06.06.2013 Views

STOCHASTIC

STOCHASTIC

STOCHASTIC

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

SEPARATION IN PORTFOLIO ANALYSIS<br />

II.2 RESTRICTION OF THE UTILITY FUNCTIONS<br />

Cass and Stiglitz [2] pointed out that separation can hold for general<br />

markets only if it holds for a convenient set of special markets. Once the utility<br />

functions have been suitably restricted by the special markets it becomes easier<br />

to examine the general market problem. It is convenient to examine first the<br />

separation property for discrete random returns. Thus, suppose that there<br />

are only finitely many states of nature, 9 = l,...,N, with Pr[0 = z] = pt. The<br />

returns pt{6) are specified completely by an nxN nonnegative matrix<br />

Py, i= \,...,n,j= l,...,N. Assuming an interior solution (or I+ =0), the<br />

following optimality conditions are obtained:<br />

E Pj u '{ Z ykPkAPij = A, i = l,...,w. (16)<br />

; = 1 \* = 1 /<br />

Consider first the case N — n and with nonsingular matrix p. Then (16) can<br />

be solved to give<br />

where<br />

«'(2>P«) = £lJfy, (17)<br />

\ * / Pi J<br />

R = p-K (18)<br />

Since u' is continuously differentiable and strictly decreasing, the inverse<br />

function g(-) = u'~ i (-) is a well-defined, strictly decreasing, continuously<br />

differentiable function on its domain. Furthermore, since u' is piecewise twice<br />

continuously differentiable, so is g. Solving (18) gives<br />

tyjPn=9[~t^\ = 9^d- d9)<br />

From the separation equation (8), it follows that<br />

where<br />

g (Afc;) = a (X) m, + b (A) nt<br />

(20)<br />

wf = £ Xj l pji, A* = £ xj 2 pji. (21)<br />

j j<br />

Clearly, the consideration which yielded Eq. (15) will apply to the vector<br />

g(J.ki), i = 1,...,«, as well as to the vector y(X), so the following necessary<br />

condition obtains:<br />

g(kik) g{kjl) g(k,Z)<br />

ktg'iktX) kjg'ikjX) k,g'(k,X)<br />

ktV(ktX) kj 2 g"(kjX) kfg'X^X)<br />

= 0 (22)<br />

3. SEPARATION THEOREMS 161

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

Saved successfully!

Ooh no, something went wrong!