I read an article (see http://www.riskandinsurance.com/070415_column_4.asp ) that poorly explains the concept of risk compensation.
All things equal, wearing a seat belt while driving is safer than not wearing a seat belt. Risk compensation results from the fact that, in the aggregate, all things are not equal. On average, people who wear seat belts due to a compulsory law feel a bit safer so they drive a little faster or more recklessly than otherwise. Thus, the number of accidents with or sans seat belts is the same. In terms of ethology, this is called behavioral adaptation. From the point of view of individual autonomy, however, risk compensation makes no sense. Indeed, if it did, this description would be akin to risk homeostasis, the controversial (and, in my opinion, incorrect) notion that we're all endowed with internal "risk thermostats" and adjust our behaviors until we reach our individual risk comfort levels (this concept was popularized by Gerald Wilde).
Risk compensation has its place, but the ideas in that article are overreaching.