06.06.2013 Views

STOCHASTIC

STOCHASTIC

STOCHASTIC

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

the utility function u over wealth w = p'x = Y.Pi*t is concave and continuously differentiable.<br />

If the investor is an expected utility maximizer, his decision problem is<br />

{maxE„u(p'x)|£x, = B, xt S 0}. (1)<br />

Suppose the return vector p has the discrete distributionpr{p = p'} — pt> 0,1 = 1,...,L < oo,<br />

and£/>, = 1. Then (1) is<br />

maxZO) = piu(p l 'x) + ••• + pLu(p L 'x), s.t. ^xt = B, xt S 0. (2)<br />

(a) Show that Z is concave and continuously differentiable and<br />

dZ(x) _ y \8u[p^x}<br />

dx, 4* I 8 w<br />

(b) Attempt to develop an efficient algorithm to solve (2) for large L that exploits the<br />

fact that each of the L terms in the objective is "similar."<br />

(c) Suppose xx is unconstrained. Show that (2) becomes<br />

maxv(x2,...,x„) = {piu[^pl l x,+p1 1 (B-Y,x,)'] + ••• + pLu[£pi L xt + pl L {B-Y1xi)~]}<br />

s.t. Xi £ 0, where i = 2,...,«.<br />

(d) Show that v is concave and continuously differentiable.<br />

(e) What is dvjdx, ? Note its similarity to the expression in (a).<br />

(f) How does the algorithm in (b) simplify when applied to (3)?<br />

(g) Investigate algorithmic simplifications when u is quadratic, u = log, p l > 0 for all /,<br />

and u is exponential.<br />

17. Under certain assumptions, such as quadratic preferences or normally distributed<br />

random returns, the mean-variance approach is consistent with the expected utility approach.<br />

To provide generality to the risk-return approach, it is desirable to be able to consider a<br />

risk-return measure that is free of such restrictions on the utility function or the probability<br />

distributions. Let 0 ( < 0) if u is strictly concave (strictly convex) and that (j> = 0 if u is<br />

linear.<br />

(b) The risk premium n is defined by the equation u{w— n) = Eu(w), so that tj> is related<br />

to n via = u(w) — u(w — n). Suppose u is strictly increasing; show that ^ = 0 if and only<br />

ifrcso.<br />

(c) Suppose u{w) is linearly transformed into v(w) — au(w) + b. Show that 4> is transformed<br />

to a.<br />

(d) Let Mk, assumed finite, be the kth moment of w about w. Suppose a Taylor's expansion<br />

of u about w exists in a neighborhood of w sufficiently large to include all w for which<br />

there is a nonzero probability of occurrence. Show that<br />

l<br />

!<br />

c=2 k { -Mk<br />

-<br />

where u lk, (w) is the &th derivative of u evaluated at w — w.<br />

(e) Suppose w is normally distributed. Show that (/> is proportional to variance, i.e.,<br />

^ = aa\ where a = - ^ ~^T n •<br />

MIND-EXPANDING EXERCISES 351

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!