1.

PROFIT MAXIMIZATION

The theory of the firm is first presented in terms of general functional forms (Lectures 1- 4) and then in Lecture 6 we consider the Cobb-Douglas production function. For Lectures 1-4 the homework is to redo the previous lecture under the assumption that the production function Q = . Log (K+1) + 2 Log (L+1). Also one should be able to reproduce the lecture without looking at your notes. Note that Log always means natural log.

Q(K, L) = output. Q, is a function of inputs, K = capital and L = labor.

Q QK K

That is, the partial derivatives are denoted by subscripts. Assume Hessian of 2nd partials is negative definite (this implies concavity of the production function).

PERFECTLY COMPETITIVE FIRM P, w and i are exogenous. P = price of output, w = wage, i = interest rate Objective Function = PQ(K, L) Lw Ki KT conditions FOC KT Conditions L = PQL w 0 K = PQK i 0 L0 K0 L . L = 0 K . K = 0 Profit = revenue minus cost

Wage Marginal Revenue Product of Labor Verbal Interpretation Interest Rate Marginal Revenue Product of Capital For an Interior Solution L = k = 0

Implied Relations (assuming L, K > 0)

QL w = QK i

Ratio of Marginal Products = Ratio of Payments to Factors

ISOQUANT

Isocurve

Q(L, K) = Q isoquant d Q = QLdL + QKdK = 0 OR

dK Q = L dL QK

slope of the isoquant

As shown above, a profit maximizing firm sets the slope of the isoquant at the profit maximizing point equals

. So for a profit maximizing firm,

dK w = dL i

This relationship can be understood via the following diagram:

ISOQUANT

3 The straight line is the budget line. A profit maximizing firm will choose the lowest budget line for any isoquant. At that point the slopes will be identical.

HOMOGENIETY

From the first order (K-T ) conditions we can establish homogeneity:

QL W P

QK i P

herefore, the First Order Conditions (FOC) are homogeneous of degree zero in w, P and i. If double w, P and i, then K and L remain the same (rts), and so does Q. Therefore, the maximum of = PQ (K, L) Lw Ki is homogeneous of degree 1 in P, i and w since Q, K, L rts (remain the same) and thus doubling P, i and w doubles . I have said that profits are homogenous of degree 1 in w, i and P. This should not be confused with constant returns to scale of the production function. Q may or may not be homogeneous of degree 1 in K and L. We will discuss returns to scale later.

IMPLICIT AND EXPLICIT DEMAND FUNCTIONS Note that the first order conditions make the profit maximizing L and K implicit functions of w, P and i. Since QLL and QKK are negative and QLLQKK > QLKQLK by assumption, in principle, L and K can be solved as explicit functions of w, P, i: L = L*(w, P, i) K = K*(w, P, i) These derived factor demand functions are the profit maximizing amounts of L and K given w, P, i. An asterisk, *, will be used in this course to denote an explicit function of the exogenous variables.

4 Characteristics of the implicit function (such as homogeniety) hold true for the explicit function as well. L*(Tw, TP, Ti) = L*(w, P, i) and K*(Tw, TP, Ti) = K*(w, P, i) because the first order condions are homogeneous of degree 0 in w, P and i

SECOND ORDER CONDITIONS

Second Order Conditions

H is negative definite since Q is negative definite by assumption H1 < 0; H2 > 0 H1 is the 1x1 determinant -- the upper left term; H2 is the 2x2 determinant

COMPARATIVE STATICS We want to find the effect of a change of an exogenous variable (w, i or P) on an endogenous variable (K or L) assuming that the firm is maximizing profits. That is, we want to find the effect of a change in w, i or P on the profit maximizing K or L, and not on just any possible K or L (we could
denote the profit maximizing K and L by K and L , but this would clutter up the notation further).

We will make use of the implicit function theorem. For our first example, we will find the effect of an exogenous change in w on L. To make things simpler, we will assume that both K and L are greater than 0. Both before and after the exogenous change, the first order conditions hold. That is the marginal profitability of increased K or L is zero. More formally:

d(L) = LLdL + LKdK + LWdw = 0 d(K) = KLdL + KKdK + KWdw = 0 d(L) = PQLLdL + PQLKdK dw = 0 d(K) = PQKLdL + PQKKdK + 0dw = 0 OR
PQLL PQKL PQLK PQKK dl dw = dk 0

W changes

Since there are 2 linear equations (linear in terms of dl and dk), we can solve using Cramers rule. First we find the effect on L
dL = dw 0 PQL K PQK K H

dL PQKK = dw H

6 QKK < 0 by assumption and H = H2 > 0 by assumption. Therefore

dL < 0 dw

Downward sloping factor demand curve

Locally the derived demand curve for the factor is always downward sloping

Next we find the effect of w on K.

dK =

PQLL dw = PQKL dw PQKL 0 H

dK = dw

PQ
H

KL

dK < 0 QK L > 0 dw

This result is surprising to many economics students. While it is true that (in the absence of crowding), more labor will increase the productivity of capital, it may not increase the marginal productivity of capital. That is, QKL, the effect of an increase in L on the marginal product of capital, may be less than 0. Let us look at the following graphs where the output function is drawn in, the higher curve is due to more labor, and the marginal product of the capital is the slope of the curve.

7 In the first diagram, more labor has no effect on the marginal productivity of capital; in the second marginal productivity increases; and in the third it decreases. When wage goes up, L goes down. If more labor decreases the marginal productivity of capital (as in the third diagram), then less labor increases the marginal product of capital, and more K is employed. If more labor increases the marginal productivity of capital (as in second diagram), then when L goes down, marginal productivity of capital will also decrease, and K will go down as well. This would be the case for a Cobb-Douglas production functionQ =AL K .
B C

Comparative statics when i changes d(L) = LLdL + LKdK + Lidi = 0 d(K) = KLdL + KKdK + Kidi = 0 d(L) = PQLLdL + PQLKdK + 0di = 0 d(K) = PQKLdL + PQKKdK di = 0

The effect of a change in i on L

Cross Demands Always Equal

PQL K dL dK = = di H dW

By a similar process one can also show that:

dK PQLL = di H

Note that we use the implicit function theorem to find dK/di. Suppose that we had solved the first order conditions explicitly for K. Then the partial of the explicit function K* with respect to the partial of i would yield the same answer as the total derivative of the implicit function with respect to the derivative of i -- dK/di. All of this can be illustrated via a simple example. Suppose that Q = Log (L+ 1) + 2 Log (K + 1). Then = P Log (L +1) + 2P Log (K + 1) Lw Ki KT conditions FOC L = P/(L+1) w 0 L0 L . L = 0 K . K = 0

K = 2P/(K + 1) i 0 K 0

We can solve for L as an explicit function of P, w and i. For L > 0,

L*(P, w, i) = (P/w) - 1.

The partial of this explicit function, L*, with respect to the partial of w = L*w = - P/w2 Alternatively, we can make use of the implicit function theorem d(L) = LLdL + LKdK + LWdw = 0 d(K) = KLdL + KKdK + KWdw = 0 d(L) = -P/(L + 1)2 dL + 0 dK dw = 0 d(K) = 0d - 2P/(K + 1)2 dK + 0 dw = 0

Hence

dL = - [(L + 1)2]/P dw dL = - P/w2 dw

But by the first order conditions L + 1 = P/w. So,

Thus the total dervivative of the implicit function L with respect to the derivative w is equivalent to the partial derivitive of the explicit function L* with respect to w.

THE EFFECT OF A CHANGE IN P ON L

d(L) = PQLLdL + PQKLdK + QLdP = 0 d(K) = PQKLdL + PQKKdK + QKdP = 0

dL =

Q L dP PQK L = Q L QK K P + QK QK L P dP Q K dP PQK K H H

Since QL =

w i and QK = P P

dL =

[QK K w

+ iQ K L] dP H

QKK < 0

Therefore QKKw > 0

If QKL > 0, then dL > 0 . Note however that dL might be greater than zero even if QKL < 0.

10

LE CHATELIER PRINCIPLE

Le Chatelier Principle

Long-Run Changes (in absolute value) > Short-Run changes We will determine the effect of a change in wage on the amount of labor employed, first in the short run and then in the long run. Recall the following: d(L) = LLdL + LKdK + LWdw = 0 d(K) = KLdL + KKdK + KWdw = 0 d(L) = PQLLdL + PQLKdK dw = 0 d(K) = PQKLdL + PQKKdK + 0dw = 0 W changes

First assume K is fixed. What is the effect of a change in w?

dL = PQLLdL dw = 0

dL 1 = < 0 dw PQL L
Now suppose K is not fixed, then again d(L) = PQLLdL + PQLKdK dw = 0 d(K) = PQKLdL + PQKKdK + 0dw = 0 OR
PQLL PQKL PQLK PQKK dl dw = dk 0

11
dw 0 PQL K PQK K H

dL =

dL = dW

PQK K PQK K = 2 < 0 H P Q L LQ K K P 2 Q 2 K L

The inequality holds Since H > 0 and Q K K < 0 .

1 1 PQ KK = < 0 2 2 2 PQL K PQL L P QLL QKK P QLK PQL L Q KK

12 CONSTANT RETURNS TO SCALE

Constant returns to Scale

(1) Q(TK, TL) = TQ(K, L) Q is homogeneous of Degree 1 in K and L

If there are constant returns to scale, then the hessian of second derivatives is negative semidefinite.

The following Cobb-Douglas production function is an example of constant returns to scale. Q = AK

2/3

. If we multiply both K and L by T we get Q = A(TK)

1/3

(TL)

2/3

= TAK

1/3

= TQ. The hessian of 2nd derivatives looks like

2 - K 5 / 3 L2 / 3 9 2 2 / 3 1/ 3 K L 9

We first take the derivative of Q(TK, TL) = TQ(K, L) with respect to K, and get:

(2) TQK(TK, TL) = TQK(K, L) or QK(TK, TL) = QK(K, L).

That is, the first derivative of a function homogeneous of degree 1 in K and L is itself homogeneous of degree 0 in K and L. This is just a special case of Euler's Theorem.

Next take derivative of both sides of equation 1, Q(TK, TL) = TQ(K, L), with respect to T:

(3) KQK(TK, TL) + LQL(TK,TL) = KQK(K, L) + LQL(K, L) = Q(K, L)

13 The first equality holds by (2). The last equality is the derivative of the right hand side of (1) with respect to t. Equivalently, PKQK + PLQL = PQ(K, L)

But by the first order conditions from profit maximization (note that when there are constant returns to scale, the first order conditions will give us ratios but not amounts of K and L), we have the following:

PKQK = iK and PLQL = Lw Therefore Cost = iK + wL = PKQK + PLQL = PQ(K, L) = Revenue

That is, there are zero profits when there are constant returns to scale.

Note that the above partial notation is somewhat sloppy. QK in TQK(TK, TL) really means the derivative of Q with respect to the first argument, TK, which is denoted by QK not Q1 for mnemonic purposes (the derivative of the argument TK with respect to K is then T).

Please note that Euler's theorem says that if the function is homogeneous of degree 1 with respect to certain variables, then the derivatives of the function are homogeneous of degree 0 with respect to the same variables and vice versa.

Do not conflate this with the homogeniety discussed earlier. Earlier we showed that the first order conditions with respect to K and L were homogeneous of degree 0 with respect to P, w and i (not with respect to K and L). Inspection of the profit equation then showed that the maximum profit equation was homogeneous of degree 1 with respect to P, w and i. However, this homogeniety would not in general be true if the firm were not maximizing profits.

14 2. PROFIT FUNCTIONS AND ENVELOPE THEOREMS

Envelope Theorems for profit maximization: The effect of an exogenous change on the objective function assuming profit maximization.

Recall the following MAX = PQ(K, L) Lw Ki KT conditions FOC KT Conditions L = PQL w 0 K = PQK i 0 L0 K0 L . L = 0 K . K = 0

We first consider an exogenous change in P on profit assuming that profit maximization holds both before and after:

d = PdP + LdL + KdK But by the First Order Conditions, L dL = 0 and K dK = 0 Hence at Profit Maximization Point

d = P dP
Total Derivative = Partial Derivative = PQ(K, L) Lw Ki Then

15 Next we consider the effect of an exogenous change in w on profits.

d = w = L dw d = i = K di
Sometimes students get confused between comparative statics and the envelope theorem. The envelope theorem deals with the objective function while comparative statics deal with the first order conditions. Furthermore under the envelope theorem the total derivative of the maximized function equals the partial of the maximized function while under the implicit function theorem, the partial of the explicit function equals locally the total derivative of the implicit function. The envelope theorem gives the effect of an exogenous change on the maximized value of the objective function (so in the end it is a kind of comparative statics result).

PROFIT FUNCTION The first order conditions make K and L implicit functions of P, w and i. Suppose that we solved for these variables explicitly and then plugged them into our original profit equations. Then we would have a profit function, which was solely a function of the exogenous variables. This profit function would give us the maximum profit for any set of w, i and P. *(w, i, P) = * profit function The profit function contains the same information as the envelope theorems for profit maximization. There profit is a function of the endogenous variables, K and L, but they are constrained to be the profit maximizing amounts by the first order conditions. The profit function has the first order condition inherent in the function itself. Thus the envelope theorems hold for the profit function since the profit function is the maximized value of the objective function.

16 The profit function is increasing in P:

* = Q. P

* is decreasing in factor prices: * = L* (w, i, P); * = K* (w,i, P) w i Another way of viewing this is that in the envelope theorem we take the total derivative of profits with respect to w (for example). The K-T conditions imply that the total derivative equals the marginal derivative. For the profit function we immediately find the partial derivative. Some students confuse maximizing profits with the profit function. One does not find the first order conditions for the profit function since they are already built in. Furthermore the profit function is an explicit function of the exogenous variables, w, i and P; while the maximizing profit equation is a function of the endogenous variables, K and L. The ideas in the previous paragraphs can be solidified by considering a particular example. Suppose that Q = Log (K+1) + 2 Log (L + 1), then = P[ log (K + 1) + 2 log (L + 1)] - Ki - wL and the first order conditions for an interior maximum are: P/(K + 1) - i = 0 and 2P/(L + 1) - w = 0 Equivalently, K* = (P/i) - 1 and L* = (2P/w) - 1.

Plugging these explicit functions back into the profit equation, we get * = P[log(P/i) + 2log(2P/w)] - i(P/i) + i - w(2P/w) + w This function is stating the same thing as subject to the first order conditions.

17 PROPERTIES OF *(w, i, P).

Note that * is a function of the exogenous variables. PROPERTIES OF * (1) * is continuous in P and w, i (2) * is non decreasing in P, non increasing in w, i (3) * is homogeneous of degree 1 in P, w, I (4) * is convex in P, w and i. Before proving this to be the case, let us consider why seeing profit maximization this way is important. First it allows us to do empirical work. For a competitive industry, I, w, and P are exogenous variables, as opposed to K and L, which are endogenous and should therefore not be treated as independent variables. Second it allows us to create profit functions even though the associated production function might not be able to be solved analytically. Third, we might have data on input prices and output prices but not on capital and labor. So this again is an empirical explanation. Finally, some theoretical arguments are most easily done using profit functions. Let us undertake a second diversion before we get to proving the four qualities. Suppose that we have an arbitrary function -- say * = P (w + i)
2 -1

-- and we want to see whether it is a profit

function. Then we test whether it satisfies the 4 qualities. (1) Is it continuous? A differentiable function is continuous. *P = 2P(w + i) 0 *w = -P (w + i) 0 *i = -P (w + i) 0 Since all the first derivatives exist for strictly positive values of w and i, * is continuous. Note that it is not generally true that *w = *i.
2 -2 2 -2 -1

18

(2) From above * is non decreasing in P, non increasing in w, I

(3) Is it homogeneous of degree 1 in P, w, and i? Yes, because

(TP) (Tw + Ti)

-1

= P (w + i) .

-1

(4) Is it convex?
* * *

PP Pw Pi * * * wP ww wi * * * iP iw ii

2(w + i)1 = 2P(w + i)2 2P(w + i)2

2P(w + i)2 2P 2 (w + i)3 2P 2 (w + i)3

2P(w + i)2 2P 2 (w + i)3 2P 2 (w + i)3

This matrix is positive semidefinite. Therefore * is convex and we have a true profit function.

One more question before we get to the proof. What is L*? It is - *w = P (w + i)

Now on to the proof 1. * is continuous in P and w, i. Follows from the Theorem of the Maximum. Q(K, L) is twice differentiable by assumption and K and L are just variables. Therefore Q, K and L are continuous. = PQ - Ki - Lw. Therefore is continuous in P, i and w.

The Theorem of the Maximum states that if a function is continuous (with a compact range) and the constraint set is a non-empty (compact valued) continuous correspondence of A, then the maximum of the function is a continuous function of A.

19

Here we have shown that continuity requirements for the Theorem of the Maximum are satisfied

2. * is non decreasing in P, non increasing in w, i. Follows from the envelope theorem. See Hotelling's lemma and the related derivations.
* = L* (w,i, P) 0; * = K* (w,i, P ) 0 ; * = Q* 0 . p w i

3. * is convex in P, w and i. I will use mathematics to prove convex in w and I. One should use a 3 by 3 determinant to show that convex for w, i and P together, but this is a hopeless complex task in this case. For the more general argument see the diagrams following the mathematical proof.
* is convex in w and i if and only if the following determinant is positive semi definite:

* ww * iw

* wi * ii

We know that * = L* (w,i, P); * = K*(w, i,P) . w i

So

* ww * iw

* wi * ii

L* w K* w

L* i K* i

Remember that Q is negative definite by assumption. Also recall from our earliest comparative statics that the following is true:

dL PQKK = dw H

PQL K dL dK = = di H dW

dK PQLL = di H

20 Therefore:

dL PQ KK = > 0; * = - L * = ww w dw H dK PQ LL = > 0. * = - * = Ki ii di H PQ LL H PQ KL H PQ LK H PQ KK H

* ww So * iw

* wi * ii

L* w * Kw

L* = i K* i

Q LL Q KL

Q LK P 2 -Q KK H 2

Because Q is negative definite and concave, it is readily demonstrated that Q is positive definite and convex. Therefore * is convex in w and i. This can also be illustrated via the following diagram.

*(P) PY* - w*X*

21
is concave in K, L but i and w have inverse relation to K, L respectively so * is convex in i, w

*(w)

Looking at the first diagram, start with a given P and maximize profits (this is the tangency point). If P changes, but K, L and Q remain the same, then profits change in a linear way (just look at the equation for profits). But * is maximal profits and must not be below this straight line of profits. Therefore * is locally convex. 4. * is homogeneous of degree 1 in P, w, i

We have already shown (on page 4) that the first order conditions are homogeneous of degree 0 in P, w and i and that is homogeneous of degree 1 in P, w and i. We will briefly sketch and review the proof here.

= PQ (K, L) Lw Ki

If P, w and i multiplied by T, then Q, L and K remain the same by the first order conditions. Therefore, *(TP, Tw, Ti) = TPQ* L*Tw K*Ti = T*(P, w, i)

22 Note well that the profit function *(P, W, i) is a function of the exogenous variables P, w and i, while (K, L) is a function of K and L.

OVERVIEW

Starting with a production function with a negative definite determinant of second order partials, we have shown that there exists an associated profit function with 4 characteristics; We could have reversed the process. Starting with a profit function that has these 4 characteristics, we could have derived an associated production function that had a negative definite determinant of second order partials.

23 DIFFERENCE BETWEEN PROFIT, MAXIMAL PROFIT AND PROFIT FUNCTIONS

(K, L) = PQ(K, L) Lw Ki is profit. It need not be a maximum. It is a function of K and L. If the K-T conditions are satisfied, and (K, L) = PQ(K, L) Lw Ki is a concave function of K and L, then we are at a maximum.

= PQ(K, L) Lw Ki subject to L = PQL w 0 K = PQK i 0 L0 K0 L . L = 0 K . K = 0

is maximal profit. Notice that L and K are variables. They are implicit functions of w, i and P.

*(w, i, P) is the profit function. It has the same characteristics as maximal profit, except that it is an explicit function of the exogenous variables. For example, in both cases, the total derivative of or * as the case may be with respect to w is the partial derivate with respect to w. *(w, i, P) is a convex function of w, i and P in contrast to profits which are a concave function of K and L.

24 PROPERTIES OF L*(w, i, P), K*(w, i, P), Q*(w, i, P)

These are obviously closely related to the properties of *. I present them here because this is the approach used by Mas-Colell, et. al.

(1) L*, K* and Q* are homogeneous of degree 0 in P, w and i. This follows from the Euler theorem. * is homogeneous of degree 1 in P, w and i, and therefore its first derivatives with respect to P, w, and i are homogeneous of degree 0. Finally, these first derivatives are equal to -L*, -K*, and Q*. That is,

* = L* (w,i,P); * = K* (w,i,P); * = Q* . w i p

Just in case you forgot, here is the formal proof of the Euler Theorem: * is homogeneous of degree 1. Then by definition T1*(w, i, P) = *(Tw,Ti, TP). Taking the derivative of this equation with respect to w, we get: T * (w, i, P) = T * (Tw, Ti, TP). Equivalently, w w T0 * (w, i, P) = * (Tw, Ti, TP). T0= 1 and is just a way of reminding us that we have w w homogeneity of degree 0.

(2) The matrix of derivatives, M*=

Lw* Kw * Qw *

Li* Ki * Qi *

LP* KP * QP *

is symmetric (by Youngs theorem) and positive semidefinite. This is just the set of second derivatives of * with minus signs. That is

25 .
* * ww wi * * iw ii * * Pw Pi L* L* L* * wP w i P * K* K* K* iP w i P = * * Q* * Qw Qi PP P

We have already shown that the upper 2 by 2 is positive semi-definite and argued that the whole matrix is positive semi-definite. Therefore M* is positive semi-definite (multiplying two rows by 1 does not change the sign).

Hessian is symmetric. And thus the other Hessians are symmetric as well (ignoring signs for M*). That is, profit-maximizing quantities respond to an increase in wage in the same way as profit maximizing demand for labor responds to a decrease in the price of the good.

(3) M*(w, i, P)' = 0. That is, the matrix of derivatives times the price vector = 0.

This is just a special case of Euler's theorem. From (1) we have T0 * (w, i, P) = * (Tw,Ti, TP). Equivalently, w w L*(w, i, P) = L*(Tw,Ti, TP).

Differentiating both sides with respect to T and evaluating at T = 1, we get: 0 = w L* + i L* + P L* w i P

26 HOMEWORK (1) In order for WAIBPC to be a profit function, what restrictions are there on A and B and C. W = Wage, I = Interest Rate, P = Price of Good. Find L*(P, i,w). Show that the profit function is continuous. (Hint: If a function is differentiable it is continuous; but not necessarily vice versa). (2) Suppose Q(K, L) =
1 1 K 3 L3

and the firm is competitive in input and output markets and is

maximizing profits.. A) Derive Profit maximizing relations. That is, derive the Kuhn-Tucker conditions and explain the relationships in words. B) Show that the objective function is concave for L, K > 0. C) Find the slope of the isoquant. D) Show that maximal is homogeneous of degree 1 in P, i and w.

(3) A monopolist has the following demand curve: The production function is Q(K, L) = K1/3 L2/3

P = A BQ

A, B > 0

The exogenous variables are A, B, w , and i

A) Is the objective function concave or quasiconcave? Write out the K-T conditions.. B) Find the effect of an increase in w on Q. C) Find the effect of an increase in w on . D) Write out the profit function, *, explicitly. Remember that a profit function is a function of the exogenous variables only.

(4)

A monopolist has demand curve:

P =D(Q)

D' (Q) < 0

The production function is Q(K, L) where Q is strictly concave. A) Find FOC, SOC. Are there additional restrictions on D that insure that the SOC hold? Only answer this yes or no. C) Find the effect of an increase in w on .