Journalistic innumeracy, hit and run edition.

I have no particular dog in this fight, I am just struck by the oddity of the reporting. The brief article is Hit-and-Run Fatalities Soar as More People Bike to Work by Scott Calvert.

Hit-and-run crash deaths are rising nationwide, and pedestrians and bicyclists account for close to 70% of the victims, according to a new report, as more people cycle to work and motor-vehicle fatalities are at a near-decade-high level.

The number of hit-and-run fatalities jumped 61% from 2009 to 2016, the most recent year for which data are available, according to the report from the AAA Foundation for Traffic Safety.

About 68% of fatal hit-and-run victims in 2016 were pedestrians or cyclists, compared with 61% a decade earlier, according to federal data cited in the report.

Click to enlarge.

In 2016, 1,980 fatal hit-and-run crashes across the U.S. resulted in 2,049 deaths—both record highs in the roughly four decades that the National Highway Traffic Safety Administration has tracked such data, the report said.

The oddity is that Calvert is focusing on a subset (Hit-and-Runs) of a superset (all traffic accidents) without paying much attention to the superset. To understand what is happening with the hit-and-run numbers, we have to understand what is happening to the overall traffic accidents numbers.

If hit-and-runs are up 10% and traffic accidents are up 10%, nothing much has changed. So are there more traffic accidents?

And why provide ten years of data and then arbitrarily start your change calculation mid-way through?

If you get the numbers from the graph, you find that hit-and-run fatalities have increased 55% between 2009 and 2016, not 61% as claimed by Calvert. Both are dramatic jumps, but one would wish that the numbers were calculated accurately.

But there is still that question of why pick 2009 as you starting point? Clearly that was a low point in the cycle which exaggerates the issue. If you look at the starting point in 2006, there is only an 11% increase in hit-and-run deaths.
Tragic, but perhaps not headline making.

Calvert lists a number of talking heads speculating why the numbers are up. Distracted driving owing to cell phones. More people biking, etc. But the biking numbers contradict Calvert's own numbers.

The number of bike commuters nationwide has ebbed in recent years, but rose nearly 40% from 2006 to 2016, when 864,000 rode to work, according to the Census Bureau.

So biking volume rose 40% from 2006 to 2016 but fatal hit-and-runs only rose 11%. Fatal hit-and-runs involving bicycles only rose 25% between 2006 and 2016. Looks as if bicycle riding is actually getting comparatively safer.

Vehicle miles driven are up in 2015 and 16 after several flat years. Traffic deaths are up strikingly from their low in 2014 of 32,744. If we use Calvert's preferred base of 2009, then traffic fatalities are up roughly 10% between 2009 and 2016. But if we use 2006, then traffic deaths are down about 12%.

Of course, if we really want to understand what is going on, we need to cast all these numbers in terms of rates. As it is, data is sloshed around with an overall implication that distracted drivers are killing more bicyclers, an interpretation which hinges on picking an aberrant year and ignoring other factors.

A dog's breakfast of an article once again demonstrating the weakness of innumeracy in journalism.

Still it is interesting for a couple of unintentional reasons. Looking at the Wikipedia entry reminds me of just how far we have come. Look at those deaths 1966 through 1980. More than 50,000 nearly every year in a country with several tens of millions fewer people. 37,461 in 2016 is still way too high but it is nearly 25% lower than those earlier dreadful decades. Nearly 15,000 more people surviving each and every year than in those earlier decades. A whole Army division. Fantastic.

If I were to impose a speculative story on Calvert's numbers, I would observe that the 2009 low in hit-and-runs probably maps to a dramatic decline in miles driven post 2007 Great Recession and when gas was at recent high. The decline in hit-and-runs in this telling is due to a decline in driving. Once the cost of gas came down and the economy picked back up, so did hit-and-runs.

I would also observe that it is possible that there was a positive outcome to the otherwise lamentable cash-for-clunkers policy failure. Note from Calvert's graph numbers, driver deaths from hit-and-runs are up from 348 to 380 between 2006 and 2016. However, passenger deaths are down to 229 from 321. Why is there a decline? If more drivers are dying, you would expect that more passengers would be as well. My guess is that cash-for-clunkers took a lot of vehicles with driver-side only airbags off the road and which were replaced with new dual airbags, thus continuing to drive down passenger deaths.