Why a strong guarantee shifts behaviour on both sides

Free returns arrive and the order rate climbs. So does the return rate, and so does the share of orders containing three sizes of the same item. Meanwhile the warehouse relaxes its final inspection, because a mistake now costs a return label rather than a lost customer.

What everyone sees

A guarantee is treated as a claim about quality — confidence made contractual. It is also a reallocation of risk, and reallocated risk does not sit still. Whoever no longer bears the cost of an outcome has less reason to spend on avoiding it, and that applies to the seller who wrote the guarantee as much as to the buyer who accepted it.

What is actually happening

Holmström’s analysis of moral hazard under imperfect observability establishes the general result: when an agent’s care is not directly observable and the consequences of low care are borne elsewhere, the equilibrium level of care falls, and no contract on outcomes alone fully restores it. The buyer side is the widely cited case; the seller side is the one usually missed, since the guarantee also insulates the seller from the reputational cost of a defect by converting it into a reversible transaction. Peltzman’s offsetting-behaviour argument, that protection induces riskier conduct, is the best-known version of this claim and remains disputed — later analyses found the compensating response real but smaller than his estimates. What holds is the direction, not the size.

Why it stays hidden

The two-sided effect hides because the two responses are measured by different departments. Return rates sit with operations and are read as a customer-behaviour problem. Inspection intensity sits with quality and is read as an efficiency gain. Nothing in either report says that both moved because the same clause changed the price of care.