Wednesday, June 08, 2016

Biscuit debt: A billion here, a billion there ...

This is a full-time job at the Nation's Newspaper of Record, right? Ensuring that corrections are so baroque that civilians will give up in frustration before they get to the error?

An article on May 13 about the sagging fortunes of the Workers’ Party of Brazil, whose standard-bearer, President Dilma Rousseff, was suspended by the Brazilian Senate, misstated the amount of foreign debt that was hobbling the Brazilian economy in 1998, when Ms. Rousseff’s mentor, Luiz Inácio Lula da Silva, ran for president. It was $250 billion, not $250 million.

In other words, an article May 13 about Brazilian politics misstated the size of the country's foreign debt in 1998, because somebody counted the toes and forgot to divide by 10 or something.

Amid the syntactic thickets, the restatement of the error takes on an interesting form: What the article misstated was "the amount of foreign debt that was hobbling the Brazilian economy in 1998," which is apparently somehow distinguishable from overall foreign debt, or from merely annoying foreign debt, or something along those lines. I think that's a cousin of the "biscuit conditional," resembling the 2+2=22 error: Foreign debt hobbles economies, Brazil had X amount of foreign debt, therefore economy-hobbling debt comes in divisible units.

The Times certainly does give the impression that no error is too slight for a correction:

A film review on Friday about “Teenage Mutant Ninja Turtles: Out of the Shadows” referred incorrectly to the turtles’ names. Three turtles are named for Renaissance artists whose major works included paintings, not four. (Donatello was a sculptor.)

No doubt "Star Wars" fans are relieved that the spotlight is on someone else. But it's fair to wonder a bit about where the Times draws the line between trivial errors of fact and substantive errors of interpretation. It'd be nice someday to see a correction along these lines:

A post on Aug. 4 described the result of a public opinion survey incorrectly. There is no such thing as a "statistical tie," and if there was, a 6-point difference (Clinton 42, Sanders 36) in a subsample of 276 likely Democratic primary voters would not be one.

My favorite point in Margaret Sullivan's farewell post as the Times's ombud was her caution about New York Times exceptionalism: "The idea that whatever The Times does is, by definition, the right thing." Reporting on survey results would be a good place to start replacing conventional Times wisdom with independent judgment.

Labels: ,


Anonymous Thomas said...

I would have said there is no such thing as a "statistical tie", but if there was, this *would* be one.

If the true support was 39% for each of them, with the remainder for 'other or don't know', you'd get a 6% point or greater difference about 25% of the time. (by simulation).

Also 274, not 276, I think.

11:50 PM, June 08, 2016  
Blogger fev said...

I think we're approaching the question from different ends of the tunnel. I'm leaving the confidence level at one SE, rather than turning it up to 95% -- not good for social science, but better than an even bet on the sample values should you be heading to the casino.

Do continue commenting, and observations on polling practice in your neighborhood would always be welcome too.

7:28 PM, June 10, 2016  

Post a Comment

<< Home