This PDF is a selection from an out-of-print volume from the National Bureau of Economic Research

Volume Title: Factors in Business Investment

Volume Author/Editor: Robert Eisner

Volume Publisher: NBER

Volume ISBN: 0-88410-484-2

Volume URL: http://www.nber.org/books/eisn78-1

Publication Date: 1978

Chapter Title: Components of Capital Expenditures—Replacement and Modernization versus Expansion

Chapter Author: Robert Eisner

Chapter URL: http://www.nber.org/chapters/c3845

Chapter pages in book: (p. 175 - 188)

Chapter Eight

Components of CapitalExpenditures—Replacement andModernization versusExpansion

INTRODUCTION AND DESCRIPTION OF DATA

Probably more than half of capital expenditures involve, inone sense or another, the replacement of existing stock.The timing and determinants of replacement expenditures

have given rise to a host of competing hypotheses, some of which arelisted below.

1. Replacement expenditures are a fairly constant proportion ofcapital.

2. Replacement expenditures substitute for expansion expenditures,thus stabilizing the annual rate of investment, falling whenexpansion increases and rising when expansion decreases.

3. They are closely tied or essentially equal to depreciation charges.4. They vary with the current rate of profit or flow of funds.5. They are positively related to the age of capital stock.

The McGraw-Hill capital expenditure survey data and collateralstatistics offer a unique opportunity to test these and relatedhypotheses. Feldstein and Foot (1971) utilized McGraw-Hill aggrega-tive reports, along with a series from the Department of Commerceon planned capital expenditures and from the Federal Trade Com-mission and Securities Exchange Commission on flow of funds, in an

Note: An earlier version of this paper was presented at the Second WorldCongress of the Econometric Society in Cambridge, England, in September1970.

175

176 Factors in Business Investment.

whereanalysis of replacement expenditures. This chapter offers a partlyparallel analysis of expansion as well as replacement expenditures on 1the basis of individual firm data.

Key to the analysis is a question that has been included in theMcGraw-Hill spring surveys in the years 1952 through 1955 and from1957 to date: "Of the total amount you now plan to invest in newplants and equipment in [the current year] how much is for: Kexpansion

_____%;

replacement and modernization

_____%?"

Ap-plying the indicated proportions to anticipated and actual capital Dexpenditures has resulted in estimates of expenditures for replace- pment and modernization as well as expenditures for expansion. Inthe case of actual expenditures, these estimates related to from 112 Rto 254 firms in each of the fourteen years from 1954 to 1968,excluding 1956.' Estimates of anticipated expenditures were avail-

upable for approximately the same firms.The basic data are as follows:2 and

Intervals forVariable Utilized Observations Superscripts e a

= [0.6, 0) zation, respectianticipated and

= [0.6, 0)indicate years (

= (Je/J)_l [1, 0] =

= [0.4, 0) and = (1 —

[0.7, 0.4] are "actual" astures.4

= [1, 0] was denol

4The McGraw-I[1.3, 0.3] tions of

all years except 19* were included or[0.7, 0.6] anticipations with

expenditures.'Small numbers of observations, forty-nine for 1952 and sixty-seven for Tables 84, 8-2,

1953, were eliminated from the final analysis when 1967 and 1968 data became total expendituresavailable because the capacity of our regression program was limited to a total of for expansionfourteen years. estimates of "àcti

modernization).2Srackets again indicate closed intervals, parentheses, semiopen intervals, expenditures do niexcluding lower bounds. Some 6 percent of observations were rejected because is inexact. We haone or more of the variables contained "extreme values," outside the indicated anticipated expendintervals. Elimination of extreme values or "outliers" in these individual firmdata has seemed prudent in order to minimize the possibility of substantialimpact due to reporting or processing errors and/or extremely low denominators

jfor variables in ratio form.

Components of Capital Expenditures 177

whereoffers a partly

expenditures Ofl capital expenditures of the year tft = capital expenditure anticipations for the yearincluded in the t+ i

1955 and from t + 1(presumably held at the end of the year t and

to invest in new reported in the spring survey of the year t + 1)

r much is for: K = gross fixed assets

_____%?"

Ap-depreciation chargesdactual capital

res for replace- P = profits after taxesr expansion. In

ted to from 112 R depreciation reserves1954 to 1968, U" = actual utilization of capacity

tures were avail-= preferred utilization of capacity

and L1s_1 = expected and ex post relative sales changes, respec-tively.3

Superscripts e and r denote expansion or replacement and moderni-zation, respectively. Time superscripts indicate that variables areanticipated and reveal the year of anticipations, and time subscriptsindicate years (or ends of years) to which variables apply.

*t—1 *

i7 = = (1 =

rt *and = (1

are "actual" and anticipated expansion and replacement expendi-tures.4

was denoted in Eisner (1967).4The McGraw-Hill surveys have included questions as to anticipated propor-

tions of expenditures for expansion and for replacement and modernization inall years except 1956, but questions as to the actual proportions, viewed ex post,were included only irregularly. Feldstein and Foot (1971) matched these6] anticipations with the Department of Commerce series for anticipated capitalexpenditures.

rid sixty-seven for Tables 8-1, 8-2, and 8-3 relate to and the products of reported actualI 1968 data became total expenditures divided by gross fixed assets and the proportions anticipatedlimited to a total of for expansion and for replacement and modernization. These are taken as

estimates of "actual" expenditures for expansion and for replacement (andmodernization). To the extent that discrepancies between actual and anticipatedsemiopen intervals, expenditures do not fall evenly on the two anticipated components our measurere rejected because is inexact. We have, however, also conducted the analysis with the data foritside the indicated anticipated expendituresLese individual firm

ility of substantial .et ty low denominators = et+i and = (1 q+1I

178 Factors in Business Investment

Capital expenditures and capital expenditure anticipations, thecapacity variables, expected sales changes, and the expansion-replace-ment ratios are taken from the McGraw-Hill surveys. The othervariables are from financial reports, generally as recorded in Moody's.All flow variables (I, sales, and P) except depreciation chargesare price-deflated, with indexes set at 1.00 in 1954. The stockvariables, gross fixed assets and depreciation reserves are not price-deflated.5

The body of data available proved sufficient to generate up to2,692 individual firm observations in six broad industry groups, asindicated in Table 8-lB. The bulk of the observations, however, wasin manufacturing.

RELATIVE STABILITY OF REPLACEMENTVERSUS EXPANSION INVESTMENT

Table 8-1 offers a variety of evidence on the relatively greaterstability of replacement expenditures as a proportion of gross fixedassets. From a quick visual inspection of the table's section A, it isclear that reported expenditures for replacement and modernizationwere not only higher than those for expansion in every year exceptone (1957), averaging some 60 percent of total expenditures, butalso markedly more stable. The standard deviation of the annualmean ratios of replacement expenditures to previous gross fixedassets was 0.0056, as against 0.01 16 for the corresponding figures forexpansion investment, and the coefficients of variation (standarddeviation divided by mean) shown in section B were 0.1241 for theformer, compared with 0.3762 for the latter. This comparison holdsup at both the industry and individual firm levels. Replacementand results were not substantially different from those for the "actual"expenditures. In Table 8.4, where we seek to isolate and compare determinantsof expansion and replacement expenditures, we report results for the anticipatedexpenditures, which should facilitate comparison with the Feldstein and Footanalysis of determinants of replacement expenditures.

Table 8-1. CarExpansion, 195

Year Replacement(t) (Ic)

1968 .04011967 .04491966 .04931965 .04861964 .04821963 .04211962 .0418

(1)

ObservationsTime Series

AggregatelndustriesbFinns

Primary metalsMetalworkingChemical processiAll otherMiningPetroleumAll industriesC

Cross SectionsAggregateIndustriesFirms, within indus

tThe use of undeflated gross fixed assets raises some problems. In principle, ameasure of net capital stock in constant prices, corresponding to current realcapacity, would be better. Measures of price-deflated gross fixed assets obtainedby utilizing ratios of accumulated previous deflated and undeflated capitalexpenditures were employed in other work (Eisner, 1967, pp. 371, 384-86), butdid not appear to affect the results sufficiently to warrant the substantialconsequent loss in observations (due to lack of full information on previouscapital expenditures). In a crude way, the failure to depreciate for decreasingcapacity or efficiency with age and the failure to appreciate for rising prices maybe taken as compensating errors, so that the gross fixed assets measure maycome as close to representing real capital stock as any other imperfect measurethat we might readily employ. We may further note the finding by Feldstein andFoot (1971, pp. 53, 55, and footnote 22) that estimates of the relations withwhich we are concerned prove insensitive to the measure of capital stock,including the substitution of a net capital series for the gross capital figures thatthey used.

aSee Chapter 1 forbN0 observations weone or more years inwere no observationscFirm time series snumber of total obselNote: Table Mi-b a

expenditures vaiaggregates, in arfirms in eachavailable.

Components of Capital Expenditures 179

latively greatern of gross fixedsection A, it is

I modernizationery year exceptpenditures, buti of the annualous gross fixediding figures foration (standard0.1241 for the

Dmparison holdss. Replacementfor the "actual"

ipare determinantsfor the anticipatedFeldstein and Foot

ems. In principle, aing to current real,ted assets obtainedundeflated capital• 371, 384-86), butList the substantialaation on previousiate for decreasing

or rising prices mayssets measure mayimperfect measure

ig by Feldstein andthe relations withof capital stock,

capital figures that

Means, All YearsStandard Deviations ofAnnual Mean Ratios

ticipatioflS, the)ansion-replace-eys. The otherled in Moody's.

eciation charges)54. The stocks are not price-

generate up toustry groups, as.s, however, was

Table 8-1. Capital Expenditures for Replacement and Modernization and forExpansion, 1954-1955, 1957-1968, as Ratios of Previous Gross Fixed Assets

A. Mean Ratios andNumber of Firms, by Year

Year

(I)

Replacement(if)

Expansion Numberof

FirmsYear

(t)Replacement Expansion

(if)

Numberof

Firms

1968196719661965196419631962

.0401

.0449

.0493

.0486

.0482

.0421

.04 18

.033 7

.0363

.0433

.0341

.023 7

.0 189

.0220

112 1961152 1960199 1959169 1958203 1957193 1955201 1954

.0385

.0434

.0411

.0397

.0493

.0566

.0558

.0 183

.0280

.02 15

.0249

.0584

.0437

.0405

244254226244175153167

.0452 .0307 2692

.0056 .0116

B. Standard Deviations, Means, and Coefficients ofVariation (a/Mean)0

(1) (2) (3) (4) (5) (6) (7) (8)

Observations. n a Mean a/Mean a Mean a/Mean

Time SeriesAggregateIndustriesb

14 .0056 .045283 .0081 .0452

.1241

.1788.0116.0153

.0307

.0307.3762.4984

Firms .

Primary metals 231 .0226 .0352 .6428 .0625 .0323 1.9353Metalworking 1084 .0334 .0528 .63 21 .0379 .0304 1.2458Chemical processing 568 .0175 .0332 .5280 .0341 .0383 .8908All other manufacturing 628 .0267 .0464 .5743 .0359 .025 2 1.4237Mining 34 .0325 .0395 .8240 .0633 .0304 2.0867Petroleum 80 .0264 .0546 .4838 .0327 .0186 1.7543All industriesC 2625 .0279 .0454 .6 140 .0398 .0307 1.2979

Cross SectionsAggregate 6 .0088 .0452 .1956 .0052 .0307 .1702Industries 83 .0108 .0452 .2385 .0118 .0307 .3827Firms, within industries 2692 .0322 .0452 .7 125 .0416 .0307 1.3544

aSee Chapter 1 for statement of the various deviations underlying the calculation of a.bNo observations were available in several regressions and in these summary statistics forone or more years in the mining and petroleum industries. For Tables 8-1, 8-2, and 8-3 therewere no observations in petroleum for 1968.CFirm time series statistics exclude firms with only one observation—hence the lessernumber of total observations.Note: Table Mi-jo appears only in microfiche.

expenditures vary less over time than expansion expenditures in theaggregates, in annual means for each industry, and within individualfirms in each of the six industry groups for which data wereavailable.

180 Factors in Business Investment

Evidence showing expenditures for replacement and moderniza-tion to be more than half of total capital expenditures and to berelatively stable should not be construed as a guarantee againstcyclical fluctuations. For one thing, the highly variable expansionexpenditures still constitute 40 percent of total expenditures. Sec-ond, variation in replacement expenditures is only relatively small.The 0.1241 coefficient of variation in the aggregative data impliesthat, in about one-third of the years, the ratio of replacement tocapital will vary by as much as one-eighth from its mean. As observedby Feldstein and Foot (1971) and argued earlier by Eisner and Nadiri(1968), this contradicts the hypothesis of a constant ratio ofreplacement to capital stock maintained by Jorgenson and hisassociates in their work on investment (Jorgenson [1963] andJorgenson and Stephenson [1967a and b], for example)—at leastinsofar as McGraw-Hill respondents can be believed. It is true thatconditions for a strict test of the Jorgenson hypothesis are not met:the gross fixed assets data cannot be taken as a measure of capitalstock necessarily consistent over time with the path of gross capitalexpenditures and replacement. However, year-to-year variation inreplacement, both in our data here and in the aggregates reported byFeldstein and Foot, is clearly greater than could be accounted for byany corrections of the relatively slow-moving capital stock series.6

A critical question is whether, in years of slackening demand forexpansion, firms fill in at least part of the slack by drawing on abacklog of needs for replacement and modernization. If this weretrue, it would suggest a substantial source of stability for capitalexpenditures as a whole. However, the evidence points the otherway. As indicated in Table 8.2, in the pooled individual firm timeseries, there is no correlation between the ratios of replacement andexpansion investment to gross fixed assets, but the industry timeseries shows a distinct positive relation. Within each industry,replacement investment moved in the same direction as expansioninvestment, with about one-fourth of its amplitude. Pooling all theobservations in the weighted aggregate time series, we find replace-ment investment varying in the same direction as expansion invest-ment, with about one-third of its amplitude and with a correctedcoefficient of determination of 0.48.

The failure of replacement and modernization expenditures tocompensate for the volatility of expansion expenditures is confirmedin Table 8-3. We note there that total investment is positively relatedto the proportion of investment designated for expansion. This

6This vitiates the objection by Jorgenson (1971, p. 1140) to the work ofFeldstein and Foot.

Table 8-2. Rep

(1)

Variable orStatistic

Constant

.eJr,2

n(—6)

r.d.f.

positive relatioiaggregate levels.expenditures f(variance in aniexpenditures fo:of that for exptures on the prosimply the negaTable 8-3.DETERMINANAND EXPANSI

Evidence enabli:tures for replacis harder toanticipated repi

•xtit+1

+

7As indicated athe actual expezl(Results, not subsUthe paper presentexpenditures herereported by Feldst

and moderniza-ures and to bearantee againstiable expansionpenditures. Sec-relatively small.ive data impliesreplacement toan. As observedisner and Nadirinstant ratio ofgenson and his)n [19631 andaniple)—at least1. It is true thatesis are not met:easure of capital

of gross capitalear variation inates reported bytccounted for bystock series.6ning demand for

drawing on a'on. If this werebiity for capitalpoints the otherividual firm timereplacement andhe industry timei each industry,ion as expansione. Pooling all thewe find replace-

expansion invest-with a corrected

• expenditures to;ures is confirmedpositively relatedexpansion. This

40) to the work of

Components of Capital Expenditures 181

Table 8-2. Replacement Expenditures as a Function of Expansion Expenditures

(1)

.Variable orStatistic

.

(2) (3)

Coefficients and Standard Errors

(4)

Time Series

Firm Industries Aggregate

Constant .0450(.007)

.0379(.0018)

.0344(.0032)

t .0126(.0148)

.2375(.0542)

.3496(.0972)

f2 —.0001 .1909 .4790n(—6) 2625 83 14

r.d.f. 2247 76 12

positive relation is highly significant at the firm, industry, andaggregate levels. At the aggregate level, indeed, the proportion ofexpenditures for expansion explains almost 80 percent of thevariance in annual capital expenditures. Since the proportion ofexpenditures for replacement and modernization is the complementof that for expansion, the regression coefficients of total expendi-tures on the proportion for replacement and modernization would besimply the negatives of those for the expansion proportion shown inTable 8-3.DETERMINANTS OF ANTICIPATED REPLACEMENTAND EXPANSION EXPENDITURES

Evidence enabling us to discriminate among determinants of expendi-tures for replacement and modernization versus those for expansionis harder to come by. Two basic parallel relations were estimated foranticipated replacement and expansion expenditures.7 These were

= ÷ + (8.1)

i=0

7As indicated above (footnote 4), a similar analysis has been conducted withthe actual expenditures which are the subject of Tables 8-1, 8-2, and 8-3.Results, not substantially different from those reported below, are described inthe paper presented at the World Econometric Congress. Use of anticipatedexpenditures here will facilitate comparison with the findings subsequentlyreported by Feldstein and Foot.

k

182 Factors in Business investment

Table 8-3. Total Investment as a Function of the Proportion of InvestmentPlanned for Expansion, Firm, Industry, and Aggregate Time Series

= b0 + b1 + Vt

(1)

.Variable orStatistic

(2)

Regression Coe(3)

fficients and Sta(4)

ndard ErrorsFinn Industry Aggregate

Constant .0483 .0156 .0012

et_lt

(.0016)

.0845(.0042)

(.0070)

.1833(.0206)

(.0115)

.2268(.0344)

f2 .1542 .5034 .7658n(—6) 2625 83 14

r.d.f. 2247 76 12Mean .3283 .3292 .3292Mean .0761 .0759 .0759Elasticity at Means .3645 .7950 .983 7

Elasticity from Logarithmicrelation with .2767

.

.6658 1.2022

where x = r for "replacement and modernization" in one case andx = e for "expansion" in the other.

There is some evidence in the firm time series shown in Table 8-4that anticipated expenditures for expansion are positively related tosales changes and utilization of capacity, while replacement expendi-tures are related to previous profits. The sum of the sales changecoefficients for expansion expenditures is distinctly higher than thecorresponding sum for replacement expenditures: the sum of esti-mated coefficients of depreciation charges and profits, d + p, how-ever, is smaller for anticipated expansion expenditures than foranticipated replacement expenditures.8 The depreciation reservevariable, rd, introduced as a proxy for age of capital, yields only avery low positive coefficient for replacement and modernizationexpenditures and shows a slightly higher value for expansion expend-itures. This may relate to its imperfect character as a proxy,reflecting, for example, the varying mix of plant and equipment or,more generally, longer- and shorter-lived capital in the investment of

8When depreciation charges and profits were introduced separately in theregressions, coefficients of the depreciation variable seemed erratic, with highstandard errors. This may have related to the very low time series variance ofd(c < 0.01, as against almost 0.05 for profits variables).

previous years. Irecent investmezrapid post-1962high even if aged

In the indusvariables in expbecause of a lachanges; the streplacement regmately zero. 'Ipositive and of sThe sum of thipositive, and mctime negative in

As indicatedoverall regressio:transitory compestimates of per"industry overasums of squaresmean, of

othe expansio

replacement ancof capacity coelregression but nConversely, theprofits is a signilzero in the expratio is alsoclose to zero in

COMPARISON

It is useful toFeldstein and I

replacement thiemphasize areplacement exIavailability of ft

Their negatiV

9Feldstein and

Components of Capital Expenditures 183

ion of InvestmentSeries

(4)

rd Standard ErrorrAggregate

.00 12(.0115)

.2268(.0344).765 8

14

12 +

.3292

.0759

.9837

1.2022

in one case and

own in Table 8-4itively related tocement expendi-the sales changehigher than the

the sum of esti-d +p, how-

ditures than forreserve

;al, yields only aI modernizationpansion expend-er as a proxy,d equipment or,le investment ofI separately in theI erratic, with highe series variance of

previous years. in a period of high expansion expenditures, if muchrecent investment has been for short-lived equipment with relativelyrapid post-1962 depreciation rates, depreciation reserves may provehigh even if age of capital is low.

In the industry time series, the decisive role of sales changevariables in expansion investment is all the more apparent, possiblybecause of a larger permanent component in the variance of saleschanges; the sum of sales change coefficients is 0.3275. In thereplacement regression, sales. change coefficients sum to approxi-mately zero. The capacity utilization coefficients, however, arepositive and of about the same (small) magnitude in both regressions.The sum of the coefficients of profits plus depreciation is againpositive, and more substantial, in the replacement regression and thistime negative in the expansion relation.

As indicated earlier (particularly in Chapter 4), cross sections andoverall regressions of grouped data, containing smaller proportions oftransitory components and transitory variations, may yield betterestimates of permanent, structural relations. This is confirmed by the"industry overall" regressions (based upon appropriately weightedsums of squares and cross products of deviations, around the overallmean, of observations that are themselves means of the individualfirm observations of an industry for a year, shown in Table 8-4).First, the sum of sales change coefficients is a fairly significant 0.32in the expansion regression but virtually zero in the regression forreplacement and modernization expenditures. The sum of utilizationof capacity coefficients is also significantly positive in the expansionregression but not significant (and slightly negative) for replacement.Conversely, the sum of coefficients for depreciation charges andprofits is a significant 0.15 in the replacement relation while virtuallyzero in the expansion relation. And now, the depreciation reserveratio is also significantly positive in the replacement relation andclose to zero in the regression for expansion expenditures.

COMPARISON WITH FELDSTEIN-FOOT RESULTS

It is useful to compare our analysis more explicitly with that ofFeldstein and Foot. They join us in rejecting "the proportionalreplacement theory as a description of short-run behavior," butemphasize a "negative short-run interdependence of expansion andreplacement expenditures" along with "the importance of internalavailability of funds."9

Their negative relation between replacement and expansion ex-9Feldstein and Foot (1971, p. 57).

Tab

le 8

-4.

Ant

icip

ated

Exp

ansi

on In

vest

men

t ver

sus

Rep

lace

men

t and

Mod

erni

zatio

n In

vest

men

t Fun

ctio

ns

=+

(d +

p1)+

14(

d1b3

r"+

Var

iabl

e

Reg

ress

ion

Coe

ffic

ient

s and

Sta

ndar

d Er

rors

Firm

and

indu

stry

tim

e se

ries"

indu

stry

ove

rall

regr

essi

ons

Firm

sin

dust

ries

Expa

nsio

nR

epla

cem

ent

Expa

nsio

nR

epla

cem

ent

Expa

nsio

nR

epla

cem

ent

orSt

atis

tic,e

t,rt

,et

"ti-i

'1.

rt "ti-I

',.e

t,.r

t

Con

stan

t—

.047

6.0

053

—.0

645

—.0

670

—.0

745

—.0

204

(.01

33)

(.01

01)

(.03

87)

(.02

59)

(.03

41)

(.03

34)

d*-

tt

.045

8(.0

223)

.057

6(.0

169)

—.1

854

(.115

2).0

630

(.077

1)—

.105

6(.1

145)

.009

4(.1

121)

d*1

t—1

—.0

198

(.021

5).0

299

(.016

3).0

883

(.110

9).0

875

(.074

3).1

005

(.107

3).1

417

(.105

1)

t.0

459

(.014

5).0

134

(.010

9).0

340

(.048

0).0

335

(.032

2).0

110

(.043

9).

.117

3(.0

430)

r.0

255

(.010

7).0

178

(.008

1).0

474

(.036

3).0

470

(.024

3).0

575

(.033

6)—

.011

3(.0

329)

1t—

.019

5(.0

106)

.004

3(.0

083)

.028

8(.0

294)

.034

8(.0

197)

.028

8(.0

262)

—.0

140

(.025

6)

t.0

368

(.014

2).0

133

(.010

7).1

428

(.048

8).0

865

(.032

7)•

.142

4(.0

436)

.051

1(.0

427)

5co

effic

ient

s/=

0r-

J.0

687

'(.03

25)

—.0

276

(.024

6).1

848

(.078

1)—

.148

7(.0

523)

.176

7(.0

613)

—.0

185

(.060

0)

-'4

Sum

ofuC

coe

ffic

ient

s

.105

5(.0

372)

.045

0(.0

138)

—.0

143

(.028

2)

.022

1(.0

104)

.327

5(.0

897)

.076

2(.0

441)

—.0

622

(.060

1)

.081

8(.0

296)

.319

1(.0

726)

.086

3(.0

317)

.032

6(.0

7 11

)

—.0

252

(.031

1)

Sum

ofd

* +P

*co

effic

ient

s.

,

.026

0(.0

255)

1054

.087

5(.0

193)

1054

—.0

971

(.083

3) 58

.150

5(.0

558) 58

—.0

051

(.036

6) 58•,

.151

1(.0

358) 58

n1O

329T

ut_I

.019

5.0

043

.028

8.0

348

.028

8—

.014

0(.0

106)

(.008

3)(.0

294)

(.019

7)(.0

262)

(.025

6).0

368

.013

3.1

428

.086

5.1

424

.051

1(.0

142)

(.010

7)(.0

488)

(.032

7)(.0

436)

(.042

7)5 1

coef

ficie

nts

.068

7—

.027

6.1

848

—.1

487

.176

7—

.018

51=

0(.

0325

)(.0

246)

(.078

1)(.0

523)

(.061

3)(.0

600)

L

Sum

and

.1055

—.0143

.3275

—0622

.3191

.0326

(.037

2)(.0

282)

(.089

7)(.0

601)

(.072

6)(.0

711)

Sum

of u

cco

effic

ient

s.0

450

.022

1.0

762

.081

8.0

863

—.0

252

(.013

8)(.0

104)

(.044

1)(.0

296)

(.031

7)(.0

311)

Sum

ofd

*+p*

.026

0.0

875

—.0

971

.150

5—

.005

1.1

511

coef

ficie

nts

(.025

5)(.0

193)

(.083

3)(.0

558)

(.036

6)(.0

358)

n(—

58)

1054

1054

5858

5858

r.d.f.

840

840

40

40

45

45

.057

7.0511

.3547

.3183

.3285

.3464

F5.35

4.82

3.38

3.02

3.32

3.52

F01

2.21

2.21

2.66

2.66

2.61

2.61

aFor r

=19

58to 1968 only; depreciation data

wer

e no

t ava

ilabl

e fo

r obs

erva

tions

of y

ears

prio

r to

1958

. Onl

y fif

ty-e

ight

indu

stry

-yea

rob

serv

atio

ns w

ere

avai

labl

e be

caus

e th

ere

wer

e no

obs

erva

tions

in m

inin

g fo

r t=

195

8,19

59, 1

960,

196

6, 1

967

and

1968

and

, sim

ilarly

, non

e fo

rpe

trole

um fo

r t=

196

7an

d 19

68.

186 Factors in Business Investment

penditures turns up in regressions where the flow of funds andutilization of capacity enter as other independent variables withpositive coefficients. Reestimates of the Feldstein-Foot relation withour data yield results somewhat different from theirs, as shown inTable 8-5. Taking the aggregate relations, we find that the positivecoefficient of planned expansion expenditures is brought close tozero when utilization of capacity and flow of funds variables areincluded, but that it does not turn sharply negative as in theFeldstein-Foot estimates. And in this case, our industry and firmregressions are roughly consistent with our aggregate regression.

The Feldstein-Foot relation indicates that higher utilization ofcapacity, as well as a greater flow of funds, brings on greaterreplacement expenditures, except to the extent that they, or otherforces, also bring on higher expansion expenditures. That expansionexpenditures might relate negatively to replacement expenditures isplausible, for expanding firms would be more likely to retain allexisting capacity. Also, Feldstein and Foot point out that while the

Table 8-5. Comparison with Feldstein-Foot Time Series Estimates, PlannedReplacement Expenditures:ft — i t't+l — —

(1)

.Variable

Statistic

(2) (3)

Regr(4) (5)

ession Coefficients an(6) (7)

d Standard Errors(8)

EisnerFeldstein-

FootAggregateFirms Industries Aggregate

Constant .044 .013(.001) (.005)

.041 .012(.002) (.014)

.040 .018(.003) (.021)

—.051(.022)

uCt — 019

— (.006)017

— (.016)015

— (.024)

104a

(.032)f(p*+d*)

I t — .097— (.011)

— .105— (.034)

— .083— (.061)

.216(.070)

.et .077 .034(.016) (.016)

.166 .057(.056) (.061)

.214 .055(.075) (.131)

—.337(.106)

n(—38) 1990 1990 81b 81b 14 14 13

r.d.f. 1682 1680 74 72 12 10 9

A) .013 .069 .094 .212 .355 .367 .85

aba, actual utilization of capacity, in the Feldstein-Foot relation.

bThe fourteen years of data for six industries generated only eighty-one industry observa-tions because there were no individual firm observations in mining for t = 1966 and 1967and in petroleum for t = 1967.

same factors irexpansion expeianother, as confithe increasing cexpansion andtion of a negatimay be related Iof replacementfrom 0.27 to 0footnote 25).

This, in turn,anticipations daexpansiondents. To the ement capital exrtures underlyin{would be asion and anticiphigher fraction iother. Thus, theminants would tcorrelation and ierror in our grosispurious positivexpenditures vaileast not sufficiein Table 8-lA, vsame in the(1954-1961).

The higher flmay reflect a fishould be moreof Commerce aMcGraw-Hill saicoefficient mayFoot use thewhile we utilizefirm's actual toobservations are

In any event,'°It should also

both our industryresponding, while li

of funds andht variables withLoot relation witheirs, as shown inthat the positivebrought close tonds variables aregative as in theidustry and firmregression.

utilization ofprings on greaterat they, or other

That expansiont expenditures isely to retain allut that while the

Estimates, Planned

(7) (8)

'rd Errors

Fe!dstein-

Foot

regate Aggregate

.018 —.051(.021) (.022)

.015(.024) (.032)

.083 .216(.061) (.070).055 —.337

(.131) (.106)14 13

10 9

.367 .85

-one industry observa-ror r = 1966 and 1967

Components of Capital Expenditures 187

same factors may contribute in part to both replacement andexpansion expenditures (causing them to be correlated with oneanother, as confirmed by their positive sample correlations and ours),the increasing cost of higher gross investment may tend to makeexpansion and replacement partial substitutes. The lack of confirma-tion of a negative coefficient of expansion investment in our datamay be related to our considerably higher positive simple correlationof replacement and expansion investment: 0.60 as compared withfrom 0.27 to 0.47 reported by Feldstein and Foot (1971, p. 54,footnote 25).

This, in turn, may stem from the fact that our capital expenditureanticipations data and anticipated proportions for replacement andexpansion involve the same coverage and, indeed, identical respon-dents. To the extent that part of the Feldstein Commerce Depart.ment capital expenditure anticipations are unrelated to the expendi-tures underlying the McGraw-Hill anticipated proportions, therewould be a negative relation between measured anticipated expan-sion and anticipated replacement: with a given (unrelated) total, ahigher fraction in one category must mean a lower fraction in theother. Thus, the simple positive correlation due to common deter-minants would be reduced, permitting more sharply negative partialcorrelation and regression coefficients. On the other hand, the likelyerror in our gross fixed assets measure of capital stock may produce aspurious positive correlation between expansion and replacementexpenditures variables; that this error in the common divisor is atleast not sufficient to create a common trend, however, can be seenin Table 8-lA, where mean values are found to be approximately thesame in the later years (1962-1968) as in the early years(1954-1961).

The higher flow of funds coefficient for the Feldstein-Foot datamay reflect a feedback of capital expenditures to income, whichshould be more conspicuous in the more comprehensive Departmentof Commerce and FTC-SEC aggregates than in our more modestMcGraw-Hill sample. The differences in the capacity utilizationcoefficient may relate to differences in the variable. Feldstein andFoot use the Federal Reserve Board index of capacity utilization,while we utilize McGraw-Hill responses to calculate the ratio for eachfirm's actual to "preferred" rate of capacity utilization; our aggregateobservations are annual means of these individual firm ratios.' °

In any event, however, there is danger of misconception in theshould also be noted that there are certain cross-sectional elements in

both our industry and aggregate time series because the samples of firmsresponding, while largely overlapping, are not identical from year to year.

188 Factors in Business Investment

Feldstein-Foot statement, "Expansion investment causes an offset-ting fall in replacement investment, supporting the view that firmspostpone replacement during periods of expansion investment andaccelerate replacement when there is less expansion investment."From the Feldstein-Foot data it would appear that, ceteris paribus,more expansion investment means less replacement investment. Butother things are not the same, and when there is more expansioninvestment there also tends to be more replacement investment. Thisis shown in the positive simple correlation between je and reportedby Feldstein and Foot, and is more sharply delineated in the simpleregression of on je that we have presented in Table 8-5.

SUMMARY AND CONCLUSIONS

1. Expenditures planned for replacement and modernization variedover time and, as observed by Feldstein and Foot, were not aconstant proportion of capital (in our case, gross fixed assets),although they were a much more constant proportion than wereexpenditures planned for expansion.

2. While varying less, replacement and modernization expenditureswere not a stabilizing substitute for expansion expenditures, butrather moved up and down with expansion expenditures.

3. Expenditures for expansion are clearly related to past and ex-pected sales changes (and to some extent utilization of capacity),particularly in cross sections and industry regressions in whichrandom or transitory components of individual firm variance overtime may cancel out.

4. Replacement and modernization expenditures, conversely, areusually more positively related to previous depreciation chargesand profits and, less certainly, to the depreciation reserve as apossible proxy for the age of capital.

Feldstein and Foot (1971, p. 54).

Our rsubstacapital

realizations. Inveefforts toof inventory saliexpenditures weof a set of sevelagged profits, aiin plant andsales, with a "hichange coefficiesections and geifirm regressions,sales that mighttions of the firstshifts in demand,

Coefficientstions were indusappeared that ivariance was

a reastaken to implyscale or declinir

189

C

A

This brief report of extensive statistical results enables us to offer thefollowing tentative conclusions.