Researchers have tested this many times, in different places, over many years, and they keep finding the same result. This one is safe to trust.
The term arrived from the insurance trade, where it described the problem of a policy changing how carefully the insured behaves. Kenneth Arrow brought it into economics in 1963, defining it in the context of medicine as the observation that insurance increases the demand for care. Five years later Mark Pauly published a correction that is still the most useful thing anyone has said about it. The behaviour is a response to a price. When somebody else pays, the price you face falls, and people buy more of things that get cheaper. Calling it hazard implies a character defect that the mechanism does not require.
The measurement came later, and it is unusually good. Between 1974 and 1982, RAND recruited thousands of families across six sites in the United States and randomly assigned them to one of five health plans. One gave care entirely free. Others required the family to pay 25, 50 or 95 percent of their own medical charges. Because assignment was random, the differences that followed could be read as caused by the plan rather than by the kind of person who chooses it, which is what makes this study the reference point four decades later.
The families with free care spent substantially more, and the later reanalysis put the gap between full coverage and near-total cost sharing at about 39 percent. So the effect is real, large, and not a matter of interpretation.
The second finding is the one that rarely gets quoted. The families paying their own way cut back on inappropriate care and on appropriate care alike. From inside a decision, the necessary and the wasteful do not come clearly labelled. Across the group as a whole, health outcomes did not differ much, with one exception: people who were both poor and already in bad health went without treatment they genuinely needed.
Every arrangement where one party decides and another pays runs this. The corporate card, the expense policy, the departmental budget that resets in January, the insurance on a company vehicle, the bailout. Nobody in any of those situations has to be dishonest for spending to rise. The price they personally face has fallen, and that is enough.
Which means the common response, tighten it up and make people feel the cost, is only half right. RAND's second finding says people cut good and bad spending together. Put a hard freeze on a team's budget and you will lose the conference nobody needed and the tool that was quietly holding the process together, both. The savings are real and so is the damage, and the damage arrives later and gets attributed to something else.
The version that costs most is at the top. A decision maker whose upside is personal and whose downside lands on the company, the shareholders or the taxpayer will take risks that look reckless from outside and perfectly reasonable from inside. That asymmetry, and what it takes to close it, is skin in the game.
One distinction worth keeping straight, because they get confused constantly. Moral hazard is about who pays. Somebody else carries the loss, so the arithmetic genuinely changes. The Peltzman effect is about how safe something feels while the same person still carries the full loss. The seatbelt driver has transferred nothing to anyone. They are siblings rather than one containing the other, and a company car with full insurance manages to be both at once.
Source: Arrow, Uncertainty and the Welfare Economics of Medical Care, American Economic Review, volume 53, 1963, pages 941 to 973. The reframing as a price response rather than a character flaw is Pauly, The Economics of Moral Hazard: Comment, American Economic Review, volume 58, 1968. The experiment is the RAND Health Insurance Experiment, 1974 to 1982, summarised by RAND itself and reanalysed in Aron-Dine, Einav and Finkelstein, The RAND Health Insurance Experiment Three Decades Later, Journal of Economic Perspectives, 2013.
Tim Harford, 2008
On behaviour that looks like a character flaw and turns out to be an ordinary response to a price, which is Pauly's correction applied across a lot more of life than health insurance.
Draw your own card. It does not take long, and it rewards taking your time.