Professional Documents
Culture Documents
4 (July, 1972)
LECHATELIER-SAMUELSON PRINCIPLE
principle" is generalized, the generalization being closer to the original formulation of the
LeChatelier principle than Samuelson's theorem itself. The results obtained are formulated
1. INTRODUCTION
IN 1884 the French chemist H. L. LeChatelier, starting from the problem of thermo-
follows:
If a system is in stable equilibrium, and one of the conditions is changed, then the
equilibrium will shift in such a way as to tend to annul the applied change in the
conditions.2
This principle, vague though it may be, often allows one to make very useful
In the 1940's, P. A. Samuelson introduced this principle into economics [7, pp.
(x1i . . . , Xn) = x and n real parameters (l . . ., An) = A. Let x(i) denote a minimum
point of F(x, A) for a given value of the parameter vector A. Then, under suitable dif-
Let us call this the "weak" LeChatelier-Samuelson principle. Samuelson [7, pp.
[x(A2) - x(Al)](A2 - A ) 0,
' The authors are indebted to a referee for his comments concerning both the thermodynamic
background and the distinction between the "weak" and "strong" versions of the principle.
2 The above wording has been compiled from several sources, principally the Encyclopedic Dictionary
of Physics and the Encyclopaedia Britannica. It would be more correct to term this the "weak" Le-
Chatelier principle, since-although only in the more restricted context of "infinitesimally small"
3 "Lenz' law: When the magnetic flux linking a closed circuit is changed, a current is introduced in
the circuit in such a direction as to produce a magnetic flux which opposes the changing flux applied"
4 We remark in passing, that Samuelson [6] has extended this theorem to the case of certain systems
which are not governed by an extremum principle. See, in this connection, also Morishima [3, pp. 3-14].
711
This content downloaded from 131.172.36.29 on Sun, 28 Feb 2016 09:27:53 UTC
All use subject to JSTOR Terms and Conditions
or briefly,
It is interesting to note that this global version of Samuelson's theorem has recently
N. Z. Shapiro [9].
Besides the theorem just stated there exists another theorem of Samuelson
Let x(A) denote a minimum point of F(x, A), and let x*(i) denote a minimum point
of F(x, A) subject to the constraint P(x) = const. If x(i) = x*(i), then under suitable
(la) X Ai ( 0o5
which constitute our main point of interest here. Therefore, in what follows, we
contains Samuelson's theorem as a special case and yet is close to the original
formulation quoted above. A part of this generalization lies in the fact that we
from the theory of cost. In Section 5 we derive additional conclusions from Theorem
1. Among other things, we show that Samuelson's theorem follows readily from
Theorem 1.
mathematical programming. This has also been done by Samuelson [5], Beckmann
F:(x,A)eA x B - F(x,A) c-
We denote by M(i) the set of all x E A which minimize F(y, A) subject to the con-
dition y E A. Thus
yeA
If, for example, in a thermodynamical context we identify x, A, and P with pressure, volume, and
temperature respectively, then (la) would correspond to the fact that the isoentropic pressure-volume
contour has a more negative slope than the isothermic pressure-volume contour, whereas (1) corresponds
only to the fact that the isoentropic pressure-volume contour is negatively sloped.
This content downloaded from 131.172.36.29 on Sun, 28 Feb 2016 09:27:53 UTC
All use subject to JSTOR Terms and Conditions
For reasons which will become clearer in Sections 3 and 5, the following statement
(3) F(X2, )2) - F(x2, Al) < F(x', A2) - F(x',, Al),
with equality if and only if xI E M(X2). Adding (4) and (5) and then subtracting
F(x', X2) + F(x2, Al) from both sides at once yields (3). Clearly equality in (3) is
define
yeA
then the inequality sign in (3) is reversed. (ii) Note that (3) is invariant under a
We may furthermore consider the difference F(x, )2) - F(x, Xl) as the "effect,"
change in the parameter A from Al to ,2 and assumes a new equilibrium state at the
which is a measure of the effect of the disturbance (i.e., of the reaction of the system
to the change) with respect to the point x2, is less than or equal to the difference
This content downloaded from 131.172.36.29 on Sun, 28 Feb 2016 09:27:53 UTC
All use subject to JSTOR Terms and Conditions
Let the state space X of Section 2 be the n-dimensional real vector space Mn
and let the parameter space A be Sm. We consider the following model from the
theory of cost.
Suppose that the variable cost F of a certain input vector (x1,.. I, xn) = x E A
depends on x and on the prices (P1, . . ., Pn) = p of the inputs, where these prices
For example, the parameters may be the world market prices of goods for which
certain input goods can be substituted. If we let A be the set of all input vectors
which can be used to produce a certain fixed output vector (a1, ... , ar) = a,
then, according to our definition x E M(i) if and only if x is an input vector which
(i) If the vector A of exogenous variables changes from Al to A' in such a way that
i.e., if the cost of the input vector x2 is increased by the change of the exogenous
variables, then the cost of the input vector xl is increased by at least the same
amount. We have
(7) 0 < F(X2, )2) - F(X2, )l) < F(x1, )2) - F(xl, )l).
(ii) Conversely, if the cost of the input vector x' is decreased by the change of the
exogenous variables, then the cost of the input vector x2 decreases by at least the
same amount.
PROOF: The theorem can be proved by substitution of the right-hand side of (8)
REMARKS (i) Samuelson [7, p. 81] relates (9) to the principle of LeChatelier:
This content downloaded from 131.172.36.29 on Sun, 28 Feb 2016 09:27:53 UTC
All use subject to JSTOR Terms and Conditions
trivially proved, but it has important consequences for the theory of chemical
(iii) If the price vector p in the model in Section 4 has the form p(x, A) = - with
corresponding costs F(x, A) = Axt, then we conclude from (9) that an increase
of this factor, provided the prices of all other factors remain constant.
linear function of x. Let us now consider the case where that term is a quadratic
form in x.
where now A = (Aij) (i,j = 1, ... , n), is a symmetric (n x n)-matrix. Let A' and )2
C(x) denote the cost of a firm which produces the output x = (x1, . .. , xn). Let
the market prices (P1, . .. , Pn) = p depend linearly on x, i.e., let pr = a' + Ax',
positive constants. Then the profit of the firm, namely xpt - C(x) = xat - C(x) +
x{xt, is a function of the form (10). Note that the matrix A in this expression can
Theorem 3 to the problem of maximizing the above profit for two different
inequality sign.
If we put
0...-O
(all parameters with the exception of A remain constant), then we conclude from
This content downloaded from 131.172.36.29 on Sun, 28 Feb 2016 09:27:53 UTC
All use subject to JSTOR Terms and Conditions
Iapj XjI
quantity xj supplied by the firm, provided that all other parameters remain fixed.
F:(x,A)eA x B F(x,A) -
c+ /3= 1.
each interior point A' of B there exists at least one vector fQJA') with the following
property:
x2 C M(A2) Then
Adding' the two inequalities and using (3), we obtain the desired result.
Theorem 4 can be applied to the model in Section 4. Let the cost function (6)
satisfy the concavity hypothesis of Theorem 4,8 and let F(x, A) be differentiable
7 A less direct, but more geometrical, proof of this theorem could be based on the facts that jr(A) =
inf {F(x, A)Jx E A} is a concave function, and that F&(A), A), originally defined as a subgradient of F,
is also a subgradient of 0. The theorem would then correspond to a well known monotonicity property
of the directional derivative of a concave function jr, and this interpretation could be of interest even
in the classical case where F is linear in A. We shall discuss such purely geometrical aspects perhaps
elsewhere.
8 This assumption can be justified, if, for example, the A's in (6) are world market prices of goods for
This content downloaded from 131.172.36.29 on Sun, 28 Feb 2016 09:27:53 UTC
All use subject to JSTOR Terms and Conditions
The marginal cost with respect to Aj, evaluated at the corresponding minimal
Because of the specific nature of the following example, this increase or decrease
is such that a relatively small variation of the minimal cost corresponds to a large
is to be minimized with respect to all input vectors which are feasible for producing
xeA
A change of the parameter by 10 per cent thus results in a 2.5 per cent change of
and
REFERENCES
[1] BAILEY, M. J.: "A Generalized Comparative Statics in Linear Programming," Review of Economic
[2] BECKMANN, M. J.: "Comparative Statics in Linear Programming and the Giffen Paradox," Review
[3] MORISHIMA, M.: Equilibrium, Stability, and Growth. A Multi-Sectoral Analysis. Oxford: Oxford
[4] PASSY, U., AND D. J. WILDE: "A Geometric Programming Algorithm for Solving Chemical Equili-
[51 SAMUELSON, P. A.: "Comparative Statics and the Logic of Economic Maximizing," Review of
[71 Foundations of Economic Analysis. 8th printing. Cambridge: Harvard University Press,
1966.
[8] SHAPIRO, N. Z.: On the Behavior of a Chemical Equilibrium System when its Free Energy Parameters
are Changed. Rep. RM-4128-PR. Santa Monica, California: The RAND Corporation, 1964.
[9] : "An Inequality Concerning the Minimum of the Sum of an Arbitrary Function and a
This content downloaded from 131.172.36.29 on Sun, 28 Feb 2016 09:27:53 UTC
All use subject to JSTOR Terms and Conditions