Solutions Manual for ANDREU MAS-GOLELE MICHAEL D. WHINSTON LUO) Lael Chiaki Hara Ilya Segal Steve TadelisCHAPTER 1 L.B.1 Since y > z implies y » 2, the transitivity implies that x » 7 Suppose that z x x. Since y x z. the transitivity then implies that y x x, But this contradicts x > y. Thus we cannot have z > x. Hence x > z. 1.B.2 By the completeness, x > x for every x € X. Hence there is no x © X such that x x ippose that x > y and y > z, then x > y yz By (iii) of Proposition 1.B.1, which was proved in Exercise 1.B.1, we have x > z. Hence > is transitive. Property (i) is now proved. As for (ii), since x x x for every x € X, x ~ x for every x € X as well. ‘Thus ~ is reflexive. Suppose that x ~ y andy ~z Then x¥ryye7%yEX and z x y. By the transitivity, this implies that x x z and z x x. Thus x ~ 2. Hence ~ ie transitive. Suppoce x that ~ y. Then x xy andy yx Thus y x x and x yy. Hence y ~ x. Thus ~ is symmetric Property (ii) is now proved. 1B.3 Let *¢ X and yc X. Since u(+) represents >, x » y if and only if u(x) = uly). Since f(-) is strictly increasing, u(x) = uly) if and only if v(x) = vly). Hence x > y if and only if v(x) = vly). Therefore vi-) represents y. 1.B.4 Suppose first that x > y. If, furthermore, y x x, then x ~ y and hence u(x) = uly). If, on the contrary, we do not have y > x, then x > y. Hence u(x) > u(y). Thus, if x » y, then ule) = uly). Suppose conversely that u(x) = u(y). If. furthermore, u(x) = u(y), then ix y and hence x » y. If, on the contrary, u(x) > uly}, then x > y. and henee x x y. uly), then x x y. So ul+) represents >. 1.B.S. First, we shall prove by induction on the number N of the elements of X that. if there is no indifference between any two different elements of X, then there exists a utility function. If N = 1, there is nothing to prove: Just assign any number to the unique element. So let N > 1 and suppose that the above assertion is true for N - 1. We will show that it is still true for N. Write X = Gy, *y-p%y) By the induction hypothesis, x can be represented by a utility funetion u(-) on the subset (xp.-y_,) - Without loss of generality we can assume that u(x,) > ulx,) >... > ulxy_,) Consider the following three cases: Case 1: For every i < N, x, > Case 2: For every 1 < Ny x, > Hy, Case 3: There exist i < Nand Jj <¢ N such that x, > Hy > x; Since there is no indifference between two different elements, these three cases are are exhaustive and mutually exclusive. We shall now show how the value of ulx,) should be determined, in each of the three cases, for u(-) to represent ¢ on the whole x. If Case 1 applies, then take u(xy) to be larger than u(x,). If Case 2 applies. take ulxy) to he smaller than uly). Suppose now that Case 9 applies. Let I= i ¢ (lu... N- Ih x) > xy,)) and J= Ge ..., N= De Xyu1 > Xj Completeness and the assumption that there is no indifference implies that 1 u J = G,..., N- 1). The transitivity implies that both I and dare “intervals,” in the sense that if i ¢ I and {’ <1, then i’ € 1; and if j€Jand j’ > j, then j © J. Let i* = max I, then i* + 1 = min J. TakeUomy) 10 tie the open interval (ulxje,,).t(%j_))- Then it is easy to see that u(-) represents x on the whole X. ‘Suppose next that there may be indifference between some two elements of x). For each n = X= &, AN, define %,, = Og, © X q, ~ %,)- There by the reflexivity of ~ (Proposition LB.1GD), UN.X, = Xs Also, by the transitivity of ~ (Proposition 1.8.16i)), if X, # X,, then X, 0%, #8 So in Xe n° Xm let M be a subset of (1,...,N) such that X Umen%nn 290 X,, # X, for any me M ‘and any ne M with m#n, Define an relation »* on {X,,: m € M} by letting X,, Fe he lalla le tal indifference between two different elements of (X,,: me M). Thus, by the preceding result, there exists a utility function u®(+) that represents 2*. Then define us XR by ulx,) = uNX,) if me Mand x, € X,, It is easy to show that, by the transitivity, u(+) represents >. LCA If y € Cl(xy,2)), then the WA would imply that y « C((x,y)), Rut contradicts the equality Cl(x,y)) = Gd. Hence y € C({x,y,2)). Thus CUx,y,2)) € (20,(2),00,2)) 1.€.2 The property in the question are cyuivalent to the following proper") Wee, B eB, xeByeB, xeBy y eB, x € OB), and y © OB), then x € C(B’) and y € CIB). We shall thus prove the equivalence between this property and the’ Weak Axion. Suppose first that the Weak Axiom is satisfied. Assume that B ¢ 8, B’ ¢ B,x* y, then there is some B € B such that x € B, y € B, x € CIB), and y € G(B). Thus x >" y. Suppose that y »* x, then there 8 Re A such that xB, ye Band x © G(B). But the Weak Axiom impliee that y € C(B), which is a contradiction. Hence if x >* y, then we cannot have yy" x. Hence x >** y. Conversely, suppose that x >** y, then x x* y but not y >* x. Hence there is some B € B such that x ¢ B, y ¢ B, x C(B) and if x € BY and yc BY for any BY © B, then y € C(B"). In particular, x € C(B) and y € C(B). Thus xt y, The equality of the two relation is not guaranteed withont the WA. As can be seen from the above proof, the WA is not necessary to guarantee that if x >" y, then x >* y, But the converse need not be true, as shown by the following example. Define X = (x,y,z), B= ((x,y)(x,y,2)), Clix.) = (x), and C((x,y,z)) = {y). Then x >* y and y >* x. But neither x >* y nor y >* x, (b) The relation >* need not be transitive, as shown by the following example Detine X = (x,y.2h, B= (Oxy) (y.2)), Clay?) = tx) and Cl{y,2)) = ty). Then x >* y and y >* z. But we do not have x >* z (because neither of the two sets in B includes (x,z}) and hence we do not have x >* z either. (c) According to the proof of Proposition 1.D.2, if B includes all three element subset of X, then »* is transitive. By Proposition 1.B.1(i), »** istransitive. Since >* ts equal to >**, >* 1s also transitive. ‘An alternative proof is as follows: Let x ¢ K, y ¢ K, 2 € X, x ># y, and y >* z, Then {x.y.z) € B and, by (a). x >** y, and y >"* z, Hence we have neither y >* x nor z >* y. Since >* rationalizes (B,C(-)), this implies that y € CUix,y,2)) and z € Clix,y,2)). Since Cl{x,y.2)) # @, CUx,y.2)) = (x). Thus x >* 2. LD.1 The simplest example is X = (x,y), B= Ux){y)), CUO) = Ge, Cy) = (y). Then any rational preference relation of X rationalizes C{-). L.D.2_ By Exercise 1.B.5, let u(+) be a utility representation of . Since X its, for any BS X with B+ @ there exists x © B such that u(x) = u(y) for all y ¢B. Then x ¢ C*(B,>) and hence C¥(B,,) + 0. (A direct proof with no use of utility representation is possible, but it is essentially the same as the proof of Exercise 1.B.5.) 1.D.3 Suppose that the Weak Axiom holds. If x € G(X), then x © C(Xx,z)), which contradicts the equality Gl(s,z)) = {2}. If y € G(X), then y € CUix,y), which contradicts Cl(x,y)) = (x). If z € C(X), then z € Clty,z)), which contradicts C((x,z)) = (y}. Thus (B,C{-)) must violate the Weak Axiom, L.D.4 Let » rationalize G(-) relative to BR, Let x € G(H, W BL) and y © GCF) v GB), then x x y bevause By UB, > G(B,) v O(B,). Thus x © O(0(B,) uv ctp,)). Let x € O(C(B,) v O(B,)) and y € B, v Bp. then there are four cases: Case 1. x € CIB), y € By.Case 2. x © OB), y © By, Case 2 x 6 CB), ¥ ¢ BY. Case 4. x € CIB,). v € By. If either Case 1 or 4 is true, then x » y follows directly from rationalizability, If Case 2 is true, then pick any z € C(B,). Then z > y. Since x € O(C(B,) u CIB,)), x x z. Hence, by the transitivity, x > y. If Case 3 is true, then pick any z € C(B,) and do the same argument as for Case 2. 1.D.5 (a) Assign probability 1/6 to each of the six possible preferences, which are x > y>Z,X>Z>Y,Y>X>2¥>Z>x,Z>x>y, andzry> (b) If the given stochastic choice function were rationalizable, then the probability that at least one of x > y, y » z, and z » x holds would be at most 3 x (1/4) = 3/4. But, in fact, at least one of the three relations ~ always holds, because, if the first two do not hold, then y » x and 2 > y. Hence the transitivity implics the third. Thus, the given stochastic choice function is not rationalizahle (c) The same argument as in (b) can be used to show that a = 1/3. Since CUx,y = CUiy,2)) = CUZ,x}) = (a 1 - a) is equivalent to Cl{y,x)) = Czy = OKx.2n " (1 ~ @ a), if we apply the same argument as in (h) to y >) Z> y, and x > 2, then we can establish I - @ = 1/3, that is, « = 2/9, Thus, in order for the given stochastic choice function is rationalizable, it is necessary that a € [1/3,2/3]. Moreover, this condition is actually sufficient: For any « € [1/3,2/3], assign probability a - 1/3 to each of x > Meé y>2,y>Z>x, and z > x > y; assign probability 2/3 - @ to each of x > 2 > yy yx z, and z>y 7x. Then we obtain the given stochastic choice function. 17CHAPTER 2 2.D.1 Let p, be the price of the consumption good in period 2, measured in units of the consumption good in period 1. Let x,, x, be the consumption levels in periods 1 and 2, respectively. Then his lifetime Walrasian budget 2. < set is equal to (x € RU: x, + ppx, = wh. 2n2 (uh) eR: b= 24, pe ens 28 2.D.3 (a) No. In fact, the budget set consists of the two points, each of which is the intersection of the budget line and an axis. () Let xe By, XB, yy and A € [0,1]. Write x" = Ax +1 - alk’. Since X is convex, x" « X. Moreover, p-x" = Ap-x) + (1 = AVlp-x') = Aw + (1 = Adw =w. Thus x" B |. paw 2.D.4 It follows from a direct calculation that consumption level M can be attained by (6 + (M ~ 85)/s") hours of labor. It follows from the definition that (24,0) and (16 - (M - 86)/c’, M) are in the budget cet. But their convex combination of these two consumption vectors with ratio M - 8s is not in the budget set: the amount of leisure of this combination equals to 16 (so the labor is eight hours), but the amount of the consumption good is 8 8 8 M >M—Po—ae = Mae = 88 ve HSE vs Hee Ws 22.E.1 The homogeneity can be checked as follows op, ew xjlepew) = a : op, + ap, + OP AP) ap, aw Xglepew) = Gaps ay BPS 2 D7 Py Py Py ep, ow xglap.aw) “ap, + ap, + =, OP, To see if the demand function satisfies Walras’ law, note that PP, + Pz + Py pextow) = Pipaeuaes) Hence p:x(p,w) = w if and only if @ = 1. Therefore the demand function satisfies Walras' law if and only if & = 1 2.E.2 Multiply by p,/w both sides of (2.£.4), then we obtain L Lge! PpylP¥0/ WG pl.) 8p, P,/x (WI) + PLY (P.WI/W Hence p_by(p.wep(p.w) + by(p.w) = 0. By (2.E.6), TE _(ppeglp. w)/ wil x ylp.v)/8N(w/xy(p.¥)) Hence Lye ¢lP-Wlepy(P.w) = 1. 2.5.3. There are two ways to verify that p:D,x(p,wip = = w. One way {s to post-multiply (2.E.5) by p, then p-D.x(p.w) p+ w= 0 by Walrac’ law. The other way is to pre-multiply (2.F.1) by pT, then p-Dpx(p,wip + p+D, x(p,w)w = 0, By Proposition 2.E.3, this is equal to p-D,x(p,wp + w An interpretation is that, when all prices are doubled, in order for the consumer to stay at the same consumption, it is necessary to increase his wealth by w. 2-2 °.2.6.4 Dy differentiating the equation x(paw) = axtp,w) with respect to a and evaluating at a = 1, we obtain wD,,x(p,w) = x(p,w). Hence D,lp,w) ~ (/w)x(p.w). Hence €,,, = (8xylp.w)/ow)lw/x,(p.w)) = 1, This means that an one-percent increase in wealth will increase the consumption level for all guods by one percent. Since (1/w)x(p,w) = x(p,1) by the homogeneity assumption, D.,x(p,w) is a function of p only. The assumption also implies that the wealth expansion path, F, = Gx(pyw): w > 0}, is a ray going through x(p,). 2.E.S Since x(p,w) is homogeneous of degree one with respect to w, x(p,aw) = ex(p,w) for every a > 0. “Thus x,(p,w) = x,(p,lw. Since ax,{p,11/8p, = Spq(p)/op, = 0 whenever k + & xp(pd) is actually @ Function of pp alone, So we can write xy(pw) = x/(P_). Since x(p,w) is homogencous of degree zero, x,(Py) must be homogeneous of degree - 1 (in p,). Hence there exists a, > 0 such that x(pj) = ay/py, By Walras’ law, Typy(cty/p,Jw = wD ya, = w. We must thus have Syay = 1. 2.E.6 When @ = 1, Walras’ nw and homogeneity hold, Hence the conclusions of Propositions 2.E.1 ~ 2.E.3 hold. 2.6.7 By Walras’ law, (w - ppqi/n, = w/p, ~ (p)/p,)ew/p,) = UL - alw/p,. ‘2 2 This demand function is thus homogeneous of degree zero. 2-32.6.8 For the first part, note that In ¥((p.w) = In yylexptin p,),....exptin p, exp(in w)), Thus, by the chain rule, ax, ax, atin x,(0.00) apie re tin Pi apeh wd aE 7 pw = eyiprwT 7 See Similarly, ax, ax, (in xylp.w)) = (p.w)-explin w) ay iP wlw Sat BT pipe * Fel Since a, = d(in, x,(p,w))/dlln pj), a, = dlln x)(p,w))/alIn pp), and a, = (in x(p,#))/dlin w), the assertion is established 2.F.1 We proved in Exercise 1.C.2 that Definition 1.C.1 and the property in the exercise is equivalent. It is easy to see that the latter is equivalent to the following property: For every Be B and B' ¢ B, if O(D) nD’ ¥ © and D 1 C(B") @, then CIB) n BY c CIB") and Bn C(B') < CIB). If Cl-) is single-valued, then this property is equivalent to the following one: For every B ¢ B and B’ € B, if C(B) < BY and Bc C(B’), then C(B) = C(B"). In the context of Walrastan demand functions, this can be restated as follows: For any (p,w) and (p',w'), if p-x(p',w') = w and p’-x(p,w) = w’, then x(p,w) = x(p',w"). But this is the contraposition of the property stated in Definition 2.F.1. Hence Definitions 1.C.1 and 2.F.1 are equivalent, 2.F.2 It is straightforward to check that the Weak Axiom holds. In fact, if = 9. since p?x! = 8, x2 Js 8 andi + 4, then p’-x’ Is revealed P preferred to x!, similarly, siuce p!-x9 = 6, x! is revealed preferred to x%,But, since p°-x" = 8, x° is revealed preferred to x 2.F.9 [First printing errata: Add the sentence "Assume that the weak axiom is satisfied." in (b) and (c).] Denote the demand for good 2 in year 2 by y. (a) His behavior violates the weak axiom if 100-120 + 100y = 100-100 + 100-100 and 100-100 + 80-100 100-120 + Soy. That is, the Weak Axiom is violated if y ¢ [75,0]. (b) The bundle in year 1 is revealed preferred if 100-120 + 100y = 100-100 + 100-100 and 100-100 + 80-100 > 100-120 + 80y, that is, y < 75. (c) The bundle in year 2 is revealed preferred if 100-100 + 80-100 = 100-120 + 8oy ana 100-120 + 100y » 100-100 + 100-100, that is, y > 80. (@) For any value of y, we have suffictent information to Justity exactly one of (a), (b), and (c). (©) We shall prove that if y < 75, then good 1 is an inferior good. So suppose that y < 75. Then100-120 + 100y = 100-100 + 100-100 and 100-100 + 80-100 > 100-120 + Boy. Hence the real wealth decreases from year 1 to 2. Also the relative price of good | increases, But the demand for good 2, y, deercases because y < 75 < 100. This means that the wealth effect on good 1 must be negative. Hence it is an inferior good. (f) We shall prove that if 80 < y < 100, then good 2 is an inferior good. So suppose that 80 < y < 100. Then 100-100 + 80-100 = 100-120 + Soy 100-120 + 100y > 100-100 + 100-100. Hence the real wealth increases from year 1 to 2. Also the relative price of good 2 decreases. But the demand for good 2, y, decreases because y < 100. This means that the wealth effect on good 2 must be negative. Hence it is an interior good. 2A (a) IE Lg <1, then (py°%)/lpy'¥q) <1 and hence pox) 1, then (p,-x,1/(p,-%p) > 1 and hence Pyx, > Py'Xy- Thus the consumer has a revealed preference for x, Over Xo A. Hence, by taking 2 (©) If py = Ap, and w, = Aw), and x, = x5, then E, Q larger or smaller than one. we can make Eq larger or smaller than one. But this obviously does nol have any revealed preference relationship.2.F.5 We shall first prove the discrete version. By the homogeneity of degree one with respect to wealth, it is enough to show that (p’ ~ p)-Ge(p',1) - x(p,1)) 0 for every p and p’. Since = x(p) = Soe (et PtP = x(p.10) 1 + Ode, Seip) > x0, it is sufficient to show that (p’ - p)-(x(p',p'+x(p,1)) ~ x(p,1)) = 0, and . 1 < @ ~ Prete, Sze TT) > XIPID) = 0. For the firet inequality, note that (p’ = p)-Ge(p',p'+x(p,0) = x(p,1)) = - pexe(p’sp!-aelpt)) + 1 If x(p',p’+x(p,1)) = x(p,1), then the value is equal to zero. If 2(p',p'+x(p,1)) # 2x(p,1), then the weak axiom implies that p-x(p',p’+x(p,1)) > 1, Henee the above value is negative. ‘As for the second inequality, . . 1 (8 ~ PG, Sestpay) ~ tea) ero 1 = prextp) - ——t Pixle. Sepp) > Pl) - Soetegy tt = 2 (pexlpD + — 6D Boxe TT 52-2 Viploxtp. MM Srsteay) 2-2=0. The infinitesimal version goes as follows. By differentiating x(p,aw) = cx(p,w) with respect to a and evaluating at @ = 1, we obtain D..x(p,wiw = x(pyw). Hence 2-7 éSlp.w) = D,x(p,w) + D, xtpywhe(p,w)™ = Dx(p.w) + (1/wlx(pywhelp, 0)" thus Dyxtpaw) = Stpiw) - arm xtpwixtpiw) By Proposition 2.F.2, S(p,w) is negative semidefinite. Moreover, since vs Gelp,whetp,w) Ww =~ (vexe(p,w))*, the matrix - (1/wix(p.whx(p,w)! is also negative semidefinite. Thus D,x(p,w) is negative semidefinite. 2.F.6 Clearly the weak axiom implics that there exists w > 0 such that for every p, p’, and w’, if p-x(p',w") = w and x(p',w') # x(p,w), then p'-x(p,w) > Conversely. suppose that such a w > 0 exists and that p:x(p’,w") = w and x(ptw!) # x(p,w). Let & = w'/w. Then x(p',w') = x(p',ew) = x(a /p',w) by the homogeneity assumption, and p-xta"'p',w) = w and x(a'p',w) # x(p,w). But this implies that (a!p')-x(p,w) > w, or, equivalently, p'-x(p,w) > aw = w'. ‘Thue the weak axiom holds. 2.F.7 By Propositions 2.E.2 and 2.E.3, P-Sip.w) = p-D,x(p.w) * p-D,x(piw)x(pw)" = peDyxlpw) + x(pw)" = 0 By Proposition 2.F.1 and Walras’ law, Ty - Sip.wip = Dyx(p.w)p + D,,x(p.wix(p.w)'p = Dpx(p.w)p + Dyx(p.w)w = 0. 2, pce 2F8 Sq(Pow) = Sopa tah) Piece Ppaeese=r - (pw) + aan (we, (pw) Ep) Op, ¥lewT Bw x ax, PX, (Pw) = ey (Pw) + We et pw) KE ePr xlp.w) ow EydP.W) + Cp (Pr¥ID, (PW)2.F.9 (a) Since x"ATx = (x"Ax)" = xTAx, a matrix A is negative definite if and gnly if xT AK + xTalx <0 for every x € K\(U). Since xtax 4 xTATs = xT(A + AT)x, this is cquivalent to the negative definiteness of A+ AT. Thus A is negative definite if and only if so is A + al ‘The case of negative definiteness can be proved similarly. ‘The following examples shows that the determinant condition is not sient fr the nsec cse. tat A= [4 2, J then Ay = 1 : - “1 and Ayy = 1 But, [4% I[ i ] = 1. Hence A is not negative semidefinite. (®) Let S(p,w) be a substitution matrix. By Proposition 2.F.3, S(p,wip = 0 and hence syo(p,w) = (- Py/pa)ay,(p.¥)- Also p-S{p,w) = 0 and hence 55,(p,¥) = p,/pQ)5,,(P.w). Thus s,(p,w) = (pt/ps)s,,(p.v). Thus, for every v = H v5)" © Now, suppose that S(p,w) ic negative semidefinite ond of rank one. According a 2 2 ,.2),2 VeSEp.We = 5, (P,WILVT = (2Py/PQ),¥> + (BI/PEIND) < = e 2. = 54 (P.wI(¥, > (Y7P2V5) to (*), the negative semidefiniteness implies that <,,(p,w) = 0. Reing of rank one implies that 5,,(p,w) * 0. Hence s,(p,w) < 0. Thus s,p{p.w) < 0. Conversely, let s,,(p,w) < 0, then, by (*), v-S(p,wlv < 0 for every v. 2.F.10 (a) If p = (1,11) and w = 1, then, by a straightforward calculation, we obtain 2-9Hence Sipiw) is not symmetric. Note that =r (pL o =a Shy) * 2 sar - 3V574. 7. =- - Bagi Be 6, 7 ‘Theorem M.D.4(iii), S(p,w) 1s negative semidefinite. (v) Let p= (tj) and w = 1. Let Sip,w) be the 2 x 2 cubmatrix: of Sip,w) obtained by deleting the last row and column. By a straightforward calculation, we obtain ° = 3e Spam = @ + e*[ 25 ® eral Thus, 7 7 1 2, 4, 4, ostam] 4 | =a, tel 1] = e+ ere ater > 0. ° if © > 0 is sufficiently small. Then Stp,w) is not negative semidefinite and hence the demand function in Exercise 2.E.1 does not satisfy the Weak Axiom. 2.F.IL By Proposition 2.F.3, Sip,w)p = 0 and hence s,,(p.w) = (- py/pp)sylp,w). Also p-S(p,w) = 0 and hence sp,(p,w) = (- p,/P9}54,(P.¥)- (We saw this in the answer for Exercise 2.F.9 as well.) Thus s,.(P,w) Spy(PW) 2.F.12 By applying Proposition 1.D.1 to the Walrasian choice structure. we know that x(p,w) satisfies the weak axiom In the sense of Definition 1.C.1. By Exercise 2.F.1, this implies that x(p,w) satisfies the weak axiom in the sense of Definition 2.F.1.2.F.13 [First printing errata: In the last part of condition (*) of (b), the inequality p-x > w chould be p'-x > w’. Also, in the last part of (c), the ~ relation x’ € x(p.w) should be x’ ¢ ¥(p,w). (a) We say that a Walrasian demand correspondence satisfies the weak axiom if the following condition ic satiefied: For any (p,w) and (p’,w'), if x € 2(b.w), x © x(p'yw'), pix = w', and p>x’ = w, then x" € x(p,w). OF equivalently, for any (p,w) and (p',w"), if x € x(pyw), x’ © alps), pox’ = w, and x’ € x(p,w), then psx > w! (b) If x © x(p,w), x’ € x(p',w’), and p-x’ < w, then x’ € x(p,w) by Walras’ law. ‘Thus p'-x > w'. (©) If x € x(p,w), x’ € x{p'w'), and p'sx = w', then (p! = p)-(x’ = x) = w= p-x'. If, furthermore, x’ « x(p,w), then Walras’ law implies that px’ = w. Hence (p' - p)-(x’ - x) = 0, If, on the contrary. x’ € x(p.w), then the generalized weak axiom implies that p-x' > w. Hence (p' ~ p)-(x' - x) <0. « (@) It can be shown in the same way as in the small-type discussion of the pranf af Proposition 2.F.1 that, in order to verify the assertion, it is sufficient to show that the generalized weak axiom holds for all compensated price changes. So suppose that x € x(p,w), x’ ¢ x(p'.w"), p'-x = w', and p+x’ Sw. Then (p' - p):(x' - x) = w - p-x' = 0. Hence, by the generalized compensated law of Demand, we must have (p' ~ p):(x' = x) = 0 and x’ € x(p,w). 2.F.14 Let p >> 0, w > 0, and @> 0, Since p-x(p,w) = w and (ap)-x(ap,aw) = aw, we have ap-x(p.w) = aw and p-x(ap,aw) = w. The weak axiom now impl au &that x(p,w) = x(ap,aw), 2.F.1S Since ax,(p,w)/ew = 0 for both £ = 1,2, we have sp(p,w) = ax(p,w)/0p, for both & = 1,2 and k = 1,2. Hence, let Stp,w) be the 2 x 2 submatria of Slp.w) obtained by deleting the last row and column, then S(p,w) = [ool 2} ri mate te seative seit tense wy, a ‘ “1 - Pau -ee- Wy, - r2Lo -1 4 ly te 7 %2 (We saw this iu the answer to Exercise 2.F.10(a).) Hence, by Theorem cee] 72) - av5/4. M.D.4liii), v-S(p,w)v <0 for alll v not proportional to p. Since Sip,w) is not symmetric, Sip.w) is not symmetric either. 2.F.16 (a) The homogeneity can be checked as follows: X,lop,aw) = apap, = Pp/Py = X{(P.W)s x,lap.aw) = ~ apy/ap, = - By/P3 ~ Xp(Pw) x,lap.aw) = aw/ap, = W/py = x3(P,W). As for Walras’ law, P,x,lb.09 + Poxplp.w) + PaXa(P.W) = (PP. ~ PyPy + PaW)/Py = W- (b) Let p = (2,0, w = 1, p’ = (11), and w’ = 2, then x(p.w) = (2, 1, 1) and x(p'w') = (1, - 1, 2). Thus p'+x(pyw) = 2 = w! and p-x(p!,w') = 1 = w. Hence the Weak Axiom is violated. (©) Denote by Dx(p,w) the 2 x 2 submatrix of the Jacobian matrix Dx(p,w) obtained by deleting the last row and column, then oe sat vii.w = ange! 4] Let Sip.w) be the 2 x 2 submatrix of S(p,w) obtained by deleting the last row 2-12and column, then S(p,w) = Dx(p,w) = wel aa ] because ax,(p,w)/aw = éxpip,wi/ow = 0. Note that v-Sipwiv = 0 for every ¥ € R°. Now let ve R 3 Note that v = (v - (vg/P3)P) + (v3/P5)P and the third coordinate of v - (vg/Pq)P is equal to zero. So denote its first two coordinates by v € R°, ‘Then, by Proposition 2.F.3, v-S(p.wiv = ¥-S(p,wiv = 0. 2.F.17 (a) Yes. In fact, xylapow) = aw/Zyap,) = w/SyPy) = x,(p.w) () Yes, In fact, p-x(p.w) = Dex (p.w) = Dppw/(Spp)) = w. (©) Suppose that p’-x(p,w) = w’ and p-x(p',w') $ w. The first inequality implies that (Eypj)w/(Eyp)) + w', that is, w/(fyp,) 3 w'/(Dypp)- The second inequality implies similarly that (J,p))w'/(Syp)) = w, that is, w'/(Sypi) = WAQypy). Therefore w/(Eyp)) = w'/(Sypj). Hence x{p,w) = xfp',w'). Thus the weak axiom holds. % (@) By calculation, we obtain eee 2, Dyelp,w) = (= w/Typg))} : ala ieee) D, (Pw) = (W/T,p,) = x(p.w). Hence Sip,w) It is symmetric, negative semidefinite, but not negative definite, 23 &CHAPTER 3 3.B.1 (a) Assume that » is strongly monotone and x >> y. Then x ® y and x # y. Hence x > y. Thus > is monotone. (b) Assume that x is monotone, x € X, and © > 0. Let e = (1,..41) € RU and y= x+(cMD)e. Then ly ~ xls © and y » x. Thus » is locally nonsatiated. 3.B.2 Suppose that x >> y. Define € = Min Gx = Yyy oir my ~ yp) > ©, then, for every 2 € X, if ly - zl < e, then x >> z, By the local nonsatiation, there exists z* ¢ X such that lly - z*ll < © and z* > y. By x >> 2* and the weak monotonicity, x > 2*. By Proposition 1.B.1(iii) (which is implied by the transitivity), x > y. Thus > is monotone r ) 3.B.3 Following is an example of a convex, locally nonsatiated preference relation that Ie not monotone in 2. For example, # >> y but J > 0 x3.C.1 Let » be a lexicographic ordering. To prove the completeness, suppose ‘that we do not have x > y. ‘Then "y, > x," : 7 Hence and “x, # y, or yp > x," EAL and ¥9 > x, either *y, 7 x)" oF Thus y x x. To prove the transitivity, cuppose that x > y and y x 2. Then x, > yy and y, = 2. Hence x, & 7). If x, > 2), then x xz. If x, = 2), then x, = ¥, = 2, Thus x, = yp and yp =z. Hence x, = zp, Thus x > 2. To show that the strong monotonicity, suppose that x = y and y # x. This , or that x, = y, and x, > yp. In implies either that x, > y, and x, > y, 7 either case x > y. To show the strict convexity, suppose that y > x, 2> x, y #2, and ae (0.1). Without loss of generality. assume that x # y. By the definition of the lexicographic ordering, we have either "y, > x," or "y, = x, and y, > x)". On the other hand, since z » x, we have either "z, > x," or "2, = x, and x, = Cv Jor” Hence, we have either “ay, + (1 - wz, 7 x," oF “ayy + (L~ az = xy and ayy + (1 - az, > xy." Thus ay + (1 - az > x. 3.C.2 Take a sequence of pairs (x) y"))"_, such that x" > y" for all n, x" + x, and y® + y, Then u(x") © uly") for all n, and the continuity of ul-) implies that u(x) = u(y). Hence x » y. Thus x is continuous. 3.€.2 One way to prave the assertion is to assume that > is monotone and notice that the proof actually make use only of the closedness of upper and lower contour sets. Then the proposition is applicable to x, implying that it has a continuous utility function. ‘Ihus, by Exercise 3.C.2, » 1s continuous. A more direct proof (without assuming monotonicity or using a utility function) gocs as follows. Suppose that there exist two sequences (x"} and 32fy") in X such that x" > y" for every n, x" > x €X, y" > ye X, and y> x. Since (2: ¥ > 2} is open, there exists a positive integer N, such that y > x” for every n> Ny. Since (2: 2 > x) is open there exists a positive integer N, such that y" > x for every n> Ny. Conceivably, there are two cases un the Ly: sequence (y"): Case 1: There exists a positive integer N such that y" » y for every n > N, Case 2: There exists a subsequence (y“) such that y > y* for every n. If Case 1 applies, then, by Propusitivi 1.B.1(iii), we have y" > x” for every n> Max (N.N3). This is a contradiction. If Case 2 applies, then there exists a positive integer m such that k(m) > Ny Since {z: z > yet), is oven, there exists postive Integer Ny such that gP > 8) top every > Ngo By x® k(n) ly xy" and Proposition 1.8.11), x > y em) for every n> Ny. Since (2: z x y""™) is closed, x x ¥ But, since k(m) > Np, this is a contradiction. .C.4 We provide two examples. The first one ie cimpler, but the second one satisfies monotonicity, which the first does not. Example 1. Let X- R, and define u(+): R, > R by letting u(x) = 0 for * <1, u(x) = 1 for x > 4, and uli) be any number in [0,1]. Denote by » the preference relation represented by u(-). We shall now prove that x is not continuous, In fact, if u(1) > 0, then consider a sequence (x) with x = 1 ~ Wn for every n, Although x" ~ 0 for every n and x" 9 1, we have 1 > 0. If u(1) <1, then consider a a sequence {x"} with x” = 1 + 1/n for every n. Although x" ~ 2 for every n and x" 4 1, we have 2 > 1. Note that if u(x) = 0, then all lower contour sets are closed. If u(1) = 1, then all upper contour sets are closed. 3-3Example 2, Take X ~ R2 and define a utility function u(-): RZ > R by the following rule: Case 1. If x +% = 2 and x # (1,1), then u(x) = Xt XD Case 2. If mintx,, oh = 1 and x # (1,0), then ulx) = min{x,,X,) +2. Case a. If x, + x, > 2, mints, x,) <1, and x, > X,, then u(x) = 3 - O - On, - 0). Case 6 + x, > 2, minty, x) <1, and x, < x5, then u(x) = 3 - (1 - xx, - 2. Case 5. u(i,1) € [2,3]. ‘The indifference curves of the preference relation » represented by u(-) are desi ‘bed in the following picture: cy 2 1 == oo 0 1 2 x Figure 3.0.4 It follows from this construction that u(-) is continuous at every x # (1,1). ‘The preference » is convex and monotone. But, whatever the choice of the value of u(l,1) is, it cannot be continuous at (1.1). In fact, (= 1/a, 1 - 1/n) > G1) and (+ 1a, 1+ W/m) > (0) as n> @, and u(l = a, 1 /n) = 2 - 2/0 9 2 3-4ull + In 1+ Yn) = 1+ in +293. Henes, if 2 < ullsl)s then (2,0) x 1. ~ W/m, d= a/m) but (141) » (2,015 If ul) <3, then (1 + In, 1+ 1/n) » (2,1) but (21) > G41). If uC = 3, ‘then all upper contour sets of » are closed: if u(l,1) = 2, then all lower contour sets of » are closed. 3.C.5 (a) Suppose first that u(+) is homogeneous of degree one and let « = 0, xen, yent, and x~y. Then ulx) = uly) and hence au(x) = auly). By the homogeneity, ulax) = ulay). Thus ax ~ ay. Suppose conversely that » is homothetic. We shall prove that the utility function constructed in the proof of Proposition 3.C.1 is homogeneous of degree one. Let x € Ry and «> 0, then ulxle ~ x and ulaxle ~ ax. Since » is homothetic, au(x)e ~ ax. By the transitivity of ~ (Proposition 1.B.1(ii)), ulcxrde ~ cu(xde. Thus u(exx) = aulx). (b) Suppose first that x is represented by a utility function of the form u(x) 4 OOK). Leta eR xe RE, ye RY, and x ~ y, Then ulx) = uly) and hence u(x) + @ = uly) + a. By the functional form, Ua) + 6 = (@ + Hy) + Hayy a) = ulx + ae), uly) +e = (@ +) + BID) = OLY + 20), where € = (1,0.....0) € RE. Hence u(x + ae,) = uly + ae,). or x + ae, ~ ¥ + ae, ‘Suppose conversely that > is quasilinear with respect to the first commodity. The idea of the proof of this direction is the same as in (a) or Proposition 3.6.1, in that we reduce comparison of cammadity bundles on a Tine by fining out indifferent bundles and then assigning utility levels along the line, But this proof turns out to exhibit more intricacies, partly because it 35depends crucially on the connectedness of RL”), which appears in X = (- @, @) xR} (Connectedness was mentioned in the first simall-lype discussion i the proof of Proposition 3.C.1.) The proof will be done in a series of steps, First, we show that comparison of bundles can be reduced to a line parallel to ej. Then we show that the quasilinearity of » implies the given functional form. Let x be a quasilinear preference and a utility function u(-) represent x. ‘The existence of such a (+) is guaranteed by Proposition 3.C.1, but, of course, it need not be of the quasilinear form. For each x € RL, define 1G) = (uGc8) © RE x, © R), then 1G) is a nonempty open interval, by the continuity and the strong monotonicity of 7 along ey, Step i For every xe RU andj € RE", if 16K) # 169), then 18) 9 1G) 2 Proof: Suppose that IG) # (7). Without loss of generality, we can assume that there exists u € I(x) such that u ¢ Ily]. Then either u = sup G7) or u + infl(G). Suppose that u = euplly). (The other case can be treated similarly.) Then let xf € R satisfy u = Glxt.x), then, for every y, € R. Ge.) > (y9). In particular, for every x, © Rand yy © Ry Gtk) > (y, - +24, 9). By the quasilinearity, this implies that (x,,%) > yD. Thus GO.8) > G9). Hence 1G0 0 19) = &. For each x € RE, define E(x) = Je RE! 1) = 1G))- Step 2: For every x € RL, EGX) is open in RL Proof: Let Ke RL, x, €R, and u = lx, x) € IG), Let © > 0 satisty (u- ec, u +e) ¢ 1G). Since ii{+) is continuous, there exists 5 > 0 such that it Fe WL and ux ~ yu <6, ten 1dlxx) ~ Glxy) < ©. Hence 3-61G) AI > G@- eure alee L-1, Thus, by Step 1, 16) =). or Fe EG. tence G eRe! ak - ju <8) ¢ E(x). Thus E(x) is open. Step 3: For every xe RE, BG ~ a, Proo; It is sufficient to show that for every xe RL and ye RE, we nave FU) = EG). Suppose net, then there exit Se RET and 3 RET auch that Elo) # E(y), then the complement RL“\E(x) is nonempty. By Step 1, RENEGH) is equal to the union of those E(3) for which ye RL NEG:. By Step 2, this implies that RL“\E(E) is open. Hence we have obtained a pertition (EGO,RE NEGO) of RL, both of whose elements are nonempty aud ‘open. This contradicts the connectedness of RL! Hence Else) = Ely) for every xe RL and jy RE By Step 3, I(x) = 1(0) for every xe RL! Thus for every xe RU, there exists a unique @ € ® such that oe, ~ (0.3). Detine ¢: wl (0,n) for every x € RL. Define us X > R Uy ula) = x, + Kye: 3K by alte, ~ ‘for every x eX Step 4: The function u(-) represents x. Proof: Suppose that x ¢ X, y € X, and x > y. By the quasilinearity, this is equivalent to (x, - Ys Xpyeees Xp) X (O,Yyyeeu¥,)- By the definition of 66), Chis ix equivalent 0 (x) = yy) Xp xp) 5 O09 yp Je Again by Je, Again ets r (Oxo XL) x Qa) + ¥, — Hey Again by the definition of ¢(+), this is equivalent to OK yooor MIO_ 2 HY gH -orVE) + Hy ~ MDOy- Hence PX p.04 Xp) = Oy ed_) + Yy ~ Xp that is, ulX) = UG. 3-7‘These properties of u(-) are cardinal, because they are not preserved under some monotone transformation, such ax Flu(a) = ufal? 3.6.6 (a) For p = 1, we have u(x) = a,x, + a,x,. Thus the indifference curves are linear. (b) Since every monotonic transformations of a utility function represents the same preference, we shall consider toa = - (a,x? + aagx?) Gx) = Inulee) = (17plin(eryxf) + a.yx8). By L’Hopital’s rule, lim U0) xd = lim (a, +a, ax + a Hin ting, + aime vin + 08) = (ayInx, + a,Inx,)/(a, + a). is Seen! . Since expl(a, + @,Vilse)) = x,'x,°, we have obtained a Cobb-Douglas utility function There is an alternative proof to this proposition: Since both the CES and the Cobb-Douglas utility functions are continuously differentiable and homothetic, it is sufficient to check the convergence of the marginal rate of substitution at every point. The marginal rate of substitution at (x,,x)) with respect to the CES utility function is equal to @,xf/a,x5"'. The marginal rate of substitution at (x,,x,) with respect to the Cobb-Douglas uullity function is equal to a,x,/a%,, Note that «x2 "'a,x2" » a,x,/2% as p>. (in tact, «x "lagx8°l ts well defined for every p and is equal 0 @,x,/a,x, when p= 1.) The proof is thus completed. Strictly speaking, there is a missing point in both proofs: We proved the convergence of preferences on the strictly positive orthant (x € R: x >> 0),but we did not prove the convergence on the horizontal and vertical axes. In fact, the convergence on the axes are obtained in such a way that all vectors there tend to be indifferent. To be more specific, compare, for example, x = (x,.0) and y = (y;.0) with x, > y, > 0, According to the CES utility function, x is preferred to y, regardless of the values of p. But, according to the Cobb-Douglas utility function, x and y are indifferent. Futhermore, the following is true: If x is in the strictly positive orthant and y is on an then x is preferred ta y for every p sufficiently close to 0. To see this, simply note that if x = (x,,0) with x, > 0 and y >> 0, then ax?’ > and ay? + a,y) > a, +a. The implication of this fact is that, as p > 0, every vector in the strictly positive orthant becomes preferred to all vectors on the axes. ‘hat is, unconditional preference towards strictly positive vectors tends to hold, as it is true for the Cobb-Douglas utility function. (©) Suppose that x, =x, We want to show that Bg hile x pee a,x5) <0, Since x, = 0 and x, = 0, we have ax? sain? + a ws Let <0. Since xj #0 and) #0, we have ay? sax? + aps Th a/Px, = (a,x? + ayx6)”P, On the other hand, since x, = x,. Pe ugh sax? + ux? = a + mL te ART + Hy Mey + abe Hence (a,x + a,x6)/P = (a, + 0,)/Px,. Therefore, Ven = + aya? & (a, + al”, a Px, = (axl + aad)? & (a, + a), Letting p * - e wovobtain Lim (ax? + anxPi!/P = x, because Lim «, "Px, pre pre lim (a, + a,)?; prot 2 3.D.1 To check condition (i), X,(Ap,AW) = a(AWIZ(Ap,) = aw/p, = x\(P.¥),x2{Ap-AW) = (I~ @(AWIZ(Ap,) = (1 = alW/p, = xp(p.w). To check condition (ii), Pyx(b.W) + poxylpow) = paw/p, + poll ~ adwrp, Condition (iii) is obvious. 3.D.2. To check condition (i), v(Ap,Aw) w eane + (1 ~ alin(L ~ @) + Inaw - aindp, ~ (1 - adinap, = alna + (1 - @)In(1 - @) + Ind + Inw = ainda - alnp) - (1. - @)ina - (1 = «inp, = a@lna + (1 - @)in(1 - a) + Inw - alnp, ~ ae inp, = vip,w). To check condition (ii), Ov(p,w)/dw = I/w > 0, avlpiwl/ap, ~~ a/p, < 0, avip.wi/ap, = - 1 - ap, < 0 Condition (iv) follows the functional form of v(-). In order to verify (iii), by property (i), it is sufficient to prove that, for any v € R and w > 0, the set {p € RL: vip,w) ¥ v} is convex. Since the logarithmic function is concave, the set 2 2 AippPp) € RL: - adnp) ~ (1 - @llnp, = v) is convex for every v € K. Since the other terms, aina + (1 - @in(l - a) + Inw, do not depend on p, thi implies that the set (pe RE: vipw) # ¥) is convex, 2.0.3. (a) We shall prove that for every pe RL, w= 0, a0, and xe Rt, ifoe(pyw), then ax = x(pyaw). Note first that p-(ax) * aw, that is, ax is affordable at (p.awl. Let y € Ry and py = aw. Then p-(a™'y) = w. Hence ula™y) = u(x). Thus, by the homogeneity, uly) = ulax). Hence ax = x(p,aw). By this result, V(paw) = ube(p,aw)) = uCare(p,w)) = aulse(pw)) = av(p,w). Thus the indirect utility function is homogeneous of degree one in w. Given the above results, we can write x(p,w) = wx(p,1) = wi(p) and v(p,w) = wv(p,1) = w¥(p). Exercise 2.£.4 showed that the wealth expansion path (elpwh: w > 0) is a ray going through slp). The wealth elasticity of demand Egy 5 equal to 1 (b) We first prove that for every p © RE iy WE 0, and a = 0, wo have x(paw) = cxc(p,w). In fact, since v(-,+) is homogeneous of degree one in w, V,v(pycw) = av,vip,w) and V,,v(p,ew) = Vv(p,w). Thus, by Roy’s identity, x(p,aw) = axc(p,w). now tet ce WE, x 6 WE, uO = ule), and «= OL Since uC) is strictly quasiconcave, by the supporting hyperplane theorem (Theorem M.G.3), there exist pe #L,, pie 6, w > 0, and w' > O euch that # = xlp.w) and x = x(p’,w'). Then u(x) = v(p,w) and u(x’) = v(p’,w’). Hence v(p,w) = v(p’.w"). Thus, by the homogeneity, vip,aw) = v(p',aw’). But as we saw above, x(p,aw) = ax and x(p',aw") = ax’. Hence vip,aw) = ulax) and vip’aw") = ulax'). Thus ulax) = ulax'). Therefore u(x) is homogeneous of degree one. 3.D4 (a) Let ¢, = (1,0,...,0) € RY. We shall prove that for every p € RL. WER GER, and xe (- 0, 0) x RL, if x = xip,w), then x + ae, xp, w + a). Note first that, by pj = 1, p-(ax + e,) < aw, that ic, x + a8, is atfordable at (p, w + an). Let y eR and pry sw +a. ‘Then poly = a) Bell= w. Hence x zy acy. Thus, by te quasilinearity, x + ae, > y. Hence . x + ae, = xlp, w + ad. Therefore, for every € €(2....L wR, and w © R, x4(p,w) = JL are x (p,w'). That is, the Walrasian demand functions for goods 2, independent of wealth. As for good 1, we have dx(p,w)/aw = 1 for every (p,w). That is, any additional amount of money is spent on good 1. (h) Define #(p) = ulx(p,0)). Since x(p,w) ~ x(p,0) + we, and the preference 1 relation can be represented by a utility function of the quasilinear form u(x) xX, + Ulx,,...%,) (Exercise 3..5), we have v(p.w) = ulx(p,w)) = x\(Rw) + Be fP,W), 26, (BAW). (p,.0) w + x4(P,0) + Ule,(p,0), w + ulx(p,0)) = w + g(p). (c) The non-negativity constraint is binding if and only if PpX5(P,0) >w, Note that x,(p,0) = (n')"“(p,), because p, = 1. Hence the constraint is binding i and only if p(n')"(p,) > w. If s0, the Walrasian demand is given by x(p.w) = (O.w/ps). Thus, as w changes, the consumption level of the first ood is unchanged and the consumption of the second good changes at rate 1/p) with w until the non-negativity constraint no longer binds. 3.0.5 (a) Since any monotone transformation of a utility function represents the same preference relation, we may as well choose Tio = pute” = ple? + x) iG) = pute? = ple? + xf). By the first-order condition of the UMP with ii(-), 3-1 xtpw) = (wtp + ponent poh,where 8 = p/(p - 1) € (- @, 1). Plug this into u(-), then we obtain vtpand = Wiig + BBV, (b) To check the homogeneity of the demand function, xlap.aw) = (aw/(ap,)® + (ap,)° (ap) lap,)°4) = ea eS yiwrtp? + pS ytp2 haps (aca? "a Kwtpy + pp)of pg) = x(p.w) To check Walras’ law, 38 a1 ae, prx(p.w) = (w/(py * pa)tpy Py + Py"Pa) = W. ‘The uniqueness is obvious. To check the homogeneity of the indirect utility function, vlap, aw) 85 ewra V808 4 528 3/8 3 = aw/(ap,)® + (ap) pai’? = wip? + pa = vip.w) To eheck the monotonicity, avip,wivaw = Vip? + pa) > 0, = wed Med 4 98/84 av(p.w)/apy = - wee p> + pb) ° ‘The continuity follows immediately from the derived functional form. Tn order to prove the quasiconvexity, by property the homogeneity, it is sufficient to prove that, for any v ¢ R and w > 0, the set (p € R* vip,w) $ Wis conver, If 8 then the utility function is a Cobb-Douglas one, and the quasiconcvity was already extabliched in Exercise 3.0.2. So we consider two cases, 8 € (0.1) and 8 £0. Ih elther ease define stp) = (98 + 38)Y®, If 8 € (0,1), then sip)” = p® + p is a concave function. Hence {p € R°: aie Ftp) = (r(p)8)8 z vy Is convex for every v. Since vlpyw) = w/S lp), this implies that {p € R%: v(p,w) < v) is convex for every v and w. 3-13If § <0, then fip)® = p> + p8 is a convex function. Hence {p € R: 81-8) wre) = (rp)? = vids convex for every v. Since vip,w) = w/y(p), this implies that {p ¢ BR: v(p,w) # v) is convex for every v and w. (c) For the linear indifference curves, we have (w/py. 0 i py < py x(p.w) = 4 (0, w/p,) if P, > Py {Cw/p A, =A: Ae (0,11) if py = poi Mipyw) = max(w/p,.w/p,) For the Leontief preference, x, (D.W) = (w/tp,+ py): vip.) = w/(p,* Pp). As for the limit argument with recpect to p. First consider the case with p< Land p31. Then d= p/p -1)>-easp>i. Case. py < Pp. 8 30. This : wp, lim pf lwp? + p8) = 11m ——_. = w/p,. 5 b9-@ 1 + (p,/p,) Since p,/p, > 1, we have (p,/p, since p/p) <1, we have (p\/pp)® > @, Thus : we, tim 8 twu8 + p= 11m ——2— = a 33-0 (p,/pQ)° +1 Thus the CES Walrasian demands converge to the Walrasian demand of the linear indifference curves. As for the indirect utility functions, we showed in the answer to Exercise 2.C.6le) that (p® + p3)/® > yy for py 4 py Heuce the CES indirect utilities converge to the indirect utility of the linear indifference curves. Case 2. py > Py Do the same argument as in the Case 1Case 3. py In this case, (w/(p> + pain" ye) ) = (wp? + prvn(p? 2 (w2p 1). This consumption bundle belongs to the set of the Walrasian demands of the linear indifference curves when pj =p. As for the indirect utility functions, we showed in the answer to Exercise 3.C.6(¢) that 5 joys z (of + py”? » p, for p, = Pp. Let’s next consider the case p > - @. Note that 5 = p/(p - 1) 91 as p > 1, Su just plug 6 = 1 into the CES Walrasian demand functions and the indirect utility functions. We then get the Walrasian demand function and the indirect utility function of the Leontief preference. (@) From the calculation of the Walrasian demand functions in (a) we get - 3-1 X(.W)/x2(P,W) = (Py/P.)° 8-2 p/P) (x (B.WI/x9(P,WII/P/Pp) = 6-2 Ale (W)/xp(eW4dlp,/pp) = (6 = Wp, 49) *- Thus €,,(p.w) = - (8 - 1) = 1/1 - p). Hence, €,>(p,w) = @ for the linear, E,2(p.w) = 0 for the Leontief, and €,,(p.w) = 1 for the Cobb-Douglas utility functions. WatBe) _ a OOOO eG OME 3.D.6 (a) Define G(x) = utx) Ox, = BY Ory = dy Gx, - d4)%, with of = aa + @ +9), B= Pa + B+ 7) v= ya+ B+ 7¥). Then a’ +p” and U(+) represents the same preferences as u(+), because the function u > ul/asey) is a monotone transformation. Thus we can assume without loss of generality that a +B + y= 1. (p) Use another monotone transformation of the glven utility function, Inu(x) = aln(x, - bj) + Bln(x, ~ by) + vinx, - b,).‘The first-order condition of the UMP yields the demand function (pw) = (Vysbyi¥) + Cw ~ prbMarp,,A/R,.¥/7,), where p:b = pjbj + poby + pads. Plug thic demand function to u(-), then we obtain the indirect utility function vip,w) = (w - p-d)a/p,)"8/p,Pta7p,)”. (©) To check the homogeneity of the demand function, X(APAW) = (bybyybg) + (Aw ~ Ap-EG/AP BAP I/A) = (by bgrb) + (w - p-bMa/pyB/pay7/Pq) = x(prw). To check Walras law, p-x(p,w) = pb + (w - p-b\(p,a/p, + p,B/P, + P52/Dq) prb + (w - pba + B+ 9) The uniqueness is obvious. To check the homogeneity of the indirect utility function, vapAw) = Aw ~ Ap-b)la/p)*(e/API arp,” ARDY — p-vya/p JB/pgIta/pq) = (w = p-b\(a/p,)%(8/p)ta/p5)* = vip,w). To check the monotonicity, avipiwi/ow = (a/p,)%(8/p9!ta/y_)" 2 0, avip,w)/apy ~ vipiw)-(- a/p,) < 0, ov(p.w)/ap, = vip.w)-(- B/p,) < 0. 1 avip.wi/ap, = vip,wi-(- ¥/pq) < 0. The continuity follows directly from the given functional form. In order to prove the quasiconvexity, It Is sufficient to prove that, for any v ¢ R and w > 0, the set (p © RY vip.w) = ¥) ia convex. Consider Invip.w) = alna + flnB + giny + In(w ~ p-b) ~ alnp, ~ Blnp, - zinpy. Since the logarithmic function is concave, the set| | \ {p € R°: In(w ~ pb) - ainp, - Binp, - yinp, = v) is convex for every vc R. Since the other terms, alna + Bin8 + yiny, do not depend on p, this implies that the set (p € R°: Inv(p,w) = v) is convex. Hence so is (p € R°: vip.w) = ¥) 3.0.7 (a) Since p!-x°< w! and x! # x°, the weak axiom implies p°-x! > w°, Thus x! has to be on the bold e in the following figure. Figure 3.D.7(a) In the following tour question, we assume the given preference can be a differentiable utility function u(-). (b) If the preference is quasilinear with recpect to the first good, then we can take a utility function ul-) so that au(x)/ax, = 1 for every x (Exercise 3.C.5(b)), Hence the first-order condition implies 6u(x')/axy = p/p; for each t = 0,1. Since p9/p° < pip} and ul-) is concave, xp > x3. Thus x! has to be on the bold line in the following figure.02 4 8 10 5 Figure 3.D.7(b) (c) If the preference is quasilinear with respect to the second good, then then we can take a utility function u(-) so that du(xl/éx, = 1 for every x (Exercise 3.C.5(b)). Hence the first-order condition implies aulx')/axy = tpt since 50/20 > ohyy? 5 Pi/P) for each t = O,1. Since py/PD > pi/p) aud ul-) is concave, we must have Oo. x} pj/Pyy x' has to be on the right side of the ray that goes through x°, Thns x! has to be on the bold line in the following figure.Figure 3.D.7(e) 3.D.8 By Proposition 3.D.3(i), vlap,aw) = v(pyw) for all a > 0. By differentiating this equality with respect to a and evaluating at a = 1, we obtain Vvip.w)-p + wavip.w)/aw = 0, Thus wavip,wi/aw = - U.v(p.w)-p. 3.E.1 The EMP is equivalent to the following maximization problem: Max = px sit. u(x) eu and x20. ‘The Kuhn-Tucker condition (Theorem M.K.2) implies that the first-order conditions are that there exists 4 > 0 and 4 RU such that p = AJu(x*) +p and y-x* = 0. That is, for some A > 0, p = AVu(x*) and x*-(p - ATu(x*)) = 0. ‘This fs the same as that of the UMP. 3.E.2 To check the homogeneity of the expenditure function, ep) = et - 0) ap Map) u T 2 = = aA yy = elp.u). To check the monotonicity, 3-20ae(p,ul/au = @ 1 - al !p tps > 0, aelp.ulvap, = alt = a pip ® > 0, Tey 70. To check the concavity, it is easy to actually calculate Dee(p.u) and then ae(piu)vep, = a1 - a) apply the condition in Exercise 2.F.9 to show that Deto.u) is negative semidefinite. An alternative way is to only calculate Tag. gytpt phe ePetp.ur/epy eal MU - apps < 0. Then note thal the homogeneity implies that DPetp.up = 0. Hence we can apply Theorem M.D.A{lii) to conclude that DZe(p,u) is negative semidefinite, The continuity follows from the functional form. To check the homogeneity of the Hicksian demand function, Iw bea “Ap, “Py hyQpwu) = ue w= hyip), Wap, Ws a, @ - aap, )* Ga - ap, )* nga = | aap |e = | ap | hae To check the no excess utility, ep, 8 = wp, uo = |e] 4) || at} 9 on (-ae-a(1-a) - od yet Trap, The uniqueness is obvious. 3.E3 Let X ¢ RU and uGR) =u. Define A = (x © RE: prx prk and ule) = vu) ‘Then A #@ by Xe A. Furthermore, A is compact: The closedness follows from that of (x ¢ RU: ulx) 2 u) and of RL; the boundedness follows from the inelustan Ac (x eR: 0 = xy = pri/p, for every F= 1,...1) 3-21Now consider the truncated EMP: Min p-x st. xe A. Since p-x is a continuous function and A is a compact set, this problem has a solution, denoted by x* € A. We shall show that this is also a solution to the original Emp. Let x ¢ RU and u(x) 2 u. If x € A, then p-x = p-x® because x* is a solution to the truncated EMP. If x € A, then p-x > p-X and hence p-x > p-x*. Thus x* is a solution of the original EMP. 3.E.4 Suppose first that x is convex and that x € h(p,u) and x’ € h(p,u). Then p-x = p+x’ and u(x) = u, u(x’) = u, Let @ € [0,1] and define x” = ax + (1 = ax’, Then p-x" = ap-x + (1 - a)p-x' = p-x = p:x’ and, by the convexity Of x. u(x") © uw. Ths x” € n(p.u) Suppose next that x is strictly convex and that x € n(p,u), x’ € n(p,u), x # x’, and u(x) 2 ulx') = u. By the argument above, x" = ax + (1 - alx’ with « € (0,1) satisfies p-x" = p-x = p-x’ and. by the strict convexity af », we have x" > x’. Since x is continuous, @x" > x’ for any B € (0,1) close enough to 1. But this implies that p-(8x") < p-x and u(x") > u(x’) 2 u, which contradicts the fact that x is a colution of the EMP. Hence A(p,u) must be a singleton. 3.E.5 [First printing errata: The equality h(p,u) = A(p)u should be h(p.u) = uftp), because u is a etalar and filp) is @ veetor.] We shall first prove that, for every p >? 0, U= 0, @ = 0, and x = 0, if x = Alp,u), then ax = Aipyew). In fact, note that ulax) = au(x) = au, that is, ax satisfies the constraint of the EMP for au. Let y ¢ RL and uly) = au. Then ula” y) = wu, Hence p-(a'y) = p+x. Thus p+y = p-(ax). Hence ax = h(p.aul. Therefore 3-22h(pju) is homogeneous of degree one in u. By this result, e(p.au) = pehlpyau) = p-(eh(pyu)) — a(p-h(p,u)) = ae(p,u) ‘Thue the expenditure function is homogeneous of degree one in u. ih(p) and Now define filp) = h(p.l) and e(p) = e(p,1), then hlp,u) = e(p,u) = uel(p). 3E.6 Define 8 = p/(p = 1), then the expenditure function and the Hicksian demand function are derived from the first-order conditions of the EMP and they are as follows: nips = ug + 98 9G, 3,8 Vs e(p.u) = ulpy + p3 To check the homogeneity of the expenditure function, 878 _ 81/8, elapyu) = ullap,)® + (ap,)°) (p> + po? = aetp.n) ‘1 2 1 2 To check the monotonicity, detp.ul/au = (p> + pb)“ > 0, de(p,ul/ap, = ups ‘ip? + py)? > 0. To check the concavity, it is a bit lengthy but casy to actually calculate ee(p,u) and then apply the condition in Exercise 2.F.9 to show that DZetp.u) is negative semidefinite, An alternative way is to only calculate aelp.ul/ape 8-2) 8, SI/S1 , F1.3s , 8y1/8-2 Bl = us ~ ph (pe + pO* + upl Mwy + pb)? Pave = Dap} 208 5S VE2 8 a = ull ~ arp ip? + pay Pept - (wt + PO) < o, by 6 <1. Then note thal the homogeneity implies that DZetpwudp = 0. Hence we can apply Theorem M.D.A(iil) to conclude that DZe(p,u) is negative semidefinite. The continuity follows from the functional form.To check the homogeneity of the Hicksian demand function, Alap.a) = ullap,)® + (ap,)8"°8”8((ap >"! (apy?) (8(1-8)/8)4(8-1),,3 , ,8)(1-81/8 8-1 8-1 ne ute? + pb) ey Py) = h(pyu). To check no excess utility. utr(pud) = leh + pp IE p/FUe , plF-eyp, Since (8 - 11/8 = - 1/p, we obtain u(h(p,u)) The uniqueness is obvious. 3.E.7_ In Exercise 3.C.5(b), we showed that every quasilinear preference with respect to good 1 can he represented by a utility function of the form u(x) = .0) € RY. We shall prove that for every x1). Let e, = (10 p> Owith p= 1 ueR we R and xe (=, 0) xR it x = Alpwul, then x7 0e, = MP, U+ a). Note first that ux + ae,) = u + a, that is, x + ae, satltes the constraint of the EMP for (pu + al. Let y @ RE and uly) = w+ & Then uly - ae) =u. Hence pely - 90,) = psx. Thus pry = pilx + ae Hence x + ae, = Alp. u + a). Therefore, for every £ € (2,...L), ue R, and v’ © R, hylpu) = ‘hj(pwu'). That is, the Hicksiai: demand functions for goods 2,...,L are independent of utility levels. Thus, if we define A(p) = h(p,0), then h(p,u) = Rip) + ue). Since h(p, u + a) = hlp.u) + ae,, we have elp, u + a) = elpyu) + & 1” Thus, if we define (p)'= e(p,0), then e(p,u) = Sp) + u. BEB We use the utility function uc) = x95". to prove (3.6.1), - Lede gylme ma anl elp(pyw)) = Mt ~ a) Ip tpd Ma - ay! pi IW) = w, na, alge ole Ia vinetp.ud) = ata = ay pi *pS MaMa ~ a Ate) Muy » u. 3-24To prove (3.£.3), & alpre(pyu)) = et - a Hyp) Burlarpy, (1 - 1p) ea a an, Gap, = || eae] [ma] nip.u), a Ina =a a1, ap, Pf a - wp hipyv(pyw)) = aU = a! “pS | | = + 1 2 = wla/py, ( ~ @/p)) = x(pw). 3.E.9 First, we shall prove that Proposition 3.D.3 implies Proposition 3.b.z via (3.E.1). Let p >> 0, p' >> 0, ueR, v' eR, and ae 0. i) Homogeneity of degree one in p: Let @ > 0. Define w = e(p,u), then u = vipw) by the second relation of (2.E.11. Hence elapyu) = elap,v(p,w)) = elap,vlapaw)) = ew = ae(p,u), where the second equality follows from the homogeneity of v(+.+) and the third 0 from the first relation of (3.E.1). (ii) Monotonicity: Let u’ > u. Define w = e(p,u) and w’ = e(p,u’), then u = vip,w) and u' = vip,w'). By the monotonicity of v(-,-) in w, we must have w" > w, that is, e(p'u) > e(p,u) Next let p' = p. Define w = e(p,u) and w’ = e(p',u), then, by the second relation of (3.E.1), u = v(p.w) = v(p',w'). By the monotonicity of v(-, must have w’ = w, that is, e(p',u) = e(p,u). (iii) Concavity: Let & € [0,1]. Define w = e(p.u) and w’ = e(p’,u), then n= vip,w) = v(p',w). Define p" = ap + (1 - ap" and w" = aw + (1 - aw’, Then, » by the quasiconvexity of v(+,+), vip"sw") (piu) and any w, if e(p au") = w for every n, then e(p,u) s w; if e(p"u") = w for every n, then e(p,u) = w. ‘Suppose that e(p",u") = w for every n. Then, by the monotonicity of v(+,-) in w, and the cecond relation of (3.E.1), we have u" < vip",w) for every a. By the continuity of v(-,-), us vip,w). By the second relation of (3.E.1) and the monotonicity of v{-,-) in w, we must have e(p,u) < w. The same argument can be applied for the case with e(p".u") = w for every n. Let’s next prove that Proposition 3.E.2 implies Proposition 3.D.3 via (3.E.1), Let p >> 0, p' >> 0, we R, w ER, and a= 0, (i) Homogencity: Let @ > 0. Refine 0 = v(p.w). Then, by the first relation of (3.E.1), e(p,u) = w. Hence vlap,aw) = v(ap,ae(p,w)) = v(ap.elap.u)) vip.w), where the second equality follows from the homogeneity of e(+,-) and the third fram the second relation of (3.E.1). Kw (ii) Monotonicity: Let w' > w. Define u = v(p,w) and u' = v(p,w'), then. = w and e(pju’) = w". et. By the monctonicity of e(-,-) and w' > w, we must have u’ > u, that is. v(p.w') > v(p.w). Next, assume that p’ = p. Define u = v(p,w) and u’ = v(p'.w), then e(p,u) = elp'ju') = w. By the monotonicity of e(-,-) and p’ = p, we must have ul su, that is, v(pyw) = v(p',w), (iii) Quasiconvexity: Let a € [0,1]. Define u = v(p,w) and u' = v(p',w'). Then e(p,u) = w and e(p,u') = w’. Without loss of generality, assume that u” =u. Define p" = ap + (1 - a)p' and w = aw + (I - aw’. Then elp"u’) = ae(p.u') + 1 ~ alelp’= ae(p,u) + (1 - ade(p’,u') saw + (1 - aw = wy where the first inequality follows from the concavity of e(-,u), the second from the monotonicity of e(-,+) in u and u' 2 u, We must thus have v(p",w") < w (iv) Continuity: It is sufficient to prove the following statement. For any a sequence ((p"™,w°)"_, with (p".w") > (p,w) and any u, if vipw") = u for every n, then v(p,w) 3 u; if v(p",w) > u for every n, then vipw) =u Suppose that vip",w') = u for every n. Then, by the monotonicity of e(+,-) in u and the first relation of (3.E.1), we have w" 3 e(p"u) for every n. By the continuity of e( ++), ws e(pu). We must thus have v(p,w) = u. The same argument can be applicd for the cace with vip",w") ® u for every n. ‘An alternative, simpler way to show the equivalence on the concavity/ quasiconvexity and the continuity uses what is sometimes called the epigraph. For the concavity/quasiconvexity, the concavity of e(+,u) is equivalent to the convexity of the set {(p,w): e(p.u) = w) and the quasi-convexity of VG) 1s equivalent ta the convexity of the set (pw): v(pw) = u) for every But (3.E.1) and the monotonicity imply that v(p,w) = u if and only if e(p.u) > w. Hence the two sets coincide and the quasiconvexity of v(-) is equivalent to the concavity of e(+,u). As for the continuity, the function e(+) is continuous if and only if both {(p,w,u): elp.u) + w) and ((p,w,w): e(p,u) = w) are closed sets. The function v(+) is continuous if and only if both {{p,w.u): vip») = u) and {(p,w,w): v(p.w) Su) are closed sets. But, again by (3.E.1) and the monotonicity, 3-27{(p.wwuh: etp.u) = w) = (p.W,w): vip.w) = Ur; A(p,w,u): efp.u) = wh = Upwyul: vipw) ud. Hence the continuity of e(+) is equivalent ta that of v(-) 3.6.10 [First printing errata: Proposition 3.f.4 should be Proposition 3.6.3.) Let's first prove that Proposition 9.D.2 implies Proposition 3.£.3 via the relations of (2.6.1) and (3.6.4), Let p ¢ RL and ue R. (1) Homogeneity: Let « > 0. Define w = e(p,u), then u = v(p,w) by the second relation of (3.E.1). Hence Alap,u) = xlap,elap,u)) = x(ap,ce(p,u)) = x(p,e(p,u)) = Alpu), where the first equality fellows from by the first relarfon of (3.E.4), the Second from the homogeneity of e(-,u), the third from the homogeneity of x(+.), and the last from by the first relation of (2.£.4) (ii) No excess utility: Let (p,u) be given and x € h(p,u). Then x € x(p.e(p,u)) by the first relation of (3.6.4). Thus u(x) = v(p,e(p,u)) = u by the second relation of (3.E.1). Convexity/Uniqueness: Obvious Let's first prove that Proposition 3.F.2 implies Proposition 2.0.2. via the relations of (3.E.1) and (3.6.4). Let p ¢ RU, and we R. (i) Homogeneity: Let « > 0 and define w = e(p,u), then vip,w) Hence xlap,ew) = hlap,viap,aw)) = hlap,v(p.w)) = hip,v(p,w)) = x(p,w), where the firet equality follows from the second relation of (3.E.4), the second from the homogeneity of v(-}, the third from the homogeneity of h(+) in b, aud the last from the first relation of (3.E.4). (ii) Walras’ law: Let (p,w) be given and x € x(p,w). Then x € hlp,v(p,w)) by 3-28the second relation of (3.6.4), Thus p'x = e(p,v(p,w)) = w by the definition of the Hicksian demand and the firet relation of (3.E.1) (iii) Convexity/Uniqueness: Obvious. 3.F.1 Denote by A the intersection of the half spaces that includes K, then clearly A > K. To show the inverse inclusion, let X ¢ K, then, since K is a closed convex set, the separating hyperplane theorem (Theorem M.G.2) implies that there exists a p * 0 and c, such that p-X <¢ < p-x for every x € K. Then (2 € RU: p-z = c} is a half space that includes K but does nol contain x. Hence X € A. Thus K > A. 3.2 If K is not a convex set, then there exists x ¢ K and y € K such that (/2)x + (W/2ly € K, as depicted in the figure below. The intersection of all the half-spaces containing K (which also means containing x and y) will contain the point (1/2)x + (1/2)y, since half-spaces are convex and the intersection of convex sets is convex. Therefore, the point (1/2)x + (1/2)y cannot be separated from K. x (iayx+ (Lay K y Figure 3.F.2 3-29As for the second statement, if K is not convex, then there exist x € K, y €K, and @ ¢ (0,1) such that ax 4 (1 - aly € K. Since every half space that includes K also cuntains ax + (1 - ay, it cannot be separated from K. 3.G.1 Since the identity v(p,e(p,u)) = u holds for all p, differentiation with respect to p ylelds V,v(p.e(p.u)) + (6v{p,e(p,u))/awI¥, (Pu) By Roy's identity, (ovlpre(p.u)/owN(- x{p.e(p.u)) + Vpe(p.u)) = 0. By avip.e(p.u))/ew > 0 and hip,u) = x(pre(p.u)), we obtain hip.u) = VLe(P.u). 3.6.2 From Examples 9.D.1 and 3.E.1, fur the utility function. u(x) = we obtain wp, Pyste = 1a alvp, | ~ aw/pe ° D,xlp.w) = 2 a 0 ~~ awrpy a a we; Velpu) = ulp,Za)(py/lt ~ a))® [ ne ep |: D,elpwu) = Drip, pel, ) D ple. u) ‘ , e og | - x - ae ell ~ @/ppp, = ulp, 20)p, iL = at iad v Hae) zi ZI el ~ ec)/y = all = a7 a PiP2 Pe ‘The indirect utility function for u(x) = xh is vip,w) = (p,Za)%(pg/t1 ~ a) Ww. (Note here that the indirect utility function obtained in Example 3.U.2 is Yor the utility function u(x) = alux, + (1 - a@)lnxy.) Thus U,MPLw) = vipwi= apy ~ (I~ ap)¥viB.w) = vip, w/w. 8 Hence: h(p,u) = Vgetp,u), Detp.u) = D,h(p.u), which is negative semidefinite and symmetric, D,hlpwulp = 0, D,plpsu) = D,xtp.w) + D,x(e.w)xtp.w)", xplb.w) ~~ (Bv(p.u)/ap )Aav(p.u)/ow) 3.6.3 (a) Suppose thal a +B + 4 Note that Inu(x) = aln(x, ~ b,) + Bln(x, ~ b,) + rinlx, ~ v4). By the first-order condition of the EMP, Alpwu) = (by byybq) + ulP,/)(Py/8 tp g/2)"(a/P PP 2 9/9) Plug this into p-h(p.u), then we obtain the expenditure function elpu) = pod + ulp,/a)"(pp/)(Py/n)”, eo where p-b = pib, + Pzb, + Pqdy To check the homogeneity of the expenditure function, e(apwu) = Ap-b + uApy/al*ap,/ePapy/)™ = Ape + Aulp/o%™p,/Bipy/a" = retp.u). To check the monotonicity, assume b, © 0, by = 0, and by = 0. Then delp.ulau = (p,/a)(p,/AIpa/1)* > 0, aelp,urvap, = 0, + prep > 1 + UD Zep BIC 4/7)" arp)) > 0, by + ulpy/a%v,/BPtPy/"(0/PQ) > 0, : 2e(p.vl/0p, Belpiul/apg = by + ulp,/al“(p,/AP(ps/a ap) > 0 To check the concavity. we can show that Dye(p.u) is equal to 331all -a/pt — 08/p,p, 23/02, ulp/a(p,/8(py/e"| —aB/p,p, BL - B/S By/ppPy & «9/P,P BY/P2P, -ra- nA, nd then apply the condition in Exercise 2.F.9 to show that Dee(pu) is negative semidefinite. An alternative way ic to only calculate the 2 x 2 submatrix obtained from Dee(p.u) by deleting the last row and the last column and apply the condition in Exercise 2.F.9 to show that it is negative definite, Then note that the homogeneity implies that De(p,ulp = 0. Hence we can apply Theorem M.D.Alii) to conclude that DZe(p,u) is negative semidefinite. The continuity follows from the functional form. To check the homogeneity of the Hicksian demand function, Opa) = (bybyby) + lap 70) apy/BIP(Apy/2)a/Ap B/APy3/APg) = (by barby) + "PF 70)%(p,/8P i g/7)"(a/P18/Py19/Pg) nip,u) To check no excess utility, & UinprU)) ~ ulp,/a)™(Py/P (pga Me’ B/P.)a/py) = The uniqueness is obvious. (b) We calculated the derivatives e(p,u)/ap, in (a). If we compare them with (P,w), then we can immediately see de(p,u)/ap, = hy(p,u). he (e) By (b), Dyhtpwul = DZe(pul, In (al, we calculated DBe(p.u). In Exon 3.D.6, we showed x(p,w) = (by,b,,d5) + (wee -b)(a/p,,B/P2,7/P,) and hence D,,x(p,w) = (@/p,,8/Py,3/Pq) and 3-32é 2 we 9 0 ap, 2 Pyxtpm = - tw = pro] © Bee 0 | - | Bie | toy bydgl 0 0 a5 ey Using these results, we can verify the Slutsky equation. (a) Vee D,hipu) = Dfetp.u) andthe explicit expression of O%tp) in ta (6) This follows from Sip,u) = D,h(p.u) = Dee(p,u) and (a), in which we showed that Dietp.u) is negative semidefinite and has rank 2. 3.G.4 (a) Let a> 0 and beR. Defined: RL > R by ilx) = aulx) + b and, for each & a, R, oR by a,lx,) auy(x,) + b/L. Then U(x) = aL yup(x,) + b= Llauglx,) + b/L) = Lyiiplx,). ‘Thus any linear (to be exact, affine) transformation of a separable utility function is again separsble. Next, we prove that if a monotone transformation of a separable utility function is again separable, then the monotone transformation must be linear (affine). To do this, let's assume that each ul) is continuous and strongly monotone, Then, for each & the range u/(R,) is a half-open interval. So let where by may be + @, Define cy = by - ay > 0, a= Spay b= and o~ Lycy (If some by is equal to + », then b and c are + @ as eee = lab). Suppose that f: [a,b) + K is strongly monotone and the utility function U(-) defined by U(x) = f(u(x)) is separable. To simplify the proof, let’s assume that f(-) is differentiable. Define g: [0,c) am by atv) = flv + a) ~ flad, 3-33then g(0) = 0, g(-) is differentiable, and glu(x) ~ uf0)) = fluGed) — flu(o)) tor every x € Ab, Thus, in order to prove that +) 1s Hnear (affine), 1s sufficient to prove that g(-) is linear. For thi tis sufficient to show that the first-order derivatives g'(v) do not depend on the choice of v € [0.c). To this end, we shall first prove that if Wwe {0,c,) for each é, then yyy Tyely). For this, it is sufficient to prove that glue) ~ ul0)) = Fygtuglxy) ~ uy(00) tor every x ¢ Hl tn fat, by the separabltyasrumpio, fer cach &, ere exists a monotone utility function il) such that ie) = fyi) for every x RL, Fix an x © RE and, for each t, detine x6 RL by yf = x) and yf = 0 for any k #2 Since iy’) = fluty’, Tylreg) + Fyceglhy Od = Flryly) + Ty gpl (ON. (0) = G(0) = F(u(O)) from both sides and noticing that o) + tk ~ Uf0) = klufx, subtracting TE sy Aagloep) + FigptylO) = upOup) = 14,(0), we obtain lx, - u,(0)). ‘Summing over &, we obtain Lyip(x,) - Lj O) = Lyetuylxy) Since Lpliplaxy) ~ LpiglO) = GOR) - (0) = ful - FlulOn = glutxd - won, we have gtulx) - u(0)) = Sygtu,(x,) ~ ulo)). We have thus proved that g(D,v)) = Syg(vy) To prove that the g’(v) do not depend on the choice of v € [0,c), note first that if vp € [0,cg) for cach & and v = Fyvy e [0,c), then g'(v) = g'(v,) 3-34for each €, This can be established by differentiating both sides of e(Dy¥,) = Tyslvp) with respect to vy So let v € [0,c) and v' € [0,c), then, for each ¢, there exist v, € {0,c4) and v; € [0,¢,) such that v = Jjvy and v’ = Jv Then g'(v) = g'(v,) and g'(v') = g"(vi). Now, for some ¥, € [0,¢,), consider v, + ¥, € [0,e) and 2 vi +H € (0c). Then si) = 8 + ¥) = 8G, gly) = glu, + V2) = g'C¥9). Thus g'(v,) = g'(v}) and hence gv) = g'(v"). Note that the above proof by means of derivatives is underlain by the cardinal property of additively separable utility function, which is that, when moving from one commodity vector to another, if the loss in utility from some commodity is exactly compensated by the gain in utility from another, ‘then this must be the case for any of its monotone transformations resulting in another additively separable utility function. (Thi fact is often much more shortly put into as: utility differences matter.) For example, consider p05) ~ unl) > 0 and xy = 0<}) + up{x3), the separability b vere - . xe RY and x’ € RY such that u(x) — uC) for every £= 2. Since uyjlx,) + ulx,) implies that u(x) = u(x’). By the equality g(F,vy) = Tyetvy) g(uj(x,) - u,(0)) + glug(x,) ~ up(0)) = gluj(x}) - u(0)) + gluy{x5) - uA(0)). Hence 8(u,0%) ~ uy(0) ~ glug) ~ uO) = glug(xs) - uA(0)) ~ glu(x,) - up{0)). We have shown that, under the differentiability assumption, if this halds for every x), x3, x}, and x}, then g(+) must be linear. 1 rt (b) Define $ = G,...,L) and let T be a subset of S. The commodity vectors 3-35for those in S are represented by 2, = (z,},., € RS! and the like, and the commodity vectors for tote outside are represented by Zp (ggg € AEM ad the like, We shall prove that for every 2, ¢ RY, 2t e RET, 2, e aL", and 23 € RY", (225) (zz,) if and only if (2,25) » (21,25). In fact, since u(+) represents », (21,25) » (21.25) if and only if Lyertelee) * Leerte%p) = Letty 2p + Leertel2y? Likewise, (21,25) x (2},25) if and only if Teer£ 20+ Baer” eer * Dye Dut both of these two inequalities are equivalent to Tyerte'ey) * Epertel@p- Hence they are equivalent to each other. (c) Suppose that the wealth level w increases and all prices remain unchanged. Then the demand for at least one good (say. good £) has to increase by the Walras’ law. From (3.D.4) we kuow that uj(c(p.w)) = (p/p) )ujbep(p.¥)) for Ne every k sL. Since x,(p,w) increased and u,(-) is strictly concave, the Tight hand side will decrease. Hence, again since u,(-) is strictly concave, ,(P,w) will have to increase. Thus all goods are normal. (@) The first-order condition of the UMP can be written as Mpwipy = Walp), where the Lagrange multiplier (pw) is a differentiable function of (p,w) This can be easily seen in the proof of the differentiability of Walrasian demand functions, which is contained in the Appendix. By differentiating the above first-order condition with respect to py, we obtain 3-36(BAlp,w)/8p,)Pp + A(p,w) (xq(p.w)(dxylp,w)/ap,) By differentiating the above first-order condition with respect to p, (k # U, we obtain “ (2n(p.w)/2p, Bp ~ UL glp, WI Bxylp.W)/8p,)- Thos dlp: x(p,wil/p,, UZ, p,x(P.w)]/dp,, 2(B.W) + Py 826, (P,WI/EPL) 1 Tyg PplB2/(P.¥)/P,)- (2lp,w)/ap, IPE + ACP. W)P, (nlp, wap, )Py = xj(prw) + ——,_*# *& Ep u" Ce (p,w)) ul ey (p.w)) 2 ACPWPY, Pe = xy.) « + (2X(p,w)/0R, Ey . UG (p yw) weg (pyw)) Lele enPr By the first-order condition, Alp, wp), = U'Gx,(p,Ww)) and hence this equals U(x (ow) x (D,wIU (xy, (D.w)) % - — (KE + 1) + (anlp,w)/8, 15, ———— UC (pw) Uw" Gx, (PW) U(x (pw) By Walras’ law, this equals zero. By the strong monotonicity and the strict Or Gey(p.w)d % concavity, —K <0 and Ey x4. < 0, By the assumption on Woy (p.w) Oe (pew)) ke t - iP WI" Og, (DAW) ul-), ——.—____ +1 > 0. Hence a (x,(P,w)) U(x (BW))xylp, wId" Cx, (PW) + = neo. UG, (BW) Ur tx, (R.W) We must thus have @(p,w/ap, <0. Hence, by (*), axylprwi/ap, > O- 3.6.5 (a) We shall show the following two statements: First, if (x*,2*) is a solution to 3-37Max, 2) He2) st. pox + az sw, then there exicte y* cuch that q-y* = z* and (x*,y*) is a solution to 7 5 w& aX y) MOLY) St pox + (agg)-y < ws second, if (x*,y*) is a solution to Maxey 9) UOKY) 8.t. pox + lagg)-y < w, then (x*,qg"y*) is a solution to Maxyy 2) Hoo) sit. px + az sw. Suppose first that (x*,2*) is @ solution to Max,, 9) W(xz) st. pox + az sw. ‘Then, by the definition of (+), there exists y* such that qg-y* = 2" and u(x*,y*) = U(x*.z"). Let (x,y) satisfy p-x + (agg)-y sw. Then ulxy) = Goygg*y) < ulxt, ulx.y"), where the first inequality follows fram the definition of ti(+) and the second Inequality follows from p-x + alqg-y) = p-x + (agg)-y = w and the definition of (x*,z*). The first statement is thus established. AAs for the second one, suppose that (x*.v*) is @ solution to Max, yy te ulxy) s.t. pex + (aqy)-¥ Sw. For every y, if Ay = doy" then Prat + (aqg)-y = p-x® + algg-y) S p-x* +-alag-y*) = w. Hence ula) 3 ubet.y®). Thus ulety*) = Glatgg:y"). Now let (x,2) satiety P’x + az = w. Then there exists y such that qo-y = z and u(xy) = U(x,z). Thus p-x + (adg)-y = px + algg-y) s p-x + az sw. Hence WO%,2) = ulxy) 5 ulxtsy*) = OOxa,-¥"). (b) (c) These are immediate consequences of the fact that the Walrasian demand functions and the Hicksian demand functions are derived in the standard way, by taking U(-) to be the (primitive) direct utility function, 3.6.6 (a) By applying Walras’ law, we obtain x, = (w - x,p, - X5P,)/P5. *%3 1Py ~ %2P2!/Po: 3-38(p) Yes. In fact, for every A > 0, 100 - 5Ap,/Apy + PAPy/Apy + SAW/AP, = 100 - Spi/py * BPa/Py + Su/Pyy a + PAD, Ap, + 7AP,/AP, + SAW/Ap, a + Bp,/p; + m,/p, + dw/p,. (c) By Proposition 3.G.2 and 3.G.3, the Slutsky substitution matrix is symmetric. Thus 8/pg + (8/P,Na + BP,/P3 * TP2/Py + 6W/P) = P/Py + (8/P3)(100 - 5p,/P3 + By/P + 3¥/y)- Hence hy putting p= 1 and rearranging terms, we obtain (B + a8) + BSD, + 25p, + S°w = (B + 1008) - S8p, + BSp, + Ow. Since this equality must hold for all p,, p and w, we have B+ ad = BP + 1005, BS = - 55, 7S = PS. Hence « = 100, 8 = - 5, and y = - 5. Therefore, X17 Xp, 7 100 - 5P,/pq + Spp/Pg + Sw/py Recall also that all diagonal entries of the Slutsky matrix must be nonpositive. We shall now derive from this fact thal 8 = 0. Let P3 = 1, then the first diagonal element is equal to - 5 + 6(100 - St Sp) + ow. If 8 # 0, then 8” > 0 and hence we can always find (pyspy,W) such that the above value is positive. We must thus have 4 In conclusion, X, =X, = 100 - 5P,/P, + SPo/Py- (a) Since x, ~ x, for all prices, the consumer's indifference curves in the (x,.x,)-plane must be L-shaped ones, with kinks on the diagonal. (So the restricted preference is the Leontief one.) They are depicted in the following figure 3-39Figure 3.G.6(4) (©) By (a), for a fixed x3, the preference for goods 1 and 2 can be represented by min(xyx,). Moreover, there is no wealth effect on the demands for goods 1 and 2. We must thus have Uber) = minty x5) 4 xy or a monotone transformation of this. 3.G.7 By the first-order condition of the UMP, there exists A > 0 such that gtx) = vu(x). Premultiply both sides by x, then Ax+g(x) = x-Vulx). By Walras’ law, x-g(x) = 1 and hence A = x-Vulx). Thus = Mute = gta outs a0) = utd = Sapte Tule) By Exercise 3.D.8, we have av(p,1)/8w = - p-V,v(p.l). By Proposition 1 . 1 3.6.4, 207) = > aap yaw Up (PH) = Ue re 3.6.8 Differentiate the equality v(p,aw) = v(p,w) + Ina with respect to « and evaluate at a = 1, then we obtain (v(p,w)/8w)w Hence év(p,1)/8w By Proposition 3.6.4, x(p,1) = Viv(p.D). 3-40ec 3.6.9 Let p >> 0 and w > 0. All the functions and derivatives below arc evaluated at (p.w) By differentiating Roy's identity with respect to p,, we obtain (2°v/8p,8p,(av/ew) ~ (8v/ap,)(3°v/ap aw) axyop, = ere (avrewy” Or, in the matrix notation, Dix = - 5 (0,,vb2y - v.y0,v,y) € a, YDB — UprD aa wy) (Recall that V.v is a column vector of RY, and Div and DV y are row vectors ? ? pw of Ri (Section M.A).) By differentiating Roy's identity with respect to w, we obtain (2? v/ep,awilav/aw) - (av/6p,)(8°v/aw") axyow = - ( av/aw) Or, in matrix notation, 1 Dx =- — Hayy - vey) Re. wv) P Pr Hence 1 2, 1 ev 1 eae Mut Ge ~ MpPPphgh — Cy MOL pia Dye * Vip Vel gy Day” i 2 1 2, Pw ~ EWG = IDV # WyYyPDpy) + az YyyDpy) w It is noteworthy that we can know directly from Roy's identity and this equality that the Slutsky matrix S has all the properties stated in Proposition 3.6.2. To, see this, note first that both V.v(p.w) and V,,v(p.w) are homogeneous of degree - 1 (in (p,w)) by Theorem M.B.1. Hence x(p,w) is homogeneous of degree O (where we are regarding Roy's identity as defining x(p.w) from Yevip.w) and V,,v(p,w)). Thus (2.E.2) follows, as proved in Proposition 2.E.1. On the other hand, by Exercise 3.D.8, p-x(p.w) 3-41Henee, 2 proved in Propositions 2.E.2 and 2.£.9, this implies (2.6.5) and (2.£.7). Now, as proved in Exercise 2.F.7, (2.£.2), (2.£.5), and (2.E.7) together imply that S(p,w)p = 0 and p-S(p,w) = 0. The matrix S(p,w) is symmetric because Dev, V.vD,¥,,v + ¥,0,¥D,¥, and V,yD,y are symmetric, ‘The negative semidefiniteness ean be shown in the following way. Since v(-) is quasiconvex, for every price-wealth pair (q,b), if Dyyq + Dyvb = 0, then (q.b)-D2y (q,b) = 0 (Theorem M.C.4). But Daya + Hye = 0 if and only if b = - Diva/V,v. Plug this into (q,b)-D7v(q.b). then 2, 2, 42 a-Diva + 20(0,0,,¥a) + Uy b' 2 2D,0 va (va) = a-Devq - —B*—p vq + vey —P » w ’ (vv) " A 2 (val? = Tp yarDiva) - 210,7yraND,va) + av —P— P Hence the quasiconvexity of v(-) implies that the above expression is nonnegative for every q. On the other hand, ; --1, -D?vq) ya) + ve a-Stp.wia (Ma Deva) - 200,¥,,vaD,va + VY (vv) (vv) Thus q-S(p,w)q 0. Therefore S(p,w) is negative semidefinite. 3.G.10 We shall prove that a(p) is a constant function and b(p) is homogeneous of degree - 1, quasicanvex, and satisfies b(p) = 0 and Uh(p) < 0 for every p >> 0. We shall first prove that a(p) must be homogeneous of degree zero. Since v(p,w) is homogeneity of degree zero, we must have a(Ap) + D(ApIAW = a(p) + biplw for all p > 0, w* 0, and A» 0. If there are p and A for which a(Ap) a(p), then this constitutes an immediate contradiction with w = 0. Thus 3642,a(p) is homogeneous of degree zero. Next, we show by contradiction that Valp) + 0 for every p > 0. Since vip,w) is nonincreasing in p, we must have Va(p) + Vb(plw = 0 tor every p >> 0 aud w 0. If there are p> © aud € for which dalpl/Opp 7 0, then this constitutes an immediate contradiction with w = 0, Thus alp) < 0 for every p > 0. We shall now prove that the only function a(p) that is homogeneous of degree zero and satisfies Va(p) = 0 for every p > 0 Is a constant function. In fact, by differentiating both sides of a(Ap) = a(p) with respect to A and evaluating them at A= 1, we obtain Valplp = Since p >> 0 and Va(p) = 0, we must have Va(p) Hence a(p) must be constant Given this result, the homogeneity of v(p,w) implies that 0(p) is homogeneous of degree - 1. Since v(p,w) is quasiconvex, so is a(p) as a Tunction of p. Finally, since V,v(p,w) = Volplw < 0 and Vvip,w) = O(p), we myst have U(p) & O aud Wolp) + 0 for every p 7 0. This result implies that, up to a constant, v(pyw) = b(p)w and hence, if the underlying utility function is quasiconcave, then it must be homogeneous of degree one. On the other hand, according to Exercise 3.D.4(b), if the underlying utility function is quasilinear with respect to good 1, then, for Pt w. You will this wander why we have ended up excluding this quasilinear case. all w and p >> 0 with p, = J, v(p,w) can be written in the form #lPpy ‘The reason is that, when we derived vip,w) = @(p) + w, we assumed that the consumption set is (- @, «) x RL. Thus x,(p,w) can be negative and, if so, then dv(p,wl/ap, is positive, which we excluded at the beginning of our analysis, following Proposition 3.0.3, (Note that, when we established ov(p,w)/op, + 0 in Proposition 9.D.3, we assumed that the consumption set is 3-43L Ry) As x,(pw) ie positive for sufficiently large w > 0, the quasilinear case could be accommodated in our analysis if we assume that the inequality 4v(p.w)/8p, $0 applies only for sufficiently large w > 0. (In this case, we can show that a(p) is homogeneous of degree zero, and b(p) is homogeneous of degree - 1, quasiconvex, and satisfies b(p) © 0 and Ub(p) = 0 for every p >> 0.) 3.G.11 Suppose that v(p.w) = a(p) + blp)w. By Roy's identity, xtp.w) = ~ (1/b(p)Talp) ~ (w7btp)7_ I p. pv, Alp (PDT, PEP). Thus the wealth expansion path is linear in the direction of ¥,b(p) and intercept (- 1/b{p))¥, alp). 3.6.12 Note first that, according to Exercise 9.G.l1, the wealth expansion path is linear in the direction of V.b(p) and intercept (— 1/b(p))0,a(p) If the underlying preference is homothetic, then (— 1/b(p))V,a(p) = Hence a(p) must be a constant function. If the underlying utility function is homogeneous of degree one in w, then vipyw) must be homogeneous of degree one in w by Exercise 3.D.3(a). Hence alp) = 0 for every p >> 0. If the preference it quasilinear in good 1, then please first go back to the proviso given at the end of the answer to Exercise 3.G.10. After doing 50, note that, since ihe demand for goods 2,...,L do not depend on w, 10), ( Yo(p)V, lp) /p,,0, or (90(p)/0p,)/0(P) = 17p, and 20(p)/9p, = O for every £ > 1. Hence v(p) bpe + x for some B+ 0, p40, and y eR. Dut, by Exercise 9.6.10, b(p) must ‘be homogeneous of degree - 1, positive, and nonincreasing. Hence p = - 1. 7 = 0, and @ > 0. That is, b(p) = B/p, with @ > 0. If the underlying utilityfunction is in the quasilinear form x, + fit...) then, by Exercise 3.0.4(b), v(p,w) must be written In the form #(p>,. pL) + w for all p> 0 with p, =. Thus 6 = 1. G3 For each 1 € {O,h....n), let @, be @ differentiable function defined on the strictly positive orthant (p< RU: p >> 0). Tat vip.w) = D*_.a\(plw! be an indirect utility function. Denoting the corresponding Walrasian demand function by x(p,w). By Roy's identity, 1 xp.) =~ pega Wath) Hence, for any fixed p, the wealth expansion path is contained the linear subspace of RE that is spanned by Pag.) As for the interpretation, recall from Exercise 3.6.11 that, an indirect utility function in the Gorman form exhibits linear wealth expansion curves. But the Gorman form is a polynomial of degree one on w and a linear wealth expansion curve is contained in a linear subspace of dimension two. Hence ‘the above result {mplies that the indirect utility functions that are polynomials on w is a natural extension of the Gorman form, 3.G.14 Define a, b,c, d, e, and f so that -10 a b ec -4 4 =o Since the substitution matrix is symmetric, we know b = ande ad By Propositions 2.F.3, p-S(p,w) Hence p,(- 10) + p,e + P38 = 0. “Thus ¢ =-4anda=c 4. For the second column, p,(- 4) + pp(- 4) + pge = 0. 3-45Hence e = 2 andd =e = 2. Finally, for the third column, we have p,3 + p,2 + psf ‘Thus f = = 7/6, Hence we have -10 -4 3 - 4 -4 0 2 3 2-1 The matrix has all the properties of a substitution matrix. which are symmetry, negative semidefiniteness, S(p,w)p = 0, and p:S(p,w) = 0. (For negative semidefiniteness, apply the determinant test of Exercise 2.F.10 and Theorem M.D.4(iii).) Pw 4pyw 3.6.15 (a) x(p,.p,.W) = z* zy PyPo + APY 4D yP2 + Pz Pou 2 pe 2) ©) MEyP” = || aap eT | | ase, | | 2 P Pat! (©) eloprpys) = apap ep’ He is then easy to show that Yhetpypyiu) = A(PyPpst- (€) vipypp,yw) = 20w/p, + 4¥/p,)". To verity Roy's identity, use av(p,.p,.wiap, = tw, + anny Ye we Ov(ppPpsWl/Opy = (w/py + 4w/py)/?(- 4m/pe), aut, wi/aw = (w/p, + aw/p,)/U/p, + 4/n,) 3.G.16 (a) It is easy to check that 8 de(p,ulven, = etp.ullo,, + vB, (Ilypy “D/P, Since e(p,u) is nondecreasing in p, this must be nonnegative for all (p,u). But, if a <0 and lpi is sufficiently small, then this becomes negative. Also, if @, < © and §pll is sufficiently big, then this becomes negative. 3-46Therefore w @, = 0 and B, = 0 for all k. It is a little bit manipulation to show that Toy BB eQrpu) =a fexplDyaylnp) +> ‘aime, 4, Ae(p.u) = Aexpl(ityesinp,) + ulTlp, D). Since e(p,u) is homogeneous of degree one with respect to p, they must be equal for every (p.u) and A> 0, Take, for example, p (41) and u = 1 Then log efpru) = (Sy loza + he log Xe(p,u) = loga + 1. They must be equal for every A > 0. Therefore @ Tete) Tete Thus Typ ~ 1s yO By Hence the expenditure function now takes the simplified form: @ elp.u) = (expudttie, Os Tyay = 1. a = This is increasing with respect to u and concave in p. “ () By equation (3.6.1), w = (exp vip,wittypy ). Hence @ vipiw) = logw ~ yszloepy- (c) By differentiating e(p,u) with respect to p, we obtain Alpwu) = el uN /Pyr.e-sth /Pp )- Since x(p,w) = h(p,v(p,w)) and e(p,v(p,w)) = w, (5) 2e(p.W) = wl 7D, PL). Use equations (3), (4), and (5) and follows the same method as in the answer 3-47to Exercise 3.6.2 to verify Roy's Identity and the Slutsky equation. 3.6.17 [Eirst printing errata: The minus sign at the beginning of the right- hand side of the indirect utility function should be deleted. That is, it should be Vip.w) = (w/p, + bMap,/p + a/b + cllexp(- bp,/p,) Also, in (b), the minus sign in front of the first term of the right-hand side of the expenditure function should be deleted. That is, it should be elp.u) = pauexplbp,/p,) - (/b\lap, + ap,/d + ppc’. Finally, in (c), the minus sign in front of the first term of the right-hand side of the Hicksian function should be deleted. That is, it should be - arb. (paw) = unexp(op, /p, (a) Use Sulp,wi/ap, = - p;'(apy/p, + bw/p, + elexpl- dp//p,) Ovlpw)/8w = pziexpl- bpy/P,)s and apply Roy's formula. (b) According to (3.£.1), we can obtain the expenditure function by colving w= (e(p.ui/p, + b'Cap,/p, + a/> + cllexpl- bp,/p,). (©) Apply Proposition 3.G.1 to obtain the given Hicksian demand function for the first good. 9.6.18 We prove the asst ion. Suppose that there exist £ € 4a connecting ¢ and k. Define J = (é) u (J € U, jon by contradi sL) and k € (1,...,L} such that there ic no chain of substitutes L): there is a chain of substitutes connecting & and j). Since £€ J and k € J, both J and its 3-48complement (I,...,.L)\J are nonempty. Moreover, for any j¢ J and any j € J, an (p.ulvap,, © 0, decause, otherwise, J < 1 Let u: RE 4 R be the underlying utility function, Following the hint, as in Exerciee 3.6.5, define G: RO > R by x L Hilyy,) = Maxtubd: x © RE, Digi ey = ¥y Dye shy ® x Gin?) + m® be the Hicksian demand function derived from i(-). Let fiw, That is, Fla,.a,,u) € RY is the solution to Ming yp) SUL ' S22 st gay) =u Define playay) >> 0 by pilayay) = ap, if je J and playa, = a,p, if j J. We shall prove that Flayagyu) = (jeg Ry (Play ag)uls Figs? hy (Play.a).0) Write x* = h(pla,.a,),u). Then iS xD QjP,x4) & ulx*) =u. Hence the ea PE jes constraint of the cost minimization problem is satisfied. Suppose that (yp¥q) © RE and ly,,y,) & u, then (assuming strong monotonicity) there 1 . a exists x € RY such that 569%; = Jy» LigsPXj = Yor and ulx) = Gy, ys). Thus u(x) © u and hence, by the cost minimization of 3", playe,)'x = pla.a,)-x*, This is equivalent to saying that 2 HD ye Pf) + HEP > Thus Flay a.0) = Lie jP X4 Lye sP XP By this equality and the chain rule (M.A.1), AR aay Wl Ba4y = Lye PD pg y(h (Pa 05).u)/0P,IP,) = ZyeBcggP Py 2 j Pla ppatghul/opp) We now derive a contradiction from this equality evaluated at (a,.0,) = (1). On the one hand, since 8h (pyu)/8p, <0 for every J © J and every k ¢ J, we must have dh,(1,1,u)/da, <0. On the other hand, note that fi(-) is theHicksian demand function of U(+) for two (composite) goods. Since GFi,(a,05,u)/8a, < 0 by the negative cemidefinitenecs, and (OF (a .05,4)/0a, Ip, + (OF, (ay.a,4)/00,)P, we must have ah,(1,1,u)/8a, = 0, We have thus obtained a contradiction. 3.H.1 By Proposition 3.H.1, e(p,u) Min(p-x: x € V,}. Thus, to complete the proof, it is sufficient to show that V, (x: u(x) =u). That is, x € Vy if and only if Sup (t: x € V,) 2 u. Clearly, if x € V, then Sup (t: x € V,} =u. Assume that Sup (t: x ¢ V,) =u. Define u* = Sup (ti x ¢ Vp. If ut > u, then there exists t € (u.u*l such that x € V,. Since e(+) is Increasing in utility levels, Vy > V, and hence x € V,. If u* = u, then, for every ne N, there exists u, € (u - 1/n, u) such that x ¢ V,, that is, px = ‘n e(p.u,) for all p. Let n+ ©, then u, + u and, by continuity of e(p.u), prx = e(p.u) for all p. Thus x € V, 3.11.2 We show the contrapositive of the assertion. If a preference is not convex, then there exists at least one nonconvex upper contour eet. Let u ¢ R be its corresponding utility level. We can choose a price vector p so that h(p,u) consists of more than one elements, as the following figure shows. According to Proposition 3.F.1, e(-) is uot differential at (p,u). 3-50Figure 3.2 3.H.3 By (3.E.1), for each p, take the inverse of e(p,u) with respect to u. ‘3.H.@ The following method {s analogous to that of "Recovering the Expenditure Function from Demand” for L = 2. Pick an arbitrary consumption vector x° and assign a utility value u° to We will now recover the indifference curve (x: u(x) = u°) going through x°, Assuming strong mono tonicity, this is equivalent to finding a function (sur: (0, @) 2 (0, a) such that ulx,§(x)) = uo for every x, > 0. Differentiate both sides of ulk,,€(%)) = u° with respect to ,, then we obtain ul, .€(%,))/6x, + (dulX,,€(%,))/6x,)6 6 0. Hence auld, .666,)/8% © Bulky ER N70K | ; since ope elses aes eG) ESS Bul ECR N/O% gl /ECE)) 89 (K EK) 3-51or, by replacing %, by x), we obtain (oy s8(25)) BRT €lx,) =~ 1 Ee By colving thie differential equation, we obtain the indifference curve yuing through: 2°. 3.H.S By (3.E.1), we can recover the expenditure function by simply inverting ‘the indirect utility function. To recover the direct utility function, define u: RI 9 R by lx) = Min(v(p,w): p+x s w). We shall prove that u(x) is the direct utility function that generates vip.w). So let x#(p,w) be the demand function and v#(p,w) be the indirect utility function generated by u(x). It is sufficient to show that v"(p,w) = v(p,w) for all p >> 0 and w = 0. Let p >> 0 and w = 0, then p-xt(p,w) = w and hence v*(p,w) = ulx®(p,w)) = vipyw). It thus remains to show that v*(p,w) = vip,w). Define 1 ==> ayia pe Ths x « at by the monotonicity. Since V,yip.m):p # Vyvipanlw = 0 by the homogeneity, p:x — It is thus sufficient to show that ulx) = v(pw). So let p' >> 0 and w’ 2 0 satisfy p'+x = w’. Then (p' - p):x = w’ - w, or, by the definition of x, V.v(p,w)-(p' - p) + V,v(p.wiw" - w) = 0. Hence, by the quasiconvexity of v(p,w), vip',w') = vip,w). Thus u(x) = v(p,w) 3.H.6 Let's first prove that the symmetry condition on S(p,w) is satisfied. Ry equation (3.H.2), Se(p,ul/dpy = ayelp,ul/py for every &. By differentiating both sides with respect to p,. we obtain aPelpiul/OppP, ~ (ay/PNBe(pvul/O,) = (aye /PyP,Je(Pv)- On the other hand, by differentiating both sides of de(p,ul/ap, = a,e(p,ul/p,, 3-52with respect to pp, we obtain * = = 2%e(p.ul/On, ay = (a,/p,)(de(D.u)/dp~) = (aya, /P,P,Je(PU Hence the symmetry condition is satisfied, We can thus apply the iterative method explained in the small-type discussion at the end of Section 3.H to derive the expenditure function. First, we shall prove by induction that, for every &, there exists a function F(Pyy-PL») such that Inetp.w) = Sy aptylnDy, + FplPpgyee-Py 2 Suppose first that £ = Since (8e(p,u)8p,)e(p,w) = &,/p,. Hence, by, integrating both sides with respect to p,, we obtain Inet(pyu) = aylnp, + £4(Pyee- Pt. Thus the equality ic verified for ¢ Suppose next that # > 1 and the equality holds for £ ~ Ry differentiating both sides of Ine(py) = ep MlOPy, + Fy y(PysssPL yw) with respect to p,, we obtain Belp.ul/8p, = Ay {yonesPy ¥)/OPp Since de(p,ul/8py = ay/Py, this is equivalent to yg = Fp Pye 0)/APp Hence, by integrating both sides with respect to py, we know that there exists FelPyyye-esPLyW) such that eplnpy = 1440 = FPpgee PLD By plugging this into Ine(p.U) = Decpas% AP + Fpy(Pp- PL we obtain Inetp.u) = YjeghylOD, + F4lPyyyreenPY: If £ = L, then this equality becomes Ine(p.u) = Tyo %lnp, + Fu). Or, 3-53equivalently, e(p.u) = (Mp, expf,, (u) In what follows, for every Increasing function f, (u), we shall find the utility function that generates the expenditure function e(p,u) o « (hyp, expf,(u). To start, consider the utility function ut(x) = Thx, & which appeared in Example 3.E.1 for the case of L = 2. Denote its expenditure function by e*(p,u), then aaeeanac es (yay “Typ, Du*. *(p,ut) (We considered a similar expenditure function in Exercise 3.G.16. Note that, these similarities incidentally show that, for every increasing function « FW), the expenditure function e(pyu) = (Ip, ‘lexpf,(u) has all the Properties of expenditure functions in Proposition 3.£.2, because it 4 corresponds one of the monotone transformations of uM(x) = Mx, 4) Let g(u*) he an monotone transformation and denote by e,(P,u) the expenditure function of the utility function (geu*)(x). Then Kg te) = Me, emp, Se tw. e,(Pu) By comparing this with e(p.u) = (Mypy Sexpy,(u), we know that (geut)(o) = ulx) if muy wy *, (nye, Dep, dew = trp, Yexpr,(w This equality is equivalent to ¢"(u) = (Myx, expf,(u). Letting ut = eu) and solving this with respect to u*, we obtain *) = 7 inut ~ gu") = F7'Unu® ~ Lyaplne) Thus wh) = glutad) = Fy "Unut(xe) - Syaylnay) = FP (Dyaylinxy ~ Ina). Of course, two possible utility functions are u(x) = Tiyx, © (corresponding to ee Fu) = Inu ~ Fyapina,) and ube) = Fyne, (corresponding to f,(u) = u - 3-54& Taina) 3.H.7 (a) Let p = (...1) Since x(P,L) = (1,...)1) and u(l,...91) according to Propositions 9.[.1 and 3.G.1, DyelPat) = Alp) Hence eff L. On the other hand, S(a,w) = DZe(a,w) = 0 for every q >> 0 by Exercise 2.F.11d). Hence elp.u) ~ equ) = Die(q.ulp - 4). for every q >> 0 and p >> 0. Now take q thus etpa) = Bhp » then e(p,l) - L = pr(p - p). (b) The upper contour set is equal to (xe Al: pox = TE py for every p >> Ob = xe RE: x 3.1.1 By the came method ae deriving equation (3.1.3), we obtain Ev(p°,pw) = e(p?,u!) - etptu!) 0 0.0 oO. 0.0 0) cll = ea - ete! p9.79,--sppul) + aes Se uePpt) = e(phu) P »: 1 neo 0.00 yl 2 nel ou Fo Moye orf amy +1 men Soranh he 1 2 cv(p®,p.w) = etp?,n°) - etp!u?) © 1) — gion? 0° = e(p°,u°) - e(ps.Pp.P5, 1 P ». 1 00 0.0 2 0, =F GA PpPa Pg PLU) ary + SO MPpPZPge PL) dP, ni Po If there is no wealth effect for either good, then, by the first relation of (3.E.4), nip, .P-PS, pPu?) for every p, > % 9 wish 1.) for evry bg > 8Thus EVip®,plw) = ov(p>, Denote the deadweight loss given in equation (3.1.5) by DW,(t) and that in equation (3.1.6) by DW,(t). Then ers tant? at B, nt) = nip? et. Bw) = (hoe tS 1 DWi(t) = Aipy +t, By w) = (rip) + t, 8. vyep)) Oneal vrontp? +t, By vhvapy. ‘Thus DW:(0) = 0 and, if ah(p B_ys8!0p, > 0 for every p, > 0, then DWi(t) > 0 for every t > 0. It can be similarly shown that NW; (0) = 0 and, if AMP B_yU°/6p, > 0 for every p, > 0, then DWg{t) > 0 for every t > 0. A possible interpretation of this result is that the first-order derivatives of the deadweight loss at t = 0 may be a bit misleading t approximation. In fact, their being zero means that, approximately, there is no deadweight loss. On the other hand, \ce those derivatives are positive at every t > 0, DW,(t) = £5 DW3{t) de > 0 and DW, {t) = S5 DWelc) dt > 0. fact positive. wl Hence the deadweight losses are i ° 3.13 Write vw vip?w) and u! = viptyw), then uv < a ‘because po . pt and Cant pos pl, Thue 0 0, ° elDp Pept) < Aylpy Pope) for every py > 0. Since good £ is inferior, 8 ° XflPppepelppP>,ul) > xylrpP,elPpP? pu). By the first relation of (3.£.4), this is equivalent to oO 0 oo, hylPpp pe) > hylpyPopu). Hence, by (3:12), (3.1.4), and p® < pl, we have cv(p%,plw) > EV(e° pw). 3.1.4 We shall give two examples, both of which have two commodities. The 3-56 Nefirst one is simpler, while the second one is more illustrative. In the first example, we consider a preference with “I.-shaped” indifference curves such that Ue vectors (1,1), (4,2), and (6,3) are kinks of indifference curves, Let u(ll) - 1. Note that if one of the two prices ‘equal to zero, then the demand is not a singleton. We thus need to consider a demand correspondence x(p,w). But this does not essentially change our argument because we are working on expenditure functions, which is single-valued by its definition. 1 Let p? = (1), pl = (12,0), p® = (0,2/3), and w= 2. Then x(p® an, xtpw) > (4,2), xlp.w) > (5,3), and vip2w) > v(ptw) Rut etn = 1/2 and elp®,1) = 2/3. Thus erty) = 2 - 72 = 3/2, cvip?,p?.w) = = 23 = 473, Hence v(p°,plw) > cvip®,p.w). It is worthwhile to remark that, although the given preference is neither smooth, strongly monotone, nor strictly convex, it can be approximated by such one. In the second example, we consider a utility function u(x) which is quasilinear with respect to the first commodity. Let v(p,w) be the corresponding indirect utility function, Starting from p° = 1,1) and w > 0, we conser two other price vectors pt = (pha) and 12 = (192) such that 0.< Ph <1, 0 < py <1, and viplw) = vip".w). Write v? = vip?w) and ub = vipt,w) = v(p%w). Then EV(p°,p.w) = EV(p°sp?w) 1 He, oven We shall now show that CV(p°,p!.w < GV(p®\p’.w). BY P| < Evip®.phw). Also, by the quasilinearity, CV(p%p?w) = EV pp?) (Exercise 3.1.5). Hence oVip° pw) < CV(p?,p%w. 357It is worthwhile to remark that, although Ev(pplw) = Ev(p?,p2,w), we can obtain the strict reverse inequality Ev(p®,plw) > EV(p?,p?w), while preserving CV(p°.p',w) < V(p”.p".w), by decreasing p% only slightly. 3.15 According to Exercice 3.£.7, we can write the expenditure function e(pu) = &(pp,....P,) + u for py Hence Ev(p?yplw) = e(p ul) - e(p?,u) p0 Co. >= G09.) + 0 e cv(p?,p yw) = elp'u') - etp'u) Gos PE) tw) Bp) +0) ° on 1 0) povw)). IP Fy6V,(p%sphw,) & 0, hen Dw, ® Delp ul). (plu), then Sw; = Ew, and v,(p',wi) = uP = v(p°.w,). 2.1.7 (a) Dy applying Walras’ law and the homogencity of degree zero, we can obtain the demand functions for all three good defined over the whole domain {(p,w) € R° x R: p >> 0}. Thus we can obtain the whole 3 x 3 Slutsky matrix as well trom the demand function. The Z x 2 submatrix of the Slutsky matrix that is obtained by deleting the last row and the last column is equal to tveg)[ 2 § ]: By the nomogeneity and Walras' law, the 3 x9 Slutsky matrix is symmetric if and only if this 2 x 2 matrix is symmetric. Moreover, just as in the proof of Theorem M.D.4{iii), we can show that the x 3 Slutsky matrix is negative semidefinite (on T,, and hence on the whole R) if and only if the 3-582% 2 matrix is negative semidefinite. Hence. utility maximization implies that ¢ bs 0, 20, and bg - c7 BO. (b) First, we verify that the corresponding Hicksian demand functions for the first two commodities are Independent of utility levels and, as functions of the prices of the first two commodities alone, they are equal to the given Walrasian demand functions. let p be any price vector and u, u' be any two utility levels. By (3.E.4), hy(p,u) = x,(p.e(p,u)) and hy(p.u') +) do not depend on wealth, xfprelpn')) for = 12, Since the x4 x,(p,e(p,u)) = x,lp,e(p.u')). Hence: (pu) = ypu). Thus the hyp.) do not depend on utility level and they are the same as the x;(p,w). If the prices change following the path (1,1) + (2,1) + (2,2), then the equivalent variation is J tpl asuddp! + 5? n?(2,p? uldp? = ra xp! 1,widp! + g 27(2,p",widp™ = (a + (3/2)b + c) + (d + 2e + (3/2)g)- If the prices change following the path (1,1) + (1,2) + (2,2), then the equivalent variation is J2 rcp wlap™ + FF np! 2,u)ep" =F Pp wiep? + 2 ipl 2widpt = (d +e + (3/2)g) + (a + (3/2)b + 20). ‘These two equivalent variations are the same if and only if ¢ (©) As we caw above, EV, = £2 x'tp'L,widp! = a + (9/2) + 6 L Pup widp™ = doe + Wye d+ e+ WB, EV, EV = (a+ (3/2)b + c) + (d + 2e + (3/2)g) 3-59a+ (3/2)b + 304d + (3/22. Hence EV — (LV, + EV,) The sum EV, + EV, does not contain the effect on equivalent variation due to the shift of the graph of the demand function for the second commodity when p! goes up to 2 (or equivalently, the shift of the graph of the demand function for the first commodity when p* goes up to 2). Graphically, Jetting ¢ = € > 0, EV contains the shaded area below but EV, + EV, does not: a 2 1 o dte d+2e x! Figure 3.1.7%(¢) (@) Since (2,1,w) - a + 2b + c, the tax revenue from the first good is equal to this. Thus DW, = (a + (3/2)b +c) ~ (a + 2b +c) = - b/2, Since x5(1,2,w) = d+ © + 2g, the tax revenue from the second good is equal to this. Thus DW, = (d +e + (3/2)g) - (4+ 6 + 26) = — gv2 Since x(2,2,w) = a + 2b + 2c and x,(2,2,w) = d+ 2e + 2g, the tax revenue from both commodities is (a+ 2b + 2c) + (d+ Ze + 2g) = a+ 24 404d + 2g. Thue DW = (a + (9/2) + 304d + (3/2)g) ~ (a + 2b + 4c +d + 2p). 3-60Hence DW - (DW, + DW,) = - c (e) Our problem is Min, DWit,,t,) tt) PWT! Bhp + ty Le ty Wty R Here, DW(t,t,) = EV(ty sty) . TR(tyty) Hell + ty 1+ ty, w= etdsLud - Spang + ty, 1+ ty, uty Set up the Lagrangean by Lty, tps) . DW(ty,to) +R TR(tyst9))- ‘Then the first-order condition with respect ta ty is aDW(t,,t. ay - STR sty) dtp = 0. But, ODWIt,.t,)/dty = Gell + th 1+ ty ul/at, - ht +t, ltt, u) ~ Sica = pe OA + ty, 14 ty, wlveryyty by Gell + ty, 14 ty, wi/oty = hell + t, (ah, (1 + ty 1+ ty, wat, ty 2! 14 ty, wy and ATR tysty)/aty = hyll + tyy 1+ ty, w+ Tym + ty 1+ ty, wl/atyty. Hence the first-order condition is written as Te planglt + ty 1+ ty, wierd + 2d tani ety Let, w= 0 for both & = 1,2, From this and R= Te hyll + ty 1+ ty ulty, we obtain be, + ety, ct, + at, “= Sper ays cs a) * See sat) + wt ty)" SOT Faye oe apy aS oe at) + BF By) (a+ bi + ty) + cll + tye + Cd 4 oft + ty) + all + tity = R 3-613.1.8 (2) Quasilinear utility functions: ulx,.x,,X_) = Glx,x,) + x, (b) As in Exercise 3.1.7(a), the symmetry implies © + ap, = by + dypp for all py > 0 and p, > 0. Thus c, = b, d, = d, = 0. Then the negative semidefiniteness implies that 2 by #0, 6) 40, byes ~ 6,7 50. (©) Since the Walrasian demand functions and Hicksian demand functions are the same as we saw in Exercise 9.1.7(b), we can defin Si! selapyowda + JP? cv = 5) wilqpy,widg + py Baarpnia + Je or, equivalently, 2 ‘1 : fp) pimp arwida + Jp, mifngwta (@) By the same calculation as in Exercise 3.1.7(c), we obtain EV, = 1/2, EV, = 1/2, EV, - 9/2. In this case, EV # EV, + EV,. In the general case in which the conditions in (b) hold, EV, = a, + (3/2)b, + ¢, v 1g + By t (BAe, ig = 8, + (3/2Ib, +e + a, + 2b, H9/2Ie,. Hence EV, + EV, = FV, if and only if b, = ¢, = 0. This condition is equivalent to saying that any change in the price of one good does not have any (cross) effect on the demand for the other. 3.19 Let Qe re be the vector of which the é-th component is one and all the other components are zero, For each t, define p(t) = p + tey Then the after-rebate income w(t) with tax t satisfies w(t) = w + x,(p(t),w(t))t 3-62Hence p-x(ptt,w(t)) = (ptt) ~ tep)-x(pte)nw(t)) = w(t) ~ xplploawende = w. Therefore x(p(t),w(t)) ig at most as good as x{p,wl. In order to prove that xe(p{t,w(t)) is strictly less preferred to x(p,w), it is sufficient to prove that these two are different, because the demand function is assumed to be single-valued. Now suppose that there exists a t > 0 such that x(p(t),w(t)) = x(p,w), Let u = v(p,w). Then we have A(p(t),u) = Alp,u). In particular, hy(p(t).u) = hglpiw). Since the Hicksian demand funetion ¢ + hy(p(s),u) is nonincreasing, this equality implies that hy(p(s),u) = hy(p.u) for every 5 € [0,t], But dlhg(pls),u)l/ds = dh,(p(s),u)/épy, Evaluating at s = 0, we have ahy(p,ul/6py = sylPw) = 0. This violates the assumption that sy(p.w) < 0. Hence Ie(p(t).w(t)) * h(pyw) for every t > 0. 3.1.10 We consider an example of a consumer who face the choices over two goods and whose preference » and demand function x(p,,P,,W) satisfy the following condition: For every p, € Ws2h (p42) = (Ul - py 1+ 2s for every py ¢ (h2h x(Py.2) = (I= €, + €)/p,) (1-2, 1+ e)>- 6, 0+ eZ). By using the figure below, you can convince yourself, perhaps with some application of the weak axiom, that there actually exists such a preference. 3-63x 2 ” \ TAA Figure 3.1.10 Define p° = (2,1) and p! = (1,2). Then x(p0,2) = (( = 072, 1+ ©), x(p2) GQ-e + ev), Thus x(p%, ) > x(p',2). However, the area variation measure following the price-change path p° » (1,1) » p! is ave vp 1 19% lPLi2V4P, + SPX (LP,.2)dp, 2 2 fy 1 ~ ele, dp, + ST (t+ evep, dn, 2 = = odin pyle y+ [+ edt py i 0 2eln2 > 0. Hence the area variation measure ranks p! over p?. BLM If (p' - p®)-x' > 0, then w > p?-x. The local non-satiation implies that x° is preferred to x'. Hence the consumer must be worse off at (p!,w). As for the interpretation in term of the first-order approximation, since e(p,u) is concave in py 1 cpu) = etpul + vetplul-(p° - ph. 3-64Since Ve(p'.u')-(p? = p') < 0, e(p.u') < eto! 1 1) = w. Thus u? = vi(p°,w) > Finally, (p! - p°)-x! > 0 if and only if w > pox’, which, in turn, Is equivalent to pox! - 2° <0. This test is depicted in the picture below: Figure 3.1.11 3112 Let v? = vip°,w®) and ul = vipliw!). Then we define EV(?.w°ipwh = e(p°.ut) - etp®.v9) = et®u!) - w°, wi etpiu®y. ° ov(p?.wiptw!) = etpiyu’) ~ etp!.u The “partial information” test can be extended as follows: If pl-x° < w', then the consumer is better off at (pw). This can be proved in three ways. The first one is the same revealed-preference argument as in the proof of Proposition 3.1.1 ‘The second way is to use the indirect utility function. Since vip.w) is quasiconvex, if tot = p%-v,vin®ws + cx! ~ wravipw®rrow > 0, then we can conclude that vip4yw!) > vip°,w°). But, by Roy's identity, this 36ssufficient condition is equal to = (wl = vd-(ort aw rrowrxty2.wd + tt = wiiorto®w°izow) (30(p9,w)7awi(= plex? + ws wl - w) (avip?,wrvawnw! ~ pl-x°) > 0, Hence, if p'+x° < w', then vipw!) > v(p°,w°). The third way is to use the expenditure function. vip',w!) > vip°,w°) if and only if e(p!,vip!w')) > etp!vip?w°)}. But e(p!,vip',w')) = w! and o e(p!vip°,w)) = pix Hence, if pl-x? < wl, then we can conclude that viptiwh > v(p,ws. 3.5.1 (First printing errata: The difficulty level should probably be B.] It follows immediately from the definition that if x(p,w) satisfies the strong axiom, then it satisfied the weak axiom. Conversely, if x( p,w ) satisfies the weak axiom (in addition to the homogeneity of degree zero and Walrac’ law), then the Slutsky matrix is negati semidefinite and, by Exercise 2.F.l, symmetric. Hence x(p,w) is integrable, implying that there exists a preference relation that generates x( p,w ). Thus x(p,w) satisfies the strong axiom as well. BAA If (pw) = (1,11), then x(p,w) = (0,1). The locally cheaper condition is not satisfied since B, , = {x € Ry: x, + x, = 1) and there is no 1 y such that p-y < w, as depicted in the following figure. 3-660 I Figure 3.AA.1 x To check that the demand function is not continuous at (J,1,1), consider the sequence (p",w") = (1 - I/n, 1, 1 - 1/n). Then (p",w") + (1,11) and x(p",w") = (1,0), but x(1,1,1) = (0,1), This discontinuous change in demands arises because the budget set D, , consists of (1,0) for every nm, but By 1) 1 = pw oe {x € RO: x, + x, = 1), so that the commodity bundle (0,1) becomes available suddenly at p = (1D. 3.AA.2. [Eirst printing errata: The upper hemicontinuity of h(p,u) cannot be guaranteed at p = 0, because the local boundedness condition in the definition of upper hemlcontinulty need not be satisfied. Hence the clause in the bracket “even if we replace minimum by infimum and allow p = 0 should be understood as concerning only with e(p,u).] We shall first prove that A(p,u) is upper hemicontinuous. Let B be a compact subset of the domain of h(p,u) (which is, in turn, a subset of {p € RU: p >> 0) x R). Then there exists a (B,u) € B such that 0 = u for every (pu) € B. Let x € h(p,u), then ulx) = u = u for every (p,u) € B. For each &, define 3-67By = Maxip-X/p, € R,: (p.u) © RD and § = (Fy.¥,) © Ry We now show thal, for every (pu) € B and x € A(p.u), we have J = x. If fact, since uli) = u, pe = px. Since p >> 0 and Xe RY, py Spex Thus pee Pyxp Divide both side by py, then we obtain Pit/py ® my and hence Jy * xp We have therefore established the local boundedness condition of upper hemicont ity. Next, let Pu), bea sequence of pairs of price vectors and utility levels, converging to (p.u). Let (x"), be a sequence in RL, x” © n(p"u") ror every n, and x" +x. It is sufficient to prove that x € h(p,u). Since u(x") = u" and u(x") » u(x) by the continuity, we obtain u(x) = u. Heuce x satisfies the constraint of the EMP at (pyu). To show that it is cost-minimizing, let y © RE and uly) =u. If uly) > u, then uly) > u" for any sufficiently large n. Hence p"+y = p™-x" for such n. By taking the limit as n @, we obtain p-y = p:x. Suppose then that uy) By the local nonsatiation, there exists a sequence {y"), in RE such that u(y) > ux) for every n and y" 4 x, Hence there exists a subsequence k(n), Kin), m CRE) of (pu), such that uy" = uk), Hence pl ptm), tn) By taking the limit ac ns m, we obtain p-y > p-x. Hence x is cost-minimizing. We now turn to the continuity of e(p,u). In fact, its continuity at every p >> 0 can be derived immediately the continuity of A(p,u), as the latter is well defined nt every p >> 0. Thus the essential part of the following proof is the case of nonnegative, but not strictly positive, vectors. We shall establish the continuity with respect to p and that with Fespect to u separately. Let u ¢ R be a utility level, p € RL be a price vector, and (p") be a vesuence of price vectors in converging to p. We need to prove that 3-68e(p"\n) + e(pu). As a preliminary result, let's first prove that if the sequence (e(p",u)) in R, converges, then it must do so to e(p,u). Let w be the limit of (e(p"u)}. Let x € Ry and u(x) = u. Then p™-x = e(p".u) for every n. Taking the limit as n + , we obtain p-x = w. Since this holds for L every x ¢ RL with ulx) & u, we have e(pw) 2 w. To prove the reverse ually elpa # wwe woe the concavity of fp.) in pc A, Take « kin) k(m) cubsequence (pK) of (p%) such that (Pp - pirg'™ - p,) 2 0 for all Le {,....L) and positive integers n and m. That is, for each & € (l,. reauire the stan ot okt” L}, we ~ pp along the subsequence to be constant (including zero). Such a subsequence does actually exist because each p® - p has one of at most 2" sign patterns. Now, for each ¢ ¢ (1 ait phim) _ p= 0 for every n; and vp =~ 1 if ph” - k(n) k(n) PE” = y+ Ing ~ Pylvy. y= 0 for every n. Then, So tet 2? = Ipk'™ - ppl = 0 and define vé € RE by letting vj = ¥, and vy = 0 for every k # &, then x(n) pe aps patel, A n Now, define 28 — 1 - Sys then Oe a pin) 2 atp + Syzpm +) since pM » p, 2B 9 0 for every £€ (ua.L). Thus 2591. Hence, for ) every sufficiently large n, 23 > 0 and pM? is a convex combination of 2°, en Therefore, by the concavity, ety) = zMe(pu) + Bevel + v4, w. k(n) Since e(p",u) w, e(pX"),u) » w. The right-hand side converges to e(p,u). Therefore w < e(p,u). We have thus proved our preliminary fact Ubat if the sequence (e(p™,u)) in R, converges, then it must do so to e(p,u). Let's now prove by 3-69contradiction that this implies that e(p"\u) + e(p,u)- So suppose not, then there are a 8 > 0 and a subsequence ip™)) of (p") such that letp™) u) - elp.u| = S for all n, Since the subsequence ¢etp“"),u)) is bounded, it has a further, convergent subsequence. On the one hand, the limit can never be e(p.u, herause tecpkin) 4) ~ e(p,u)| > 8 for alll n. On the other hand, our preliminary result implies that the limit must be e(p.u). This is 2 contradiction. We must thus have e(p",u) > e(p,u). Let's now turn to the continuity of e(p,u) with respect to u. Let p € Rt be a price vector, u € R be a utility level, and {u") be a sequence of utility levels in R converging tou. We need to prove that e(pu") » e(p,u). Just ae before, it is sufficient to prove that if the sequence (e(p,u")) in R, converges, then it must do so to e(p.u). Let w be the limit of {e(p,u™). Let © > 0, x € RE, u(x) = u, and p-x < e(p,u) + €. By the local nonsatiation, we can make u(x) > u while preserving p+x < e(p,u) + ¢. Then there exicts a positive integer N such that u(x) > u" for every n > N. Thus, for such n, p+x = e(p.u"), Take the limit as n> ©, then p-x = w. Thus (pu) t © > w. Since this holds for every © > 0, we must have e(p,u) 2 w. To show the reverse inequality e(p,u) = w, we can assume that u" < u for every n. (The reason is as follows: If there is a subsequence such that u = u for every n in the subsequence, then we can apply this ease to subsequence. If there is no such subsequence, then there is a subsequence such that u" = u for every n in the subsequence, Hence e(p,u) < e(p,u") for such n. Taking the limit, we obtain e(p,u) = w.) Now let ¥ ¢ RY and ule) = u. Define B= (x € RL: RE xd, then B is compsct. This and u" = u implies that the truncated EMP Min p-x s.t u(x) = u™ 3-10has a solution, denoted by x" © B. Then x" © hi(p,u"), that is, x" a solution ee eens car femmeaesl pa oe siete aleeemer cee is a convergent subsequence (x*')) of (x), Denote its limit by x. Since u(-) is continuous, u(x) = u and hence p-x = e(pu). Moreover, p-x“™ = kin), etp.u*!™)) and p-x! px. Thus w = p-x and hence w = e(p,u). Suppose that u(x) is strictly quasiconcave, twice continuously differentiable and that Ju(x) * 0 for all x. Then we know that hip,u) is a function and the Lagrange multiplier A of the EMP must be positive. The first-order condition for the EMP can be considered as a system of L +1 equations and L + 1 unknowns: p- avu(x) = 0 ux) =u = 0 By the implicit function theorem (Theorem M.E.1), the solution h(p,u) as a function of the parameters (p.u) of the system is differentiable if the Jacobian of this system has a nonzero determinant = Dux = p vulx™ 0 at (p.x) satisfying the above two equatians. Fut, then, p= AVu(x) and hence #0 this condition is equivalent to - Dud = wud |g fn , Puls) ° that is, 2, pu) vax) | yg, vutx™ 0 By Theorem M.D.2(i), thie inequality holds if D°u(x) is negative definite on ty < RL: vulx)-y = 0). This sufficient condition is a stronger differential version of quasiconcavity, as the latter is equivalent to the condition that 2, D°u(x) Is negative semidefinite on {y € RU: Vulx)-y = 0). 71CHAPTER 4 4.B.1 By Roy's identity (Proposition 9.6.4) and vj(p.w,) = ai(p) + (pw, 1 x00) = aT en? 1 NG Bey pA”) ~ Bepy MPP) fl Thus ¥y (009) =~ p9,$) for al |. Since the right-hand side i i identical for every i, the set of consumers exhibit parallel, straight expansion paths. As for the second part, by (3.E.1), e\(b.u,) = Cu, ~ a,(P))/0(0). Hence, by letting clp) = 1/b(p) and d,(p) = - a,(p)/olp), we obtain e,(p,u,) = wha, + ain). 4.B.2 (a) Let pe RU he a price vector and w > 0 be an aggregate wealth. Consider two consumers, i and j. Consider two wealth distributions Wy) and (wi,...,wy) such that w; w= 0, w, = 0 for any k # i, and w; = 0 for any k # j, Since the preferences are homothetic, x(p,0,5,) = 0 for every k. Thus the aggregate demand with (w,,...,W;) is x(p,w,5,) and the aggregate demand with (w, W)) 18 x(p,w.s,). Since aggregate demand depends only on prices and aggregate wealth, we have xlp.w,s,) = (pws 3) Since p and w were arbitrarily chosen, this means that i and j have the same demand function. Hence they have the same preference. Since i and j were arbitrarily chosen, we conclude that all consumers have the same preference. (b) By analogy to the Gorman form, consider the following form of indirect utility runctions: adNote that b(p) and e(p) do not depend on i. By this and Roy's identity (Proposition 3.6.4), 1 x(p.Wys) =~ V,y,(Pw,) Pie Wai pw) philPM ¥, s 1 ‘ : 2 pepe Mal) = orp MP) = Bepy Pele) Thus = ghee tivated - FE once - BE gees eer HPAI) = Bey PNP) “ot NP Thus the aggregate demand depends only on J,w, and Js, (and p). Tptpwye 4.C.1 By the definition of a directional partial derivative, D,x(ewhly = lim, ,o(l/elixly + edp, w) ~ x(P)- Hence ap-Dyxtp.wép = dp-(lim, g(1/eNixtp + edp, w) ~ xlp.w))) = lim, ,o(l/eldp- Getp + edp, w) ~ x{p,w)) But the ULD property implies that dp-(x(p + edp, w) - x(p,w)) = 0 for all © > 0. Hence, by taking the limit © > 0, we obtain dp-D.x(p,widp + 0. Thus D,x(p,w) is negative semidefinite rl, a We shall prove the converse by contradiction. Suppose that the Jacobian D,elp.w) is negative definite for all (p,w) and that there exist p € RY, pie RY, and w, ¢ R such that x;(p,w,) # x,(p',w,) and (Pw) - xp.) = 0. Let A > 1 be sufficiently close to 1 that for every A € [0,A], demand is well defined at (1 - A)p + Ap’. (If demand ic well defined at strictly positive price vectors, 1 is determined so that (1 - Alp + Ap’ >> O for every A € f0.41.) Define pla) = (1 - Alp + Ap’ and 4-2wOD = (Pt = P)-Ge(POI,w) ~ 26;(P.¥7)) ‘Then the function w(-) is differentiable. w,(0) = 0, w,(1) = 0, and Wi) = (p" = p)-Dyx,(pA,wHP" - p). We consider two cases: Gase 1: w(A) + 0 for every A € (0, ‘Then ) = 0 and it is a maximum. Thus wi(1) = 0, that is, (p" ~ p)-D,x(p",wilp" - p) = 0. This is a contradiction to the negative definiteness. Case 2: wi(2) > 0 for some A € [0,A]. ‘Then, by the mean-value theorem, there exists A* € (0,A) such that w,(A) ~ w)(O) = WANE ~ 0), By wi(0) = 0 and wi(A) > 0, wi(A") > 0. That is, ("= p)-D,x(pOa*),wilp" ~ p) > 0. This is a contradiction to the negative definiteness. Our proof is thus completed. 4.0.2 If Dyx\(Paiw) is negative definite on the whole Re for every i, then ‘the sum J;D_x/(P.a;w) is negative definite on the whole RY. since Dox(Pw) = 7. plying that TD, ;(0.a,w), De x(p.w) is negative definite on the whole Ri x(p,w) satisfies the ULD property. To establish the WA, one way is simply to notice that the ULD property implies the WA, as the latter considers only compensated price changes. Another way is to prove the xiven differential sufficient condition. Let’s assume that w > 0. Define H = (v ¢ RY: v-x(p,w) = 0}, thet is, H is the ‘hyperplane with normal x(p,w) that goes through the origin. Then p ¢ H because p-x(pyw) = w > 0. Thus, if v € RI and v is not proportional to p, then there exist v, ¢ RU and v, € RY such that v, €H, v, #0, v, is 3proportional to p, and y= ¥ + ¥p. Since S(pywlvy = 0 and vy:S{p,w) = 0 by Proposition 2.F.3, we have veS(p.wlv = (vy, + v,)Slp.wily, + ¥)) vy Siwy, + v-Slpiwlyy + vy-Stp.why, + ¥5"SIp,¥v, vy Sipwy,. But here, by v, € H, JT, Slpswoy = (Dyxlp.w) + Dyxtprwhetp.w) Wyy= Delp.) Hence ¥,-S(p.wiy, ~ y-D,x(p.w)y, <0 because v, # 0 and D,xip,w) is negative definite. Thus the WA holds. 4.C.3 A Giffen good will be a most familiar example. In the figure below, good 1 Is a Giffen good, Figure 4.0.3 This example shows that the ULD property is actually not derived from the utility maximization. It ic a restriction on preferences. 4.0.4 The L-shaped indifference curves imply that for every strictly positiveprice vector, the consumer's demand is always be at the corner of a upper contour set. Hence no compensated price change will change the demand. Thus S,(P.w) = 0 for every (p,w,) and D.2;(p,0,) = Dy x(p.w,)x\(p.w,)" Suppose i that there exists (p,w,) such that Dy, sa(p.w,) * (7, (ow). Since PD, x,(p,w, ; = prllyw xp) = this implies that. Dy (Pw) and i x,(P.w,) are not proportional. Hence there exists av € R2 such that v'D,, %(P/¥,) < 0 and v-x,(p.w,) > 0, as illustrated in the figure below: i (Awyx(p.wi) P x Figure 4.0.4 Thus ee wiEy = = Ov WD LxylPw Wy = — wD, aelprwa(pw) Ty = - (Dy (bw 0 WD) > 0, which implies that the ULD property is not satisfied. Thus, If it is in fact satisfied, then we must have D,, x,(p,w;) = (I/w,)x(p,w,) for every (p,w)- ‘Thus the unique wealth expansion path, which is the set of the corners of the “upper contour sets, is a ray going through the origin. Henee the preference is homothetic.4.CS Following the hint, we fix w = 1 and write x,(p) = x;(p.1). Consider 1 the indirect demand function g0) = Sorte i Wujx). Since x\(p) = x if and only if g(x) = p, the ULD property of x;(p) is equivalent to the following property: if x # yy then (g;(3) ~ g,(y))+(x = y) <0. For this latter property, it is sufficient to show that D?g,0x) is negative definite. We shall now establish this. By the chain rule (Appendix M.A), (0-9 Dg, = O60 anu, ~ vu,000u,007 ~ vu,00x7D7u,00) Let q = Vu,(x) and C = D’u,(x), then this can be rewritten as Dgi6o) = (eg) Abx-@)C ~ aa” - ax") We need to show that v-Da,(xlv < 0 for every v # 0. If v-q = 0, then v-Dg,(xWv = (x-q) "v-Cv < 0, (This property is equivalent to the negative definiteness of the bordered Hessian of u,(+) and used to guarantee the differentiabllity of the dewand function, as explained in the Appendix to Chapter 3). So cuppoce that v-q * 0. By multiplying a ecalar to v if necessary, we can assume that viq = xq. Then v-Dg,(xlv = (x-q)""(vGv ~ veg ~ x-Cv). By x-q > 0, we need to show that v-Cv - vq - x-Cv < 0. Since C is symmetric, vite = xeGv = (v ~ (172)x)-Clv ~ (1/2)x) = 1/4 )K-Cx. Since u,(-) is concave, C is negative semidefinite and the first term in the above expression is non-positive. Thus, vety = xeCv 5 ~ (U/Ax-ox. Hence veGy = veg = x+Cv S ~ (U/AK-CX ~ ax, since - °C < 4, the right-hand side is nexative, Hence so Is the left=hand side. 4.C.6 By differentiating both sides of u(x) = Au(x) with respect to A and taking A = 1, we obtain Vu(x)-x = u(x). Then by differentiating both sides of this equality with respect to x, we obtain D°ulx)x + u(x) = Yulx). Thus D7uGdx = 0 and hence of) 4.C.7 Suppose that the distribution of wealth has a differentiable, nonincreasing density function f(-) over the interval [0,i]. Let ve Rt and v * 0, then, just as in the proof of Proposition 4.C.4, we have v-Dxlply = SMor-ste,wnrtwaw a Serb, Zp wn Sp, wi Flwidw. Here, the first term is negative, unleas v is proportional to p. (This Property is equivalent to the negative definiteness of the bordered Hessian of u,(-) and used to guarantee the differentiability of the demand function, as explained in the Appendix to Chapter 3). As Tor the second term, just as in the proof of Proposition 4.¢.4, 3 p.w) dw Sl Dy Hew ly Hp,whsewdaw = (72) 3 Stwnaw, By integration by parts and X(p,0) = 0, this is equal to Aas2yv ABW) = 720% xt, WI)? F wldw The first part of this is always nonnegative, and it is positive when v is Proportional to p. ‘Ihe integral of the second part is nonpositive because 2 Flwidw = 0. Hence v-Dx(plv < 0. X(p.wi) dw F'lw) $0. Thus (i/2)s% 2 To see that there are unimodal density functions for which the conclusions of this propositions do not hold, recall that viDx(ply = Sly Stp,wie)flwldw ~ Shlv-D, Mp.wiliv-Xip. wns twraw. To be specific, let v = (1,0,...,0) € al, thenDap) = Shs, (p.whflwlaw ~ £4D, %(0.w)X| os11 (pw f (whdw. Suppose also that the graph of the function w + 5(p.w) is ac chown below. 0 we w Figue 4.07 Then D,,X,(p,w) = 0 for every w € [0,w*] and D,,(P,w) = 0 for every w € (wf. Thus 5"D,,,(p.wi¥(pwiF(wiaw = 0 and 5% 0, 5 (0.09%, Fd =0. If the density function (+) is uyuincreasing, the weight placed on the interval [0,w*] is sufficient to dominate the negative effect of the interval {w41, implying thar SD, 5 (p.wlS(p.wiflwldw = 0, If the distribution function is not nonincreasing, then the weight on the interval [w*,w] may dominate that on the interval [0,w®l, so that SYD, 5 (P.wIE,(, IF (wd <0. It could even dominate the substitution effect, in which case we have WeDxiply = SG(-Stp.wwlflwldw ~ SG (v-D,,Xp.wlv:Xlpyw))Fw)aw > 0 4.68 By substituting (4.C.6) into (4.C.5), we obtain Stp,w) = 5,5; co 9) = Fy (Dawa DH) + aD (PAI W (On the other hand, the right-hand side of (4.C.7) equals to yy 0,0) ~ F:Dy, x(D.awhelp.aw) + (Te,D, X,(p.a,w))x(p,w)™ i i 4-8+ Dy (pw x, (5a), x(P.0 xp, wT = F5,eq0) - Ey, Bawa (ra) + ied, x; (0.9; ¥))x(p,w)7, We have thus proved (4.C.8). 4.9 The homotheticity implies that Dy, Paw) = (/e;whx, (paw) and hence D,,x(p,w) = Fay (wix(p,w). Thus Ol, w) = Fie, ((1/a,whx;(0.0,99) - U/wlxtp.wa)e,(p,a,) = x(pwnT = xtp,wyt, > Tyla wx, (B.a40) ~ slp, w)NC1/e, (0, Thus, for every v € kt, Y-Olpiwy = Tila wil (t/a, (9,0) ~ xip,wi)-v)? = 0 Therefore, Clp,w) is positive somidefinite. 4.C:10 Let's start by formulating our continuum-of-consumers situation We take the interval 0,2] as the set of (the names of) the consumers, The Population density is equal to 1/2 uniformly on [0,2]. We assume that the Proportion of each consumer's wealth to the average wealth (which is, in this situation, a counterpart of the aggregate wealth) is constant regardless of the amount of the aggregate wealth. We further assume that when the "average’ wealth is equal to w > 0, the wealth of consumer 7 € [0,2] is equal to nw. Since Jnw(/2)in = W, the term “average” is justified. The average demand is then defined by x(p.i) = J2%(p.niv)lL/2idn and the Slutsky matrix S(p,w) is defined as in (4.C.4). The Slutsky matrix of consumer 7 is denoted by Sip.nw). We then define C(p.w) by ep, ) = SoStp.nw\(1/2)dn - S(p.w). Just as in the small-type discussion following the proof of Proposition 4.C.4, oywe have orp. ~ eared, Mermw) ~ Dy xp. CLM nD — stp) an = evan, Mp.ninepnivldn ~ Dy xlp,iilxtD.i0™ = £21172), o.niv) Sp nivddn ~ (SEln/20D,,(P,nivddn(p Thus, for each v € R, veClp.wv = Rasare-n, panini: Spyniinan ~ (FGla/20"-D, He niaMV-2(0. 9). For the first term, we know that Berard, Spm Zpanivnian = W/4HN0-Klp,20. [As for the second term, by integration by parts, Soln/2lv-D,, Ep. nivden tensencv-Stp niin iP’ ~ Bavany- XowMGWen 1 AWN Xp,20)) = C/W(v- xf 9), Hence veOlp. wy = (/AU/AYy-Htp, 20) ~ (WFP, 2HNKv-vKp. HD) + (w-xetp,)7) = (W/WNMU/2N¥-Hp,28) ~ v-xe(p,H))? = 0 Thus C(p,W) is positive semidefinite. 4.1 (a) When deriving individual demands from the first-order conditions of utility maximization, we will neglect the nonnegativity constraints (which is investigated in Exercise 3.D.4(c)). In fact, we will later see that, for prices and wealths under consideration, the demands are always in the interior of the nonnegative orthant. It follows directly from the first-order conditions that x lPyw/2) = (oe (Pr ¥/2).xp(P.W/2)) = (8720, ~ 4D4/D2~ 495/05). = Uapern2 x P.W/2) = (,9{D.W/2).XpqlP,W/2)) = (4P2/Pj+ W/2P, ~ 4PQ/P))-Hence 2elp.w) = 2e(p,w/2) + x9(pyw/2) 2 = (0/20, So/05 + 48/08, wg Any + Prd (b) Denote the (4k) entry of the Slutsky matrix S,(p,w) of consumer i by FygiP-w)- Since dx,(p,w/2)/8w, = 0, s29,(p,W/2) = Bx,,(p,w/2)/ep, 2 rf /e8, Hee by Proposton 2. 8 (P.W/2) 2 Siq(P1¥/2) = Bp,/p5, and hence s11)(P.w/2) = - 8/p). Thus 2 - 8p, 8p,/p5 S,(p,w/2) = i 2 2,3 8p,/P5 - 8pa/pS Similarly, we can show that 2,9 2 ~ 87% 8P2/P S,(p,w/2) = 2! 2 Bpy/Pp = 8/p, We can also apply Proposition 2.F.3 to derive the Slutsky matrix S(p,w) of the aggregate demand function: 2 2,3 2 2 ~ WAR, = 6/p, ~ 685/m, w/4P yD, + 6P,/p% + p/p? 2 S(p,w) = 2 a W/4P,P2 + 6P,/P3 + 6P,/P; = W/4D5- 6/p, 3 = 808 By Exercise 2.F.9(b) (and S(p,w)p = 0), if dp € R°, dp # 0, and dp is not proportional to p, then dp+S(P,w)dp < 0. Thus, according to the ‘small-type discussion after Proposition 2.1 , the aggregate demand function x(p,w) satisfies the WA. (c) By substituting p = (1,1), we obtain Wh 4 4 - wre Clp,w) = T.S;(p,w72) - Sipw) = 7 4- wih w/t - 4 Thus, if w > 16, then it is positive semidefinite, and if 8 < w < 16, then it is negative semidefinite. (This can be shown by applying Theorem M.D.2(ii), or by noticing that C(p,w)p = 0 and p-C(p,w) = 0 and then applying theargument in the proof of Exercise 2.F.9(b).) Far example, if w = 12 and v = (2,0), then v-Clp.wly = = ‘Thus C(p,w) is not positive semidefinite. We saw in (b) that the aggregate demand function satisfies the WA. We can thus conclude that in order for an aggregate demand function to satisfy the WA, It ig not necessary that the matrix C(p,w) is positive semidefinite (@) Here ic a figure depicting the wealth expansi ns paths of the two consumers for p = (1.1). They intersect each other at (4,4) because x,(1,1,8) = xp{li,8) = (4,4). Note that if 8 < w < 16, the Engel curves resemble those of figure 4.C.2(b); if w > 16, the Engel curves resemble those of figure 4.C.20a). consumer 2 xh 4 consumer 1 0 4 xi Figure 4.C.11(@) 4.C.12 As suggested in the hint, our example is a two-commodity, two-consumer economy in which the twa consumers have the same preference and the wealth distribution rule is such that when the aggregate wealth is equal to 4, the wealth of consumer 1 is equal to 1 and the wealth of consumer 2 is equal ta 3. The example here is essentially the same as Example 4.C.1 (which is 4-12iustrated in Figure 4.C.1). Example 4.C.1 is not directly applicable to the Present context, because the two consumers Imave the same wealth but the different demands. However, if wealth levele are different between the two consumers, then it is possible that they have the same preference and yet demands similar to those in Figure 4.C.1, yielding a violation of the WA in the aggregate. This is illustrated in the figure below: Figure 4.C.12 4.6.13 We consider a two-commodity, two-consumer economy with the given wealth distribution rule w,(p,w) = wp,/(p, + pp) and wa(p,w) = wep/(p, + pp). The preferences of the consumers are represented by the following utility functions: Ujbg) = Min (4 ),2%5)), Up(X5) = Min (2x,5,%59). So the preferences are homothetic aud have L-shaped indifference curves. The unique wealth expansion paths are depicted in the figure below:by consumer 2 consumer 1 x Figure 4.0.13 Just as we saw in the answer to Exercise 3.D.5(c), the individual demand functions are xy(pow,) = (2w,/(2p, + pp). wy/(2D, + PL) XplPrWy) = (W/(D, + 2Pp) 2wy/(p, + 2P>))- The aggregate demand function is given by xlp.w) = ¥,(p.w(p.w)) + x,(p.w,(P,W)). We claim that the aggregate demand function does not satisfy the WA. To see this, define = (p € R2,: p, + Py = 1) and our restrict attention to p € a and w= 1. Then wi(p,w) = p; for both i and hence lim, ‘p31, 07 paf0,1yetPl) ~ Him, ¢o,ypKplPomptP.0) = (1/212. Thus, if p< 4, q€ A, and p, and a, are sufficiently small, then q-x(p,1) <1 (B.A) = img sc gy%y(Pr¥7 (PD) = (11/2), and px(ql) < 1. Hence the aggregate demand function does not satisfy the WA. Proposition 4.C.1 does not apply to this example because the wealth distribution rule of this example depends on prices, while the proposition does not allow thi4:D.1 It is casy to check from the budget constraints that the distribution (x,(p.wy(p.W),....{(P.w4(p.¥))) of commodity bundles satisfies the constraints of the maximization problem in this exercise, and, from the definition of the indirect utility functions, that it attains the value v(pw). It thus remains to show that if (,,...,%,) satisfies the constraints of the maximization problem in this exercise, then W(u(x,),...sty(x,)) < Mipiw). For each i, define w; = prxy, then Sw, = Eip-x, = p-(Fix) = w, that is. (wy,...,W)) satisfies the constraint of the maximization problem of (4.D.1). Hence, by the definition of v(p,w), WO (p.w,),-...¥ (Pw) 5 vip,W). ye On the other hand, by the definition of the indirect utility functions, u,(x,) = vi(prw;) for every i. This wid the Increasingness of w+) imply that Wu Cees OED) = WO lpr Ire emp). Hence W(u,(x,),....U,(x))) $ vip.w). This completes our proof. 4:D.2 To check that vip,w) is increasing in w, let p be a price vector, w and w' be two aggregate wealth levels with w sw’, and (w,,....W;) be a solution to the social welfare maximization problem of (pw). Then ¥(P.w) = WUyy(P,W,)oooorv gC). Also, yy; = w and hence Tw, Sw’. Thus, by the definition of v(p,w'), Wor y(P.)o-a¥y(P.w) = vipAW'). Hence v(p,w) 3 vip,w’). To check that v(p,w) is nonincreasing in p, let w be an aggregate wealth level, p and p' be two price vectore with and p’ and let (wy,..4Wy) be a solution to the social welfare maximization problem of (p',w). Then YEP W) = Wiv(B',W,)..--.¥P'w)).Also, v,(p',w,) # vj(p,w;) for every i because p’ = p. Since W(-) 1s increasing, this implies that WY wens ¥yP DP) = WE (Pow so sy (PH) By the definition of vip,w), v,(p.w)) = vipw). WO {p.w)). Hence v(p',w) 5 v(p,w). To verify the homogeneity of degree zero and the quasiconvexity, we apply the equivalence of the two maximization problems established in Exercise 4.D.1. For any (p,w) and A > 0, the two price-wealth pairs (p,w) and (Ap,Aw) give the same constraints to the maximization problem of Exercise 4.D.1. Hence vip.w) = v(ap.Aw), implying the homogeneity. As for the quasiconvexity, let v eR. Let (pw) and (p’,w’) be two price-wealth pairs such that v(p,w) 5 v and v(p',w') =v. Let A € (0,1] and define p" = ap + (1 - Alp' and w" = Aw += ADW. Let Oxy..sy%)) be a solution for (p'yw"), Then p'(Fj3e)) < w" and hence ap-(Dx) + (1 - Alp (x,) s aw + U - aw’. We must thus have either w = p-(J,x,) or w = p'-(J,x,). Hence we must have either Wu lx)soetylep) # vip.w) or WUE). t)O%)) € MPIW). Tn either case, we have W(u,(x,),-...ty(%,)) v. Hence v(p,w) is quasiconvex. 4.D.3. The welfare maximization problem is now rewritten with nonnegativity constraints: MAK) Wy (Pw ¥ (PH) sit. ry, s™, w, = 0 for all i We assume that x,(p.0) = 0 for every i and every p >> 0. This assumption is 4-16satsied if for example, the consumption ste are ll SL. Then Yyf.0) © 0 for every i and every p 2 0. ‘The Kulu-Tuckereondtlons for the Social welfare maximization (Theorem M.K.2) are that there exist A > O and 4H; 0 such that (aw /oaptoryow)) — A+ py = 0 for every t= am w, = 0 and uw, = 0 for every i = 1. where all derivatives are evaluated at the solution (eyes) = (04 (P.W)..os¥7G(B.¥O), i The Envelop Theorem (Theorem M.L.1) implies that av/aw v7 8p, = Yi(BW/0u,1av,/p,). Now define J = : w,(p.w) > 0). Since 0 for every i € J, av/aw = (ewsa (ov,/w,) for every i ¢ J. Since av,/épy = 0 for any i € J, Bv/apy = Tyes(OW70uMOv70p,). Thus, 8v/aPy (aW/éu. ) (av OP») (aW/8u;)(av,/8p ») ~ wee ~~ Tes ——ae79w ——* > Ties Ta 785, Ter ERT av, JAD» ' t i = Sie ar raw = Lie rilPylrw) = Lyx(p.w,(p,w)) i Hence the Walrasian demand function derived from v(p,w) equals Y,x;(p.w, (ew) and the proof is completed. 4.D.4 (a) If the u;{+) and W(-) are monotone, then so is u(-). If the u,(-) and W(-) are continuous, then the composite map (x,,....%;) > Wuj(24)),---.uj(4p) is also continuous. Thus, by the Theorem of the Maximum (Theorem M.K.6), u(+) is also continuous. If the uj(-) are concave and W(+) are monotone and concave, then u(+) isconcave, This cau Ue proved as follows. Let x eR’, x © wb, aud A © [0.1] La Mm ax + (= Axe RE, 2t,) € ato) € RE Define x" ~ ax + (1 Ade RY Let Otro) © RM Gch axt) RU Typ =x ph = x? wld = Wl Oe). .ugbe), and ube) = pean te erin eee eee ee eae Wluylxp)ensti(e)). Define Oxfs..axt) € RT by xt = ax, + (1 = Alx} for each i. Then Six! = ax + (1 - A)x’ and hence wax +L = AR) = WCU. XD) By the concavity of the uj(+) and the monotonicity of W(-), Wu) uC) = wau(x)) + 1 - Alu 0x). coe AU) + ee Adu Og). By the concavity of W(-), Wu bx) + UL ~ Adu) OG), vy Au_OG) + CL = AluyOG)) = Au(x) + (1 - Alu(x') Hence u(Ax + (1 - A)x’) = Au(x) + (1 - A)u(x’). Hence u(-) is concave. It is worthwhile to point out that the quasiconcavity of W(-) and the u,(-) does not imply that of u(-). As an example, let L = 2, 1= 2, u,(x,) = 2 2 - X}y Ug(%y) = XQg and W(u,,u,) =u, + u,. Then W(Z,0) = W(u,{2,0),U(0,01) = W14,0) = 4, W(0,2) = Wluy(0,0),u,(0,2)) = WO, WOLD = W(u,(1,0),u,0,0) = WEL) = 2. (b) We shall first prove that, for every x € RY, if there exists (x,,....%,) € MT that satisfies J,x, = x and Is a solution to the maximtzation problem of Exercise 4.D.1, then x is @ solution to the maximization problem of this part. In fact, suppose that Oxy) © RU pix, x, and Oxy.) Is a solution to the maximization problem of Exercise 4.D.1. Let x’ ¢ RU and p-x’ Swe Let Oxo) € RM, Soe s xt, and ula") = Wiley Op),..th CP). Then p-(J.x}) = w. Hence Wu, (x)),...y0e)) = (uy (4). 406). Since a8TX; X, Wlur, (4), (%))) ¢ ulx). Thus ux’) = ube), Hence x is a solution to the maximization problem of this part. By Exercise 4.D.1, (x)(b04j(P,¥y---ey(Pr¥(P,W)) is @ solution to its maximization problem. Thus, by the above recult, J.x(p,w;(p,w)) is a solution to the maximization problem of this part. Hence the Walrasian demand function generated from it is equal to the aggregate demand function. 4D5 The ie no positive representative consumer if the WA is violated. Example 4.C.1 thus serves as an example for this exercise. 4.D.6 The social welfare maximization problem is now written as Max, Tajlnv,(p.w) (ype) SEEM st. ty sw. ‘The first-order conditions are that there exists A > 0 such that (aw /au,18v,/0w,) ~ A for every i, where all derivatives are evaluated at a w solution (w,, Ww). By the definition of W(-), aW7éu, = a,/v,{p,w,). By Exercise 3.D.9(b), vi(pyw;) ia homogencous of degree one in w; and lence 8v,(p.w,1/aw, = v,(p.w,)/w,. Hence the left-hand sides of the above first-order conditions equal (a,/v,(p.w,))(vj(p.w,)/w,) = a,/w, for every i Thus w, = @/A. Since Ya, = 1 and Sw, = w, w= 1/A. Hence w, = aw. Thus Wi(P.w) = aw. 4.D.7 The social welfare maximization problem is now written as Mau, aw) Yap) + byw) ae Ty, = Thus any (w,....,w,) with Siw, = w is a solution to this problem and v(pw) = 4-19Tap) + otpiw. 4.D.8 Suppose that (p',w') passes the potential cumpensation test over eee ee (iw). ‘Then there exists (rj...wf) © RU ouch that Fyw! < w and vi(p'.w)) > vdp.w;(p.w)). Since W(+) is increasing, W(v,(p',w}),....v,(p'.w;)) > v(p.w). Ry the definition of v(p'.w'), vip'w') = W(v,(p'sW))s....¥{(P',wp))- Hence vipyw') > vipyw). 4.D.9 Define A, (x,) = ujG%)), then A= FiA;. Then, for every p >> Ooxe RE is @ solution to Mingey PY if and only if there exists (xy,....%)) © Ay x. x Ay such that D2 and, for every i, x, is @ solution to for every 1. Hence Minlp-y: y € A) = ie\(p.u,() On the other hand, by the definition, Min(p-y: y € B) = e(p,u(x)). since A ¢ B, Minfp+y: y © A) = Min(p-y: y € B}. Hence g(p) = elp,uGo) - Fe, (p.uy! for every p. By the definition, g(f) = 0. Hence, by the second-order necescary condition for a maximum (Section M.K), D@g(p) is negative semidefinite. Since D”gip) = S(p,uG) - F,S,(p.uy ,)) by Propositions 3.C.3 and 3.6.3, S(p,ulx)) - J,S,(p.u(%,)) is negative semidefinite. 204.D.10 As we saw in the answer to Exercise 4.C.1l(b), for every (p,w), S(p,w) is symmetric, ond if dp R, dp ¥ 0, and dp is not proportional to p, then dp-S(P.whdp < 0. Thus, according to Section 3.H, there exists a positive Tepresentative consumer. But according to Exercise 4.C.lllc), if 8 < w < 16, then the matrix C(p,w) is not positive semidefinite. Hence, according to the small-type discussion in Section 4.D, there is no normative representative consumer. 4D. We shall give an example in which L = 3 and C,,(p,w) # Cy\(p,w). By the definition, ey00.80 = SBlureNek aontneame,tp.niirdn ~ (520/218 tp. ni1/0w\dnI%(0. ni) Here, by integration by parts, Ien/2nak (p.niv\/awldn = Spln/2ieN a (p,nie)/andan ne ar = G/W)X(p,2W) - (ew) (p,W). = lonandktpawit SGW/2WE,(p-nwan Hence =.) _ 2, Se Cyalp.w¥) = SGUW/2Va%(pmW)/awiz,(p,nWwIan + CINE, (p,28 x, 19,0) + (1/28 (Py p Py Similarly, mye ae Coil.) = So(1/2)%e (p,mw)L8%,(P.nw)/awldn + UA PK (p,2H) + Waive, (p,NX,(,W). Again by integration by parte, (see SHUN (7. nia, (paw) /awdn 4-21SR 2N% (min lp mis) = 172% (wom, Kp-MWITEG ~ SGU/2WNAE (pomw Zaz y(v,4W dn = (/2WNK (pw) (0,20) ~ IZU/2N2%,p.ndn7aW)x,(p.nindn It is thus sufficient to obtain an example in which Dae a wm we ne So(l/2NOx, (pinw)/ow)xp(penw)dn - (Wx (P2wxy(p,w) + WRK (P.AMElp.aie) ~ LG /21E% tp, nied awry {o.nidhin = W/ Wx (.¥p(P.20, or, equivalently, S302 8% (p.niv) awl. mind + (ZINE .AIE,(p.2H) + UE (p.2vx,l9,8) ~ U/ x ,(p,WEP.20. so consider a preference, a price vector p, and the average wealth w such that: K,(p.2i) = 2 for both £ = 1,2, xp.) = SGlu/2)%,(penwldn = 1 for both £ = 1,2. (Another reetriction will be given shortly. ‘The demand for good 3 ts cetermined by Walras’ law.) Then the right-hand side of the above inequality is equal to 2%. It ie then cufficient to show that SABE (p.-nWV/awlk,(p.nvldn # 2/1. But if the graphs of the functions » > X,(p,nW) are as in the figure below, then the first term can be made as close to zero as needed. The above inequality ean thus be ectabliched.Figure 4.0.11CHAPTER 5. 5.B.1 The first example violates irreversibility and the second one satisfies this property, Figure 5.B.1 5.8.2 Suppose first that ¥ exhibits constant returns to scale. Let ze wi? and «> 0. Then (- z, f(2)) ¢ ¥. By the constant returns to ecale, (- az, afi) € Y. Hence af(2) = flaz). By applying this inequality to az in place of 2 and a” in place of a, we obtain @“f(az) = fla” “az)) = f(2), or flaz) = af(2). Hence faz) = af(z). The homogeneity of degree one is thus obtained, Suppose conversely that f(+) is homogeneous of degree one. Let (- 2, q) ¢ Y and @ = 0, then q = f(z) and hence aq < af(2) = flaz). Since (- az, flaz)) € Y, we obtain (- az, aq) € Y. ‘The constant returns to scale is thus obtained. 5.B.3 Suppose first that Y is convex. Let z, 2’ FL! and a « (0,11, then(2 FN © ¥ and (- 2", sz) € ¥. By the convexity, ( (az + 1 - az), aff) + 1 - wifiz) © y, Thus, af(2) + (1 ~ @lf(2) = Flaz + (1 - ade). Hence F(z) is concave Suppose conversely that f(z) is concave. Let (a, - 2) € Y. (= 2) Ys and a ¢ (0,11, then q = s(z) and q = /(c’), Hence eq + (1 ~ alg’ s aflz) + (1 - afte’), By the concavity, f(z) + (1 - adf(2') s flaz + (1 ~ ale’) Thus aq + (1 - aq’ = flaz + (1 - ale’), Hence (faz + (1 a2"), ag + = alg’) = af 2, a) + (1 ale 2’, ae. Therefore Y is convex. S-B4 Note first that if Y itself! Is additive, then ¥" = Y by the definition \ of the additive closure. This applies to Figures 5.B.4, 5.B.6\a), 5.B.6(b), 5.B.7, and 5.B.8. So we shall not depict them, {Let's now take up the cases in which the production set is not additive, Note first that Y* is equal to the set of vectors of AY that can be Fepresented as the sum of finitely many vectors of Y. If the production set 's convex, ac in Figures 5.B.1, 5.8.2(a), 5.B,2(b), 5.B.3(a), 5.B.3(b), and S.B.5(a) (which is the same as 5.B,2(b)), then we have the following stronger Property: Y* consists of all "multiplied" production Plans. To be more Precise, for each positive integer n, define ny ¢ RE by nY = (uy © al ye yy We then claim that ¥* mY. In fact. if y € nY for some n, then (1/n)y € Y- By the definition, (1/nly © ¥* and hence y = nl(i/nly) € Y*. Thus ¥* > UpePY. To prove that ¥* ¢ Upei"®, it is sufficient to show that UR AY is € 5-2additive, by the definition of Y". So let y € UP ynY and y' € UY ,n¥. Then thre exist positive integers n and n’ cuch that y ¢ nY and y' ¢ n'y. Thus (my © ¥ and G/n' iy’ € ¥. Since ¥ is convex and u/(u + a") + a(n + a") = 4 (on + any) + (niin + nINLZAIY) € Y. That is, (1/(a + n'y + y") € Y. Thus y + y" € (n+ ny. If the production set is not convex, as in Figure 5.B.5(b), then we no longer have ¥" = U_n¥. As we can see in the above proof, while we still nave Y* > uP snY, we need not have Y" ¢ U" jnY. That is, it may be true that some production plans in ¥* can be attained only by allocating different inputs to different production units. This point can be formulated as follows. Let (y*,y$) be Ute production plan at which the average return is maximized. That is, y$/Iy$l > yp/ly,! for any other y = (yyyp) € Y. Assume that the function that associates each y, <0 lo Ue average return at y, Is quasiconcave. (This oppears to be truc from the figure, It is equivalent to saying that the average return function is single-peaked.) Let y, <0 be an (aggregate) input level and n be the positive integer such that ny} < y, < (n= Dyf. Let (yyo--¥,,) be an outpur-maxtmizing allocation of aggregate input y,. (The number J of production units to be used is also being optimized here.) We shall prove that one of the following three cases must then apply: (@) J =n Land yy) = yy/tn ~ 1) for every J Gi) J-= n and yy; = y,/n for every j. (ili) J = nm and there exists k € (J,...,J) such that yy = y, ~ (nm - Dyy and y, 4 TT for any J =k. To prove this, note first that we must have either y, = yf for every J, or ¥,j <¥p for every j. In fact, if neither of these applies, then a small J 5-3Input reassignment from a production unit with y, < y} to one with y, > y> increases the (aggregate) output level, because the average return increases. fy) 5 yf for every J, then the average return is decreasing at every y,; and hence the y, ; must be all the same. By the quasiconcavity. J must be the maximum integer that satisties yj = Jy, (andy, $ yf). Hence J =n 1 and @) applies. If y,, = yf for every J, then the average return is increasing al every yy. If the y,; are all the same, then, by the quasiconvexity, 3 must be the minimum integer that satisfies yf = Dy, (and y, = yf). tence 4 =n and (i) applies. So suppose that some of the y, are different. If there exist two production units for which y,,> yf, then a small input reassignment from one to the other would increase the output level, because the average return is increasing. Hence there must exist at most one k for which yy 7 y+ By the guasiconcavity, yy) = yf for any J * k. Thus (ii applies. This last case happens when the average return decreases very fast after the input level goes beyond yf. For Figure 5.B.1 For Figure 5.8.2(a)For Figure 5.B.2(b) For Figure 5.B.3(a) 55For Figure 5.B.3(b) For Figure 5.B.5(b) 5-6SBS Lety © Y andve- sb, Then, for every ne, nv e~ BE and hence nv ev tye - ab, since ¥ is conven, = umy + usm = = my + ve Since ¥ is closed, y + ¥ = lim, (1 - Waly + v) € ¥. 5.B.6 (a) From the given functions #,(+) (i = 1.2), the production set is defined as Y = (ly,,yqyq): there exist q, = 0 and q, = 0 such that q, + 4, = 4 and - y, = 9,(q,) for both 1}, A three-dimensional production sel is depicted in the following picture, assuming that the 9,(-) are convex. output input 2 input 1 Figure 5.B.6(a) (U) We claim that the condition that #9, + gi) #(a,) + $,(aj) for all g; = 0, gj > 0, and i = 1,2, is sufficient for adé ity. In fact, let (y)¥2.a) € Y and (yjyp.@") € ¥. ‘Then there exist (q,,q3) such that , + q, = q and - ; ® #4(a,). and (a), a3) such that q+ yD = q' and - ie ¥lay). Then (+a ttt aratar = (y+ YP = Mlqd + Aa}) = Ala, + ai) 5-7Thus (y, + ¥}. Yo + 93. 9 +41) © Y, establishing the additivity. (©) Let the output price be p. The first-order necessary conditions for Profit maximization are that, for both i, w,éi(q,) ©. The interpretation is that marginal cost (in monetary term) due to a unit increase of output level must be smaller than ar equal to the output price, and the former must be enual to the latter if the output level Is positive. If both #(-) and $4(-) are convex (so that the corresponding production Set is convex), then, according to Theorem M.K.3, these first-order necescary conditions are also sufficient. (€) Let q > 0. By renumbering i = 1,2 if necessary, we can assume that 16(@) = w29,(q). In order to prove the statement, it is sufficient to show that, for any q, > 0 and a, > 0 with 4, + 4) = a, we have ¥,9,(4) + *29,(4,) ? W,0,(q). In fact, since the #.(+) are strictly concave (and #0) = 0) and aa + wea 1, W19(4) + WyF9la9) > wy(a,/a)04(a) + wela,/al¥,(@) = (ay/aw 9) + (a,/aw,6,0) = w,4,(q). The statement is thus proved. The strict concavity af the OC) te interpreted as the increasing returns to scale, which makes the statement quite plausible: Under the increasing returns to scale, it is better to concentrate on one technique. ‘The isoquants of the input use is drawn in the following figure. Note ‘that the additive separability imposes the same restriction on the isoquants as that alluded to in Exercise 3.G.4(b). 5-8input 2 ; input 1 Figure 5.B.6(d) S.C.1 If there ic a production plan y ¢ Y with p-y > 0, then, by using ay ¢ Y with a large a > 1, it is possible to attain any sufficiently large profit level. Hence n(p) = . If, on the contrary, p-y s 0 for all y ¢ Y. then a(p) 0. Thus we have either n(p) = + or nip) = 0. 5.6.2 Let p >> 0, p’ >> 0, a [0,1], and y € ylap + (1 - ap"), then pry nlp) and p'-y < r(p’). Thus, (ap + (I~ @)p)+y = ap-y + (1 ~ ap'+y < amlp) + (1 ~ adn(p’?. Since (ap + (1 - a)p')-y = wlap + (1 - ap"), map + (1 = ap’) = ax(p) + (1 - adn(p’). Hence m(-) is convex. 5.C.3. The homogeneity of c(-) in q is implied by that of 2(+), We shall thus prove this latter homogeneity only. Let w >> 0, q=0, anda>0. Letze z(w.q). Since f(+) 1s homogeneous of degree one, faz) = aftz) = ag. For every 2 € RL it giz’) = aq, then flalz’) = a ly(2") = q. Thus, by 2 €2long)s welale") = wz, Heuce wz" 2 we(az). Thus a2 € 2(w,eq). So az(w,q) © z(w.aq). By applying this inclusion to @”! in place yf @ and aq in place of we obtain a'z(w,aq) ¢ ztw.a”Maqll, or 2(w,aq) ¢ az(vr,q) ond thus conclude that z(w.aq) = az(wq). We next prove property (vill). Let we x\7', q 20, q' 0, andae (0,1. Let 2 € 2tw.q) and 2’ € z(w.q'). Then f(z) = q, fl2') = @', clw.q) = wz, and elwsq!) = wee’. Hence ac(w.g) + (1 = adelw,q") = a(w-2) + (1 - adlw-2? = woz + (1 = az"), Since f(-) is concave, Flaz + (I~ ale") = aflz) + (1 ~ adfle’) = aq + (1 - ada’. Thus w-(az + (1 - az") = cw, aq + (1 - alg’ That is, ee(w, @) + (I~ alow, 9") 2 c(w, aq + (1 = ald’), 5.0.4 irst printing errata: When there are multiple outputs, the function f(z) need not be well defined because it is conceivably possible to produce different combinations of outputs from a single combination of inputs. Assuming that the first L - M commodities are inputs and the last M commodities are outputs, we should thus underctand the set (z 2 0: f(z) © q) as tz © RY: (- 2, q) € YI.) For each q = 0, define Y= @eER™M Gz ge vee ™, F229). Then c¢(-,q) is the support function of Y(q) for every q. Hence property (ii) follows from the discussion of Section 3.F. Moreover, according tu Exercise 3.F.1, if ¥(q) is closed and convex, then L-M, Yiq) = @ « wt wee ® olwyq) for all w >> 0}. Since Y = {(- z, a): q = 0 and z © Yiq)), this implies property ( To prove property (iv), let w >> 0, q = 0, a> 0, z € zlw.q), and 2 « Yq). Then w-z = w-z’, Hence (aw)-2 < (ew)+z'. Thus z € zlew,q). Therefore 5-10z(w,q) ¢ z(aw,q). By applying this inclusion to aw in place of w and a in place of a, we obtain zlaw.g) © zlu faw)y) - z(mq). Property (iv) thus follows. Property (iv) implies the homogeneity of degree one of c(-) in w, which is the first part of property (i). As for its second part, let w >> 0, q = 0, q’ 2 0, and q’ = q. Then Y(q’) ¢ Y(q). Since c(-,q) and c(-,q’) are the support functions, this inclusion implies that c(w,q’) = c(w,q). Hence e+) i nondecreasing in q. = YQ) ae RE: As for property (v), note that 2(w,q) wez = clw.q)} for every w >> 0 and q 20. Since both of the two sets on the right-hand side is convex, so is the intersection, and hence so is z(w.q). As for the single-valuedness, let q = 0, w > 0, z € zlw,q) , 2° € 2(w,q), and 7 # 7. Also suppose that Y(q) is strictly convex. By the convexity of z(w.a), (22 + (72)2" € 2lwrg)- By the strict convexity of Yiq), there exists a 2" € Y(q) such that (u/2)z + (172)2" >> 2". Hence w:((1/2)z + (1/2)2") > w-2", which contradicts (1/2)z + (I/2Iz" € zlw,q). Thus z(w,q) must be single-valued. Property (vi) follows from the fact that c(-,q) is the support function of Y(q) and the duality theorem (Proposition 3.F.1) Property (vi) implies that if 2(+.q) is differentiable at W, then p,2W.q) = D,(V,clW.4)) = DZclW.q). As a Hessian matrix, thic ic eymmetric. By property (ii), it is negative semidefinite. By property (iv), D,2(,q)W = 0. Property (vil) 1s thus established. 5.C.5 If the production function f(+) is quasiconcave, then the set Y(q) = (z L+1, e RL: f(z) = a} is convex for any q, and thus property (iii) holds. (Notethat we uscd the convexity of Yiq), not of Y.) If there is a single input and the production function is given by f(z) = 2°, then it is quasiconcave but the corresponding production set exhibits increasing returns to scale. Quasiconcavity is thus compatible with increasing returns. 5.C.6 Throughout the following answers, the input prices are denoted by w >> 0 and the output price is denoted by p > 0. For convenience, we denote by 2lp,w) the input demands at prices (pw). As a preliminary result, by using the implicit function theorem (Theorem M.E.1), we shall prove that 2(-) is a continuously differentiable function and give Its derivatives in terms of F(-). (To be rigorous, we need to assume that f(-) ia twice continuously differentiable and the input demands are always strictly positive, so that the nonnegativity constraints never bind.) Since D(z) is negative definite for all z, the first-order necessary and sufficient condition (Theorems M.K.2 and M.K.3) for profit maximization is then that z Is an input demand vector at prices (p,w) if an only it pvr(a) - w= 0 If we regard the left-hand side as defining the function defined over (p.w,z), then the function is continuously differentiable and its derivative with respect to z is equal to pD*f{(2). It is negative definite, and hence has the inverse matrix (at every z). Thus, by the implicit function theorem (Theorem M.E.1), for each (pw), there is a unique z for which pUf(z) - w = 0 and the mapping from (p,w) to z is continuously differentiable. This is equivalent to saying that z(+) is a continuously differentiable function. The implicit function theorem also tells us that YpIb*s(2(p,w VFzip,w)), a2(p,w)/apvy2tP.w) = (/pID*ste(p,w))™ Note here that, since ne y(ztpw)) is negalive definite, so is its inverse De ytz(pw)t by Theorem M.D.1Gii). (a) By the chain rule (Section M.A), So Us tetp win = UFlelp.0) FE (oo) L/plUftatp,wi)-b?s(2¢p,w)VFCalp.w)) Since D2f(z\p,w))"" is negative definite, uls(z(p,w))I/dp > ©. () Since af(z)/az, = 0 for all £ and diflz(p,w))I/dp > 0, as the output price increases, the demand for some input must increase. (6) Since V,2(p.w) = (/pID*statp.w)Y', dztp,wi/ép, is equal to the &h diagonal entry of (1/pID2stetp.w)1. Since D?f(2(p,w))"" is negative definite, the diagonal entry is negative. Hence so is d2(p,w)/épy. 5.0.7 (First printing errata: The condition o”/(2)/6z,3z, < 0 should be 2 (z\/ézq02, > 0. That Is, al inputs are complementary to one another.) As we saw in the answer to Exercise §.C.6, az - 2 rtatp.wiit 22 p.m = Crd rlatpswNl and (7p it0? slalp.w VF 20.) oz Hem az, Hence, in order to prove that 6z,(p.w)/aw, < 0 for all k # land e (pew) > op © for all t, it is sufficient to show that all entries of (D@p(a(p,wl” are negative. This is an immediate consequence of the celebrated Hawkins-Simon condition, which can be found, for example, in "Convex Structures and Economic Theory" by Hukukane Nikaido. Here, instead of simplemindediy quoting the Hawkine-Simon condition, we shall provide a direct proof that relies on the symmetry of D“F(z(p,w)) (which is not assumed in the Hawkins-Simon condition).Write H = D?f(z(p,w)). To show that all entries of H” are negative, it is suffirient ta prove that, tor every ve RM", if Hy = O and Hy = 0, then v La << 0. In fact, then, for each &, we can choose v € RU! so that Hv is the vector whose &th coordinate is equal to one and the other coordinates are equal to zero. Then H (Hv) is equal to the éth column of Hof course, it is also equal to v, which is claimed to be strictly negative. Thus every column of H™ is strictly negative. Hence all entries of H”! are negative. For this property, in turn, it is sufficient to prove that for every v © nil if Hv = 0, then vs 0. (That is, weak inequalities suffice.) In fact, it there excte av 6 RL! uch that HY 20, Hy # 0, + #0, andy = 0 for some £, Then v # O and hence Ye glv.w)/Ow, yy. = Deel BzylPiw)/dw, vy <0, wich contradits Hy & 0. We shall uow prove by contradiction that for every v € at if Hv 2 0, then v 5 0. Than there exits vc RE ouch thet Hv = 0, and vy > 0 for some & By re-ordering the inputs if necessary, we can assume that the firet M entries of v are positive and the last L - 1 - M entries are nonpositive hx, Write write v = [ tM 1 M >» < =< iso, 2) TH, H, write H = [ , Hy H, M) matrix 1} where H, is an Mx M matrix, Hp is Mx (L - ‘whose entries are all positive, H, is an (L - 1 - M) x M matrix whose entries are all positive, and H, is an (L - 1 ~ M) x (L - 1 - M) matrix, Let Hv = L-1-M y, finesse pe ty. J, : BE) te d-C 9 = 0 because y, = 0, x, $0, and all entries of H, are Hence Hyx, = y, ~ H, S14positive. Thus, by x, >> 0, we obtain x,-H,x, s 0, which is a contradiction to the negative definiteness of H, aid hence that of Hy. 5.C.8 The cost that AI incurred in month 95 is 2-55 + 2-40 = 190, but it could attain the same output level with a lower cost by using the input combination of month 3: 2-40 + 2-50 = 180. Thus the problem we will encounter is that, perhaps due to mis-observation and/or some restrictions that Al faced outside the market, the profit-maximizing production plans are not observed tu have been used and it is Impossible to use those observations in order to recover its technology based on Proposition 5.C.2(iii) or 5.C.1( iii). 5.C.9. To find n(+) and y(+) for (a), the first-order condition (5.C.2) is nol very useful, because one of the nonnegativity constraint binds. Also, to find n(+) and y(-) for (b), it is not even applicable because s(+) is not differentiable. In Uuth cases, however, because of the nature of the is similar to production functions, it is quite easy to solve their CMP (wi those in Exercise 5.C.10.), and the cost functions c{-) turn out to be differentiable with respect to output levels q. We can thus apply the first- order condition (5.C.6) (which requires only Une differentiability of the cost function with respect to output levels) to find profit maximizing production levels, and hence the profit functions and supply correspondences. Throughout the answer, the output price is fixed to be equal to one. Vaw, if w, 5 W, 1 = 2 (a) aw) i L/w, if w, > Wy 2 , (- Vvawy, 0, 1/2W, 1) if w, < w, = = ) w, ylw) = 4 - > 2p, Vaw,): 2, = 0, 22% 0, 2, + 22 = WV4wy) it wy = woi 2 ' (0, - 1/4w5, L/2w)) if w, > We.(©) Ow) = Law, + wp) 2 yo) = © 1780 ow), 4lur, + w, 1 W72(w, + wo). 2! 2) (c) Note first that this production function exhibits constant returns to scale. Moreover, if p <1, then the nonnegativity constraint does not bind. If p = 1, then this production funetion gives rise te the same isoquants as that of (a), and hence one of the nonnegativity constraints binds. It is thus casy to apply (5.C.6). If p <1, then ig WOON 4 yO) ¢ , nw) = . 0 it WP 4 p/P sy, 2 if wi ppl), ypAle-L) ¢ 1 2 gto =f ale WPD, - ylMOD, PPL), yPMEV)1/0, g = o) if wi 4 (o it WOE) 4 WhO) yy PPL) y yp p-1) 1 Wo If p = 1, then { © if Min(w,,w} = 1 RO) =) if Mintw,.w,) <1 10) If Min(wy.wy? ® is 2 if Min(w yw} <5 {a{- 1, 0, 1): w= 0} ifl=w, < wo: FO =) (a(v, = 1, 1): a= O} ifw,> wy = 1. {al- 2), = 25» I: a= 0, 2,2 0, 2,20, 2, +2, » 1 5.c.10 aw, if ws Wi @ewaey 1 1 2 aw, if w, > wy(4.0 zlwia) = 4 (24.29) © RE (0,q) (©) cOw.a) = (wr, + wp) ZhHa) = (qa). (0) ctwig) = awh OD 5 wPMern ve? 1 twa) = atwhP"D were) Ve ye ye-D) : . aa 5.C.11 Ascume that ¢(+) is twice continuously differentiable. By Proposition 5.C.z(vi), 2(+) is continuously differentiable and aaylwr.gl/aq ~ (B/8q)(8elw.q)/ow,) = (4/8u,)dclw.0)/80). Hence a2, e marginal cost is increasing in Wy. (w,q)/aq > 0 if and only if (8/8W,)(8C(w,q)/aq).> 0, that is, 5.C.12 Suppose first that y © Y maximizes profit at p, then (p.m(p))-(y. - 1) Also, for every a = 0 and y’ € Y, (pym(p))-(aly’, ~ 1) = alp-y" - x(p)) = 0. Thus (y', - 1) maximizes profit Y' at prices (p.n(p)). Conversely, if y € Y and (y, ~ 1) maximizes profit in Y' at (P,P, ,) then (P.Ppqy)"(s ~ 1) = PY ~ PLyy * © by the constant returns to srale. Also, for every y’ € Y, (P,P, 4))°(¥'s ~ 17 = Pry’ ~ Ppyy 50 Hence y maximizes profit in Y at prices p and m(p) = p-¥ ~ Pryy 5.C.13 Denote the production function of the firm by f(-), then its optimization problem is Maris zpdeo Uae) St MR * Mah * c ‘This is analogous to the utility maximization problem in Section 3.D and the function R(-) corresponds to the indirect utility function. Hence,analogously to Roy's identity (Proposition 3.G.4), the input demands are obtained as WyR(P.W.C) = (aC/w,, (= aCe, w 1 2 S.D.1 We chall use the differentiability of O(-) only at §. The everywhere differentiability is not necessary. Ry the definition, wor) = 2 SD). SG zi q q ‘Thus, if the average cost is minimized at q ~ G then AC'(@) = 0 and hence og) og) = Cl@Za = Acta). S.D.2 Figure 5.D.2 S.D.3 Let q(w,p) be the profit-maximizing output level when the input prices are w and the output price is p. Lat # he the initial input prices, p be the initial output price, Z be the initial long-run input demand at (w,p), and q be the initial long-run output level at (W.p). By (5.C.6), acW.a(W.p))/éq P for every p (assuming that q(w,p) > 0). Thus, by differentiating both sideswith respect to p and then evaluating at p we obtain (07 eGwalwpn/0uNoql%)/0P) = ty that is, dq(WnB)/ap = (80754189). (On the other hand, define the short-run cost function function ¢.(w,q1z,) and the short-run output function ¢.(w,qlz,) as suggested in the hint. Just as we did above, we can show that aq (W,512,)/ap = (8c2(W.912,)/8"@)* (WP 2, Maly (assuming that q,(W,p1z,) > 0). Now, by the definition, c(w,q) = e(mal?,) for all q and of c,(%,a12,). Hence the function g(q) = e(w.q) - e,(v,q12,) is maximized at q = ] and, by the second-order necessary condition (see Section M.K), g"(q) = 0, that is, ac(W,4)/8%q = 8c2(¥i.41z,)/0%q. Thus aqlwsp)/ap = 24, (W,512,)/6p. 5.0.4 (a) Suppose that a = Ta) = 0) (ay 0 summing over j, we obtain Clq) < Ey_,cla,) (a faiota) = O1G,). By g over a) * Fay) By the decreasing average costs (and 6(0) Hence there ig no way to break up the production of q among multiple firms and lower the cost of production. Hence C(-) is subadditive. (b) Let M = 2 and define Clq) = Vmintg,,q,), then C(+) exhibits decreasing ray average cost, Rit let q, = (1B), a, = (81, and a ~ a) + Gy = (9,9). Then G(4,) = Ol4y) = 1 and G{q) = 9. Hence Clq) > Cla,) + Clay). Hence (+) is not subadditive (©) We shall first prove that if q = ja, and a, > 0 for every j, then Clq) Zeta). In fact, then, for every J, there exists ¥; > 0 suen that 79, 7 4 By the increasingness, Gla ,q,) 7 Gly). Thus, by the continuity and Clq) > here exists @, € (l4y,) such that Cla,q,) = Clq). Define Va. ela) th +; € (ha) such that Cla ja OF ine 8 = Ty i J then Tyl/Aa, ~ 1 and E,(v/Baiea,) = (/A)a. Thus, by the quasiconvexity, C((1/B)q) = Clq). By the increasingness, @ = 1. Hence, by the decreasing rayaverage cost, 2 Poy (q) = AGI) & Gt Tyelap = Lyla peiaa) = Lye eta) = acta) ota) For the general ease in which some q, = 0 need not be strictly positive, apply the shove resnlt to the qy +e, where © > 0 and © = (11,..1) 4 RM, and then take the limit as ¢ 3 0. The continuity of C(-) then implies that Tosa, = cla). 5.D.5 (a) The production function f(+) exhibits increasing returns if and only if f(Az) = Af(z) for all z and all A= 1. Hence, if 2’ = 2 > 0, then U2 Ft2) = 1/2) fll2! ada) & (1/22! /2\fl2) = (1/2) (2) Thus the average product is nondecreasing. The marginal product may however be decreasing on come region of output levels, as the following example shows: fz) : > 0 Figure 5.D.5(a) 2 (b) Mathematically, the consumer's maximization problem is max, uS(2) ~ z, The first-order necessary condition is u’(f(z))f"(z) = 1, which can be rewritten as u'(f(2)) = s7(2)"' Since the cost function is given by 2 = £(q), the marginal cost is equal to "(2)"! and the equality of marginal 5-20cost and marginal utility is thus a necessary condition for a maximum, Economically, the consumer will choose the output level at which the marginal utility of an extra unit of the output is exactly equal to the di tility incurred by giving up the necessary amounts of input to produce it. But the latter is nothing but the marginal cost. Hence the marginal utility is equal to the marginal cost. (©) This assertion is wrong, As the following figure shows, even if marginal cost and marginal utility are equal at an input level, there may be another input level at which the consumer attains higher utility fe a Indiffereass Curves 0 , z Figure $.D.5(6) ‘The reason for suboptimality is that the first-order necessary conditions are not sufficient when the production function exhibits increasing returns (which gives rise to nonconvexity of the feasible set). S.E.1 By applying Hatelling's lemma (Proposition 5.Gil(vi)) twice, we ret yp) = ORM) = WL (PN = Live lp) = Loy iP). 5.E.2 This is just @ matter of going through the proof of Proposition 5.E.1 5-21and checking that convexity was never used. Its interpretation was given before the statement of the proposition (p. 148), It is a consequence af The very definition of the aggregate production set, that is, it is the sum of the J firms’ production sets. It is thus independent of conve: 'y or any other properties of the firms’ production sets. 5.E.3 [First printing errata: We should assume that there is a p* >> O and y* © FY) such that pty? = pty for every y © EY, Otherwise, denoting the Profit function of J.¥; by m(-), the set (y € RY: pry < x*(p) for all p >> 0} may be empty, and all we can obtain is the equality between SY, and (y ¢ RY: Pry = n*(p) for all p= 0). This assumption is also necessary for the validity of Proposition 5.¢.1(I11). In fact, its proof should go as follows: It ig sufficient to prove that, for every ze RIVY, there exist a p >> 0 such that p-z > m(p). Since Y is closed and convex, the separating hyperplane theorem implies the existence of such a nonzero vector p. The free disposal Property implies that p must actually be nonnegative. If it is not strictly positive, then take the convex combination (1 - elp + ep* with a sufficiently small ¢ > 0, Then it is strictly positive, and satisfies ((1 - elp + ept)-2 > n((1 ~ elp + ep*) by the upper semicontinuity of m(-), which is implied by its convexity.] Since each Y, is convex and satisfies the free disposal property, L)¥; is also convex and satisfies the free disposal property. Since it is also assumed to be' closed, Proposition §.C.1(iii) implies that Uy=06 a: psy = np) for all p >> 0} fon S.E.1, w¥(p) = Pi (p) and hence L, = ye RY py < Din (p) for all p >> 0) Wee Pry = Dimitp) Pp >> 0}, S.E.4 (a) Denote by y(w) ¢ R° the set of the supplies of the technology with