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 function t/,(C,_,, wt|a,, ft) in (9)<br />

provides the maximum expected utility of<br />

lifetime consumption and bequests if the<br />

consumer is alive in state 0t at period t, his<br />

wealth is w,, his past consumption was<br />

Ct_i, and optimal consumption-investment<br />

decisions are made at the beginning of<br />

period t and all future periods at which he<br />

is alive.<br />

Since (7)-(9) are just a special case of<br />

the model summarized by (4), Proposition<br />

1 applies directly to the probabilistic<br />

death model. In this case the proposition<br />

implies that for all t and &, Ut(Ct-i,<br />

v)t Zt, &), £/t(C«_i, wt\at, $»), and l/t(Ct-i,<br />

z"t dt, (3t) are monotone increasing and<br />

strictly concave in (Ct_i, wt).<br />

Finally, expressions (7)-(9) suggest an<br />

alternative to the "sure death" interpretation<br />

of the horizon T+1. For given<br />

wealth uii, and state of the world |3i, the<br />

consumer's problem at period' 1 is to<br />

choose ci and Hj, which<br />

max I U,(Ci, BR' | s,, fodF^lfi,)<br />

- max I [xtUACi, HR' | a,, &)<br />

FAMA: MULTIPERIOD DECISIONS 171<br />

+ (1 - x,)Ut(Cu HR' | dt, fofldFutft).<br />

Using (8) and (9) to expand this expression,<br />

it can be shown that the decision<br />

problem of period t has weight<br />

*t*t_i •••*, = x(at | a,)<br />

in the expected utility for the decision of<br />

period 1. The probability *(at|oi) of being<br />

alive at t will decrease with t. Thus in<br />

general for some t=T+l, the decisions of<br />

periods T+1 and beyond will have negligible<br />

weight in the expected utility for the<br />

decision at period 1, so that in the decision<br />

at period 1 it is unnecessary to look beyond<br />

T+1.<br />

More simply, since the consumer is<br />

likely to be dead, the effects of distant<br />

future decisions, which are unlikely to be<br />

made, can be ignored. And this result does<br />

not arise from discounting of future consumption,<br />

though the effect is the same. At<br />

period 1, consumption in period T+1 may<br />

be regarded as equivalent to consumption<br />

in period 1. But in the decision of period 1<br />

the decision of T+1 is weighted by the<br />

probability that the consumer will be alive<br />

at that time, which reduces the importance<br />

of the future consumption in the current<br />

decision.<br />

V. Conclusion<br />

In sum, assuming only that markets for<br />

consumption goods and portfolio assets<br />

are perfect and that the consumer is risk<br />

averse in the sense that his utility function<br />

for lifetime consumption is strictly concave,<br />

it has been shown that though he<br />

faces a multiperiod problem, in his consumption-investment<br />

decision for any<br />

period the consumer's behavior is indistinguishable<br />

from that of a risk averter who<br />

has a one-period horizon. It was then<br />

shown how this result can be used to provide<br />

a multiperiod setting for the more<br />

detailed hypotheses about risk averse<br />

consumer behavior that are traditionally<br />

derived in a one-period framework.<br />

APPENDIX<br />

Proposition J. If t/t+^Ct, a>l+i|/3t+i) is<br />

monotone increasing, and strictly concave<br />

(henceforth m.i.s.c.) in (Ct, i«t+i), then<br />

Ut(Ct-u wt\pt) is m.i.s.c. in (Ct-i, ait).<br />

Proof: The proof of the proposition relies<br />

primarily on straightforward applications of<br />

well-known properties of concave functions<br />

(cf. Iglehart). We first establish:<br />

Lemma 1. If U^.i(Ct, u>t+i|0tfi) is m.i.s.c.<br />

in (Ct, tt>t+i), the expected utility function<br />

j tf„.,(C„ w,+l | ft+O'WutCft+i)<br />

•'d+i<br />

(10) - f £/«+i(Ct, HnR{fi,+1)' I /3l+1)<br />

•''t+l<br />

is strictly concave in (Ct, Hn).<br />

• (WfltOSt+i)<br />

2. RISK AVERSION OVER TIME IMPLIES STATIC RISK AVERSION 397

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

Saved successfully!

Ooh no, something went wrong!