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.

78 THE REVIEW OF ECONOMICS AND STATISTICS<br />

the investor's portfolio must simultaneously lie<br />

within the opportunity set and satisfy the probability<br />

constraint (6), it must lie above the<br />

line AB and below the line RT. The feasible<br />

set of portfolios is therefore given by the set<br />

RST and is illustrated in figure 3.<br />

FIGURE 3. — "MAXIMUM /I" SAFETY-FIRST<br />

CRITERION : No RISKLESS ASSET<br />

1 ' »<br />

i M<br />

It is easy to see that the portfolio which maximizes<br />

expected returns among those included<br />

in this set is the one represented by T, which is<br />

therefore the one selected by the Telser safetyfirster.<br />

It also can be seen that this portfolio<br />

is less conservative (in the sense that it leads<br />

the investor to select a portfolio with a larger<br />

mean and larger standard deviation) than that<br />

selected by a "minimum a" investor who selects<br />

the same disaster level z, or by a "maximum<br />

2" investor who selects the same probability<br />

level a. They will select the points W,S respectively.<br />

On the other hand, the "maximum /u"<br />

approach does create problems of feasibility<br />

that the other two variants do not have. By<br />

selecting both a and 2, the investor fixes the<br />

position of the line RT and it is quite possible<br />

that the values of these parameters he selects<br />

are such that this line fails to intersect the<br />

opportunity locus. In this case the "maximum<br />

fi." principle breaks down for there is no portfolio<br />

which will permit the investor to meet his<br />

disaster level with the required probability.<br />

indifference curves in the mean-standard deviation plane for<br />

the "maximum p." approach as could be done for the other<br />

approaches. The reason is that in this approach the optimality<br />

criterion, which is to maximize the expected return,<br />

does not involve a. Consequently, one cannot get a very<br />

sensible "balancing off" between fi and cr such as portrayed<br />

by normal mean-standard deviation indifference curves.<br />

l/s<br />

We are now in a position to study the relationship<br />

between portfolio analysis based on<br />

expected utility maximization and that based<br />

on the safety-first principle. Referring to figure<br />

1, imagine that an investor who maximizes expected<br />

utility chooses the portfolio P. Since the<br />

set of attainable portfolios bounded by APB is<br />

a convex set and since the set of points preferable<br />

to those lying along Ij also forms a convex<br />

set, by the separating hyperplane theorem<br />

there exists a straight line CD passing through<br />

P which is tangential to both APB and I,. If<br />

we use the "minimum a" safety-first criterion,<br />

we can equate the slope of this line to the slope<br />

given by (9),

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

Saved successfully!

Ooh no, something went wrong!