It's an issue many safety professionals ignore or deny. Obviously, the notion that personal protective equipment could encourage risky behavior is extremely repugnant. But we need to explore this issue: Do we compensate -adjust our behavior- for perceptions of safety?

In recent years, the issue of compensating for safety has been given several different labels, including risk homeostasis, danger compensation, risk-offsetting behavior, and perverse compensation. Whatever the label, the basic idea of risk compensation is quite simple and straightforward. People are presumed to change their behavior to compensate for changes in perceived risk. If a job is made safer with machine guards or the use of PPE, employees might perceive less risk and actually work more dangerously.

This idea certainly seems intuitive. In fact, I bet every reader has experienced the phenomenon. I clearly remember the increased risks I took when wearing a standard high school football uniform. With the protective helmet and shoulder pads, I'd willingly throw my body at another player or leap high to catch a pass. I didn't perform these behaviors until perceiving security from the personal protective equipment.

I'm sure you can apply my experience to a number of competitive or recreational activities in your life. So how can the phenomenon of risk compensation be denied?

Research findings

As predicted by risk compensation theory, Peltzman found that these vehicle-manufacturing safety standards did not reduce the frequency of crash fatalities per miles driven. Perhaps the most convincing evidence of risk compensation was that the cars equipped with safety devices were involved in a disproportionately high number of crashes.

Peltzman's findings have been criticized on a number of counts, primarily statistical; but they did stimulate a number of follow-up investigations. Dr. John Adams of University College, London, UK, for example, compared traffic fatality rates between countries with and without safety belt use laws. His annual comparisons (from 1970 to 1978) showed dramatic reductions in fatal vehicle crash rates after countries introduced seat belt use laws.

Taken alone, this data would lend strong support to the seat belt legislation. However, the drop in fatality rates was even greater in countries without safety belt use laws.

These studies compared risk behaviors across large data sets. Behavioral scientists call this a between-group design, and it can only indirectly test the occurrence of risk compensation. Since the theory predicts that individuals take more chances after perceiving an increase in safety or security, it can only be tested by comparing the same group of individuals under different conditions. Behavioral science researchers call this a within-subject design.

Most within-subject tests of risk compensation theory have been restricted to simulated laboratory investigations. This is because within-subject observations under different risk conditions are time consuming and quite difficult to pull off in a real-world situation. One such study was conducted by Dr. Fredrick Streff and myself in 1987. To conduct this research we built an oval clay go-kart track (about 100 meters in circumference) and equipped a 5-hp go-kart with an inertia reel-type combination shoulder-lap safety harness.

Subjects were told to drive the go-kart around the track quickly, but at a "comfortable" speed. The 56 subjects were either buckled or unbuckled in the first of two phases of 15 driving trials. After the first phase, the safety condition was switched for half the subjects. That is, the safety belt was unbuckled for subjects previously buckled up, or the belt was used by subjects who previously did not use it. The speed and accuracy of each subject's driving trial were systematically measured. Following the first and second phases (of 15 trials each), the subjects completed a brief questionnaire that assessed their perceived risk while driving the go-kart.

The between-subject comparisons showed no risk compensation. That is, the subjects who used the safety belt for all trials did not drive faster than subjects who never used the safety belt. And, the perceptions of risk were not different across these groups.

On the other hand, the within-subject differences did show the predicted changes in risk perception and behavior. That is, subjects reported feeling safer when they buckled up, and subsequently drove the go-kart significantly faster than subjects who used the safety belt during both phases. The subjects who took off their safety belts reported a significant decrease in perceived safety, but this change in risk perception was not reflected in slower driving speeds (compared to drivers who never buckled up in the go-kart).

Our go-kart study was later followed up in the Netherlands using a real car on real roads. Specifically, habitual, "hard-core" non-users of safety belts buckled up at the request of the experimenter. Compared to measures taken when not using a safety belt, these buckled-up drivers drove faster, followed more closely behind vehicles in front of them, changed lanes at higher speeds, and braked later when approaching an obstacle.

I'm convinced from personal experience and reading the research literature that risk compensation is a real phenomenon.

Safety implications

First, the off-setting or compensating behavior doesn't negate the benefits of safety protections. Although football players increase risky behaviors when suited up, they sustain far fewer injuries than they would without the PPE. More importantly, Professor Gerald Wilde, the leader in risk homeostasis theory and research, says safety excellence cannot be achieved through top-down rules and mandates. Under these conditions, some people only follow the rules when they are supervised and might take greater risks when they feel "free." Such behavior is predicted by risk compensation theory.

As the title of his 1994 book, "Target Risk," indicates, Wilde advocates that safety interventions need to reduce the level of risk people are willing to tolerate. This requires a change in values. Wilde claims that improvements in safety cannot be achieved through training, engineering or enforcement. In his words, "the extent of risk taking with respect to safety and health in a given society, therefore, ultimately depends on values that prevail in that society, and not on the available technology" (Wilde, 1994, p. 223).

This means, of course, our vision should be to make safety a value, not a priority. When safety becomes a value, risks will not be tolerated, regardless of how protected and safe workers feel.