Steven Levitt

Stephen Dubner

Superfreakonomics

In recent crashes and old ones, in vehicles large and small, in single-car crashes and pileups, there is no evidence that car seats are better than seat belts in saving the lives of children two and older. In certain kinds of crashes—rear-enders, for instance—car seats actually perform slightly worse.

So maybe the problem is, as NHTSA admits, that too many car seats are installed improperly. (You might argue that a forty-year-old safety device that only 20 percent of its users can install correctly may not be a great safety device to begin with; compared with car seats, the condoms worn by Indian men seem practically infallible.) Could it be that the car seat is a miracle device but that we just haven't learned to use it properly?

To answer this, we sought out crash-test data for a side-by-side comparison of seat belts and car seats. You wouldn't think this would be hard to find. After all, every car seat brought to the market must undergo crash testing to gain federal approval. But it appears that researchers have rarely, if ever, run parallel tests on child-sized crash-test dummies. So we decided to do it ourselves.

The idea was simple: we would commission two crash tests, one with a three-year-old-sized dummy in a car seat versus a three-year-old dummy in a lap-and-shoulder belt, the other with a six-year-old-sized dummy in a booster seat versus a six-year-old dummy in a lap-and-shoulder belt. In each case, the test would simulate a thirty-mile-per-hour frontal collision.

We had a hard time finding a crash-test lab that would do our tests, even though we were willing to pay the $3,000 fee. (Hey, science doesn't come cheap.) After being turned down by what felt like every facility in America, we finally found one willing to take our money. Its director told us we couldn't name the lab, however, out of concern he might lose work from the car-seat manufacturers that were the core of his business. But, he said, he was 'a fan of science,' and he too wanted to know how things would turn out.

After flying in to this undisclosable location, we bought some new car seats at a Toys 'R' Us and drove to the lab. But once the engineer on duty heard the particulars of our test, he refused to participate. It was an idiotic experiment, he said: of course the car seats would perform better—and besides, if we put one of his expensive dummies in a lap-and-shoulder belt, the impact would probably rip it to pieces.

It seemed odd to worry over the health of a crash-test dummy—aren't they made to be crashed?—but once we agreed to reimburse the lab if the seat-belted dummy was damaged, the engineer got to work, grumbling under his breath.

The lab conditions guaranteed that the car seats would perform optimally. They were strapped to old-fashioned bench-style rear seats, which give a flush fit, by an experienced crash-test engineer who was presumably far better at securing a car seat than the average parent.

The chore was gruesome, from start to finish. Each child dummy, dressed in shorts, T-shirt, and sneakers, had a skein of wires snaking out of its body to measure head and chest damage.

First came the pair of three-year-olds, one in a car seat and the other in a lap-and-shoulder belt. The pneumatic sled was fired with a frightening bang. In real time, you couldn't see much (except that, to our relief, the seat-belted dummy remained in one piece). But watching the super-slow-motion video replay, you saw each dummy's head, legs, and arms jerk forward, fingers flailing in the air, before the head snapped back. The six-year-old dummies were next.

Within minutes, we had our results: the adult seat belts passed the crash test with flying colors. Based on the head-and chest-impact data, neither the children in the safety seats nor those in the seat belts would likely have been injured in this crash.

So how well did the old-fashioned seat belts work?

They exceeded every requirement for how a child safety seat should perform. Think of it this way: if we submitted our data from the seat-belted dummies to the federal government and said it came from the latest and greatest car seat, our 'new' product—which is pretty much the same nylon strap Robert McNamara pushed for back in the 1950s—would easily win approval. Since a plain old seat belt can meet the government's safety standard for car seats, perhaps it's not very surprising that car-seat manufacturers turn out a product that can't beat the seat belt. Sad, perhaps, but not surprising.

As one can imagine, our lack of appreciation for car seats places us in a slim minority. (If we didn't have six young children between us, we might well be labeled child haters.) One compelling argument against our thesis is called 'seat-belt syndrome.' A group of prominent child-safety researchers, warning that crash-test dummies typically don't have sensors to measure neck and abdomen injuries, tell grisly emergency-room tales of the damage seat belts inflict upon children. These researchers gathered data by interviewing parents whose children were in car accidents, and concluded that booster seats reduce significant injury by roughly 60 percent relative to seat belts.

These researchers, many of whom actively care for injured children, are surely well-meaning. But are they right?

There are a variety of reasons why interviewing parents is not the ideal way to get reliable data. Parents may have been traumatized by the crash and will perhaps misremember details. There's also the question of whether the parents—whose names the researchers harvested from an insurance company's database—are being truthful. If your child was riding unrestrained in a car crash, you might feel strong social pressure (or, if you think the insurance company will raise your rates, financial pressure) to say your child was restrained. The police report will show whether or not the vehicle had a car seat, so you can't readily lie about that. But every backseat has a seat belt, so even if your child wasn't wearing one, you could say he was, and it would be difficult for anyone to prove otherwise.

Are there data sources other than parent interviews that could help us answer this important question about child injuries?

The FARS data set won't work because it covers only fatal accidents. We did, however, locate three other data sets that contain information on all crashes. One was a nationally representative database and two were from individual states, New Jersey and Wisconsin. Together, they cover more than 9 million crashes. The Wisconsin data set was particularly useful because it linked each crash to hospital-discharge data, allowing us to better measure the extent of the injuries.

What does an analysis of these data reveal?

For preventing serious injury, lap-and-shoulder belts once again performed as well as child safety seats for children aged two through six. But for more minor injuries, car seats did a better job, reducing the likelihood of injury by roughly 25 percent compared with seat belts.

So don't go throwing out your car seats just yet. (That would be illegal in all fifty states.) Children are such valuable cargo that even the relatively small benefit car seats seem to provide in preventing minor injuries may make them a worthwhile investment. There's another benefit that's hard to put a price tag on: a parent's peace of mind.

Or, looking at it another way, maybe that's the greatest cost of car seats. They give parents a misplaced sense of security, a belief they've done everything possible to protect their children. This complacency keeps us from striving for a better solution, one that may well be simpler and cheaper, and would save even more lives.

Imagine you were charged with starting from scratch to ensure the safety of all children who travel in cars. Do you really think the best solution is to begin with a device optimized for adults and use it to strap down some second, child-sized contraption? Would you really stipulate that this contraption be made by dozens of different manufacturers, and yet had to work in all vehicles even though each vehicle's seat has its own design?

So here’s a radical thought: considereing that half of all passengers who ride in the backseat of cars are children, what is seat belts were designed to fit them in the first place? Wouldn't it make more sense to take a proven solution—one that happens to be cheap and simple—and adapt it, whether through adjustable belts or fold-down seat inserts (which do exist, though not widely)—rather than relying on a costly, combersome solution that doesn't work very well?

But things seem to be moving in the opposite direction. Instead of pushing for a better solution to child auto safety, state governments across the United States have been raising the age when kids can graduate from car seats. The European Union has gone even further, requiring most children to stay in booster seats until they are twelve.