The recent issue of "Status Report" (42:10, pp 6-8) from the Insurance Institute for Highway Safety has a polemical article against "risk compensation" approaches to safety restraint legislation. In the article, "Risk Compensation Theory Keeps Popping Up Where It's Wholly Irrelevant," authors argue that the introduction of new safety features does not inherently alter driving behaviors for the worse. It is noted that, in response to a recent article on risk compensation in the statistical analysis journal "Significance,"
To believe [risk compensation concerning safety restraints in automobiles] you'd have to believe that people have a certain tolerance for risk and that their levels of risk are regulated by a homeostatic mechanism so that, if forced to "consume" more safety than they voluntarily would choose, people will balance the safety increase by taking more risk. It's a stretch, isn't it? (p 6)
The article being argued against, "Seat Belt Laws- Repeal Them?" (pdf format, requires subscription) is written by a professor emeritus from the geography department of UCL, John Adams. Prof. Adams explains that
"the evidence with respect to seat belts suggests that the law had no effect on total fatalities but was associated with a redistribution of danger from car occupants to pedestrians and cyclists. (p 89)
Finally, an article in Accident Analysis and Prevention presents some quantitative research on the matter (found here, in pdf and subscription required).