Production Function, Lecture Notes - Economics, Study notes for Economics. The London School of Economics and Political Science (LSE)

Economics

Description: Production functions
Showing pages  1  -  2  of  47
The preview of this document ends here! Please or to read the full document or to download it.
Document information
Uploaded by: floweryy
Views: 1000+
Downloads : 5
University: The London School of Economics and Political Science (LSE)
Address: Economics
Subject: Economics
Upload date: 08/09/2011
Embed this document:
ProductionFunctions.DVI

PRODUCTION FUNCTIONS

1. ALTERNATIVE REPRESENTATIONS OF TECHNOLOGY

The technology that is available to a firm can be represented in a variety of ways. The most general are those based on correspondences and sets.

1.1. Technology Sets. The technology set for a given production process is de- fined as

T = {(x, y) : x ∈ Rn+, y ∈ Rm :+ x can produce y} where x is a vector of inputs and y is a vector of outputs. The set consists of

those combinations of x and y such that y can be produced from the given x.

1.2. The Output Correspondence and the Output Set.

1.2.1. Definitions. It is often convenient to define a production correspondence and the associated output set.

1: The output correspondence P, maps inputs x ∈Rn+ into subsets of outputs, i.e., P: Rn+ → 2R

m + . A correspondence is different from a function in that a

given domain is mapped into a set as compared to a single real variable (or number) as in a function.

2: The output set for a given technology, P(x), is the set of all output vectors y ∈ Rm+ that are obtainable from the input vector x ∈ Rn+. P(x) is then the set of all output vectors y ∈ Rm+ that are obtainable from the input vector x ∈Rn+ . We often write P(x) for both the set based on a particular value of x, and the rule (correspondence) that assigns a set to each vector x.

1.2.2. Relationship between P(x) and T(x,y).

P (x) = (y : (x, y ) ∈ T )

1.2.3. Properties of P(x). P.1a: Inaction and No Free Lunch. 0 ∈ P(x) ∀ x ∈ Rn+ . P.1b: y 6∈ P(0), y > 0. P.2: Input Disposability. ∀ x ∈ Rn+ , P(x) ⊆ P(θx), θ ≥ 1. P.2.S: Strong Input Disposability. ∀ x, x’ ∈ Rn+ , x’ ≥ x⇒ P(x) ⊆ P(x’). P.3: Output Disposability. ∀ x ∈ Rn+ , y ∈ P(x) and 0 ≤ λ ≤ 1⇒ λy ∈ P(x). P.3.S: Strong Output Disposability. ∀ x ∈ Rn+ , y ∈ P(x) ⇒ y’ ∈ P(x), 0 ≤ y’ ≤

y. P.4: Boundedness. P(x) is bounded for all x ∈ Rn+ . P.5: T is a closed set P: Rn+ → 2R

m + is a closed correspondence, i.e., if [x` →

x0, y` → y0 and y` ∈ P(x`), ∀ `] then y0 ∈ P(x0). P.6: Attainability. If y ∈ P(x), y ≥ 0 and x ≥ 0, then ∀ θ ≥ 0, ∃ λθ ≥ 0 such

that θy ∈ P(λθx).

Date: August 29, 2005. 1

2 PRODUCTION FUNCTIONS

P.7: P(x) is convex P(x) is a convex set for all x ∈ Rn+ . This is equivalent to the correspon-

dence V:<n+ → 2< m + being quasiconcave.

P.8: P is quasi-concave. The correspondence P is quasi-concave on Rn+ which means ∀ x, x’ ∈

Rn+ , 0 ≤ θ ≤ 1, P(x) ∩ P(x’) ⊆ P(θx + (1-θ)x’). This is equivalent to V(y) being a convex set.

P.9: Convexity of T. P is concave on Rn+ which means ∀ x, x’ ∈ Rn+ , 0 ≤ θ ≤ 1, θP(x)+(1-θ)P(x’) ⊆ P(θx + (1-θ)x’)

1.3. The Input Correspondence and Input (Requirement) Set.

1.3.1. Definitions. Rather than representing a firm’s technology with the technol- ogy set T or the production set P(x), it is often convenient to define an input corre- spondence and the associated input requirement set.

1: The input correspondence maps outputs y ∈ Rm+ into subsets of inputs, V: Rm+ → 2R

n + . A correspondence is different from a function in that a

given domain is mapped into a set as compared to a single real variable (or number) as in a function.

2: The input requirement set V(y) of a given technology is the set of all com- binations of the various inputs x ∈ Rn+ that will produce at least the level of output y ∈ Rm+ . V(y) is then the set of all input vectors x ∈ Rn+ that will produce the output vector y ∈ Rm+ . We often write V(y) for both the set based on a particular value of y, and the rule (correspondence) that assigns a set to each vector y.

1.3.2. Relationship between V(y) and T(x,y).

V (y) = (x : (x, y) ∈ T )

1.4. Relationships between Representations: V(y), P(x) and T(x,y). The technol- ogy set can be written in terms of either the input or output correspondence.

T = {(x, y) : x Rn+, y Rm+ , such that x will produce y} (1a)

T = {(x, y) ∈ Rn+m+ : y P (x), x Rn+} (1b)

T = {(x, y) ∈ Rn+m+ : x V (y), y Rm+} (1c) We can summarize the relationships between the input correspondence, the

output correspondence, and the production possibilities set in the followingpropo- sition.

Proposition 1. y P(x)x V(y)(x,y) T

2. PRODUCTION FUNCTIONS

2.1. Definition of a Production Function. To this point we have described the firm’s technology in terms of a technology set T(x,y), the input requirement set V(y) or the output set P(x). For many purposes it is useful to represent the re- lationship between inputs and outputs using a mathematical function that maps vectors of inputs into a singlemeasure of output. In the casewhere there is a single

PRODUCTION FUNCTIONS 3

output it is sometimes useful to represent the technology of the firm with a math- ematical function that gives the maximum output attainable from a given vector of inputs. This function is called a production function and is defined as

f (x) = max y

[y : (x, y) ∈ T ]

= max y

[y : x V (y)]

= max y ∈ P (x)

[y]

(2)

Once the optimization is carried out we have a numerically valued function of the form

y = f (x1, x2 , . . . , xn) (3)

Graphically we can represent the production function in two dimensions as in figure 1.

FIGURE 1. Production Function

x

y

fHxL

In the case where there is one output, one can also think of the production func- tion as the boundary of P(x), i.e., f(x) = Eff P(x).

2.2. Existence and the Induced Production Correspondence. Does the produc- tion function exist. If it exists, is the output correspondence induced by it the same as the original output correspondence fromwhich f was derived? What properties does f(x) inherit from P(x)?

a: To show that production function exists and is well defined, let x ∈ Rn+. By axiom P.1a, P(x) 6= ∅. By axioms P.4 and P.5, P(x) is compact. Thus P(x) contains a maximal element and f(x) is well defined. NOte that only these three of the axioms on P are needed to define the production function.

b: The output correspondence induced by f(x) is defined as follows

Pf (x) = [y R+ : f(x) ≥ y], x Rn+ (4)

4 PRODUCTION FUNCTIONS

This gives all output levels y that can be produced by the input vector x. We can show that this induced correspondence is equivalent to the output correspondence that produced f(x). We state this in a proposition.

Proposition 2. Pf (x) = P (x), x Rn+.

Proof. Let y ∈ Pf (x), x ∈ Rn+ . By definition, y ≤ f(x). This means that y ≤ max {z: z ∈ P(x)}. Then by P.3.S, y ∈ P(x). Now show the other way. Let y ∈ P(x). By the definition of f, y ≤ max {z: z ∈ P(x)} = f(x). Thus y ∈ Pf (x).



Properties P.1a, P.3, P.4 and P.5 are sufficient to yield the induced pro- duction correspondence.

2.2.1. Relationship between P(x) and f(x). We can summarize the relation- ship between P and f with the following proposition:

Proposition 3. y P(x)f(x)y, x Rn+ 2.3. Examples of Production Functions.

2.3.1. Production function for corn. Consider the production technology for corn on a per acre basis. The inputs might include one acre of land and various amounts of other inputs such as tillage operations made up of tractor and implement use, labor, seed, herbicides, pesticides, fertilizer, harvesting operations made up of dif- ferent combinations of equipment use, etc. If all but the fertilizer are held fixed, we can consider a graph of the production relationship between fertilizer and corn yield. In this case the production function might be written as

y = f (land, tillage, labor, seed, fertilizer, . . . ) (5)

2.3.2. Cobb-Douglas production function. Consider a production function with two inputs given by y = f(x1, x2). A Cobb-Douglas [4] [5] represention of technology has the following form.

y = Axα11 x α2 2

= 5x 1 3 1 x

1 4 2

(6)

Figure 2 is a graph of this production function.

Figure 3 shows the contours of this function.

With a single output and input, a Cobb-Douglas production function has the shape shown in figure 4.

2.3.3. Polynomial production function. We often approximate a production function using polynomials. For the case of a single input, a cubic production function would take the following form.

y = α1 x + α2 x2 + α3 x3

= 10x + 20x2 − 0.60x3 (7)

The cubic production function in equation 7 is shown in figure 5.

PRODUCTION FUNCTIONS 5

FIGURE 2. Cobb-Douglas Production Function

5

10

15

20

x1

5

10

15

20

x2

0

10

20f Hx1,x2L

FIGURE 3. Contours of a Cobb-Douglas Production Function

5 10 15 20 25 30

5

10

15

20

25

30

Notice that the function first rises at an increasing rate, then increases at a de- creasing rate and then begins to fall until it reaches zero.

6 PRODUCTION FUNCTIONS

FIGURE 4. Cobb-Douglas Production Function with One Input

x

y

FIGURE 5. Cubic Production Function

x

y

fHxL

2.3.4. Constant elasticity of substitution (CES) production function. An early alterna- tive to the Cobb-Douglas production function is the constant elasticity of substi- tution (CES) production function [1]. While still being quite tractable, with a min- imum of parameters, it is more flexible than the Cobb-Douglas production func- tion. For the case of two inputs, the CES production function takes the following form.

y = A [ δ1 x

−ρ 1 + δ2 x

−ρ 2

]−1 ρ

= 5 [ 0.6x−21 + 0.2x

−2 2

]−1 2

(8)

The CES production function in equation 8 is shown in figure 6. The production contours of the production function in equation 8 are shown in figure 7. If we change ρ to 0.2, the CES contours are as in figure 8.

2.3.5. Translog production function. An alternative to the Cobb-Douglas production function is the translog production function.

PRODUCTION FUNCTIONS 7

FIGURE 6. CES Production Function

5

10

15

20

x1

5

10

15

20

x2

0 25 50 75

100

f Hx1,x2L

FIGURE 7. CES Production Function Contours - ρ=2

2.5 5 7.5 10 12.5 15 17.5 20

2.5

5

7.5

10

12.5

15

17.5

20

ln y = α1 ln x1 + α2 ln x2 + β11 ln x21 + β12 ln x1 ln x2 + β22 ln x 2 2

= 1 3 ln x1 +

1 10 ln x2 −

2 100

ln x21 + 1 10 ln x1 ln x2 −

2 10

ln x22 (9)

The translog can also be written with y as compared to ln y on the left hand side.

8 PRODUCTION FUNCTIONS

FIGURE 8. CES Production Function Contours - ρ=0.05

2.5 5 7.5 10 12.5 15 17.5 20

2.5

5

7.5

10

12.5

15

17.5

20

y = A , xα11 x α2 2 e

β11 ln, x 2 1 + β12 ln x1 lnx2 + β22 lnx

2 2

= x1/31 x 1/10 2 e

−.02 ln x21 + 0.1 ln x1 ln x2 − 0.2 ln x 2 2

(10)

Figure 9 shows the translog function from equation 9 while figure 10 shows the contours of the translog function.

2.4. Properties of the Production Function. We can deduce a set of properties on f that are equivalent to the properties on P in the sense that if a particular set holds for P, it implies a particular set on f and vice versa.

2.4.1. f.1 Essentiality. f(0) = 0.

2.4.2. f.1.S Strict essentiality. f(x1, x2, . . . , 0, . . . , xn) = 0 for all xi.

2.4.3. f.2 Monotonicity. ∀ x ∈ Rn+ , f(θx) ≥ f(x), θ ≥ 1.

2.4.4. f.2.S Strict monotonicity. ∀ x, x’ ∈ Rn+ , if x ≥ x’ then f(x) ≥ f(x’).

2.4.5. f.3 Upper semi-continuity. f is upper semi-continuous on Rn+ .

2.4.6. f.3.S Continuity. f is continuous on Rn+.

2.4.7. f.4 Attainability. If f(x) > 0, f(λx) → +∞ as λ→ +∞.

2.4.8. f.5 Quasi-concavity. f is quasi-concave on Rn+.

2.4.9. f.6 Concavity. f is concave on Rn+.

2.5. Discussion of the Properties of the Production Function.

PRODUCTION FUNCTIONS 9

FIGURE 9. Translog Production Function

5

10

15

20

x1

5

10

15

20

x2

0

10

20 f Hx1,x2L

FIGURE 10. Contours of Translog Production Function

2.5 5 7.5 10 12.5 15 17.5 20

2.5

5

7.5

10

12.5

15

17.5

20

2.5.1. f.1 Essentiality. f(0) = 0.

10 PRODUCTION FUNCTIONS

This assumption is sometimes called essentiality. It says that with no inputs, there is no output.

2.5.2. f.1.S Strict essentiality. f(x1, x2, . . . , 0, . . . , xn) = 0 for all xi.

This is called strict essentiality and says that some of each input is needed for a positive output. In this case the input requirement set doesn’t touch any axis. Consider as an example of strict essentiality the Cobb-Douglas function.

y = Axα11 x α2 2 (11)

Another example is the Generalized Leontief Function with no linear terms

y = β11x1 + 2β12 x1 x2 + β22 x2 (12)

2.5.3. f.2 Monotonicity. ∀ x ∈ Rn+ , f(θx) ≥ f(x), θ ≥ 1.

This is amonotonicity assumption that sayswith a scalar expansion of x, output cannot fall. There is also a strong version.

2.5.4. f.2 Strict monotonicity. ∀ x, x’ in Rn+ , if x ≥ x’ then f(x) ≥ f(x’).

Increasing one input cannot lead to a decrease in output.

2.5.5. f.3 Upper semi-continuity. f is upper semi-continuous on Rn+ .

The graph of the production function may have discontinuities, but at each point of discontinuity the function will be continuous from the right. The prop- erty of upper semi-continuity is a direct result of the fact that the output and input correspondences are closed. In fact, it follows directly from the input sets being closed.

2.5.6. f.3.S Continuity. f is continuous on Rn+.

We often make the assumption that f is continuous so that we can use calculus for analysis. We sometimes additionally assume the f is continuously differen- tiable.

2.5.7. f.4 Attainability. If f(x) > 0, f(λx) → +∞ as λ→ +∞.

This axiom states that there is always a way to exceed any specified output rate by increasing inputs enough in a proportional fashion.

2.5.8. f.5 Quasi-concavity. f is quasi-concave on Rn+.

If a function is quasi-concave then

f(x) ≥ f(x0) ⇒ f(λx + (1− λ) x0) ≥ f(x0) (13) If V(y) is convex then f(x) is quasi-concave because V(y) is an upper contour

set of f. This also follows from quasiconcavity of P(x). Consider for example the traditional three stage production function in figure 11. It is not concave, but it is quasi-concave.

PRODUCTION FUNCTIONS 11

FIGURE 11. Quasi-concave Production Function

x

y

If the function f is quasi-concave the upper contour or isoquants are convex. This is useful in problems of cost minimization as can be seen in figure 12.

FIGURE 12. Convex Lower Boundary of Input Requirement Set

x1

x2

2.5.9. f.6 Concavity. f is concave on Rn+. If a function is concave then

f(λx + (1 − λ)x′) ≥ λ f(x) + (1 − λ)f(x′) (14) Concavity of f follows from P.9 (V.9) or the overall convexity of the output and

input correspondences. This means the level sets are not only convex for a given level of output or input but that the overall correspondence is convex. Contrast the traditional three stage production function with a Cobb-Douglas one. Concavity is implied by the function lying above the chord as can be seen in figure 13 or below the tangent line as in figure 14.

2.6. Equivalence of Properties of P(x) and f(x). The properties (f.1 - f.6) on f(x) can be related to specific properties on P(x) and vice versa. Specifically the following proposition holds.

12 PRODUCTION FUNCTIONS

FIGURE 13. Concavity Implies that a Chord Lies below the Function

x

y

FIGURE 14. A Concave Function Lies Below the Tangent Line

x

y

Proposition 4. The output correspondence P: Rn+ → 2R+ satisfies P.1 - P.6 iff the pro- duction function f : Rn+ → R+ satisfies f.1 - f.4. Furthermore P.8 f.5 and P.9f.6.

Proofs of some of the equivalencies between properties of P(x) and f(x).

2.6.1. P.1 f.1. P.1a states that 0 ∈ P(x) ∀ x ∈ Rn+ ; y 6∈ P(0), y > 0. Let x ∈ Rn+ so that 0 ∈ P(x). Then by Proposition 3 f(x) ≥ 0 . Now if y > 0 then y 6∈ P(0) by P.1b. So y > f(0). But f(x) ≥ 0 by Proposition 3. Thus let y→ 0 to obtain f(0) = 0.

Now assume that f.1 holds. By Proposition 3 it is obvious that 0 ∈ P(x), ∀ x ∈ Rn+. Now compute P(0) = [y∈R+: f(0) ≥ y ]. This is the empty set unless y = 0. So if y > 0, then y 6∈ P(0).

2.6.2. P.2f.2. P.2 states that ∀ x ∈Rn+, P(x)⊆ P(θx), θ ≥ 1. Consider the definition of f(x) and f(θx) given by

f(x) = max [y R+ : y P (x)] x Rn+ f(θx) = max [y R+ : y P (θx)] x Rn+

PRODUCTION FUNCTIONS 13

Nowbecause P(x)⊆ P(θx) for θ ≥ 1 it is clear that themaximumover the second set must be larger than the maximum over the first set.

To show the other way remember that if y ∈ P(x) then f(x) ≥ y. Now assume that f(θx) ≥ f(x) ≥ y. This implies that y ∈ P(θx) which implies that P(x) ⊆ P(θx).

2.6.3. Definition of f P.3. Remember that P.3 states ∀ x ∈ <n+ , y ∈ P(x) and 0 ≤ λ ≤ 1 ⇒ λy ∈ P(x). So consider an input vector x and an output level y such that f(x) ≥ y. Then consider a value of λ such that 0 ≤ λ ≤ 1. Given the restrction on λ, y λy. But by Proposition 3 which follows from the definition of the producion function in equation 2 and Proposition 2, λy P (x).

2.6.4. F.3 P.4. Recall that P.4 is that P(x) is bounded for all x ∈ Rn+. Let x ∈ <n+ . The set

M (x) = {u ∈ <n+ : u x} (15) is compact as it is closed and bounded. F.3 says that f(x) is upper-semicontinuous,

thus the maximum

f(u∗) = max{f(u) : u M (x)} ≥ f(x) (16) u∗ ∈ M (x) exists. The closed interval [0,f(x)] is a subset of the closed interval

[0,f(u∗], i.e, [0, f(x)] ⊆ [0, f(u∗)]. Therefore P(x) = [0,f(x)] is bounded.

2.6.5. P.4F.3. Recall that f(x) is upper semi-continuous atx0 iff lim supn → ∞ f(xn) ≤ f(x0) for all sequences xn → x0.

Consider a sequence {xn} → x0. Let yn ≡ f(xn). Now suppose that lim supn → ∞ yn = f(x0). Then {yn} → y0 ≥ f(x0) because the maximum value of the sequence {yn} is greater than f(x0). Because P : <n+ → 2<+ is a closed correspondence, y0 ∈ [0, f(x0)] and y0 ≤ f(x0), a contradiction. Thus lim supn → ∞ f(xn) ≤ f(x0).

2.6.6. Other equivalencies. One can show that the following equivalences also hold. a: P.6⇒ f.4 b: P.7 follows from the definition of P(x) in terms of f in equation 4.

2.7. Marginal and Average Measures of Production.

2.7.1. Marginal product (MP). The firm is often interested in the effect of additional inputs on the level of output. For example, the field supervisor of an irrigated crop maywant to know howmuch crop yield will rise with an additional application of water during a particular period or a district managermaywant to knowwhatwill happen to total sales if she adds another salesperson and rearranges the assigned areas. For small changes in input levels this output response is measured by the marginal product of the input in question (abbreviated MP or MPP for marginal physical product). In discrete terms the marginal product of the ith input is given as

MPi = ∆y xi

= y2 − y1

x2i − x1i (17)

where y2 and x2 are the level of output and input after the change in the input level and y1 and x1 are the levels before the change in input use. For small changes

14 PRODUCTION FUNCTIONS

in xi the marginal physical product is given by the partial derivative of f(x) with respect to xi, i.e.,

MPi = ∂f (x) ∂xi

= ∂y

∂xi (18)

This is the incremental change in f(x) as xi is changed holding all other inputs levels fixed. Values of the discrete marginal product for the production function in equation 19 are contained in table 2.7.1.

y = 10x + 20x2 − 0.60x3 (19) For example the marginal product in going from 4 units of input to 5 units is

given by

MPi = ∆y xi

= 475 − 321.6

5 − 4 = 153.40

The production function in equation 19 is shown in figure 15.

FIGURE 15. Cubic Production Function

10 20 30 x

1000

2000

3000

4000 y

fHxL=10x+20x2-0.6x3

PRODUCTION FUNCTIONS 15

TABLE 1. Tabular representation of y = 10 x + 20 x2 − 0.60 x3

Input (x) Output (y)

Average Product y/x

Discrete Marginal Product

∆y ∆x

Marginal Product

0.00 0.00 10.00 1.00 29.40 29.40 29.40 48.20 2.00 95.20 47.60 65.80 82.80 3.00 193.80 64.60 98.60 113.80 4.00 321.60 80.40 127.80 141.20 5.00 475.00 95.00 153.40 165.00 6.00 650.40 108.40 175.40 185.20 7.00 844.20 120.60 193.80 201.80 8.00 1052.80 131.60 208.60 214.80 9.00 1272.60 141.40 219.80 224.20 10.0 1500.00 150.00 227.4 230.0 11.0 1731.40 157.40 231.4 232.2 12.0 1963.20 163.60 231.8 230.8 13.0 2191.80 168.60 228.6 225.8 14.0 2413.60 172.40 221.8 217.2 15.0 2625.00 175.00 211.4 205.0 16.0 2822.40 176.40 197.4 189.2 17.0 3002.20 176.60 179.8 169.8 18.0 3160.80 175.60 158.6 146.8 19.0 3294.60 173.40 133.8 120.2 20.0 3400.00 170.00 105.4 90.0

We can compute themarginal product of the production function given in equa- tion 19 using the derivative as follows

dy dx

= 10 + 40 x − 1.80 x2 (20)

At x = 4 this gives 141.2 while at x = 5 this gives 165.0. The marginal product function for the production function in equation 19 is shown in figure 16.

Notice that it rises at first and then falls as the production function’s rate of increase falls. Although we typically do not show the production function and marginal product in the same diagram (because of differences in scale of the ver- tical axis), figure 17 shows both measures in the same picture to help visualize the relationships between the production function and marginal product.

2.7.2. Average product (AP). The marginal product measures productivity of the ith input at a given point on the production function. An average measure of the relationship between outputs and inputs is given by the average product which is just the level of output divided by the level of one of the inputs. Specifically the average product of the ith input is

16 PRODUCTION FUNCTIONS

FIGURE 16. Marginal Product

5 10 15 20 25 30 x

-200

-100

100

200

MP

f’HxL=40-0.36x

FIGURE 17. Production and Marginal Product

10 20 30 x

1000

2000

3000

4000 y

APi = f (x) xi

= y

xi (21)

For the production function in equation 19 the average product at x=5 is 475/5 = 95. Figure 18 shows the average and marginal products for the production func- tion in equation 19. Notice that the marginal product curve is above the average product curve when the average product curve is rising. The two curves intersect where the average product reaches its maximum.

We can show thatMP =AP at themaximumpoint ofAP by taking the derivative of APi with respect to xi as follows.

PRODUCTION FUNCTIONS 17

FIGURE 18. Average and Marginal Product

5 10 15 20 25 30 x

-200

-100

100

200

MP AP

f’HxL=40-0.36x fHxLx= 10 x + 20 x2 - 0.6€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€

x

(

f(x) xi

)

∂ xi = xi

∂ f ∂ xi

f(x) x2i

= 1 xi

( ∂ f

∂ xi − f(x)

xi

)

= 1 xi

(MPi − APi )

(22)

If we set the last expression in equation 22 equal to zero we obtain

MPi = APi (23) We can represent MP and AP on a production function graph as slopes. The

slope of a ray from the origin to a point on f(x) measures average product at that point. The slope of a tangent to f(x) at a point measures the marginal product at that point. This is demonstrated in figure 19.

2.7.3. Elasticity of ouput. The elasticity of output for a production function is given by

i = ∂ f

∂ xi

xi y

(24)

3. ECONOMIES OF SCALE

3.1. Definitions. Consider the production function given by

y = f (x1, x2, ... xn) = f (x) (25) where y is output and x is the vector of inputs x1...xn. The rate at which the

amount of output, y, increases as all inputs are increased proportionately is called the degree of returns to scale for the production function f(x). The function f is said to exhibit nonincreasing returns to scale if for all x ∈ Rn+ , λ ≥ 1, and µ ≤ 1,

f (λx) ≤ λf(x) and µf(x) ≤ f (µx) (26)

18 PRODUCTION FUNCTIONS

FIGURE 19. Average and Marginal Product as Slopes

x

y

Slope = Average Product

Slope = Marginal Product fHxL

Thus the function increases less than proportionately as all inputs x are in- creased in the same proportion, and it decreases less than proportionately as all x decrease in the same proportion. When inputs all increase by the same pro- portion we say that they increase along a ray. In a similar fashion, we say that f exhibits nondecreasing returns to scale if for all x ∈ Rn+ , λ ≥ 1, and 0 < µ ≤ 1.

f (λx) ≥ λf(x) and µf(x) ≥ f (µx) (27)

The function f exhibits constant returns to scale if for all x ∈ Rn+ and θ > 0.

f (θ x) = θf(x) (28)

This global definition of returns to scale is often supplemented by a local one that yields a specific numerical magnitude. This measure of returns to scale will be different depending on the levels of inputs and outputs at the point where it is measured. The elasticity of scale (Ferguson 1971) is implicitly defined by

 = ln f (λx) ln λ

| λ = 1 (29)

This simply explains how output changes as inputs are changed in fixed pro- portions (along a ray through the origin). Intuitively, this measures how changes in inputs are scaled into output changes. For one input, the elasticity of scale is

 = ∂f(x) ∂x

x

f(x) (30)

We can show that the expressions in equations 29 and 30 are the same as follows.

PRODUCTION FUNCTIONS 19

ln f(λx) ln λ

|λ=1 = ∂f(λx) ∂λ

λ

f(λx) |λ=1

= ( ∂f

∂λx x

) λ

f(λx) |λ=1

= ∂f

∂x

1 λ

· x λ f(λx)

|λ=1

= ∂f

∂x ·

x

f(λx) |λ=1

= ∂f

∂x

x

f(x)

(31)

So the elasticity of scale is simply the elasticity of the marginal product of x, i.e.

 = ∂f (x) ∂x

x

f(x) =

∂y

∂x

x

y =

∂ln y

∂ln x (32)

In the case of multiple inputs, the elasticity of scale can also be represented as

 = ∑

n i=1

∂f

∂xi

xi y

= n∑

i=1

∂f ∂xi y xi

= n∑

i=1

MPi APi

(33)

This can be shown as followswhere x is now an n element vector:

ln f(λx) ln λ

|λ=1 = ∂f(λx) ∂λ

λ

f(λx)

= n∑

i=1

( ∂f

∂λxi xi

) λ

f(λx) |λ=1

= n∑

i=1

∂f

∂xi

1 λ

· xi λ

f(λx) |λ=1

= n∑

i=1

∂f

∂xi

xi f(λx)

|λ=1

= n∑

i=1

∂f(x) ∂xi

xi f(x)

= n∑

i=1

MPi xi y

(34)

20 PRODUCTION FUNCTIONS

Thus elasticity of scale is the sum of the output elasticities for each input. If  is less than one, then the technology is said to exhibit decreasing returns to scale and isoquants spread out as output rises; if it is equal to one, then the technol- ogy exhibits constant returns to scale and isoquants are evenly spaced; and if  is greater than one, the technology exhibits increasing returns to scale and the iso- quants bunch as output expands. The returns to scale from increasing all of the inputs is thus the average marginal increase in output from all inputs, where each input is weighted by the relative size of that input compared to output. With de- creasing returns to scale, the last expression in equation 33 implies thatMPi <APi for all i.

3.2. Implications of Various Types of Returns to Scale. If a technology exhibits constant returns to scale then the firm can expand operations proportionately. If the firm can produce 5 units of output with a profit per unit of $20, then by dou- bling the inputs and producing 10 units the firm will have a profit of $40. Thus the firm can always make more profits by expanding. If the firm has increasing returns to scale, then by doubling inputs it will have more than double the output. Thus if it makes $20 with 5 units it will make more than $40 with 10 units etc. This assumes in all cases that the firm is increasing inputs in a proportional manner. If the firm can reduce the cost of an increased output by increasing inputs in a manner that is not proportional to the original inputs, then its increased economic returns may be larger than that implied by its scale coefficient.

3.3. Multiproduct Returns to Scale. Most firms do not produce a single product, but rather, a number of related products. For example it is common for farms to produce two or more crops, such as corn and soybeans, barley and alfalfa hay, wheat and dry beans, etc. A flour miller may produce several types of flour and a retailer such as Walmart carries a large number of products. A firm that produces several different products is called a multiproduct firm. Consider the production possibility set of the multi-product firm

T = {(x, y) : x ∈ Rn+, y ∈ Rm :+ x can produce y}

where y and x are vectors of outputs and inputs, respectively. We define the mul- tiproduct elasticity of scale by

m = sup{r : there exists a δ > 1 such that (λx, λ yr) ∈ T for 1 ≤ λ δ} (35)

For our purposes we can regard the sup as a maximum. The constant of pro- portion is greater than or equal to 1. This gives the maximumproportional growth rate of outputs along a ray, as all inputs are expanded proportionally [2]. The idea is that we expand inputs by some proportion and see how much outputs can pro- portionately expand and still be in the production set. If r = 1, then we have con- stant returns to scale. If r < 1 then we have decreasing returns, and if r > 1, we have increasing returns to scale.

PRODUCTION FUNCTIONS 21

4. RATE OF TECHNICAL SUBSTITUTION

The rate of technical substitution (RTS) measures the extent to which one input substitutes for another input, holding all other inputs constant. The rate of techni- cal substitution is also called the marginal technical rate of substitution or just the marginal rate of substitution.

4.1. Definition of RTS. Consider a production function given by

y = f(x1, x2 . . . , xn) (36) If the implicit function theorem holds then

φ(y, x1, x 2, . . . , xn) = y f(x1, x2 . . . , xn) = 0 (37) is continuously differentiable and the Jacobian matrix has rank 1. i.e.,

∂φ

∂xj =

∂f

∂xj 6= 0 (38)

Given that the implicit function theorem holds, we can solve equation 38 for xk as a function of y and the other x’s i.e.

x∗k = ψk(x1, x2, . . . , xk−1, xk+1, . . . , y ) (39) Thus it will be true that

φ(y, x1, x2, . . . , xk−1, x∗k, xk+1, . . . , xn ) ≡ 0 (40) or that

y f(x1, x2, . . . , xk−1, x∗k, xk+1, . . . , xn ) (41) Differentiating the identity in equation 41 with respect to xj will give

0 = ∂f

∂xj +

∂f

∂xk

∂xk ∂xj

(42)

or

∂xk ∂xj

= − ∂f

∂xj ∂f ∂xk

= RTS = MRS (43)

Or we can obtain this directly as

∂φ(y, x1, x2, . . . , xk−1, ψk, xk+1, . . . , xn) ∂xk

∂ψk ∂xj

= −∂φ(y, x1, x2, . . . , xk−1, ψk, xk+1, . . . , xn)

∂xj

∂φ ∂xk

∂xk ∂xj

= − ∂φ ∂xj

∂xk ∂xj

= − ∂f

∂xj ∂f ∂xk

= MRS

(44) The above expression represents the slope of the projection of the boundary of

the input requirement (or level/contour) set V(y) into xk, xj space. With two in- puts, this is, of course, just the slope of the boundary. Its slope is negative. It is

22 PRODUCTION FUNCTIONS

convex because V(y) is convex. Because it is convex, there will be a diminishing rate of technical substitution. Figure 20 shows the rate of technical substitution.

FIGURE 20. Rate of Technical Substitution

x1

x2

RTS = ¶x2 €€€€€€€€€ ¶x1

Dx2

Dx1

4.2. Example Computation of RTS.

4.2.1. Cobb-Douglas. Consider the following Cobb-Douglas production function

y = 5x 1 3 1 x

1 4 2 (45)

The partial derivative of y with respect to x1 is

∂y

∂x1 =

5 3 x

−2 3

1 x 1 4 2 (46)

The partial derivative of y with respect to x2 is

∂y

∂x2 =

5 4 x

1 3 1 x

−3 4

2 (47)

The rate of technical substitution is

∂x2 ∂x1

= − ∂f

∂x1 ∂f ∂x2

= − 5 3 x

−2 3

1 x 1 4 2

5 4 x

1 3 1 x

−3 4

2

= − 5 3 5 4

x2 x1

= − 4 3 x2 x1

(48)

PRODUCTION FUNCTIONS 23

4.2.2. CES. Consider the following CES production function

y = 5 [ 0.6x−21 + 0.2x

−2 2

]−1 2 (49)

The partial derivative of y with respect to x1 is

∂y

∂x1 =

−5 2 [ 0.6x−21 + 0.2x

−2 2

]−3 2 (−6/5)x−31 (50)

The partial derivative of y with respect to x2 is

∂y

∂x2 =

−5 2 [ 0.6x−21 + 0.2x

−2 2

]−3 2 (−2/5)x−32 (51)

The rate of technical substitution is

∂x2 ∂x1

= − ∂f∂x1

∂f ∂x2

= − −5 2

[ 0.6x−21 + 0.2x

−2 2

]−3 2 (−6/5)x−31

−5 2

[ 0.6x−21 + 0.2x

−2 2

]−3 2 (−2/5)x−32

= − 2 5 6 5

x−31 x−32

= − 3 x 3 2

x31

(52)

5. ELASTICITY OF SUBSTITUTION

The elasticity of substitution is a unitlessmeasure of how various inputs substi- tute for each other. For example, how does capital substitute for labor, how does low skilled labor substitute for high skilled labor, how do pesticides substitute for tillage, how does ethanol substitute for gasoline. The elasticity of substituion at- tempts to measure the curvature of the lower boundary of the input requirement set. The most commonly used measure of the elasticity of substitution based on the slope of an isoquant is due toHicks [13, p. 117, 244-245], [17, p.330]. He defines the elasticity of substitution between x2 and x1 as follows.

σ = d(x2/x1) d(f1/f2)

· (f1/f2) (x2/x1)

(53)

This is the percentage change in the input ratio induced by a one percent change in the RTS.

5.1. Geometric Intuition. We can better understand this definition by an appeal to geometry. Consider figure 21. First consider the factor ratio of x2 to x1. Along the ray labeled a, the ratio x2x1 is given by the tangent of the angle θ. Along the ray labeled b, the ratio x2x1 is given by the tangent of the angle φ. For example at point

d, x 0 2

x01 = tan φ along the ray b.

Now consider the ratio of the slopes of the input requirement boundary at two different points. Figures 22 and 23 show these slopes. The slope of the curve at point c is equal to minus the tangent of the angle γ in figure 22. The slope of the curve at point d is equal to minus the tangent of the angle δ in figure 23.

24 PRODUCTION FUNCTIONS

FIGURE 21. Elasticity of Substitution (Factor Ratios)

Consider the right triangle formed by drawing a vertical line from the inter- section of the two tangent lines and the x1 axis. The base is the x1 axis and the hypotenuse is the tangent line between the other two sides. This is shown in fig- ure 24. Angle αmeasures the third angle in the smaller triangle. Angle α + angle γ equal 90 degrees. In figure 25, the angle between the vertical line and the tan- gent at point d is represented by α + β. Angle β + angle α+ angle δ also equals 90 degrees, so that α + γ = α +β + δ or γ = β + δ. This then means that β, the angle between the two tangent lines is equal to γ - δ.

FIGURE 22. Elasticity of Substitution (Angle of Tangent at Point c)

PRODUCTION FUNCTIONS 25

FIGURE 23. Elasticity of Substitution (Angle of Tangent at Point d)

FIGURE 24. Elasticity of Substitution (Angle Tangent at Point d)

If we combine the information in figures 22, 23, 24 and 25 into figure 26, we can measure the elasticity of substitution. Remember that it is given by

σ = d(x2/x1) d(f1/f2)

· (f1/f2) (x2/x1)

The change in x2x1 is given by the angle ξ which is θ - φ. The marginal rate of technical substitution is given by the slope of the boundary of V(y). At point d the slope is given by δ while at c it is given by γ. The change in this slope is β. The rate of technical substitution is computed as

26 PRODUCTION FUNCTIONS

FIGURE 25. Elasticity of SubstitutionMeasuring Slope Change

RTS = ∂x2 ∂x1

= − ∂f

∂x1 ∂f ∂x2

Therefore we can say that the ratio of the change in the input ratio and the change in the slope of the lower the boundary of V(y) is given by

σ ξ β

PRODUCTION FUNCTIONS 27

FIGURE 26. Geometric Interpretation of the Elasticity of Substitution

5.2. Algebraic Expression for σ12. In the case of two inputs, the elasticity of sub- stitution is given by

σ12 = −f1 f2 (x1 f1 + x2 f2)

x1 x2 (f11 f22 − 2f12 f1 f2 + f22 f21 ) (54)

5.3. Derivation of the Algebraic Expression for σ12. By definition

σ12 = d(x2/x1) d(f1/f2)

· (f1/f2) (x2/x1)

(55)

First compute d (

x2 x1

) . This is a differential. Computing this differential we

obtain

d

( x2 x1

) =

x1 dx2 − x2 dx1 x21

(56)

Then compute the differential d (

∂ f ∂ x1 ∂ f ∂ x2

) .

d

( ∂ f ∂ x1 ∂ f ∂ x2

) = d

( f1 f2

) =

( f1f2 )

∂x1 dx1 +

( f1f2 )

∂x2 dx2 (57)

With the level of y fixed as along the boundary of V(y), we have via the implicit function theorem or equation 43

28 PRODUCTION FUNCTIONS

∂x2 ∂x1

= − ∂f∂x1

∂f ∂x2

(58)

Rearranging equation 58 (and being a bit sloppy with differentials and partial derivatives) we can conclude

dx2 dx1

= −f1 f2

dx2 = ( − f1 f2

) dx1

(59)

and

dx1 = ( − f2 f1

) dx2 (60)

This then implies that

d

( x2 x1

) = x1

( − f1 f2

) dx1 − x2 dx1 x21

=

( x1

( − f1 f2

) − x2

) dx1

x21

(61)

and that

d

( ∂ f ∂ x1 ∂ f ∂ x2

) = d

( f1 f2

) =

( f1f2 )

∂x1 dx1 +

( f1f2 )

∂x2

( − f1 f2

) dx1

=

( ( f1f2 )

∂x1 +

( f1f2 )

∂x2

( − f1 f2

)) dx1

(62)

Now compute ∂

( f1 f2

)

∂x2 .

(

f1 f2

)

∂x2 =

f2 f12 − f1 f22 f22

(63)

and ∂

( f1 f2

)

∂x1

(

f1 f2

)

∂x1 =

f2 f11 − f1 f21 f22

(64)

Now replace ∂

( f1 f2

)

∂x2 and

∂ (

f1 f2

)

∂x1 in equation 62with their equivalent expressions

from equations 63 and 64 in equation 65.

d

( ∂ f ∂ x1 ∂ f ∂ x2

) = d

( f1 f2

) = ( f2f11 − f1f21

f22 +

f2 f12 − f1f22 f22

( −f1 f2

)) dx1 (65)

PRODUCTION FUNCTIONS 29

We now have all the pieces we need to compute σ12 by substituing from equa- tions 61 and 65 into equation 55.

σ12 = d(x2/x1) d(f1/f2)

· (f1/f2) (x2/x1)

=

( x1

( − f1 f2

) − x2

) dx1

x21( f2 f11 − f1 f21

f22 + f2 f12 − f1 f22

f22

( − f1 f2

) ) dx1

· (f1/f2) (x2/x1)

=

( x1

( − f1 f2

) − x2

) f1 x1

x21 f2 x2

( f2 f11 − f1 f21

f22 + f2 f12 − f1 f22

f22

( − f1 f2

) )

= (− f1 x1 − f2 x2 )

f2 f1

x1 x2 f2

( f2 f11 − f1 f21

f22 + f2 f12 − f1 f22

f22

( − f1 f2

) )

= − f1 ( f1 x1 + f2 x2 )

x1 x2 f22

( f2 f11 − f1 f21

f22 + f2 f12 − f1 f22

f22

( − f1 f2

) )

(66)

Consider the expression in parentheses in the denominator of equation 66. We can rearrange and simplify it as follows

( f2f11 − f1f21

f22 +

f2 f12 − f1f22 f22

( −f1 f2

)) =

f22 f11 − f1f2f21 f32

+ −f1f2f12 + f21 f22

f32

= f22 f11 − f1f2f21 − f1f2f12 + f21 f22

f32

= f22 f11 − 2f1f2f12 + f21 f22

f32 by Young’s Theorem

(67) Now substitute equation 67 into equation 66

σ12 = − f1 ( f1 x1 + f2 x2 )

x1 x2 f22

( f22 f11−2f1f2f12+f21 f22

f32

)

= − f1 f2 ( f1 x1 + f2 x2 )

x1 x2 f22 (f11 f22 − 2 f12 f1 f2 + f22 f21 )

(68)

5.4. Matrix Representation of σ12. In matrix notation σ12 is given by

σ12 = (x1 f1 + f2 x2)

x1 x2

F12 F

(69)

where

F =

∣∣∣∣∣∣

0 f1 f2 f1 f11 f12 f2 f21 f22

∣∣∣∣∣∣ = det (HB) (70)

30 PRODUCTION FUNCTIONS

whereHB is the bordered Hessian of the production function. F12 is the cofactor of f12 in the matrix F, i.e.,

f12 = (−1)5 ∣∣∣∣ 0 f1 f2 f21

∣∣∣∣ = (−1) (−f1 f2) = f1 f2 (71)

The determinant of F computed by expanding along the first row is given by

F =

∣∣∣∣∣∣

0 f1 f2 f1 f11 f12 f2 f21 f22

∣∣∣∣∣∣ = 0 · (f11f22 − f212)− f1 · (f1f22 − f2f12) + f2 · (f1 f21 − f2f11)

= −f21 f22 + 2f1 f2 f12 − f22 f11 (72)

Substitute equations 71 and 70 into the last expression of equation 69 to find F12F

F12 F

= −f1 f2

f21 f22 − 2f1 f2 f12 + f22 f11 (73)

Now substitute F12F from equation 73 into equation 69.

σ12 = (x1 f1 + f2 x2)

x1 x2

f1 f2 f21 f22 − 2f1 f2 f12 + f22 f11

(74)

Then compare with equation 68

σ12 = − f1 f2 ( f1 x1 + f2 x2 )

x1 x2 f22 (f11 f22 − 2 f12 f1 f2 + f22 f21 ) Note that from equation 74, σ12 is symmetric. We can also show that for a quasi-

concave production function that F12F is always negative so that the elasticity of substitution is also negative. This also follows from the fact that quasi-concave functions have convex level sets.

6. HOMGENEITY

6.1. Definition of Homogeneity. A function is homogeneous of degree k in x if

f(λx, z) = λkf(x, z), λ > 0, x Rn+, z Rm+ (75)

6.2. Some Properties of Homogeneous Functions. If f(x1, . . ., xn) is homoge- neous of degree k in x, then with suitable restrictions on the function f in each case,

a: ∂f

∂xi is homogeneous of degree k-1, i = 1, 2, . . . , n (76)

b:

f(x) = xki f ( x1 xi , x2 xi , . . . , 1, x i+1

xi

, . . . , xn xi

) ,

xi > 0, i = 1, 2, . . . , n. (77)

PRODUCTION FUNCTIONS 31

6.3. An Euler Equation for Homogeneous Functions. f is homogeneous of de- gree k if and only if

n∑

i=1

∂f

∂xi xi = k f(x) (78)

Proof of Euler equation. Define the following functions g(λ, x) and h(x).

g(λ, x) = f(λ x) − λk f(x) ≡ 0 (79a)

h(x) = n∑

i=1

∂f

∂xi xi − k f(x) (79b)

Now differentiate g with respect to λ using the chain rule

∂g(λ, x) ∂λ

= n∑

i=1

∂f(λx) (λxi)

(λxi) ∂ λ

k−1 f(x)

= n∑

i=1

∂f(λx) (λxi)

xi − k−1 f(x) (80)

Now multiply both sides of 80 by λ to obtain

λ ∂g(λ, x) ∂λ

= λ [ Σni=1

∂f(λx) ∂λxi

xi − k−1 f(x) ]

= [ Σni=1

∂f(λx) ∂λxi

λxi − k f(x) ]

= [ Σni=1

∂f(λx) ∂λxi

λxi − k f(λ x) ]

= h(λx)

(81)

The last two steps follows because f(λx)= λk f(x). By assumption g(λ, x) ≡ 0 so ∂ g(λ, x)

∂ λ ≡ 0. Now set λ = 1 in the equation 81 obtain

0 = h(λx) |λ=1 = [ Σni=1

∂f(x) ∂xi

xi − k f(x) ]

⇒ Σni=1 ∂f(x) ∂xi

xi = k f(x) (82)



A more complete disussion of homogeneity is contained in Eichorn [6, p. 69- 77],Sydsaeter [20, p. 125-131] and Sydsaeter and Hammond. [21, p. 432-442].

6.4. Geometry ofHomogeneous Functions. There are a number ofways to graph- ically characterize a homothetic function. The first is that the points on the graph above a ray form a smooth curve in n + 1 space. The second considers the shape of level sets and the third the slope of level sets along a given ray.

32 PRODUCTION FUNCTIONS

FIGURE 27. Graph of a PLH Function

x1

x2

f@x1,x2D

6.4.1. Graph of a homogeneous function as lines through the origin. Consider a positive linear homogeneous production (PLH) function with two inputs. The graph of the function is shown in figure 27.

In figure 28 we plot the points on the surface of the graph which lie above the ray running through the point (x10, x20, 0). As can be clearly seen, the graph above this ray is a straight line. The height of the function at this point is c. Points along the ray through the point (x10, x20, 0) are generated as λ (x10, x20, 0). The height of the function above any point along the ray is then given by λk c, where k is the degree of homogeneity of the function. Figure 29 shows the straight line as part of the graph.

Thus the graph of f consists of curves of the form z = λk c, above each ray in the hyperplane generated by λ (x10,x20,0). Thus if we know one point on each ray we know the function. If k = 1 then z = λ c and the graph is generated by straight lines through the origin. This is the case for the function in this example as demonstated in figure 30. Figures 31, 32, 33, and 34 show the same relationships when the function is homogeneous of degree 0 < k < 1.

6.4.2. Shape of level sets. Suppose the and are both elements of the lower bound- ary of a particular level set for a function which is homogeneous of degree k. Then we know that

f(x̃) = f(z̃)

f(λ x̃) = λkf(x̃)

f(λ z̃) = λkf(z̃)

f(λ x̃) = f(λ z̃)

(83)

PRODUCTION FUNCTIONS 33

FIGURE 28. Graph of PLH Function as a Straight Line Above Any Ray

x1

x2

f@x1,x2D

Hx10,x20,0L

FIGURE 29. Graph of a PLH Function Along with Function Value Above a Ray

x1

x2

f@x1,x2D

Thus the two points λ x̃ and λ z̃ are on the same level set. In figure 35we can see how to construct the level set containing λ x̃ and λ z̃ and the one containing and z̃. If we scale points a and b on the intial lower isoquant by 2, we arrive on the same level set at y = 108.15. Given that this is less than twice 60, this technology exhibits decreasing returns to scale. We can construct all other level sets from this initial one in a similar manner. The conclusion is that if the function is homogeneous

34 PRODUCTION FUNCTIONS

FIGURE 30. Graph of a PLH function Along with Function Values Above Rays

x1

x2

f@x1,x2D

FIGURE 31. Graph of a General Homogeneous Function

x1

x2

f@x1,x2D

and we know one level set, we can construct all the others by a radial expansion or contraction of the set.

6.4.3. Description using slopes along rays. Remember from equation 76 that the de- rivative function homogeneous of degree k is homogenous of degree k−1. Specif- ically

PRODUCTION FUNCTIONS 35

FIGURE 32. Graph of General HomogeneousFunction as a Curve Above Any Ray

x1

x2

f@x1,x2D

Hx10,x20,0L c

FIGURE 33. Graph of a general homogeneous function alongwith function value above a ray

x1

x2

f@x1,x2D

∂f

∂xi (λx1, λx2, · · · , λxn) = λk−1

∂f

∂xi (λx1, λx2, · · · , λxn) (84)

This is true for all i. Define the gradient of f(x1, x2, . . ., xn ) as the vector of first derivatives, i.e.,

36 PRODUCTION FUNCTIONS

FIGURE 34. Graph of a general homogeneous function alongwith function values above rays

x1

x2

f@x1,x2D

FIGURE 35. Constructing One Level Set from an Initial One for Homogeneous Functions

x1

x2

a

b

c

d

y=60

y=108.15

f(x) =

 

∂ f(x) ∂ x1 ∂ f(x) ∂ x2 ...

∂ f(x) ∂ xn

 

(85)

PRODUCTION FUNCTIONS 37

We can then write for all λ > 0 and i = 1,2,. . ., n

f(λx1, λx2, · · · , λxn) = λk−1 ∇f(λx1, λx2, · · · , λxn) (86)

So along a given ray the gradients are parallel. Gradients are orthogonal (or at right angles) to level sets. So along a given ray if the gradients are all parallel, then the level sets are parallel along the ray.

FIGURE 36. Level Sets for a Function which is Homogeneous of Degree One

5 10 15 20 25 30

5

10

15

20

25

30

6.5. Returns to Scale and a Generalized Euler Equation.

6.5.1. Definition of elasticity of scale.

(λ, x) = ln f(λx) ln λ

= ∂ f(λx) ∂ λ

λ

f(λx)

= n∑

i=1

∂f(λx) ∂λxi

(λxi) ∂λ

λ

f(λx)

= n∑

i=1

∂f(λx) ∂λxi

λxi f(λx)

(87)

Now evaluate at λ = 1 to obtain

38 PRODUCTION FUNCTIONS

(1, x) = n∑

i=1

∂f(x) ∂xi

xi f(x)

⇒ n∑

i=1

∂f(x) ∂xi

xi = (1, x) f(x)

(88)

The sum of the partial derivatives of f(x) multiplied by the levels of the x’s is equal to (1,x) multiplied by the value of the function f(x). Remember from equation 24 the output elasticity is given by

j = ∂f(x) ∂xj

xj f(x)

, f(x) 6= 0. (89)

Replace 1 in equation 88 with λ and rewrite it as follows

(λ, x) = n∑

i=1

∂f(λx) ∂λxi

λxi f(λx)

(90)

Now substitute equation 89 in equation 90 to obtain

(λx) = 1(λx) + 2(λx) + · · · + n(λx) (91) If λ =1 , then we obtain the result we previously obtained in equation 33.

(1, x) = 1(x) + 2(x) + · · · + n(x) (92) And we usually write (x) for (1,x).

6.5.2. Properties of returns to scale. Remember from equations 26, 27, and 28 that 1: (x) = 1⇒ constant returns to scale (CRS) 2: (x) < 1⇒ decreasing returns to scale (DRS) 3: (x) > 1⇒ increasing returns to scale (IRS)

6.5.3. General nature of returns to scale. In general (x) is a function depending on x. With different levels of xj , (x) will change.

6.6. Returns to Scale for Homogeneous Technologies. We know from equation 88 that returns to scale are given by

(1, x) = n∑

i=1

∂f(x) ∂xi

xi f(x)

⇒ n∑

i=1

∂f(x) ∂xi

xi = (1, x) f(x)

(93)

For a homogeneous function (equation 78) the Euler equation implies

n∑

i=1

∂f

∂xi xi = r f(x) (94)

where r is the degree of homogeneity. Combining the results in equations 93 and 94 we obtain

 (1, x) = r (95)

PRODUCTION FUNCTIONS 39

Alternatively it can be shown directly by using the definition of homogeneity and the definition of . First note that if a function is homogeneous f(λx) = λr f(x). Now write the definition of the elasticity of scale, substitute λr f(x) for f(λx) and then take the derivative.

(λ, x) = ln f(λx) ln λ

= ∂ f(λx) ∂ λ

λ

f(λx)

= [λr f (x)]

∂ λ

λ

λr f (x)

= r λr−1 f(x) · λ λrf(x)

= r

(96)

This means that the isoquants spread or bunch at a constant rate and are paral- lel. Figure 37 shows the isoquants for a CES production function that is homoge- neous of degree one.

FIGURE 37. Level Sets for CES Production Function that is Ho- mogeneous of Degree 1

5 10 15 20 25 30

5

10

15

20

25

30

Figures 38 and 39 show the isoquants for production functions that are homo- geneous of a degree less than one.

Figure 40 shows the isoquant for a production function that is homogeneous of a degree greater than one.

40 PRODUCTION FUNCTIONS

FIGURE 38. Level Sets for Production Function that is Homoge- neous of Degree Less than 1

10 20 30 40 50 60

10

20

30

40

50

60

7. HOMOTHETIC FUNCTIONS

7.1. Definition of a Homothetic Function. A function f(x, z) is homothetic in x if it can be written

y = f(x, z) = H(x, z) = F (φ(x, z), z) (97)

where z ∈ Rm+ and F is continuously differentiable to the second degree (C2), finite, non-negative and non-decreasing. Furthermore

(a) lim µ→∞

F (µ, z) = ∞

(b) F (0, z) = 0

(c) φ(x, z) is positively linear homogeneous (PLH) in x

7.2. AlternativeDefinition of aHomothetic Function. Let y = f(x)= F(φ(x)) where F has the same properties as in the original definition. Then consider the function h(y) = φ(x) where φ(x) is PLH and h( ) = F−1( ) and it is assumed that F−1 exists. The idea is that a transformation of y is homogeneous. Specifically the function in equation 98 is a homothetic function.

h(y) = φ(x), where φ(x) is positively linear homogeneous (PLH) in x (98)

7.3. Some Properties of Homothetic Functions.

PRODUCTION FUNCTIONS 41

FIGURE 39. Level Sets for CES Production Function that is Ho- mogeneous of Degree Less than 1

20 40 60 80 100 120 140

20

40

60

80

100

120

140

7.3.1. Rate of technical substitution for homothetic functions. Define y as follows

y = h(x) = F (φ(x)) (99)

Using the implicit function theorem (equation 43 ), find ∂ xi∂ xj .

∂xi ∂xj

= − ∂F∂xj

∂F ∂xi

= − F ′ ∂φ

∂xj

F ′ ∂φ ∂xi

= − ∂φ ∂xj ∂φ ∂xi

(100)

φ is homogeneous of degree 1 so its derivatives are homogeneous of degree zero by the results in section 6.2. The ratio of these derivatives is also homogeneous of degree zero so that the rate of technical substitution does not change as x changes; i.e. the isoquants are parallel because RTS is constant along a ray from the ori- gin just as with positively linear homogeneous functions. Lau [14] also shows that this condition is sufficient for a production function to be homothetic. This implies that we can obtain the isoquant for any y ≥ 0 by a radical expansion of the unit

42 PRODUCTION FUNCTIONS

FIGURE 40. Level Sets for Production Function that is Homoge- neous of Degree Greater than 1

2.5 5 7.5 10 12.5 15 17.5 20

2.5

5

7.5

10

12.5

15

17.5

20

isoquant in the fixed ratio h(y)/h(1). Figure 41 shows the isoquants for a Cobb- Douglas production function that is homogeneous of degree one.

y = ( 5x.251 x

.75 2

) (101)

Note that the level sets are evenly spaced {20, 40, 60, 80, 100, 120}. Also notice that along a ray through the origin the tangent lines are parallel. We can construct the level set for all levels of output by expanding the level set for any one level of output inward or outward.

Now let’s consider a transformation of a linear homogeneous CES production function to create a homothetic function.

f(x) = [ 0.3x−21 + 0.1x

−2 2

]−1 2

y = F [f(x)] = f(x)2 + f(x)

= [ [

0.3x−21 + 0.1x −2 2

]−1 2

]2 + [ 0.3x−21 + 0.1x

−2 2

]−1 2

(102)

The function is shown in figure 42 while the level sets are contained in figure 43. The function is not concave and has increasing returns to scale. The level sets are convex and parallel but become closer together for equal increases in output.

PRODUCTION FUNCTIONS 43

FIGURE 41. Level Sets for Cobb-Douglas Prodcution Function that is Homogeneous of Degree 1

5 10 15 20 25 30

5

10

15

20

25

30

The spacing depends on the level of output and is not a fixed rate of increase or decrease.

FIGURE 42. Homothetic Production Function

5

10

15

20

x1

5

10

15

20

x2

0

250

500

750

1000

fHx1,x2L

7.3.2. Elasticity of scale for homothetic functions. Let y = f(x) = F(φ(x)) or h(y) = φ(x) as in equation 98.

44 PRODUCTION FUNCTIONS

FIGURE 43. Level Sets for Homothetic Function with Increasing Returns to Scale

5 10 15 20 25 30 35 40

5

10

15

20

25

30

35

40

h(y) = φ(x), where φ(x) is positively linear homogeneous (PLH) in x

Differentiate both sides of the identity with respect to xi

h′(y) ∂y

∂xi =

∂φ

∂xi ⇒ ∂y

∂xi =

∂φ ∂xi

h′(y) (103)

Then write the definition of  and substitute

PRODUCTION FUNCTIONS 45

(1, x) = n∑

i=1

∂F (φ(x)) ∂xi

· xi F (φ(x))

= n∑

i=1

∂y

∂xi · xi y

= n∑

i=1

∂φ ∂xi

h′(y) xi y

= h (y) h′(y) y

by Eulers Theorem

= ̃(y)

(104)

so  is a function only of y and not a function of x. This is clear from figure 43.

7.4. A Generalized Euler Equation for Homothetic Functions. Let y = f(x) = F(φ(x)) or alternatively let h(y) = φ(x). Then from equation 104 we know that

 = h(y) h′(y)y

(105)

where

n∑

i=1

∂f

∂xi

xi y

= h(y) h′(y) y = (y) (106)

This implies then that

n∑

i=1

∂f

∂xi xi = (y) y (107)

So  depends on y and not on the mix of the x’s. With homogeneous functions  is a constant not depending on y.

7.5. Multiplication property of homothetic functions. If h represents a homo- thetic function then

Theorem 1.

h(y) = xi · V (

xi

) = xi V (ψ) (108)

where x̃ is (x1, x2, xi−1, xi+1, . . . , xn) and

ψj = xj xi , j 6= i (109)

Here h is a transform like F, in fact h(z) = F−1(z) and V is a nondecreasing concave function.

Proof. Let y = F(φ(x)) where φ(x) is PLH. Then

46 PRODUCTION FUNCTIONS

y = F (φ(x))

= F (xi · φ[ψ1, ψ2 . . . 1 . . . ψn]) (by PLH) ⇒ F−1(y) = xi · φ[ψ1, ψ2 . . . 1 . . . ψn])

xi = F−1(y)

φ(ψ1, ψ2 ... 1 ... ψn)

= h (y) V (ψ)

h(y) = xi V (ψ)

(110)



PRODUCTION FUNCTIONS 47

REFERENCES

[1] Arrow, K. J., H. B. Chenery, S. Minhas, and R. M. Solow. ”Capital-Labor Substitution and Economic Efficiency.” Review of Economics and Statistics XLII(1961):225-50.

[2] Baumol, W. J., J. C. Panzar, and R. D. Willig. Contestable Markets and the Theory of Industry Structure. NewYork: Harcourt Brace Jovanovich Inc., 1982.

[3] Chambers, R. G. Applied Production Analysis. Cambridge: Cambridge University Press, 1988. [4] Cobb, C. W., and P. H. Douglas. ”A Theory of Production.” American Economic Review 18

((1928)(Suppl.):139-165. [5] Douglas, P. H. ”Are There Laws of Production.” American Economic Review 38(1948):1-41. [6] Eichorn. W. Functional Equations in Economics. Reading, Massachusetts: Addison Wesley, 1978. [7] Fare, R. Fundamentals of Production Theory. New York: Springer-Verlag, 1988. [8] Fare, R., L. Jansson, and C. A. Knox Lovell. ”Modeling Scale Economies with Ray Homothetic

Production Functions.” Review of Economics and Statistics 67(1985): 624-629. [9] Fare, R. and D. Primont. Multi-output Production and Duality: Theory and Applications. Boston:

Kluwer Academic Publishers, 1995. [10] Ferguson, C. E. The Neoclassical Theory of Production and Distribution. Cambridge: Cambridge Uni-

versity Press, 1971. [11] Frisch, R. Theory of Production. Dordrecht: D. Reidel Publishing, 1965. [12] Fuss, M. and D. McFadden. Production Economics: A Dual Approach to Theory and Application. Ams-

terdam: North Holland, 1978. [13] Hicks, John R. The Theory of Wages. London: Macmillan Company, Ltd, 1932. [14] Lau, Lawrence J. ”Duality and the Structure of Utility Functions.” Journal of Economic Theory

1:(1969):374-396. [15] McElroy, F. W. ”Returns to Scale, Euler’s Theorem, and the Form of Production Functions.” Econo-

metrica 37(1969):275-279. [16] Panzar, J. C., and R. Willig. ”Economies of Scale in Multi-Output Production.” Quarterly Journal of

Economics 91 (1977): 481-493. [17] Robinson, Joan. The Economics of Imperfect Competition. London: Macmillan Company, Ltd, 1933. [18] Shephard, R. W. Cost and Production Functions Princeton New Jersey: Princeton University Press,

1953. [19] Shephard, R. W. Theory of Cost and Production Functions. Princeton New Jersey: Princeton Univer-

sity Press, 1970. [20] Sydsaeter K. Topics in Mathematical Analysis. New York: Academic Press, 1981. [21] Sydsaeter, Knut and Peter Hammond. Essential Mathematics for Economic Analysis, Harlow, Eng-

land: Pearson Education Limited, 2002. [22] Varian, H. R. Microeconomic Analysis. 3rd Edition. New York: Norton, 1992.

Docsity is not optimized for the browser you're using. In order to have a better experience please switch to Google Chrome, Firefox, Internet Explorer 9+ or Safari! Download Google Chrome