MACRO-ECONOMIC THEORY

A Mathematical Treatment
 Some Other ELBS Low-Priced Editions

Allen MATHEMATICAL ANALYSIS FOR

ECONOMISTS Macmillan
Allen MATHEMATICAL ECONOMICS Macmillan
Bridger and FAMINE IN RETREAT? Dent
de Soissons

Dasgupta and COST-BENEFIT ANALYSIS Macmillan

Pearce
Giles MARKETING Macdonald & Evans
Hague MANAGERIAL ECONOMICS Longman
Hanson A DICTIONARY OF ECONOMICS
AND COMMERCE Macdonald & Evans
Hanson MONETARY THEORY AND PRACTICE Macdonald & Evans
Hanson TEXTBOOK OF ECONOMICS Macdonald & Evans
Hicks VALUE AND CAPITAL Oxford
University Press
Lipsey AN INTRODUCTION TO POSITIVE
ECONOMICS Weidenfeld
Meade THE THEORY OF INTERNATIONAL Oxford
ECONOMIC POLICY, Vols. I and II University Press
Marshall PRINCIPLES OF ECONOMICS Macmillan
Paish and Culyer BENHAM'S ECONOMICS Pitman
Prest PUBLIC FINANCE IN THEORY
AND PRACTICE Weidenfeld
Robinson THE ACCUMULATION OF CAPITAL Macmillan
Stafford MATHEMATICS FOR ECONOMISTS Macdonald & Evans
Stillwell, Lipsey WORKBOOK TO ACCOMPANY THE
and Clarke FOURTH EDITION OF 'AN
INTRODUCTION TO POSITIVE
ECONOMICS' Weidenfeld
Stone and Stone NATIONAL INCOME AND EXPENDITURE Bowes & Bowes
Stonier and Hague A TEXTBOOK OF ECONOMIC THEORY Longman
Thirlwall GROWTH AND DEVELOPMENT Macmillan
Second Edition
MACRO-ECONOMIC
THEORY

A M athelnatical Treatnzent

R.G.D.ALLEN

[Jj[5J
~
Palgrave Macmillan
 © R. G. D. Allen 1967

All rights reserved. No part of this publication

may be reproduced or transmitted, in any form
or by any means, without permission.

First edition 1967

Reprinted 1968, (with corrections) 1970, (with corrections) 1973

ELBS edition first published 1973

Reprinted 1975, 1979

Published by
THE MACMILLAN PRESS LTD
London and Basingstoke
Associated companies in Delhi Dublin
Hong Kong Johannesburg Lagos Melbourne
New York Singapore and Tokyo

ISBN 978-1-349-81543-2 ISBN 978-1-349-81541-8 (eBook)

DOI 10.1007/978-1-349-81541-8

The paperback edition of this book is sold subject to the condition that it
shall not, by way of trade or otherwise, be lent, re-sold, hired out, or other-
wise circulated without the publishers' prior consent in any fonn of binding
or cover other than that in which it is published and without a similar con-
dition including this condition being imposed on the subsequent purchaser.
 CHAPTER 13

Technical Progress
13.1 Various Forms of Technical Progress
STEADY-STATE growth can be generated, as we have seen, by capital
accumulation in conjunction with a growing labour force, even in the
absence of technical progress. But technical progress is an important and
perhaps the main factor making for growth. Technical progress simply
implies that increased output can be obtained over time from given
resources of men and machines. But it can take a variety of forms and our
purpose now is to define and distinguish some of the simpler varieties.
A main distinction is between ' disembodied' technical progress and
various 'embodied' forms. We start with disembodied technical
progress which applies equally and alike to all resources of men and
machines in current use. To use a well-known simile, such technical
progress represents technical know-how falling like manna from heaven.
Disembodied technical progress is here first incorporated in a simplified
model that takes a homogeneous commodity, for consumption and as
infinitely durable capital, and a smooth production function in which
output is obtained from continuously substitutable inputs of capital and
labour. No lag is assumed between output and income, both represented
by the variable Y. We write the production function Y =F(K, L, t) where
Y, K, and L are continuous variables over time, where F is a given
continuous and differentiable function, and where the variable t is
introduced explicitly to allow the whole production function to shift over
time.
The form of the production function F may be completely unspecified.
But it is often taken as linear and homogeneous in the variables K and L,
the case of constant returns to scale. Then, as particular cases, algebraic
forms with appropriate parameters may be taken for F, e.g. the fixed-
coefficients, Cobb-Douglas, or Constant Elasticity of Substitution (C.E.S.)
forms of Chapter 3. Later we can give up the assumption of a production
function, at least in explicit form, and replace it by such other constructions
as KaIdor's technical-progress function.
We turn next to technical change in some variety or other of embodied
 TECHNICAL PROGRESS 237
technical progress. This applies, not to the whole range of available
resources, but only to certain tranches of capital equipment, usually
machines produced and installed currently, together with the associated
labour crews. Capital is no longer assumed to be homogeneous. On the
contrary, capital becomes essentially a mixed stock of different' vintages '.
Machines of one vintage are different in kind from those of another;
because of embodied technical progress, new machines are more productive
than older (if otherwise similar) machines. Further, labour may be treated
in a similar way by giving up the assumption of a homogeneous labour
force and by taking men of different ' vintages', distinguished by age and
training. Men of the current vintage, e.g. those recently trained, are then
more productive than those of earlier vintages. The analysis can again
proceed with a production function, either of fixed-coefficient form or in
some smooth version, or in terms of a technical progress function as an
alternative construct.
Since we are exploring technical change in the context of steady-state
growth at a constant proportional rate, we assume generally that technical
progress proceeds at some given proportional rate. We write m for the
rate of technical progress, given by exogenous factors.

13.2 Neutrality of Disembodied Technical Progress

We look particularly for those forms of technical change that shift the
production function in such a way to leave undisturbed over time the
balance between capital and labour in current production. For this
purpose, we need to specify neutral technical progress in some precise
way, to represent technical change with no bias in the direction of either
capital-saving or labour-saving.
Consider the several ways in which disembodied technical progress can
be incorporated in a smooth production function Y =F(K, L, t) to make it
shift over time. We need to specify a neutral shift, to say precisely how it
is to be included in the function for neutral technical progress.
(1) Harrod-neutral technical progress is the first case to formulate.
It obtains when the production function shifts over time according to the
form:
Y=F(K,aL)
\'.:hereoc=oc(t) subjectto oc(t)=1 att=O
and oc(t»I, oc'(t) >0 for t>O.
This definition follows Joan Robinson's concept of an 'all-round increase
238 MACRO-ECONOMIC THEORY

in the efficiency oflabour ',t rather than Harrod's original (and equivalent)
definition which we shall derive later (13.3 below). We can write the
production function:
Y=F(K,L) where L=(X(t)L
and in effect we have a fixed form for the production function but with
the labour input L measured in efficiency units rather than the natural
units of the labour force, e.g. the number of man-years L. In exposition,
however, we generally refer rather loosely to ' men' as the natural unit in
which L is measured.
Hence we have two alternatives: either we can take a shifting produc-
tion function Y =F{K, (X(t)L} with the natural units for labour (men) as
well as for capital; or we can write a fixed production function Y =F(K, L)
but adjust the units for labour to vary over time. In the second alternative,
L is labour in efficiency units, connected with the natural units (men) by
means of the relation L = IX( t)L at each point of time t.
Technical progress of this form is labour-augmenting in the sense that
it is equivalent to a corresponding increase in the labour force. Since IX
increases over time, Y = (K, (XL) implies that a given output can be obtained
from a given capital input combined with a labour input L in men that
decreases as time goes on. Alternatively, Y =F(K, L) implies that a given
output can be obtained from a given capital input and the same labour
input L over time but measured in efficiency units. As time goes on, a
given labour force L in men represents an increasing number of efficiency
units because of technical progress (L =(XL); so the same input in efficiency
units represents a decreasing number of men. In Harrod-neutral technical
progress, one man at time t does (X times as much work as he did originally.
For example, technical progress may operate so that one man does as much
as two men used to do, then as much as three men, and so on.
In this form, technical progress proceeds at a proportional rate
dlX/lXdt, generally varying over time. If it is at a constant proportional
rate m, then
1 dlX
-IXdt
- =m with (X = 1 at t =O.

Hence: lX=eml •

Harrod-neutral technical progress at the constant rate m is given by:

Y =F(K, emtL)
t Joan Robinson, , The Classification of Inventions ',Review of Ecollomic Studie;
(February 1938), p. 140.
 TECHNICAL PROGRESS 239
or by Y=F(K,L) where L=emtL.
We are considering growth models in which we envisage Y and K as
growing at the same warranted rate of growth. Here we expect the output-
capital ratio Y/K to be constant over time, or at least to settle down to a
constant level in the limit. Harrod-neutral technical progress is of a kind
to ensure that this is possible. As we have seen, such technical progress
corresponds to an increase in the labour force. If a model without technical
progress has steady-state growth when the natural rate is n (rate of growth
of labour force), then the same model with Harrod-neutral technical
progress at rate m should have steady-state growth for a natural rate of
m + n (rate of growth of labour force and technical progress together). This
we shall find to be the case. t
(2) Solow-neutral technical progress is of the same form, except that
it is capital-augmenting instead of labour-augmenting. The shifting
production function is:
Y=F(IXK,L)
where IX = IX( t) subject to the same conditions as before. Alternatively:
Y =F(K, L) where K =IXK
in terms of a fixed production function but with capital measured in
efficiency units which vary over time.
The properties of technical progress in this form are similar to those of
the Harrod-neutral type with K and L interchanged. It follows, however,
that it is not an appropriate form of technical progress for growth models
in which the output-capital ratio is constant, at least as long as the technical
change is of the disembodied kind now considered. It becomes more
relevant, and was in fact so considered by Solow,! to models of the vintage
type in which technical change is embodied in successive vintages of
machines that become more efficient as time goes on.
(3) Hicks-neutral technical progress is the form of disembodied
technical progress that appears to conform most naturally to the idea of
neutrality, and it was the first type to be proposed.§ It obtains when the
t Indeed it can be argued (see F. H. Hahn and R. C. O. Matthews, ' The Theory
of Economic Growth: A Survey', Economic Journal (December 1964), p. 829) that
technical progress which is not Harrod-neutral is inconsistent with steady-state
growth and a constant output-capital ratio. For example, a capital-saving bias
must lead to a running-down in the growth of output.
t R. M. Solow, Capital Theory and the Rate of Return (1963).
§ J. R. Hicks, The Theory of Wages (1932), first gave the definition, not in this
form, but in equivalent terms (see Exercise 13.1).
240 MACRO-ECONOMIC THEORY

production function shifts over time by a uniform upward displacement of

the whole function:
Y=rxF(K,L)
where rx=rx(t) suhject to the same conditions as before. If the inputs K
and L, in their natural units, remain unchanged, then a product is obtained
that increases over time at the proportional ratedrx/rx dt. This is 'manna
from heaven' in the most obvious sense.
If the fixed production function F is linear and homogeneous in K and L,
the case of constant returns usually taken, then Hicks-neutral technical
progress is a straightforward combination of the other two types proceeding
at the same rate; it is equally labour- and capital-augmenting. The linear
and homogeneous property is that F(rxK, rxL)=rxF(K, L) for any rx>O.
Hence the shifting production function in the Hicks-neutral, constant-
returns case is:
Y =F(K, L) where K = rxK and L = rxL.
It follows again that neutrality in this Hicks sense is not the appropriate
concept for a steady-state growth model with disembodied technical
progress. It is not consistent with a constant output-capital ratio.

13.3 Constant Returns

The case to be explored in greatest detail is that in which the production
function is subject to constant returns. It is then oflinear and homogeneous
form so that, apart from technical progress, the function Y =F(K, L) can
be reduced to the two-dimensional form in per capita terms: y = f(k) where
y= Y/L and k=K/L. We consider disembodied technical progress with
particular reference to the Harrod-neutral type at a constant rate m.
The features of the case of constant returns that make the analysis so
much easier are examined in 3.4 and 3.5 above. The production function
y=f(k), relating output per head to capital input per head, can be repre-
sented as a curve on the (k, y) plane, as illustrated in Fig. 13.3A in the case
of the well-behaved function (3.4 above). The two marginal products are:

OY =f'(k) and OY =f(k) -kf'(k)

oK oL
to be equated under perfect competition to the profit rate p and the wage
rate w respectively. The product then distributes: y = pk + w, as repre-
sented by ON = WN + OW in Fig. 13.3A, the profit rate (marginal product
of capital) being interpreted as the slope of the tangent at P.
 TECHNICAL PROGRESS 241
We now add technical progress at the Harrod-neutral rate m. The
shifting production function is:
Y=F(K, erntL)=F(K, L) where L=emtL.

Write: y=!"= Ye_mt=ye- mt}

~ ~ ........................... (1)
and k = - = - e-mt =ke-mt
L L
where y and k are output and capital per head y and k are output and
capital per labour efficiency unit. SinceF(K, L) is linear and homogeneous
in K and L, we write Y =F(K, L) as y=f(h). Hence the production
function in the case of constant returns and Harrod-neutral technical
progress at rate m is:
y = f(h) where y = ye-rnt and k = ke-mt .................. (2)
or alternatively:
y = ern tf(ke-rnt) . ................................. (3)
The first of these representations has f fixed and the variables y and k in
units that vary over time as shown by (1). The second shows a shifting
function and variables y and k in fixed (natural) units.
y

M k
FIG. 13.3A

An essential property of Harrod-neutrality in the case of constant returns

is now derived. At any time t, the marginal and average products of
capital are:

and
242 MACRO-ECONOMIC THEORY

Hence if the marginal product iJ Y/iJK remains constant over time, then k is
constant and so is the output-capital ratio Y/K. The converse also holds.
Hence the result:
If technical progress is Harrod-neutral with constant returns to
scale, then the marginal product of capital (iJY/iJK) remains
constant over time whenever the output-capital ratio (Y/K) is
constant over time.
Moreover, if the distribution of income is determined under perfect
competition on the basis of marginal productivity, the marginal product of
capital is the profit rate p at any time t. The result then becomes:
If technical progress is Harrod-neutral under perfect competition
with constant returns to scale, then the profit rate p remains
constant over time wherever the output-capital ratio is constant
over time.
This property is, in fact, the one used as a definition of neutral technical
progress by Harrod.t Further, Harrod would describe non-neutral
technical progress as being either labour-saving if a constant p over time
is associated with a falling output-capital ratio, or capital-saving if with a
rising output-capital ratio.
The production function under constant returns with Harrod-neutral
technical change is represented either by (2) or by (3). There are two
corresponding, and alternative, methods of diagrammatic representation in
two dimensions. Take the form (2) and draw the diagram Fig. 13.3A with
k and y measured along the axes instead of k and y. The variables of the
diagram are then capital and output per labour efficiency unit. The unit
varies over time with technical progress. The variables k and y are
expressed in per capita terms by discounting for technical progress up to
time t:
k=ke-mt and y=ye-mt•
All the variation over time, due to technical progress, appears in k and y.
The production function itself remains fixed, i.e. the curve of Fig. 13.3A
t R. F. Harrod, in a review of Joan Robinson's essays of 1937, Economic Journal
Gune 1937), pp. 328-9, and in Towards a Dynamic Economics (1948), pp. 22-27.
Joan Robinson has pointed out that Harrod's definition is equivalent to an ' all-
round increase in the efficiency of labour', i.e. to the definition taken here (13.2).
We have now established that our definition implies (is a sufficient condition for)
Harrod's property. It can also be established that our definition is a necessary
condition for Harrod's, that the two are equivalent. See H. Uzawa, 'Neutral
Inventions and the Stability of Growth Equilibrium " Review of Economic Studies
(February 1961).
 TECHNICAL PROGRESS 243
does not shift over time. From the result just obtained, the essential
property of Harrod-neutral technical progress is that, as long as k remains
unchanged as at M of Fig. 13.3A, then the marginal product of capital
(equal to the profit rate under perfect competition) is unchanged over time,
shown by the slope of the tangent at P; and the output-capital ratio is also
unchanged over time, shown by the slope of OP.
Now take the alternative representation of the production function (3)
and draw a diagram Fig. 13.3B in which k and yare measured along the

/"/ PI
/
/
/
/
/

o k
FIG. 13.3B

axes, i.e. capital and output per man in natural units fixed over time. Then
the curve of the production function shifts over time, according to the
variable t in the form (3), as shown in Fig. 13.3B where two positions at tl
and t2 (tl <t2) are represented. The essential property of Harrod-neutral
technical progress now appears rather differently. If PI and P 2 are the
points on the two curves which correspond in the sense that they have the
same output-capital ratio (YjK =yjk), then PI andP2 lie on the same radius
from O. The property of Harrod-neutral technical progress is that the
profit rate is the same at PI and P2 , i.e. the marginal product of capital is
unchanged and the slope of the tangent at PI is the same as the slope of the
tangent at P 2• Hence as time goes on, from tl to t2 for example, as long as
the output-capital ratio is unchanged, then the tangents to the shifting
production curve remain parallel and the profit rate (slope of tangent) is
unchanged.
The wage rate can also be read off the diagrams in alternative forms.
The wage bill is W = wL = wL, in terms either of the wage rate per man w
244 MACRO-ECONOMIC THEORY

or the wage rate per labour efficiency unit w. Since L =emtL, it follows
that:
w=we-mt
which corresponds exactly to (1) for y and k. For a shifting production
function under Harrod-neutral technical progress, suppose that the
output-capital ratio remains constant and hence the profit rate remains
constant. Then, from Fig. 13.3A (with k and y as the variables), the wage
rate w per efficiency unit is given by 0 Wand w also remains constant.
Then w=wemt , as the wage rate per man, increases over time. This is
shown in Fig. 13.3B, where the wage rate per man w is given by the
intercept on Oy; the value of w increases from OWl to OW2 from time tl
to time t 2 •
Take the shifting production function in form (3) and, for brevity, write
I and I' for the function and its derivative at the value k =ke-mt at any
time t. Take logarithms: logy =mt + log I, and differentiate:

! dy = m +L dk where dk = (dk _ mk)e-mt•

ydt I dt dt dt
On substituting/=ye-mt from (3), we get:

! dy = m+I' ~ (! dk _ m)
ydt y kdt

I.e. }t = m( 1-/, J) +I'J G~:)· ..................(4)

This is a linear relation between the two growth rates, dy/ydt and
dk/kdt, permitted by the shifting technology. The coefficients, however,
are not constant in (4); they depend explicitly (through l' valued at
k = ke-'Tnt) on time t. Hence the linear relation shifts over time.
We can ask the question: under what circumstances does the shifting
technology permit the growth rates in the per capita variables y and k to be
equal and constant over time? In (4), try dy/ydt=dkjkdt=>", constant
for all t. Then: >..=m[1-j'(kjy)]+j'(kjy)>... This gives >"=m, provided only
that 1 - j'(kjy) ¥- O. The proviso is yjk ¥- j'(ke-mt ), i.e. y/k ¥- j'(k); in the
diagrammatic terms of Fig. 13.3A, we require that the slope ofOPis not equal
to the slope of the tangent at P. For a well-behaved function, as illustrated in
the diagram, this is true everywhere since the slope of OP is greater than
the tangent slope. Hence we conclude that, at least for the kind of well-
behaved production function we assume, the technology permits output
per head and capital per head to grow at the same rate, the common value
 TECHNICAL PROGRESS 245
being the rate (m) of technical progress. It then follows that, if the labour
force grows at the given rate n, then output and capital can grow at the
same rate IL = m + n, constant over time. This is the natural rate of growth,
simply the sum of a given rate (m) of technical progress and a given rate (n)
of growth in the labour force. So:
If technical progress is Harrod-neutral and if there are constant
returns to scale, then the technology allows output and capital
to have a common steady-state rate of growth, the natural rate
j.I. = m + n obtained by adding the rate of technical progress and
the rate of growth of the labour force. Output or income per
head then grows at the same rate (m) as technical progress.
It is this result that makes the concept of Harrod-neutral technical progress
relevant to models of steady-state growth. It is an expression of the fact
that such technical progress is labour-augmenting, equivalent to an
addition to the labour force.

13.4 Case: Fixed Coefficients

Our first particular case of a production function is that with fixed
coefficients (u and v). We have then the special case where there are
constant returns but no substitution between capital and labour inputs.
The concepts of neutral technical progress, though developed in 13.2 for a
smoothly substitutable production function, are easily applied to the fixed-
coefficients case.
Harrod-neutral technical progress. In general, the shifting production
function is:
K L
y=-=- whereL=cx(t)L
v u
and where cx(t) is subject to the same restrictions as before (13.2 above).
However, we take technical progress at the given (proportional) rate m so
that cx( t) = e"":
K L
Y =-=- where L=e"'tL.........................(1)
v u
Taking (1) as our shifting technology, we are in a position to introduce
technical progress both into the demand-oriented models of capital
accumulation (Chapter 10) and into the Harrod-Domar growth models
(Chapter 11). These models are based on a production function with fixed
coefficients, now shown as (1).
246 MACRO-ECONOMIC THEORY

The demand-oriented models are completely unaffected by technical

progress, except for an obvious change in the demand for labour carried
forward to the labour market. The equilibrium conditions (K =V Y and
DK =s Y) are unchanged since the first comes from (1) whatever the rate of
(Harrod-neutral) technical progress. Equally the disequilibrium model
used to assess stability of steady-state growth is unchanged. The equili-
brium solution is still unique: growth in output (and capital stock) at the
warranted rate g = sIv; and the growth is stable or unstable under exactly
the same circumstances as before. All that is changed is that the demand
for labour (given output Y) is L=uY, measured in efficiency units. To
get the demand in natural units (men) to set against the labour force, we
write L = Le-mt and so
L = (ue- mt ) Y. .. ............................... (2)
The demand for labour is reduced by the existence of technical progress.
This is again a reflection of the fact that the type of technical progress
assumed is labour-augmenting.
The application to the Harrod-Domar model of growth is immediate.
The demand for labour (2) is set against the given labour force, assumed to
grow at the constant rate n, in the equilibrium condition of full employment:
L=(ue-mt)Y=Loent
I.e. u Y = Loe"t where I-' = m + n.
Here I-'=m+n is the natural rate of growth in the model. On the other
hand, in equilibrium, output Y grows at the warranted rate g =s/v. For
steady-state growth, the parameters of the model must satisfy the condition:
slv=m+n .................................... (3)
which expresses the need for agreement between warranted and natural
growth rates. There are now four parameters involved. When m = 0, the
case of no technical progress, we re-obtain the familiar condition: sIv = n.
When m>O, the condition (3) is more likely to be satisfied for reasonable
values ofthe parameters sand v. If s is between 0·1 and O· 2 and if v ranges
from 2 to 4 on annual data, then the natural rate I-' = m + n needs to be in the
range from 2-!% to 10% per year. This is quite possible, even for slow
rates of technical progress and of growth in population. To summarise:
If the technology has fixed coefficients and Harrod-neutral technical
progress at the given rate m, then the warranted rate of growth
g=sjv in a Harrod-Domar growth model is independent of m,
 TECHNICAL PROGRESS 247
and the condition for steady-state growth when the labour force
grows at the given rate n becomes g = . . where .... = m + n, i.e. the
four parameters need to conform to the condition slv = m + n.
We see, in short, that technical progress of the labour-augmenting type fits
easily into the framework of growth models based on fixed production
coefficients.
Solow-neutral technical progress at rate m. The shifting production
function is:
Y=~=~
v u
where K=emIK.........................(4)

The immediate consequence now is that steady-state growth is no longer

consistent with the Harrod-Domar models extended to include technical
progress of type (4). We look for a solution of the models in which the
rates of growth of output Yand capital stock K are both constant over time
and equal. The relation between Y and K permitted by the production
function (4) in logarithmic form is:
log Y = mt + log K - log v
1 dY 1 dK
and so: Ydt=m+ K dt'
Hence the rate of growth of output must exceed that in capital stock by the
extent of technical progress. This is a simple reflection of the capital-
augmenting nature of this form of technical progress. The actual equili-
brium rates of growth are to be found by solving the equations of the
product market (see Exercise 13.4) and they confirm that, though the rates
vary over time, their difference is a constant m.
The conclusion is that, at least in this case of fixed coefficients, technical
progress of a capital-augmenting (Solow-neutral) type is not consistent
with steady-state growth. The same conclusion is reached for technical
progress of the Hicks-neutral form, giving a shifting production function
(with fixed coefficients) that is a combination of (1) and (4).
Capital-augmenting technical progress of linear-trend form. A
rougher approximation to technical progress in capital-augmenting (but
not in neutral) form is provided by a shifting production function:
K+mt L .
Y=--=- where m>O IS constant................ (5)
v u
248 MACRO-ECONOMIC THEORY

This is a form suggested by Goodwin. t It is a type of technical progress

that can be fitted into the framework of a Harrod-Domar growth model
more easily than the Solow-neutral form. The equilibrium conditions of
the product market become from (5):
K=vY -mt and DK=sY
giving: sY=D(vY -mt)=vDY -m.
Hence the differential equation for output Y is:
vDY -sY=m
with solution: Y= -~+( Yo+ ~)eg, where g=slv ................... (6)
Hence output Y, in equilibrium, grows at the warranted rate g=slv,
independent of the parameter m for technical progress. The stationary
level of output is no longer Y =0 (corresponding to autonomous expen-
diture ignored); it is Y= - mIs, dependent on the parameter m.
The corresponding equilibrium paths of capital stock and investment
follow from the product-market conditions. The initial values are Ko = v Yo
and 10 =gKo =sYo- The paths to match (6) for output are found to be:

K= -mt-i+(Ko+i)e g, and 1= -m+(Io+m)ell ' •••••••(7)

Hence both K and I grow at the warranted rateg=slv, except that K also
has a downward trend term ( - mt) arising from the capital-augmenting
technical progress.
The solution (6) and (7) has a simple interpretation. The variables Y,
K+mt, and I all grow at the warranted rateg=s/v. The stationary levels
are Y = - mIs, K + mt = - mIg, and 1= - m; if these values obtain initially,
then they are maintained for all time. The clue is seen in the stationary
level of investment. Because of the capital-augmenting technical progress,
there arises an autonomous disinvestment of amount m. When this
(negative) autonomous expenditure is multiplied up, output Y is decreased
by mIs, as shown in (6). The interpretation of the path of output is
precisely that of the simple growth model of 10.3 above when autonomous
expenditure is taken as A = - m.

13.5 Case: Cobb-Douglas Production Function

The Cobb-Douglas form of the production function with constant
returns to scale can be written with a single parameter at (O<at<l):
t R. M. Goodwin, ' The Nonlinear Accelerator and the Persistence of Business
Cycles'. Econometrica Ganuary 1951).
 TECHNICAL PROGRESS 249
Y = KrzV-. A multiplicative constant is here absorbed into the units for
measuring the homogeneous commodity in output and capital (3.7 above).
When there is technical progress, the production function shifts over
time. The obvious case to consider, as representing neutral technical
progress, is:
Y=AKrzLl-rz
where A=A(t) subject to A(t)=1 at t=O
and A(t» 1, A'(t»O for t>O.
For technical progress proceeding at a given proportional rate, we are led
to write:
Y =eAtKrzV-rz for some constant A>O. •.............. (1)
It remains to interpret A.
We take each of the types of neutral technical progress in turn, using the
definitions of 13.2 and applying them to the Cobb-Douglas function.
Harrod-neutral technical progress at rate m:
Y=KrzV-rz where L=emtL
I.e.
So:
which is (1) with A=m(1 - IX).
Solow-neutral technical progress at rate m:
Y = KrzLl-rz where K =emtK
I.e.
So:
which is (1) with A=mrx.
Hicks-neutral technical progress at rate m:
Y = KrzLl-rz where K =emtK and L =emtL
I.e.
So:
which is (1) with A=m.
Hence the form (1) of the shifting Cobb-Douglas production function is
250 MACRO-ECONOMIC THEORY

at the same time Harrod-neutral, Solow-neutral, and Hicks-neutral. t All

that needs to be done, to distinguish the three cases, is to interpret the
parameter ,\ appropriately in terms of the constant rate m of technical
progress. The result is:
The Cobb-Douglas production function (1) represents neutral
technical progress at the rate m where
,\=m(1-oc) for Harrod-neutrality
'\=moc for Solow-neutrality
and '\=m for Hicks-neutrality.
Hence we can use the Cobb-Douglas form as a particular production
function for anyone of the three types of neutral technical progress. We
need only specify which and interpret the parameter'\ of (1) accordingly.
The Cobb-Douglas production function under constant returns in per
capita form is:
Y K
y = k" where y = Land k = L .
Take technical progress at the Harrod-neutral rate m, so that (1) is the
production function with ,\ = m( 1 - oc). Hence, provided that technical
progress is Harrod-neutral, the Cobb-Douglas form in per capita terms
can be written as:
y=k« wherey=ye-mt and k=ke- mt }
.................. (2)
or as: y=em(l-«)tk«
The Cobb-Douglas form (2) is of the well-behaved type, as illustrated in
Fig. 13.3A above.
The relation between the growth rates in the per capita variables y and k
permitted by the Cobb-Douglas form (2) is obtained by logarithmic
differentiation:
logy=m(l- oc)t + oc log k

giving: ~ t =m(1 - oc) + ocG ~:) ......................... (3)

The relation is linear, as in the general case of constant returns, equation (4)
of 13.3 above. The simplification here is that the coefficients in the relation
t It is well known that the ollly production function which is both Harrod- and
Hicks-neutral is the Cobb-Douglas form. See H. Uzawa, Review of Ecollomic
Studies (February 1961), pp. 120-1.
 TECHNICAL PROGRESS 251
are constant, given by the two parameters m and tx. On substituting
dy/ydt=dk/kdt='A in (3), we get 'A=m(l-tx)+tx'A, so that 'A=m.
Again output or income per head and capital per head can both grow at the
steady-state rate m of technical progress.

13.6 Non-Constant Returns

The analysis is more complicated if there are not constant returns to
scale. The shifting production function is Y =F(K, L) where L =emtL and
where technical progress is Harrod-neutral at rate m. The function F is
fixed but, since it is not linear and homogeneous, it is not generally possible
to switch from three variables to two variables in per capita form.
The analysis is manageable when F is of Cobb-Douglas form:
Y =K«LfJ without technical progress.
Then tx + f3 > 1 represents increasing returns, tx + f3 = 1 constant returns,
tx+ f3 <1 decreasing returns. With Harrod-neutral technical progress at
rate m:
Y =K«(emtL)fJ
i.e. Y =emfJtK«LfJ .................................. (1)
Suppose that the labour force L grows at the constant proportional rate n,
so that dL/Ldt=n. By derivation from (1) in logarithmic form:
log Y = mf3t + tx log K + f3 log L
1 dY 1 dK
we get: Ydt =mf3+tx Kdi+f3n
1 dY 1 dK
I.e. Ydi=(m+n)f3+ tx l( dt ......................... (2)

This is a fixed linear relation between the rates of growth of output and
capital, given the rate of growth n of the labour force and the rate m of
technical progress.
Put the two rates of growth equal, dY/Ydt=dKjKdt='A in (2):
'A = (m + n)f3 + tx'A.
Hence there is a unique rate at which both Y and K can grow:

'A=(m +n) 1 ~ ex = /!:ex where fL=m +n ............... (3)

provided that tx < 1 (which is usually assumed). The common rate is

fL = m + n under constant returns (ex + f3 = 1). In this case, if Y and K grow
252 MACRO-ECONOMIC THEORY

at the rate (m+n) whileL grows atthe rate n, theny= Y/L andk=K/L both
grow at the rate m of technical progress. This confirms the results of 13.5
above.
In the case of increasing returns (IX +f3 > 1), the common rate of growth
of Y and K given by (3) is greater than /L. Hence increasing returns tend
to supplement the other two factors (technical progress and population
growth) in producing steady-state growth in output and capital. This is
the well-known case of external economies.
In the case of decreasing returns (IX + f3 < 1), the common rate of
growth now possible in Y and K by (3) can be written:

(m+n) _1 ~::f3 (m+n)

and the corresponding rate of growth in output per head is:
l-IX-f3
m- I-IX (m+n).

The effect of decreasing returns is to offset the effect of technical progress

in raising income per head.

13.7 Technical Progress Functions

The Cobb-Douglas production function, with constant returns and
Harrod-neutral technical progress, implies a fixed linear relation, (3) of
13.5 above, between the rates of growth of output per head and capital per
head. If a more general production function is taken, the relation, (4-) of
13.3 above, remains linear but the coefficients are not fixed, being dependent
on the capital-output ratio (k/y) and varying over time.
Kaldor adopts a different formulation in defining his Technical Progress
Function.t This is effectively another generalisation of the fixed linear
relation between rates of growth given in the Cobb-Douglas case. Instead
of generalising the form of the production function, Kaldor dispenses with
the production-function concept (at least explicitly) and assumes instead a
relation, generally non-linear, between the rates of growth of output per
head and capital per head:
! dy =F(! dk).
ydt ,kdt
tN. Kaldor, 'A Model of Economic Growth', Economic Journal (December,
1957). This is the formulation given here. Later, in N. Kaldor and J. A. Mirrlees,
, A New Model of Economic Growth', Review of Economic Studies (June 1962),
another formulation is given in the context of vintage capital.
 TECHNICAL PROGRESS 253
Here the Technical Progress Function F is assumed to be well behaved:
F'(x) > 0, F"(x) <0, F'(x)->O as x~ 00,

where x=dk/kdt. If in addition F(x)~O at x=O, as illustrated by the

form of the Technical Progress curve of Fig. 13.7 then there is a unique rate
of growth of capital per head, given by M, which has an equal rate of growth
of output per head to correspond. The essential feature of Kaldor's
formulation is that the relation between dy/ydt and dk/kdt, and their
unique common value, are independent of the output-capital ratio y/k.

I dy
y,k

o M I dk
k dt
FIG. 13.7

As a generalisation of the Cobb-Douglas case, the Kaldor Technic:l!

Progress Function has the property:
The Technical Progress Function is linear if and only if there is a
production function of Cobb-Douglas form with constant
returns and Harrod-neutral technical progress at rate m.
Proof. For the sufficient (if) condition suppose there is a Cobb-Douglas
production function y=em(l-a)tk a • Then by (3) of 13.5 there is a fixed
linear relation between dy/ydt and dk/kdt. The Technical Progress
Function is linear. To prove the necessary (only if) condition: suppose
that the Technical Progress Function is linear:
1dy Idk
--=a+ex--
ydt kdt
where a>O and O<ex<l if there is to be a unique point P in Fig. 13.7.
d
Hence dt (log y - ex log k) =a
I2
254 MACRO-ECONOMIC THEORY

I.e. ~ (log 1'

dt k'
) =a
i.e. y=Aeatk a (A arbitrary constant).
This is a Cobb-Douglas production function, with constant returns, and
with Harrod-neutral technical progress at rate m =a/(1 - ex). Proved.
The result raises the question: if the Technical Progress Function is
non-linear, does it imply a production function of form other than Cobb-
Douglas. The answer is that the non-linear Technical Progress Function
is not generally integrable and there is no production function lying
behind it.t

13.8 Embodied Technical Progress

We turn finally to the situation that arises when technical progress does
not drop like manna from heaven on all men and machines but only on
certain types of capital equipment and on certain sections of the labour
force. The assumption that suggests itself is that technical progress is
embodied in the current vintage of machines being installed (as opposed to
existing machines) and/or in the currently-trained men of the labour force
in contrast to those trained at earlier times. Broadly speaking, we can say
that it is the age of men and machines which we have to distinguish.
However, the more important feature of embodied technical progress
appears in the various vintages of capital equipment rather than in the
training of the labour force. For simplicity, therefore, we confine our
exposition to the case of technical progress embodied in vintage capital.
It is quite possible to adjust the analysis to the more general case in which
both the labour force and the aggregate stock of capital are split up into
vintages. Indeed, when vintage labour is linked directly to vintage
equipment, the following analysis applies at once to the combination of
vintage capital and labour.
The assumption of a homogeneous capital stock is relaxed by taking the
stock as made up of machines of different vintages. Each vintage consists
of a homogeneous set of machines produced for installation at one time;
successive vintages relate to a sequence of times. We need two time
variables, one (t) for time in the usual sense and the other (T) for the dating
of vintages of machines in use at time t. In a discrete analysis, with equal
spacing of vintages (say each year), the machines in use at time t are of
vintages T=t (new), t-1 (one year old), t-2 (two years old), .... In a
t See J. Black, 'The Technical Progress Function and the Production Function',
Economica (May 1962).
 TECHNICAL PROGRESS 255
continuous analysis, adopted here, there are machines of all vintages T~t
in use at any time t.
However, machines may be subject to physical deterioration (deprecia-
tion) and/or to economic deterioration (obsolescence). In the latter case,
as the machines get older, their quasi-rent in the Marshallian sense falls
and eventually becomes zero, when the machines are scrapped. In order
to simplify the analysis, we assume that there is no depreciation, i.e. that
machines are infinitely durable. This is not an important restriction and
it is generally possible to introduce, quite easily, a term for depreciation in
vintage models. On the other hand it is essential to investigate obsolescence
and to find the economic life T of machines of a particular vintage as a
variable of the models considered. Generally machines in use at time tare
of vintages T, where t - T~T~t.
The assumption is that technical progress proceeds at a given propor-
tional rate m, but falls like manna from heaven only on new machines.
Once installed, the technical features of machines remain frozen, later
technical progress passing them by. At time t, machines of vintage T have
benefited from technical progress up to time T but not thereafter.
Passing to the production function appropriate to machines of different
vintages, we must distinguish between substitutability between machines
and labour before the installation of new machines on the one hand, and
that after the machines are installed on the other hand. When machines
are being designed and the decision taken to install, we may assume
substitutability according to a smooth production function. In general,
the function will vary from one vintage to another. At time T, when
machines of vintage T are new, the production possibilities apart from
technical progress can be shown by:
QT =FT(K., LT)
where KT is the number of machines, while LT and QT are labour inputs
and product outputs in the same units as before. Technical progress, in
Harrod-neutral and in Solow-neutral form respectively, can be included:

QT=Fi~T' LT) where ~ =emTLT} .................. (1)

or QT=FT(KT,LT) where KT=emTKT
Simplifying assumptions are now needed in the interests of manageable
analysis. First, the same production function F is taken for all vintages;
second, the Cobb-Douglas form with constant returns and a single
parameter ex is taken for F. So:
QT=eATK~~-a. ................................. (2)
256 MACRO-ECONOMIC THEORY

gives the output Q, from a labour input L, with a number K, of machine~

of vintage 'T, when new. The coefficient A, allowing for technical progres:-
at rate m (up to time 'T), is to be interpreted (13.5 above) as A=m(1 - 7
when the Harrod-neutral case of (1) is taken, or as A= ma when the
neutrality is of Solow form.
It is a different question how far machines and labour can be substituted
at any time t > 'T after the machines of vintage 'T are installed. At time t = 'T,
according to the conditions of the model considered, a decision is taken to
use a labour crew of L, with K, machines of the new vintage, the output
Q, being given by (2). At any time t >'T, two alternative cases suggest
themselves: either substitution between machines and labour continues
according to a production function, Qt=eA'KtL~-", similar to (2), or the
K, machines installed at time 'T continue to be used with the same labour
crew L, that goes with them. On the first alternative, the employment of
machines of vintage 'T and the labour used with them can and does vary
over time. On the second alternative, once the machines are installed, they
remain for all time (until scrapped as obsolescent) with the same team of
men and producing the same output.
Using the graphic terms introduced by Phelps,t we can distinguish the
two alternative cases as putty-putty when there is substitution before and
after installation, and as putty-clay when there is substitution before but
fixed labour requirements after installation. There is a special case of the
latter in which the capital-labour ratio is fixed both before and after
installation, the clay-clay case. The production function (2) is then
replaced by one of fixed-coefficients form (13.4 above).
Apart from the usual conditions for equilibrium, two additional matters
come up for consideration in constructing models with vintage capital.
One is the determination of the variable economic life T of machines of any
vintage and its relation to steady-state growth. The other is to see what
measure (if any) can be assigned to the total stock of capital. We have a set
of machines of different vintages at any time t, i.e. K, machines of vintage
'T for t - T~'T~t. It is a matter for investigation whether the set can be
aggregated into a capital stock K at time t.
The analysis envisaged here is based on the assumption of a smooth
production function, at least before installation of machines of vintage 'T,
incorporating embodied technical progress up to time T, either in the
rather general form (1) or in the more manageable Cobb-Douglas form (2).
As with disembodied technical progress (13.7 above), there is the alternative
t E. S. Phelps, ' Substitution, Fixed Proportions, Growth and Distribution',
International Economic Review (September 1963).
 TECHNICAL PROGRESS 257
of replacing the production function by some form of Technical Progress
Function. Such a function can relate the rate of change of output per man
from new machines either to investment (in new machines) per man or to
the rate of change of investment per man.
Whichever method of analysis is followed, capital and investment is in
terms of numbers of machines of various vintages. Hence it is gross
investment in new machines which is relevant to the analysis, in contrast
with the net investment 1= dKJdt appropriate to the analysis with homo-
geneous capital. To get from gross investment to capital stock K, if it can be
measured, is still a further step to be taken in a vintage analysis with
embodied technical progress.

EXERCISES
13.1 Show that the constant-returns production function with Hicks-neutral
technical progress (as defined in 13.2 above) can be written 51 = f(k), and
interpret k and y. Obtain expressions for the marginal products 0 YloK and
oYloL and show that the ratio depends only on k. Hence establish that, in
the case of constant returns, our definition of Hicks-neutral technical progress
implies Hicks' original definition (ratio of marginal products unchanged when
capital employed per head is unchanged).
13.2 Represent the constant-returns production function with Hicks-neutral
technical progress at the rate m on a diagram similar to Fig. 13.3B. Use the
property of the previous exercise to show that, at a fixed capital per head (k),
the tangent to the shifting curve always passes through a fixed point on Ok.
13.3 Write the constant-returns production function with Solow-neutral technical
progress at the rate m as y = f(ke mt ) and deduce that the linear relation similar
to (4) of 13.3 above is:
! dy =/' kemt
ydt y
(m +!k dt~).
For the Cobb-Douglas form, show that
! dy = mod a(! c{k)
y dt k dt '
with constant coefficients.
13.4 Consider the Harrod-Domar model with fixed coefficients and Solow-neutral
technical progress at the rate nl. From the product market (equilibrium)
conditions, K=ve-mty and DK=sY, obtain the varying rates of growth:
(1 IK)DK =ge mt and (1 I Y)D Y =nl + ge mt . What is the general nature of the
equilibrium paths of Y and K?
13.5 When the production function is Y=(K +mt)fv=Llu, write the product-
market conditions in the Harrod-Domar model, eliminate Y, and obtain an
equation for K. Solve to get: K + mt = - mIg + (Ko + mlg)egt where Ko =V Yo.
Hence write an expression for investment I at any time.
13.6 In the model of the previous exercise, the demand for labour is u Y where
Y = - mls + (Yo + mls)egt . Add the full-employment condition and show
that, for a sensible result, labour supply L (like output Y) must grow from
some stationary level. Write L = A + Bent and show that appropriate constants
258 MACRO-ECONOMIC THEORY

are A = - mu/s and B =Lo +mu/s. Is this treatment needed in all Harrod-
Domar models in which there is a constant autonomous investment?
13.7 Take Y=e At KaL{3 and assume that L grows at the given rate n. Show that
the relation between growth rates of the per capita variables is
! dy = A + ('" +,8 -1)n +
ydt
"'(!
dk).
kdt
Put fJ = 1 - '" and A = m(1 - ",) and check (3) of 13.5 above. Why is this latter
result independent of n?
13.8 Show that the production function of the previous exercise permits y and k
to grow at the same rate v where v = [A/(1 - 0:)] + [(0: +fJ -1)/(1 - o:)]n. Deduce
that. when there are increasing returns (0: +fJ> 1), then v> oeven when A=0, i.e.
that increasing returns can result in rising income per head even in the absence
of technical progress.
13.9 Show that the C.E.S. production function with Harrod-neutral technical
progress at the rate m can be written in per capita terms:
y =emt {a(ke- mt )-{3 + 1}-1/{3
and that:
{a(ke-mt)-.a + I} (!y dt1)~)
m +a(ke-mt)-.a(! dk).
=
kdt
What common rate of growth in y and k is possible?