Peltzman Effect

If you look at the data on traffic deaths it is obvious that we are doing something right in the US.
As far back as the data goes traffic deaths have been falling at a -3.2% annual rate.

But if your source of information was economics you would be hard pressed to know this. It has been long shown that driving/riding in a small car was dangerous and led to more deaths. But now we have a new study at
where Michael Anderson found that driving a large SUV led to the same results — more deaths.

It is as if economists found that living was bad for your health.

The approach that I’ve long had troubles with is the Peltzman Effect that concludes that seat belts cause more deaths. Peltzman agrees that wearing a belt and having air bags sharply reduces the changes of death if you are in an accident. He does question the government estimate that these reduce your chance of death by 50%. Rather Peltzman argues that wearing a seatbelt makes you feel safer and so drive more dangerously and this causes more accidents
that offsets the gains from wearing seatbelts. They conclude that one should ignore the direct evidence that seatbelt use and traffic deaths have a -0.9 correlation.

Rather they model traffic deaths as a function of real income, demographics, crime, race, congestion, and other similar developments and develop models where seat belt use has a positive rather then a negative sign.

Of course if you look at the -3.2% trend and regress it against almost any real economic variable it would show up as a powerful variable that is likely to swamp other variables. But I’m not going to get into that now.

However, there are studies that use the same approach and find a much weaker impact and even find no evidence for the Peltzman Effect. On is a study by Alma Cohen and Liran Einav,
Cohen and Einav even suggest an alternative thesis that the act of putting on a seat belt makes a driver think about how dangerous it is on the road and leads to them driving more carefully–
just the opposite of the Peltzman Effect.

I’m not going to go in to the econometric arguments behind these two opposing thesis. Rather I’m going to take a different approach.

The Peltzman Effect contents that sealtbelts cause more accidents. But when I look at the accident data I find that the data shows falling accident rates. The first data series is a very old series on the number of accidents from the National Safety Council — I found it in the Statistical Abstract. It shows that accidents have fallen since sealtbelts were first started to be widely used in the early 1960 and really fell sharply after laws were passed at the state level in the 1980s making sealtbelt use mandatory. That is when seatbelt use also jumped sharply.

The second set of data is a new series from the National Highway Traffic Safety Administration
that shows traffic crashes since 1989. It can be found at

The people doing the Pelzman type studies think that the accident data is unreliable and do not use it in their studies. Maybe, but if it is biases, it probably has the same biases year after year. For example both data sources could easily under-report minor fender binders that are never reported to the police.

But I have now found two independent sources of data that clearly show that since 1964 when sealtbelts were first required to be installed in new cars or since the early 1980s when state laws requiring sealtbelt use became widespread that the traffic accident rate has been falling. This data or evidence directly contradicts the Peltzman Efeect thesis that depends on rising accident rates.

So, for all the people who believe in the Peltman Effect and are teaching their students that seatbelts cause more deaths then they save I have two questions. One, why should I ignore this evidence that traffic accidents have been falling? Two, where is your evidence of more accidents.

If Peltzman is right, it should not be hard to find evidence of a sharp rise in accidents that the thesis implies, right?

So where is it?

P.S. I failed to point out that the accident data is based on data collected by the insurance industry while the crash data is based on police reports to the Transportation Department.
So they really are independent data.

PPS.. Several people have raised he issue of demographics. this Chart shows that in the mid-1960s when seatbelts were required to be installed in cars their should have been a big jump in deaths due to young drivers. But it barely happened — the jump in the early 1960s precedes seatbelt installation and should have continued into the mid-1960s to the late 1960s. this data implies that seat belts actually prevented higher deaths in the 1960s due to baby boomers starting to drive — again jus the opposite of the Peltzman Effect.