BROOKE GLADSTONE: Most numbers, however shoddy, have at the very least an original source. There are some, though, that appear to lack even that. Four years ago, we delved into the mysterious number, said to be 50,000, of child predators online at any given time. It was cited by the NBC Dateline program To Catch a Predator and also by then Attorney General Alberto Gonzales. But spokespersons for the FBI, the National Center for Missing and Exploited Children and the Crimes against Children Research Center said it was not based on any research they were aware of. The A.G.’s office at the time, well, they said it came from Dateline.

CHRIS HANSEN:

It was attributed to, you know, law enforcement, as an estimate, and it was talked about as sort of an extrapolated number.

BROOKE GLADSTONE: Chris Hansen is a reporter for the Dateline series.

CHRIS HANSEN: So when we went to interview Ken Lanning, who was the expert we talked to in our first piece, I said, look, this number keeps surfacing, do you think that it's accurate, it's reliable? And he essentially said to me, I've heard it but depending on how you define what is a predator, it could actually be a very low estimate.

BROOKE GLADSTONE: He took that as confirmation, but maybe he shouldn't have.

KEN LANNING: I didn't know where it came from. I couldn't confirm it, but I couldn't refute it either, but I felt it was a fairly reasonable figure.

BROOKE GLADSTONE:

That's Hansen's source, FBI veteran Ken Lanning.

KEN LANNING: I was somewhat curious about the fact [CHUCKLES] that it was 50,000. That number had popped in the past, because I had been an FBI agent for over 30 years. In the early 1980s, this was the number that was most often used to estimate how many children were kidnapped or abducted by strangers every year. But the research that was done in the early 1990s found that somewhere in the neighborhood of 2 to 300 children every year were abducted in this manner.

BROOKE GLADSTONE: It seems bad things don't come in threes – they come in 50,000.

KEN LANNING: The other one that I specifically [LAUGHS] remembered kind of came in the late '80s, where there were a lot of people who were talking about satanic cults that were supposedly running around the country engaging in human sacrifices. And when you'd try to say, well, how much of this is going on, once again, [LAUGHS] the same number popped up, 50,000 a year.

BROOKE GLADSTONE: Sacrificed?

KEN LANNING: Yes. That's what they were alleging. [LAUGHS] This one here was a little bit more obviously problematic to me, because we do have good data on homicide. And at that time, there was somewhere in the neighborhood of 20 to 23,000 murders every year, so this meant that the satanists all by themselves were killing twice as many [LAUGHING] people as all the other murderers combined.

BROOKE GLADSTONE: So why is 50,000 such an unaccountably sticky figure?

KEN LANNING: Maybe the appeal of the number was that it wasn't a real small number. It wasn't like 100, 200. And it wasn't a ridiculously large number, like 10 million. It was like a Goldilocks number, not too hot, not too cold.

BROOKE GLADSTONE: Carl Bialik tracks down dubious numbers in his column for The Wall Street Journal.

CARL BIALIK: An interesting phenomenon of these numbers is that they'll often be cited to an agency or some government body, and then a study will pick it up, and then the press will repeat it from that study. And then once it appears in the press, public officials will repeat it again, and now it's become an official number.

BROOKE GLADSTONE: Steve Ross, a former professor at Columbia University's School of Journalism, who taught a class on reporting numbers, says journalists have their own reasons for ornamenting their stories with digits.

STEVE ROSS: Look, 30, 40 years ago, ever since I've been in the business, the uh, editor will come down to you and say, add a number, it builds credibility – got to have a number in there.

BROOKE GLADSTONE: But when it comes to crime, a good number is hard to find.

STEVE ROSS: The only reasonably accurate national crime statistics come out of something called the Uniform Crime Report. The Uniform Crime Report only tracks eight different crimes – rape, murder, auto theft, that sort of thing. If it's not a crime that is tracked – child pornography is not tracked, for instance – there is no hard and fast national number that comes out of that. At the very best, it's a number that's extrapolated from a more limited survey.

BROOKE GLADSTONE: Dateline reporter Chris Hansen.

CHRIS HANSEN: There's a natural tendency to try to quantify a problem. I think we all do it.

BROOKE GLADSTONE: So you think you'll continue to use that number?

CHRIS HANSEN: We used it in the first two stories, and we haven't used it in the last three.

BROOKE GLADSTONE: But in terms of changing the previous stories - or I know that they may be rebroadcast at MSNBC - you'll just leave them as they stand.

CHRIS HANSEN: Well, I don't, I don’t think that decision's been made yet.

CARL BIALIK:

The strongest form of media bias is probably a reporter's bias for his or her own story.

BROOKE GLADSTONE: Wall Street Journal numbers guy, Carl Bialik.

CARL BIALIK: And when you find a number that backs up the thesis you've adopted for your story, it can be really hard to pass it up.

BROOKE GLADSTONE: Add to that the mysterious allure of 50,000, applied to mayhem ranging from Korean War casualties to the annual death toll from second-hand smoke. I tossed it to former journalism prof, Steve Ross. If I were to throw that number out to you as if it were a Jeopardy answer, what might you guess the question would be?

STEVE ROSS: How many traffic deaths are there in the United States every year?

BROOKE GLADSTONE: How about how many people are killed a year because of uh, satanic human sacrifice?

STEVE ROSS: I would doubt that it is very many. [LAUGHTER] But I've heard 50,000.

[LAUGHTER]

BROOKE GLADSTONE: The FBI's Fanning says there may well be 50,000 sexual predators trawling the net. He doesn't know. He does know that 50,000 is the Goldilocks of crime stats. But, as experience shows, that doesn't mean it's just right.