You are on page 1of 11

CHAPTER 12

Exercise 12.2
1. (a) Z = xy + λ(2 − x − 2y). The necessary condition is:
Zλ = 2 − x − 2y = 0

Zx = y − λ = 0

Thus λ∗ = 12 , x∗ = 1, y ∗ =

1
2

Zy = x − 2λ = 0

— yielding z ∗ = 12 .

(b) Z = xy + 4x + λ(8 − x − y). The necessary condition is:
Zλ = 8 − x − y = 0

Zx = y + 4 − λ = 0

Zy = x − λ = 0

Thus λ = 6, x = 6, y = 2 — yielding z = 36.
(c) Z = x − 3y − xy + λ(6 − x − y). The necessary condition is:
Zλ = 6 − x − y = 0

Zx = 1 − y − λ = 0

Zy = −3 − x − λ = 0

Thus λ∗ = −4, x∗ = 1, y ∗ = 5 — yielding z ∗ = −19.
(d) Z = 7 − y + x2 + λ(−x − y). The necessary condition is:
Zλ = −x − y = 0

Zx = 2x − λ = 0

Thus λ∗ = −1, x∗ = − 12 , y ∗ =

1
2

Zy = −1 − λ = 0

— yielding z ∗ = 6 34 .

2.


1
(a) Increase; at the rate dz
dc = λ = 2 .
dz ∗
(b) Increase;
= 6.
dc
dz ∗
(c) Decrease;
= 4.
dc
dz ∗
(d) Decrease;
= −1.
dc

3.
(a) Z = x + 2y + 3w + xy − yw + λ(10 − x − y − 2w). Hence:
Zλ = 10 − x − y − 2w = 0

Zx = 1 + y − λ = 0

Zw = 3 − y − 2λ = 0
Zy = 2 + x − w − λ = 0
¡
¢
(b) Z = x2 + 2xy + yw2 + λ 24 − 2x − y − w2 + µ (8 − x − w). Thus
Zλ = 24 − 2x − y − w2 = 0

Zµ = 8 − x − w = 0

Zx = 2x + 2y − 2λ − µ = 0

Zy = 2x + w2 − λ = 0

Zw = 2yw + 2λw − µ = 0
76

¯ ¯ 1 0 −1 ¯ ¯ ¯ g1 ¯ ¯ = −g12 < 0 ¯ Z11 ¯ 3. y) + λ [0 − G(x. g2 and g3 (in that order appearing in the last column and in the last row. In contrast. Since the constraint g = c is to prevail at all times in this constrained optimization problem. the equation takes on the sense of an identity. ¯ ¯ 1 1 0 ¯ ¯ ¯ 0 1 1 ¯ ¯ ¯ 1 0 −1 ¯ = −2. Thus d2 z does not have to be zero as a matter of course.4. z ∗ = 36 is a maximum. y) − λG(x. 77 . 6. Z = f (x. The first-order condition becomes: Zλ = −G(x. Then it follows that d2 g must be zero. the sign of λ∗ will be changed. The zero can be made the last (instead of the first) element in the principal diagonal. and it follows that dg must be zero. Exercise 12. z ∗ = 6 34 is a minimum. z ∗ = −19 is a minimum. with g1 . y)] = f (x. z ∗ = 12 is a maximum. y). but is being set equal to zero to locate the critical values of the choice variables. ¯ ¯ 2 1 0 ¯ ¯ ¯ 0 1 1 ¯ ¯ ¯ 1 0 1 ¯ = 2. y) = 0 Zx = fx − λGx = 0 Zy = fy − λGy = 0 5. No. the equation dz = 0 is in the nature of a first-order condition — dz is not identically zero.3 1. ¯ ¯ 1 −1 0 ¯ ¯ ¯ 0 1 1 ¯ ¯ ¯ 1 2 0 ¯ = −1. too. The new λ∗ is the negative of the old λ∗ . (a) Since (b) Since (c) Since (d) Since ¯ ¯ ¯ ¯ ¯ ¯¯ ¯¯ = ¯ ¯H ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯¯ ¯H ¯¯ = ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯¯ ¯H ¯¯ = ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯¯ ¯¯ = ¯ ¯H ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ 0 ¯1¯ = ¯ 2. ¯H ¯ ¯ g1 ¯ ¯ 0 1 2 ¯ ¯ ¯ 1 0 1 ¯ = 4.

A sufficient condition for maximum z is ¯H ¯ ¯ ¯ ¯ ¯ 3 ¯ > 0 and ¯H ¯ ¯ > 0. Examples of acceptable curves are: 78 . H = ¯¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ 0 0 g11 g21 g31 0 0 g12 g22 g32 g11 g12 Z11 Z12 Z13 g21 g22 Z21 Z22 Z23 g31 g32 Z31 Z32 Z33 g41 g42 Z41 Z42 Z43 ¯ ¯ g41 ¯ ¯ ¯ g42 ¯ ¯ ¯ Z14 ¯ ¯ ¯ Z24 ¯ ¯ ¯ Z34 ¯ ¯ ¯ Z44 ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ 3 ¯ < 0 and ¯H ¯ 4 ¯ = ¯H ¯ ¯ > 0.¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ 4.4 1. A sufficient condition for minimum z is ¯H Exercise 12.

and thus f [θu + (1 − θ)v] = a. we must actually have v > u. 1. (b > 0). and strictly so. . is strictly . 79 Hence f (x) = a + bx. which is equal to (not greater than) f (u). (a) Quasiconcave. or v ≥ u. f (v) ≥ f (u) means that a+bv ≥ a+bu.Fig. quasiconcave. 2. Since f [θu + (1 − θ)v] = a + b [θu + (1 − θ)v] = a + b [θu + (1 − θ)v] + (bu − bu) = a + bu + b (1 − θ) (v − u) = f (u) b (1 − θ) (v − u) = f (u) + some positive term it follows that f [θu + (1 − θ) v] > f (u). to have u and v distinct. Moreover. This is because f (v) = f (u) = a. but not strictly so. In the present case. (b) Quasiconcave.

is strictly quasiconcanve. neither S ≥ nor S ≤ is a convex set. Here. (c) Setting x2 − ln x1 = k. If we pick k = 0. f (v) ≥ f (u) means a + cv2 ≥ a + cu2 . we can rewrite the above expression as a + cu2 + c [θu + (1 − θ) v + u] [θu + (1 − θ) v − u] = f (u) + c [(1 + θ) u + (1 − θ) v] [(1 − θ) (v − u)] = f (u) + some positive term > f (u) Hence f (x) = a + cx2 . this in turn means v < u. Now we have ¡ ¢ 2 f [θu + (1 − θ)v] = a + c [θu + (1 − θ) v] + cu2 − cu2 n o 2 = a + cu2 + c [θu + (1 − θ) v] − u2 Using the identity y 2 − x2 ≡ (y + x) (y − x). or v 2 ≤ u2 (since c < 0). The function is neither quasiconcave nor quasiconvex. 2. In the x1 x2 plane. 4. with a hill in the second quadrant and valley in the fourth. (a) This cubic function has a graph similar to Fig.(c) Quasiconcave. Both f (x) and g (x) are monotonic. f (x) + g (x) displays both a hill and a valley. for instance. and thus quasiconcave. this plots for each value of k as a log curve shifted upward vertically 80 . (b) This function is linear. we get the isovalue equation x2 = ln x1 + k. and strictly so. 3.8c. and hence both quasiconcave and quasiconvex. Therefore f (x) + g (x) is not quasiconcave. If we pick k = 5 12 . neither S ≥ nor S ≤ will be a convex set. (c < 0). For nonnegative distinct values of u and v. and solving for x2 . However.

Since f 0 (x) = 2x. fxx = −2. The set S ≤ .25’) for quasiconvexity. then the cubic function will be upward-sloping for nonnegative x. b < 0. fy = −2y. confirming the conclusion in Example 1. the function is quasiconcave. consists of the points lying on or below a rectangular hyperbola — not a convex set. we know that if a. fyy = −2. Then. c. by (12. by (12. Let u and v be two values of x. and fyy = 0. 8. (a) A cubic curve contains two bends.26). since fx = y. 7.21). Thus the function is quasiconvex (but not quasiconcave). d > 0. The set S ≤ = {(x1 . 9. which violates the necessary condition (12. (a) Since fx = −2x. we have |B1 | = −y 2 ≤ 0 and |B2 | = 2xy ≥ 0. fxx = −2. we have |B1 | = −4 (x + 1)2 < 0 |B2 | = 8 (x + 1)2 + 8 (y + 2)2 > 0 By (12. fy = −2 (y + 2).26). x2 ) | f (x1 . and let f (v) = v2 ≥ f (u) = u2 .21).21). the function is quasiconcave. fxy = 1. and b2 < 3ac.22). it is both quasiconcave and quasiconvex. which implies v ≥ u. Alternatively. we have ¡ ¢ |B1 | = −4x2 < 0 |B2 | = 8 x2 + y 2 > 0 By (12. 5.by the amount of k. Hence the function is quasiconvex by (12. x2 ) ≤ k} — the set of points on or below the isovalue curve — is a convex set. fyy = −2. and would thus violate both parts of (12. fy = x. involving the inequality xy ≤ k. (b) From the discussion of the cubic total-cost function in Sec. fxy = 0. fxx = 0. we find that f 0 (u) (v − u) = 2u (v − u) ≥ 0 f 0 (v) (v − u) = 2v (v − u) ≥ 0 Thus. 81 . (b) Since fx = −2 (x + 1). the function is both quasiconcave and quasiconvex. fxy = 0. 6.4.

³ ∗´ ³ ∗´ ³ ∗´ B+Py ∂x ∂x ∂x < 0. but raises the optimal purchase of the other commodity. and y ∗ = 11.5 1. we get λ∗ = 3. 2Px Py Thus: 82 . Uxy = Uyx = 1. Hence utility is maximized. Zx = y + 1 − 4λ = 0. we can find that λ∗ = B+2Px +Py 2Px Py ¯ ¯ ¯ 0 ¯ ¯ ¯¯ ¯¯ = ¯ P (c) ¯H ¯ x ¯ ¯ Py Px 0 1 +y +x = −B = −1 = −2 B+2P −P B−2P +P x y x y x∗ = y∗ = 2Px 2Py ¯ ¯ Py ¯ ¯ ¯ 1 ¯ = 2Px Py > 0.Exercise 12. We have Uxx = Uyy = 0. (a) Z = (x + 2) (y + 1) + λ (130 − 4x − 6y) (b) The first-order condition requires that Zλ = 130 − 4x − 6y = 0. Thus we have ¯ ¯ ¯ 0 4 ¯ ¯ ¯¯ ¯ ¯ ¯ (c) H = ¯ 4 0 ¯ ¯ ¯ 6 1 (d) No. ¯ ¯ 0 ¯ 2. x∗ = 16. Py = 6. y An increase in income B raises the level of optimal purchases of x and y both. an increase in the price of one commodity reduces the optimal purchase of that commodity itself. Yes. and B = 130. (a) Z = (x + 2) (y + 1) + λ (B − xPx − yPy ) (b) As the necessary condition for extremum. 1 2Px > 0. ∂P = − = 3. These check with the preceding problem. ∂Py 2P 2 < 0. 2 ∂B 2P ∂Py x x ³ ∗´ ³ ∗´ ∂y ∂y 1 x = − B+2P ∂Px = Py > 0. Zy = x + 2 − 6λ = 0 λ∗ = 3. Utility is maximized. 4. 2Px and λ∗ = (B+2Px +Py ) . ³ ∂y∗ ∂B ´ = 1 2Py > 0. = 2P1x > 0. |J| = ¯H x∗ = (B−2Px +Py ) . ¯ ¯ ¯ ¯ = 2Px Py . x∗ = 16 and y ∗ = 11. ¯ ¯ 6 ¯ ¯ ¯ 1 ¯ = 48 > 0. ¯ ¯ 0 ¯ (d) When Px = 4. we have Zλ = B − xPx − yPy = 0 or Zx = y + 1 − λPx = 0 −Px x −Py y −Px λ Zy = x + 2 − λPy = 0 −Py λ By Cramer’s Rule.

homogeneous of degree one. Thus dU ¯ is constant. and hence ¯H ¯ ¯ ¯ ¯. homogeneous of degree one. The statement is not valid.330 ). ¯ =U ¯ (x∗ .33’). Therefore we cannot be sure that ¯H dx (12. From (12. (a) No. 6.(a) (b) ³ ³ ∂x∗ ∂B ∂x∗ ∂Px ´ ´ = 1 2Px . 1. jw w ¡ ¢ 4 3 (f) (jx) − 5 (jy) (jw) = j 4 x4 − 5yw3 . if d2 y dx2 Exercise 12. we have dU ¯ = 0. express dU 7. Let j = k1 .6 p √ (jx) (jy) = j xy. nothing definite be said about the sign of Uxx and > 0. we have Ux Uy = Px Py at the optimum. 2. 5. ¯ ¯ ¯ ¯ > 0. then 3. K ¢ ¡ L¢ ¡L¢ ¡L¢ = f 1. but says ¯ ¯ ¯ ¯ > 0 in (12. or where Ux and Uy are evaluated at the optimum. homogeneous of degree one. ∂y∗ ∂B and ´ ³ = 1 2Py . When U Ux dx∗ + Uy dy ∗ = 0. diminishing marginal utility means only that Uxx and Uyy are negative. A negative sign for that derivative can mean either that the income effect (T1 ) and the substitution effect (T2 ) in (12. Q K =f ¡K L K. ´ ³ 2 2 2 xy (e) (jx)(jy) + 2 (jx) (jw) = j + 2xw . =− and ³ (B+Py ) 2Px2 . ∂y∗ ∂Px ´ = 1 Py . Thus Q = Kψ K . p ¡ √ ¢ (d) 2jx + jy + 3 (jx) (jy) = j 2x + y + 3 xy . homogeneous of degree four. K =ψ K . Uyy . i 12 h ¢1 ¡ (b) (jx)2 − (jy)2 = j x2 − y 2 2 . or that the income effect is positive (inferior good) but is overshadowed by the negative substitution effect. The optimal utility level can be expressed as U ¯ = Ux dx∗ + Uy dy ∗ .32) and d2 y2 > 0 in nothing about Uxy .420) are both negative (normal good). y ∗ ). Thus we can also ¯ = 0 by Px dx∗ + Py dy ∗ = 0. These answers check with the preceding problem. 83 . (a) (c) Not homogeneous. or -Px dx∗ − Py dy ∗ = 0. because Uxy also appears in ¯H (b) No. homogeneous of degree two.

or MPPK = APPK . a + b + c > 1 implies increasing returns to scale. we have K ∂K = Q. implying decreasing returns to scale. implying increasing returns to scale. hence APPL indeed can be plotted against k. Thus ²QK = ∂(ln Q) ∂(ln K) b c =α and ²QL = ∂(ln Q) ∂(ln L) =β 7. 4. or ∂Q ∂L ∂Q (b) When MPPL = 0. Slope of a radius vector = Akα k = Akα−1 = APPK . APPL − k · MPPK = Akα − kAαkα−1 = Akα − Aαkα = A (1 − α) kα = MPPL . (a) Since the function is homogeneous of degree (α + β). a (a) A (jK) (jL) (jN ) = j a+b+c · AK a Lb N c = j a+b+c Q. thus the slope of APPL = Aαkα−1 = MPPK . (c) APPK = φ(k) k = APPL k = ordinate of a p oint on the APPL curve abscissa of that p oint =slope of radius vector to the APPL curve (d) MPPL = φ (k) − kφ0 (k) = APPL − k · MPPK 5. or MPPL = APPL . or ∂Q ∂K = = Q L. homogeneous of degree a + b + c.(a) When MPPK = 0. d. APPL = Akα . the value of the function will increase more than j-fold when K and L are increased j-fold. (c) Share for N = N ∂Q ( ∂N ) Q = N AK a Lb cN c−1 AK a Lb N c =c 84 . (c) Taking the natural log of both sides of the function. Yes. b. (b) a + b + c = 1 implies constant returns to scale. c. the value of the function will increase less than j-fold when K and L are increased j-fold. they are true: (a) APPL = φ (k). Q K. if α+β > 1. 6. we have ln Q = ln A + α ln K + β ln L. we have L ∂Q ∂L = Q. (b) If α + β < 1. (b) MPPK = φ0 (k) = slope of APPL .

4. On the basis of (12. it will plot as a straight line passing through the origin. no. Then the equation in (a) yields: ¡K ¢ ¡K ¢ Q 2 L2 = g L . because QLL and QKK have both been found to be negative. we have h ¡ ¢ρ ¡ ¢ i δ−1 d2 K d δ−1 K 1+ρ = = δ (1 + ρ) K 2 dL dL δ L L ¡ K ¢ρ 1 ¡ dK ¢ δ−1 = δ (1 + ρ) L L2 L dL − K > 0 5. with a (positive) slope equal to β α.66). A larger δ implies a larger capital share in relation to (b) No. the output levels are in the ratio of 1 : 22 : 32 . jL) (b) Let j = L1 . Since ab ∗ = α Pb . (b) With second-degree homogeneity. ³ ∗´ ³ ´³ ´ β Pa 2. This result does not depend on the assumption α + β = 1.7 1. The elasticity of substitution is merely the elasticity of this line. 8. (a) Linear homogeneity implies that the output levels of the isoquants are in the ratio of 1 : 2 : 3 (from southwest to northeast).8. This implies that Q = L φ (k). (a) Lab or share Capital share = LfL Kfk the labor share. we get ³ ´ ¡ ¢ 0 (jL) φ0 jK = jLφ0 K jL L = jLφ (k) = j · MPPK Thus MPPK is homogeneous of degree one in K and L. Now MPPK depends on L as well as k.2) to be unity at all points. Exercise 12. or 1 : 4 : 9. (d) If K and L are both increased j-fold in the MPPK expression in (c). Yes. which can be read (by the method of Fig. 85 . = 1−δ δ ¡ K ¢ρ L d dL ¡K ¢ L £ because dK dL ¤ <0 . ¡ ∂k ¢ ¡1¢ 0 2 0 2 0 (c) MPPK = ∂Q ∂k = L φ (k) ∂K = L φ (k) L = Lφ (k). (a) j 2 Q = g (jK. 3. 1 = φ L = φ (k).

If both K and L are changed j-fold.. 7. output will change from Q to: h i− ρr −r −r −ρ −ρ A δ (jk) + (1 − δ) (jL) = A {j −ρ [δK −ρ + (1 − δ) L−ρ ]} ρ = (j −ρ ) ρ Q = j r Q Hence r denotes the degree of homogeneity. xx must tend to e0 = 1 as x → 0+ . By L’Hôpital’s Rule.6. (a) By successive applications of the rule. With r > 1 (r < 1).68). The isoquants would be downwardδ sloping straight lines. But as ρ → −1. and n (x) = x1 . and since. σ → ∞. (c) Linear isoquants and infinite elasticity of substitution both imply that the two inputs are perfect substitutes. the expression x ln x tends to zero as x → 0+ . we have increasing (decreas- ing) returns to scale.66) yields dK dL = − (1−δ) = constant. we have: (a) limx→4 x2 −x−12 x−4 (b) limx→0 ex −1 x (c) limx→0 5x −ex x (d) limx→∞ ln x x = limx→4 = limx→0 ex 1 = limx→0 =7 =1 5x ln 5−ex 1 1 x = limx→∞ 2x−1 1 1 = ln 5 − 1 =0 9. from (b) above. σ is not defined for ρ = −1. (12. (b) By (12. 8. (a) If ρ = −1.. we find that limx→∞ xn ex = limx→∞ nxn−1 ex = limx→∞ n(n−1)xn−2 ex = . = limx→∞ n! ex =0 (b) By taking m (x) = ln x. 86 . we have limx→0+ ln x 1 x = limx→0+ 1 x − x12 = limx→0+ −x = 0 (c) Since xx = exp (ln xx ) = exp (x ln x).