A new book by journalist Charles Seife examines how figures and statistics are poorly used in public debates. I like his idea of “proofiness” which seems similar to the concept of “truthiness.” Here are some of the types of bad statistics he points out:
Falsifying numbers is the crudest form of proofiness. Seife lays out a rogues’ gallery of more subtle deceptions. “Potemkin numbers” are phony statistics based on erroneous or nonexistent calculations. Justice Antonin Scalia’s assertion that only 0.027 percent of convicted felons are wrongly imprisoned was a Potemkin number derived from a prosecutor’s back-of-the-envelope estimate; more careful studies suggest the rate might be between 3 and 5 percent.
“Disestimation” involves ascribing too much meaning to a measurement, relative to the uncertainties and errors inherent in it. In the most provocative and detailed part of the book, Seife analyzes the recounting process in the astonishingly close 2008 Minnesota Senate race between Norm Coleman and Al Franken. The winner, he claims, should have been decided by a coin flip; anything else is disestimation, considering that the observed errors in counting the votes were always much larger than the number of votes (roughly 200 to 300) separating the two candidates.
“Comparing apples and oranges” is another perennial favorite. The conservative Blue Dog Democrats indulged in it when they accused the Bush administration of borrowing more money from foreign governments in four years than had all the previous administrations in our nation’s history, combined. True enough, but only if one conveniently forgets to correct for inflation.
Books like these are needed in our society as politicians often debate through numbers. Without a proper understanding of who is using these numbers, where they come from, and what they mean, the public will have difficulty understanding what is going on. (And this may be the aim of politicians.)
(Based on this review, his arguments and concepts seem similar to those of sociologist Joel Best.)