Two companies, $A$ and $B$, compete for customers. In each period, a fraction of each company's customers may switch to the other. Company $A$ retains $80\%$ of its customers and loses
0\%$ to $B$. Company $B$ retains $70\%$ of its customers and loses $30\%$ to $A$.
The initial market shares are $s_A^{(0)}$ and $s_B^{(0)}$ with $s_A^{(0)} + s_B^{(0)} = 1$.
Write down the transition matrix $P$ and the market share update rule.
Find the steady-state (long-run) market shares $\pi_A$ and $\pi_B$.
Do the steady-state shares depend on the initial market shares? Why or why not?
Hints
Model this as a Markov chain. What is the transition matrix, and what equation does the steady-state distribution satisfy?
At equilibrium, the flow of customers from $A$ to $B$ must equal the flow from $B$ to $A$. Set $0.2\,\pi_A = 0.3\,\pi_B$.
Use the constraint $\pi_A + \pi_B = 1$ along with the balance equation to solve for both shares. Then argue why the initial distribution does not matter.
Worked Solution
How to Think About It: This is a textbook Markov chain problem. Customers switch between companies according to fixed probabilities each period, so market shares evolve as a matrix multiplication. The steady state is the left eigenvector of the transition matrix with eigenvalue
$ -- it is the distribution that, once reached, does not change. The key practical insight: for an ergodic (irreducible, aperiodic) chain, the steady state is unique and independent of where you start.
Quick Estimate: Company $A$ loses
0\%$ per period, $B$ loses $30\%$. Since $B$ leaks faster, you would expect $A$ to end up with a larger share. Roughly, the steady state should balance the flows: $0.2 \pi_A = 0.3 \pi_B$, giving $\pi_A / \pi_B = 3/2$, so $\pi_A = 0.6$, $\pi_B = 0.4$.
Approach: Set up the balance equations from the transition matrix and solve.
Formal Solution:
Part 1 -- Transition matrix:
The transition matrix (rows are "from", columns are "to") is:
Yes, the steady state is independent of the initial market shares. This is because the chain is ergodic: it is irreducible (any state can reach any other with positive probability) and aperiodic (the diagonal entries are positive, so the chain can stay in a state). For ergodic Markov chains, the stationary distribution is unique, and the chain converges to it from any starting distribution.
Answer: The long-run market shares are $\pi_A = 0.6$ (60%) and $\pi_B = 0.4$ (40%), regardless of the initial market shares.
Intuition
The steady state of a Markov chain is where inflows and outflows balance for every state. For two companies, this balance condition is simple: the rate at which $A$ loses customers ($0.2 \pi_A$) must equal the rate at which it gains them ($0.3 \pi_B$). The company with the lower churn rate ends up with a larger market share -- which makes intuitive business sense. The independence from initial conditions is the remarkable property of ergodic Markov chains: no matter how lopsided the starting position, the long-run equilibrium is fully determined by the transition probabilities. This is why Markov chain models are so popular in finance for modeling credit ratings, customer behavior, and regime-switching -- the long-run distribution tells you where things settle regardless of today's snapshot.