Notes Markov Analysis
Notes Markov Analysis
Markov Process
1. An example
1.1 Urban and rural populations are 48 million and 116 million respectively. Historical data
reveals that 5 per cent of urban population migrates to rural area and 20 per cent rural
population migrates to urban area. Determine urban-rural population after one year.
(Assuming zero per cent growth)
1.2 Simple arithmetic:
Urban: 48*0.95 + 116*0.3 = 68 million. Rural: 116*0.80 + 48*.05 = 92
1.3 Above example with percentage
Present urban-rural population is 30/70 and next year it would be:
Urban: 30*0.95 + 70*0.3 = 42.50%. Rural: 70*0.80 + 30*.05 = 57.50%
2. Generalization:
2.1
2.2
6. Transition matrix P =
7. We have two states and sum of each row = 1 as it is mutually exclusive and completely
exhaustive.
8. Probabilities of states is a row vector
n is period and k number of companies.
9. Π(1) = Π(0)P
Π(2) = Π(1)P
Π(3) = Π(2)P
.
.
.
Π(n + 1) = Π(n)P
18.4 Note that the probability of a dollar in the 0 – 30 days category moving to the paid
category in the next period is 40%. Note also that a dollar in a 0 – 30 days category
cannot make the transition to a bad debt state in one week.
19. Absorbing States
We saw (Sec 17.3) that a dollar can change state from 2 to 3 and vice versa. However,
once a dollar makes a transaction to state – 1, the paid category, the probability of making
a transition to other state is zero. Similarly, once a dollar is in state – 2 (bad debt),
probability of transition to any other state is zero. Thus, once a dollar reaches state 1 or 2,
the system will remain in that state indefinitely. We conclude that all AR eventually be
absorbed into state – 1 or state – 2, hence the name absorbing state.
20. Whenever a Markov process has absorbing states, we do not compute steady-state
probabilities because each unit ultimately ends up in one of the absorbing states.
21. Now we want to know what percentage of 0 – 30 state will go to paid state and what goes
to bad debt and similarly for 31 – 90 state.
22. The Fundamental Matrix
22.1 Computation of absorbing state probabilities requires the determination of a fundamental
matrix. We begin the partitioning the matrix of transition probabilities into the following
four parts:
22.4
22.5 Interpretation: 89% of 0 – 30 days category will eventually be paid and 11% will become
bad debt. 74% of 31 – 90 days category will eventually being paid and 26% will remain
unpaid.
23. Establishing the Allowance for Doubtful Accounts – An Example
Suppose that 31 December balance sheet shows that AR in 0 – 30 days category is $1000
and 31 – 90 days category is $2000. In matrix form B = [1000 2000]. Multiply B with
NR i.e. BNR = [2370 630]. Thus out of $3,000, 2370 will be collected and 630
uncollected. Therefore, an allowance for bad debt should be $630.