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.

390 GORDON PTE<br />

Inspection of this last expression shows that zT-i(n) is given by<br />

2?_I(TO) = (a - n)/(a + b) 0 S n £ a<br />

— 0 n > o.<br />

Thus, 2r_i(n) is a decreasing function of n as required. Using this it is also easily verified<br />

that LT(a — n, z) — az = LT(n + b, z) + bz for all n and all 2:0 £ z £ zr_i(«)<br />

as required. This completes the proof.<br />

It might be conjectured that the critical values are a decreasing function of t as<br />

well as n. This, however, is not true. A counterexample is provided by the minimax<br />

policy calculated for the example in the following section. On the basis of examples it<br />

appears that the critical values are much less sensitive to t than n except near T.<br />

This, taken together with Theorem 5, has the following implication.<br />

After some of the asset is sold initially, little more will be sold if the price rises or<br />

stays the same. In this case n will remain equal to zero. On the other hand, more of<br />

the asset will be sold as soon as the price drops because then n will increase. In general,<br />

for a sequential pohcy, more is sold on price dips than rises because the former increase<br />

n while the latter do not. An example for a simple case is provided in the next<br />

section.<br />

4. An Example of a Sequential Policy<br />

As a simple example of an arithmetic random walk consider the following process.<br />

Suppose that each period the price can either go up by one unit, down by one unit or<br />

stay the same. This means that a = 6 = 1 and that n can take on only integer values.<br />

Suppose furthermore that there are three points of time available to sell the asset<br />

before it must all be sold so that T = 3. When t = 2 and n S 1 it is already known<br />

from Theorem 2(a) that the minimax policy is to sell all of the asset which is left. For<br />

t = 2 and n = 0, (13) and (14) give the following:<br />

L2(0, z) = MinoS*s, Max [L3( — 1, x) — x, Lj(l, x) + x]<br />

= MinoS»s, Max [1 — x, x]<br />

= 1 — z z S | x = z z •& \<br />

= J z>\ = i « > i-<br />

Thus, the critical value, Z2*(0), is f. Any asset which is held above this amount should<br />

be sold. If less than this amount is held none should be sold.<br />

Using this result, the minimax strategies for t = 1 and n = 0, 1 can next be derived.<br />

LxtO, z) = Min0 I<br />

ii(l, z) = Min0SlS2 Max [L2(0, x) — x, Li(2, x) + x]<br />

= Minos*s, Max IL _ * * = \ ,x\<br />

= l-2z « i l x* = z zfij<br />

588 PART V. DYNAMIC MODELS

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

Saved successfully!

Ooh no, something went wrong!