NYT’s Richtel traffic hype soars at historic rate

What if everything you learned about traffic fatalities you got from Matt Richtel at the New York Times?

On April 27, Richtel wrote:

Over the last seven years, most states have banned texting by drivers, and public service campaigns have tried an array of tactics — “It can wait,” among them — to persuade people to put down their phones when they are behind the wheel.

Yet the problem, by just about any measure, appears to be getting worse. Americans confess in surveys that they are still texting while driving, as well as using Facebook and Snapchat and taking selfies. Road fatalities, which had fallen for years, are now rising sharply, up roughly 8 percent in 2015 over the previous year, according to preliminary estimates.

I left the paragraph breaks as they were, so you can see the connection he implied. Sure seems that all that texting is driving up the rate of fatalities, although there is no evidence offed for that. Of course, since you only read Richtel, you don’t know that since 1994, cell phone subscription rates have risen 1200% while traffic fatalities have fallen 13%. (My series on this, with all these facts, is under the texting tag.)

But what about this “now rising sharply” fact? The same fact – an estimated 8% increase in one year, grew from “now rising sharply” on April 27 to “soaring at a rate not seen in 50 years” by May 22.

When I complained to the NYT Science Desk that this was a misleading representation of a traffic fatality rate that is still at historically low levels, someone checked it out and nicely informed me they had “confirmed the article accurately states the fact: preliminary estimates indicate road fatalities are rising at a rate not seen in 50 years.” Complaint denied.

Assuming you share my obsession with this problem of hyping traffic fatalities – and distracting the public with stories of bad drivers instead of paying attention to the real problem, which is rampant car culture itself – then you’ll want to make a distinction between the facts themselves and the NYT representation of them.

Facts

The fact here is actually kind of weird. Instead of using the official traffic fatality rates, which the National Highway Traffic Safety Administration hasn’t released yet for 2015, Richtel here is reporting the preliminary 2015 estimate from a private group, the National Safety Council. The weird thing is that NSC uses a different method, counting people as dead from a traffic incident if they die from any cause within a year of the accident, while NHTSA only counts them as dead from the incident if they die within 30 days. The rule is arbitrary either way, but I prefer to NHTSA method in the absence of a compelling argument. (As any Law & Order fan can tell you, people who die can have that death attributed to something that happened years earlier if the medical examiner owes the detectives a favor.) Not surprisingly, NSC produces estimates that are higher – about 8% higher, about 3,000 deaths more than the roughly 35,000 NHTSA reports. The longer death window seems bad for comparing rates over time, because the population is aging and therefore the death rate will probably rise among people who have had an accident in the previous year just because they’re older on average.

Anyway, since NSC doesn’t report their long-term trend (at least on the free part of their website), I applied their estimate of the “soaring” 2015 change – an 8.2% increase in total deaths – to the NHTSA series (helpfully recorded on this Wikipedia page), to extend the series to 2015. We also now have the Federal Highway Administration’s report on vehicle miles traveled in 2015 (+4% from 2014), so I can use that estimate of total deaths to calculate deaths per mile for 2015, as well as deaths per person (using the average of the Census Bureau’s monthly estimates for the year, which was 321.4 million, +0.8%). By these calculations, Richtel’s soaring 8.2% increase in total deaths becomes a 4% increase in deaths per mile, and a 7.3% increase in deaths per person. Here are the 50-year trends, with recessions shown:

The dramatic increase in deaths for 2015, which is quite large on a relative scale during this time period – in fact, at no time in last 50 years has the number of deaths increase by 8.2% in one year – looks kind of small. But perhaps more important, it seems in line with the cyclical nature of the trend. Death rates fall during recessions and rebound afterwards. In fact, the declines in 2008 and 2009 – which were 9.3% and 9.5% respectively – are also unprecedented during this period, so a larger rebound is not surprising.

Representations

So how should Richtel describe the trend? Keep in mind this is a reporter who is still promoting a book premised on the crisis of distracted driving – the homepage for which, despite notifications to the author and publisher, still repeats the bogus internet meme that 11 teens per day die from texting while driving.

After I complained on Twitter, Richtel tweeted to his followers:

Fatalities soaring at rate not seen in 50 years. (Friends, please ignore white noise) https://t.co/MHjOYEE554. https://t.co/uZM26zvhCa

— matt richtel (@mrichtel) May 23, 2016

So I put the description question to the test of my Twitter readers, offering them a poll. I made two figures, one from the actual death trend (with my 2015 estimate), and one with the same trend in reverse, and asked people (without comment) which one they thought was better described by Richtel’s phrase:

Here are two trends. Based only on the description given, which series do you think fits best? (poll follows) pic.twitter.com/l8ooOIETbv

— Philip N Cohen (@familyunequal) May 24, 2016

My readers were closely divided, but gave Richtel the edge, 56% to 44%:

Here's the poll: The description given best fits…

— Philip N Cohen (@familyunequal) May 24, 2016

I don’t think it’s very strong vindication to barely eke out a majority in a poll where people are asked to choose which is right, what you said or the exact opposite of what you said. I would think real reporting would have a higher rate of concordance than that.