You are on page 1of 8

JOURNAL OF ECONOMIC THEORY 1, l-8 (1969)

Disembodied Technological Change


with Several Factors

EDWIN BURMEISTER
Department of Economics, Universify of Pennsylvania,
Philadelphia, Pennsylvania 19104

RODNEY DOBELL
Institute for the Quantitative Analysis of Social and Economic Policy,
University of Toronto, Toronto 5, Canada

1. INTRODUCTION AND REPRESENTATION LEMMA


Study of economic models involving technological change entails
introduction of a whole family of production functions, each member of
the family corresponding to a different state of technology. But such a
level of generality is unsatisfactory; empirical work demands some further
assumptions about the specific character of technological change. The
factor-augmentation hypothesis is one convenient restriction. Because
this hypothesis is restrictive, but not particularly transparent in its impli-
cations, some exploration of its meaning seems worthwhile. Therefore
in this paper we give a necessary and sufficient condition that a single
production function with several factor inputs, shifting over time, admits
a representation in the usual factor-augmenting form, and we derive
from this result natural characterizations of Hicks-, Harrod-, and Solow-
neutrality for the case of several factors.
Solow has observed [4] that the hypothesis that technological change
can be represented in such a form is definitely restrictive, and this is
obviously true. On the other hand, Phelps [2] has asserted that time
series data can not suffice to refute this hypothesis unless further speci-
fication is imposed. We seek first a mathematical characterization and an
1 Burmeisters research was supported by the National Science Foundation under
Research Grant GS-2304; Dobells research was supported in part by Harvard Univer-
sitys Program on Society and Technology under a long-term grant from the IBM
Corporation. While retaining sole responsibility for the contents and conclusions of
this paper, the authors wish to acknowledge, with thanks, this support.
0 1969 by Academic Press, Inc.
i 1
2 BIJRMEISTER AND DOBELL

economic interpretation of the sense in which the hypothesis is restrictive,


and it is then clear why complete time-series data for one history of the
world could never suffice to refute the factor-augmenting hypothesis.

LEMMA. Let a given production function F(K,, . . . , K,,, L; t) be homo-


geneous of degree one and concave in (K,, . , . , K,,, L), and let the ratios
KilL be denoted ki. There exist positive functions a(t) and hi(t), i = 1,. . . , n,
of time alone, having continuous jirst derivatives (with a(0) = hi(O) = 1)
and a function G homogeneous of degree one in all its arguments, such that
F may be represented in the factor-augmenting form
Q = WLK,,. . ., K,, L; 0 = G[bl(Wl,. . . , b&K,, 4Wl
if and only if
there exist positive functions hi(t), i = 1,. . . , n, of time alone, having
continuous$rst derivatives (with hi(O) = l), such that when all the expres-
sions hi(t)ki are constant (identically in t), then all the relative shares 7Li
are constant.

Proof. (Necessity.) Assume that the factor-augmenting representation


is possible and define hi(t) = bi(t)/a(t), i = 1,. . . , n; then for all i, hi(t)
is clearly unique, positive, possesses continuous first derivatives, and
satisfies hi(O) = 1. Further rci = f,hikilf is simply a function of the
expressions hiki, i = 1,. . . , n. Hence, if all such expressions hiki are
constant, all relative shares 71i are constant.
(Sufficiency.) Assume the existence of a set {hi(t); i = 1,. . . , n} of
functions hi(t) having the required properties.
Define
Zi(ki, t) = hi(t)ki, i = 1,. . .) n.
Then certainly no generality is lost in writing
q =f&,.. ., k,;t)=Q(Z1,Z2 ,..., Z,;t).
Since rri = ~~Zi/~, if all ni are constant whenever all Zi are constant,
then
cP,/Q?i = QJcP for all Zi and t.
With adequate continuity, assumed here, this condition implies

from which we conclude that @#I) is independent of 2, and hence is a


function of time alone. Integrating, we find

log @ = jx(C)dT+M(Z,,. . .,Z,>,


0
DISEMBODIED TECHNOLOGICAL CHANGE 3

and consequently,
@(Z,,. . *,Z,;t) = g(r) p-=1,...=,)
where M(Z,, . . . , Z,) is an arbitrary function of Zi to be determined by
boundary conditions. Defining
a(t) = eNcr), hi(f) = hi( > 0,
and
g(Z,,. . . , Z,) = eMcZ1*. ,n),
we obtain
q =ftkl,. . .,k,;t)=@(Z, ,..., Z,;t)

where by construction
= a(t)g ~b,(t) k
[ a(t) I. . .
b,ok
u(t) 1

the function a(t) is positive and satisfies a(0) = 1.
Let
b,(W, . . , -bn(Wn .
G[b,(W,, . . . >4,(W,, 4Wl = 4OG -9. u(t)L
G is then, by construction, a function of the required form, homogeneous
1
of degree one in its arguments. This completes the proof.

To illustrate the use of this criterion, consider some initial point


(k,(O), k,(O), . . . >k,(O)) for an economy, which therefore has a per capita
output f(k,(O), MO), . . ., k,(O); 0) at time t = 0, distributed according
to marginal productivity pricing into relative income shares
ni=KiFi/F=kifi/f, i=l,...,n
and
l-i;xi=LF+l/F
i=l

where subscripts on the functions F or f denote partial differentiation.


As the value of the index t changes, it is always possible to select some
path (WI, b(t), . . . , k,(t)) such that all relative income shares remain
constant at the initial level. Considering any different initial point, with
different relative income shares, it is again possible to find a path
Gl (0, h(t), . * * 9 E,(t)) such that all relative income shares remain constant
(at the new initial levels). Then our lemma asserts that the function F
may be represented in factor augmenting form if and only if for all i and
all t,

That is, the factor-augmenting form is possible only if all paths showing
constant income shares are, in effect, scale replicas of one another. Thus
the nature of the strong regularity imposed by the factor-augmenting
4 BURMEISTER AND DOBELL

hypothesis is clear, and it is also clear why no single history could suffice
to refute the hypothesis.
Particular interest attaches to special cases in which the paths maintain
all income shares constant, and we now turn to these.

2. NEUTRAL TECHNOLOGICAL CHANGE


Early discussion of technological change focused on the question of
what would happen to labors share of national product as a result of
technical progress. This emphasis makes it natural to treat cases in which
relative income shares remain constant as boundary lines between different
categories of technological change. Of course with a given production
function family, what happens to income shares depends on what path
in the input space is followed; if on particularly regular paths in the
input space, relative income shares remain constant, then we may say
that the production function map displays neutral technological
change.
Let us call technological change Harrod-neutral if when all output-
capital ratios Q/lvi (denoted Xi) remain constant, so also do all capital
rentals Fj (and hence all profit shares xi). Similarly, let us call technologi-
cal change Hicks-neutral if when all capital-labor ratios remain constant,
so also do all profit shares 71i. Then we may prove:

THEOREM 1. Technological change is Harrod-neutral on the above dej?ni-


tion if and only if there exist a(t) and b,(t) such that F may be represented
in the factor-augmenting form with all bi constant. That is, the family of
production surfaces represented by the function F has the property that
when all capital-output ratios Xi are constant, all projit shares ni are also
constant if and only if F may be represented in the form
F&I> . . ., K,, L; t) E G[K,,. . ., K,, a(t)L].
Proof. (a) Sufficiency follows by direct calculation.
(b) (Necessity). Assume that on any path with all Xi constant, all ni
are likewise constant. Then clearly along the same path all Fi are also
constant.
Consider some given initial vector k = (ky, . . . , kz); per capita output
q(k; 0) is then determined, as are all output-capital ratios Xp = q(kO ; 0)/k:.
Moreover, for any fixed t and any other vector k, the ratio
q(k; t)/q(k; 0) 2 F(k, 1; t)/F(k, 1; 0)
2 A referee points out that this result, of course, does not imply that the assumption
of factor augmentation at exponential rates could not be refuted by a single time-
series history of the world.
DISEMBODIED TECHNOLOGICAL CHANGE 5

is well-defined. Using the homogeneity of F, we may rewrite this identity


as
dkO;O) k q(k; 0) k dkO; 0)
F(k,l;O)-F ___ ___ -;t .
[ q(k;f) q(k;t) q(k;f) I
Now let us confine attention to those vectors k which maintain all
output-capital ratios constant at the initial levels X0; i.e. we consider
only vectors which satisfy, for some t, the condition
q(k; t)/ki = q(k;O)/kp, i = 1,. . . , n.
By hypothesis it is possible to find some vector k which satisfies the latter
equations for any given t, and since the function F is strictly concave in
k,,. . ., k,, any such solution k is unique. Consequently, we may write
k = k(k, t) as an alternative form of the imposed path restriction.
For this selection of vectors k, we have

and differentiating
F(k,l;O)=F 0 dk;O),
k , -.
q(kQ
this identity with respect to ko, we obtain
f
1

o q(k;O), t +
Fi(ko,l;O) = Fi k , -.
q&i t> I

+Fn+, ko dkO;O);t . a[dk; Wdk ; 01


dk;O I ak;
Using the fact that the functions Fi are homogeneous of degree zero and
substituting the path restriction ki = koq(k; t)/q(k; 0), we obtain

1; t) . GdkO;Wdk
Fi(kO, l;O) = Fi(k, 1; f)+F,+,(k,
01
c3k; -
The selection of vectors k guarantees that all output-capital ratios remain
unchanged; by hypothesis, therefore, all marginal products Fi also remain
constant. Since F,, 1 > 0 by assumption, we conclude that
41q(k;Wdk; 01 = oI i = I
,. . *, n,
i3k;
along any admissible path. That is, on all admissible paths the ratio
q(k; t)/q(k; 0) is a function only of t. This function we call u(t), and
we define
hi(f) = l/a(t), i = 1,. . ., n.
By the uniqueness of the solution k(k, t) obtained above, we see that,
given any initial point k, of all h,k, remain constant, then all output-
capital ratios Xi = q(k; t)/k, remain constant and equal to X7. Therefore,
the hypothesis of Theorem 1 implies that all xi are likewise constant
t, BURMEISTER AND DOBELL

along this path. Consequently the hypothesis of our lemma is satisfied,


and we may represent F in the factor-augmenting form.
Since the factor-augmenting form leads to bi/bi = fi/fi along a path
with Xi constant, and because we have seen that rental rates are constant
along such a path, the constancy of all hi(t) follows. This completes
Theorem 1.
THEOREM 2. Technological changeis Hicks-neutral on the above definition
if and only ifthe function F may be written in thefactor-augmenting form with
b,(t) = a(t), i = 1,. . ., n.
That is, the family of production surfaces represented by the function F
has the property that when all capital-labor ratios ki are constant, then
all profit shares xi are constant if and only if F may be represented in the
form F(K,,. . ., K,, L; t) = a(t)G(K,,. . ., K,, L).
Proof. (a) Sufficiency follows by direct calculation.
(b) (Necessity.) Define hi(t) = 1 for all i. Then it is clear that when all
hiki are constant, all ki are themselves constant, and by hypothesis all
Xi also. Thus the hypothesis of the theorem is met, and the factor-
augmenting representation is implied. Since by construction bi = hia = a
for all i, we have the special representation cited above. This completes
Theorem 2.
Because the labelling of inputs is arbitrary in this study of a single
production function, so-called Solow-neutral technical progress may
be viewed as one case of the above Harrod-neutral change (in which only
one factor is augmented).
[There is, however, also a further possibility when the num-
ber of factors exceeds two, as the following result indicates. (In
the following, the expression R is used to denote n/x defined along
some specified path.)
PROPOSITION. Let an idex set I c (1,. . . , n} be given. A pro-
duction function F may be representedin factor-augmenting form
with bi E 0 for i E I lf and only {f there exist functions hi(t) such
that when all hiki are constant, all 71i are constant and also
zi s 0 for all i E I. That is, if there exists a set (hi(t); i = 1,. . . ,n>
of positive functions h,(t) such that ail projit shares Iii and also
somecapital-output ratios Xi are constant along any path on which
all expressions h,k, are constant, then the function F may be
represented in factor-augmenting form, and for any i such that
l%?iz 0 along the designatedpath, also ai = 0. The converse also
holds.
DISEMBODIED TECHNOLOGICAL CHANGE 7

PROPOSITION. As a special caseof the above result, a production


function may be representedin a purely capital-augmenting form
(i.e. 2 E 0) if and only if there exist functions h,(t) such that when
all h,k, are constant, all zi are constant and also q is constant.
The above results indicate that the usual measures of
bias in technological change may be extended in an obvious way.
Defining the index of bias
l3=;
[l-&,
i=l 1
P

where P indicates that the partial derivative is taken along some


designated path, we might then say that technological progress
is labor-saving, neutral, or capital-saving as B is negative,
zero, or positive. The path restriction P determines the sense
in which the bias is defined.]
Thus we may consider Robinson-Uzawa-Phelps-type theorems in two
parts. The first problem is to express various economic definitions of
neutral technological change by path restrictions in terms of capital-labor
ratios. Once this is accomplished the representation lemma provides the
factor-augmenting form.

3. CONCLUSION
For each fixed t, the function F describes a conventional production
surface (giving Q as a function of the K,, . . . , K,, and L) in n + 1 indepen-
dent variables; as t changes, the function F generates a whole family of
such surfaces. In discussing technical progress, we suppose the entire
family to be completely known and attempt to describe some features of
the family itself, To assert, for example, that, Technical progress is
everywhere Hicks-neutral, is to assert that along any particular path
in the class of paths having all capital-labor ratios constant, factor shares
are likewise constant. In other words, the particular Hicks-neutral
family has the property that factor shares, which are in general functions
of the capital-labor ratios and the index t, are in fact functions of the
capital-labor ratios alone, independent of the index t. More generally, our
lemma states that technological change may be represented as factor aug-
menting if and only if there exists some class of such paths which may
be represented

where h,(t) is independent of all initial values k,(O), along which competi-
tive relative income shares are all constant. That is, the family of surfaces
represented by F must be such that the paths having all factor shares
8 BURMEISTER AND DOBELL

constant are proportional in the sense that at any time t all ratios
ki(t)/ki(O) are completely independent of the actual values of ki(t) and
k,(O). Thus, just as constant returns to scale is a restriction which
requires that as soon as we know one isoquant we know them all, factor-
augmenting technical progress is a restriction which requires that as
soon as we know one path on which all relative income shares are constant,
we know all such paths, simply as scale replicas of the path we have.
Standard categories of neutral technological change for the case of
several inputs follow directly by specifying the form of these paths.

REFERENCES
1. HICKS, J. R., Theory of Wages. St. Martins Press, New York (Second Edition), 1966.
2. PHELPS, EDMUND S., Golden Rules of Economic Growth. Norton & Co., New York,
1966.
3. ROSE, HUGH, The condition for factor-augmenting technical change. Economic J.,
December, 1968.
4. SOLOW, ROBERT M., Some recent developments in the theory of production, in The
Theory and Empirical Analysis of Production (Murray Brown, Editor), NBER
Studies in Income and Wealth. Columbia University Press, New York, 1967.
5. UZAWA, HIROFUMI, Neutral inventions and the stability of growth equilibrium.
Review of Economic Studies, February, 1961.
6. UZAWA, HIROFUMI, Production functions with constant elasticities of substitution.
Review of Economic Studies, October, 1962.

RECEIVED: July 31, 1968.

You might also like