In 2017 academic researchers got a lot of press when they showed putting a tiny sticker on a stop sign could turn it into a speed limit sign. The human eye was not fooled. Machine vision was.
In 2019 ticket industry lobbyists got a lot of press when they turned a change in speed limits into a public safety emergency. Statisticians were not fooled. Reporters were.
In machine learning we call it overfitting. In crash data analysis it’s a blend of overfitting and its twin underfitting. People pretend the flawed model is reality, just like they keep citing a discredited mathematical model of pedestrian fatalities.
Recall that last year the Insurance Institute for Highway Safety lied about a speed limit reduction in Boston. The data showed no change in traffic speed after the city speed limit was reduced from 30 to 25. Grasping for anything to publish, they looked at cars that were going 36 mph when the speed limit was 30. A few of those slowed down to 35. Average speed didn’t change so a few must have sped up to 37. The press release boasted of a big decrease in traffic going over 35. In fact the evidence showed even that ±1 mph change was not caused by the speed limit reduction.
If you prune the data set and explore enough parameters, you will find a model to make your data fit your conclusion.
This spring the IIHS dumped another load of statistical BS: speed limit increases killed tens of thousands of people. This press release is regurgitated from 2016. As long as it gets attention they’ll keep publishing it. It’s been debunked before and I’m mostly not getting into the statistics.
Instead I’m going to ask, “so what?” What if we lived in a fantasy world where people drive the speed limit and deaths go up with speed limit increases?
Among the government’s many statistical reports is one called “The Economic and Societal Impact Of Motor Vehicle Crashes.” You can use it to put a dollar value on car accidents and traffic delay.
At $25 per hour the time we spend driving adds up to around a trillion dollars per year.
The IIHS counted fatalities. Fatal motor vehicle crashes cost us $176 billion per year. (Most of that is noneconomic cost, the $100,000 per year society values being alive and well instead of cold and dead.)
The ratio of those costs is over 5:1. If we drive 1% faster and traffic deaths increase by 5%, the time saved cancels out the increase in deaths.
The IIHS claims increasing the maximum speed limit in a state by 5 mph (about 8%) increases total fatalities by less than 4%. In our fantasy world where people drive the speed limit and faster driving is deadly, the time savings from speeding up is worth ten times as much as the risk. In other words, IIHS says speed saves. Floor it.
We’ve been here before. Many years ago Scottsdale, Arizona got some publicity by paying for a report saying speed cameras were good. Subjected to a similar cost-benefit analysis the report actually said we should ban speed cameras.
I said the IIHS lives in a fantasy world. Here’s a clue. Increasing the highest speed limit on any freeway in a state was claimed to increase fatalities by 2.8% on roads other than freeways. When Texas raised the speed limit on a toll road near Austin, people died on city streets in El Paso 500 miles away. Increasing the speed limit in West Texas killed people in Houston, where speed limits had been reduced under orders from the EPA.
The model is a broken exercise in self-gratification. The IIHS is not making a serious attempt to find out why people die on the highway. They are not doing a cost-benefit analysis of traffic policy.
All we have here is another self-serving press release from the ticket industry.
The opinions expressed in this post belong to the author and do not necessarily represent those of the National Motorists Association or the NMA Foundation. This content is for informational purposes and is not intended as legal advice. No representations are made regarding the accuracy of this post or the included links.
