With bike and motorcycle helmets and seatbelts, you can compare the stats between jurisdictions that require their use and jurisdictions that don't. Interestingly what you get is a splitting of risk clusters. For most of the population, the introduction of the safety equipment reduces overall injury and death. But there's a subgroup for whom the stats move slightly towards increased fatalities at higher speeds. Turns out this subgroup figures they're safer now so take even more risks.
Overall though, they decrease injury and death.