1

Homogenous and Homothetic Functions

Reading: [Simon], Chapter 20, p. 483-504.

1.1

Homogenous Functions

Denition 1 A real valued function f (x1 , ..., xn ) is homogenous of degree k if for all t > 0 f (tx1 , ..., txn ) = tk f (x1 , ..., xn ). (1) Examples. 3 a) A monomial of degree 6 f (x1 , x2 , x3 ) = x2 1 x2 x3 is a homogenous function of degree 6: (tx1 ) (tx2 ) (tx3 ) =
2 3

f (tx1 , tx2 , tx3 ) = 3 2 2 3 t x1 t x2 tx3 = t6 x2 1 x2 x3

= t6 f (tx1 , tx2 , tx3 ).

(b) The function f (x1 , x2 ) = 3/2: f (tx1 , tx2 ) =

3 x3 1 + x2 is a homogenous function of degree

(tx1 )3 + (tx2 )3 =

3/2 3 3 t3 (x3 x3 1 + x2 ) = t 1 + x2 .

(c) A linear function f (x1 , ..., xn ) = n k=1 ak xk is homogenous of degree 1 (prove it). (d) A quadratic form f (x1 , ..., xn ) = n k=1 aij xi xj is homogenous of degree 2 (prove it). (e) If f is a homogenous function of degree k and g is a homogenous function of degree l then f g is homogenous of degree k + l and f is homogenous g of degree k l (prove it). (f) If f and g are homogenous functions of same degree k then f + g is homogenous of degree k too (prove it). (g) Only homogenous function of degree k of one variable is f (x) = axk (prove it). (h) Only homogenous function of degree 0 of one variable is constant function (prove it). (i) However there exist nonconstant homogenous functions of degree 0: is homogenous of degree 0 (prove it). the function f (x, y ) = x y 1.1.1 Economical Examples

Constant return to scale - production function which is homogenous of degree k = 1. Increasing return to scale - production function which is homogenous of degree k > 1. Decreasing return to scale - production function which is homogenous of degree k < 1. 1

Cobb-Douglas function
n 1 q (x1 , ..., xn ) = Ax 1 ... xn

is homogenous of degree k = 1 + ... + n . q p p Constant elasticity of substitution (CES) function A(a1 xp 1 + a2 x2 ) is homogenous of degree q . Demand functions that are derived from utility functions are homogenous of degree 0: if the prices (p1 , ..., pn ) and income I change say 10 times all together (as it was in 1961), then the demand will not change. More precisely, let U (x1 , ..., xn ) be the utility function, p = (p1 , ..., pn ) be the price vector, x = (x1 , ..., xn ) be a consumption bundle and let p x = p1 x1 + ... + pn xn I be the budget constraint. The demand function x = D(p1 , ..., pn , I ) associates to each price vector p and income level I the consumption bundle x which is a solution of the following constraint maximization problem Maximize U (x) subject of p1 x1 + ... + pn xn I . So we claim that demand function x = D(p1 , ..., pn , I ) is homogenous of degree 0, that is D(tp1 , ..., tpn , tI ) = t0 D(p1 , ..., pn , I ) = D(p1 , ..., pn , I ). Indeed, the bundle D(tp1 , ..., tpn , tI ) is a solution of the following constraint maximization problem Maximize U (x) subject of tp1 x1 + ... + tpn xn tI . As we see this is the same problem, thus D(tp1 , ..., tpn , tI ) = t0 D(p1 , ..., pn , I ) = D(p1 , ..., pn , I ). Prot and cost functions that are derived from production function are homogenous too. 1.1.2 Properties of Homogenous Functions

Theorem 1 If f (x1 , ..., xn ) is homogenous of degree k then its rst order partial derivatives are homogenous of degree k 1. Proof. Dierentiate f (tx1 , ..., txn ) = tk f (x1 , ..., xn ) by xi . Theorem 2 Level sets of a homogenous function are radial expansions of one another, that is for arbitrary x = (x1 , ..., xn ), y = (y1 , ..., yn ), t > 0 if x and y are on the same level set then their radial expansions tx and ty are on the same level set too. 2

Proof. We must show that f (x) = f (y ) implies f (tx) = f (ty ), indeed f (tx) = tk f (x) = tk f (y ) = f (ty ). Theorem 3 The tangent lines of the level curves of a homogenous function f (x, y ) have constant slopes along each ray from the origin. Proof. The slope of the tangent line of the level curve at (x0 , y0 ) is
f x k = f

(x0 , y0 ) (x0 , y0 ) y

and the slope of the tangent line of the level curve at (tx0 , ty0 ) is
f x kt = f

tk1 f (tx0 , ty0 ) (x0 , y0 ) x = k1 f =k ( tx , ty ) ( x , y ) t 0 0 0 0 y y

1.1.3

Calculus Criterion for Homogeneity

Observation. Let f (x) = axk be a homogenous function of one variable of degree k. Then x f (x) = kf (x) (check it!). The following theorem generalizes this fact for functions of several variables. Theorem 4 (Eulers theorem)Let f (x1 , ..., xn ) be a function that is homogenous of degree k . Then x1 f f (x) + ... + xn (x) = kf (x), x1 xn

or, in gradient notation, x f (x) = kf (x). Proof. Dierentiate f (tx1 , ..., txn ) = tk f (x1 , ..., xn ) by t and then set t = 1. Note that the converse result is also valid: x f (x) = kf (x) implies the homogeneity of f . 1.1.4 Economic Application of Eulers Theorem

Suppose the production function q = f (x1 , ..., xn ) is homogenous of degree k and p be the prise of one unit of the product. Then, by Eulers Theorem x1 p f f (x) + ... + xn p (x) = kpq. x1 xn 3

Here pq is the total value of output; is amount of additional output when the input xi is increased by 1, i.e. the marginal product of the i-th input;
f p (x) is price of additional output when the input xi is increased by 1, i.e. xi marginal revenue of the i-th input, assuming x is prot maximizer, it equals to marginal cost of the i-th input; f (x) is the total cost of the i-th input; xi p xi f x1 p (x) + ... + xn p xfn (x) is the total cost of production at prot maxix1 mizer. f (x) xi

Thus from Eulers Theorem follows: If k = 1 the rm spends whole revenue on input, so no prot in this case. If k > 1, then the total payment exceeds the value of output. If 0 < k < 1, then the total payment is less then the value of output, so the rm makes a positive prot. 1.1.5 Homogenizing of a Function

Theorem 5 Suppose a function of n + 1 variables F (x1 , ..., xn , z ) is homogenous of degree k , that is F (tx1 , ..., txn , tz ) = tk F (x1 , ..., xn , z ), and let f (x1 , .., xn ) be the function of n variables which is the restriction of F on arguments with z = 1: f (x1 , .., xn ) = F (x1 , .., xn , 1). Then this restriction allows to reconstruct whole F : F (x1 , ..., xn , z ) = f ( Proof. F (x1 , ..., xn , z ) = F (z x1 xn x1 xn x1 x1 , ..., z , z 1) = z k F ( , ..., , 1) = f ( , ..., ). z z z z z z x1 x1 , ..., ). z z

Theorem 6 For a function f : Rn R and an integer k the n + 1 variable function x1 xn F (x1 , ..., xn , z ) = z k f ( , ..., ) z z is homogenous of degree k . This is unique extension of f to a homogenous function: if G(x1 , ..., xn , z ) is homogenous of degree k and G(x1 , ..., xn , 1) = f (x1 , ..., xn ), then G = F . Proof. Direct calculation (do it). Examples. (a) Let f (x) = x , then its homogenization of degree 1 is x x F (x, y ) = yf ( ) = y = x y 1 . y y (b) Consider nonhomogeneous function f (x) = x ax2 . Its homogenization of degree one is x x x2 x2 F (x, y ) = yf ( ) = y [ a 2 ] = x a . y y y y (c) Let D1 (p1 , p2 ) be the demand function for the good 1, where p1 and p2 are prices for the goods 1 and 2 respectively. As we know D1 is homogenous of degree 0. Suppose we just know the demand function of the good 1 when the price of good 2 is xed at p. Can we reconstruct D1 for arbitrary price vector D1 (p1 , p2 )? Yes: D1 (p1 , p2 ) = D1 (p1 , p2 pp1 p2 pp1 p2 p ) = D1 ( , p) = D1 ( , p). p p p2 p p2

1.2

Cardinal vs Ordinal

A property of a function is called ordinal if it depends only on the shape and location of level sets and does not depend on the actual values of the function. A property is called cardinal if it also depends on actual values of the function. In this context two functions are equivalent if they have the exact same level sets (they represent the same preferences), although they may assign dierent numbers to a given level sets. For example the functions u(x, y ) and u(x, y ) + 2 dier as functions but they have exact same level sets. Denition 2 Let u : Rn R be a function. Its monotonic transformation is the composition g u : Rn R R where g : R R is a strictly increasing function.

In this case u and g u are equivalent in the above sense: they have the exact same level sets. Examples. Let U (x, y ) = xy . Then the following functions 3xy + 2, (xy )2 , exy , lnx + lny are monotonic transformations of u, the suitable strictly increasing g -s are respectively 3z + 2, z 2 , ez , lnz. Monotonic transformations preserve ordinal properties of u but do not preserve cardinal properties. Example. Let u(x, y ) be a utility function which is monotone in x, that is u is an increasing function of x. It means u (x, y ) > 0. x This property is ordinal: if we replace u by the function v (x, y ) which is the composition v (x, y ) = gu(x, y ) with strictly increasing g then v (g u) u (x, y ) = (x, y ) = g (u(x, y )) (x, y ) > 0 x x x since both factors here are positive. Example. Marginal Utility (MU) is cardinal but marginal Rate of Substitution (MRS) is ordinal. Indeed, let u(x, y ) be a utility function and v (x, y ) = g (u(x, y )) be its monotonic transformation. The functions u and v have dierent partial derivatives (dierent marginal utilities), but their MRS equal: v (x,y ) g (u(x,y )) x M RS (v (x, y )) = x = = v (x,y ) g (u((x,y ))
y y

(x,y ) g (u(x,y )) u x g (u(x,y ) u ( x,y ) y

u (x,y ) x u (x,y ) y

= M RS (u(x, y )).

1.3

Homothetic Functions

Denition 3 A function : rn R is called homothetic if it is a monotonic transformation of a homogenous function, that is there exist a strictly increasing function g : R R and a homogenous function u : Rn R such that = g u. It is clear that homothetiticy is ordinal property: monotonic transformation of homothetic function is homothetic (prove it!). Examples. Let u(x, y ) = xy , a homogenous function of degree 2. Then the monotonic transformations g1 (z ) = z + 1, g2 (z ) = z 2 + z, g3 (z ) = ln z generate the following homothetic (but not homogenous) functions 1 (x, y ) = xy + 1, 2 (x, y ) = x2 y 2 + xy, 3 (x, y ) = ln x + ln y. 6

1.3.1

Properties of Homothetic Functions

Theorem 7 Level sets of a homothetic function are radial expansions of one another, that is (x) = (y ) implies (tx) = (ty ) for arbitrary x = (x1 , ..., xn ), y = (y1 , ..., yn ), t > 0. Proof. (tx) = g (u(tx)) = g (tk u(x)) = g (tk u(y )) = g (u(ty )) = (ty ). Theorem 8 For a homothetic function the slopes of level sets along rays from the origin are constant, that is for all i, j and t > 0. In other words Marginal Rate of Substitution (MRS) for a homothetic function is a homogenous function of degree 0. Proof.
(tx) xi (tx) xj

(tx) xi (tx) xj

(x) xi ( x) xj

g (u(tx)) xi g (u(tx)) xj

u g (u(tx)) x (tx)

i u (u(tx)) x j

(tx)

u (tx) xi u (tx) xj

i u tk1 x j

u tk1 x (x)

(x)

u (x) xi u (x) xj

Substituting t = 1 we obtain
(x) xi (x) xj

=
u

u (x) xi u (x) xj

Thus

i x

(tx) (tx)

i = x u

( x) ( x)

i = x

(x) (x)

xj

xj

xj

this completes the proof.

Exercises 1. Which of the following functions is homogenous? What are degrees of homogenous ones? (a) 3x5 y + 2x2 y 4 3x3 y 3 , (b) 3x5 y + 2x2 y 4 3x3 y 4 , (c) x1/2 y 1/2 + 3xy 1 + 7, (d) x3/4 y 1/4 + 6x, x2 y 2 (e) x3/4 y 1/4 + 6x + 4, (f ) + 3. x2 +y 2 2. If f (x1 , x2 ) is homogenous of degree r then
2 fx1 x1 x2 1 + 2fx1 x2 x1 x2 + fx2 x2 x2 = r (r 1)f.

3. Write the degree one homogenization of each of the following functions 3 2 2 (a) ex , (b) ln x, (c) 5, (d) x2 1 + x2 , (e) x1 + x2 . 4. Is the zero function f (x) = 0 homogenous? 5. Show directly that each of the ve utility functions 3xy + 2, (xy )2 , (xy )3 + xy, exy , lnx + lny are equivalent to xy . Show that they have the same marginal rates of substitution at the bundle (2, 1). Show that they have dierent marginal utilities (of good one) at (2, 1). 6. Use the monotonic transformation z k to prove that every homogeneous function is equivalent to a homogeneous function of degree one. 7. Is having decreasing marginal utility, property? Why?
2U x2 i

< 0 for all i, an ordinal

8. Prove that any function f : R1 R1 with f > 0 everywhere is equivalent to a homogeneous function of degree one. 9. Which of the following functions are homothetic? Give a reason for each answer. (a) ex y exy , (d) x2 y + xy,
2 2

(b) 2logx + 3logy, x2 y 2 (e) xy . +1

(c) x3 y 6 + 3x2 y 4 + 6xy 2 + 9,

Homework: Exercises 2, 4, 8, 9 8