You are on page 1of 30

The IUP Journal of Managerial Economics, Vol. VIII, No.

4, 2010
6
A Brief History
of Production Functions
S K Mishra*
Introduction
Production function has been used as an important tool of economic analysis in the
neoclassical tradition. It is generally believed that Wicksteed (1894) was the first economist
to algebraically formulate the relationship between output and inputs as
P = f (x
1
,x
2
,...,x
m
), although there are some evidences suggesting that Johann von Thünen
first formulated it in the 1840s (Humphrey, 1997).
It is relevant to note that among others there are two leading concepts of efficiency
relating to a production system—the one often called the ‘technical efficiency’ and the
other called the ‘allocative efficiency’ (Leibenstein et al., 1988). The formulation of production
function assumes that the engineering and managerial problems of technical efficiency
have already been addressed and solved, so that analysis can focus on the problems of
allocative efficiency. That is why a production function is (correctly) defined as a
relationship between the maximal technically feasible output and the inputs needed to
produce that output (Shephard, 1970). However, in many theoretical and most empirical
studies it is loosely defined as a technical relationship between output and inputs, and the
assumption that such output is maximal (and inputs minimal) is often tacit. Further,
although the relationship of output with inputs is fundamentally physical, production
function often uses their monetary values. The production process uses several types of
* Professor, Department of Economics, North-Eastern Hill University, Shillong, India.
E-mail: mishrasknehu@gmail.com
© 2010 IUP. All Rights Reserved.
This paper gives an outline of evolution of the concept and econometrics
of production function, which was one of the central apparatus of
neoclassical economics. It shows how the famous Cobb-Douglas
production function was indeed invented by von Thünen and Wicksell,
how the Constant Elasticity of Substitution (CES) production function
was formulated, how the elasticity of substitution was made a variable
and finally how Sato’s function incorporated biased technical changes.
It covers almost all specifications proposed during 1950 to 1975, as well
as the linear-exponential (LINEX) production functions and incorporation
of energy as an input. The paper is divided into single product functions,
joint product functions, and aggregate production functions. It also
discusses the ‘capital controversy’ and its impacts.
7 A Brief History of Production Functions
inputs that cannot be aggregated in physical units. It also produces several types of output
(joint production) measured in different physical units. There is an extreme view that (in a
sense) all production processes produce multiple outputs (Faber et al., 1998). One of the
ways to deal with the multiple output case is to aggregate different products by assigning
price weights to them. In so doing, one abstracts away from essential and inherent aspects
of physical production processes, including error, entropy or waste. Moreover, production
functions do not ordinarily model the business processes, thereby ignoring the role of
management, of sunk cost investments and the relation of fixed overhead to variable costs
(Wikipedia-a).
It has been noted that although the notion of production function generally assumes
that technical efficiency has been achieved, this is not true in reality. Some economists and
operations research workers (Farrell, 1957; Charnes et al., 1978; Banker et al., 1984; Lovell
and Schmidt, 1988; Seiford and Thrall, 1990; and Emrouznejad, 2001) addressed this
problem by what is known as the Data Envelopment Analysis (DEA). The advantages of
DEA are: here one need not specify a mathematical form for the production function
explicitly; it is capable of handling multiple inputs and outputs and being used with any
input/output measurement; and efficiency at technical/managerial level is not presumed.
It has been found useful for investigating into the hidden relationships and causes of
inefficiency. Technically, it uses linear programming as a method of analysis. We do not
intend to pursue this approach here.
Starting in the early 1950s until the late 1970s, production function attracted many
economists. During the said period, a number of specifications or algebraic forms relating
inputs to output were proposed, thoroughly analyzed and used for deriving various
conclusions. Especially after the end of the ‘capital controversy’, search for new specification
of production functions slowed down considerably. Our objective in this paper is to briefly
describe that line of development. In the schema of Frisch (1965), we shall first concentrate
on ‘single-ware’ or single-output production function. Then we shall move to ‘multi-ware’
or multi-output production function. Finally, we shall address the pros and cons of the
aggregate production function.
Single-Output Production Function
Humphrey (1997) gives an outline of historical development of the concept and
mathematical formulation of production functions before the enunciation of Cobb-Douglas
function in 1928. Paul Douglas, on a sabbatical at Amherst, asked mathematics professor
Charles W Cobb to suggest an equation describing the relationship among the time series
on manufacturing output, labor input, and capital input that Douglas had assembled for
the period 1889-1922, and this led to their joint paper.
An implicit formulation of production functions dates back to Turgot. In his 1767
Observations on a Paper by Saint-P´eravy, Turgot discusses how variations in factor proportions
affect marginal productivities (Schumpeter, 1954). Malthus introduced the logarithmic
production function (Stigler, 1952) and Barkai (1959) demonstrated how Ricardo’s quadratic
The IUP Journal of Managerial Economics, Vol. VIII, No. 4, 2010
8
production function was implicit in his tables. Ricardo used it to predict the trend of rent’s
distributive share as the economy approaches the stationary state (Blaug, 1985).
Johann von Thünen was perhaps the first economist who implicitly formulated the
exponential production function as
[
÷
÷ = =
3
1
) 1 ( ) (
i i
F a
e A F f P , where F
1
, F
2
and F
3
are the
three inputs, labor, capital and fertilizer, a
i
are the parameters and P is the agricultural
production. Lloyd (1969) provides a complete account of von Thünen’s exponential
production functions and their derivation. He was also the first economist to apply the
differential calculus to productivity theory and perhaps the first to use calculus to solve
economic optimization problems and interpret marginal productivities essentially as partial
derivatives of the production function (Blaug, 1985). Mitscherlich (1909) and Spillman
and Lang (1924) rediscovered von Thünen’s exponential production function.
In The Isolated State, Vol. II, von Thünen wrote down the first algebraic production
function as: p = hq
n
, where p is output per worker (Q/L), q is capital per worker (C/L) and
h is the parameter that represents fertility of soil and efficiency of labor. The exponent n is
another parameter that lies between zero and unity. Multiplying both sides of von Thünen’s
function by L (labor), we have
n n n
L hC L hq Lp
÷
= =
1
= P = Output. Thus, we have the
Cobb-Douglas production function hidden in von Thünen’s production function (Lloyd,
1969). The credit for presenting the first Cobb-Douglas function, albeit in disguised or
indirect form, must go to von Thünen in the late 1840s rather than to Douglas and Cobb in
1928 (Humphrey, 1997). Further, von Thünen was uneasy to realize the implication of his
production function in which labor alone cannot produce anything. Hence, he modified
his function to . ) (
1 ÷
+ =
n n
L C L h P This equation, which von Thünen estimated empirically
for his own agricultural estate and which he declares he discovered only after more than
20 years of fruitless search, states that labor produces something even when unequipped
with capital (Humphrey, 1997). It is surprising, however, that modern economists never
formulate a production function in which labor alone can produce something.
Velupillai (1973) points out how Wicksell formulated his production function in
1900-01 that is identical to Cobb-Douglas function. In his 1923 review of Gustaf Akerman’s
doctoral dissertation Realkapital und Kapitalzins, Wicksell (1923) wrote his function as
| o
C cL P =
with the exponents adding up to unity.
Works of Turgot, von Thünen, and Wicksell might not have been known to Paul Douglas,
but it is surprising to know that before his collaboration with Cobb, Sidney Wilcox,
a research assistant of Douglas, had formulated in 1926 a production function of which
the Cobb-Douglas function is only a special case (Samuelson, 1979). Wilcox’s production
function was, perhaps, ignored by Douglas and till date it has remained in obscurity.
Likewise, what is now well-known as the Leontief production function was formulated
by Jevons, Menger and Leon Walras. In the Walrasian model of general equilibrium, the
proportions of output to inputs are fixed and no substitution among inputs is entertained.
9 A Brief History of Production Functions
Bertrand Russell tells us that certain academic achievements of individual scholars are
a culmination of the research efforts of an entire epoch. Such achievements are superb in
exposition though containing only a little of originality, and yet they are known to the
posterity by the name of those scholars in whose work they found their final expression.
These achievements set forth a paradigm and thus put a halt on the further progress of
science for quite a long period. This is true of Aristotle, Newton and many others (Russell,
1984, p. 521). So, it is also true of Cobb-Douglas. A notable change in their formulation of
production function came only in 1961—after a gap of 33 years—with the work of Arrow
et al. (1961), which, however, is only an extension, not an alternative paradigm.
Of course, two generalizations of the Cobb-Douglas production function appeared in
the literature before 1961. One of them is described as
] [
1 ) / ( | | o ÷
= L K Ae P
L K
and the
transcendental production function of Halter et al. (1957) specified as
. 1 , 0 ]; [ < < =
+
| o
| o
L K Ae P
bL aK
The first of these functions is a neoclassical production
function only in a restricted region in the non-negative orthant of the (K, L) plane, the
region in which the marginal products are nonnegative and diminishing marginal rate of
substitution holds. When o > 0, 0 < | < 1 the marginal product of labor is nonnegative only
if . / / L K > o | Moreover, it does not contain a linear function as its special case. The
linear production function is important in view of the Harrod-Domar fixed coefficient
model of an expanding economy and therefore, every neoclassical production function,
the Cobb-Douglas or its generalizations, must contain the linear production function
bL aK P + =
to be consistent with it (Revankar, 1971). The transcendental function is a
neoclassical production function if a, b < 0 and b L / | ÷ s and a K / o ÷ s (Revankar,
1971). The transcendental function also does not contain a linear function. However, these
functions allow for variable elasticity of substitution.
In the Cobb-Douglas production function the elasticity of substitution of capital for
labor is fixed to unity, e.g., 1% for 1%. The production function formulated by Arrow et al.
(1961) permitted it to lie between zero and infinity, but to stay fixed at that number along
and across the isoquants, irrespective of the size of output or inputs (capital and labor)
used in the production process. This function is well-known as the Constant Elasticity of
Substitution (CES) production function. It encompasses the Cobb-Douglas, the Leontief
and the linear production functions as its special cases.
Two mutually interrelated difficulties with the CES production function came to light
very soon. The first is in the constancy of the elasticity of substitution (between inputs)
along and across the isoquants, and the second is in defining the said elasticity when
more than two inputs are used in production. For three inputs there would be three
elasticities (say, o
ij
, o
ik
, o
jk
) and for more inputs there would be many more. Uzawa (1962)
and McFadden (1962 and 1963) proved that it is impossible to obtain a functional form for
a production function that has an arbitrary set of constant elasticities of substitution, if the
number of inputs (factors of production) is greater than two. Mathematical enunciations of
these assertions are now known as the impossibility theorems of Uzawa and McFadden.
The IUP Journal of Managerial Economics, Vol. VIII, No. 4, 2010
10
It is rather predictable what economists would do afterwards. That is: to find a functional
form of (a neoclassical) production function that would permit variable elasticities of
substitution (among different inputs) along and across the isoquants as well as larger
number of inputs to be included in the recipe.
Mukerji (1963) generalized CES for constant ratios of elasticities of substitution.
Bruno (1962) suggested a generalization of CES production function to permit the elasticity
of substitution to vary. Liu and Hildebrand (1965) formulated a variable elasticity of
substitution production function as
| | . ) 1 (
/ 1
) 1 (
q
q q q
o o
m m
L K K A P
÷
+ ÷ =
It reduces to CES
for m = 0. For this function, the elasticity of substitution is given as
1
) / 1 (
÷
+ ÷ =
k
S mq q o ,
where S
k
is the share of capital. In 1968, Bruno formulated his Constant Marginal Share
(CMS) production function , ) / ( / or
1
m L K A L P mL L AK P ÷ = ÷ =
÷ o o o
which implies
that productivity of labor increases with capital-labor ratio at a decreasing rate. The CMS
production function contains a linear production function and defines the elasticity of
substitution as, | | ). / ( ) 1 /( 1 P L m o o o ÷ ÷ = As the output-labor ratio increases (e.g., with
economic growth), the elasticity of substitution in this function tends to unity and thus the
CMS tends to the Cobb-Douglas production function. The CMS function has nonnegative
marginal productivity of labor over the entire (K, L) orthant only if 0 s m (and 0 < o < 1).
But then the elasticity of substitution will never be less than unity. This feature of the CMS
function puts some limitations on it.
Lu and Fletcher (1968) generalized the CES production function to permit variable
elasticity of substitution. Their Variable Elasticity of Substitution (VES) function is specified
as
| | . ) / ( ) 1 (
/ 1
) 1 (
|
| | |
q o o L L K K A P
c ÷ ÷
÷ + =
Assuming that competition has led to minimal
cost conditions in production, the elasticity of substitution is obtained as,
)}] /( 1 { 1 [ ) 1 (
1
rK wL c + ÷ + =
÷
| o where wL and rK are the shares of labor and capital in
output (net value added) respectively. It may be noted that the assumption of minimal cost
conditions and use of factor shares in defining the elasticity of substitution limit the
importance of this function in empirical investigation. Sato and Hoffman ((1968) also
introduced their variable elasticity of substitution production function.
In his doctoral dissertation, Revankar (1967) expounded his generalized production
functions that permit variability to returns to scale as well as elasticity of substitution.
In contrast with the production functions that (rather unrealistically) assume the same
returns to scale at all levels of output, Zellner and Revankar (1969) found a procedure to
generalize any given (neoclassical) production function with specified constant or variable
elasticities of substitution such that the resulting production function retains its
specification as to the elasticities of substitution all along, but permits returns to scale to
vary with the scale of output. Their Generalized Production Function (GPF) is given as,
h h P
f c Pe =
u
where f is the basic function (e.g., Cobb-Douglas, CES, etc.) as the object of
11 A Brief History of Production Functions
generalization, c is the constant of integration, and u, h relate to parameters associated
with the returns-to-scale function. In particular, if the Cobb-Douglas production function
is generalized, we have
.
) 1 ( o µ µo u ÷
= L AK Pe
P
This function is interesting from the viewpoint
of estimation also. It has to be estimated so as to maximize the likelihood function since the
least squares and maximum likelihood estimators of parameters do not coincide (see Mishra,
2007a). The returns to scale function is given by ). 1 /( ) ( P P u µ µ + = Depending on the sign
of u, the returns-to-scale function monotonically increases or decreases with increase in P.
However, as we know, the returns to scale first increases with output, remains more or less
constant in a domain and then begins to fall. This fact is not captured by the Zellner-
Revankar function since it gives us a linear returns-to-scale function. Revankar (1971)
presented his VES production function as
µoµ oµ µ
µ ] ) 1 ( [
) 1 (
K L AK P ÷ + =
÷
with restriction
on parameters: 1 0 ; 1 0 ; 0 , s s < < > oµ o µ A and ). 1 /( ) 1 ( / oµ µ ÷ ÷ > K L The elasticity of
substitution function is given as . / 1 / )] 1 /( ) 1 [( 1 ) , ( L K L K L K | oµ µ o + = ÷ ÷ + = Revankar’s
VES does not contain the Leontief production function (while the CES does contain it), but
it contains Harrod-Domar fixed coefficient model, the linear production function and the
Cobb-Douglas function.
Brown and Cani (1963) generalized the CES production to allow for non-constant
returns to scale. Nerlove (1963) generalized the Cobb-Douglas production function to
allow for variable returns to scale. Ringstad (1967) advanced on the same line. Their
production function is given as , 0 ;
) ln 1 (
> =
+
c K AL P
P c | o
that contains the Cobb-Douglas
production function for c = 0.
On the side of generalizing the CES production function to more than two factors of
production, Uzawa (1962) made fundamental contributions. Further, Sato (1967)
generalized the CES production function so as to incorporate more than two inputs in the
specification. To illustrate Sato’s schema, let ) , , , (
4 3 2 1
x x x x f P = where P is output and
x
i
is an input. Now combine x
1
and x
2
in a manner of CES to obtain Z
1
and similarly,
combine x
3
and x
4
to obtain Z
2
. At the second level, combine Z
1
and Z
2
to obtain P. This
schema would provide (constant) elasticities of substitution between x
1
and x
2
(say, o
12
)
and between x
3
and x
4
(say, o
34
) at the first level and between Z
1
and Z
2
(say, s
12
) at the
second level. The nesting schedule of inputs would depend on the nature of inputs and
production technology.
There are ample empirical evidences that suggest capital-skill complementarity
(Griliches, 1969), or the wage differential between skilled and unskilled workers. It requires
two types of labor (skilled and unskilled) to be separately dealt with in specifying the
production function. Obviously, the elasticity of substitution of capital for skilled labor is
very little—they are complementary to each other, while the case of unskilled labor is
entirely different. To specify such models, the two-level CES production technology with
capital, skilled labor and unskilled labor as inputs may be more suitable.
The IUP Journal of Managerial Economics, Vol. VIII, No. 4, 2010
12
Diewert (1971) made two very important generalizations of production functions. First,
he obtained a functional form that can incorporate many inputs in the specification and
second, such a functional form permitted variable elasticities of substitution.
He generalized the Leontief production function to attain an arbitrary set of Allen-Uzawa
elasticities or shadow elasticities of substitution. He also obtained the generalized linear
production function that can attain an arbitrary set of direct elasticities of substitution at a
given set of inputs and input prices. Diewert’s generalized linear production function is:
0 ;
1 1
2 1 2 1
> =
|
|
.
|

\
|
=
¿¿
= =
ji ij
n
i
n
j
j i ij
a a x x a h P
where h is a continuous, monotonically increasing function that tends to +· and has
h(0) = 0. Restricting the sum of coefficients to unity, this function provides convexity to the
isoquants and implies a well-behaved cost function.
Griliches and Ringstad (1971), Berndt and Christensen (1973) and Christensen et al.
(1973) introduced their translog production functions that permit more than two inputs as
well as variable elasticities of substitution (which is a necessity in view of Uzawa-McFadden
theorems). For n inputs (x
i
), the translog production function is specified as:
¿ ¿ ¿
= = =
= + + =
n
i
n
j
ji ij j i ij
n
i
i i
b b x x b x a a P In
1 1 1
0
); ( ln ) ( ln 5 . 0 ) ( ln ) (
Unfortunately, this function is not invariant to the units of measurement of inputs and
output (Intriligator, 1978). Further, on account of inclusion of ln(x
i
), ln(x
j
), their product
and their squared values, estimation of parameters of this function often suffers from
multicollinearity problem.
Kadiyala (1972) proposed a production function that includes Cobb-Douglas, CES,
Lu-Fletcher, Revankar and Sato-Hoffman production functions as its special cases. The
function is defined as:
0 ; 1 2 ; ] 2 )[ (
22 12 11
) /( ) (
22 12
) (
11
2 1 2 1 2 1 2 1
> = + + + + =
+ + +
ij
K K L L t E P e e e e e e e
| | µ | | | | | |
In the function above, E(t) is the efficiency parameter, which also absorbs the neutral
technical progress. Further, |
1
and |
2
bear the same sign as (|
1
+ |
2
). Kadiyala showed that
for e
12
= 0 his function is CES, for e
22
= 0 it is Lu-Fletcher, and for e
11
= 0 it is Sato-Hoffman
production function. He also showed that the function may be generalized for more than
two inputs, but to take care of the Allen or Uzawa-McFadden partial elasticities of
substitution one has to involve all the input ratios and the functional form may thus be
quite lengthy and complicated. In spite of its generality, Kadiyala’s production function
has remained in obscurity.
At this juncture it would be appropriate to point out that in 1967 Kmenta had proposed
a production function whose second-order Taylor’s series approximation (around
substitution parameter, | = 0) could be used to estimate the CES production function,
13 A Brief History of Production Functions
| µ | |
o o
÷ ÷ ÷
÷ + = ] ) 1 ( [ L K A P . Kmenta obtained:
2
)] ( ln ) ( ln )[ 1 ( 5 . 0 ) ( ln ) 1 ( ) ( ln ) ( ln ) ( ln L K L K A P ÷ ÷ ÷ ÷ + + = o µ|o o µ µo
McCarthy (1967) pointed out that there could exist another production function,
0 , 0 , ; ] [
3 2 1
= > + + =
÷
b a A L K a L a K a A P
j
b c b c b b µ
whose second-order Taylor’s series approximation would be econometrically
indistinguishable from Kmenta’s approximation. Thus, if one uses Kmenta’s approximation
to estimate the parameters of a production function, one is not sure whether those
parameters would really pertain to the CES or the McCarthy’s production function.
However, enough attention has not been given to this issue, at least in the textbooks.
In most cases, McCarthy’s work is ignored and unmentioned.
Färe and Mitchell (1989) modified McCarthy’s production function by constraining
a
1
+ a
2
= 1; 0 < a
1
< 1 and defining ¸ = a
3
/ b; b = 0 to obtain:
b c b c b b
L K b L a K a A P
µ
¸ ] ) 1 ( [
1 1
÷
+ ÷ + =
and showed that this specification, which we would call the McCarthy-Färe-Mitchell (MFM)
production function, has eight production functions as its special cases (see Figure 1).
A note on estimation of MFM production function is deemed necessary. It appears that if
one estimates the MFM function, one can identify a variety of its special cases according to
the values taken on by the estimated parameters. However, in spite of being an excellent
Figure 1: Family Tree of Linearly Homogenous Production Functions
MFM Production Function
Quadratic Mean of Order Alpha Transcendental
Generalized Leontief
Generalized Square
Root Quadratic
Linear Cobb-Douglas
c’ = 0 lim b0
b = 1
¸ = 0
lim b0
¸ = 0
or c = 0
CES Translog
Second-Order
Approx.
around c = 0
lim b0 c = b/2
c = 0
or ¸ = 0
¸ = 0
b = 1
b = 2
The IUP Journal of Managerial Economics, Vol. VIII, No. 4, 2010
14
generalization of numerous production functions, econometric estimation of MFM from
empirical data is problematic due to the fact that some parameters (especially b and c) tend
to be zero for some special cases. The parameter ¸ is a ratio ) 0 ; / (
3
= = b b a ¸ and it appears
as a factor of one of the coefficients. These peculiarities may have destabilizing effect on the
algorithm that estimates them.
By the middle of the 1970s, generalization of Cobb-Douglas and CES production
functions (in their classical form) was almost complete. In the classical form, these functions
assume that the marginal rate of substitution between any two factors of production is
associated only with relative factor prices and it is independent of technical progress or
level of output or, in technical terms, the technological progress is Hicks-neutral. To explain
this concept a little more, we note that a technological change describes a change in the set
of feasible production possibilities. A technological change is Hicks-neutral (Hicks, 1932),
if the ratio of capital’s marginal product to labor’s marginal product is unchanged for a
given capital-labor ratio. A technological change is said to be Harrod-neutral, if the
technology is labor-augmenting, and it is Solow-neutral, if the technology is capital-
augmenting. It is very easy to incorporate Hicks-neutral technical change into (a classical)
production function. The production function P = f(x) is modified as, P = e
¸ t
f(x) where e
¸ t
captures the technological change that does not modify the elasticity of substitution between
the factors of production.
Sato (1975) observed that so far the marginal rate of substitution function had
been specified as, ) / ( ln ) / 1 ( ) ( ln ) / ( ln L K a r w o + = where w and r are the prices of labor
(L) and capital (K), respectively, and o is the elasticity of substitution. In view of the
homotheticity assumption (assertion that a function g is a continuous positive
monotonically increasing function of a homogenous function f, such that if
) 0 / then )) , ( ( )) ( ( ) ( ); , ( ) (
2 1 2 1
> = = = = = dx dg x x f g x f g y g z x x f x f y implicit in the
(classical versions of) production functions (Cobb-Douglas or CES, etc., discussed so far),
the marginal rate of substitution was considered to be independent of the level of
production or neutral technical change. Empirical data in many cases, however, suggested
that the factor price ratio varies even at a constant input ratio. An introduction of
factor-augmenting technical progress (of Harrod or Solow) fails to perform due to
impossibility of identification of the bias (of technical progress) and substitution effect
(Sato, 1970). Therefore, Sato (1975) relaxed the homotheticity assumption so that the
level of output and the degree of neutral technical progress explicitly affect the factor
combinations or )), ( ( ln ) ( ln ) / ( ln ) / 1 ( ) ( ln ) / ( ln t T c P b L K a r w + + + = o where P is the
production and T(t) is the time dependent index of biased technical progress. From this
specification he obtained the ‘most general’ class of CES function, of which the classical
CES (as well as the Cobb-Douglas) and the ‘non-homothetic Cobb-Douglas’ production
functions are only special cases. Sato’s generalization permitted decomposition of income
and substitution effects (in the factor market) and distinction between normal and ‘inferior’
inputs.
15 A Brief History of Production Functions
Sato’s CES function: , 0 ) ( ) ( ) ( ) , , (
2 2 1 1 2 1
= + + = f H X f C X f C f X X F where
i i i i
x X u o
|
+ =
÷
(for o = 1) or
i i i i
x X u o + = ) ( ln (for o = 1), o and u are appropriate constants,
o o | / ) 1 ( ÷ = for a non-unitary (and non-zero) elasticity of substitution (o ), x
i
is a factor of
production, and ) ( ), ( f H f C
x
and f are defined appropriately according to homotheticity
(or otherwise) and separability. Since all homothetic functions are separable, Sato obtained
a threefold classification of CES. First, when 0 / ÷ df dC
i
and , 0 / = df dH we obtain
ordinary CES and Cobb-Douglas functions depending on whether o is non-unitary or
unitary. Second, when , 0 ) ( ) , , (
2 1 2 1
= + ÷ = X f C X f X X F the constants
i i i
o | o µ u ÷ ÷ = ) / 1 ( and , 0 , 0 ), / (
2 1 2 1
> < = |o |o X X V f where µ is the non-
homogeneity parameter, we have separable non-homothetic CES function or Cobb-Douglas
function depending on whether o is non-unitary or unitary (Sato, 1974). In this case,
0 / , 0 / = = df dH df dC
i
and
2 1
mC C = where m is a constant. Separable CES functions
are linear solutions of F(.). Finally, we have non-separable CES functions as the nonlinear
solutions of 0 ) ( ) ( ) , , (
2 1 2 1
= + + ÷ = f H X f C X f X X F in terms of f or C(f). In this case too,
2 1
and 0 / , 0 / mC C df dH df dC
i
= = = , where m is a constant. Examples of non-separable
CES are: if C(f) = af and H(f)=bf
2
, then f is given by ; 0 ) 2 /( ] } 4 ) {( [
2 1
1
2
2 2
> + ± ÷ b bX aX aX
if C(f)=af
2
and H(f )=b f, then we have . 0 ) 2 /( }] 4 { [
2 2 1
2
> + ± ÷ aX X aX b b f Note that, in
general, the function C(f ) is:
i
x
f h x x
f h x x
f C , 1 , / ) 1 ( ,
)) ( (
)) ( (
) (
2
1
2
1
1
1
1
1
= ÷ =
+ ÷
÷
=
÷ ÷ ÷ ÷
÷ ÷ ÷ ÷
o o o |
| |
| |
| |
| |
are
initial values of x
i
.
Sato showed that his generalized CES might easily be extended to n-inputs case as well
as to variable elasticity of substitution. If the elasticity of substitution depends on the level
of output then the ratio of factor prices , 0 ) ( , 0 ) ( ), ( ) / ( ) / (
) ( / 1
2 1
> > = f C f f C x x r w
f
o
o
in which case ). ( / )) ( 1 ( f f o o | ÷ = The elasticity of substitution is constant along an
isoquant, but it varies across the isoquants, as output varies. It allows for a case when an
isoquant in the (x
1
, x
2
) plane may be a Cobb-Douglas or (an ordinary) CES but another
isoquant can be a non-homothetic CES. For n-inputs case, if we define for any two inputs
i and j the ratio of factor prices as e
ij
and the elasticity of substitution between them as
n j i j i x x
ij j i ij
,..., 2 , 1 , , ), ( ln / ) / ( ln = = c c = = e o o
, then . 0 ), ( ) / (
1
> =
ij ij j i ij
C f C x x
o
e
We then have
¿
=
= + =
n
i
i i
f H X f C f X F
1
0 ) ( ) ( ) , ( where
i i i i
x X u o
|
+ =
÷
i i
X o = or
i i
x u + ) ( ln for a non-unitary or unitary o respectively. In the n-inputs case, we
may permit variability of the elasticity of substitution across the isoquants by defining
o = o (f ) = constant at any f. This grand generalization of the CES (and Cobb-Douglas)
functions possibly concluded an era of investigations on this topic.
The IUP Journal of Managerial Economics, Vol. VIII, No. 4, 2010
16
The importance of energy in the economic system was well-stressed by Georgescu-
Roegen (1971), well-known for versatility and competence in economics, mathematical
sciences, physical sciences, life sciences and philosophy alike. However, the energy crisis
due to Jom-Kippur and Iraq-Iran wars that threatened the US and many other economies
(in the mid 1970s and thereafter) led many economists to formulate energy-dependent as
well as other production functions that included energy and materials (besides conventional
labor and capital) as inputs (Hudson and Jorgenson, 1974; Tintner et al., 1974; and Berndt
and Wood, 1979). Kümmel (1982) and Kümmel et al. (1985 and 1998/2000) introduced the
linear-exponential (LINEX) production function that was based on physics and technology
as much as on economic considerations. It may be noted that the economists with technology
or physics background have almost always pleaded for incorporation of technological
considerations into production function, while the economists of the other category have
often limited themselves to mere economic considerations. Koopmans (1979) appears to be
in agreement with an engineer’s view of economics, saying: “Economics is not dismal but
incomplete. The things missed are very important” and a physical scientist saying:
“Economists are technologically radical. They assume everything can be done.” By the
way, it is worth noting that Koopmans was a physical scientist (as well as an economist)
of high order (Wikipedia-b). He is well-known for his theorem in molecular orbital theory
in quantum chemistry.
Chenery (1949 and 1950) was, perhaps, the first economist to demonstrate how
engineering information could be used to improve the empirical studies of production and
bridge the gap between theoretical and empirical analysis of production (Wibe, 1984).
In his paper, Soren Wibe has presented a detailed survey of engineering production
functions, so we do not repeat them here.
We return to the LINEX function (Lindenberger and Kümmel, 2002), which depends
linearly on energy (E) and exponentially on quotients of capital (K), labor (L) and
energy (E). The constant P
0
is a technology parameter indicating changes in the monetary
valuation of the original basket of goods and services making up the output unit, P. All
inputs and output are measured relative to some fixed quantity, E
0
, K
0
, L
0
and P
0
(and thus
all of them are indices with a fixed base). Creativity-induced innovations and structural
change make a
0
, c
0
, and P
0
time-dependent. With these variables and parameters, the function
is specified as: )]. 1 / ( } / ) ( 2 { exp[
0 0 0 0
÷ + + ÷ = E L c a K E L a E P P
Lindberger (2003) extended the above LINEX function to ‘service production functions’
that may be defined as }] / ) ( 2 { exp[ ) / (
0 0
0
K E L a L E L P P
m
c a
+ ÷ = or alternatively,
)}] / ( 1 { )} / ( ) / ( 2 3 { exp[
2
0
2
0 0
E L c a K LE K L a L P P
m
÷ + ÷ ÷ = , that emphasizes labor-
dependence of service production, and yields (non-negative) production elasticities,
. 1 )}; / ( ) / ( { ), 1 / )( / ( 2
2 2
0 0
¸ o | ¸ o ÷ ÷ = ÷ = + = K LE E L c a K E K L a
m
One may estimate the
function by any suitable algorithm (Mishra, 2006).
17 A Brief History of Production Functions
Multiple-Output Production Function
In the preceding narrative, we have dealt with the case when a producing agent produces
a single product. Now we turn to a case of multiple or joint production. It is worth mentioning
that Kurz (1986) presents a very lively account of multiple-output production function as
visualized by the classical and the (early) neoclassical economists. Salvadori and Steedman
(1988) review the concept in the Sraffian framework. A number of studies have been carried
out that deal with this topic. In particular, studies in agricultural (or farm) economics have
addressed this problem more frequently (Jawetz, 1961; Mundlak, 1963; Mundlak and Razin,
1971; Chizmar and Zak, 1983; Just et al., 1983; and Weaver, 1983). Studies in the economics
of household production and allocation of time between work and leisure (Becker, 1965;
Pollak and Wachter, 1975; Gronau, 1977; and Graham and Green, 1984) have also dealt
with joint production function. Of late, it has found recognition in the upcoming field of
environmental economics (Baumgärtner, 2004).
The first important question regarding a joint production function is its definition and
existence. While in case of a single-output technology, the production function was defined
as the maximal output obtainable from a given input vector (Shephard, 1970), no simple
maximal output exists for a multi-output technology, so that a multi-output production
function could be defined and its existence proved readily (Pfouts, 1961). Building upon
the work of Bol and Moechlin (1975), Al-Ayat and Färe (1977) examined necessary and
sufficient conditions for the existence of a joint production function within a general
framework of production correspondences. Without enforcing the strong disposability of
inputs or outputs it was shown that a joint production function exists if and only if both
input and output correspondences are strictly increasing along rays (implies that a
technology is both input and output ray homothetic). Subsequently, Färe (1986) put forth
three alternative ways in which multi-output production function might be defined.
As defined by Shephard (1970), it is an isoquant joint production function in the manner
that for a given pair of input and output vectors, the input vector and the output vector
belonged to the isoquant of input correspondence and the isoquant of output
correspondence respectively. The second follows Hanoch (1970) in which an efficient
joint production function characterizes input and output vectors that are simultaneously
input and output efficient. The third concept of joint production might relate weakly efficient
input vectors to weakly efficient output vectors.
The main finding of Färe is: let L(u) and L(v) denote all input vectors yielding
output u and v (respectively), and P(x) and P(y) denote all output vectors obtainable by
inputs x and y (respectively), then, for non-negative inputs and outputs (such that P(x) is
not zero and L(u) is not null) a joint production exists if and only if the sets S{L(u)} and
S{L(v)} are disjoint and the sets S{P(x)} and S{P(y)} are disjoint. More formally, a joint
production function exists if the following conditions are satisfied:
|
|
|
= = >
=
=
) ( }, 0 { ) ( that such 0 ,
, )} ( { )} ( {
, )} ( { )} ( {
u L x P u x
xRy y P S x P S
uRv v L S u L S


The IUP Journal of Managerial Economics, Vol. VIII, No. 4, 2010
18
The relationship R is ìx (ì = 1), >
*
and >, respectively (see Färe, 1986, p. 673). Weak
disposability of inputs and outputs is assumed. Färe also proved that under certain
conditions all the three definitions of joint production function are equivalent.
Econometric analysis of joint production perhaps dates back to the works of von
Stackelberg (1932) and Klein (1947). Since then a number of studies have been conducted
to estimate the parameters of multi-output or joint production functions. Methodologically
those studies may be classified under four heads: those formulating process analysis
models; those formulating simultaneous equations systems; those formulating composite
macro function; and those formulating composite implicit macro function. Some important
works are briefly reviewed as follows.
Baumgärtner (2001) points out the contributions of von Stackelberg (1932) to the theory
of joint production (and joint cost) function. Von Stackelberg conceptualized joint production
by introducing two new theoretical tools: the notions of length (r) and direction (¢) of the
production vector x = (x
1
, x
2
, …, x
m
). They are obtained by replacing the Cartesian coordinates
x
1
, x
2
, etc., by polar coordinates. In the case of just two outputs, for which von Stackelberg
defines these terms, x
1
and x
2
are replaced by r and ¢ with and
2
2
2
1
2
x x r + =
). / ( tan
1 2
1
x x
÷
= ¢ A production vector x is then uniquely determined by its length r and
its direction ¢. From this, two cases emerge: the first, when ¢ is a constant, in which case
the production function behaves like a single-output production function, and the second
in which ¢ is variable. In the latter case one may obtain the ‘curve of the most favorable
directions’ in output space. This curve is defined as the locus of all those production
programs (x
1
, x
2
), which yield a given revenue at the lowest possible total costs.
In the von Stackelberg schema, estimation of joint production function is rather simple
(Barrett and Hogset, 2003). Let n i p p p P
im i i i
,..., 2 , 1 ), ,..., , (
2 1
= = be n observations on m
joint products. Then, a point r
i
in m-dimensional (real) space is defined as
. ) ... (
5 . 0 2 2
2
2
1 im i i i
p p p r + + + = Only the positive value is taken (since r
i
is a length). This is used
as the dependent variable. The direction vectors are obtained as ). / ( cos
1
i ij ij
r p
÷
= u One of
the outputs (say, p
i1
) is considered as a standard measure or reference. These direction
vectors together with the inputs (x) are used as regressors. Thus,
n i x x x f r
ik i i im i i i
,..., 2 , 1 ), ,..., , ; ,..., , (
2 1 3 2
= = u u u are used for estimation of the parameters
of production function. Note that the first output is used as a standard, and hence u
i1
is not
included among the regressors. Implicitly it is assumed that all the products are identically
related to the inputs. It is also assumed that all the products are measured in the same
(Euclidean) space.
Mundlak (1963) approached estimation of joint production function through a different
sort of aggregation. His method lies in specifying the individual micro production function
for each (joint) product as well as the manner of aggregating them to an analogous macro
production function. The macro production function is then estimated and its relationship
19 A Brief History of Production Functions
with the micro production functions is investigated. However, the possibilities of
establishing the relationship among the macro and micro production functions depend
on availability of information on allocation of inputs used for different (joint) products.
Mundlak also proposed formulation and estimation of a general implicit production
function. This led to his further work (Mundlak, 1964) in which he formulated the problem
of estimation of multiple/joint production functions as an exercise in estimation of an
implicit function. If X are inputs and Yis output, then the implicit function g(f(X) – ¢(Y)) = 0
is expressed in terms of the composite input function f(X) and the composite output function
¢(Y). Mundlak illustrated his approach by the transcendental specification (proposed by
Halter et al., 1957) of the composite functions ), exp( ) (
2 2 1 1 2 1 0
2 1
x b x b x x a X f
a a
+ =
) exp( ) (
2 2 1 1 2 1
2 1
y d y d y y Y f
c c
+ = and the simple implicit function g(X, Y) = f(X) – ¢(Y) = 0.
It may be noted, however, that generally output is considered to contain errors due to
specification of f(X) such that any output vector y
k
= f(X) + u
k
, but inputs are considered
non-stochastic. This consideration would lead to the specification c ¢ = ÷ ))
ˆ
( ) ( ( Y X f g
where c is the disturbance term. The least squares estimation of such functions has remained
problematic. Mundlak and Razin (1971) also basically attempted aggregation of micro
functions to macro function.
Vinod (1968) addressed the problem of estimation of joint production function by
Hotelling’s canonical correlation analysis (Hotelling, 1936; and Kendall and Stuart, 1968).
Later, he improved his method to take care of the estimation problem if the data on output
(of different products) or inputs were collinear (Vinod, 1976). His method summarily lies,
first, in transforming the input vectors (X) and the output vectors (Y) into two composite
(weighted linear aggregate) vectors, U = Xw and V = Ye respectively, where the weights,
w and e, are (mathematically derived) so as to maximize the squared (simple product
moment) coefficient of correlation between U and V, and then transforming U and V back
into X and Y respectively. He showed that the back transformation of the composite vectors
U and V into X and Y poses no problem when the number of inputs is equal to or larger
than the number of output. However, when that is not the case, one has to resort to some
sort of least squares estimation (resulting from his suggested use of the least squares
generalized inverse in the transformation process).
There were strong reactions to Vinod’s method of estimation of joint production functions
(Chetty, 1969; Dhrymes and Mitchell, 1969; and Rao, 1969). Rao pointed out that to be
economically meaningful, the production function must be convex and the transformation
curve concave. However, the method proposed by Vinod did not yield composite output
function (transformation function) that satisfied these requirements. Dhrymes and Mitchell
(much like Chetty) pointed out that Vinod’s formulation was partly erroneous and partly
a “very complicated way of performing ordinary least squares.” If the ordinary least squares
method applied to estimate each production function separately and independently (ignoring
the fact that they relate to joint products) was inconsistent, then so would be the canonical
correlation method. While acceding to the errors pointed out by the critics, in his reply
The IUP Journal of Managerial Economics, Vol. VIII, No. 4, 2010
20
Vinod (1969) disagreed on the inconsistency issue shown to exist in his method and argued
that the critics (Dhrymes and Mitchell) had to establish the necessity and would not merely
put up some particular cases thereof. It is interesting, however, to note that Vinod undermined
the role of a single counterexample in demolishing the mathematical property of a method.
Apart from the problems pointed out above, Vinod’s method cannot be useful when
production functions are intrinsically nonlinear such that it is not possible to transform
them (by some simple procedure such as log-linearization) into linear equations. Secondly,
it may not be correct to form the composite output function in Vinod’s manner. Thirdly, it
is not necessary that the specification of production functions is identical for all products.
It is possible that while one of the products follows the CES, another follows the nested
CES (Sato, 1967) and yet another follows the Diewert (1971) or any other specification.
However, it is possible to specify production functions of different types for different
products and estimate their parameters jointly, although without making any composite
output function (Mishra, 2007b and 2007c).
Since the early work of Manne (1958), process analysis has amply exhibited its ability
to deal with the economics of joint products. However, it requires a large database and
solving large programming models. Further, it precludes the calculation of price and
substitution elasticities that may have important policy implications. Griffin (1977) used a
method similar to process analysis supplementing it with pseudo data to ascertain
appropriate types of production frontier functions for different joint products of petroleum
refinery. A pseudo data point shows the optimal input and output quantities corresponding
to a vector of input and output prices. By repetitive solution of the process model for
alternative price vectors, the shape of the production possibility frontier may be determined.
However, as pointed out by Griffin himself, the efficiency of pseudo data approach to
estimation of joint production functions ultimately rests on the quality of the engineering
process model often difficult for an economist to build or evaluate. Even then, this approach
does not rule out the possibilities of aggregation bias completely.
Just et al. (1983) formulated and estimated their multicrop production functions as a
system of nonlinear simultaneous equation model. The methods of estimation were
nonlinear two-stage and three-stage least squares. Chizmar and Zak (1983) discussed the
appropriateness of simultaneous equation modeling of multiple products raised or
manufactured simultaneously. However, they held that in case of joint products the implicit-
form single-equation modeling would be appropriate.
Aggregate Production Functions
Now we turn to the most turbulent area of research in the economics of production.
As long as a production function describes a relation that relates output and inputs of a
firm, the matters are more or less straight. However, when one makes an attempt to identify
such a relation at the industry, sector or economy-level, the matters may be quite different.
An industry is generally made up of numerous firms producing similar products in which
each firm uses inputs in its own accord with its own cost, returns to scale and market
21 A Brief History of Production Functions
implications. A production function (or a cost curve for that matter) at an industry level is
obtained by using the quantities of inputs and output aggregated over the constituent
firms. The question arises: does this aggregate production function represent the core
technological relation between inputs and output of a majority of firms? Or does it represent
the production function of a ‘Platonic’ or ‘archetypal’ firm that might not be a real firm, but
the empirically observed real firms are only the imperfect instances of that archetypal
firm? A little further, a firm might not possess certain characteristics but an industry may
possess them. A firm might be a price-taker in the factor market, but an industry might be
a price-maker. Rigidities might not permit a real firm to adjust to the changes at the industry
level and all input-output relations at the firm level would be suboptimal. Deliberate
keeping of room for adjustments and maneuvers and perpetual X-efficiency may invalidate
the assumption of technical efficiency and vitiate a search for allocative efficiency. These
issues and many others turn out to be more and more significant when one moves higher
to macroeconomic levels.
Since Adam Smith (or even before him) economists of a particular hue have largely
been preoccupied with intellectual work to prove that an autonomous management (and
organization) of a society by capitalistic principles (characterizing self-interest guided
agents operating in an institutional framework of private property, market economy,
competition, accumulation of capital, etc.) is the best, just, stable and most viable among all
possible approaches to manage (and organize) a society. A society organized on that
principle would grow (expand) indefinitely and deliver justice to all the agents. However,
Karl Marx questioned the efficacy of the capitalistic system in not only delivering justice
but also guaranteeing indefinite expansion and stability. After Marx, the (neoclassical)
economists, therefore, had to invent strong arguments to defend the legitimacy, efficacy
and supremacy of the capitalistic system. The Walrasian general equilibrium, Pareto-
optimality of the competitive economy, aggregate production function, marginal
productivity theory of distribution, product exhaustion theorem, and ultimately Harrod’s,
Solow’s and von Neumann’s paths to expansion are only some major lemmas to prove the
said grand proposition.
Whitaker (1975) brings to the light some unpublished works of Alfred Marshall and
points out how Marshall formulated his aggregate production function P = f (L.E,C,A,F),
where L is labor and E its efficiency, C is Capital, A is level of technology and F is fertility
of soil. In this aggregate production function, Marshall used the time derivative of the
variables and therefore it may be regarded as the first neoclassical growth model,
foreshadowing the Tinbergen-Solow growth model.
Knut Wicksell made significant contribution to demonstrate that non-homogeneous
production functions for firms are perfectly compatible with a linear homogeneous function
for the entire industry. Suppose industry output expands and contracts through the entry
and exit of identical firms, each operating at the same minimum unit cost. The result is to
trace out a horizontal long-run industry supply curve that looks like it came from a constant-
The IUP Journal of Managerial Economics, Vol. VIII, No. 4, 2010
22
returns production function. Thus, he justified the use of aggregate linear homogeneous
functions such as the Cobb-Douglas function (Humphrey, 1997).
The Cobb-Douglas production function gave a readily convincing proof that in
competitive equilibrium all inputs are paid their marginal product (and hence their
respective real price), the entire product exhausts (as the sum of input elasticities of product
sum up to unity), constant returns to scale prevails, and the empirically observed constancy
of relative shares of factors of production for long periods is fully explicable, justified and
natural. This finding strengthened the foundations of using aggregate production function
at the macroeconomic level. Humphrey (1997) has presented this development so lucidly
that it is best to quote him: “Each stage saw production functions applied with increasing
sophistication. First came the idea of marginal productivity schedules as derivatives of a
production function. Next came numerical marginal schedules whose integrals constitute
particular functional forms indispensable in determining factor prices and relative shares.
Third appeared the path-breaking initial statement of the function in symbolic form. The
fourth stage saw a mathematical production function employed in an aggregate
neoclassical growth model. The fifth stage witnessed the flourishing of microeconomic
production functions in derivations of the marginal conditions of optimal factor hire.
Sixth came the demonstration that product exhaustion under marginal productivity requires
production functions to exhibit constant returns to scale at the point of competitive
equilibrium. Last came the proof that functions of the type later made famous by Cobb-
Douglas satisfy this very requirement. In short, macro and micro production functions
and their appurtenant concepts—marginal productivity, relative shares, first-order
conditions of factor hire, product exhaustion, homogeneity and the like—already were
well advanced when Cobb and Douglas arrived”.
During the post-Great Depression period until the end of the Second World War,
economists investigated into the possibilities of growth without violent fluctuations. In
this investigation the aggregate production function proved to be very useful. The Cobb-
Douglas production function was found quite amenable to incorporation of technical
change introduced into the production system from time to time without altering the basic
conclusions on factor shares. The growth model of von Neumann (1937) deviated from
using the Cobb-Douglas production function but retained the practice of aggregation.
Consequent upon the development of linear programing as a method of optimization, this
line of investigation progressed very rapidly. Activity analysis of Koopmans, input-output
analysis of Leontief, aggregate linear production function of Georgescu-Roegen (1951),
separation theorems and generalization of von Neumann’s model by Gale (1956), and
careful proofs given by Nikaido (1968) all strengthened the foothold of aggregate production
function in economic analysis.
However, the 1950s did not welcome the aggregate production function wholeheartedly.
Robinson (1953) viewed aggregate production function with a remark: “. . . the production
function has been a powerful instrument of miseducation. The student of economic theory
23 A Brief History of Production Functions
is taught to write Q =f (L, K ) where L is a quantity of labor, K a quantity of capital and Q a
rate of output of commodities. He is instructed to assume all workers alike, and to measure
L in man-hours of labor; he is told something about the index-number problem in choosing
a unit of output; and then he is hurried on to the next question, in the hope that he will
forget to ask in what units K is measured. Before he ever does ask, he has become a professor,
and so sloppy habits of thought are handed on from one generation to the next.”
A controversy began, especially regarding the measurement of capital, in which Piero
Sraffa, Joan Robinson, Luigi Pasinetti and Pierangelo Garegnani (among others) argued
against the use of aggregate production function, and Paul Samuelson, Robert Solow, Frank
Hahn and Christopher Bliss (among others) argued in favor of using aggregate production
function for explaining relative factor shares. As argued by the first group, it is impossible to
conceive of an abstract quantity of capital which is independent of the rates of interest and
wages. However, this independence is a precondition of constructing an isoquant (or
production function). The isoquant cannot be constructed and its slope measured unless the
prices are known beforehand, but the protagonists of aggregate production function use the
slope of the isoquant to determine relative factor prices. This is begging the question.
This controversy (that began in 1953) lasted until the mid-1970s. A very lively account
of it was presented by Harcourt (1969), who was himself a participant in the controversy
(from the Robinson’s side), and as Stiglitz (1974) notes, may not possibly have been impartial
in giving an unbiased account. However, it may be noted that Stiglitz admits that he was
a participant from the other side. A much more comprehensive account of the controversy
may be found in Cohen and Harcourt (2003).
The controversy exposed all aggregate production functions, the Cobb-Douglas
production function in particular, and proved almost conclusively that it has no economics
in it as its properties stem from mere algebra. A series of three papers by Anwar Shaikh
(Shaikh, 1974, 1980 and 2005) on his ‘humbug production function’ should have ousted
the Cobb-Douglas production function from all serious endeavors in economic analysis.
Felipe, Fisher and their associates in a number of papers (Felipe and Fisher, 2001; Felipe
and Adams, 2005; Felipe and McCombie, 2005; and Felipe and Holz, 2005) have exposed
the validity of an aggregate production function.
Once Samuelson (1966) wrote: “Until the laws of thermodynamics are repealed,
I shall continue to relate outputs to inputs—i.e., to believe in production functions. Unless
factors cease to have their rewards to be determined by bidding in quasi-competitive
markets, I shall adhere to (generalized) neoclassical approximations in explaining their
market remunerations.” One fails to understand as to how the validity of laws of
thermodynamics and relating outputs to inputs entails validity of the proposition that the
aggregate of functions would be the function of aggregates, and if there is some particular
type of function that has this property, then how that particular function is the correct
function describing the production technology of any economy and the relative factor
The IUP Journal of Managerial Economics, Vol. VIII, No. 4, 2010
24
shares. Solow (1957) made an impressive empirical study to demonstrate how the aggregate
production function fits to the US data for 1909-49, which confirms neutral technical
change, shift in production function, and therefore validates the artifact of aggregate
production function as a powerful tool of analysis. Samuelson (1962) invented the
so-called ‘surrogate production function’ which related the total quantity of output per
capita to the total quantity of capital per capita, which can be used to predict all behavior
in the sense of the wage and profit rates that would prevail in different long-run equilibria,
which, according to Harris (1973), was based on too extremely restrictive assumptions
and therefore was too fragile to hold true in reality.
Nonreswitching theorem was a major plank to support measurability of capital and its
aggregation. Levhari and Samuelson published a paper which began, “We wish to make
it clear for the record that the nonreswitching theorem associated with us is definitely
false. We are grateful to Dr. Pasinetti...” (Levhari and Samuelson, 1966). The final outcome
of the controversy may be best concluded in the words of Burmeister (2000): “… the damage
had been done, and Cambridge, UK, ‘declared victory’: Levhari was wrong, Samuelson
was wrong, Solow was wrong, MIT was wrong and therefore neoclassical economics was
wrong. As a result there are some groups of economists who have abandoned neoclassical
economics for their own refinements of classical economics. In the US, on the other hand,
mainstream economics goes on as if the controversy had never occurred. Macroeconomics
textbooks discuss ‘capital’ as if it were a well-defined concept—which it is not, except in a
very special one-capital-good world (or under other unrealistically restrictive conditions).
The problems of heterogeneous capital goods have also been ignored in the ‘rational
expectation revolution’ and in virtually all econometric work”. The rational expectation
theory was propounded by Lucas (1976).
The effects of ‘capital controversy’ beginning with a criticism of using aggregate
production function for explaining (and justifying) relative factor shares had a disastrous
effect on neoclassical economics. Lavoie (2000) observed: “Capital reversing renders
meaningless the neoclassical concepts of input substitution and capital scarcity or labor
scarcity. It puts in jeopardy the neoclassical theory of capital and the notion of input
demand curves, both at the economy and industry levels. It also puts in jeopardy the
neoclassical theories of output and employment determination, as well as Wicksellian
monetary theories, since they are all deprived of stability. The consequences for neoclassical
analysis are thus quite devastating. It is usually asserted that only aggregate neoclassical
theory of the textbook variety—and hence macroeconomic theory, based on aggregate
production functions—is affected by capital reversing. It has been pointed out, however,
that when neoclassical general equilibrium models are extended to long-run equilibria,
stability proofs require the exclusion of capital reversing…. In that sense, all neoclassical
production models would be affected by capital reversing.”
Gehrke and Lager (2000) observed: “These findings destroy, for example, the general
validity of Heckscher-Ohlin-Samuelson international trade theory …, of the Hicksian
25 A Brief History of Production Functions
neutrality of technical progress concept …, of neoclassical tax incidence theory …, and of
the Pigouvian taxation theory applied in environmental economics …”
What has remained with the believers in neoclassical aggregate production function is
the complaint that neither Joan Robinson nor her associates developed an alternative set
of theoretical (as opposed to descriptive) tools that avoid her concerns about the limitations
of equilibrium analysis.
Conclusion
It is said that once Stanislaw Ulam very earnestly sought for an example of a theory in
social sciences that is both true and nontrivial. Paul Samuelson, after several years supplied
one: the Ricardian theory of comparative advantages. If this is true then it speaks enough
about the position of ‘production function’ in economics. Anwar Shaikh almost conclusively
proved that the properties of the Cobb-Douglas production function are devoid of any
economic content; they stem from the algebraic properties of the function. Measurement of
capital was put in jeopardy by the capital controversy, not only for aggregate production
function, but also for any production function even at the firm level. If one has interacted
with engineers and the managers who are decision makers at certain level, one might
know their view of the utility of production functions. Possibly, Joan Robinson was right
to comment that the production function has been a powerful instrument of miseducation.
The student of economic theory is taught to write Q =f (L, K ), and that is its sole utility. The
era of classical economics had ended (sometime in 1870s) with its criticism by Marx and
the birth of neoclassicism. The era of neoclassical economics possibly ended with the
capital controversy sometime in 1970s.
Bibliography
1. Al-Ayat R and Färe R (1977), “On the Existence of Joint Production Functions”, Research
Report, Operations Research Center, California University, Berkeley; subsequently
published in 1979 in Naval Research Logistics Quarterly, Vol. 26, No. 4, pp. 627-630.
2. Arrow K J, Chenery H B, Minhas B S and Solow R M (1961), “Capital Labour Substitution
and Economic Efficiency”, Review of Econ. and Statistics, Vol. 63, No. 3, pp. 225-250.
3. Banker R D, Charnes R F and Cooper W W (1984), “Some Models for Estimating
Technical and Scale Inefficiencies in Data Envelopment Analysis”, Management Science,
Vol. 30, No. 9, pp. 1078-1092.
4. Barkai H (1959), “Ricardo on Factor Prices and Income Distribution in a Growing
Economy”, Economica, Vol. 26, No. 103, pp. 240-250.
5. Barrett C and Hogset H (2003), “Estimating Multiple-Output Production Functions for
the CLASSES Model”, Available at http://aem.cornell.edu/special_programs/
AFSNRM/Basis/Documents/Memos/Multi-outputProductionFunctionEstimation
forCLASSES.pdf
The IUP Journal of Managerial Economics, Vol. VIII, No. 4, 2010
26
6. Baumgärtner S (2001), “Heinrich von Stackelberg on Joint Production”, Euro. J. History
of Economic Thought, Vol. 8, No. 4, pp. 509-525.
7. Baumgärtner S (2004), “Price Ambivalence of Secondary Resources: Joint Production,
Limits to Substitution, and Costly Disposal”, Resources, Conservation and Recycling,
Vol. 43, No. 1, pp. 95-117.
8. Becker G S (1965), “A Theory of the Allocation of Time”, The Economic Journal, Vol. 75,
No. 299, pp. 493-517.
9. Berndt E and Christensen L (1973), “The Translog Function and the Substitution of
Equipment, Structures and Labor in US Manufacturing, 1929-1968”, Journal of
Econometrics, Vol. 1, No. 1, pp. 81-114.
10. Berndt E R and Wood D O (1979), “Engineering and Econometric Interpretations of
Energy-Capital Complementarity”, American Economic Review, Vol. 69, No. 3,
pp. 342-354.
11. Blaug M (1985), Economic Theory in Retrospect, 4
th
Edition, Cambridge University Press,
Cambridge.
12. Bol G and Moechlin O (1975), “Isoquants of Continuous Production
Correspondences”, Naval Research Logistics Quarterly, Vol. 22, No. 3, pp. 391-398.
13. Brown M and Cani J S de (1963), “Technological Change and the Distribution of
Income”, International Economic Review, Vol. 4, No. 3, pp. 289-309.
14. Bruno M (1962), “A Note on the Implications of an Empirical Relationship Between
Output Per Unit of Labour, the Wage Rate and the Capital-Labour Ratio”, Unpublished
Mimeo, Stanford University.
15. Bruno M (1968), “Estimation of Factor Contribution to Growth Under Structural
Disequilibrium”, International Economic Review, Vol. 9, No. 1, pp. 49-62.
16. Burmeister E (2000), “The Capital Theory Controversy”, in H D Kurz (Ed.), Critical
Essays on Piero Sraffa’s Legacy in Economics, Cambridge University Press, Cambridge.
17. Center for Economic Policy Analysis (CEPA), “The Production Function”, Available
at http://cepa.newschool.edu/het/essays/product/prodfunc.htm
18. Charnes A, Cooper W W and Rhodes E (1978), “Measuring the Efficiency of Decision-
making Units”, European Journal of Operational Research, Vol. 2, No. 6, pp. 429-444.
19. Chenery H B (1949), “Engineering Production Functions”, Quarterly Journal of
Economics, Vol. 63, No. 4, pp. 507-531.
20. Chenery H B (1950), “Engineering Bases of Economic Analysis”, Ph.D. Dissertation
(Supervisor W W Leontief), Harvard University, Harvard.
21. Chetty V K (1969), “Econometrics of Joint Production: A Comment”, Econometrica,
Vol. 37, No. 4, p. 731.
27 A Brief History of Production Functions
22. Chizmar J F and Zak T A (1983), “Modeling Multiple Outputs in Educational
Production Functions”, The American Economic Review, Vol. 73, No. 2, pp. 18-22.
23. Christensen L R, Jorgenson D W and Lau L J (1973), “Transcendental Logarithmic
Production Frontiers”, The Review of Economics and Statistics, Vol. 55, No. 1, pp. 28-45.
24. Cobb C W and Douglas P H (1928), “A Theory of Production”, American Economic
Review, Vol. 18, No. 1, pp. 139-165.
25. Cohen A J and Harcourt G C (2003), “Retrospectives: Whatever Happened to the
Cambridge Capital Theory Controversies?”, Journal of Economic Perspectives, Vol. 17,
No. 1, pp. 199-214.
26. Dhrymes P J and Mitchell B M (1969), “Estimation of Joint Production Functions”,
Econometrica, Vol. 37, No. 4, pp. 732-736.
27. Diewert W E (1971), “An Application of the Shephard Duality Theorem: A Generalized
Leontief Production Function”, The Journal of Political Economy, Vol. 79, No. 3,
pp. 481-507.
28. Emrouznejad A (2001), “An Extensive Bibliography of Data Envelopment Analysis
(DEA)”, Vol. I, Working Paper, Business School, University of Warwick.
29. Faber M, Proops J L R and Baumgärtner S (1998), “All Production is Joint Production:
A Thermodynamic Analysis”, in S Faucheux, J Gowdy and I Nicolaï (Eds.),
Sustainability and Firms: Technological Change and the Changing Regulatory Environment,
pp. 131-158, Edward Elgar, Cheltenham.
30. Färe R (1986), “On the Existence and Equivalence of Three Joint Production Functions”,
Scand. J. of Economics, Vol. 88, No. 4, pp. 669-674.
31. Färe R and Mitchell T M (1989), “A Family Tree of Linearly Homogenous Production
Functions”, Scand. J. of Economics, Vol. 91, No. 4, pp. 749-757.
32. Farrell J M (1957), “The Measurement of Productive Efficiency”, Journal of the Royal
Statistical Society, Vol. 120, No. 3, pp. 253-281.
33. Felipe J and Fisher F M (2001), “Aggregation in Production Functions: What Applied
Economists Should Know”, Metroeconomica, Vol. 54, pp. 208-262, Available at http://
ssrn.com/abstract=422067
34. Felipe J and Adams F G (2005), “‘A Theory of Production’: The Estimation of the Cobb-
Douglas Function: A Retrospective View”, Eastern Economic Journal, Vol. 31, No. 5,
pp. 427-445.
35. Felipe J and Holz C A (2005), “On Production Functions, Technical Progress and
Time Trends”, Available at http://repository.ust.hk/dspace/bitstream/1783.1/
2200/1/FelipeHolzprerefereedAggregateProdFcn3Nov98.pdf
36. Felipe J and McCombie J S L (2005), “How Sound are the Foundations of the Aggregate
Production Function?”, Eastern Economic Journal, Vol. 31, No. 5, pp. 467-488.
The IUP Journal of Managerial Economics, Vol. VIII, No. 4, 2010
28
37. Fioretti G (2006), “Production Function”, Working Paper, Preprint available at
http://arxiv.org/PS_cache/physics/pdf/0511/0511191.pdf
38. Fisher F M (1993), Aggregation, Aggregate Production Functions and Related Topics, The
MIT Press, Cambridge, MA.
39. Fisher F M (2005), “Aggregate Production Functions: A Pervasive, but Unpersuasive,
Fairytale”, Eastern Economic Journal, Vol. 31, No. 5, pp. 489-491.
40. Frisch R (1965), Theory of Production, D Reidel, Dordrecht.
41. Gale D (1956), “A Closed Linear Model of Production”, in H W Kuhn et al. (Eds.),
Linear Inequalities and Related Systems, pp. 285-303, Princeton University Press.
42. Gehrke C and Lager C (2000), “Sraffian Political Economy”, Encyclopedia of Political
Economy, Routledge.
43. Georgescu-Roegen N (1951), “The Aggregate Linear Production Function and Its
Applications to von Neumann’s Economic Model”, in T C Koopmans (Ed.), Activity
Analysis of Production and Allocation, No. 13, Cowles Commission Monograph, Wiley,
New York.
44. Georgescu-Roegen N (1971), The Entropy Law and the Economic Process, Harvard
University Press, Cambridge, Mass.
45. Graham J W and Green C A (1984), “Estimating the Parameters of a Household
Production Function with Joint Products”, The Review of Economics and Statistics,
Vol. 66, No. 2, pp. 277-282.
46. Griffin J M (1977), “The Economics of Joint Production: Another Approach”, The
Review of Economics and Statistics, Vol. 59, No. 4, pp. 389-397.
47. Griliches Z (1969), “Capital-Skill Complementarity”, Review of Economics and Statistics,
Vol. 51, No. 4, pp. 465-468.
48. Griliches Z and Ringstad V (1971), Economies of Scale and the Form of Production Function,
North-Holland Publishing Co., Amsterdam.
49. Gronau R (1977), “Leisure, Home Production, and Work: The Theory of the Allocation
of Time Revisited”, The Journal of Political Economy, Vol. 85, No. 6, pp. 1099-1123.
50. Halter A N, Carter H O and Hocking J G (1957), “A Note on the Transcendental
Production Function”, Journal of Farm Economics, Vol. 39, No. 4, pp. 966-974.
51. Hanoch G (1970), “Homotheticity in Joint Production”, Journal of Economic Theory,
Vol. 2, No. 4, pp. 423-426.
52. Harcourt G C (1969), “Some Cambridge Controversies in the Theory of Capital”, Journal
of Economic Literature, Vol. 7, No. 2, pp. 369-405.
53. Harris D J (1973), “Capital, Distribution, and the Aggregate Production Function”,
The American Economic Review, Vol. 63, No. 1, pp. 100-113.
29 A Brief History of Production Functions
54. Hicks J R (1932), The Theory of Wages, 2
nd
Edition (1963), St. Martin’s Press, New York.
55. Hotelling H (1936), “Relations Between Two Sets of Variates”, Biometrica, Vol. 28,
Nos. 3-4, pp. 321-377.
56. Hudson E A and Jorgenson D W (1974), “US Energy Policy and Economic Growth,
1975-2000”, The Bell Journal of Economics and Management Science, Vol. 5,
No. 2, pp. 461-514.
57. Humphrey T M (1997), “Algebraic Production Functions and Their Uses Before Cobb-
Douglas”, Economic Quarterly, Vol. 83, No. 1, pp. 51-83, Federal Reserve Bank of
Richmond, Available at http://ideas.repec.org/a/fip/fedreq/y1997iwinp51-83.html
58. Intriligator M D (1978), Econometric Models, Techniques and Applications, Prentice Hall
Inc., New Jersey.
59. Jawetz M B (1961), “The Joint Production Function of Starch Equivalent and Protein
Equivalent in Feeding Dairy Cows”, Department of Economics, University College of
Wales, Aberystwyth.
60. Just R E, Zilberman D and Hochman E (1983), “Estimation of Multicrop Production
Functions”, American Journal of Agricultural Economics, Vol. 65, No. 4, pp. 770-780.
61. Kadiyala K R (1972), “Production Functions and Elasticity of Substitution”, Southern
Economic Journal, Vol. 38, No. 3, pp. 281-284.
62. Kendall M G and Stuart A (1968), The Advanced Theory of Statistics, Vol. 3, Charles
Griffin & Co., London.
63. Klein L R (1947), “The Use of Cross-Section Data in Econometrics with Application to
a Study of Production of Railroad Services in the United States” (Mimeographed),
National Bureau of Economic Research, Washington DC.
64. Kmenta J (1967), “On Estimation of the CES Production Function”, International
Economic Review, Vol. 8, No. 2, pp. 180-189.
65. Koopmans T C (1979), “Economics Among the Sciences”, The American Economic Review,
Vol. 69, No. 1, pp. 1-13.
66. Kümmel R (1982), “The Impact of Energy on Industrial Growth”, Energy:
The International Journal, Vol. 7, No. 2, pp. 189-203.
67. Kümmel R, Strassl W, Gossner A and Eichhorn W (1985), “Technical Progress and
Energy Dependent Production Functions”, Journal of Economics, Vol. 45, No. 3,
pp. 285-311.
68. Kümmel R, Lindenberger D and Eichhorn W (2000), “The Productive Power of Energy
and Economic Evolution”, Indian Journal of Applied Economics, Vol. 8, No. 2, pp. 1-26,
Available at ftp://ftp.physik.uni-wuerzburg.de/pub/preprint/1998/WUE-ITP-98-
033.ps.gz
The IUP Journal of Managerial Economics, Vol. VIII, No. 4, 2010
30
69. Kurz H D (1986), “Classical and Early Neoclassical Economists on Joint Production”,
Metroeconomica, Vol. 38, No. 1, pp. 1-37.
70. Lavoie M (2000), “Capital Reversing”, Encyclopedia of Political Economy, Routledge.
71. Leibenstein H, Blair G and Hodgson M (1988), “Allocative Efficiency versus
X-Efficiency”, American Economic Review, Vol. 56, No. 3, pp. 392-415.
72. Levhari D and Samuelson P A (1966), “The Nonswitching Theorem is False”, The
Quarterly Journal of Economics, Vol. 80, No. 4, pp. 518-519.
73. Lindenberger D (2003), “Service Production Functions”, EWI Working Paper
No. 03.02, Institute of Energy Economics, University of Cologne (EWI), Cologne,
Available at http://www.ewi.unikoeln.de/ewi/content/e266/e283/e281/
Ewiwp0302_ger.pdf
74. Lindenberger D and Kümmel R (2002), “Energy-Dependent Production Functions
and the Optimization Model ‘PRISE’ of Price-Induced Sectoral Evolution”, International
Journal of Applied Thermodynamics, Vol. 5, No. 3, pp. 101-107.
75. Liu T C and Hildebrand G H (1965), Manufacturing Production Functions in the United
States, 1957, Cornell University Press, Ithaca.
76. Lloyd P J (1969), “Elementary Geometric/Arithmetic Series and Early Production
Theory”, Journal of Political Economy, Vol. 77, No. 1, pp. 21-34.
77. Lovell C A L and Schmidt P (1988), “A Comparison of Alternative Approaches to the
Measurement of Productive Efficiency”, in A Dogramaci and R Färe (Eds.), Applications
of Modern Production Theory: Efficiency and Productivity, pp. 3-32, Kluwer, Boston.
78. Lu Y C and Fletcher L B (1968), “A Generalization of the CES Production Function”,
Review of Economics and Statistics, Vol. 50, No. 4, pp. 449-452.
79. Lucas R (1976), “Econometric Policy Evaluation: A Critique”, Carnegie-Rochester
Conference Series on Public Policy, Vol. 1, No. 1, pp. 19-46.
80. Manne A S (1958), “A Linear Programming Model of the US Petroleum Refinery
Industry”, Econometrica, Vol. 26, No. 1, pp. 67-106.
81. McCarthy M D (1967), “Approximation of the CES Production Function: A Comment”,
International Economic Review, Vol. 8, No. 2, pp. 190-192.
82. McFadden D (1962), Factor Substitution in the Economic Analysis of Production, Ph.D.
Dissertation, University of Minnesota.
83. McFadden D (1963), “Constant Elasticity of Substitution Production Function”, Review
of Economic Studies, Vol. 30, No. 1, pp. 73-83.
84. Mishra S K (2006), “A Note on Numerical Estimation of Sato’s Two-Level CES
Production Function”, Available at http://www.ssrn.com/author=353253
31 A Brief History of Production Functions
85. Mishra S K (2007a), “Estimation of Zellner-Revankar Production Function Revisited”,
Economics Bulletin, Vol. 3, No. 14, pp. 1-7.
86. Mishra S K (2007b), “Least Squares Estimation of Joint Production Functions by the
Differential Evolution Method of Global Optimization”, Economics Bulletin, Vol. 3,
No. 51, pp. 1-13, Available at http://economicsbulletin.vanderbilt.edu/2007/
volume3/EB-07C10007A.pdf
87. Mishra S K (2007c), “Performance of Differential Evolution Method in Least Squares
Fitting of Some Typical Nonlinear Curves”, Available at http://ssrn.com/
abstract=1010508
88. Mitscherlich E A (1909), “Das Gesetz des Minimums und das Gesetz des abnehmenden
Bodenertrages”, Landw. Jahrb., Vol. 38, No. 1009, pp. 537-552.
89. Mukerji V (1963), “A Generalized SMAC Function with Constant Ratios of Elasticities
of Substitution”, Review of Economic Studies, Vol. 30, No. 2, pp. 233-236.
90. Mundlak Y (1963), “Specification and Estimation of Multiproduct Production
Functions” (Mimeographed), Paper Read at the Meetings of the Econometric Society
and the American Farm Economic Association, Pittsburgh, USA, December 1962,
Summarized in the Journal of Farm Economics, Vol. 45, No. 2, pp. 433-443.
91. Mundlak Y (1964), “Transcendental Multiproduct Production Functions”, International
Economic Review, Vol. 5, No. 3, pp. 273-284.
92. Mundlak Y and Razin A (1971), “On Multistage Multiproduct Production Functions”,
American Journal of Agricultural Economics, Vol. 53, No. 3, pp. 491-499.
93. Nerlove M (1963), “Returns to Scale in Electricity Supply”, Reprinted in M Nerlove
(1965), Estimation and Identification of Cobb-Douglas Production Functions, North-Holland
Publishing Co., Amsterdam.
94. Nikaido H (1968), Convex Structures and Economic Theory, Academic Press, New York.
95. Pfouts R W (1961), “The Theory of Cost and Production in the Multiproduct Firm”,
Econometrica, Vol. 29, No. 3, pp. 65-68.
96. Pollak R A and Wachter M L (1975), “The Relevance of the Household Production
Function and Its Implications for the Allocation of Time”, The Journal of Political
Economy, Vol. 83, No. 2, pp. 255-278.
97. Rao P (1969), “A Note on Econometrics of Joint Production”, Econometrica, Vol. 37,
No. 4, pp. 737-738.
98. Revankar N S (1967), “Production Functions with Variable Elasticity of Substitution
and Variable Returns to Scale”, Doctoral Dissertation, University of Wisconsin, USA.
99. Revankar N S (1971), “A Class of Variable Elasticity of Substitution Production
Functions”, Econometrica, Vol. 39, No. 1, pp. 61-71.
The IUP Journal of Managerial Economics, Vol. VIII, No. 4, 2010
32
100. Ringstad V (1967), “Econometric Analysis Based on Production Function with
Neutrally Variable Scale Elasticity”, Swedish Journal of Economics, Vol. 69, No. 2,
pp. 115-123.
101. Robinson J (1953), “The Production Function and the Theory of Capital”, Review of
Economic Studies, Vol. 21, No. 2, pp. 81-106.
102. Russell B (1984), A History of Western Philosophy, Counterpoint Edition, Unwin
Paperbacks, London.
103. Salvadori N and Steedman Z (1988), “Joint Production Analysis in a Sraffian
Framework”, Bulletin of Economic Research, Vol. 40, No. 3, pp. 165-196.
104. Samuelson P A (1962), “Parable and Realism in Capital Theory: The Surrogate
Production Function”, Review of Economic Studies, Vol. 29, No. 3, pp. 193-206.
105. Samuelson P A (1966), “Rejoinder: Agreements, Disagreements, Doubts, and the Case
of Induced Harrod-Neutral Technical Change”, Review of Economics and Statistics,
Vol. 48, No. 4, pp. 444-448.
106. Samuelson P A (1979), “Paul Douglas’s Measurement of Production Functions and
Marginal Productivities”, Journal of Political Economy, Vol. 87, No. 5, pp. 923-939.
107. Sato K (1967), “A Two-Level Constant-Elasticity-of-Substitution Production Function”,
Review of Economic Studies, Vol. 34, No. 2, pp. 201-218.
108. Sato R (1970), “The Estimation of Biased Technical Progress and the Production
Function”, International Economic Review, Vol. 11, No. 2, pp. 179-208.
109. Sato R (1974), “On the Class of Separable Non-Homothetic CES Functions”, Economic
Studies Quarterly, Vol. 25, No. 1, pp. 42-55.
110. Sato R (1975), “The Most General Class of CES Functions”, Econometrica, Vol. 43,
Nos. 5-6, pp. 999-1003.
111. Sato R and Hoffman R F (1968), “Production Functions with Variable Elasticity of
Factor Substitution: Some Analysis and Testing”, Review of Economics and Statistics,
Vol. 50, No. 4, pp. 453-460.
112. Schumpeter J A (1954), History of Economic Analysis, Allen and Unwin, London.
113. Seiford L M and Thrall R M (1990), “Recent Developments in DEA: The Mathematical
Programming Approach to Frontier Analysis”, Journal of Econometrics, Vol. 46,
No. 1, pp. 7-38.
114. Shaikh A (1974), “Laws of Production and Laws of Algebra: The Humbug Production
Function”, The Review of Economics and Statistics, Vol. 56, No. 1, pp. 115-120, Reprint
available at http://homepage.newschool.edu/~AShaikh/humbug.pdf
115. Shaikh A (1980), “Laws of Production and Laws of Algebra: Humbug II”, in E J Nell
(Ed.), Growth, Profits and Property, Cambridge University Press, Cambridge, Reprint
available at the website http://homepage.newschool.edu/~AShaikh/humbug2.pdf
33 A Brief History of Production Functions
116. Shaikh A (2005), “Nonlinear Dynamics and Pseudo-Production Functions”, Eastern
Economic Journal, Vol. 31, No. 5, pp. 447-466.
117. Shephard R W (1970), Theory of Cost and Production Functions, Princeton University
Press, Princeton, NJ.
118. Solow R M (1957), “Technical Change and the Aggregate Production Function”, The
Review of Economics and Statistics, Vol. 39, No. 3, pp. 312-320.
119. Spillman W J and Lang E (1924), The Law of Diminishing Returns, World Book Co.,
Yonkers-on-Hudson, New York.
120. Stigler G J (1952), “The Ricardian Theory of Value and Distribution”, Journal of Political
Economy, Vol. 60, No. 3, pp. 187-207.
121. Stiglitz J E (1974), “The Cambridge-Cambridge Controversy in the Theory of Capital:
A View from New Haven”, The Journal of Political Economy, Vol. 82, No. 4, pp. 893-903.
122. Thünen Johann von (1930), Der isolierte Staat in Beziehung auf Landwirthschaft und
National¨okonomie, 3 Volumes, Fischer, Jena, Germany.
123. Tintner G, Deutsch E and Rieder R (1974), “A Production Function for Austria
Emphasizing Energy”, in F L Altman, O Kyn and H J Wagener (Eds.), On the
Measurement of Factor Productivities, pp. 151-164, Vandenhoek & Ruprecht, Göttingen.
124. Uzawa H (1962), “Production Functions with Constant Elasticities of Substitution”,
Review of Economic Studies, Vol. 29, No. 4, pp. 291-299.
125. Velupillai K (1973), “The Cobb-Douglas or the Wicksell Function? A Comment”,
Economy and History, Vol. 16, No. 1, pp. 111-113.
126. Vinod H D (1968), “Econometrics of Joint Production”, Econometrica, Vol. 36, No. 2,
pp. 322-336.
127. Vinod H D (1969), “Econometrics of Joint Production: A Reply”, Econometrica,
Vol. 37, No. 4, pp. 739-740.
128. Vinod H D (1976), “Canonical Ridge and Econometrics of Joint Production”, Journal
of Econometrics, Vol. 4, No. 2, pp. 147-166.
129. von Neumann J (1937), “Über ein ökonomisches Gleichungssystem und eine Verallgemeinerung
des Brouwerschen Fixpunktsatzes”, in Ergebnisse eines Mathematischen Kolloquiums, Vol. 8,
1935-1936 (Franz-Deuticke), pp. 73-83, Translated: A Model of General Economic
Equilibrium, Review of Economic Studies, Vol. 13, No. 1 (1945-1946), pp. 1-9.
130. von Stackelberg H (1932), Grundlagen einer reinen Kostentheorie (Foundations of a Pure
Theory of Costs), Verlag von Julius Springer, Wien.
The IUP Journal of Managerial Economics, Vol. VIII, No. 4, 2010
34
131. Weaver R D (1983), “Multiple Input, Multiple Output Production Choices and
Technology in the US Wheat Region”, American Journal of Agricultural Economics,
Vol. 65, No. 1, pp. 45-56.
132. Whitaker J K (Ed.) (1975), The Early Economic Writings of Alfred Marshall, 1867-1890,
Vol. 2, The Free Press, New York.
133. Wibe S (1984), “Engineering Production Functions: A Survey”, Economica, New Series,
Vol. 51, No. 204, pp. 401-411.
134. Wicksell K (1923), “Real Capital and Interest”, Review of Gustaf Akerman’s Realkapital
und Kapitalzins, in L Robbins (1934) (Ed.), Lectures on Political Economy,
Vol. 1, General Theory, Appendix 2, Routledge & Kegan Paul, London.
135. Wicksteed P H (1894), An Essay on the Co-ordination of the Laws of Distribution, Macmillan
& Co., London, Available at http://cepa.newschool.edu/het/texts/wicksteed/
wickess.pdf
136. Wikipedia-a, “Production Function”, Available at http://en.wikipedia.org/wiki/
Production_function
137. Wikipedia-b, “Koopmans’ Theorem”, Available at http://en.wikipedia.org/wiki/
Koopmans’_theorem
138. Zellner A and Revankar N S (1969), “Generalized Production Functions”, The Review
of Economic Studies, Vol. 36, No. 2, pp. 241-250.
Reference # 21J-2010-11-01-01
Reproducedwith permission of thecopyright owner. Further reproductionprohibited without permission.

inputs that cannot be aggregated in physical units. It also produces several types of output (joint production) measured in different physical units. There is an extreme view that (in a sense) all production processes produce multiple outputs (Faber et al., 1998). One of the ways to deal with the multiple output case is to aggregate different products by assigning price weights to them. In so doing, one abstracts away from essential and inherent aspects of physical production processes, including error, entropy or waste. Moreover, production functions do not ordinarily model the business processes, thereby ignoring the role of management, of sunk cost investments and the relation of fixed overhead to variable costs (Wikipedia-a). It has been noted that although the notion of production function generally assumes that technical efficiency has been achieved, this is not true in reality. Some economists and operations research workers (Farrell, 1957; Charnes et al., 1978; Banker et al., 1984; Lovell and Schmidt, 1988; Seiford and Thrall, 1990; and Emrouznejad, 2001) addressed this problem by what is known as the Data Envelopment Analysis (DEA). The advantages of DEA are: here one need not specify a mathematical form for the production function explicitly; it is capable of handling multiple inputs and outputs and being used with any input/output measurement; and efficiency at technical/managerial level is not presumed. It has been found useful for investigating into the hidden relationships and causes of inefficiency. Technically, it uses linear programming as a method of analysis. We do not intend to pursue this approach here. Starting in the early 1950s until the late 1970s, production function attracted many economists. During the said period, a number of specifications or algebraic forms relating inputs to output were proposed, thoroughly analyzed and used for deriving various conclusions. Especially after the end of the ‘capital controversy’, search for new specification of production functions slowed down considerably. Our objective in this paper is to briefly describe that line of development. In the schema of Frisch (1965), we shall first concentrate on ‘single-ware’ or single-output production function. Then we shall move to ‘multi-ware’ or multi-output production function. Finally, we shall address the pros and cons of the aggregate production function.

Single-Output Production Function
Humphrey (1997) gives an outline of historical development of the concept and mathematical formulation of production functions before the enunciation of Cobb-Douglas function in 1928. Paul Douglas, on a sabbatical at Amherst, asked mathematics professor Charles W Cobb to suggest an equation describing the relationship among the time series on manufacturing output, labor input, and capital input that Douglas had assembled for the period 1889-1922, and this led to their joint paper. An implicit formulation of production functions dates back to Turgot. In his 1767 Observations on a Paper by Saint-P´eravy, Turgot discusses how variations in factor proportions affect marginal productivities (Schumpeter, 1954). Malthus introduced the logarithmic production function (Stigler, 1952) and Barkai (1959) demonstrated how Ricardo’s quadratic
A Brief History of Production Functions 7

production function was implicit in his tables. Ricardo used it to predict the trend of rent’s distributive share as the economy approaches the stationary state (Blaug, 1985). Johann von Thünen was perhaps the first economist who implicitly formulated the exponential production function as P  f ( F )  A

 (1  e
3 1

 ai Fi

) , where F1, F2 and F3 are the

three inputs, labor, capital and fertilizer, ai are the parameters and P is the agricultural production. Lloyd (1969) provides a complete account of von Thünen’s exponential production functions and their derivation. He was also the first economist to apply the differential calculus to productivity theory and perhaps the first to use calculus to solve economic optimization problems and interpret marginal productivities essentially as partial derivatives of the production function (Blaug, 1985). Mitscherlich (1909) and Spillman and Lang (1924) rediscovered von Thünen’s exponential production function. In The Isolated State, Vol. II, von Thünen wrote down the first algebraic production function as: p = hqn, where p is output per worker (Q/L), q is capital per worker (C/L) and h is the parameter that represents fertility of soil and efficiency of labor. The exponent n is another parameter that lies between zero and unity. Multiplying both sides of von Thünen’s function by L (labor), we have Lp  hq n L  hC n L1n = P = Output. Thus, we have the Cobb-Douglas production function hidden in von Thünen’s production function (Lloyd, 1969). The credit for presenting the first Cobb-Douglas function, albeit in disguised or indirect form, must go to von Thünen in the late 1840s rather than to Douglas and Cobb in 1928 (Humphrey, 1997). Further, von Thünen was uneasy to realize the implication of his production function in which labor alone cannot produce anything. Hence, he modified his function to P  h(L  C )n Ln1 . This equation, which von Thünen estimated empirically for his own agricultural estate and which he declares he discovered only after more than 20 years of fruitless search, states that labor produces something even when unequipped with capital (Humphrey, 1997). It is surprising, however, that modern economists never formulate a production function in which labor alone can produce something. Velupillai (1973) points out how Wicksell formulated his production function in 1900-01 that is identical to Cobb-Douglas function. In his 1923 review of Gustaf Akerman’s doctoral dissertation Realkapital und Kapitalzins, Wicksell (1923) wrote his function as P  cL C  with the exponents adding up to unity. Works of Turgot, von Thünen, and Wicksell might not have been known to Paul Douglas, but it is surprising to know that before his collaboration with Cobb, Sidney Wilcox, a research assistant of Douglas, had formulated in 1926 a production function of which the Cobb-Douglas function is only a special case (Samuelson, 1979). Wilcox’s production function was, perhaps, ignored by Douglas and till date it has remained in obscurity. Likewise, what is now well-known as the Leontief production function was formulated by Jevons, Menger and Leon Walras. In the Walrasian model of general equilibrium, the proportions of output to inputs are fixed and no substitution among inputs is entertained.
8 The IUP Journal of Managerial Economics, Vol. VIII, No. 4, 2010

these functions allow for variable elasticity of substitution. 1% for 1%. the region in which the marginal products are nonnegative and diminishing marginal rate of substitution holds. So. every neoclassical production function. and the second is in defining the said elasticity when more than two inputs are used in production. 521). 0 < < 1 the marginal product of labor is nonnegative only if  /  K / L. Of course. jk) and for more inputs there would be many more. Uzawa (1962) and McFadden (1962 and 1963) proved that it is impossible to obtain a functional form for a production function that has an arbitrary set of constant elasticities of substitution. The transcendental function also does not contain a linear function. ik. two generalizations of the Cobb-Douglas production function appeared in the literature before 1961. In the Cobb-Douglas production function the elasticity of substitution of capital for labor is fixed to unity. The production function formulated by Arrow et al. but to stay fixed at that number along and across the isoquants. must contain the linear production function P  aK  bL to be consistent with it (Revankar.g. L) plane. However. 0   . Newton and many others (Russell.   1. p. The transcendental function is a neoclassical production function if a. irrespective of the size of output or inputs (capital and labor) used in the production process. Mathematical enunciations of these assertions are now known as the impossibility theorems of Uzawa and McFadden. This function is well-known as the Constant Elasticity of Substitution (CES) production function. Moreover. For three inputs there would be three elasticities (say. b < 0 and L    / b and K   / a (Revankar. (1961) permitted it to lie between zero and infinity.. and yet they are known to the posterity by the name of those scholars in whose work they found their final expression. is only an extension. the Cobb-Douglas or its generalizations. which. 1971). The first is in the constancy of the elasticity of substitution (between inputs) along and across the isoquants. The linear production function is important in view of the Harrod-Domar fixed coefficient model of an expanding economy and therefore. ij. The first of these functions is a neoclassical production function only in a restricted region in the non-negative orthant of the (K. 1971). the Leontief and the linear production functions as its special cases. When > 0. It encompasses the Cobb-Douglas. not an alternative paradigm. (1961). Such achievements are superb in exposition though containing only a little of originality. e. it is also true of Cobb-Douglas. This is true of Aristotle. (1957) specified as P  Ae aK bL [K  L ].Bertrand Russell tells us that certain academic achievements of individual scholars are a culmination of the research efforts of an entire epoch. it does not contain a linear function as its special case. however. Two mutually interrelated difficulties with the CES production function came to light very soon. if the number of inputs (factors of production) is greater than two. A Brief History of Production Functions 9 . 1984. One of them is described as P  Ae (K / L ) [K  L1  ] and the transcendental production function of Halter et al. A notable change in their formulation of production function came only in 1961—after a gap of 33 years—with the work of Arrow et al. These achievements set forth a paradigm and thus put a halt on the further progress of science for quite a long period.

In 1968.. Assuming that competition has led to minimal cost conditions in production.. In his doctoral dissertation. VIII. where Sk is the share of capital. As the output-labor ratio increases (e. Zellner and Revankar (1969) found a procedure to generalize any given (neoclassical) production function with specified constant or variable elasticities of substitution such that the resulting production function retains its specification as to the elasticities of substitution all along. PeP  c h f h where f is the basic function (e. No. L) orthant only if m  0 (and 0 < < 1). Revankar (1967) expounded his generalized production functions that permit variability to returns to scale as well as elasticity of substitution.g. CES. It reduces to CES for m = 0. This feature of the CMS function puts some limitations on it. which implies that productivity of labor increases with capital-labor ratio at a decreasing rate. Liu and Hildebrand (1965) formulated a variable elasticity of substitution production function as P  A (1   )K   K m L(1m )   1/ . 2010 . Bruno formulated his Constant Marginal Share (CMS) production function P  AK  L1  mL or P / L  A(K / L)  m. The CMS production function contains a linear production function and defines the elasticity of substitution as. Lu and Fletcher (1968) generalized the CES production function to permit variable elasticity of substitution. Their Generalized Production Function (GPF) is given as. In contrast with the production functions that (rather unrealistically) assume the same returns to scale at all levels of output. with economic growth). The CMS function has nonnegative marginal productivity of labor over the entire (K. That is: to find a functional form of (a neoclassical) production function that would permit variable elasticities of substitution (among different inputs) along and across the isoquants as well as larger number of inputs to be included in the recipe. 4. Mukerji (1963) generalized CES for constant ratios of elasticities of substitution.It is rather predictable what economists would do afterwards. Vol.   1  m /(1   )(L / P).   (1   ) 1 [1  c{1  wL /(rK )}] where wL and rK are the shares of labor and capital in output (net value added) respectively.g. But then the elasticity of substitution will never be less than unity. Cobb-Douglas. Their Variable Elasticity of Substitution (VES) function is specified as P  A K   (1   ) ( K / L) c (1  ) L   1/  . etc.) as the object of 10 The IUP Journal of Managerial Economics. For this function. the elasticity of substitution is given as   (1    m / Sk )1 . the elasticity of substitution in this function tends to unity and thus the CMS tends to the Cobb-Douglas production function. It may be noted that the assumption of minimal cost conditions and use of factor shares in defining the elasticity of substitution limit the importance of this function in empirical investigation. the elasticity of substitution is obtained as. but permits returns to scale to vary with the scale of output. Sato and Hoffman ((1968) also introduced their variable elasticity of substitution production function. Bruno (1962) suggested a generalization of CES production function to permit the elasticity of substitution to vary.

the returns to scale first increases with output. It requires two types of labor (skilled and unskilled) to be separately dealt with in specifying the production function. the two-level CES production technology with capital. if the Cobb-Douglas production function is generalized. Revankar’s VES does not contain the Leontief production function (while the CES does contain it). Further. Revankar (1971) presented his VES production function as P  AK  (1 ) [L  (   1)K ]  with restriction on parameters: A. It has to be estimated so as to maximize the likelihood function since the least squares and maximum likelihood estimators of parameters do not coincide (see Mishra. Sato (1967) generalized the CES production function so as to incorporate more than two inputs in the specification. x 2 . as we know. On the side of generalizing the CES production function to more than two factors of production. Brown and Cani (1963) generalized the CES production to allow for non-constant returns to scale. Now combine x1 and x2 in a manner of CES to obtain Z1 and similarly. At the second level. x 4 ) where P is output and xi is an input. To specify such models. Uzawa (1962) made fundamental contributions. and .the returns-to-scale function monotonically increases or decreases with increase in P. skilled labor and unskilled labor as inputs may be more suitable. h relate to parameters associated with the returns-to-scale function. The elasticity of substitution function is given as  (K . remains more or less constant in a domain and then begins to fall. that contains the Cobb-Douglas production function for c = 0. but it contains Harrod-Domar fixed coefficient model. 12) and between x3 and x4 (say. Their production function is given as P (1 c ln P )  AL K  . The returns to scale function is given by  (P )   /(1  P ).   0 . or the wage differential between skilled and unskilled workers. Nerlove (1963) generalized the Cobb-Douglas production function to allow for variable returns to scale. x 3 . 2007a). Obviously. This function is interesting from the viewpoint of estimation also. combine x3 and x4 to obtain Z2. In particular. A Brief History of Production Functions 11 . This fact is not captured by the ZellnerRevankar function since it gives us a linear returns-to-scale function. L )  1  [(   1)/(1   )]K / L  1  K / L. The nesting schedule of inputs would depend on the nature of inputs and production technology. 34) at the first level and between Z1 and Z2 (say. This schema would provide (constant) elasticities of substitution between x1 and x2 (say. the linear production function and the Cobb-Douglas function. c  0 . c is the constant of integration. Ringstad (1967) advanced on the same line. There are ample empirical evidences that suggest capital-skill complementarity (Griliches. Depending on the sign of .generalization. 0    1. the elasticity of substitution of capital for skilled labor is very little—they are complementary to each other. To illustrate Sato’s schema. let P  f ( x1 . 0    1 and L / K  (1   )/(1   ). while the case of unskilled labor is entirely different. s12) at the second level. However. 1969). combine Z1 and Z2 to obtain P. we have Pe P  AK  L (1 ) .

He generalized the Leontief production function to attain an arbitrary set of Allen-Uzawa elasticities or shadow elasticities of substitution.Diewert (1971) made two very important generalizations of production functions. 12 The IUP Journal of Managerial Economics. bij  b ji Unfortunately. aij  a ji  0  j 1  n where h is a continuous. Kadiyala showed that for 12 = 0 his function is CES. ln(xj). such a functional form permitted variable elasticities of substitution. For n inputs (xi). which also absorbs the neutral technical progress. monotonically increasing function that tends to + and has h(0) = 0. Further. Griliches and Ringstad (1971). on account of inclusion of ln(xi). 2010 . E(t) is the efficiency parameter. the translog production function is specified as: In( P )  a0   n a i 1 i ln( x i )  0. 11  212   22  1. The function is defined as: P  E(t )[11 L( 1   2 )  212 L1 K  2   22 K (  1   2 ) ]  /( 1   2 ) . He also showed that the function may be generalized for more than two inputs. Revankar and Sato-Hoffman production functions as its special cases. He also obtained the generalized linear production function that can attain an arbitrary set of direct elasticities of substitution at a given set of inputs and input prices.  = 0) could be used to estimate the CES production function. he obtained a functional form that can incorporate many inputs in the specification and second. their product and their squared values. (1973) introduced their translog production functions that permit more than two inputs as well as variable elasticities of substitution (which is a necessity in view of Uzawa-McFadden theorems). No.  ij  0 In the function above. Restricting the sum of coefficients to unity. and for 11 = 0 it is Sato-Hoffman production function. At this juncture it would be appropriate to point out that in 1967 Kmenta had proposed a production function whose second-order Taylor’s series approximation (around substitution parameter. VIII. Lu-Fletcher. Further. CES. 1978). First. this function provides convexity to the isoquants and implies a well-behaved cost function.5   n i 1 n b j 1 ij ln ( x i ) ln ( x j ). Diewert’s generalized linear production function is:  n P  h   i 1   12 12 aij x i x j . 1 and 2 bear the same sign as (1 + 2). but to take care of the Allen or Uzawa-McFadden partial elasticities of substitution one has to involve all the input ratios and the functional form may thus be quite lengthy and complicated. Vol. 4. Berndt and Christensen (1973) and Christensen et al. In spite of its generality. estimation of parameters of this function often suffers from multicollinearity problem. Kadiyala’s production function has remained in obscurity. for 22 = 0 it is Lu-Fletcher. Kadiyala (1972) proposed a production function that includes Cobb-Douglas. this function is not invariant to the units of measurement of inputs and output (Intriligator.

one can identify a variety of its special cases according to the values taken on by the estimated parameters. Thus. if one uses Kmenta’s approximation to estimate the parameters of a production function. b  0 to obtain: P  A[a1K b  (1  a1 )Lb  b K c Lb c ]  b and showed that this specification. In most cases. 0 < a1 < 1 and defining  = a3 / b. enough attention has not been given to this issue. P  A[ a1K b  a 2 Lb  a 3 K c Lb c ]  b . b  0 whose second-order Taylor’s series approximation would be econometrically indistinguishable from Kmenta’s approximation. Kmenta obtained: ln(P)  ln( A)   ln(K )   (1   ) ln(L)  0.P  A[K    (1   )L  ]   . a j  0 . It appears that if one estimates the MFM function. McCarthy’s work is ignored and unmentioned. in spite of being an excellent Figure 1: Family Tree of Linearly Homogenous Production Functions MFM Production Function lim b0 Transcendental Second-Order Approx. However. one is not sure whether those parameters would really pertain to the CES or the McCarthy’s production function. A. which we would call the McCarthy-Färe-Mitchell (MFM) production function. at least in the textbooks.5  (1   )[ln(K )  ln(L)]2 McCarthy (1967) pointed out that there could exist another production function. Färe and Mitchell (1989) modified McCarthy’s production function by constraining a1 + a2 = 1. However. around c = 0 Translog CES lim b0 c = b/2 Quadratic Mean of Order Alpha b=2 Generalized Square Root Quadratic c=0 or = 0 = 0 b=1 Generalized Leontief = 0 b=1 = 0 or c = 0 c’ = 0 Cobb-Douglas Linear lim b0 A Brief History of Production Functions 13 . has eight production functions as its special cases (see Figure 1). A note on estimation of MFM production function is deemed necessary.

respectively. however. By the middle of the 1970s. b  0) and it appears as a factor of one of the coefficients. A technological change is said to be Harrod-neutral. the technological progress is Hicks-neutral. these functions assume that the marginal rate of substitution between any two factors of production is associated only with relative factor prices and it is independent of technical progress or level of output or. x 2 ). A technological change is Hicks-neutral (Hicks. discussed so far). The production function P = f(x) is modified as. of which the classical CES (as well as the Cobb-Douglas) and the ‘non-homothetic Cobb-Douglas’ production functions are only special cases. Empirical data in many cases. 1932). Vol. econometric estimation of MFM from empirical data is problematic due to the fact that some parameters (especially b and c) tend to be zero for some special cases. An introduction of factor-augmenting technical progress (of Harrod or Solow) fails to perform due to impossibility of identification of the bias (of technical progress) and substitution effect (Sato. and it is Solow-neutral. etc. In the classical form. Sato (1975) relaxed the homotheticity assumption so that the level of output and the degree of neutral technical progress explicitly affect the factor combinations or ln (w / r )  ln ( a)  (1/ ) ln (K / L)  b ln (P )  c ln (T (t)). P = e t f(x) where e t captures the technological change that does not modify the elasticity of substitution between the factors of production. It is very easy to incorporate Hicks-neutral technical change into (a classical) production function. Sato (1975) observed that so far the marginal rate of substitution function had been specified as. 14 The IUP Journal of Managerial Economics. No. in technical terms. if the technology is labor-augmenting. if the technology is capitalaugmenting.. Sato’s generalization permitted decomposition of income and substitution effects (in the factor market) and distinction between normal and ‘inferior’ inputs. generalization of Cobb-Douglas and CES production functions (in their classical form) was almost complete.generalization of numerous production functions. From this specification he obtained the ‘most general’ class of CES function. ln (w / r )  ln ( a)  (1/ ) ln (K / L) where w and r are the prices of labor (L) and capital (K). where P is the production and T(t) is the time dependent index of biased technical progress. In view of the homotheticity assumption (assertion that a function g is a continuous positive monotonically increasing function of a homogenous function f. To explain this concept a little more. 1970). we note that a technological change describes a change in the set of feasible production possibilities. z  g( y )  g( f ( x))  g( f ( x1 . 2010 . if the ratio of capital’s marginal product to labor’s marginal product is unchanged for a given capital-labor ratio. Therefore. and  is the elasticity of substitution. VIII. suggested that the factor price ratio varies even at a constant input ratio. 4. such that if y  f ( x )  f ( x1 . The parameter  is a ratio (  a 3 / b . x 2 )) then dg / dx  0) implicit in the (classical versions of) production functions (Cobb-Douglas or CES. the marginal rate of substitution was considered to be independent of the level of production or neutral technical change. These peculiarities may have destabilizing effect on the algorithm that estimates them.

1974). and C x ( f ). the constants  i  (1/    i )   i and f  V (X 1 / X 2 ). where m is a constant.  1  0 . i . Examples of non-separable CES are: if C(f) = af and H(f)=bf 2. as output varies. where X i   i x i    i (for   1) or X i   i ln( x i )   i (for  1). we may permit variability of the elasticity of substitution across the isoquants by defining = (f ) = constant at any f. dC i / df  0.   1. we have non-separable CES functions as the nonlinear solutions of F(X 1 .Sato’s CES function: F(X1 ..  ( f )  0 . The elasticity of substitution is constant along an isoquant. but it varies across the isoquants. It allows for a case when an isoquant in the (x1. i  j . if we define for any two inputs i and j the ratio of factor prices as ij and the elasticity of substitution between them as   1 x1    1 x1 (h( f ))  . This grand generalization of the CES (and Cobb-Douglas) functions possibly concluded an era of investigations on this topic. x i are    1 x 2    1 x 2 (h( f ))    ij     ln( xi / x j )/  ln(ij ).. n . X 2 . C ij  0. Finally. Since all homothetic functions are separable. xi is a factor of production. X 2 .  2  0. dH / df  0 and C1  mC 2 . f )   n i 1 Ci ( f )X i  H( f )  0 where X i   i x i    i or X i   i ln( xi )   i for a non-unitary or unitary  respectively. In this case. In this case too. Sato showed that his generalized CES might easily be extended to n-inputs case as well as to variable elasticity of substitution. we obtain ordinary CES and Cobb-Douglas functions depending on whether  is non-unitary or unitary.   (1   )/ for a non-unitary (and non-zero) elasticity of substitution (). then f is given by [ aX 2  {(aX 2 ) 2  4bX1 }1 2 ]/(2b)  0 . f )  X1  C( f )X 2  H ( f )  0 in terms of f or C(f). If the elasticity of substitution depends on the level of output then the ratio of factor prices (w / r )  ( x1 / x 2 )1/ ( f ) C( f ). A Brief History of Production Functions 15 . if C(f)=af 2 and H(f )=b f.). j  1. C( f )  0 . in general.. when F(X 1 . In the n-inputs case. then we have f [b  {b 2  4 aX1X 2 }]/(2 aX2 )  0. Separable CES functions are linear solutions of F(. dC i / df  0. Second. H( f ) and f are defined appropriately according to homotheticity (or otherwise) and separability. then  ij  ( x i / x j )1  C ij ( f ). when dCi / df  0 and dH / df  0.   (1   )/ . First. For n-inputs case..2. x2) plane may be a Cobb-Douglas or (an ordinary) CES but another isoquant can be a non-homothetic CES. where  is the nonhomogeneity parameter. We then have F(X .  and are appropriate constants. we have separable non-homothetic CES function or Cobb-Douglas function depending on whether  is non-unitary or unitary (Sato. dH / df  0 and C1  mC 2 where m is a constant. in which case   (1   ( f )) / ( f ). f )  X 1  C( f )X 2  0. Sato obtained a threefold classification of CES. Note that. f )  C1 ( f )X1  C 2 ( f )X 2  H ( f )  0. the function C(f ) is: C( f )  initial values of xi. X 2 .

well-known for versatility and competence in economics. Lindberger (2003) extended the above LINEX function to ‘service production functions’ that may be defined as P  P0 L(E / L) a0 cm exp[a0 {2  (L  E)/ K}] or alternatively.   a0 {cm (L / E)  (LE / K 2 )}. In his paper. One may estimate the function by any suitable algorithm (Mishra. saying: “Economics is not dismal but incomplete. 2010 . He is well-known for his theorem in molecular orbital theory in quantum chemistry.The importance of energy in the economic system was well-stressed by GeorgescuRoegen (1971). the first economist to demonstrate how engineering information could be used to improve the empirical studies of production and bridge the gap between theoretical and empirical analysis of production (Wibe. They assume everything can be done. the function is specified as: P  P0 E exp[a0 {2  (L  E)/ K }  a0 c 0 (L / E  1)]. 1984). It may be noted that the economists with technology or physics background have almost always pleaded for incorporation of technological considerations into production function. while the economists of the other category have often limited themselves to mere economic considerations. Kümmel (1982) and Kümmel et al. and yields (non-negative) production elasticities. (1985 and 1998/2000) introduced the linear-exponential (LINEX) production function that was based on physics and technology as much as on economic considerations. We return to the LINEX function (Lindenberger and Kümmel.   1     . 1974. 16 The IUP Journal of Managerial Economics. perhaps. The constant P0 is a technology parameter indicating changes in the monetary valuation of the original basket of goods and services making up the output unit. L0 and P0 (and thus all of them are indices with a fixed base). Chenery (1949 and 1950) was. With these variables and parameters. that emphasizes labor- dependence of service production. 2006). which depends linearly on energy (E) and exponentially on quotients of capital (K). All inputs and output are measured relative to some fixed quantity. 2002). it is worth noting that Koopmans was a physical scientist (as well as an economist) of high order (Wikipedia-b). K0. P. the energy crisis due to Jom-Kippur and Iraq-Iran wars that threatened the US and many other economies (in the mid 1970s and thereafter) led many economists to formulate energy-dependent as well as other production functions that included energy and materials (besides conventional labor and capital) as inputs (Hudson and Jorgenson. Koopmans (1979) appears to be in agreement with an engineer’s view of economics.” By the way. so we do not repeat them here. 1974.. However. 2   2a0 (L / K )(E / K  1). 2 P  P0 L exp[a0 {3  2(L / K )  (LE / K 2 )}  a0 cm {1  (L / E)}] . No. Vol. c0. 4. 1979). life sciences and philosophy alike. and P0 time-dependent. and Berndt and Wood. E0. Creativity-induced innovations and structural change make a0. Tintner et al. Soren Wibe has presented a detailed survey of engineering production functions. VIII. The things missed are very important” and a physical scientist saying: “Economists are technologically radical. physical sciences. mathematical sciences. labor (L) and energy (E).

u  0 such that P( x)  {0}. for non-negative inputs and outputs (such that P(x) is not zero and L(u) is not null) a joint production exists if and only if the sets S{L(u)} and S{L(v)} are disjoint and the sets S{P(x)} and S{P(y)} are disjoint. 1961). 1984) have also dealt with joint production function. In particular. it has found recognition in the upcoming field of environmental economics (Baumgärtner. Building upon the work of Bol and Moechlin (1975). 1963. Chizmar and Zak. A number of studies have been carried out that deal with this topic. and P(x) and P(y) denote all output vectors obtainable by inputs x and y (respectively). Subsequently. xRy S{L(u)}  S{L(v)}   . More formally. It is worth mentioning that Kurz (1986) presents a very lively account of multiple-output production function as visualized by the classical and the (early) neoclassical economists. L(u)   17 A Brief History of Production Functions . Without enforcing the strong disposability of inputs or outputs it was shown that a joint production function exists if and only if both input and output correspondences are strictly increasing along rays (implies that a technology is both input and output ray homothetic). uRv x . While in case of a single-output technology. Gronau. the input vector and the output vector belonged to the isoquant of input correspondence and the isoquant of output correspondence respectively. 1971. Mundlak and Razin. 1983. 1961. 1977. no simple maximal output exists for a multi-output technology. 1983. 1975. and Graham and Green. The third concept of joint production might relate weakly efficient input vectors to weakly efficient output vectors. The first important question regarding a joint production function is its definition and existence. so that a multi-output production function could be defined and its existence proved readily (Pfouts. Just et al. The second follows Hanoch (1970) in which an efficient joint production function characterizes input and output vectors that are simultaneously input and output efficient. Al-Ayat and Färe (1977) examined necessary and sufficient conditions for the existence of a joint production function within a general framework of production correspondences. The main finding of Färe is: let L(u) and L(v) denote all input vectors yielding output u and v (respectively). As defined by Shephard (1970). Pollak and Wachter. studies in agricultural (or farm) economics have addressed this problem more frequently (Jawetz. Färe (1986) put forth three alternative ways in which multi-output production function might be defined. and Weaver. Now we turn to a case of multiple or joint production. Mundlak. Studies in the economics of household production and allocation of time between work and leisure (Becker. Salvadori and Steedman (1988) review the concept in the Sraffian framework. the production function was defined as the maximal output obtainable from a given input vector (Shephard. 2004). a joint production function exists if the following conditions are satisfied: S{P( x)}  S{P( y )}   . 1970). 1983). it is an isoquant joint production function in the manner that for a given pair of input and output vectors. then.. 1965.Multiple-Output Production Function In the preceding narrative. Of late. we have dealt with the case when a producing agent produces a single product.

x 1 and x 2 are replaced by r and  with r 2  x1  x 2 and   tan 1 ( x 2 / x1 ).. Baumgärtner (2001) points out the contributions of von Stackelberg (1932) to the theory of joint production (and joint cost) function. a point r i in m-dimensional (real) space is defined as 2 ri  ( pi21  pi22  . x ik ).. x i 2 ... …. 1986.. It is also assumed that all the products are measured in the same (Euclidean) space. by polar coordinates.. Econometric analysis of joint production perhaps dates back to the works of von Stackelberg (1932) and Klein (1947).... ri  f ( i 2 . xi 1 . x2. A production vector x is then uniquely determined by its length r and its direction . 2.  pim )0. They are obtained by replacing the Cartesian coordinates x1. Implicitly it is assumed that all the products are identically related to the inputs. In the latter case one may obtain the ‘curve of the most favorable directions’ in output space. xm). Vol. estimation of joint production function is rather simple (Barrett and Hogset.... 4.. pi 2 . those formulating simultaneous equations systems... Von Stackelberg conceptualized joint production by introducing two new theoretical tools: the notions of length (r) and direction () of the production vector x = (x1. for which von Stackelberg 2 2 defines these terms. pi1) is considered as a standard measure or reference. Since then a number of studies have been conducted to estimate the parameters of multi-output or joint production functions. This curve is defined as the locus of all those production programs (x1. Färe also proved that under certain conditions all the three definitions of joint production function are equivalent. and hence i1 is not included among the regressors. etc. Methodologically those studies may be classified under four heads: those formulating process analysis models. VIII. Weak disposability of inputs and outputs is assumed. when  is a constant. His method lies in specifying the individual micro production function for each (joint) product as well as the manner of aggregating them to an analogous macro production function. Then. Some important works are briefly reviewed as follows. n be n observations on m joint products. pim ). > and  respectively (see Färe. x2. Note that the first output is used as a standard..  i 3 . Let Pi  ( pi1 . Only the positive value is taken (since ri is a length).. and the second in which  is variable. Mundlak (1963) approached estimation of joint production function through a different sort of aggregation.. This is used 1 as the dependent variable. The direction vectors are obtained as  ij  cos ( p ij / ri ). These direction vectors together with the inputs (x) are used as regressors.. In the von Stackelberg schema. p. and those formulating composite implicit macro function.5 .  im .. i  1.* The relationship R is x (  1).. i  1. two cases emerge: the first. From this.. n are used for estimation of the parameters of production function. 2010 . 2003). those formulating composite macro function. x2). No. One of the outputs (say. in which case the production function behaves like a single-output production function. Thus.. The macro production function is then estimated and its relationship 18 The IUP Journal of Managerial Economics.2 . 673). In the case of just two outputs. which yield a given revenue at the lowest possible total costs.

but inputs are considered ˆ non-stochastic. 1969. where the weights. However. and Rao. Later. the method proposed by Vinod did not yield composite output function (transformation function) that satisfied these requirements. U = Xw and V = Y respectively. He showed that the back transformation of the composite vectors U and V into X and Y poses no problem when the number of inputs is equal to or larger than the number of output. and then transforming U and V back into X and Y respectively. 1969. 1957) of the composite functions f (X )  a0 x11 x 22 exp(b1 x1  b2 x 2 ). c c f (Y)  y11 y 22 exp(d1 y1  d2 y 2 ) and the simple implicit function g(X. His method summarily lies. 1976). then so would be the canonical correlation method. There were strong reactions to Vinod’s method of estimation of joint production functions (Chetty. and Kendall and Stuart. The least squares estimation of such functions has remained problematic.with the micro production functions is investigated. 1969). first. However. then the implicit function g(f(X) – (Y)) = 0 is expressed in terms of the composite input function f(X) and the composite output function (Y). Y) = f(X) –  (Y) = 0. This led to his further work (Mundlak. Mundlak also proposed formulation and estimation of a general implicit production function. Rao pointed out that to be economically meaningful. While acceding to the errors pointed out by the critics. However. when that is not the case. that generally output is considered to contain errors due to specification of f(X) such that any output vector yk = f(X) + uk. It may be noted. in transforming the input vectors (X) and the output vectors (Y) into two composite (weighted linear aggregate) vectors. he improved his method to take care of the estimation problem if the data on output (of different products) or inputs were collinear (Vinod. Mundlak illustrated his approach by the transcendental specification (proposed by a a Halter et al. Vinod (1968) addressed the problem of estimation of joint production function by Hotelling’s canonical correlation analysis (Hotelling. w and . Dhrymes and Mitchell.. the possibilities of establishing the relationship among the macro and micro production functions depend on availability of information on allocation of inputs used for different (joint) products. Mundlak and Razin (1971) also basically attempted aggregation of micro functions to macro function. This consideration would lead to the specification g( f (X )   (Y ))   where  is the disturbance term. 1936. one has to resort to some sort of least squares estimation (resulting from his suggested use of the least squares generalized inverse in the transformation process). are (mathematically derived) so as to maximize the squared (simple product moment) coefficient of correlation between U and V. 1964) in which he formulated the problem of estimation of multiple/joint production functions as an exercise in estimation of an implicit function. the production function must be convex and the transformation curve concave. however. in his reply A Brief History of Production Functions 19 . Dhrymes and Mitchell (much like Chetty) pointed out that Vinod’s formulation was partly erroneous and partly a “very complicated way of performing ordinary least squares.” If the ordinary least squares method applied to estimate each production function separately and independently (ignoring the fact that they relate to joint products) was inconsistent. If X are inputs and Y is output. 1968).

Thirdly. Further. As long as a production function describes a relation that relates output and inputs of a firm. sector or economy-level. However. when one makes an attempt to identify such a relation at the industry. returns to scale and market 20 The IUP Journal of Managerial Economics. they held that in case of joint products the implicitform single-equation modeling would be appropriate. However. Secondly. Griffin (1977) used a method similar to process analysis supplementing it with pseudo data to ascertain appropriate types of production frontier functions for different joint products of petroleum refinery. VIII. the efficiency of pseudo data approach to estimation of joint production functions ultimately rests on the quality of the engineering process model often difficult for an economist to build or evaluate. the shape of the production possibility frontier may be determined. A pseudo data point shows the optimal input and output quantities corresponding to a vector of input and output prices. it requires a large database and solving large programming models. However. 1967) and yet another follows the Diewert (1971) or any other specification. although without making any composite output function (Mishra. as pointed out by Griffin himself. 2010 . (1983) formulated and estimated their multicrop production functions as a system of nonlinear simultaneous equation model.Vinod (1969) disagreed on the inconsistency issue shown to exist in his method and argued that the critics (Dhrymes and Mitchell) had to establish the necessity and would not merely put up some particular cases thereof. it may not be correct to form the composite output function in Vinod’s manner. 2007b and 2007c). the matters are more or less straight. Chizmar and Zak (1983) discussed the appropriateness of simultaneous equation modeling of multiple products raised or manufactured simultaneously. It is possible that while one of the products follows the CES. Since the early work of Manne (1958). process analysis has amply exhibited its ability to deal with the economics of joint products. it is possible to specify production functions of different types for different products and estimate their parameters jointly. to note that Vinod undermined the role of a single counterexample in demolishing the mathematical property of a method. By repetitive solution of the process model for alternative price vectors. 4. Aggregate Production Functions Now we turn to the most turbulent area of research in the economics of production. another follows the nested CES (Sato. However. it is not necessary that the specification of production functions is identical for all products. However. An industry is generally made up of numerous firms producing similar products in which each firm uses inputs in its own accord with its own cost. Vol. Vinod’s method cannot be useful when production functions are intrinsically nonlinear such that it is not possible to transform them (by some simple procedure such as log-linearization) into linear equations. it precludes the calculation of price and substitution elasticities that may have important policy implications. Even then. It is interesting. the matters may be quite different. No. however. Just et al. The methods of estimation were nonlinear two-stage and three-stage least squares. Apart from the problems pointed out above. this approach does not rule out the possibilities of aggregation bias completely.

foreshadowing the Tinbergen-Solow growth model. and ultimately Harrod’s. just.E. These issues and many others turn out to be more and more significant when one moves higher to macroeconomic levels. Paretooptimality of the competitive economy. A firm might be a price-taker in the factor market. marginal productivity theory of distribution.implications. A is level of technology and F is fertility of soil. The question arises: does this aggregate production function represent the core technological relation between inputs and output of a majority of firms? Or does it represent the production function of a ‘Platonic’ or ‘archetypal’ firm that might not be a real firm. the (neoclassical) economists. C is Capital.F). Solow’s and von Neumann’s paths to expansion are only some major lemmas to prove the said grand proposition. However. but the empirically observed real firms are only the imperfect instances of that archetypal firm? A little further. A society organized on that principle would grow (expand) indefinitely and deliver justice to all the agents. Deliberate keeping of room for adjustments and maneuvers and perpetual X-efficiency may invalidate the assumption of technical efficiency and vitiate a search for allocative efficiency. Suppose industry output expands and contracts through the entry and exit of identical firms. A production function (or a cost curve for that matter) at an industry level is obtained by using the quantities of inputs and output aggregated over the constituent firms. market economy. Knut Wicksell made significant contribution to demonstrate that non-homogeneous production functions for firms are perfectly compatible with a linear homogeneous function for the entire industry. After Marx. product exhaustion theorem.C.A. Rigidities might not permit a real firm to adjust to the changes at the industry level and all input-output relations at the firm level would be suboptimal.) is the best. Since Adam Smith (or even before him) economists of a particular hue have largely been preoccupied with intellectual work to prove that an autonomous management (and organization) of a society by capitalistic principles (characterizing self-interest guided agents operating in an institutional framework of private property. In this aggregate production function. therefore. etc. aggregate production function. a firm might not possess certain characteristics but an industry may possess them. competition. Karl Marx questioned the efficacy of the capitalistic system in not only delivering justice but also guaranteeing indefinite expansion and stability. had to invent strong arguments to defend the legitimacy. accumulation of capital. each operating at the same minimum unit cost. The Walrasian general equilibrium. The result is to trace out a horizontal long-run industry supply curve that looks like it came from a constantA Brief History of Production Functions 21 . Marshall used the time derivative of the variables and therefore it may be regarded as the first neoclassical growth model. but an industry might be a price-maker. stable and most viable among all possible approaches to manage (and organize) a society. Whitaker (1975) brings to the light some unpublished works of Alfred Marshall and points out how Marshall formulated his aggregate production function P = f (L. where L is labor and E its efficiency. efficacy and supremacy of the capitalistic system.

1997). Humphrey (1997) has presented this development so lucidly that it is best to quote him: “Each stage saw production functions applied with increasing sophistication. Third appeared the path-breaking initial statement of the function in symbolic form. the 1950s did not welcome the aggregate production function wholeheartedly. this line of investigation progressed very rapidly. homogeneity and the like—already were well advanced when Cobb and Douglas arrived”. The fifth stage witnessed the flourishing of microeconomic production functions in derivations of the marginal conditions of optimal factor hire. The CobbDouglas production function was found quite amenable to incorporation of technical change introduced into the production system from time to time without altering the basic conclusions on factor shares. the entire product exhausts (as the sum of input elasticities of product sum up to unity). economists investigated into the possibilities of growth without violent fluctuations.returns production function. No. the production function has been a powerful instrument of miseducation. he justified the use of aggregate linear homogeneous functions such as the Cobb-Douglas function (Humphrey. 2010 . 4. In this investigation the aggregate production function proved to be very useful. The fourth stage saw a mathematical production function employed in an aggregate neoclassical growth model. Consequent upon the development of linear programing as a method of optimization. first-order conditions of factor hire. In short. Last came the proof that functions of the type later made famous by CobbDouglas satisfy this very requirement. product exhaustion. VIII. aggregate linear production function of Georgescu-Roegen (1951). Vol. However. and the empirically observed constancy of relative shares of factors of production for long periods is fully explicable. First came the idea of marginal productivity schedules as derivatives of a production function. The Cobb-Douglas production function gave a readily convincing proof that in competitive equilibrium all inputs are paid their marginal product (and hence their respective real price). and careful proofs given by Nikaido (1968) all strengthened the foothold of aggregate production function in economic analysis. . . Activity analysis of Koopmans. Thus. The growth model of von Neumann (1937) deviated from using the Cobb-Douglas production function but retained the practice of aggregation. This finding strengthened the foundations of using aggregate production function at the macroeconomic level. During the post-Great Depression period until the end of the Second World War. Next came numerical marginal schedules whose integrals constitute particular functional forms indispensable in determining factor prices and relative shares. relative shares. macro and micro production functions and their appurtenant concepts—marginal productivity. Sixth came the demonstration that product exhaustion under marginal productivity requires production functions to exhibit constant returns to scale at the point of competitive equilibrium. justified and natural. The student of economic theory 22 The IUP Journal of Managerial Economics. input-output analysis of Leontief. constant returns to scale prevails. separation theorems and generalization of von Neumann’s model by Gale (1956). Robinson (1953) viewed aggregate production function with a remark: “.

and proved almost conclusively that it has no economics in it as its properties stem from mere algebra. it is impossible to conceive of an abstract quantity of capital which is independent of the rates of interest and wages. who was himself a participant in the controversy (from the Robinson’s side). K a quantity of capital and Q a rate of output of commodities. 2005. The controversy exposed all aggregate production functions. this independence is a precondition of constructing an isoquant (or production function).” A controversy began. and Paul Samuelson. 2001. then how that particular function is the correct function describing the production technology of any economy and the relative factor A Brief History of Production Functions 23 . Joan Robinson. may not possibly have been impartial in giving an unbiased account. A series of three papers by Anwar Shaikh (Shaikh. He is instructed to assume all workers alike. the Cobb-Douglas production function in particular.” One fails to understand as to how the validity of laws of thermodynamics and relating outputs to inputs entails validity of the proposition that the aggregate of functions would be the function of aggregates. he is told something about the index-number problem in choosing a unit of output. in the hope that he will forget to ask in what units K is measured. Luigi Pasinetti and Pierangelo Garegnani (among others) argued against the use of aggregate production function. Frank Hahn and Christopher Bliss (among others) argued in favor of using aggregate production function for explaining relative factor shares. A very lively account of it was presented by Harcourt (1969). in which Piero Sraffa. 2005) have exposed the validity of an aggregate production function. However. I shall continue to relate outputs to inputs—i. Robert Solow. he has become a professor. This is begging the question. especially regarding the measurement of capital. to believe in production functions. Unless factors cease to have their rewards to be determined by bidding in quasi-competitive markets. Once Samuelson (1966) wrote: “Until the laws of thermodynamics are repealed. 1974. This controversy (that began in 1953) lasted until the mid-1970s. and to measure L in man-hours of labor. and if there is some particular type of function that has this property. and Felipe and Holz. Fisher and their associates in a number of papers (Felipe and Fisher. K ) where L is a quantity of labor. Felipe and Adams. Felipe. However. but the protagonists of aggregate production function use the slope of the isoquant to determine relative factor prices. The isoquant cannot be constructed and its slope measured unless the prices are known beforehand. and then he is hurried on to the next question. Before he ever does ask.e. I shall adhere to (generalized) neoclassical approximations in explaining their market remunerations. and as Stiglitz (1974) notes.is taught to write Q = f (L. As argued by the first group. 2005. Felipe and McCombie. 1980 and 2005) on his ‘humbug production function’ should have ousted the Cobb-Douglas production function from all serious endeavors in economic analysis. it may be noted that Stiglitz admits that he was a participant from the other side. and so sloppy habits of thought are handed on from one generation to the next.. A much more comprehensive account of the controversy may be found in Cohen and Harcourt (2003).

mainstream economics goes on as if the controversy had never occurred. It is usually asserted that only aggregate neoclassical theory of the textbook variety—and hence macroeconomic theory. It puts in jeopardy the neoclassical theory of capital and the notion of input demand curves.. for example. “We wish to make it clear for the record that the nonreswitching theorem associated with us is definitely false. which can be used to predict all behavior in the sense of the wage and profit rates that would prevail in different long-run equilibria.. The problems of heterogeneous capital goods have also been ignored in the ‘rational expectation revolution’ and in virtually all econometric work”. It also puts in jeopardy the neoclassical theories of output and employment determination. The final outcome of the controversy may be best concluded in the words of Burmeister (2000): “… the damage had been done. ‘declared victory’: Levhari was wrong. however. and Cambridge. We are grateful to Dr. Samuelson was wrong. was based on too extremely restrictive assumptions and therefore was too fragile to hold true in reality. on the other hand. UK. of the Hicksian 24 The IUP Journal of Managerial Economics. which confirms neutral technical change. since they are all deprived of stability. 4.shares. all neoclassical production models would be affected by capital reversing. 2010 . stability proofs require the exclusion of capital reversing…. 1966). No. Macroeconomics textbooks discuss ‘capital’ as if it were a well-defined concept—which it is not. The rational expectation theory was propounded by Lucas (1976). Lavoie (2000) observed: “Capital reversing renders meaningless the neoclassical concepts of input substitution and capital scarcity or labor scarcity. In that sense. both at the economy and industry levels. Nonreswitching theorem was a major plank to support measurability of capital and its aggregation. Levhari and Samuelson published a paper which began. VIII. Solow (1957) made an impressive empirical study to demonstrate how the aggregate production function fits to the US data for 1909-49. It has been pointed out. The consequences for neoclassical analysis are thus quite devastating. as well as Wicksellian monetary theories. according to Harris (1973). the general validity of Heckscher-Ohlin-Samuelson international trade theory …. As a result there are some groups of economists who have abandoned neoclassical economics for their own refinements of classical economics. which. Samuelson (1962) invented the so-called ‘surrogate production function’ which related the total quantity of output per capita to the total quantity of capital per capita. shift in production function.” Gehrke and Lager (2000) observed: “These findings destroy. Vol. MIT was wrong and therefore neoclassical economics was wrong. except in a very special one-capital-good world (or under other unrealistically restrictive conditions). The effects of ‘capital controversy’ beginning with a criticism of using aggregate production function for explaining (and justifying) relative factor shares had a disastrous effect on neoclassical economics. that when neoclassical general equilibrium models are extended to long-run equilibria.” (Levhari and Samuelson. and therefore validates the artifact of aggregate production function as a powerful tool of analysis. In the US. based on aggregate production functions—is affected by capital reversing. Pasinetti. Solow was wrong.

63.edu/special_programs/ AFSNRM/Basis/Documents/Memos/Multi-outputProductionFunctionEstimation forCLASSES. No. 5. Arrow K J. pp. The era of classical economics had ended (sometime in 1870s) with its criticism by Marx and the birth of neoclassicism. and that is its sole utility. Operations Research Center. No. pp. pp. but also for any production function even at the firm level. one might know their view of the utility of production functions. Vol. 103. 225-250. Barkai H (1959). Vol. “Ricardo on Factor Prices and Income Distribution in a Growing Economy”. Barrett C and Hogset H (2003). If this is true then it speaks enough about the position of ‘production function’ in economics. “Some Models for Estimating Technical and Scale Inefficiencies in Data Envelopment Analysis”. The era of neoclassical economics possibly ended with the capital controversy sometime in 1970s. 627-630. Vol. Chenery H B. not only for aggregate production function. 30. Berkeley. K ). Conclusion It is said that once Stanislaw Ulam very earnestly sought for an example of a theory in social sciences that is both true and nontrivial. they stem from the algebraic properties of the function.cornell. Available at http://aem. “Capital Labour Substitution and Economic Efficiency”.neutrality of technical progress concept …. California University. subsequently published in 1979 in Naval Research Logistics Quarterly. and Statistics. Management Science. Measurement of capital was put in jeopardy by the capital controversy. after several years supplied one: the Ricardian theory of comparative advantages. and of the Pigouvian taxation theory applied in environmental economics …” What has remained with the believers in neoclassical aggregate production function is the complaint that neither Joan Robinson nor her associates developed an alternative set of theoretical (as opposed to descriptive) tools that avoid her concerns about the limitations of equilibrium analysis. 3. Economica. Paul Samuelson. Joan Robinson was right to comment that the production function has been a powerful instrument of miseducation. “On the Existence of Joint Production Functions”. No. “Estimating Multiple-Output Production Functions for the CLASSES Model”. 9. Minhas B S and Solow R M (1961). 4. Charnes R F and Cooper W W (1984). 26. No. Review of Econ. Al-Ayat R and Färe R (1977).pdf A Brief History of Production Functions 25 . Banker R D. Possibly. Bibliography 1. pp. Research Report. 26. 1078-1092. The student of economic theory is taught to write Q = f (L. 3. of neoclassical tax incidence theory …. Anwar Shaikh almost conclusively proved that the properties of the Cobb-Douglas production function are devoid of any economic content. If one has interacted with engineers and the managers who are decision makers at certain level. 4. Vol. 2. 240-250.

3. “Technological Change and the Distribution of Income”. “A Theory of the Allocation of Time”. Vol. 4. “Engineering Bases of Economic Analysis”. Bruno M (1962). 14. 4th Edition. 1. Cambridge. No. No. Euro. Ph. 1. Chenery H B (1949). pp. “Estimation of Factor Contribution to Growth Under Structural Disequilibrium”. pp. 10. No. Center for Economic Policy Analysis (CEPA). No. 1. J. European Journal of Operational Research. Critical Essays on Piero Sraffa’s Legacy in Economics. Berndt E and Christensen L (1973). 731. pp. Vol. in H D Kurz (Ed. Chetty V K (1969). American Economic Review. “Econometrics of Joint Production: A Comment”. Vol. 17. pp. 391-398. Baumgärtner S (2004). the Wage Rate and the Capital-Labour Ratio”. Baumgärtner S (2001). 22. Conservation and Recycling. 1929-1968”. pp. International Economic Review. Becker G S (1965). “Engineering and Econometric Interpretations of Energy-Capital Complementarity”. 4. and Costly Disposal”. No. 493-517. 11. Cambridge University Press. Charnes A. “Measuring the Efficiency of Decisionmaking Units”. pp. “The Capital Theory Controversy”. Berndt E R and Wood D O (1979). 63. “Heinrich von Stackelberg on Joint Production”. 8. No. Unpublished Mimeo. 507-531. 75. Vol. Naval Research Logistics Quarterly. No. pp. Vol. The Economic Journal. 342-354.htm 18.). 19. No. Chenery H B (1950). Cambridge. 1. 9. No. 15. 43. 4. “Price Ambivalence of Secondary Resources: Joint Production. 429-444.edu/het/essays/product/prodfunc. Harvard University. Vol. No. pp. Vol. 69. Cambridge University Press. 81-114. No. Structures and Labor in US Manufacturing. Vol. History of Economic Thought. VIII. 6. Burmeister E (2000). 95-117. 3. pp. 299. 26 The IUP Journal of Managerial Economics. Journal of Econometrics. 7. Brown M and Cani J S de (1963). Economic Theory in Retrospect. “Isoquants of Continuous Production Correspondences”. “The Translog Function and the Substitution of Equipment. Available at http://cepa. International Economic Review. “A Note on the Implications of an Empirical Relationship Between Output Per Unit of Labour. 4. 289-309.newschool. “Engineering Production Functions”. Cooper W W and Rhodes E (1978). Stanford University. 4. Dissertation (Supervisor W W Leontief). 37. pp. Blaug M (1985).6. 2010 . Harvard. 13. 3. 20. 8. Bol G and Moechlin O (1975). 16. Vol. 2. Quarterly Journal of Economics. Bruno M (1968). “The Production Function”. Limits to Substitution. p. No. Vol. Econometrica. 9. 509-525.D. Resources. Vol. 49-62. Vol. 12. 21.

Cohen A J and Harcourt G C (2003). American Economic Review. Scand. 4. “‘A Theory of Production’: The Estimation of the CobbDouglas Function: A Retrospective View”. Färe R and Mitchell T M (1989). pp. Christensen L R. Journal of Economic Perspectives. Econometrica. 28-45. No. “Estimation of Joint Production Functions”. Dhrymes P J and Mitchell B M (1969). pp.22. Scand. 4. pp. 481-507. Available at http://repository. Technical Progress and Time Trends”. Vol. pp. No. pp. 139-165. pp. “All Production is Joint Production: A Thermodynamic Analysis”. Felipe J and McCombie J S L (2005). 27. J. 18. 31. pp. pp. 3. pp. 749-757. Farrell J M (1957). 29. pp. 31. No. 5. 32. 669-674. “A Theory of Production”. 467-488. 79.ust. 73. Vol. Cheltenham. No. 199-214.pdf 36. Business School. 25. 1. 88. “An Extensive Bibliography of Data Envelopment Analysis (DEA)”. 427-445. 23. 91. 37. The Review of Economics and Statistics. No. Vol. 24. Vol. 18-22. Emrouznejad A (2001). 131-158. Vol. No. No. “A Family Tree of Linearly Homogenous Production Functions”. Felipe J and Fisher F M (2001). Chizmar J F and Zak T A (1983). Metroeconomica. “Retrospectives: Whatever Happened to the Cambridge Capital Theory Controversies?”. Eastern Economic Journal. “How Sound are the Foundations of the Aggregate Production Function?”. 54. Vol. in S Faucheux. 253-281. J. I. Vol. 30. 1. Jorgenson D W and Lau L J (1973). Eastern Economic Journal. No. 55. Vol. Diewert W E (1971). Färe R (1986). 732-736. 4.hk/dspace/bitstream/1783. Vol. 1.com/abstract=422067 34. No. Journal of the Royal Statistical Society. Vol. Edward Elgar.). Available at http:// ssrn. No. Vol. Sustainability and Firms: Technological Change and the Changing Regulatory Environment. “Transcendental Logarithmic Production Frontiers”. “The Measurement of Productive Efficiency”. Cobb C W and Douglas P H (1928). “On the Existence and Equivalence of Three Joint Production Functions”. “Modeling Multiple Outputs in Educational Production Functions”. 17. 31. 120. Vol. 3. 28.1/ 2200/1/FelipeHolzprerefereedAggregateProdFcn3Nov98. Felipe J and Adams F G (2005). The American Economic Review. The Journal of Political Economy. 26. Vol. Proops J L R and Baumgärtner S (1998). 35. “An Application of the Shephard Duality Theorem: A Generalized Leontief Production Function”. 208-262. Working Paper. pp. Faber M. University of Warwick. “Aggregation in Production Functions: What Applied Economists Should Know”. “On Production Functions. pp. of Economics. No. 5. A Brief History of Production Functions 27 . 2. J Gowdy and I Nicolaï (Eds. of Economics. pp. 33. Felipe J and Holz C A (2005).

VIII. No. North-Holland Publishing Co. Dordrecht. D Reidel. Encyclopedia of Political Economy. The Review of Economics and Statistics. Graham J W and Green C A (1984). 465-468. 50. Cambridge. No. Routledge. Vol. pp. Fairytale”. Vol. Frisch R (1965). The American Economic Review. Distribution. 966-974. Harvard University Press. 51. pp. “Some Cambridge Controversies in the Theory of Capital”. pp. Review of Economics and Statistics. 5. Vol. 100-113.). 51. 4. Journal of Economic Theory. Fioretti G (2006). 53. 1099-1123. Cambridge. Gronau R (1977). “Leisure. Amsterdam. “A Closed Linear Model of Production”. (Eds. Eastern Economic Journal.org/PS_cache/physics/pdf/0511/0511191. Activity Analysis of Production and Allocation. 2010 . No. Gehrke C and Lager C (2000). 2. pp. Halter A N. 59. “Capital-Skill Complementarity”. Journal of Economic Literature. in T C Koopmans (Ed. 4. Aggregation. “Aggregate Production Functions: A Pervasive. 423-426. New York. 277-282. Carter H O and Hocking J G (1957). pp. 48. but Unpersuasive. 39. No. Vol. Preprint available at http://arxiv. pp. Cowles Commission Monograph. 52. 63. “Production Function”. Griliches Z and Ringstad V (1971). 41. Journal of Farm Economics.). 2. 285-303. “Estimating the Parameters of a Household Production Function with Joint Products”. The MIT Press. 85. Vol. 40. 39. Aggregate Production Functions and Related Topics. “A Note on the Transcendental Production Function”. Economies of Scale and the Form of Production Function. 4. Fisher F M (2005). 44. Harris D J (1973). No. 4. Hanoch G (1970). Georgescu-Roegen N (1971). No. Vol. 489-491. “The Economics of Joint Production: Another Approach”. “Homotheticity in Joint Production”.37. pp. Griliches Z (1969). MA. Fisher F M (1993). 66. Princeton University Press. Vol. No. “The Aggregate Linear Production Function and Its Applications to von Neumann’s Economic Model”. No. pp. 46. Wiley. 7. The Review of Economics and Statistics. 49. Home Production.pdf 38. Georgescu-Roegen N (1951). 31. 1. No. 13. 43. Vol. pp. 42. Working Paper. Theory of Production. 389-397. Linear Inequalities and Related Systems. 45. Harcourt G C (1969). 4. Vol. “Sraffian Political Economy”. Vol. 369-405. “Capital. 6.. Mass. 2. and Work: The Theory of the Allocation of Time Revisited”. No. in H W Kuhn et al. Griffin J M (1977). 47. and the Aggregate Production Function”. The Entropy Law and the Economic Process. pp. No. 28 The IUP Journal of Managerial Economics. The Journal of Political Economy. Gale D (1956).

Hotelling H (1936). Strassl W. 1-13. 57. Vol. Kümmel R. 61. 63. No. “Production Functions and Elasticity of Substitution”. pp. 2nd Edition (1963). Kadiyala K R (1972). “The Joint Production Function of Starch Equivalent and Protein Equivalent in Feeding Dairy Cows”. 2. pp. Available at http://ideas. 56. Lindenberger D and Eichhorn W (2000). Klein L R (1947).ps. 59. Charles Griffin & Co.gz A Brief History of Production Functions 29 .physik. Vol. Journal of Economics. Indian Journal of Applied Economics. Martin’s Press. Available at ftp://ftp. pp. “Estimation of Multicrop Production Functions”. No. Aberystwyth. 68. No. Gossner A and Eichhorn W (1985). 2. Econometric Models. pp. Intriligator M D (1978). No. 69. 1. pp. 65. Hicks J R (1932). St. 67. Vol. 64. 285-311.uni-wuerzburg. 8. Jawetz M B (1961). The Advanced Theory of Statistics. Vol. Energy: The International Journal. Hudson E A and Jorgenson D W (1974). Vol. No. “The Impact of Energy on Industrial Growth”. 83. London. The Bell Journal of Economics and Management Science. No. 2. Vol. 3. “On Estimation of the CES Production Function”. Kümmel R. Vol. 1-26. The American Economic Review. Southern Economic Journal. Federal Reserve Bank of Richmond. 189-203. 66. 8.html 58. The Theory of Wages. 1975-2000”. Kendall M G and Stuart A (1968). 3.54. 51-83. University College of Wales. Vol. New Jersey.repec.. No. pp. pp. International Economic Review. 180-189. Kmenta J (1967). 770-780. New York. “The Productive Power of Energy and Economic Evolution”. 7. 45. Techniques and Applications. pp. Economic Quarterly. 321-377. Humphrey T M (1997). Zilberman D and Hochman E (1983). Vol. 38. No. Washington DC. 4. 5. Department of Economics. Nos. 60. 3.org/a/fip/fedreq/y1997iwinp51-83. pp. “The Use of Cross-Section Data in Econometrics with Application to a Study of Production of Railroad Services in the United States” (Mimeographed).de/pub/preprint/1998/WUE-ITP-98033. Vol. “Economics Among the Sciences”. “US Energy Policy and Economic Growth. “Technical Progress and Energy Dependent Production Functions”. 1. Vol. Kümmel R (1982). “Algebraic Production Functions and Their Uses Before CobbDouglas”. 55. 28.. No. 62. pp. 281-284. 3-4. Prentice Hall Inc. 2. Just R E. Koopmans T C (1979). Biometrica. 461-514. National Bureau of Economic Research. “Relations Between Two Sets of Variates”. American Journal of Agricultural Economics. 65.

Available at http://www. Review of Economics and Statistics. 71. Institute of Energy Economics. Vol. 73-83. McCarthy M D (1967). Carnegie-Rochester Conference Series on Public Policy. 3. 1. Vol. McFadden D (1963). No. pp.ewi. pp. Review of Economic Studies. 2. Vol. McFadden D (1962). No. pp. Factor Substitution in the Economic Analysis of Production. Metroeconomica. No. “The Nonswitching Theorem is False”. pp. Manne A S (1958). 1. Cologne. 4.com/author=353253 30 The IUP Journal of Managerial Economics. 77. “Capital Reversing”. “Elementary Geometric/Arithmetic Series and Early Production Theory”. “Service Production Functions”. “Energy-Dependent Production Functions and the Optimization Model ‘PRISE’ of Price-Induced Sectoral Evolution”. Boston. pp. Lovell C A L and Schmidt P (1988). 38. 449-452. Dissertation. 21-34.de/ewi/content/e266/e283/e281/ Ewiwp0302_ger. Journal of Political Economy. “Allocative Efficiency versus X-Efficiency”. No. Vol. VIII. 82. Encyclopedia of Political Economy. 77. Econometrica. pp. 392-415. “A Generalization of the CES Production Function”. Leibenstein H. 56. Available at http://www. Vol. 1. 8. 101-107. Levhari D and Samuelson P A (1966). University of Cologne (EWI). 1. in A Dogramaci and R Färe (Eds. 1. 80. Vol. Lucas R (1976). Vol. “A Comparison of Alternative Approaches to the Measurement of Productive Efficiency”. International Economic Review. No. “A Linear Programming Model of the US Petroleum Refinery Industry”. Vol. pp. Routledge. 3. No. Kluwer. 83. Vol. Lavoie M (2000). pp.ssrn. 67-106. 1957. 1-37. Cornell University Press. 80. The Quarterly Journal of Economics. “Approximation of the CES Production Function: A Comment”. Lindenberger D and Kümmel R (2002). 4. University of Minnesota. Lindenberger D (2003). Vol. 19-46. 78. Ph. 190-192. Kurz H D (1986). 84. 76. No. 50. Vol. 72. 3-32. Lloyd P J (1969). Lu Y C and Fletcher L B (1968). pp. Mishra S K (2006). 1. 4. “Classical and Early Neoclassical Economists on Joint Production”.). 26. pp.unikoeln. “Econometric Policy Evaluation: A Critique”. International Journal of Applied Thermodynamics. No. No.69. Blair G and Hodgson M (1988).D. EWI Working Paper No. 30. Liu T C and Hildebrand G H (1965). 75. No.02. Ithaca. Applications of Modern Production Theory: Efficiency and Productivity. No. American Economic Review. “Constant Elasticity of Substitution Production Function”. 5. 2010 . “A Note on Numerical Estimation of Sato’s Two-Level CES Production Function”. Manufacturing Production Functions in the United States. 81. 73. 70.pdf 74. 79. 518-519. 03. pp.

pdf 87. No. 37.. “Performance of Differential Evolution Method in Least Squares Fitting of Some Typical Nonlinear Curves”. pp. Vol. 90. No. 45. Mitscherlich E A (1909). Pollak R A and Wachter M L (1975). No. “The Relevance of the Household Production Function and Its Implications for the Allocation of Time”. Vol. No. Economics Bulletin. 98. “A Generalized SMAC Function with Constant Ratios of Elasticities of Substitution”. pp. 273-284. 96. 233-236. pp. 537-552. 53. Econometrica. 91. Pittsburgh. Available at http://economicsbulletin. “Production Functions with Variable Elasticity of Substitution and Variable Returns to Scale”. 38. 3. Vol. Mundlak Y (1964). A Brief History of Production Functions 31 . Mishra S K (2007c). 3. “Returns to Scale in Electricity Supply”. 3. Vol. 5. Econometrica. 14. Mundlak Y and Razin A (1971). pp. 433-443. “Estimation of Zellner-Revankar Production Function Revisited”. Doctoral Dissertation. 89. USA. pp. “Transcendental Multiproduct Production Functions”. Paper Read at the Meetings of the Econometric Society and the American Farm Economic Association. “Specification and Estimation of Multiproduct Production Functions” (Mimeographed). Vol. 1. Mishra S K (2007b). pp. Review of Economic Studies. Landw. No. Vol.. Vol. 83. 737-738. Econometrica. 39. 94. 51.vanderbilt.edu/2007/ volume3/EB-07C10007A. “On Multistage Multiproduct Production Functions”. 29. No. Vol. 2. 4. New York. pp. 3. Summarized in the Journal of Farm Economics. Mundlak Y (1963). 1009. December 1962. Convex Structures and Economic Theory. pp. Economics Bulletin. Mukerji V (1963). 3. 2. Vol. 1-13. Rao P (1969). 491-499. American Journal of Agricultural Economics. pp. No. North-Holland Publishing Co. University of Wisconsin. 99.com/ abstract=1010508 88. 30. 93. 1-7. “A Note on Econometrics of Joint Production”. 65-68. “A Class of Variable Elasticity of Substitution Production Functions”. Estimation and Identification of Cobb-Douglas Production Functions. pp. Reprinted in M Nerlove (1965). “Das Gesetz des Minimums und das Gesetz des abnehmenden Bodenertrages”.85. 95. Pfouts R W (1961). Available at http://ssrn. Jahrb. Academic Press. USA. No. Revankar N S (1967). Nikaido H (1968). pp. Amsterdam. 92. Vol. “The Theory of Cost and Production in the Multiproduct Firm”. Nerlove M (1963). Revankar N S (1971). Mishra S K (2007a). The Journal of Political Economy. 255-278. 97. No. International Economic Review. 2. No. No. Vol. “Least Squares Estimation of Joint Production Functions by the Differential Evolution Method of Global Optimization”. 61-71. 86.

Vol. Sato R (1975). 81-106. “Recent Developments in DEA: The Mathematical Programming Approach to Frontier Analysis”. Vol.edu/~AShaikh/humbug2. Vol. 46. pp. A History of Western Philosophy. No. 69. Bulletin of Economic Research. 48. Vol. Reprint available at the website http://homepage.100. 11. “Rejoinder: Agreements. Counterpoint Edition. No. “A Two-Level Constant-Elasticity-of-Substitution Production Function”. Vol. pp. 106.pdf 115. “Parable and Realism in Capital Theory: The Surrogate Production Function”. Review of Economic Studies. 453-460. No. 110. Nos. No. London. Econometrica. No. 50. The Review of Economics and Statistics. Salvadori N and Steedman Z (1988). Vol. Journal of Econometrics. 1. Growth. Swedish Journal of Economics. 923-939. Shaikh A (1974). No. 4. 105. pp. pp. Sato K (1967). Unwin Paperbacks. Samuelson P A (1979). History of Economic Analysis. Economic Studies Quarterly. “The Most General Class of CES Functions”. pp. 114. Robinson J (1953). 2010 . No. “Joint Production Analysis in a Sraffian Framework”. VIII. 193-206. 111. 5. Doubts.newschool. pp. 1. Cambridge University Press. 999-1003. 101. 179-208. 2. Vol. 115-123. London. “Paul Douglas’s Measurement of Production Functions and Marginal Productivities”. 102. 56. 34. “Production Functions with Variable Elasticity of Factor Substitution: Some Analysis and Testing”. Shaikh A (1980). “The Production Function and the Theory of Capital”. No. “On the Class of Separable Non-Homothetic CES Functions”. 4. Samuelson P A (1962). Profits and Property. 109. Review of Economic Studies. 108.newschool. 113. Russell B (1984). 5-6. Sato R and Hoffman R F (1968). 2. No. No. Review of Economics and Statistics. Vol. 104. “Laws of Production and Laws of Algebra: Humbug II”. 444-448. Vol. 165-196. pp.). 3. and the Case of Induced Harrod-Neutral Technical Change”. Vol. pp. Ringstad V (1967). Vol. Seiford L M and Thrall R M (1990). 3. Review of Economic Studies. Journal of Political Economy. Reprint available at http://homepage. 2. Disagreements. pp. Vol. 40. 107. pp.pdf 32 The IUP Journal of Managerial Economics. International Economic Review. Sato R (1974). “Laws of Production and Laws of Algebra: The Humbug Production Function”. 43. pp. Samuelson P A (1966). No. 201-218. Schumpeter J A (1954). 2. 115-120. 21. Vol. Vol. pp. in E J Nell (Ed. Sato R (1970). pp. Review of Economics and Statistics. 87. 103. “The Estimation of Biased Technical Progress and the Production Function”. 112. 7-38. 42-55. No. 4. Allen and Unwin. No. Cambridge. “Econometric Analysis Based on Production Function with Neutrally Variable Scale Elasticity”. 29. 25. 1.edu/~AShaikh/humbug.

1935-1936 (Franz-Deuticke). 1-9. “Econometrics of Joint Production”. Review of Economic Studies. Vol. 126. pp. pp. “Über ein ökonomisches Gleichungssystem und eine Verallgemeinerung des Brouwerschen Fixpunktsatzes”. 39. 128. Eastern Economic Journal. No. 4. Deutsch E and Rieder R (1974). On the Measurement of Factor Productivities. pp. 37. 121. Shephard R W (1970). Grundlagen einer reinen Kostentheorie (Foundations of a Pure Theory of Costs). pp. 118. von Neumann J (1937). Vol. Wien. “The Cobb-Douglas or the Wicksell Function? A Comment”. 8. The Law of Diminishing Returns. 5. 122. Princeton University Press. 120. 893-903. The Review of Economics and Statistics. No. Germany. “Canonical Ridge and Econometrics of Joint Production”. Jena. Stigler G J (1952). 447-466. Vinod H D (1969). No.). Vol. 312-320. 125. 31. 4. 123. Translated: A Model of General Economic Equilibrium. No. in F L Altman. pp. 739-740.. Theory of Cost and Production Functions. 291-299. 60. Shaikh A (2005). Vol. 4. The Journal of Political Economy.116. 129. Vol. 124. World Book Co. O Kyn and H J Wagener (Eds. Journal of Political Economy. No. Review of Economic Studies. Velupillai K (1973). Vol. Vol. No. 322-336. Thünen Johann von (1930). 117. No. 147-166. 151-164. Fischer. 130. pp. 3. Journal of Econometrics. pp. “Nonlinear Dynamics and Pseudo-Production Functions”. 16. Tintner G. Econometrica. 187-207. Yonkers-on-Hudson. 2. Vandenhoek & Ruprecht. pp. Der isolierte Staat in Beziehung auf Landwirthschaft und National¨okonomie. Spillman W J and Lang E (1924). A Brief History of Production Functions 33 . pp. in Ergebnisse eines Mathematischen Kolloquiums. Göttingen. 1. 36. Econometrica. pp. 2. New York. Stiglitz J E (1974). 82. “Technical Change and the Aggregate Production Function”. 119. 127. 111-113. Vol. Vol. No. Vinod H D (1976). Uzawa H (1962). 4. Verlag von Julius Springer. Vol. No. Princeton. Economy and History. “The Ricardian Theory of Value and Distribution”. pp. 13. 3. Vol. Vinod H D (1968). NJ. Solow R M (1957). “Production Functions with Constant Elasticities of Substitution”. 73-83. “The Cambridge-Cambridge Controversy in the Theory of Capital: A View from New Haven”. 3 Volumes. “A Production Function for Austria Emphasizing Energy”. pp. No. 29. 1 (1945-1946). von Stackelberg H (1932). “Econometrics of Joint Production: A Reply”.

2010 . 51. “Production Function”. “Engineering Production Functions: A Survey”. 65. No. Multiple Output Production Choices and Technology in the US Wheat Region”. 135. General Theory. VIII. 2. An Essay on the Co-ordination of the Laws of Distribution. “Real Capital and Interest”. Lectures on Political Economy. 132. pp. 1. Vol. Vol. London. No. The Review of Economic Studies. No. New York. 2. “Generalized Production Functions”. 45-56. Wibe S (1984). Wicksteed P H (1894).org/wiki/ Koopmans’_theorem 138. Appendix 2. The Free Press. London.wikipedia.. Vol.newschool. pp. Zellner A and Revankar N S (1969). Wikipedia-a. Reference # 21J-2010-11-01-01 34 The IUP Journal of Managerial Economics. 1. Wicksell K (1923). pp. Economica. Routledge & Kegan Paul. 1867-1890.pdf 136. Wikipedia-b. Whitaker J K (Ed.131. Vol.). in L Robbins (1934) (Ed.wikipedia. Macmillan & Co. New Series. The Early Economic Writings of Alfred Marshall. 204. Weaver R D (1983). Vol. Available at http://cepa. 133. “Koopmans’ Theorem”. 4. Vol. Review of Gustaf Akerman’s Realkapital und Kapitalzins. Available at http://en. 134. “Multiple Input. 401-411. American Journal of Agricultural Economics. Available at http://en.edu/het/texts/wicksteed/ wickess.) (1975). No. 36.org/wiki/ Production_function 137. 241-250.

.Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.