Science

These Amtrak Ridership Charts Give Data Vis a Bad Name

How not to preach to the choir

You know what I’d like to see? Trains. Fast trains. Trains that get people off of highways and from San Diego to Sacramento in three hours. Trains that make a spooky wooshing sound as they glide on magnetic tracks to the station. Failing that, I’m in favor of trains that can automatically brake if they’re in danger of tumping off the rails on tight curves.

Because I want to see trains do well — not killing people, obviating the need for cars, getting me to where I need to go at airplane speeds — I'm inclined to cotton to pro-train arguments. This is because I'm a human being and I love to feel as if the opinions I already hold are the correct ones. Reinforcing my existing beliefs makes clicking around the Internet an exercise in smug self-congratulation.

But when Think Progress stepped into the aftermath of the Amtrak derailment last week to shame Congress for skinflinting the trains, I was stuck looking at this supposed proof that Congress has epically jobbed American train-takers. Again: I agree with the premise and conclusion of this argument! Life, on balance, would improve if we invested in rad train technology. Three hundred-some retweets suggest I’m not the only one who feels this way. So why, for the love of Cartesian coordinates, are we lying with lines here?

Eyeball these graphs. Don't look at the numbers yet — just check out the ratios of those inclines and line segment lengths. What’s that leap we’ve taken in 14 years there, in ridership. How much higher is that end square from the first. Ten times higher? Fifteen? Clearly Amtrak ridership has gone from Joe Biden, the occasional acrophobe, well-heeled hobos, and the occasional wander-in to a circuslike proportions, Tokyo subway, uptown 6 at rush hour, rear cars of Snowpiercer ridership. From driverless ghost train to a Mumbai commute. This is a fact that looks insane, and so probably is, but go ahead and retweet ASAP!

In fact the gain is more along the lines of, what, 50 percent, as the Y-axis scale explains. The difference between a 50 percent gain and an apparent 10-fold (1,000 percent) gain is an exaggeration of roughly 20. That seven-year crater job in federal funding, too — that’s, what, a 20 percent decline illustrated to look like 80 percent or so? Humans use drawings to communicate quantities because they put huge, abstract numbers into a scale we can interpret. Intuitively a sloped line makes quicker sense than a series of eight-digit figures or a list of year-over-year percentage changes. The easy uptake of a sloped line, however, makes it a really great delivery system for bullshit. And these lines are bullshit.

What would non-bullshit lines look like? Someone quick on the draw with graphing software offered this example. It describes the same phenomenon, but tells a different story. In this version, the world swings not on cataclysmic shifts in money and human behavior, but on smaller increments and little details wherein national debates may take on a shade more nuance.