Before Charles Seife was a journalist, he studied mathematics. In Proofiness: The Dark Arts of Mathematical Deception, the New York University professor and science writer studies our faith in numbers and how it can be manipulated. He chatted recently with U.S. News about why numbers often lie. Excerpts:
Why write about lying statistics?
When I studied mathematics, I also read a lot of science journalism. It seemed to me that there were lots of errors. In 1992, for instance, there was a study that came out in the journal Nature that was incredibly wrong. It said that female runners would outpace male runners by 1998. The whole study was based on a formula that, if extended a bit further, showed that female runners would eventually run faster than the speed of sound. Over the years, I began collecting stories like these. In time, it was clear that a sinister pattern was emerging. These stories were deliberate, meant to get attention or convince people of something incorrect. There are similar studies that say that natural blonds will be extinct in 200 years or that a Hummer is more environmentally friendly than a Prius.
Yes, the idea that you can use the language of mathematics to convince people something is true even when it is not.
A short time ago, I went out to Minnesota to watch a recount of the Senate race between Norm Coleman and Al Franken. Just like the 2000 election in Florida, it was extremely close, eventually decided by 312 votes out of millions. From a mathematician's point of view, the margin between the two candidates was much, much, much smaller than can be measured, given the uncertainties of counting. From a mathematician's point of view, it was a tie.
But it was lawyers doing the counting.
And from a lawyer's point of view, their side had won.
Are people too trusting of numbers?
When people present numbers, there is a presumption that the numbers are correct. Numbers may be the purest representation of truth that we have, but they are not pure. A person's height, for instance, is fairly reliable. But counting the number of people in the state is not accurate.
Yet, we do so regularly with the census.
When the Census Bureau goes out to count people, it is fairly effective through mailed surveys, door-to-door counters, and all the rest. But no matter how hard you try, you are going to miss people—those who move around, those who avoid contact with the government, illegals. In the late 1990s, the Census Bureau moved to change the way the count is done to incorporate statistical models. The Republicans sued to prevent this, because those who are not counted tend to lean toward the Democrats. If it had been the other way around, I'm sure the Democrats would have sued.
But it's not just domestically. Apparently, proofiness also works on the world stage.
In the late 1970s, we were negotiating with the U.S.S.R. to stop all nuclear tests. The Reagan administration opposed the treaty but couldn't pull out unilaterally because it would have made the U.S. look hawkish. So, they ginned up a violation to accuse the Soviet Union. But the accusation of cheating was based on a calculation that the administration knew was wrong. As a result, the test ban treaty failed. Obama has even talked about bringing up the treaty again, but it is so far outside the realm of possibility because of one little mathematical lie.
What's wrong with political polling?
Polling is the largest manufacturing method for phony numbers known to man. Polls are also pseudo events because you actually create the news by asking people their opinions. We can commission an event, then report on it. Exit polls, for instance, give journalists something to talk about in the five hours before all the votes are counted. Journalists like to think that they're not in the business of selling a candidate or product, but when journalists create polls, they're putting ourselves in the same group.