A Journalistic Math Problem, Times Two

From Dan Seligman:

I need to talk back to Rob Walker…[who] was responding to my recent Forbes article on innumeracy in the media and in the process heavily hinting that I had swiped one of my best examples from statistics professor John Allen Paulos without giving him credit. Both Paulos and I had pointed derisively to a William Safire column that offered up odds on assorted Democratic politicians winning the 2004 nomination for the Presidency. And both of us made the point that the probabilities represented by Safire's odds summed to 168 percent, which is mathematically impossible. (You cannot go above 100 percent in dealing with mutually exclusive possibilities.) Walker says correctly that Paulos was the first to make this point, and encourages Slate readers to think that in copying his piece I was "showing off quantitative prowess" I didn't really have.

I have two grievances. One is that, in the course of fostering the notion that I'd copied from Paulos, the Slate article develops at some length the proposition that his writings appear all over the place, including Forbes and the New York Times, and he is a major figure in "American public discourse generally." But the article nowhere mentioned that his piece on Safire did not appear in print, or on radio, or on television, and only readers who clicked on the "link" discovered that it had appeared in abcNews.com a site not visited by numerous people, including me. Before I did the Forbes article, I performed a Nexis search on all my innumeracy examples, in an effort to be sure I wasn't repeating what others had said. But Nexis doesn't cover abcNews.com.

My other grievance is that is that I actually have some quantitative skills, and have been writing about odds and statistics longer than John Allen Paulos. The stuff in question has appeared in Forbes, Fortune, the Wall Street Journal and Slate (which had assigned me to explain why the bookies took such a clobbering on the 1996 Tyson Holyfield fight). I have also been analyzing amateurs' betting lines for many years and claim I was the first to point out Richard Nixon's gaffe in 1967, when (according to Theodore H. White's The Making of the President: 1968) he set the odds on the Republic nomination at 50 percent for George Romney, 33 1/3 percent for himself, another 33 1/3 percent for Chuck Percy, and 20 percent for Ronald Reagan. It comes to 136.7 percent.

--Dan Seligman

(To find or answer this post, click here .)


From John Allen Paulos:

My little Safire piece seems to have attracted more attention than I would have thought. For the record, I don't think Safire is innumerate in the least. Although I intended my piece to be somewhat whimsical, I wanted to at least hint at three serious points: if Safire's assessment was an estimate of probabilities (which it should have been since he is a pundit and not a bookie), then the probabilities are wrong; even if, as is arguable, he was writing as a bookie, the profit margin was excessive; the relationship between odds and probability can be misleading (especially in medical and crime reporting).

--John Allen Paulos

(To find or answer this post, click here .)

(1/10)

Notes From The Fray Editor:

There is some disagreement in the Fray on whether Safire is right or wrong, and whether calculating probabilities and handicapping are different things. Although the Fray team did study Math (and indeed once got a journalism job partly on the strength of spotting a mathematical mistake in a piece), we're saying nothing: we're certainly not rating our input above that of someone called Dutch Schultz. [Added, 1/10: Not going to argue with John Allen Paulos, either--see above.]

Comments From The Fray:

What Forbes, John Allen Paulos and Moneybox miss in their mathematical criticism of Safire's column is that odds-makers have no obligation to establish probabilities equaling 100%. Indeed, if they did, there would be no way for the house or track to make any profit. The Morning Line odds at a track generally total 125% and I calculate the odds from the 2001 Kentucky Derby, from www.derbypost.com , to add up to about 127%.

Sure, Safire's running a pretty steep game with 168%, but this doesn't necessarily show that he lacks the math skills.

--David L. Duncan

(To find or answer this post, click here.)

You are correct that the odds of a morning line on a horserace do not add up to 100%. However, this is because this is a betting, or "action" line, intended to get even pari-mutuels, it is not a prediction or assessment of the real odds of any of the horses. The final odds of any horserace will add up to 100%, and Safire's odds should have also.

--Dutch Schultz

(To find or answer this post, click here.)

Safire was talking about odds of a political nomination, a process in which the field is not fixed and the bettors' commitments are not set in stone once the race has begun. In the Kentucky Derby, jockeys don't run out of funding or give up halfway through the race, exhorting their supporters to get behind one of the other horses. Somebody smarter than me would have to put that into mathematical terminology, though.

--Ex-fed

(To find or answer this post, click here.)

(1/9)