Car seat laws as contraception

The headline version went like this. Car seat laws have stopped 145,000 American babies from being born. Safety rules meant to protect children are quietly shrinking families. That is roughly what the paper says, and the way it gets there is more careful than the retweets.

The claim

Jordan Nickerson (then at MIT Sloan, now at the University of Washington) and David Solomon (Boston College) wrote a paper with the best title in recent economics, “Car Seats as Contraception.” Their argument is simple. Most cars fit two child seats across the back row, not three. When a state raises the age up to which a child must be in a car seat, some families with two kids under that age face a real cost to having a third. You need a bigger car, or you stop at two. The paper finds that they stop at two more often than they otherwise would.

How they found it

The trick is that states changed their car seat laws at different times. In 1980 the typical mandate stopped at age three or four. By the late 2010s most states required a seat, or a booster, up to age eight. That staggered rollout gives you a natural experiment. Compare a woman with two kids under the mandate in a state that just raised its age limit to a similar woman in a state that has not yet.

The data is the American Community Survey from 2000 through 2017 plus 5 percent samples from the 1990 and 2000 censuses. The survey records whether a woman gave birth in the last year, how many children she has, their ages, and whether the household owns a car. That last piece matters. If the story is about back seats, the effect should show up in households with a car and vanish in households without one.

The headline number is 0.73 percentage points. Women with two children both younger than their state’s car seat cutoff are that much less likely to give birth in a given year. For context, the paper’s own estimate of the baseline annual birth probability for those women puts the drop in the range of a several percent relative decline, not a collapse.

The mechanism

The authors build a case that this is about seats, not about something vaguely correlated with safety laws. The effect shows up only for third births, not first or second. It shows up only in households with access to a car. And it is larger when a man is present in the household, which the authors read as both front seats being occupied so there is no escape hatch of putting a kid up front.

They also priced it. Using vehicle fatality data, they estimate the laws prevented about 57 child car crash deaths in 2017. The same year they estimate 8,000 fewer births. You can argue with the comparison, and people did, but the paper puts the trade-off on the table rather than hiding it.

How much to believe it

One paper. This is a single study. It circulated as an SSRN working paper from July 2020 and was published in the Journal of Law and Economics in 2024. I could not find an independent replication using different data, and I looked.

Large sample, good design. The ACS is a big, well-understood survey and the staggered state rollout is a standard identification strategy. Nobody has argued the numbers are wrong. The arguments are about what they mean.

The strongest criticism. Two lines of attack. First, the workaround problem. Narrow car seats exist, and three of them do fit in many mid-size sedans. If families can buy a slim seat for a fraction of the price of an SUV, the cost of a third child is smaller than the paper implies, and the mechanism gets shakier. A commenter at LessWrong and Zvi Mowshowitz both raised this. Second, the timing problem. The steepest fertility declines in the data coincide with the 2000s and the post-2008 collapse in US birth rates that has many causes. The staggered design is supposed to net that out, but a skeptic will note that the 145,000 figure is an extrapolation from a small per-year coefficient across four decades, and extrapolations compound whatever error is in the coefficient.

Does it matter in real life? For any individual family, a 0.73 percentage point change in annual birth odds is tiny. Summed across every two-kid household in the country over 37 years it becomes a number with six digits, which is where the headline comes from. Both facts are true at once. The result is more interesting as a demonstration that fertility responds to small, dumb costs than as a policy lever anyone should pull.

What it means for you

If you build models of anything with a long tail of small per-unit effects, this is the pattern to notice. A coefficient that is invisible at the household level can be enormous at the population level, and the confidence interval on the aggregate is wider than the confidence interval on the coefficient suggests.

Suggested chart

A step chart of the median state car seat age mandate from 1980 to 2017, with the third-birth rate for two-child households overlaid. Source would be the state law data in the paper’s internet appendix plus the ACS.

Sources

1. Nickerson, J. and Solomon, D. H., 2024, “Car Seats as Contraception,” Journal of Law and Economics, 67(3). journals.uchicago.edu. Working paper version on SSRN.
2. Nickerson, J., Internet Appendix for “Car Seats as Contraception.” jordannickerson.com.
3. Mowshowitz, Z., 2020, “On Car Seats as Contraception.” thezvi.substack.com.
4. Reason, 2020, “Are Car Seat Laws Driving Down America’s Birthrate?” reason.com.