Frankie Flack: Lies, damn lies, statistics and business journalism
If you’re a business reporter and don’t love numbers, I don’t want anything to do with you.
I love selling stories with facts and figures a lot more than I like selling sizzle. Show me market share or same-store sales or margins data, and I’ll go to town. In contrast, today’s great profile of a brash young CEO might be tomorrow’s cautionary tale. But the numbers, they don’t lie.
(This isn’t, technically, true. Number lie all the time. They also mislead. But — and this is what I love about numbers — they can be fact checked, vetted, fisked, taken apart, put back together. You want to question my assumptions? Bring it on.)
That said, “data-driven journalism” has now officially jumped the shark. Election-predicting Nate Silver has re-launched fivethirtyeight.com for ESPN, but he’s not content to just crunch baseball stats or election polls. Nope. His site is going to bring a data-driven approach to everything, no matter how silly.
I mean, the site has been up for only a few days, and two things have already happened. First, reporters are now claiming that statistics can solve any problem. That missing airliner? If only the nerd-kings were in charge! They’d apply Bayesian statistics and — voila — have a much better idea of where the plane went down. Never mind that Bayesian statistics don’t work well in an information vacuum. Just wave that magic wand!
Second, reporters are now violating most of the primary rules of working with numbers, such as the general principle that the bigger the dataset, the better the results. Yet 24 hours after the site launched, the “lead writer for news” used this phrase: “My experiment had a sample size of one.” That is not a sentence that inspires confidence.
And it’s not just fivethirtyeight. The New York Times is launching “The Upshot” to do essentially the same thing. Ezra Klein has a similar belief in the edifying effect of charts and data. Some of this stuff will be really good. But a ton of it will be horseshit, dressed up as science.
This is going to boomerang back to haunt you business reporters. You heard it here first: It’s going to take about six weeks for the PR brain trust to decide that we need less pitches with infographics (thank God) and more datasets that we can dump on unsuspecting reporters.
You all will have to start double-checking my numbers because no one wants to let a juicy data story slip away. And we’ll all be worse off, spending our days hunched over Excel spreadsheets, looking for trickery.
So here’s my offer, and we need to agree on this quickly before things spin even more out of control: on behalf of all flacks, we will not pivot to sending you crappy “data” stories if you all agree not to hop on the bandwagon of dressing up garbage numbers as some sort of absolute truth.
That doesn’t mean I’m abandoning numbers. Far from it.
I’m still going to push for GAAP earnings reports and sensible information on marketing trends or whatever to send you.
But I won’t claim that Bayesian statistics are the key to understanding the frozen concentrated orange-juice futures if you don’t publish “experiments with a sample size of one.”