Pattern/Operations and Risk/No. 0861
Risk Compensation
Risk compensation is a change in behavior in which people take more risks when they feel safer, potentially offsetting some benefits of safeguards. Also called the Peltzman effect after economist Sam Peltzman, it varies by setting and does not imply that safety gains are fully lost.
Also called Peltzman Effect
- Evidence
- Mixed evidence
- Read
- 7 min
- Links
- 7 connections
01You've seen this when…
- in life
You put on ice cleats and take the steep shortcut you avoided yesterday. The extra grip changes your choice of route.
- at work
A warehouse installs automatic braking on its forklifts. Drivers who used to slow well before a crossing start approaching it faster.
- out in the world
A town widens the shoulders on a winding road. Drivers feel they have more room for mistakes, and traffic speeds creep up.
02The idea
A safeguard changes more than the consequences of an accident. It can also change what people choose to do before one happens.
Risk compensation is the possibility that feeling safer leads people to take more risks. They may drive faster, leave less space, attempt a harder route, or check less often. That response can offset some of the protection. The important word is some. A measure can change behavior and still deliver a large net safety benefit.
Keep three things separate: what the safeguard does at unchanged behavior, how behavior changes after its introduction, and the final effect on harm. Those are different questions. Evidence for a behavioral change leaves the question of the safeguard’s success open.
This is also weaker than risk homeostasis, the theory that people adjust their behavior to maintain a preferred level of risk. Compensation can be partial and vary across situations without a deliberate decision to take more risks. Evidence has yet to establish that people spend every safety gain on extra danger.
The practical lesson is to evaluate the protection with people using it. Perfect compliance and inevitable recklessness are assumptions that require evidence.
03Why it happens
- The margin feels larger. Better brakes or a protective barrier can make an action seem more forgiving. People may use that margin for speed, convenience, or a more demanding activity.
- The reward stays visible. Finishing a delivery sooner provides an immediate benefit. The additional chance of a collision remains uncertain. Existing production pressure can encourage people to use new protection to work faster rather than more safely.
- The protection becomes a substitute for attention. Someone who expects a warning system to catch mistakes may check less carefully. This overlaps with automation complacency, though compensation can happen without any automation.
- Perception matters more than engineering specifications. Behavior responds to how protected people think they are. A conspicuous but limited safeguard may encourage more adjustment than an effective safeguard they barely notice.
None of these responses requires a conscious decision to accept more danger. People can gradually change their pace or habits without calculating the trade-off.
04A worked example
Consider an illustrative warehouse introducing automatic collision braking on its forklifts. Management expects fewer collisions and keeps the same delivery targets. The following numbers are invented to show the distinction between protection and compensation.
What it looks like The equipment assessment predicts that braking would reduce collisions from 20 to 10 per year if vehicles kept using the same routes at the same speeds and traffic volume stayed unchanged. After installation, the warehouse records 14. Managers conclude that the technology underperformed.
What’s actually going on Suppose observation also shows faster approaches to crossings and shorter following distances. Those changes could explain part of the gap. In this simplified example, behavior offsets four of the ten collisions the equipment would otherwise prevent. But the warehouse still improves from 20 to 14. Partial compensation is not the same as no benefit. The numbers alone would leave the explanation uncertain because the warehouse might also have changed its reporting or experienced shifts in workload and traffic patterns.
What would have helped Counting close calls before rollout while recording approach speeds and vehicle movements, then tracking the same measures afterward. Management could retain speed limits and separate pedestrian routes while keeping delivery targets from rising merely because automatic braking is available. That is defense in depth: the new protection works together with the other safeguards, which remain in place.
05How to spot it
Check who received the protection, too. Higher-risk people may be more likely to adopt it. Finding more accidents among protected users can reflect selection bias, with the difference arising from their higher starting risk.
06What to do about it
- Keep the effective protection. Judge helmets, seat belts, barriers, and other safeguards with demonstrated benefits by their overall effects, even when compensation is plausible.
- Measure behavior as well as injuries. Track the margins people control: speed, spacing, checking, exposure, and rule compliance. These leading indicators can reveal adjustment before enough injuries accumulate to show a pattern.
- Test the complete intervention. Where feasible and ethical, use a staged rollout or an A/B test. Compare similar users and account for activity levels. Don’t remove established protection just to create a comparison group.
- Explain the protection’s limits. Make clear what it catches, what it misses, and which precautions remain necessary. Training should address the mistaken belief that the system makes an activity safe under all conditions.
- Protect the safety margin from production pressure. Don’t automatically shorten schedules or increase throughput after adding a safeguard. Check whether managers, incentives, or informal expectations are encouraging people to consume the new margin.
The aim is to preserve worthwhile gains while preventing avoidable harm as behavior changes.
07When it isn’t risk compensation
More activity can occur while people remain just as cautious. A protected cycle route may encourage people to ride more often. That increases exposure, but it can be a valuable outcome, especially if risk per trip falls. Assess total harm separately from risk per unit of activity, then consider the activity’s benefits on their own.
Compensation also differs from moral hazard. Moral hazard concerns incentives created when someone else bears part of the downside. This often happens under insurance or a contract. Risk compensation concerns behavior responding to perceived protection. Each can occur on its own, and they can overlap.
Most importantly, the effect is not universal. A systematic review of bicycle helmet research found little support for the claim that wearing a helmet encourages riskier behavior. Broader public-health reviews likewise warn against treating compensation as an automatic objection to prevention.
A safeguard may go unnoticed, while people who notice it may have no reason to change behavior or may become more cautious because its introduction draws attention to danger. Measure the response.
08Roots
In 1975, economist Sam Peltzman examined American automobile safety regulation. Cars were gaining features such as seat belts and padded interiors. The engineering question was whether these features protected occupants. Peltzman added a behavioral question: would drivers respond to that protection by driving differently?
His analysis argued that behavioral adjustment offset benefits from vehicle safety regulation, potentially shifting some danger onto other road users. The interpretation became influential and controversial. It helped establish the problem as something safety evaluations should investigate, but it did not settle how much compensation occurs across interventions.
Psychologist Gerald Wilde proposed a stronger account in 1982: people regulate their behavior around a target level of risk. His risk-homeostasis theory brought the idea into wider safety debates. Later reviews, including James Hedlund’s, emphasized examining whether people notice a change, believe it affects their safety, and have an opportunity or motive to adjust their behavior.
One adaptive hypothesis is that people use protection to pursue more reward, accepting some danger in the process. Extra security might make exploration or a demanding task worthwhile. That is a plausible trade-off story. Establishing it as an evolutionary explanation would require more evidence. The story leaves open whether people accurately judge risk or maintain a fixed risk target.
09How solid is this?
Behavioral adjustment occurs in some settings, but its size and effect on overall harm vary. Reviews of bicycle helmets and public-health interventions find little support for assuming compensation; the stronger claim of a stable target level of risk is not established.
10Connections
- Often confused with Moral Hazard
- Countered by Defense in Depth, A/B Testing
- Part of Unintended Consequences
- IncludesAutomation Complacency
- See also Margin of Safety, Selection Bias
11Origin and sources
Sam Peltzman’s 1975 analysis of automobile safety regulation gave an influential economic account. Gerald Wilde’s 1982 risk-homeostasis theory proposed the stronger claim that people adjust behavior around a target level of risk.
- [1]Peltzman, S. (1975). The Effects of Automobile Safety Regulation. Journal of Political Economy, 83(4), 677–725.
- [2]Wilde, G. J. S. (1982). The theory of risk homeostasis: implications for safety and health. Risk Analysis, 2(4), 209–225.
- [3]Hedlund, J. (2000). Risky business: safety regulations, risk compensation, and individual behavior. Injury Prevention, 6(2), 82–90.
- [4]Esmaeilikia, M., Radun, I., Grzebieta, R., & Olivier, J. (2019). Bicycle helmets and risky behaviour: A systematic review. Transportation Research Part F: Traffic Psychology and Behaviour, 60, 299–310.
- [5]Mantzari, E., Rubin, G. J., & Marteau, T. M. (2020). Risk compensation is not a reason to abandon public health interventions. BMJ, 370, m2913.
Suggest an edit· Updated 2026-10-02