1
Search and Stop Problem
Consider a game: At each period of the game, you draw a random offer x from a uniform distribution from 0 to 1. You can then choose either: A) Accept the payoff x immediately and end the game. B) Reject the payoff and draw again a period later (Repeat game). All draws are independent. Rejections are costly because the agent discounts the future at an exponential discount factor ρ = 0.01. At what value of x should you accept the offer? Solution 1. Write problem in Bellman form. v(xt ) | {z }
= max {xt ,
V alue f rom strategy choice of x
δEv(xt+1 ) | {z }
}
discounted f uture expected value
2. Realise that goal is to identify a threshold value x∗ . Thus, your optimal strategy v(x) would be to: ACCEP T and get x if x ≥ x∗ v(x) = REJECT and get x∗ if x < x∗ 3. Solve: If you randomly draw the offer x∗ , you should be indifferent between stopping and continuing. v(x∗t ) = x∗
= δEv(xt+1 ) Z x=x∗ Z ∗ = δ x f (x)dx + δ
x=1
xf (x)dx
x=x∗
x=0
1
1 2 x 2 x∗ 1 1 = δx∗2 + δ − δ (x∗ )2 2 2 δ = (1 + x∗2 ) 2 x∗
= δ [x∗ x]0 + δ
Solve: x∗ =
√ δ 1 − 1 − δ2 (1 + x∗2 ) ⇒ x∗ = 2 δ
4. Since your discount factor ρ (i.e. cost of retrying) is 0.01 every period, Discount rate δ = exp(−ρ) = exp(−0.01) = 0.99 ⇒ x∗ = 0.87 Optimal strategy: Keep playing until you draw at least 0.87. Then you would accept and end the game. 1