Concave functions of two variables While we will not provide a proof here, the following three definitions are equivalent if the function f is differentiable.1 Definition 1: Concave Function The function f is concave on X 2 if for any x0 , x1 X and any (0,1) f ( x ) (1 ) f ( x0 ) f ( x1 ) Definition 2: Concave function The differentiable function f is concave on X if for any x0 , x1 X and any (0,1) f ( x1 ) f ( x 0 ) f 0 1 0 f 0 1 ( x )( x1 x1 ) ( x )( x2 x20 ) x1 x2 Definition 3: Concave function A twice continuously differentiable function f is concave if and only if 2 f 2 f 2 f 2 f 2 f (i) ( x) 0, i 1, 2 and (ii) ( x) ( x) ( x) ( x) 0 xi xi x1x1 x2x2 x1x2 x2x1 In the one variable case a function is concave if the derivative of the function is decreasing. We can use this result and the following proposition to define a class of concave function in higher dimensions. Proposition 1: The sum of concave functions is concave 1 See Module 4 for a proof of the equivalence of D1 and D2. 1 Exercise: Complete the following argument to prove Proposition 1. Suppose that f1 ( x) and f 2 ( x) are both concave. Then for any x 0 , x1 and convex combination x , f1 ( x ) (1 ) f1 ( x0 ) f1 ( x1 ) and f 2 ( x ) (1 ) f 2 ( x0 ) f 2 ( x1 ) Define g ( x) f1 ( x) f 2 ( x) . You need to show that g ( x ) (1 ) g ( x0 ) g ( x1 ) . Remark: If follows that if u1 ( x1 ) and u2 ( x2 ) are both concave then U ( x) u1 ( x1 ) u2 ( x2 ) is concave. In economics we very often assume that a set is convex. Consider for example a consumer whose preferences are represented by the utility function U ( x1 , x2 ) . For any x̂ , the set of points having utility equal to that at x̂ is called a contour set. C {x | U ( x) U ( x0 )} Economists call this an indifference curve. The set of points that are preferred to x̂ is called an upper contour set. CU {x | U ( x) U ( xˆ)} . Consider the figure below showing three contour sets of an increasing function. The upper contour set is the shaded region. 2 Economists typically assume that upper contour sets are strictly convex. The following result establishes that the is property holds for all convex function Proposition 2: The upper contour sets of a concave function are convex Exercise: Complete the following argument to prove Proposition 2. The upper contour set CU {x | U ( x) U ( xˆ )} is convex if for any x 0 , x1 in this set each convex combination x lies in the set as well. Thus it must be shown that if U ( x0 ) U ( xˆ ) and U ( x1 ) U ( xˆ) then U ( x ) U ( xˆ) . Now appeal to Definition 1 for a concave function. Example: Concave function Suppose f ( x) 1 ln x1 2 ln x2 . Note that d 1 ln( x j ) . This is a decreasing dx j xj function of x j . Therefore j ln x j is concave and so, by Proposition 1, f ( x) is concave. By Proposition 2 it follows that the upper contour sets of f are convex. Proposition 3: Sufficient condition for the upper contour sets of a function to be convex Suppose that g () is a strictly increasing function and that h( x) g ( f ( x) is concave. Then the upper contour sets of f are convex. Exercise: Proof of Proposition 3 Complete the following argument to prove Proposition 3 Proof: Appealing to Proposition 2, the upper contour sets of h are convex. That is, for any x̂ the set CU {x | h( x) h( xˆ)} is convex. That is, for any x̂ the set CU {x | g ( f ( x)) g ( f ( xˆ))} is convex. Then argue that g ( f ( x)) g ( f ( xˆ )) if and only if f ( x) f ( xˆ) . 3 Example: f ( x) x11 x22 where x 2 Define h( x) ln f ( x) 1 ln x1 2 ln x2 . As argued above, this is concave and hence has convex upper contour sets. Since ln() is strictly increasing it follows from proposition 3 that the upper contour sets of f are convex. Just as in the one variable case we now show that if f is concave, the FOC for a maximum are both necessary and sufficient. Proposition 4: Necessary and sufficient conditions for a maximum If f is a differentiable concave function then the following conditions are both necessary and sufficient for f to take on its maximum value at x 0 . f 0 f 0 (x ) (x ) 0 . x1 x2 FOC Proof: If f takes on its maximum at x 0 consider changes only in x j , j 1, 2 . This reduces the problem to a one variable problem so we know that the FOC must hold. f 0 f 0 (x ) (x ) 0 . x1 x2 From Definition 2, for any x1 x0 , f ( x1 ) f ( x 0 ) f 0 1 0 f 0 1 ( x )( x1 x1 ) ( x )( x2 x20 ) f ( x 0 ) x1 x2 Thus f takes on its maximum at x 0 . QED 4