Proofiness - Charles Seife [22]
For example, in 2007, an editorial in the Wall Street Journal slammed the United States for having such a high corporate tax rate—paradoxically, the editorial argued, cutting the tax rate would increase tax revenues. It’s a counterintuitive position; if you’re a government that wants to collect more money from your citizens, you should raise taxes, not lower them. However, the claim that cutting taxes generates more revenue was a key element of Reaganomics (which George H. W. Bush once called “voodoo economics”) and has since been adopted by many conservatives. The idea can be depicted graphically by a mountain-shaped Laffer curve, named after economist Arthur Laffer: as you increase taxes, revenue rises, hits a maximum, and then falls to nothing.
When plotting tax rates versus revenues, the Wall Street Journal editorial board clearly saw a Laffer curve in the data.
Figure 8. A laugher curve.
It’s obvious that the Wall Street Journal would have drawn a leprechaun in the dots if that’s what Reaganomics had predicted. There isn’t any justification for drawing a Laffer-like curve that rises and falls within this particular data set; if anything, a straight line going up and to the right fits the data best. But by zipping the curve through an outlier—Norway—the Journal used randumbness to make Arthur Laffer look like a prophet even as the data made him look like a fool.
There’s danger of proofiness even when data do seem to fall neatly along lines or curves, even when there seems to be a pattern nestled in the numbers. Just because a statistician or an economist or a scientist has discovered a real, bona fide relationship between sets of data doesn’t mean that that relationship has any meaning. A line or curve on a graph, equation, or formula can represent a tight relationship among a vast amount of data—yet it might have no value whatsoever.
A good example of this is a 2004 Nature paper written by a motley collection of zoologists, geographers, and public health experts. This illustrious group of scientists analyzed athletes’ Olympic performance on the 100-meter dash over the years and found some striking patterns. Male sprinters were getting faster and faster over the years; their times on the 100-meter dash were decreasing so steadily that you could draw a straight line through the data (which the researchers did). Female sprinters were also getting faster in a similar manner, also explained nicely by a line.
These graphs seemed to explain the data beautifully; the data never strayed far from the lines, so the researchers expressed high confidence that the lines described how men and women perform on the 100-meter dash, even far into the future. And if you follow those lines out, you see that they cross—women match and then surpass men—around the year 2156. The conclusion: women sprinters will be faster than men sometime in the middle of the next century.23 After all, the lines fit the data, and when the lines cross, women outrace men.
Figure 9. A graph that shows women outrunning men after 2156.
However, those lines hide an absurdity. Follow them out for a while and the silliness becomes apparent. By the year 2224 or so, according to those lines, women will be running the 100-meter dash in seven seconds—a speed of roughly 32 miles per hour. Still plausible . . . barely. But the line keeps going. If you were to keep following it, you’d see that 150 years later, women would be sprinting at about 60 miles per hour. In roughly the year 2600, they would break the sound barrier. Shortly thereafter, they’d break the speed of light and travel backward in time, winning races before they actually begin. It’s impossible that those lines truly represent what is going on in the data—they’re faulty