What Is Moral Hazard? Behaving Differently When Someone Else Bears the Cost

Moral hazard is a change in behaviour that occurs because someone else bears the consequences, arising after an arrangement is entered into rather than before. A person insured against theft may take less care, a driver with comprehensive cover may drive less cautiously, a firm knowing it will be rescued may take greater risks. The word moral is historical and misleading, since the effect does not require any intention to exploit and follows from ordinary incentives, so a careful person responding to reduced consequences is displaying the phenomenon without doing anything wrong. It is distinguished from adverse selection, which operates before the arrangement and concerns who chooses to enter it, with people at higher risk being more likely to buy insurance. Both arise from asymmetric information and they are frequently confused.

Contracts are designed around the problem and the devices are recognisable:

• •Deductibles and excesses, which leave the insured bearing the first portion of any loss
• •Co-payments and coinsurance, leaving a proportion of every cost with the insured
• •Coverage limits, capping what is paid
• •Premiums that respond to claims history, which restores a consequence over time
• •Monitoring and required precautions, which make cover conditional on behaviour
• •Exclusions for deliberate acts and for specified risky activities

The concept dominates argument about financial regulation. A bank whose failure would damage the wider economy may be rescued by a government, and if that is anticipated, the bank and its creditors face less downside from risk than they otherwise would, which encourages taking more of it. The 2008 crisis and the rescues that followed sharpened the argument considerably, and subsequent regulation attempted to address it through higher capital requirements, through arrangements for resolving failing institutions without public money, and through rules imposing losses on creditors before any public support. Whether those arrangements are credible is the central question, since a commitment not to rescue is only effective if it is believed, and a government facing an actual collapse has strong incentives to intervene regardless of what it previously said, which is a commitment problem rather than a design flaw.

The practical question is not whether the effect exists but how big it is in a given setting, and that has been measured in several cases. A large randomised study of health insurance in the 1970s assigned families to plans with different levels of cost sharing and found that those paying more used fewer services, with the reduction spread across care judged necessary and unnecessary alike, which complicates the usual argument that cost sharing targets waste. Studies of unemployment support find modest effects on the duration of job search, varying with the generosity and the labour market. Insurance research finds measurable effects in some lines and negligible ones in others. The general pattern is that the effect is real, is smaller than rhetorical use assumes, and varies enough between contexts that assuming a magnitude rather than measuring it is not defensible.

The concept is invoked far beyond situations where it applies and the misuse is worth identifying. It is used to oppose insurance and support of all kinds on the assumption that any reduction in consequences produces carelessness, which converts an empirical question into an assumption, and the empirical evidence varies enormously by context. Health insurance does increase use of services, and distinguishing genuinely unnecessary use from care people were previously going without is the substantive question rather than a detail. Unemployment support does lengthen job search modestly, and whether that is waste or better matching is contested. Applying the term to situations where people cannot meaningfully control the risk, including many illnesses and disasters, is simply wrong, since there is no behaviour to change. And invoking it rhetorically to oppose helping people while presenting the objection as technical is a recognised move.

Behaviour changes when someone else bears the consequences, which follows from incentives rather than dishonesty, and the word moral is historical and misleading. Deductibles, co-payments and experience-rated premiums restore some consequence. In banking the difficulty is that a commitment not to rescue is only effective if believed. The term is frequently applied where people cannot control the risk at all.