ratemeasures. On the basis of these results, a price forecaster would want topick and choose among the alternative measures of utilization. No doubt,the results for any one index or for any one industry could be improvedwith a more elaborate specification of the price equation. But even with thesimple form used here, all three of the measures do well enough to be taken

seriously.Why does the Wharton index do as well as any in explaining prices when

it failed in explaining capacity growth and was not quite up to its competi-tors in explaining investment? An explanation seems to lie in the fact thatmost of the price increase in the period 1954:4 through 1971:2 occurredduring the years 1966-71. As Table 3 showed, manufacturing operatingrates by all four measures are highly correlated with each other in thisinterval. Wharton's jump to a new plateau of operating rates relative to the

16. Richard Benson of Harvard University, who estimated some of the price equa-tions for this paper, also reports this result for autos and tobacco in equations that in-clude profit rates as explanatory variables or prices.


30/42


other measures n 1966does not interfere with its predictive bility n priceequations.

Operating Rate Levels

How tight has capacity utilization been n recent quarters? efore earch-ing for the elusive answer o that question, ome udgments must be madeabout the operating ates shown by the different measures. At a disaggre-gated level, the candidates are the two McGraw-Hill measures and theWharton ndex.

Several of the utilization rate series calculated rom the McGraw-Hillcapacity survey exhibit marked time trends over most of the 1954-72period. In some industries, he trends were so pronounced s to swampany cyclical variation n operating ates: in food, utilization declined nall but two years of the 1955-72 nterval; n chemicals, t rose n all but oneyear of the 1956-68 nterval; n rubber, t rose in all but one year of the1955-69 nterval.

These cases, and a few others hat are not so conspicuous, ast doubt onthe reliability f the capacity urvey or measuring he level of utilizationdespite ts usefulness or other purposes. t is significant hat two principalusers of this survey benchmark t periodically o utilization ates from theMcGraw-Hill tilization urvey. As noted earlier, he FRB capacity ndexis constructed in part) by adjusting he capacity urvey o match he trendof capacity growth mplied n the utilization survey. McGraw-Hill tselfderives a special series of monthly utilization ates by a method hat links

capacity rowth stimates rom he capacity urvey o utilization ates romthe utilization survey.17 he particular way in which these estimates arelinked does not equate utilization rom the two sources every year; but itkeeps the capacity survey estimate from wandering very far over anyperiod of time.

The equation results reported n Tables 4, 5, and 6 show that operatingrates from the capacity urveycan be useful predictors. n such equations,exponential rifts n the operating ate ndex can be compensated or in theestimated onstant erm of the regression. But the level of operating ates

17. See, for example, he bulletin, McGraw-Hill Measure of the Industrial OperatingRate (McGraw-Hill Publications Company, June 1972; processed).


31/42

George L. Perry 731

derived rom this survey cannot be relied on for assessing he currentsituation.

That eaves he Wharton ndex and the McGraw-Hill tilization urvey.Table 7 compares operating ates recorded by these two measures or in-dividual ndustries. t shows the peak rates achieved during he first halfof 1973and the difference etween hese rates and the peaks achieved nthe 1968-69 and 1966 expansionary eriods.

WHARTON

As a result of the considerable ise in production hat had occurredthrough mid-1973, most industries n the Wharton measure howed op-erating ates at 100 at that time. It would be comforting o believe n suchan apparently well-balanced xpansion, but no other evidence supportssuch a view. The table also shows hat by Wharton data, operating ates n1973have been above 1966 peaks in most industries. Yet there is wide-spread agreement hat capacity was being utilized very intensely n mostindustries during at least part of 1966.18These facts, revealed n Table 7,reflect he two basic weaknesses f the Wharton methodology: ts inabilityto distinguish ny difference n the intensity of utilization achieved at dif-ferent yclicalpeaks; and ts need to wait on a subsequent eak before set-tling on what operating rates have been-even by its own definitions-during n expansion. f the present xpansion were o continue at an above-average ace for some time, the current apacity stimates n the Whartonindex would be revised upward and the estimates of recent operating ateswouldbe reduced.

MCGRAW-HILL UTILIZATION

All in all, the McGraw-Hill tilization urvey eems he most believableof the available measures. Unlike the capacity urvey, t can be expected obe reasonably ree of drift over time. A priori, one would expect thatchanges n obsolescence of facilities, n their capital-labor atios, or inother characteristics f production echniques hould be accounted or by

18. This is true even though cyclical peaks were not recorded n every industry hatyear; otherwise Table 7 would never indicate a 1966 peak operating rate below a 1973peak.


32/42

X~~~~~~~~~~~~~~~~~~~ 1% O >x tO o eof) oo=

2

S > b i q I X n t

to _ ON m Cm 09 ..Iab ~~~~~~~~~~~

t> ? x

~~~~t N oN oo- 0 o oT 00 C

CX en k > 8 mmm~cO 0 0 Od 3 X

ti ;l -| W,|2iz|OIOnNXN^b.-.*****

ii-0r=

; 2 ,@?C| iS | ? O O s ? v > g s xc >0 ch O O > O i Oo a0

*^ r; , ZO0* *C* . . . . . ) . 0? .

PC Ct-?; W;5o *P

capacity in manufac

Documents