Density Estimation and Combination under Model Ambiguity



Lets consider an individual making portfolio choice at time T , this choice involves two kind of assets: a
risky asset which consists of a broad portfolio of stocks (S&P500 index), whose gross return at time t per
f

unit invested at time t - 1 is 1+Rt , and a riskless asset whose gross return is 1+Rt . The decision maker
has access to the return histories over T periods, he knows in advance the future return of the riskless asset
that in accordance with standard practice is assumed to be constant.

Lets define rt = log(1 + Rt) and rtf = log(1 + Rtf),then we can describe the investor’s information set in
the following way:

It ≡ {rt}tT=0 , {rtf}tT=0 and {rf}tH=T +1 .

He invests one unit of saving, divided between an amount 1 - a in the safe asset and a in the risky asset,
and then he holds on to the portfolio until date H.

Let W (rt ,a) denote the value of the portfolio and suppose also that we are considering a self financing
portfolio. Thus, the value of the portfolio at time t
= H is given by:

HH    H

W(rt,a)=(1-a) Y (1+Rtf)+a Y (1+Rt)=(1-a)Hrtf +a X rt.

t=T +1             t=T +1                           t=T +1

Lets also assume that utility depends only on the final value of the portfolio: U(W (rt, a)).Then, the
problem is to choose the best decision rule d that maps the observations contained in I
t into actions a, in
other words: the optimal share to invest in the risky asset. This decision rule is obtained by maximizing the
following expected utility:

Egr|s[U(W(rt,a))]

H

s.t Wt+H =(1-a)Hrtf+a      rt and a [0, 1].

t=T +1

In order to simplify the analysis lets assume that H=1, such that the wealth form reduces to

W (ret ,a)=(1- a)rtf + art = rtf + a(rt - rtf )=C + aret ,

where C is a constant and ret is the excess return.

Example 1: (CARA investor) Assume that U(W (ret,a)) is the utility function of an investor with negative
exponential utility:

U(W (ret,a)) = - exp(-δW (ret, a))

Egrs [U(W(et,a))] = -KEgrs exp(-δaet^,

where δ is the risk aversion parameter, K is the expected utility relative to the riskless asset and gr|s is the
distribution of the return
rt given the regime s, which is unknown. Typically, the return distribution gr|s is
assumed to be Normal, consequently the expected utility results to be:

Eg l [U(W(et, a))] = - exp(-aδμ + 1 δ2a2σ2),
r|s                                                         2

where μ = E(rt) and σ2 = Var(et) are the mean and the variance of the Normal distribution. In this case
the optimal share to invest in the risky asset is given by:

a(gr|s) = arg maax Egr|s [U(W(ret,a))]

18



More intriguing information

1. The name is absent
2. Pass-through of external shocks along the pricing chain: A panel estimation approach for the euro area
3. The name is absent
4. The name is absent
5. The name is absent
6. Explaining Growth in Dutch Agriculture: Prices, Public R&D, and Technological Change
7. The name is absent
8. A parametric approach to the estimation of cointegration vectors in panel data
9. The East Asian banking sector—overweight?
10. The effect of globalisation on industrial districts in Italy: evidence from the footwear sector