A Growth Cycle

R. M. Goodwin

July 30, 2014

† Presented at the First World Congress of the Econometric Society, held

in Rome, 1965.

The Model

Presented here is a starkly schematized and hence quite unrealistic model

of cycles in growth rates. This type of formulation now seems to me to
have better prospects than the more usual treatment of growth theory or of
cycle theory, separately or in combination. Many of the bits of reasoning are
common to both, but in the present paper they are put together in a different
way.
The following assumptions are made for convenience:
(1) steady technical progress (disembodied);
(2) steady growth in the labour force;
(3) only two factors of production, labour and ‘capital’ (plant and
equipment), both homogeneous and non-specific;
(4) all quantities real and net;

1
 (5) all wages consumed, all profits saved and invested.
These assumptions are of a more empirical, and disputable, sort:
(6) a constant capital-output ratio;
(7) a real wage rate which rises in the neighbourhood of full employ-
ment.
No. (5) could be altered to constant proportional savings, thus changing
the numbers but not the logic of the system. No. (6) could be softened but
it would mean a serious complicating of the structure of the model.
Symbols used are:
q is output;
k is capital;
w is wage rate;
a = a0 eαt is labour productivity; α constant;
σ is capital-output ratio (inverse of capital productivity);
w/a is workers’ share of product, (l − w/a) capitalists’;
Surplus = profit = savings = investment= (l − w/a)q = k̇.
Profit rate = k̇/k = q̇/q = (l − w/a)/σ.
n = n0 eβt is labour supply, β constant;
l = q/a is employment.
Writing (q/l)· for d/dt(q/l), we have
˙ = α,
(q̇/l)/q/l = q̇/q − l/l

so that
˙ = (l − w/a)/σ − α.
l/l

Call
u = w/a, v = l/n,

2
 Figure 1:

so that
v̇/v = (l − u)/σ − (α + β).

Assumption (7) may be written as

ẇ/w = f (v)

as shown in fig.1.
The following analysis can be carried out using such an f (v), with a
change in degree but not in kind of results. Instead, in the interest of lucidity
and ease of analysis, I shall take a linear approximation (as shown in fig.1),

ẇ/w = −γ + ρv

and this does quite satisfactorily for moderate movements of v near the point
+1. Both γ and ρ must be large. Since

u̇/u = ẇ/w − α, u̇/u = −(α + γ) + ρv.

From this and the equation above for v, we have a convenient statement

3
of our model.

v̇ = [(1/σ − (α + β)) − (1/σ)u]v, (1)

u̇ = [−(α + γ) + ρv]u. (2)

In this form we recognize the Volterra case of prey and predator (Theorie
Mathematique de la Lutte pour la Vie, Paris, 1931). To some extent the
similarity is purely formal, but not entirely so. It has long seemed to me
that Volterra’s problem of the symbiosis of two populations - partly com-
plementary, partly hostile - is helpful in the understanding of the dynamical
contradictions of capitalism, especially when stated in a more or less Marxian
form.
Eliminating time and performing a first integration we get

(1/σ)u + ρv − [1/σ − (α + ρ)] log u − (γ + α) log v = constant.

Letting

θ1 = 1/σ; η1 = 1/σ − (α + β),

θ2 = ρ; η2 = γ + α,

we can transform this into

φ(u) = uη1 e−θ1 u = Hv −η2 eθ2 v = Hψ(v), (3)

where H is arbitrary constant, depending on initial conditions. Since 1/σ >

α + β, all coefficients are positive. By differentiating,

dφ/du = (−θ1 + η1 /u)φ, dψ/dv = (θ2 − η2 /v)ψ,

so that we can see that these functions have the sorts of shapes given in fig.2.

4
 Figure 2:

Our problem as stated in (3) is to equate φ(u) to ψ(v) multiplied by a

constant H. This can be done neatly in the four quadrant positive diagram
in fig.3. We draw through the origin a straight line, A, with the slope
φ/ψ = H (arbitrary since dependent on the given initial condition). Then in
symmetrical quadrants we place the two curves and φ and ψ equating these
two through the constant of proportionality gives a possible pair of values
for u and v. All possible pairs of u and v constitute a solution, which may
be plotted in the remaining quadrant. It can be shown, and indeed is quite
obvious, that these solution points lie on a closed, positive curve, B, in u, v
space. By going back to equations (1) and (2) we can find in what order the
points succeed each other and hence in what direction we traverse curve B,
as indicated by arrows in fig.3. A second integration will yield u and v as
functions of time, thus allowing us to determine the second arbitrary factor,
the point on B at which we start. By varying the slope of A we can generate
a family of closed curves broadly similar to B, thus yielding all the possible
solutions. One initial condition selects the curve, a second fixes the starting
point, and then we traverse some particular curve B in the direction of the

5
Figure 3:

6
arrows forever, in the absence of given outside changes. There remains only
to spell out the meaning of the motion.
Hence we may classify our model as a non-linear conservative oscillator
of, fortunately, a soluble type. As the representative point travels around
the closed curve B, u vibrates between ξ1 and ξ2 , and v between ξ1 and
ξ2 . Both u and v must be positive and v must, by definition, be less than
unity; u normally will be also but may, exceptionally, be greater than unity
(wages and consumption greater than total product by virtue of losses and
disinvestment). Over the stretch 0 to + 1 on the u axis, the point u indicates
the distribution of income, workers’ share to the left, capitalists’ to the right.
The capitalists’ share, multiplied by a constant, 1/σ, gives us the profit rate
and the rate of growth in output, q̇/q. When profit is greatest, u = ξ1 , em-
ployment is average, v = ηθ2 ,, and the high growth rate pushes employment
to its maximum ξ2 which squeezes the profit rate to its average value η1 /θ1 .
The deceleration in growth lowers employment (relative) to its average value
again, where profit and growth are again at their nadir ξ2 . This low growth
rate leads to a fall in output and employment to well below full employment,
thus restoring profitability to its average value because productivity is now
rising faster than wage rates. This is, I believe, essentially what Marx meant
by the contradiction of capitalism and its transitory resolution in booms and
slumps. It is, however, un-Marxian in asserting that profitability is restored
not (necessarily) by a fall in real wages but rather by their failing to rise with
productivity. Real wages must fall in relation to productivity; they may fall
absolutely as well, depending on the severity of the cycle. The improved prof-
itability carries the seed of its own destruction by engendering a too vigorous

7
expansion of output and employment, thus destroying the reserve army of
labour and strengthening labour’s bargaining power. This inherent conflict
and complementarity of workers and capitalists is typical of symbiosis.
An undisturbed system has constant average values η1 /θ1 for u and for
v, hence a constant long-run average distribution of income and degree of
unemployment. Much more remarkable is the fact that a disturbed system
still has the same constant long-run values. The time averages of u and of
v are independent of initial conditions. We can see this from the fact that
a rotation of A (an outside change) will only make the curve B larger or
smaller but will not alter its central point. Therefore continual shocks will
alter the shape of the cycle but not the long-run average values. Output
and employment both will show alternating rates of growth. Whether they
actually decrease or merely rise less rapidly will depend on the severity of
the cycle. For a mild cycle the growth rate may decrease but never become
negative, in other cases there may be a sharp fall. However, the increases
must predominate over the decreases, since the time average of 1 − u is
positive and hence so also is that of q̇/q. Likewise employment grows in
the long-run at the same rate as labour supply, since the time average of v
is constant. Similarily the equality of the growth rate in wages to that in
productivity follows from the constancy of u. By contrast the profit rate is
equal to 1 − u and therefore tends to constancy. We may look at this as
standing Ricardo (and Marx) on his head. Progress first accrues as profits
but profits lead to expansion and expansion forces wages up and profits down.
Therefore we have a Malthusian Iron Law of Profits. This is because of the
tendency of capital, though not capitalists, to breed excessively. By contrast

8
labour is something of a rent good since the supply, though variable, does
not seem to be a function of wages. Hence it is the sole ultimate beneficiary
from technical progress. By now there would, I suppose, be considerable
agreement that what happened in history is: wage rates went up; profit
rates stayed down. It is to the explanation of this that the present paper is
addressed.