Buckle up your children? Endogeneity and car seats

I was in an online discussion about car seats yesterday and recalled a paper by Freakeconomist, Steve Levitt, that for 2-6 year olds car seats are as effective at preventing deaths as seat belts.  As Levitt points out you can't just look at car seat use since people who tend to use them are wealthy and people who take less risks (likely to have better safety features on their car and get in fewer serious accidents).  In other words the choice to use a car seat is endogenous to other factors that also influence the likely hood of crash. A more recent follow up paper by Doyle and Levitt show with a different data set  a similar result in terms of serious injury 9no difference between safety seat and belt), but they also show car seats for 2-6 year olds reduce less serious injuries.

Classical economics suggests that you shouldn't tell people what to do, since you might not know their preferences. Assuming Levitt's results are correct what is the trade off between a reduction in less serious injury versus the expense and hassle of a car seat. Most economists would point out we all take risks, not driving is likely less risky than driving.

But the law and social norms say I should put my 2 year old daughter in a car seat, and I haven't really researched it so as a risk averse dad, I'm going to stick with the car seat.