Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am not sure to solve the following investment problem:

I have an investor which receives an income $I_n\ge 0$ at the start of year $n$. The investor chooses a proportion $p_n\in[0,1]$ of this in the stock market and consumes the remainder. At the start of the next year his income is of course $I_{n+1}=I_n+p_nI_nX_n$, where $X_n>0$ is the retrn on the stock market in year $n$ and is IID with mean $\mu$.

The investor now chooses $p_n$ to maximise $\mathbb E[\sum_{n=0}^{N-1}(1-p_n)I_n+I_N]$

This equation obviously represents the consumption over $N$ years, which the investor wants to maximise.

My solution: Let $F_s(i)$ be the maximal reward obtainable in state $i$ when there is time $s=N-t$ to go, then the dynamic equation is $F_s(i)=\max_p[(1-p)i+F_{s-1}(i+piX)]$ with $F_0(i)=0$ because at time $N$ there is no consumption.

Now $F_1(i)=i, F_2(i)=\max[(1-p)i+i+piX]=\max[2i,i+piX]=i\max[2,1+pX]=i\rho_2$

There I get by induction $F_s(i)=i\rho_s$

How should the investor now choose his strategy? I am almost done but I do not see thr last crucial step.

share|cite|improve this question
up vote 1 down vote accepted

For the last two years you want to maximise the expectation of $$I_N+I_{N-1}(1-p_{N-1}) = I_{N-1}(2+p_{N-1}(X_{N-1}-1))$$ and the expectation is maximised: if $\mu=E[X_{N-1}] \gt 1$ when $p_{N-1}=1$; and if $\mu=E[X_{N-1}] \lt 1$ when $p_{N-1}=0$.

If you invest everything in the penultimate year then it seems obvious you will also invest everything in all previous years.

So let's look at the $\mu=E[X_{N-1}] \lt 1$ so $p_{N-1}=0$ and $I_N+I_{N-1}(1-p_{N-1}) = 2I_{N-1}$: then the last three years are about maximising the expectation of $$I_N+I_{N-1}(1-p_{N-1}) + I_{N-2}(1-p_{N-2}) = I_{N-2}[3+ p_{N-2}(2X_{N-2} -1)]$$ and whether optimal $p_{N-2}$ is $1$ or $0$ depends on whether $\mu \gt \frac12$ or $\mu \lt \frac12$.

Taking this back in a similar manner, the overall optimal strategy becomes clear:

  • let $p_n=1$ (i.e. invest everything) when $\mu \gt \frac{1}{N-n}$, i.e. when $n \lt N - \frac{1}{\mu}$

  • let $p_n=0$ (i.e. invest nothing) when $\mu \lt \frac{1}{N-n}$, i.e. when $n \gt N - \frac{1}{\mu}$.

  • You can choose any $p_n\in [0,1]$ (i.e. invest anything) when $\mu = \frac{1}{N-n}$, i.e. when $n = N - \frac{1}{\mu}$

I would not recommend this as a personal investment strategy. Risk should have an impact too.

share|cite|improve this answer
Thanks for your answer. Do you also have an idea how to use my approach using the dynamic programming equation? When you are saying "Taking this back in a similar manner" it is a more intuitive approach, I think using the Bellman Equation would be satosfactory in this case. – Alkibiades Nov 22 '13 at 1:51
When I say "Taking this back in a similar manner", I am saying that more generally given $I_{N-k+1} = I_{N-k}+p_{N-k}I_{N-k}X_{N-k}$ and if the sum of the last $k$ terms is $kI_{N-k+1}$ then the sum of the last $k+1$ terms is $k(I_{N-k}+ p_{N-k}I_{N-k}X_{N-k})+(1-p_{N-k})I_{N-k}$ which simplifies to $I_{N-k}[k+1 + p_{N-k}(kX_{N-k}-1)]$ which will be maximised by having $p_{N-k}=1 \text{ or }0$ depending on $E[kX_{N-k}-1]$, and if $E[kX_{N-k}-1] \lt 0$ then $p_{N-k}=0$ and the sum of the last $k+1$ terms is $(k+1)I_{N-k}$, thereby completing the induction. – Henry Nov 22 '13 at 7:37

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.