# Chapman-Kolmogorov Equation

## Theorem

Let $X$ be a discrete state-space Markov chain with $n$-step transition probability matrix:

$\mathbf P^{\paren n} = \sqbrk {\map {p^{\paren n} } {j, k} }_{j, k \mathop \in S}$

where:

$\map {p^{\paren n} } {j, k} = \condprob {X_{m + n} = k} {X_m = j} = {p_{j k} }^{\paren n}$ is the $n$-step transition probability.

Then:

$\mathbf P^{\paren {n + m} } = \mathbf P^{\paren n} \mathbf P^{\paren m}$

or equivalently:

$\ds {p_{i j} }^{\paren {n + m} } = \sum_{k \mathop \in S} {p_{i k} }^{\paren n} {p_{k j} }^{\paren m}$

## Proof

We consider the conditional probability on the left hand side:

 $\ds$  $\ds \map \Pr {X_{m + n} = j \mid X_0 = i}$ $\ds$ $=$ $\ds \condprob {\paren {\bigcup_{k \mathop \in S} \sqbrk {X_{n + m} = j, X_n = k} } } {X_0 = i}$ $\ds$ $=$ $\ds \sum_{k \mathop \in S} \condprob {X_{n + m} = j, X_n = k} {X_0 = i}$ Definition of Countable Additivity $\ds$ $=$ $\ds \sum_{k \mathop \in S} \condprob {X_{n + m} = j} {X_n = k, X_0 = i} \times \condprob {X_n = k} {X_0 = i}$ Chain Rule for Probability $\ds$ $=$ $\ds \sum_{k \mathop \in S} \condprob {X_{n + m} = j} {X_n = k} \times \condprob {X_n = k} {X_0 = i}$ Markov Property $\ds$ $=$ $\ds \sum_{k \mathop \in S} {p_{i k} }^{\paren n} {p_{k j} }^{\paren m}$ Definition of Stationary Transition Probabilities

$\blacksquare$

## Source of Name

This entry was named for Sydney Chapman and Andrey Nikolaevich Kolmogorov.