[Math] Markov Chain: Pensioner Problem

markov chainsprobabilityprobability theorystochastic-processes

A pensioner receives 2000 dollars at the beginning of each month. The amount of money he needs to spend during a month is independent of the amount he has and is equal to i (i.e. i thousand dollars) with probability Pi, i = 1, 2, 3, 4, $\displaystyle\sum\limits_{i=1}^4P_{i}=1$. If the pensioner has more than 3000 dollars at the end of a month, he gives the amount greater than 3000 to his son.

Q1.If, after receiving his payment at the beginning of a month, the pensioner has a capital of 5000, what is the probability that his capital is ever 1000 or less at any time within the following four months?

Best Answer

Call $X_n$ the number of thousands of dollars the pensioner has at the end of the $n$th month just before he receives his payment. Modify the dynamics by considering that if the pensioner has or has had $0$ dollar or $1000$ dollars at time $n$ or anytime before time $n$, then $X_n=1$.

The process $(X_n)$ is a Markov chain on the state space $S=\{1,2,3\}$ starting from $X_0=3$ and one asks for the probability $p$ that $X_4=1$. This is $Q^4(3,1)$, where $Q$ is the transition matrix of $(X_n)$ indexed by $S\times S$. Thus $$ Q=\begin{pmatrix}1&0&0\\ 1/2&1/4&1/4\\ 1/4&1/4&1/2\end{pmatrix}. $$ From here, the quickest route might be to (somewhat tediously) compute the first column and the third line of $Q^2$, namely, $$ Q^2=\begin{pmatrix}1&*&*\\ 11/16&*&*\\ 1/2&3/16&5/16\end{pmatrix}, $$ and to deduce the coefficient $$ Q^4(3,1)=\sum_{i=1}^3Q^2(3,i)\times Q^2(i,1)=\frac12\times1+\frac3{16}\times\frac{11}{16}+\frac5{16}\times\frac12=\ldots $$

Related Question