ams-511-sol-05-m.nb 1 AMS 511 - Foundations Utility Theory and Portfolio Choice Robert J Frey, Research Professor Stony Brook University, Applied Mathematics and Statistics frey@ams.sunysb.edu Solutions to exercises for Class 5. Chapters refer to Luenberger’s Investment Science. Chapter 9, Problem 2 Let f(x) be the pdf of the payoff x. Note that the terminal wealth given x is W – w + x. The expected utility of terminal wealth is, therefore, -‡ ‰ ¶ -a HW -w+xL f HxL „ x = - ‡ ‰ -¶ ¶ -a W -a Hx-wL ‰ f HxL „ x = -‰ -¶ -a W -a Hx-wL f HxL „ x ‡ ‰ ¶ -¶ Thus, if we are comparing two investments with price and payoff {w1 , x1 } and {w2 , x2 }, then their relative ranking will be independent of of W. Chapter 9, Problem 6 We'll start out with the HARA expression itself. g 1-g ax y ; j hara = ÅÅÅÅÅÅÅÅÅÅÅÅ i j ÅÅÅÅÅÅÅÅÅÅÅÅ + bz z g k 1-g { ‡ Linear If we set a = 1 and then take the limit as g Ø 1, then Limit@hara ê. a Ø 1, g Ø 1D x ‡ Quadratic If we look at HARAs with g = 2, then we get the quadratic polynomial ams-511-sol-05-m.nb 2 g 1-g ax y ê. g Ø 2E j ExpandA ÅÅÅÅÅÅÅÅÅÅÅÅ i j ÅÅÅÅÅÅÅÅÅÅÅÅ + bz z g k 1-g { b2 a2 x2 - ÅÅÅÅÅÅÅ + a b x - ÅÅÅÅÅÅÅÅÅÅÅÅÅ 2 2 An affine transformation puts U(x) into the form we want with c = a / 2b. 1 b SimplifyA% ÅÅÅÅÅÅÅÅ + ÅÅÅÅÅÅÅÅ E ab 2a a x2 x - ÅÅÅÅÅÅÅÅÅÅ 2b ‡ Exponential If we set b =1 and then take the limit as g Ø -¶,then Limit@hara ê. b Ø 1, g Ø -¶D -‰-a x ‡ Power For simplicity we set a =1 and then take the limit as b Ø 0. Limit@hara ê. a Ø 1, b Ø 0D -1+g x x H ÅÅÅÅÅÅÅ 1-g L ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ÅÅÅÅÅÅÅÅÅÅÅÅ g We can further simplify this to H1 - gL1-g ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ xg g H1-gL and then make the substitution that c = ÅÅÅÅÅÅÅÅ ÅgÅÅÅÅÅÅÅÅÅÅÅ to bring it into its final form. 1-g ‡ Logarithmic The text suggests looking at the function below. We'll take its limit as g Ø 0 ams-511-sol-05-m.nb 3 H1 - gL1-g LimitA ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ Hxg - 1L, g Ø 0E g Log@xD Note that for g ≠ 0 this function is equivalent to a power law utility (with an affine transformation of its standard form). Thus, the log utility function is the dividing line between power laws with positive and negative exponents. ‡ Arrow-Pratt The Arrow-Pratt risk aversion coefficient is shown below. Note that this can easily be restated in form desired. D@#, 8x, 2<D aHara = SimplifyA- ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ E &@haraD D@#, xD a-ag ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ b+ax-bg a b-bg We now make the substituton c = ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ and d = ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ a-ag a-ag 1 a b-bg SimplifyAaHara ã ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ê. 9c Ø ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ , d Ø ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ =E cx +d a-ag a-ag True Chapter 9, Problem 14 Let p = 0.25 be the probability of winning the race. For a $1 bet we have w = 5 as the payoff on a win and l = 0 as the è!!! payoff on a loss. The individual's utility function is U(x) = x . The initial wealth is W. Our objective is to determine the optimal bet b. The expected utility is E@UHxLD = H1 - pL è!!!!!!!!!!!!! è!!!!!!!!!!!!!!!!!!!!!!!!!! W - b + p W + bHw - 1L We differentiate with respect to b and solve for the root. ams-511-sol-05-m.nb 4 sol = SolveA∂b IH1 - pL è!!!!!!!!!! è!!!!!!!!!!!!!!!!!!!!!!!!!!! W - b + p W + b Hw - 1L M ã 0, bEP1, 1T H-1 + 2 p - 2 p2 w + p2 w2 L W b Ø ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ H-1 + wL H1 - 2 p + p2 wL This is the general solution. We can now substitute the actual parameters into the rule and then recover the actual value of b. b ê. sol ê. 8p Ø 0.25, w Ø 5< 0.134615 W Thus, Gavin ought to bet roughly 13.5% of his wealth on the bet. A plot of expect utility illustrates the result. PlotA IH1 - pL è!!!!!!!!!! è!!!!!!!!!!!!!!!!!!!!!!!!!!! W - b + p W + b Hw - 1L M ê. 8W Ø 1, p Ø 0.25, w Ø 5<, 8b, 0, 0.3<E; 1.008 1.006 1.004 1.002 0.05 0.1 0.15 0.2 0.25 0.3 As an additional illustration, we'll repeat the analysis with log utility. Note that this gives a more conservative result. ams-511-sol-05-m.nb 5 sol = Solve@∂b HH1 - pL Log@W - bD + p Log@W + b Hw - 1LDL ã 0, bDP1, 1T b ê. sol ê. 8p Ø 0.25, w Ø 5< Plot@HH1 - pL Log@W - bD + p Log@W + b Hw - 1LDL ê. 8W Ø 1, p Ø 0.25, w Ø 5<, 8b, 0, 0.15<D; -W + p w W b Ø ÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅÅ -1 + w 0.0625 W 0.006 0.004 0.002 0.02 0.04 0.06 0.08 -0.002 -0.004 0.25 H5 - 1L + 0.75 H-1L 0.25 0.1 0.12 0.14