STRATHMORE UNIVERSITY SCHOOL OF FINANCE AND APPLIED ECONOMICS Bachelor of Business Science – Actuarial Science, Finance & Financial Economics BSM 2215: INTRODUCTION TO STOCHASTIC MODELLING Hitting and Exit-Times of a Stochastic Process Syllabus objectives: The student should be able to: 1. Define a Stopping Time 2. Explain and apply Doob’s Optional Stopping theorem 3. Use Doob’s Optional Stopping theorem to calculate the probability of Gambler’s Ruin. 4. Use Doob’s Optional Stopping theorem to calculate the expected duration of Gambler’s Ruin. 5. Use Doob’s Optional Stopping theorem to calculate the probability of Exit from a standard Brownian from an interval. 6. Use Doob’s Optional Stopping theorem to calculate the expected duration of Exit from a standard Brownian from an interval. Stopping Time When you are playing a gambling game, you have to decide when to stop, when you are speculating on the price of a share you have to decide when to sell or buy, or when you are holding an American option you must decide when to exercise it. Normally, this decision is based on the occurrence of an event happening a random time ๐ before or on time ๐ก therefore (๐ ≤ ๐ก) hence (๐ ≤ ๐ก) ∈ ๐น( . Meleah Oleche Page 1 Doob’s Optional stopping theorem Let T be a stopping time and M a Martingale, then: ๐ธ ๐+ = ๐ธ ๐As long as any of the following conditions is satisfied: 1. T is bounded 2. |๐(/0 − ๐( | ≤ ๐พ 3. ๐ธ[๐] < ∞ Applications of the Doob’s Optional Stopping Theorem Gambler’s Ruin problem Let us consider two players A and B. They toss a fair coin. If the coin is a head, A wins 1 dollar from B, otherwise A losses 1 dollar to B. Given that A and B start with ๐ and ๐ initially. We are interested in the following: 1. The wealth process of player A 2. The probability that A wins all the money from B 3. Probability that A loses all the money to B 4. The duration of the game The Wealth process of player A The wealth process of player A, the wealth of player A after the ๐ − ๐กโ ๐๐๐๐ can be modelled as follows: ๐A = ๐ + Where: ๐D = Meleah Oleche A DE0 ๐D 1 ๐ด ๐ค๐๐๐ 1 −1 ๐ด ๐๐๐ ๐๐ 1 Page 2 The wealth process of player A is a Martingale Proof: ๐A/0 = ๐ + A/0 DE0 ๐D = ๐A + ๐A/0 ๐ธ ๐A/0 ๐นA = ๐ธ ๐A + ๐A/0 ๐นA Applying measurability = ๐A + ๐ธ ๐A/0 ๐นA Applying independence = ๐A + ๐ธ[๐A/0 ] ๐ธ ๐A/0 = 1×0.5 + −1×0.5 = 0 = ๐A + 0 = ๐A Hence the Wealth process of player A is a Martingale Meleah Oleche Page 3 Another Martingale associated with Player A’s wealth process is the following: ๐A = (๐A − ๐)Q − ๐ Proof: ๐A/0 = (๐A/0 − ๐)Q − (๐ + 1) = (๐A + ๐A/0 − ๐)Q − ๐ − 1 = ๐A + ๐A/0 Q − 2๐ ๐A + ๐A/0 + ๐Q − ๐ − 1 = ๐A Q + 2๐A ๐A/0 + ๐A/0 Q − 2๐ ๐A + ๐A/0 + ๐Q − ๐ − 1 ๐ธ[๐A Q + 2๐A ๐A/0 + ๐A/0 Q − 2๐ ๐A + ๐A/0 + ๐Q − ๐ − 1|๐นA ] Applying measurability, linearity and constant property of conditional expectation ๐ธ[๐A Q + 2๐A ๐A/0 + ๐A/0 Q − 2๐ ๐A + ๐A/0 + ๐Q − ๐ − 1|๐นA ] ๐A Q + 2๐A ๐ธ ๐A/0 + ๐ธ ๐A/0 Q ๐นA − 2๐๐A − 2๐๐ธ ๐A/0 ๐นA + ๐Q − ๐ − 1 Applying independence ๐A Q + 2๐A ๐ธ[๐A/0 ] + ๐ธ[๐A/0 Q ] − 2๐๐A − 2๐๐ธ[๐A/0 ] + ๐Q − ๐ − 1 ๐ธ ๐A/0 = 1×0.5 + −1×0.5 = 0 ๐ธ [๐A/0 Q = 1Q ×0.5 + (−1)Q ×0.5 = 1 ๐A Q + 1 − 2๐๐A + ๐Q − ๐ − 1 ๐A Q − 2๐๐A + ๐Q − ๐ = ๐A − ๐ Q − ๐ = ๐A Hence it is a Martingale Gambler Ruin Probability: Since ๐ธ[|๐A/0 − ๐A |] < 1 we can apply the Doob’s optional stopping theorem Define ๐ = inf {๐ก > 0|๐+ = ๐ + ๐ ๐๐ ๐+ = 0} ๐ธ ๐+ = ๐ธ ๐- = ๐ ๐ธ ๐+ = ๐ธ(๐+ |๐ด ๐ค๐๐๐ ๐๐๐ ๐๐๐๐ ๐ต)๐(๐+ = ๐ + ๐) + ๐ธ(๐+ |๐ด ๐๐ ๐๐ข๐๐๐๐)๐(๐+ = 0) ๐ = ๐ + ๐ ×๐ ๐+ = ๐ + ๐ + (0)×๐(๐+ = 0) ๐ + ๐ ×๐ ๐+ = ๐ + ๐ = ๐ Probability that A wins ๐ ๐+ = ๐ + ๐ = ๐ ๐+๐ The probability of Ruin of player A Meleah Oleche Page 4 But ๐ ๐+ = ๐ + ๐ + ๐ ๐+ = −๐ = 1 ๐ ๐+ = 0 = 1 − ๐ ๐ = ๐+๐ ๐+๐ Duration of a fair game Define ๐ = inf {๐ก > 0|๐+ = ๐ + ๐ ๐๐ ๐+ = 0} ๐+ = (๐+ − ๐)Q − ๐ ๐ธ ๐ = ๐ธ[ ๐+ − ๐ Q − ๐+ ] ๐ธ ๐ ≤ ๐Q We can therefore apply the Doob’s optional stopping theorem ๐ธ ๐+ = ๐ธ ๐- = 0 ๐ธ ๐ = ๐ธ ๐+ − ๐ ๐ธ ๐ = ๐Q × Meleah Oleche Q = ๐Q ๐ ๐ด ๐๐ข๐๐๐๐ + ๐ Q ๐ ๐ด ๐ค๐๐๐ ๐๐๐ ๐ ๐ (๐ + ๐) + ๐Q × = ๐๐ = ๐๐ ๐+๐ ๐+๐ (๐ + ๐) Page 5 Hitting and Exit Time of a Standard Brownian Motion from an interval Let ๐( be a standard Brownian Motion with ๐- = 0, and ๐ > 0 ๐๐๐ ๐ > 0. We want to calculate two things: 1. The exit probability of the Standard Brownian Motion from an interval Ø The ๐(๐ต+ ≤ −๐) Ø The ๐(๐ต+ ≥ ๐) 2. The expected duration of the exit time from an interval The Exit probability of a standard Brownian Motion from an interval Define ๐ = inf {๐ก > 0 ๐ต( ∉ (−๐, ๐) ๐ต( is a Martingale and is bounded and is bounded, so applying Optional Stopping theorem ๐ธ ๐ต- = 0 = ๐ธ ๐ต+ = −๐×๐ ๐ต+ = −๐ + ๐×๐ ๐ต+ = ๐ But ๐ ๐ต+ = −๐ + ๐ ๐ต+ = ๐ = 1 −๐[1 − ๐ ๐ต+ = ๐ ] + ๐×๐ ๐ต+ = ๐ = 0 −๐ + ๐๐ ๐ต+ = ๐ + ๐×๐ ๐ต+ = ๐ ๐๐ ๐ต+ = ๐ + ๐×๐ ๐ต+ = ๐ = ๐ ๐ ๐ต+ = ๐ ๐ + ๐ = ๐ ๐ ๐ ๐ต+ = ๐ = ๐+๐ ๐ ๐+๐−๐ ๐ ๐ต+ = ๐ = 1 − = = ๐+๐ ๐+๐ ๐+๐ Expected duration of standard Brownian Motion from an interval Consider a Martingale ๐( = ๐ต( Q − ๐ก ๐+ = ๐ต+ Q − ๐ ๐ธ ๐ = ๐ธ ๐ต+ Q − ๐ธ[๐+ ] Since ๐ต+ Q is bounded ๐ธ ๐ < ∞ has upper bound and hence the Optional stopping Theorem applies ๐ธ ๐- = ๐ธ ๐+ = 0 ๐Q × Meleah Oleche ๐ ๐ + ๐Q × =0 ๐+๐ ๐+๐ ๐๐(๐ + ๐) = ๐๐ ๐+๐ Page 6