Question
Two orchards send apples to a small market every week. Each privately chooses to deliver either one crate or two; the market price is £\(11-2(q_A+q_B)\) per crate when the total deliveries are \(q_A+q_B\), and neither orchard has any other cost. Choices become public after that week's sales.
They propose to cooperate by each delivering one crate. If anyone instead delivers two during a cooperative week, both must deliver two crates for the next four weeks, after which they return to one crate each. The four-week punishment runs its full course even if someone changes their delivery during it; no new punishment is added. Future weekly profits are discounted by a common factor \(\delta\), with \(0\le\delta<1\).
For what values of \(\delta\) can this strategy be a subgame-perfect equilibrium? Show that an orchard cannot gain by cheating during cooperation or during the punishment weeks. Explain why four punishment weeks, rather than a permanent price war, matter.
Get hint
Hint 1
Compute one week's profit for each orchard at delivery pairs (1,1), (2,1), (1,2), and (2,2). Compare immediate temptation with the discounted four-week loss.
Check your answer
Solution
At \((q_A,q_B)=(1,1)\), the total is two crates, the price is £7 per crate, and each orchard earns £7. At \((2,1)\), price is £5: the deviator earns \(2\times\text{£}5=\text{£}10\) and the cooperator £5. At \((2,2)\), price is £3 and each earns £6. Thus cooperation pays \(R=7\), a one-week deviation against cooperation pays \(T=10\), and mutual high output pays \(P=6\). If the other orchard produces two, choosing one yields £5, less than the £6 from producing two.
Suppose a cooperative history has been reached. Cooperating gives the stream of £7 weekly. Deviating once raises today's profit by \(T-R=\text{£}3\) but creates four punishment weeks paying £6 instead of £7, starting next week. After those weeks the streams coincide again. Therefore the no-deviation condition is
\[3\le(7-6)(\delta+\delta^{2}+\delta^{3}+\delta^{4})=\delta+\delta^{2}+\delta^{3}+\delta^{4}\].
Equivalently, the gain difference between cooperation and deviation is \(-3+\delta+\delta^{2}+\delta^{3}+\delta^{4}\). The right side is strictly increasing on \([0,1]\), starts at zero and approaches four. Consequently, there is a unique threshold \(\delta^{*}\in(0,1)\), satisfying \((\delta^{*})^{4}+(\delta^{*})^{3}+(\delta^{*})^{2}+\delta^{*}-3=0\). Numerical evaluation gives \(\delta^{*}\approx0.8881796676\). The incentive holds at equality and for every δ above that threshold, up to but excluding \(\delta=1\).
During any of the four punishment periods, both are prescribed two crates. Because the other orchard is producing two, choosing two earns £6, versus £5 for choosing one. The punishment counter advances regardless of that choice, so neither the subsequent punishment periods nor the return date is affected. Delivering two is therefore strictly optimal in every punishment subgame. At cooperative histories the discounted one-shot deviation test above applies after any past return from punishment. These observations establish sequential incentives for the complete finite-state strategy by the one-deviation principle.
As a check, at \(\delta=0.9\) the four-week loss is \(0.9+0.81+0.729+0.6561=3.0951>\text{£}3\); at \(\delta=0.8\) it is \(2.3616<\text{£}3\). A finite punishment can deter cheating only if players value the four future losses enough. With permanent punishment the relevant discounted loss would be \(\frac{\delta}{1-\delta}\), a different condition.
Assumptions
Each orchard chooses only \(q=1\) or \(q=2\) every week, with the stated price formula and zero other costs.
Actions and deviations are observed after weekly sales; punishment begins the following week and lasts exactly four weeks.
During punishment, the countdown and return to cooperation do not change with further deviations.
Both agents use the same discounted-sum objective with \(0\le\delta<1\).
Final Answer
The strategy is subgame perfect for \(\delta\in[\delta^{*},1)\), where \(\delta^{*}\) is the unique root of \(\delta+\delta^{2}+\delta^{3}+\delta^{4}=3\) in (0,1), approximately 0.8881796676.