Foray 445 · Essai 02 · A constructed exchange
Must it be right first time?
A correction promise can remove the incentive for first-pass quality—and still make a good final answer cheaper. The purchase needs to say which quality it is buying.
A bounded question about the exchange
The first essai separated a checkable dependency audit from overall project readiness and counted the work around delivery. This trial asks a different question: how does the acceptance arrangement change the supplier's best production choice?
All figures below are invented value units. Careful production costs 60 and always meets the written requirement. Quick production costs 20 and initially fails it. A perfect independent check costs 20 when used. The buyer values a conforming final answer at 100. A 40-unit supplier bond is returned on success. Nobody has bought this offer, and these are not estimated market prices.
If a check catches a bad answer, the supplier either forfeits the price and bond or, where allowed, pays to correct it and keeps the price. The check cannot be gamed in this model. A missed bad answer has zero value. These strong assumptions make the mechanism inspectable; they are not evidence about professional judgement in practice.
Change the acceptance arrangement
| Supplier choice | Expected surplus | Final conformity |
|---|
Highlighted choices maximise the supplier's expected surplus. Ties use the least favourable quality and buyer outcome; we do not assume the supplier picks the convenient option.
This is a temporary calculation. No offer, payment, bond or client record is created. Refresh restores the example.
Where is an acceptable exchange possible?
The map varies price(left to right) and checking probability(bottom to top). Green means careful production; blue means quick production followed by correction if checked. A dark mix is a tie that still meets the selected quality target. Pale cells fail at least one requirement. The marker is your current arrangement.
Every accepted cell requires nonnegative supplier surplus, nonnegative expected buyer value after price and checking, and the required quality under every profit-maximising choice. A zero supplier surplus is a participation boundary, not guaranteed willingness to accept.
Grid resolution: one price unit and one percentage point. The map is not a continuous optimisation proof.
The advance: specify final quality separately from production effort
With the worked assumptions and no correction option, a price of 60 needs a checking probability strictly above 40% to make careful work uniquely preferable. At exactly 40%, careful and defective delivery tie. The continuous buyer-cost infimum is 68, unattained under strict first-pass incentives; the grid finds 68.2.
Now allow correction for 20. No available checking probability makes careful production preferable: the expected correction penalty is at most 20, below the 40 extra cost of careful work. But checking every answer and correcting failures delivers 100% final conformity. The participation boundary is a 40-unit price plus 20 checking cost, or 60 total; price 41 leaves the supplier one unit and costs the buyer 61.
“The correction clause undermines incentives” is therefore an incomplete conclusion. It undermines one production method. If the buyer needs a conforming final answer and full checking is dependable and cheap enough, planned correction can be better. If the buyer needs first-pass quality, timely acceptance, or a quality dimension the check cannot verify, this result does not carry over.
Raise the checking or repair cost to see that exchange disappear. The useful offer boundary concerns the verifiable outcome, the checking technology and who pays for repair. A nominally small deliverable alone does not define it.
What can the buyer afford to inspect?
At repair cost 20 and a 100% final-quality target, compare two purchasing arrangements at the live inspection cost. The planned repair route costs 40 + k at the participation boundary. For first-pass care, a price of 60 needs inspection probability strictly above 0.4. Its cost approaches 60 + 0.4k. Across the displayed price and inspection ranges a higher price cannot lower that bound. They cross when the inspection costs 33⅓.
The buyer must be able to commit to the stated inspection rate. If an assurance of careful work leads the buyer to skip a costly check, the supplier faces different incentives. These figures concern verifiable full conformity with no delay cost.
Source → mechanism → exact limits
Hart and Moore(1988), Incomplete Contracts and Renegotiation, especially pp 755–757, separates what parties observe from what an outsider can verify and examines how later renegotiation affects earlier incentives. That distinction motivated the question. The three-strategy calculation here is our deliberately smaller model; it is not their theorem or a fitted economic model.
Careful surplus is P−60. Defect-and-forfeit surplus is (1−q)P−40 q−20. Defect-and-correct surplus is P−20−qr. Thus careful work needs q·min(P+40,r)>40 when both deviations are available. Without correction the condition is q(P+40)>40. The equality boundary needs explicit tie treatment. Under the stated parameters, minimising P+20 q subject to P≥60 gives the 68 infimum without correction.
Risk neutrality, enforceable payments, no checking error, complete correction, no delay or reputational effects, no adverse selection, no discretionary renegotiation and known costs are assumptions. If quality is merely observable but no contractible check exists, set q=0: this simple performance mechanism cannot distinguish careful from cheap defective production. Relational contracting or other signals are outside the construction.
Model · Independent event-tree accounting · Executed checks and frontier. The checker calculates inspected/uninspected event outcomes separately and compares them with the closed-form model over 288 parameter settings. No demand, sale or realised saving is claimed.