Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
Wiley and Royal Economic Society are collaborating with JSTOR to digitize, preserve and extend access to The
Economic Journal.
http://www.jstor.org
y+aq,
I,
(2)
892
THE
ECONOMIC
JOURNAL
[DECEMBER
in any situation in which e is either zero or a constant there mustbe a perfect correlation between p and q - but one which does not assert anything, since it is the
automatic consequence of measuring the same thing twice over.'
In this way I found that in at least two sectors, Agriculture and Commerce,
the Verdoorn equation did not produce meaningful results. In each case the
regression coefficient of p on q was around I, with R2s between o-8 and og9,
and t values of 7-IO; but the corresponding relationship of e on q produced
R2s of ooI-o o4, with t values of 1-1
On the other hand in manufacturing, where the regression coefficients for
both p on q and e on q were around o 5, the R2s were very similar (o-826 in the
one case and o 844 in the other) and the t values of the coefficients were as high
as 7 in both cases. These findings were not dependent on the inclusion ofJapan
in the sample. IfJapan is excluded from the sample, the results for the remaining
I I countries (based on the data shown in table 2, p. I2, of my Lecture) are as
follows:
p =
IP359 +o 4I7q,
R2 = o0536,
(O* I 29)
e = -I'33I
+o574q,
R2 = o-685.
(0 I 30)
The exclusion ofJapan reduces the closeness of the fit (and also the numerical
value of the Verdoorn coefficient, from o 5 to 0.4) but the results, in terms of t
values and R2s, are still sufficiently good to convey something significant.
The coefficient of e on q has a t value of over 4, and is significantly smaller than
I, by the test of the t value related to the differenceof the coefficient from unity.
On the other hand I nowhere suggested in my lecture that a statistically
significant positive correlation between p and e is a necessarytest of the Verdoorn
Law. The reason for this was a simple one. Since I regarded output as being
in general the exogenous variable (determined by demand) any error or disturbance would be associated with the employment term; and all such "disturbances" would automatically be reflected - with the opposite sign - in the
productivity series, thereby generating a spurious negative correlation between
p and e.
It follows that the existence of statistically significant relationships between
p and q and e and q does not carry with it that the relationship between p and e is
also statistically significant. The latter may happen, if the relationship between
e and q gives a sufficiently close fit, but it would not hold if the latter relationship
is not close enough. There is nothing very surprising therefore in the fact that it
is only by including Japan that the regression equation between p and e (as
calculated by Rowthorn) is statistically significant; even so, the R2 in the latter
equation is only o0447 as against o-844 on the basic equation of e on q, while
the t value of the regression coefficient is less than 3 (as against over 7 in the
basic equation).
1 Rowthorn is correct in saying that the coefficients of (2) can be algebraically derived from (i), or
vice versa; but whereas a significant relationship between e and q (with o < a < i) automatically
ensures that equation (i) (the "Verdoorn equation") is also significant, this is not true the other way
round - not unless one also specifies that the coefficient ,B in that equation is significantly less than i,
which has not hitherto been regarded as an integral property of the Verdoorn Law.
1975]
ECONOMIC
GROWTH
AND
THE
VERDOORN
LAW
893
hhouir
ahbsorhed
in mrniifhctiirincr
in the coiirse
of indnstri1ialtion
does
It was certainly unfortunate that Cripps and Tarling, in dealing with my hypothesis that there is
a positive association between productivity and employment in manufacturing, produced a correlation
in support of it between p and e, the validity of which depended (apparently without their realising it)
on a single extreme observation, Japan, and the significance of which (given the low value of R2) was in
any case dubious. If my above argument is correct this was not necessaryfor supporting the hypothesis
that increasing returns prevail in manufacturing. For that they had very much stronger evidence for the
1950-65
period in terms of the nature of the correlation between e and q from which the relationship
between p and q automatically follows. For the I965-70 period, on the other hand, they found no
significant relationship between e and q and hence no statistical support for the Verdoorn Law. (Cf.
T. F. Cripps and R. J. Tarling, Growthin AdvancedCapitalistEconomies,1950-1970, D.A.E. Occasional
Paper, no. 40, Cambridge University Press, 1973).
2 This would be the case, for example, if one assumed (a) that the total labour force effectively
employed grows at an exogenous rate; (b) the proportion of the labour force available to the manufacturing sector is given. Both these assumptions are patently untenable, especially if we take into account
inter-regional and inter-national migration of labour (which can be shown to have been largely demandinduced) as well as the very large changes (over time) in the inter-sectoral distribution of the labour
force of any particular region.
3 It could be argued that since in the lecture I regarded U.K. manufacturing output as being constrained by labour shortage, this is tantamount to saying that in the case of the United Kingdom
I regarded e as exogenous. However this is irrelevant, since the regression equations of e on q and p on
q were derived from a sample of countries for which e (i.e. the rate of growth of employment in manufacturing industry) was not exogenously given.
4 Since readers could hardly be expected to remember this equation, published more than I0 years
ago, it is worth reproducing it here (using the notation adopted in the Cripps-Tarling paper):
qGDP
II53+0
6I4qMF,
R2 =
0-959.
(o o8o)
This relationship has since been confirmed by other investigators such as the ECE (EconomicSurvey
of Europe... (I969), p. 78), UNCTAD, Cripps and Tarling, etc., and I am sure that it holds equally for
Gomulka's sample of 39 countries as for my sample of I2 countries; and that it holds for the I965-70
period, as well as earlier periods. An important property of the equation is that the regression coefficient
is significantly less than unity (implying that for growth rates exceeding 3 % a year, industrial production
894
THE
ECONOMIC
[DECEMBER
JOURNAL
not diminish production in the rest of the economy, owing to the existence of
surplus labour in agriculture (and also, though I did not say so explicitly, in
services) which is only eliminated at a late stage of industrial development,
at the stage of "economic maturity".
This view has been strikingly confirmed by Cripps and Tarling's findings in
two remarkable correlations (not mentioned by Rowthorn) which explain the
overall productivity growth of the economy (the rate of growth of GDP per
head) in terms of the rate of growth of industrial output and the (relative)
diminution of non-industrial employment. This relationship has in no way been
impaired by the failure of the Verdoorn Law in manufacturing in the post- I965
period; indeed the correlation coefficient is even higher for the I965-70
I950-65
period.
PGDP
PI 72 +o0534qlND-
(oo58)
I965-70:
PGDI
I I53+0
o8I2eNJ,
R2 =
o805,
0958,
(0o202)
R2
642qlND-o872eNI,
(O- I 25)
(oo058)
where PGDI1,, qIND, eNI stand for the rate of growth of GDP per employed person,
the rate of growth of industrial production and the rate of growth of nonindustrial employment respectively.'
The important thing to note is - and herein lies Rowthorn's misunderstanding that the existence of increasing returns to scale in industry (the Verdoorn Law)
is not a necessary or indispensable element in the interpretation of these equations.
Even if industrial output obeyed the law of constant returns, it could still be
true that the growth of industrial output was the governing factor in the overall
rate of economic growth (both in terms of total output and output per head) so
long as the growth of industrial output represented a net addition to the effective
use of resources and not just a transfer of resources from one use to another. This
would be the case if (a) the capital required for industrial production was
(largely or wholly) self-generated - the accumulation of capital was an aspect,
invariably rises faster than the GDP as a whole); the standard error is very small-t = I 5 in the above
equation; Cripps and Tarling (p. 22) fouind that t = 20 in a corresponding equation for a bundle of
43 observations-and as I have shown in the Appendix to my lecture, the equation owes nothing to
"auto-correlation" since the structure and the coefficient of the equation remain much the same if
manufacturing is excluded from the GDP on the LHS of the equation.
I Cf. Cripps and Tarling, Op.cit. p. 30. To see how far (if at all) these equations would be affected by
the exclusion of Japan, I asked Roger Tarling to re-compute the two regressions by excluding Japan
from the sample. The results are as follows.:
Elevencountries,excludingJapan
I 950-65:
PGDP =
I.768+o0369qIND-o
(o-o63)
647eNI,
R2 = o0678,
(0o I 7 )
I 965-70:
PGDP = o 8I9+O07IoqIND-o0848eNI,
(O I 24)
R2 = 0-930.
(0 I 35)
It is interesting to note that whereas the exclusion ofJapan somewhat reduces the fit, and the explanatory power of the equation (as measured by R2 and the size of the constant) for the I95o-65 period, it
makes virtually no difference for the I965-70 period.
I975]
ECONOMIC
GROWTH
AND
THE
VERDOORN
LAW
895
or a by-product, of the growth of output; and (b) the labour engaged in industry
had no true opportunity-cost outside industry, on account of the prevalence
of disguised unemployment both in agriculture and services. There is plenty
of direct evidence to substantiate both of these assumptions.
The important implication of these assumptions is that economic growth is
demand-induced, and not resource-constrained - i.e. that it is to be explained by
the growth of demand which is exogenous to the industrial sector' and not by
the (exogenously given) growth rates of the factors of production, labour and
capital, combined with some (exogenously given) technical progress over time.
While in the Lecture I gave the main emphasis to the Verdoorn Law as an
explanation for the difference in growth rates, and still believe in its importance, I would now regard the existence of surplus labour, and the critical role
of profits and profit expectations in capital accumulation as the more basic
cause of the difference of view between the neo-classical and Keynesian (or
post-Keynesian) schools of thought: the question, that is, whether one regards
economic growth as the resultant of demand (i.e. the growth of markets) or of
(exogenously given) changes in resource-endowment.
On the other hand, I now believe that I was wrong in thinking in I966 that
the United Kingdom had attained the stage of "economic maturity" (in the
sense I defined that term) and that her comparatively poor performance was
to be explained by inability to recruit sufficient labour to manufacturing
industry rather than by poor market performance due to lack of international
competitiveness. Statistical studies that have since come to light2 make it
doubtful whether I was correct in thinking that earnings in the service trades of
the United Kingdom had come to be fully competitive with earnings in manufacturing or that the growth of manufacturing industry in the United Kingdom
was constrained by labour shortages other than in a purely short-term sense
- e.g. of not having sufficient skilled labour in engineering to sustain a rapid
expansion of engineering production (which from a long-run point of view is
itself a consequence of a low trend rate of growth of demand).3 But while
1 In saying that growth is explained by the increase in demand which is " exogenous " to the growing
sectors I am conscious of the fact that this statement in itself is a simplification but one which does not
invalidate the statistical inferences derived from it. The growth of industrial output for any region is
governed in part by the growth in productivity which itself influences demand through the change in
competitiveness which is induced by it. It is this reverse link which accounts for the cumulative and
circular nature of growth processes. There is a two-way relationship from demand growth to productivity
growth and from productivity growth to demand growth; but the second relationship is, in my view, far
less regular and systematic than the first.
2 Sleeper, R., "Manpower Redeployment and the Selective Employment Tax," Bulletin of the
Oxford UniversityInstituteof Economicsand Statistics, November, I970.
3 The belief that the expansion of manufacturing production and thus of exports was hindered by the
inelastic supply of labour to manufacturing industry undoubtedly played a role in the introduction of
Selective Employment Tax (as was explained in the Government White Paper issued on its introduction).
But Rowthorn is wrong in thinking that the existence of increasing returns in manufacturing industry
was a necessary part of its justification. Given the fact that over 850% of U.K. exports were manufactured goods, and that the U.K. economy was threatened by a balance-of-payments crisis due to an
insufficiency of exports, the existence of a labour shortage would have been a perfectly adequate reason
for securing the release of labour from services, irrespective of whether increasing returns or constant
returns prevailed in industry. (In the actual case, as events have shown, the improvement of export
performance needed a devaluation, however, to improve the competitiveness of British goods in foreign
markets.)
896
THE
ECONOMIC
JOURNAL
[DECEMBER
I975]
I would now modify the story concerning the United Kingdom, such modification would definitely not be in the direction of Rowthorn, Gomulka or the neoclassicals. In particular, I would now place more, rather than less, emphasis on
the exogenous components of demand, and in particular on the role of exports,
in determining the trend rate of productivity growth in the United Kingdom
in relation to other industrially advanced countries.1
NICHOLAS
KALDOR
King's College
Cambridge
Date of receiptqf typescript:April 1975
1 Gomulka's thesis, favoured by Rowthorn - that the more rapid growth of productivity of latecorners like Japan was to be explained by the diffusion of technical knowledge - could hardly explain
how the higher productivity growth rates could have continued after the productivity levels of the
diffusees came to surpass those of the diffusants.