Direct and Indirect Effects of Industrial Research and ... · PDF file361 Direct and Indirect...

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

Volume Title: New Developments in Productivity Measurement

Volume Author/Editor: John W. Kendrick and Beatrice N. Vaccara, eds.

Volume Publisher: University of Chicago Press

Volume ISBN: 0-226-43080-4

Volume URL: http://www.nber.org/books/kend80-1

Publication Date: 1980

Chapter Title: Direct and Indirect Effects of Industrial Research and Development on the Productivity Growth of Industries

Chapter Author: Nestor Terleckyj

Chapter URL: http://www.nber.org/chapters/c3917

Chapter pages in book: (p. 357 - 386)

III. The Effects of Research andDevelopment, Energy, andEnvironment on ProductivityGrowth

6 Direct and Indirect Effectsof Industrial Research andDevelopment on theProductivity Growthof IndustriesNestor E. Terleckyj

6.1 Introduction

In an earlier study by the author, rather high estimates were obtainedof the effeëts of industrial R and D on the rate of productivity growthof industries (Terleckyj 1974). Two kinds of such effects were identifiedusing the total factor productivity data compiled by John W. Kendrick(1973): direct effects in the industries in which the R and D is con-ducted, and indirect effects in the industries purchasing intermediate andcapital inputs from the industries conducting the R and D. Indications ofpresence of these effects were obtained for the privately financed R andD, but not for government financed R and D, and for the manufacturingindustries but not for the nonmanufacturing industries. Because humancapital was not included in earlier research, the resulting estimates ofproductivity returns to R and D might be inflated by the possible effectsof increases in the employment of human capital if such increases werecorrelated with increased use of R and D inputs.

In this paper these results are tested for independence from possibleeffects of increased use of human capital and other input characteristics,first by introduction of an explicit variable measuring increases in theuse of human capital by industry and then by repeating estimation ofthe R and D effects using the measures of total factor productivity

Nestor E. Terleckyj is with the National Planning Association.The author greatly appreciates the helpful suggestions and comments received

from Milton Moss, Neil McMullen, and B. I. Stone, and the help in obtainingdata given by Dale W. Jorgenson, John W. Kendrick, and Elizabeth Wehle. Thepresent revised version of this paper incorporates changes made in response tothe original comments by Steven Globerman, who discussed this paper at the con-ference. The research for this paper was supported by grant GSOC74—17574 fromthe National Science Foundation.

359

360 Nestor E. Terleckyj

growth prepared by Gollop and Jorgenson (1979) which are alreadyadjusted for the use of human capital, as well as for other characteristicsof input.

6.2 The Analytical Model

The theoretical model underlying the present analysis treats the re-search capital as a third input in addition to the labor and capital inputs.This model has been formulated by Griliches (1973). Its adaptation tothe present analysis has been discussed in earlier work by the author(Terleckyj 1974, pp. 3—8, 19). Hence, only a very brief discussion ofthe model will be provided here.

In this model the production function for time, t, is represented by:(1)where Q, L, K, and R are the output, and the inputs of labor, tangiblecapital, and R and D capital, respectively; p, and a are therespective elasticity parameters for the three inputs; A is a constantspecifying the level of output in the base year; and the parameter Arepresents the disembodied growth of productivity.

The productivity ratio at time t can then be expressed as a productof a component cumulative effects of disembodied techni-cal change and the stock of R and D capital raised to an exponentrepresenting the elasticity of output with respect to research capital:

(2) = = AeXt

Differentiating this equation with respect to time, one can show that therate of growth in productivity is the sum of the autonomous disembodiedcomponent and a component representing the product of the relativegrowth of research capital and the elasticity of output with respect tothat capital:

(3) p4=A+c4.Because the rate of growth of research capital usually cannot be

directly observed, an alternative formulation of equation (3) may beused, provided one assumes that the gross investment in R and D,i.e., the expenditure for R and D in the base year, also representsthe net R and D investment (i.e., that there is no depreciation of Rand D or that it can be ignored). One can then express the second termon the right-hand side of equation (3) by a product of the marginalproduct of capital, v, and the ratio of R and D investment to output, 1:(4) p=A+vi,(because v = dQ/dR, I = R/Q, and a = dQ/dR R/Q, aR/R = vi).

361 Direct and Indirect Effects of Industrial Research and Development

The present research results are based on the statistical estimates ofan expanded version of equation (4). The equation is expanded byintroducing successively different components offirst, by separating R and D conducted in industry according to itssource of financing—into the cost of research conducted with privatefunds of the industry and research conducted with the federal govern-ment funds—and then by estimating the R and D embodied in purchasesfrom other industries separately for the privately financed and the gov-ernment-financed R and D. The equation is also expanded by the intro-duction of variables other than R and D which have been found to becorrelated with productivity growth and the omission of which mightdistort estimates of the net effect of R and D on productivity. The gen-eral form of the equations estimated in this study can be stated asfollows:(5) p = a0 + aIX,. + b,l,.

Here a0 is the constant of regression representing (when normalized)the remaining unexplained residual growth. The as's are the regressioncoefficients of the variables which represent factors other than R andD, and the b,'s are the coefficients of the research intensity ratios, I,'s.

The coefficients b, are the estimates of the marginal products of therespective types of R and D capital. They also represent the produc-tivity rates of return on the different types of R and D expenditures.Because the costs of labor and of capital used in R and D are alreadyincluded in the input index used to estimate productivity, the regressioncoefficients b, measure the "excess" or additional rates of return toR and D, i.e., net of its cost. Thus, the statistical estimates of the pro-ductivity rates of return approximate the concept of "internal rates ofreturn." The accuracy of this statistical approximation actually achieveddepends among other things on the extent to which the cost and theeffects of same R and D projects are included in the data for the periodused. The period 1948—66 used in the present analysis covers 18 years.It should be sufficient to include a predominance of complete lifetimesof R and D projects and their productivity effects. The time series dataon costs and on private and social returns to innovations developed byMansfield and his associates in their case studies of specific innovationssupport this assumption (Mansfield Ct al. 1975).

6.3 Previous Estimates

In the following discussion, only the main results and the basic dataof previous research are summarized. The methods of estimation ofdata, various qualifications, and details regarding specific assumptionswere discussed in the publication in which these results were first given(Terleckyj 1974).


The basic data are shown in tables 6.1 and 6.2. Kendrick's (1973,pp. 78, 79) data for the rates of change in total factor productivity forthe period 1948—66 which were used as the dependent variable, areshown in table 6.1 together with the author's estimates of the four Rand D investment intensity ratios for the base year 1958. Estimates ofR and D embodied in purchased goods were derived by summing theR and D of the conducting industries, redistributed in proportion totheir sales as given in the 1958 Department of Commerce input-outputmatrices for intermediate and capital flows among industries. Table 6.2contains the data for the three non—R and D variables introduced tohold productivity effects of R and D constant. One of these variablesis the percent of sales to the private sector, which tends to have positiveeffects on productivity. Another is the unionization rate of the workforce of the industry. It was found by Kendrick (1973) to have signifi-cant negative correlation with productivity growth. It is also includedhere for a different year (1953 rather than 1958, because nonmanufac-turing data could not be obtained for 1958). The third variable is anindex of cyclical instability of output of the industry which in anotherstudy by the present author was found to have a negative effect on pro-ductivity (Terleckyj 1960).

The highlights of the results obtained are shown in table 6.3, whichcontains a series of estimates of equation (5) for twenty manufacturingindustries. The table follows the sequence of analysis of the R and Dvariables. First, the total ratio for all R and D conducted in the industrywas introduced. Its coefficient was not statistically significant at the 5%level. After dividing the total R and D into privately financed and gov-ernment-financed R and D, the estimated rate of productivity return toprivate R and D was highly significant and amounted to 37%, whilethe return estimated for the government-financed R and D was notsignificant (and numerically near zero). Based on this result, the ratioof government-financed R and D was omitted from further analyses,and the total R and D embodied in purchases from other industries wasintroduced. The coefficient for industry's own R and D continued to besignificant, but its magnitude dropped from 37 to 28%. The estimate 2of return to total imputed R and D was 45% and significant at the 5%level. External R and D was then divided into government-financedand privately financed components. This division of the embodied R andD resulted in an overall improvement in the fit of the equation andincreases in the significance of almost all coefficients but one (salesnot to government). The coefficient obtained for privately financedR and D embodied in purchased inputs is equivalent to a 78% rate ofproductivity return, while the coefficient estimated for the government-financed purchased R and D is near zero, negative, and not significant.

pT

able

6.1

Rat

esof

Gro

wth

in T

otal

Fac

tor P

rodu

ctiv

ity, 1

948-

66, a

nd th

e 19

58R

and

D V

alue

Add

ed R

atio

s for

Priv

atel

y Fi

-na

nced

R a

nd D

and

for G

over

nmen

t Fin

ance

d R

and

D C

ondu

cted

in th

e In

dusf

ry a

nd fo

r Priv

atel

y Fi

nanc

ed a

ndG

over

nmen

t Fin

ance

d R

and

D E

mbo

died

in P

urch

ased

Goo

ds, T

hirty

-thre

e In

dust

ries i

n th

e Pr

ivat

e D

omes

tic E

con-

omy

(R a

nd D

am

ount

s as p

erce

nt o

f val

ue a

dded

)

R a

nd D

Inte

nsity

Rat

ios,

1958

R a

ndD

Con

duct

ed in

R a

nd D

Em

bodi

ed in

Prod

uctiv

ityG

row

th R

ate,

Ann

ual

Indu

stry

Purc

hase

d G

oods

Priv

atel

yG

over

nmen

tPr

ivat

ely

Gov

ernm

ent

Ave

rage

Fina

nced

Fina

nced

Fina

nced

Fina

nced

Indu

stry

1948

—66

R a

nd D

R a

nd D

R a

nd D

R a

nd D

(arr

ange

d in

ord

er o

f des

cend

ing

prod

uctiv

ity g

row

th)

(per

cent

)(p

erce

nt)

(per

cent

)(p

erce

nt)

(per

cent

).

4Air

trans

porta

tion

*Coa

l min

ing

*Raj

I roa

dsC

hem

ical

s*E

lect

ric a

nd g

as u

tiliti

esTe

xtile

sR

ubbe

r pro

duct

s*C

omm

unic

atio

n ut

ilitie

sEl

ectri

cal m

achi

nery

Lum

ber p

rodu

cts

*Far

min

g*O

il an

d ga

s ext

ract

ion

Tran

spor

tatio

n eq

uipm

ent a

nd o

rdna

nce

Food

sPe

trole

um re

finin

gFu

rnitu

reIn

stru

men

ts

8.0

5.2

5.2

4.9

4.9

4.0

3.9

3.8

3.7

3.5

3.3

3.2

3.2

3.0

3.0

2.9

2.9

0 0 06.

8 .1 .4 1.1 .3 5.5 .2 .5 .2 6.9 .5

4.3 .2 4.2

0 0 01.

8 0 0 .4 .112

.0 0 0 018

.6 0 .2 03.

6

1.6 .3 .2 1.8 .3 2.1

2.3 .6 1.9 .3 .6 .1 3.8 .6 .9 .4 2.1

1.5 .1 .1 .8 .3 .1 .3 1.1

3.3 .1 .1 .1 3.8 .1 .3 .1 2.2

R a

nd D

Inte

nsity

Rat

ios,

1958

R a

nd D

Con

duct

ed in

R a

nd D

Em

bodi

ed in

Prod

uctiv

ityG

row

th R

ate,

Ann

ual

Indu

stry

Purc

hase

d G

oods

Priv

atel

yG

over

nmen

tPr

ivat

ely

Gov

ernm

ent

Ave

rage

Fina

nced

Fina

nced

Fina

nced

Fina

nced

Indu

stry

1948

—66

R a

nd D

R a

nd D

R a

nd D

R a

nd D

(arr

ange

d in

ord

er o

f des

cend

ing

prod

uctiv

itygr

owth

)(p

erce

nt)

(per

cent

)(p

erce

nt)

(per

cent

)(p

erce

nt)

Prin

ting

and

publ

ishi

ng2.

70

0.2

.2M

achi

nery

, exc

ludi

ng e

lect

rical

2.6

3.7

2.5

1.1

.9*N

onm

etal

min

ing

2.6

00

.4.1

Pape

r2.

5.9

0.5

.1tra

de2.

50

0.1

.1*M

etal

min

ing

2.4

00

.4.2

*Ret

ail t

rade

2.4

00

.1.1

Ston

e, c

lay,

and

gla

ss p

rodu

cts

2.4

.90

.5.1

Bev

erag

es2.

20

0.6

.1

App

arel

1.9

.10

.30

Fabr

icat

ed m

etal

pro

duct

s1.

9.7

.5.7

.3Le

athe

r Pro

duct

s1.

7.2

0.2

.1

Prim

ary

met

al p

rodu

cts

1.6

.9.1

.6.2

*Con

tract

con

stru

ctio

n1.

50

0.6

.3To

bacc

o pr

oduc

ts1.

10

0.3

0*W

ater

tran

spor

tatio

n0.

50

0.4

0

*Non

man

ufac

turin

g in

dust

ries.

SOU

RC

ES: P

rodu

ctiv

ity G

row

th R

ate

data

repr

oduc

ed b

y pe

rmis

sion

from

the

auth

or: K

endr

ick

(197

3) ta

ble

5—1,

pp.

78—

79; R

and

Din

tens

ity ra

tios f

rom

Ter

leck

yj (1

974)

, pp.

13,

15.

Not

e re

visi

on o

f dat

a fo

r bev

erag

es in

dust

ry R

& D

con

duct

ed in

indu

stry

ratio

s.

jT

able

6.2

-V

aria

btes

Oth

er th

an R

and

I) T

este

d fo

r Pos

sibl

e Ef

fect

s on

Prod

uctiv

ity

r-T

able

6.1

(con

t.)

——

— /

fiji.

&%

4*.

.SJ.

tin

tens

ityra

tios f

rom

Ter

leck

yj (1

974)

, pp.

13,

15. N

ote

revi

sion

of d

ata

for b

ever

ages

indu

stry

R &

D c

ondu

cted

in in

dust

ry ra

tios.

Tabl

e 6.

2V

aria

bles

Oth

er th

an R

and

D T

este

d fo

r Pos

sibl

e Ef

fect

s on

Prod

uctiv

ity G

row

th,

Thirt

y-th

ree

Indu

strie

s in

the

Priv

age

Dom

estic

Eco

nom

y

Uni

on M

embe

rsas

Per

cent

of A

llSa

les O

ther

than

Wor

kers

inIn

dex

of C

yclic

alto

Gov

ernm

ent

Prod

ucin

gIn

stab

ility

of

as P

erce

nt o

fEs

tabl

ishm

ents

Indu

stry

Out

put

Tota

l Sal

es 1

958

1953

1948

—66

Indu

stry

PVTS

UN

CY

C(a

rran

ged

in o

rder

of d

esce

ndin

g pr

oduc

tivity

gro

wth

)(p

erce

nt)

(per

cent

)(p

erce

nt)

trans

porta

tion

9551

0*C

oal m

inin

g98

8416

.1*R

ailro

ads

9595

10.4

Che

mic

als

955.

7*E

lect

ric a

nd g

as u

tiliti

es96

410

Text

iles

100

307.

4R

ubbe

r pro

duct

s97

5495

*Com

mun

icat

ion

utili

ties

9752

0El

ectri

cal m

achi

nery

9056

11.0

Lum

ber p

rodu

cts

100

218.

7

tFar

min

g98

13.

2SO

il an

d ga

s ext

ract

ion

100

145.

4Tr

ansp

orta

tion

equi

pmen

t and

ord

nanc

e77

6511

.0Fo

ods

100

450

Petro

leum

refin

ing

9467

2.5

Furn

iture

9529

8.7

Inst

rum

ents

8550

9.2

Prin

ting

and

publ

ishi

ng98

383.

2M

achi

nery

, exc

ludi

ng e

lect

rical

9545

13.1

*Non

met

al m

inin

g99

324.

6

P

Tabl

e 6.

2 (c

ont.)

Uni

on M

embe

rsas

Per

cent

of A

llSa

les O

ther

than

Wor

kers

inIn

dex

of C

yclic

alto

Gov

ernm

ent

Prod

ucin

gIn

stab

ility

of

as P

erce

nt o

fEs

tabl

ishm

ents

Indu

stry

Out

put

Tota

l Sal

es 1

958

1953

1948

—66

Indu

stry

PVTS

UN

CY

C(a

rran

ged

in o

rder

s of d

esce

ndin

g pr

oduc

tivity

gro

wth

)(p

erce

nt)

(per

cent

)(p

erce

nt)

Pape

r99

456.

7*W

hole

sale

trad

e99

44.

1*M

etal

min

ing

9270

10.6

*Ret

ajj t

rade

9914

3.6

Ston

e, c

lay

and

glas

s pro

duct

s10

045

7.4

Bev

erag

es10

044

3.2

App

arel

9953

4.1

Fabr

icat

ed m

etal

pro

duct

s99

.45

8.4

Leat

her p

rodu

cts

9939

5.0

Prim

ary

met

al p

rodu

cts

9855

lS.2

*Con

tract

con

stru

ctio

n71

724.

1To

bacc

o pr

oduc

ts10

058

2.0

*Wat

er tr

ansp

orta

tion

9576

7.0

indu

strie

s.SO

UR

CES

: Dat

a on

indu

stry

sale

s to

gove

rnm

ent i

s fro

m U

.S. D

epar

tmen

t of C

omm

erce

, Off

ice

of B

usin

ess E

cono

mic

s, "T

he T

rans

actio

nsTa

ble

of th

e 19

58 In

put-O

utpu

t Stu

dy a

nd R

evis

ed D

irect

and

Tot

al R

equi

rem

ents

Dat

a,"

Surv

ey o

f Cur

rent

Bus

ines

s, 9

(196

5): 3

4—39

;pe

rcen

t uni

oniz

atio

n is

from

H.G

. Lew

is, U

nion

izat

ion

and

Wag

es in

the

Uni

ted

Stat

es (C

hica

go: U

nive

rsity

of C

hica

go P

ress

, 196

3), p

p.28

9—29

0; c

yclic

al in

stab

ility

inde

x of

indu

stry

out

put i

s bas

ed o

n an

nual

out

put d

ata

from

Ken

dric

k (1

973)

cal

cula

ted

in a

man

ner d

escr

ibed

in T

erle

ckyj

(196

0). S

ee n

ote

6, p

. 16,

in T

erle

ckyj

(197

4).

4——

—

ItioL

eUIII

Iy II

IUC

A U

t UlU

US

ttjf

15 U

4SC

U U

fl dt

IlIU

iiua

ia It

OH

!ca

iiuia

teu

ina

man

ner

Uca

s.lo

UeU

inTe

rleck

yj (1

960)

. See

not

e 6,

p. 1

6, in

Ter

leck

yj (1

974)

.

Tabl

e 6.

3R

egre

ssio

n C

oeff

icie

nts O

btai

ned

in th

e A

naly

sis o

f the

Ann

ual R

ates

of C

hang

e in

Tot

al F

acto

r Pro

duct

ivity

for t

hePe

riod

1948

—66

, Tw

enty

Man

ufac

turin

g In

dust

ries (

t—ra

tios i

n pa

rent

hese

s)

Cos

t of R

and

D/V

alue

-add

ed R

atio

s for

Diff

eren

t R a

nd D

Com

pone

nts

Oth

er V

aria

bles

Ran

d D

Con

duct

edR

and

D E

mbo

died

in P

ur-

in In

dust

rych

ases

from

Oth

er In

dust

ries

Con

stan

t of

Reg

ress

ion

Tota

l,19

58

Priv

atel

yG

over

nmen

tPr

ivat

ely

Gov

ernm

ent

Fina

nced

,Fi

nanc

ed,

Tota

l,Fi

nanc

ed,

Fina

nced

,19

5819

5819

5819

5819

58

Perc

ent o

f Sal

esN

ot to

Gov

ernm

ent,

1958

Uni

on M

embe

rs a

sPe

rcen

t of W

orke

rsin

Pro

duci

ng E

stab

-lis

hmen

ts, 1

953

Ann

ual R

ate

ofC

yclic

al C

hang

ein

Out

put,

1948

—66

R2

Cor

-re

cted

a0IR

DI

IRD

NIR

DG

PRD

IPR

DN

PRD

GPV

TSU

NC

YC

2.29

(0.3

1)1.

62(0

.25)

0.12

(1.7

6)0.

370.

01(3

.31)

(0.1

2)

0.02

(0.3

3)0.

03(0

.46)

—0.

04(2

.38)

—0.

05(2

.94)

—0.

03(0

.56)

—0.

03(0

.56)

0.39

0.44

—7.

83(1

.36)

0.28

0.45

(2.8

4)(2

.49)

0.12

(2.2

0)—

0.05

(3.5

4)—

0.05

(1.2

5)0.

62

—3.

72(0

.71)

0.29

0.78

—0.

09(3

.43)

(3.7

9)(0

.34)

0.08

(1.5

4)—

0.05

(4.0

3)—

0.04

(1.3

0)0.

72

Sous

ca: T

erle

ckyj

(197

4), p

p. 2

0, 2

2, 2

6, 2

9.

—-i


The effects estimated for the three non—R and D variables remainedrather stable, in the course of these substitutions of the R and D vari-ables.

These results for the manufacturing industries constituted the mainfindings of the earlier study. The results for nonmanufacturing (notreproduced here) were rather erratic. No indication of positive returnsto R and D conducted in the industry was obtained. But, for this groupof industries, it is not surprising. It may simply reflect the fact that mostnonmanufacturing industries conduct little or no R and D. (Also, theR and D data for the nonmanufacturing industries were derived fromthe NSF data by a series of additional assumptions.) On the other hand,a statistically significant coefficient was obtained for indirect returns,suggesting a very high rate of productivity return of 187%. Dividingthe indirect R and D by sources of financing gave large and positivecoefficients for both components which, however, were not statisticallysignificant.

The overall fit of the equation for all industries combined was con-sistently lower than for either manufacturing or nonmanufacturing in-dustries alone. Among the coefficients for the R and D intensity ratios,only the one for the total R and D cost embodied in purchases wasstatistically significant.

6.4 Productivity Returns to R and D with Considerationof Human Capital

Human capital is not included in the measures of input used in esti-mating productivity growth. The underlying indexes of labor input usedin constructing the indexes of total factor productivity from whichthe growth data are derived are based on the number of man-hours.Consequently, a part of growth in productivity may represent produc-tivity returns to a growing stock of human capital. Moreover, if in-creases in the use of human capital are correlated with increased useof R and D capital, the regression coefficients intended to measure theproductivity return to R and D may be biased upward to the extent thatthey (also) reflect returns to human capital. Use of human capital maybe correlated either with own R and D, if use of human capital inproduction is complementary with the conduct of R and D, or withpurchased R and D, if human capital is complementary with the useof R and D—intensive, "high-technology" inputs. Thus, the regressioncoefficients for both direct and indirect return to R and D are subjectto potential bias from this source.

In testing the hypothesis that the previously obtained statistical esti- tmates of returns to R and D do not include the returns to human capitalas well, the estimating equations are revised to include human capital


investment intensity as an additional variable. This variable is formu-lated in the same manner as the R and D investment intensity variable.After the introduction of human capital, equations (2), (3), and (4)discussed earlier become equations (6), (7), and (8):(6) Pt AeAtRatHYt,

J;T

(7)

(8)

Here y is the elasticity parameter, analogous to a, and r in equation(8) is the productivity rate of return to human capital, analogous to vfor research capital. Both parameters are estimated as regression coeffi-cients of the respective net investment intensity ratios in the estimatingequation for productivity growth. However, while the basic cost ofRand D activities conducted in an industry is included in the labor andcapital input indexes (except for the effect of differences in labor costper man-hour), the cost of human capital is not included in the inputdata.

No direct measures of human capital stock or investment exist thatcould be readily applied in estimating its productivity effects. Humancapital has to be measured indirectly, by schooling or by an indicatorof its market value. In this paper, the indicator of human capital is in-vestment intensity. It is based on the increases in real wages per man-hour worked.

This indicator of growth of human capital is based on its evidentmarket value. The advantage of basing the measure of human capitalon wages rather than on schooling is that such a measure includes bothschooling and experience components of earning power (Mincer 1974),which presumably reflects productivity and at the same time excludesconsumption components of schooling and the variability of the learn-ing content of years of schooling over time. However, there are alsodisadvantages in using the relative real wage growth as a measure ofhuman capital investment. One is that other factors unrelated to workerproductivity affect this growth to an unknown extent. Such factors aswage bargaining or legislation have autonomous effects on wages.

Also, because, statistically, the growth in labor compensation is amajor component of growth in output and output per man-hour is bydefinition a large part of the total factor productivity growth, the humancapital investment rate estimated from growth in real wages is subjectto some possibility of the simultaneous equation bias. However, thisbias would not arise if the competitive working of the labor marketequalized the earnings within occupations rapidly relative to the length


of time over which the productivity growth is measured. Then theobserved long-term increases in wages in the individual industries wouldbe independent of the increases in productivity in the same industriesbecause the period of observation would be sufficiently long for thecorrelation to reflect only the effects of increases in human capital onproductivity.

The best approach to resolve the uncertainties resulting from thenature of the indicator used for human capital is to test the resultsobtained by alternative indicators. Use of alternative indicators was notpossible within the scope of this paper because of the lack of the avail-able data in comparable industry detail or for the period studied. Itshould be possible in future research to develop or adapt indicators ofaverage education and of the skill mix of labor input in individualindustries. Here, a more general further test of independence of theestimated effects of R and D from previously unmeasured inputs isundertaken by estimating the effects of industrial R and D on the totalfactor productivity growth measured after an adjustment for humancapital and other inputs.

The real wage growth was calculated from unpublished NBER dataused in Kendrick's study. The original data were in the form of periodaverages of actual hourly earnings for the period 1948—5 3 and againfor the period 1960—66. These averages were converted to 1958 dollarsby the Consumer Price Index. The difference between the two periodaverages was divided by 12.5, the number of years between the mid-points of the two periods. The result is shown in the first column intable 6.4. Equation (8) requires the net rate of investment in humancapital in the aggregate in the industry rather than per hour. Therefore,the amount per hour in the first column is multiplied by the man-hoursworked in the industry in the base year, 1958, used in this study. Theresult, giving the investment in human capital in millions of 1958 dol-lars, is shown in the second column. This result is then divided by the1958 value added by industry in order to obtain the desired variable,H/Q, i.e., the human capital investment intensity ratio which is shownin percentages in the third column.

It may be noted that, while, theoretically, human capital investmentintensity ratios are commensurate with the R and D investment intensityratios used in the regression analysis, statistically their estimates aremuch stronger. The human capital data are actually based on changefor the entire period rather than on the experience in the base year.Also, they are based on net investment in human capital rather than onthe gross investment as was the R and D ratio.

The results of introducing human capital investment into the estimat-ing equation for productivity growth for the twenty manufacturing in-dustries are given in table 6.5. The estimated effect of the human capital

I

Tabl

e 6.

4Es

timat

es o

f the

Net

Inve

stm

ent R

ate

in H

uman

Cap

ital,

This

ty-th

ree

Man

ufac

turin

g an

d N

onm

anuf

actu

ring

Indu

strie

s,19

58 (d

ofla

r am

ount

s in

1958

doH

ars)

Ave

rage

Ann

ual

Indu

stry

Inve

stm

ent

Incr

ease

in R

eal

Estim

ated

Indu

stry

as P

erce

nt o

fW

age

per H

our

Inve

stm

enta

1958

Val

ue A

dded

,In

dust

ry(li

sted

in o

rder

of d

esce

ndin

g ra

te o

f pro

duct

ivity

gro

wth

)W

orke

d, 1

948—

66(1

)(m

illio

ns)

(2)

1958

(3)

*Air

trans

porta

tion

$.10

15$

36.5

2.4%

*Coa

l min

ing

.071

127

.61.

6*R

ajlro

ads

.111

221

6.4

2.6

Che

mic

als

.108

017

2.7

1.9

*Ele

ctric

and

gas

util

ities

.095

212

7.4

1.2

Text

iles

.034

060

.91.

5R

ubbe

r pro

duct

s.0

678

46.2

1.6

*Com

mun

jcat

ions

.081

714

1.2

1.6

Elec

trica

l mac

hine

ry.0

751

182.

22.

0Lu

mbe

r pro

duct

s.0

503

67.9

2.1

*Far

mjn

g.0

166

208.

81.

0SO

il an

d ga

s ext

ract

ion

.059

144

.5.5

Tran

spor

tatio

n eq

uipm

ent a

nd o

rdna

nce

.fl71

414.

42.

6Fo

ods

.074

323

5.2

1.9

Petro

leum

refin

ing

.143

063

.52.

0

Furn

iture

.047

736

.21.

9In

stru

men

ts.0

9 17

59.4

2.3

Prin

ting

and

publ

ishi

ng.0

433

77.1

1.3

Mac

hine

ry, e

xclu

ding

ele

ctric

al.0

847

232.

62.

15N

onm

etal

min

ing

.069

519

.31.

9

Tabl

e 6.

4 (c

ont.)

Ave

rage

Ann

ual

Indu

stry

Inve

stm

ent

Incr

ease

in R

eal

Estim

ated

Indu

stry

as P

erce

nt o

fW

age

per H

our

Inve

smen

ta19

58V

alue

Add

ed,

Indu

stry

Wor

ked,

194

8—66

(mill

ions

)19

58(a

rran

ged

in o

rder

of d

esce

ndin

g pr

oduc

tivity

grow

th)

(1)

(2)

(3)

Pape

r.0

769

89.5

1.9

*Who

lesa

le tr

ade

.070

646

8.7

1.6

*Met

al m

inin

g.0

800

15.3

1.4

trade

.041

589

2.4

2.0

Ston

e, c

lay,

and

gla

ss p

rodu

cts

.08

1293

.42.

0

Bev

erag

es.0

569

23.8

1.0

App

arel

.028

460

.01.

3Fa

bric

ated

met

al p

rodu

cts

.073

615

9.6

2.0

Leat

her p

rodu

cts

.036

924

.11.

6

Prim

ary

met

al p

rodu

cts

.114

225

0.7

2.3

cons

truct

ion

.085

860

7.6

3.0

Toba

cco

prod

ucts

.090

716

.2.6

*Wat

er tr

ansp

orta

tion

.045

543

.02.

9

*Non

man

ufac

turin

g in

dust

ries.

aAve

rage

ann

ual i

ncre

ase

in re

al w

ages

per

hou

r wor

ked

from

col

umn

(I) m

ultip

led

by th

e nu

mbe

r of m

an-h

ours

wor

ked

in 1

958,

as

SOU

RC

ES: J

ohn

W. K

endr

ick

and

the

Nat

iona

l Bur

eau

of E

cono

mic

giv

en in

Ken

dric

k (1

973)

, pp.

196

—97

.R

esea

rch,

unp

ublis

hed

data

on

real

wag

es.

Tabl

e 6.

5R

egre

ssio

nC

oeff

icie

nts O

btai

ned

in th

e A

naly

sis o

f the

Ann

ual R

ates

of C

hang

e in

Tot

al F

acto

r Pro

duct

ivity

for t

hePe

riod

1948

—66

with

Add

ition

al V

aria

bles

, Tw

enty

Man

ufac

turin

g In

dust

ries (

t-rat

ios i

n pa

rent

hese

s)

Cos

t of R

and

D/ V

alue

-add

ed R

atio

s for

Diff

eren

t R a

nd D

Com

pone

nts

R a

nd D

Em

bodi

ed in

R a

nd D

Con

duct

edPu

rcha

ses f

rom

in In

dust

ryO

ther

Indu

strie

sO

ther

Var

iabl

es•

Hum

anC

apita

lIn

vest

men

t Int

ensi

tyPe

rcen

t of S

ales

Not

to.

Uni

onM

embe

rs a

sPe

rcen

t of W

orke

rsA

nnua

l Rat

e of

Cyc

lical

Cha

nge

Priv

atel

yG

over

nmen

tPr

ivat

ely

Gov

ernm

ent

Con

stan

t of

Reg

ress

ion

Fina

nced

,Fi

nanc

ed,

Fina

nced

,Fi

nanc

ed,

1958

1958

1958

1958

Gov

ernm

ent,

1958

in P

rodu

cing

Est

ab-

lishm

ents

,195

3in

Out

put,

1948

—66

Rat

io B

ased

on

Gro

wth

ofR

ealW

ages

l958

R2

Cor

-re

cted

a0IR

DN

IRD

OPR

DN

PRD

GPV

TSU

NC

YC

HC

IW—

6.58

.25

.82

.10

—.0

4—

.07

.39

.73

(1.5

3)(3

.00)

(3.9

9)(2

.62)

(3.3

8)(1

.73)

(1.0

4)—

6.17

.25

—.0

5.8

5.1

7.1

0—

.04

—.0

7.3

8.6

9(1

.02)

(2.5

0)(.5

9)(3

.78)

(.37)

(1.7

0)(3

.41)

(1.5

6)(.9

3)


variable is not statistically significant, while, on the whole, the previousresults remain unaffected by its introduction.

6.5 Results with Productivity Data Based on an InclusiveConcept of Input

In their paper, Gollop and Jorgenson have developed a set of esti-mates of output, input, and total factor productivity for individualindustries (as well as economic sectors and the economy) which arebased on different measurement concepts than the estimates by Ken-drick, which are used to derive the estimates of the effects of R and Don productivity. Thus, while Kendrick measures the net output and thenet input, Gollop and Jorgenson use gross measures of output and input,i.e., including intermediate inputs.

Also, the Gollop-Jorgenson method attempts to account for qualitycharacteristics of inputs, and their input index accordingly includes"quality" index components for labor and for capital.

To the extent that research and development activities result in im-proved intermediate and capital inputs, one would expect the estimatesof productivity rates of returns to R and D to be lower when produc-tivity is derived from the input data already adjusted for the qualitychanges than when the calculation of productivity growth is based onthe unadjusted inputs.

The two sets of productivity data also differ in the underlying theo-retical formula. Kendrick implicitly used a reduced form of the Cobb-Douglas production function. The Gollop-Jorgenson data are derivedfrom a translog production function. The two sets of productivity esti-mates also differ in the method of estimating the man-hours input andin the estimate of depreciation of capital. But in contrast to theinput quality adjustments, there is no apparent reason to expect thesedifferences in measurement to have a systematic effect on the observedcorrelations between productivity growth and the R and D variables.

The results of substituting Gollop-Jorgenson estimates for Kendrickestimates of productivity change' for the twenty manufacturing indus-tries and with the same independent variables are shown in table 6.6.The Gollop-Jorgenson estimates are not available for the nonmanufac-turing industries.

1. With two minor changes in the Gollop-Jorgenson data to make consistentthe industry definitions: (1) Productivity growth for "food and kindred products"is used for both "food products" and "beverages," and (2) the arithmetic averageof growth rates for "motor vehicles" and for "transportation equipment, otherthan motor vehicles, and ordnance" was used for "Transportation equipment andordnance."

0.I

ici

ici

i.—

.i

Tab

le6.

6R

egre

ssio

n C

oeff

icie

fits O

btai

ned

with

Alte

rnat

ive

Estim

ates

of t

he A

nuua

l Rat

es o

f Cha

nge

in T

otal

Fac

tor P

rodu

ctiv

ity,

Twen

ty M

anuf

actu

ring

Indu

strie

s, 19

48—

66 (i

-rat

ios i

n pa

rent

hese

s)

Con

stan

t of

Reg

ress

ion

Cos

t of R

and

D/V

alue

-add

edR

atio

sfo

rD

iffer

ent R

and

D C

ompo

nent

s

Oth

er V

aria

bles

R a

nd D

Em

bodi

edR

and

D C

ondu

cted

in P

urch

ases

from

in In

dust

ryO

ther

Indu

strie

sPe

rcen

t of S

ales

Not

toG

over

nmen

t,19

58

Uni

on M

embe

rs a

sPe

rcen

t of W

orke

rsin

Pro

duci

ng E

stab

-lis

hmen

ts, 1

953

Ann

ua' R

ate

ofC

yclic

al C

hang

ein

Out

put,

1948

—66

R2

Cor

-re

cted

Priv

atel

yG

over

nmen

tPr

ivat

ely

Gov

ernm

ent

Fina

nced

,Fi

nanc

ed,

Fina

nced

,Fi

nanc

ed,

1958

1958

1958

1958

a0JR

DN

IRD

GPR

DN

PRD

GPV

TSU

NC

YC

Ken

dric

kda

ta—

4.01 (.73)

.27

—.0

5.8

1.1

2(2

.92)

(.53)

(3.6

6)(.

26)

.08

(1.5

2)—

.04

(3.6

9)—

.05

(1.3

7).6

9G

ollo

p-Jo

rgen

son

dala

—32

.26

(1.7

9).2

0—

.18

1.83

1.67

( .64

)(.6

3)(2

.53)

(1.1

0).3

3(1

.87)

03 (.88)

m02 (.1

4).3

0


Compared to the estimates based on Kendrick's data, the estimatesderived with the Gollop-Jorgenson data in table 6.6 have a much poorerstatistical fit. But the two sets of estimates are qualitatively consistent.The signs of all coefficients are the same in both equations, and theirgeneral magnitudes are comparable. Among the four R and D variables,only the coefficient for privately financed purchased R and D is statis-tically significant at the 5% level. It is also considerably larger (1.83)than the coefficient derived with the Kendrick data (0.81), despite themuch lower estimates of productivity growth by Gollop-Jorgenson(averaging 0.9% a year for the twenty manufacturing industries duringthe period 1948—66) than by Kendrick (averaging 2.8%).

Somewhat surprisingly, perhaps, the equation based on Gollop-Jor-genson data suggests the possibility of positive "spillover" effects ofgovernment R and D to the industries purchasing inputs from the in-dustries conducting government-financed R and D. In this equation therespective coefficient for the government-financed R and D is larger andstronger than in all the earlier analyses, but it is still not statisticallysignificant.

Among the non—R and D variables, the estimated coefficients aregenerally similar, but their statistical significance is eliminated, exceptperhaps for the share of sales in private markets.

6.6 Conclusions

Significant effects of the privately financed industrial R and D onproductivity growth of manufacturing industries were found in earlierresearch by the author. These effects were two-fold: (1) direct increasesin productivity of industries conducting the privately financed R and D,and (2) indirect increases in productivity of industries purchasing capi-tal and intermediate inputs from the industries conducting the privatelyfinanced R and D. The estimated indirect effects were considerablylarger, per dollar of R and D expenditure, than the direct effects. Nocomparable effects were found for government-financed industrial Rand D.

These findings were tested in this paper for independence of the Rand D effects from the possible effects of the previously unmeasuredinputs and input characteristics, in general, for human capital, interme-diate goods, and composition of physical capital, and for human capitalin particular.

The results continue to uphold strongly significant and large estimatesof indirect effects of privately financed R and D. There was a weakeningof the significance of the estimated direct effects of privately financedR and D, a continued absence of any indication of direct productivityeffects of government-financed R and D, and some indication of pos-

A


sible indirect effects of the government-financed industrial R and D.Qualitatively, the results of the tests were consistent with the earlierfindings. More detailed research is needed in the future to permit anevaluation of the possible effects of the previously unmeasured inputs,one at a time.

References

Gollop, Frank, and Jorgenson, Dale W. 1979. U.S. total factor produc-tivity by industry, 1947—73, this volume.

Griliches, Zvi. 1973. Research expenditures and growth accounting. InScience and technology in economic growth, ed. B. R. Williams. NewYork: John Wiley & Sons, Halsted Press.

Kendrick, John W. 1973. Postwar productivity trends in the UnitedStates, 1948—1969. New York: National Bureau of Economic Re-search.

Mansfield, Edwin; Rapoport, John; Romeo, Anthony; Wagner, Samuel;and Beardsley, George. 1975. Returns from industrial innovation:Detailed description of 17 case studies. University of Pennsylvania,preliminary and unpublished.

Mincer, Jacob. 1974. Schooling, experience, and earnings. New York:National Bureau of Economic Research.

Terleckyj, Nestor E. 1960. Sources of productivity advance. A pilotstudy of manufacturing industries, 1899—1953. Ph.D. diss., ColumbiaUniversity.

1974. Effects of R&D on the productivity growth of industries:an exploratory study. Washington: National Planning Association.

Comment Steven Globerman

IntroductionTerleckyj's paper is an extension of a rich and interesting study of a

slightly earlier vintage (Terleckyj 1974). The earlier study sought toidentify separately returns to the R and D conducted within industriesand the R and D "purchased" from other industries in the form of em-bodied technology in capital and intermediate goods. A related concernwas to estimate the separate returns to privately financed R and D (both

V

Steven Globerman is at York University.

•

.378 Nestor E. Terleckyj

"own" and "purchased") and government-financed R and D. The mainfindings of that study are as follows:

1. For manufacturing industries, direct returns to private R and Dwere on the order of 30%, in terms of productivity growth; indirectreturns were on the order of 80%. The productivity returns, both directand indirect, to government-financedR and D were estimated at zero.

2. For nonmanufacturing industries, no indication of positive returnsto R and D conducted in the industry was obtained. Indirect returnswere on the order of 187%.

This conference paper extends earlier findings by attempting to holdconstant (either explicitly or implicitly) the effects of increased use ofhuman capital, relative increases in the use of intermediate inputs, andchanges in the composition of labor and capital which may have beencorrelated with increased direct and indirect investment in R and D bythe industry.

Before summarizing Terleckyj's empirical results, the basic modelunderlying the estimation procedure should be briefly reviewed. Theunderlying model for most estimations is a Cobb-Douglas productionfunction with labor, physical capital, research capital, and a "disem-bodied" rate of growth of productivity as arguments of the function.By making the usual assumption about equality of factor prices to mar-ginal value products and taking time derivatives of the variables, areduced-form equation is derived in which the rate of change in anindustry's total factor productivity is a linear function of the rate ofgrowth of the intangible stock of R and D capital. Since the growth rateof the R and D capital stock cannot be directly measured, the ratio ofgross investment in R and D to output is substituted into the equation.For purposes of estimation, it is assumed that depreciation in R and Dcapital can be ignored and that the gross investment in R and D capitalover the sample period can be measured by the industry's R and D in-tensity in 1958.

In the estimating equations, the aggregate R and D intensity variableis decomposed by source of financing (private versus public) and bylocation of conduct (within or without the industry) Additional non—R and D standardizing variables included in all reported estimatingequations are percent of sales not made to government, union membersas a percent of workers in producing establishments, and annual rateof cyclical change in output. The initial dependent variable is Kendrick'sindex of total factor productivity compiled for thirty-three manufactur-ing and nonmanufacturing industries, covering the period 1948—66.

The initial estimation results reported for the sample of twenty man-ufacturing industries are essentially those of Terleckyj's earlier study,and exclude the effect of changes in input quality and in the use of


intermediate inputs. Results for the nonmanufacturing sample are notreproduced, but are reported as above.

In the first extension of his preceding findings, Terleckyj introducesa measure of human capital intensity, based upon the increase in realwages per man-hour worked over the period 1948—66, into earlier esti-mating equations. He finds that introduction of the human capital mea-sure leaves previous results essentially unaffected, and the coefficientfor the human capital variable itself is statistically insignificant.

In a second extension, Gollop-Jorgenson total factor productivityestimates for the twenty manufacturing industries are substituted forKendrick's estimates in the initial set of estimating equations. The over-all statistical results are much poorer than those obtained employingthe Kendrick estimates: the adjusted R2 coefficients decrease substan-tially, and the "own" privately financed R and D coefficient becomesstatistically insignificant. Interestingly, the "purchased" privately fi-nanced R and D coefficient remains statistically significant using theGollop-Jorgenson measure of total factor productivity, and, furthermore,has a greater impact upon productivity than in the earlier equations.Thus, the various specifications of the productivity/R and D relationshipprovide a range of estimates of the direct and indirect effects of R andD expenditures.

The following discussion will focus primarily upon issues relating tothe model specifications, measurement of variables, and potential prob-lems associated with the single-equation estimation procedure. I haveno great difficulty in accepting the general nature of Terleckyj's findings.Specifically, to the extent that the performance (or purchase) of R andD is associated with improved quality of conventional factor inputs, onewould expect the estimated returns to R and D to be lower when in-creases in the "quality" of labor and capital are otherwise accountedfor. Furthermore, since the numerator of Kendrick's total factor pro-ductivity measure is gross output while the denominator excludes inter-mediate inputs, one would expect estimates of returns to "purchased"R and D to be biased upward if embodied R and D expenditures arepositively correlated with the relative use of intermediate inputs. Thisis bound to be the case for Terleckyj's sample since the imputations ofembodied R and D were done by redistributing the R and D expendi-tures of each industry in proportion to the distribution of sales of thatindustry to other industries and them summing the amounts attributedto each of the sample industries. The fact that the "purchased" privatelyfinanced R and D coefficient increases substantially in the equationemploying the Gollop-Jorgenson data suggests the possibility that other,unspecified sources of bias may be present in the estimations.


The observation that the productivity effect of "own" privately fi-nanced R and D essentially disappears when "quality" adjustments forlabor and capital and the relative use of intermediate inputs are incor-porated into the dependent variable (i.e., the Gollop-Jorgenson index)is somewhat difficult to accept on intuitive grounds. If labor and capitalresources employed in R and D activities have less risky employmentalternatives, one would expect R and D resources to earn their users"excess" returns or be shifted into other activities over time. Part of theexplanation for the different results reported in table 6.6 may, indeed,rest in the different methodologies employed to construct the dependentvariables. In any case. since the attribution of the productivity effectsof R and D will ultimately depend upon how the factor productivityresidual is defined, we are somewhat less concerned about explanationsof differences in estimated rates of return to R and D across differentproductivity measures than we are with the identification of returns toR and D employing any given productivity measure. It is the latterconcern we will primarily address in our discussion.

Model SpecificationThe inclusion of input "quality" measures reflects Terleckyj's concern

about possible estimation biases arising from the omission from theestimating equation of one (or more) variables whose values differedacross sample industries. To the extent that differences in the values ofother omitted variables are unsystematically related to the includedvariables over the sample period, the slope parameters would remainunaffected. However, the brace of standardizing variables employed byTerleckyj excludes certain variables that may be systematically relatedto the rate of growth in R and D investment.

One such variable is the rate of growth in output in the sample indus-tries over the estimation period. There is ample evidence that industriesenjoying above-average productivity growth rates (associated in part,with investments in R and D) also enjoy above-average rates of growthin output. Relative increases in output growth rates could, in turn,affect industry differences in productivity through differential scale andlearning economies. Including an output growth rate variable in theestimating equation would raise an identification problem; however,one suspects that its exclusion leads to an upward bias in the estimatedR and D parameters. it

Another potential source of bias arises from changes in the relativedegree of product specialization within industries over time. The deriva-tion of productivity and R and D data along establishment (and prod- duct) lines reduces but does not obviate this possibility, particularlygiven the level of aggregation of the sample industries. One might hy-pothesize that over the sample period, R and D—intensive industries

t


were becoming relatively more specialized (on an establishment basis),both because a rapidly expanding market for their output facilitatedincreased specialization and because specialization facilitated entry intohigh-technology industries.' The production and, perhaps more impor-tantly, the learning economies associated with increased product spe-cialization would support the productivity effects of both disembodiedand embodied R and D, leading to upward biases in the estimatedparameters.

The effect of differential rates of growth in "x-efficiency," in turnrelated to changes in industrial market structure, might also be em-bodied in the estimated R and D parameters. An upward bias could beimparted to the R and D coefficients if technological change bringsabout a decline in average plant size relative to market size. Such adecline could facilitate easier entry into the industry, thereby fosteringgreater competition and a more efficient allocation of resources, includ-ing faster interfirm rates of technological diffusion. Evidence relatingchanges in concentration ratios to technological change is far from con-clusive but, on balance, points to the existence of a negative relationshipbetween the two variables.2

The variable used by Terleckyj to standardize for changes in laborquality is the increase in real wages per man-hour worked. The originaldata for this variable were in the form of period averages of actualhourly earnings for the periods 1948—53 and 1960—66, deflated to 1958dollars. The annualized percentage change between the two periods wasmultiplied by the man-hours worked in the industry in the base year,1958, to obtain the rate of growth in aggregate human capital. Theresulting term was then divided by the 1958 value added by the industryto obtain the desired variable.' The author acknowledges the possibilityof a simultaneous-equation bias arising from the feedback of increasedproductivity to higher wages, but argues that this bias would not ariseif the competitive working of the labor market equalized the earningswithin occupations rapidly relative to the length of the period over whichthe productivity growth is measured. Even if adjustments in the relativesupply of labor for different occupations proceeded rapidly (a phenom-enon which is certainly at variance with recent evidence on the signifi-

1. The process of entry through specialization in the innovation process is illus-trated in the case of the semiconductor industry. See John Tilton, InternationalDiffusion of Technology: The Case of Semi-Conductors (Washington: The Brook-ings Institute), 1971.

2. Examining concentration ratios at the establishment level for different indus-tries over time might provide some feel for the magnitude of this potential bias.

3. Since the numerator of the productivity measure is gross output, it is notevident why the human capital variable, as well as the other standardizing vari-ables, are deflated by value-added in the base year rather than by gross output.

t


cance of information costs in factor markets), a simultaneous-equationbias might still be obtained if wages (to any significant extent) incor-porate expected productivity gains. While it is likely that the Gollop-Jorgenson labor quality measure, based on the shift of workers amongdifferent categories, is subject to a smaller simultaneity bias, some biaswill still be present if higher wages, in part, are realized in the form ofjob upgrading.

In light of the acknowledged shortcomings of the labor-quality vari-able employed, other measures, including the average education levelof employees, might have been tested. Median years of schooling of thelabor force by industry are available for censal years. While the avail-able censal years 1959 and 1969 do not provide a perfect overlap withthe estimation sample period, such exact concordance might not benecessary if relative differences in education levels across industries arereasonably stable over time. The assumption that differences in educa-tion levels at a point in time reflect differences in growth rates over thepreceding period seems no more heroic than a similar assumption in-voked by the author in specifying the R and D variables.

In fact, we reestimated the basic productivity equation across thetwenty manufacturing industry sample using Kendrick's index of averageeducation per employee for 1959 (Kendrick 1973) and obtained resultsquite similar to those obtained by Terleckyj for the growth-in-real-wagesvariable. The failure to identify a statistically significant positive returnto human capital investment for those estimations employing Kendrick'sproductivity index as the dependent variable is a surprising result. Apossible explanation of this result is the existence of multicollinearitybetween the human capital variables and other included variables. In-deed, the zero-order correlation coefficient between Terleckyj's humancapital measure and "own" privately financed R and D equals .58,while the correlation coefficient between the human capital measure andthe cyclical change in output is .69. Thus, the influence of human capi-tal investment on productivity growth may be largely assigned to col-linear variables in the estimation process.

Measurement of VariablesI will not comment extensively on the problems associated with mea-

surement of variables since Terleckyj considers most of the obviousproblems in some detail in his earlier study. Many of the problems are,in any case, intractable given the state of the available data.

The failure of cotiventional productivity measurements to adequatelyreflect product quality changes is well-recognized and presumably leadsto an underestimation of real rates of productivity change. Insofar as thebias differs among products, the industry comparisons would be affectedand, on balance, this probably imparts a downward bias to the estimated

nI'

I383 Direct and Indirect Effects of Industrial Research and Development

R and D parameters. To the extent that federally financed R and D is.primarily directed towards improvements in product quality as opposedto cost reduction, the methodology used in deriving industry produc-tivity estimates could contribute to the finding that government-financedR and D is not significantly related to productivity change. Furthermore,while the period 1948—66 might be long enough compared to the timelag one would expect between privately financed R and D intensityratios in 1958 and their productivity effects, it might be too short tofully incorporate the effects of federally financed R and D which ispresumably aimed at effecting greater changes in underlying productionconditions. While Terleckyj's finding of no productivity return to gov-ernment-financed R and D accords with similar results obtained byLeonard (1971) in a study of interindustry output growth differences,the latter study is subject to the same sorts of criticisms.

The imputations of purchased R and D to the sample industries weredone by redistributing the R and D expenditures of each industry inproportion to the distribution of sales of that industry to other indus-tries and then summing the amounts attributed to each of the industries.While imputing embodied R and D as a strict proportion of sales mightbe no more arbitrary than any other procedure, I would expect thatindustries performing R and D intensively are more likely to purchasecomplex and technically advanced equipment than are those industrieswhich perform little R and D. Another imputation procedure mighttherefore attribute the R and D transferred from a supplier industry toany given purchasing industry as a proportion of the percentage of totalsales made to the purchasing industry weighted by the relative "own"R and D intensity of the purchasing industry. In any case, it would beenlightening to evaluate the sensitivity of the statistical results to theimputation procedure used.

A more serious concern is the use of a single year's set of observa-tions to calculate interindustry flows of purchased inputs.4 Albeit theyare dictated by available data, it is certainly possible that observedfactor ratios in the given year were not long-run equilibrium values.Indeed, a similar concern might be expressed about most of the inde-pendent variables. For example, the "own" R and D intensity variablesare measured for a given base year, 1958. In his earlier study, Terleckyjargues that differences in base year R and D intensity levels are likelyto reflect differences in preceding R and D growth rates. While thisassumption might be quite tenable when comparing high and low R andD—intensive industries, it is somewhat more objectionable for the seven

4. Data for 1958 were employed in developing the "purchased" privately if.nanced R and D intensity variable and the relative growth rate in intermediateinputs variable.

-

I384 Nestor E. Terleckyj

manufacturing industries with "own" privately financed R and D inten-sity ratios ranging between .2 and .5.

Estimation ProcedureGiven the substantial differences in the parameters estimated across

the separate samples of manufacturing and nonmanufacturing industries,it is inappropriate to fit common slope parameters for the R and Dvariables across the full sample of thirty-three industries.

Terleckyj notes that the results for nonmanufacturing industries wereerratic, and they can be shown to be highly sensitive, in particular, tothe inclusion or exclusion of the air transportation industry. Part of thereason for the sampling instability of the parameters for the nonmanu-facturing industry sample might be the substantial collinearity that existsalong the basic set of independent variables. Substantial collinearity alsoexists among specific independent variables for the manufacturing in-dustry sample. Specifically, the zero-order correlation coefficients forpairs of the four R and D intensity variables provided in Terleckyj'stable 6.1 (for the sample of twenty manufacturing industries) rangebetween .73 and .95. The simple correlation coefficients between salesother than to government as a percent of total sales and the variousR and D intensity variables range between —.79 and — .91. The factthat coefficients for the privately financed R and D variables tend to beunaffected by inclusion or exclusion of the government-financed R andD variables provides a reason for optimism about the reliability of theestimated returns to privately financed R and D. However, this stabilitymight simply reflect the fact that rate-of-return estimates for govern-ment-financed R and D are confounded by collinearity with other van-ables, and particularly with the nongovernment sales variable, so thatthe private R and D variables are assigned the joint effect of the en-tire set.

One might also conjecture that the relationship between productivitychange and returns to federally funded R and D depends upon thenature and the level of the R and D performed. Specifically, contractR and D performed in the electrical machinery and transportationequipment industries is largely defense-related. Thus, the effects ofR and D performance on productivity may be atypical for those twoindustries. Our own estimation provides some indication that returnsto government-financed R and D (both embodied and disembodied) arehigher in the above-mentioned industries than in other manufacturingindustries. The difference might reflect the existence of increasing re-turns to federally financed R and D or the possibility that governmentR and D support to other manufacturing industries is, in effect, a sub-sidy to offset low rates of productivity growth. These possibilities maybe worthy of further investigation.


Concluding CommentsTerleckyj's general findings (for the specifications employing Ken-

drick's productivity index) are directionally in accord with existingevidence in the literature; however, differences among the various studiesin the nature of the samples, the measurement of variables, and the tech-niques of estimation make specific comparisons rather tenuous. Forexample, Griliches (1973) obtained an estimated productivity returnto R and D of around 40% for a sample of eighty-five two, three, andfour-digit manufacturing industries. This estimate includes both direct(i.e., intraindustry) and indirect (i.e., interindustry) productivity effects.Terleckyj's estimates of total R and D returns (employing Kendrick'sproductivity index) range between 100 and 110% and lie substantiallyabove Griliches's estimated total returns. In another study using a par-tial productivity measure, Griliches (1975) estimated the rate of returnto R and D for 883 large R and D—performing U.S. manufacturingcompanies. The average gross excess rate of return to R and D was27% in 1963. It should be noted that this estimate includes the produc-tivity effects of intercompany as well as interindustry technology trans-fers. Mansfield (1965) found that marginal rates of return to R and Daveraged about 40 to 60% for a sample of petroleum firms; for asample of chemical firms, returns averaged about 30% if technicalchange was assumed to be capital-embodied but only about 7% if itwas disembodied.5

Thus, various alternative estimates of the productivity returns to Rand D tend to lie somewhat below estimates obtained by Terleckyj,although all of the estimated returns are substantial. Terleckyj's studyperforms the valuable service of demonstrating that these substantialreturns (at least for one productivity measure) cannot be significantlyreduced by statistically standardizing for human capital investments inan industry. It also provides us with dramatic evidence that estimatedreturns to R and D are extremely sensitive to the way in which produc-tivity is measured. However, Terleckyj's study shares a weakness withother studies in failing to hold constant the effects of such factors aschanges in average plant size, changes in plant-level specialization, andchanges in organizational structure (including intrafirm and interfirmrates of technology diffusion) which may be related to the performanceof R and D. An attempt to incorporate the influence of the above-mentioned factors into the basic productivity equation might make asignificant contribution to our understanding of the influence of R andD on industrial performance.

5. Mansfield's estimated returns also include the indirect effects of the R and Dconducted outside the sample firms.


Finally, Terleckyj should be complimented for the thoughtful inclu-sion and careful description of the data series included in his estimationwork. The inclusion facilitates extension of his interesting work byother researchers.

References

Griliches, Zvi. 1973. Research expenditures and growth accounting. InScience and technology in economic growth, ed. B. R. Williams. NewYork: John Wiley and Sons.

1975. Returns to research and development expenditures inthe private sector. Paper presented at Conference on Research in In-come and Wealth.

Kendrick, John. 1973. Post-war productivity trends in the UnitedStates, 1948—1959. New York: National Bureau of Economic Re-search.

Leonard, W. N. 1971. Research and development in industrial growth.Journal of Political Economy 79:232—56.

Mansfield, Edwin. 1965. Rates of return from industrial research anddevelopment. American Economic Review 55:310—22.

Terleckyj, Nestor E. 1974. Effects of R & D on the productivity growthof industries: an exploratory study. Washington: National PlanningAssociation.

Direct and Indirect Effects of Industrial Research and ... · PDF file361 Direct and Indirect...

Documents

Transcript of Direct and Indirect Effects of Industrial Research and ... · PDF file361 Direct and Indirect...