Counting Crowds: Results May Vary
IRA FLATOW, host:
You're listening to SCIENCE FRIDAY. I'm Ira Flatow.
When Jon Stewart held his rally in D.C., he said that the success of the event would be measured by how many people showed up, and that seems true. We always watch for the official count. But just how do they count the crowd? For example, how many people attended the Million Man March? It depends on whom you ask. According to the U.S. Park Police, about 400,000. But the organizers of the march took issue with that number and asked for a recount. And using different images, a different image, a crowd counting expert at Boston University estimated the crowd to be closer to 800,000 - almost twice the number that the Park Service had. So what followed? The Park Service stopped publicizing its crowd counts and did not give them out anymore.
But the controversy over the headcount didn't end or start there. According to my next guests, nearly every major protest or rally sparks a similar dispute over the numbers. There were disagreements over how many people attended President Obama's inauguration, Glenn Beck's rally at the Mall.
How do you count a crowd? Is there a scientific way to do it? Well, it's not as easy you think, according to my guests.
Clark McPhail is professor emeritus of sociology at the University of Illinois at Urbana-Champaign.
Thanks for joining us today.
Professor CLARK McPHAIL (Sociology, University of Illinois at Urbana-Champaign): Thank you.
FLATOW: You're welcome. Stephen K. Doig is Knight Chair in Journalism at the Cronkite School of Journalism at Arizona State University in Tempe. He's also part of a Pulitzer Prize-winning group in the Miami Herald for their 1993 coverage of Hurricane Andrew.
Welcome to SCIENCE FRIDAY.
Professor STEPHEN K. DOIG (Knight Chair in Journalism, Cronkite School of Journalism, Arizona State University): Thank you, Ira.
FLATOW: Clark McPhail, is there a good way to count the number of people in the crowd? I mean, why is it so hard that we get so many different numbers for these things?
Prof. McPHAIL: Part of it is a matter of the perspective of the organizers who are frequently looking at the rally from only a very slightly elevated vantage point on a rally platform, and right in front of them, the gathering tends to be very, very dense. And then if they look down through the depth of the gathering, toward the end, they're tricked by something that's called foreshortened perspective. It appears that the density of the crowd immediately in front of them extends throughout the length of the gathering.
And so what people have tried to do to overcome that is to get an overhead or an aerial, an elevated vantage point from which to look at the gathering. And there are three kinds of variables. You need to know how much - what the carrying capacity of the space is in which the gathering is arrayed, and what the square footage is. What's the carrying capacity? What proportion of that is occupied? Is it a hundred percent, 75 percent, 50 percent? And then what is the density of the occupation?
And that method was invented, really, by a journalism professor - Steve Doig, I'm sure, is familiar with the story - Herb Jacobs, who had an office overlooking Sproul Plaza during the Free Speech Movement back in the '60s. And he was curious about the size of the gathering in front, and so he invented this procedure. I don't know if we can call it scientific or not, but it is an objective procedure of the three variables that I've mentioned: carrying capacity, the proportion of the space occupied and the density of the occupation.
FLATOW: So, Steve, if you know the square footage of the area and you know how - the size of person, can't you make a pretty good guess at how many people are there?
Prof. DOIG: Right. You can certainly set limits, at least, you know, a rational area in which the crowd size is likely to be. And that's usually the problem with crowd estimation, is in advance of the crowd, the organizers - no matter what the event - all too often come up with sort of hyped projections of what's going to be. And then they wind up having to show that their event was even more popular than they predicted. And so that's why you wind up with the overhyped ones. But, you know, the point of doing what Professor McPhail talks about is to set sort of rational boundaries on what actually could be there, and it is - it's actually a simple math problem, as he points out.
FLATOW: How do you know how much space to allow an individual in those pictures?
Prof. DOIG: What I do when I train students to make observations is to use the front page of newspapers. If you look at - below the fold, that's about one-and-a-quarter square feet. The entire page is about two-and-a-half square feet. Two pages are five square feet, and so on. At the front of any gathering, the density is very high. It can be pretty close to two-and-a-quarter, or even more dense than that. But as you move to the back and to the sides, it sparses out.
So what the Park Service, the U.S. Park Police has done over the years, working with the formulation that Jacobs' developed, they just average out and use five square feet.
Now what Steve has been doing with Kurt Westergaard in the air photo live organization, is to do a little more discretionary estimation of density and, perhaps, he can tell you about that.
Prof. DOIG: Sure. Yeah. It's the advantage you have when you have a nice overhead image of it is you can break the crowd up. You can take the image. You can grid it off. You can make estimates and see the differences that Professor McPhail was talking about, you know, clearly up in the front or in front of JumboTrons or whatever you're getting to get much density. But if you have a grid, you can make reasonable estimates of the density based on the number of square feet per person that are being taken up, and use that to get a, I guess, a more nuanced density, rather than just coming up a single guess of, say, five square feet per person you know.
FLATOW: You know, I was wondering, in this age of spy in the sky satellites, having been up in the air for decades now, and facial recognition, you would think that some people in the military, in the intelligence service have already figured out some way of really getting an accurate account we just don't know about.
Prof. McPHAIL: You're quite correct about that. During the run up into the 2009 inauguration, there were a number of people who were using satellite imageries. And one of the people that I came in contact with, her name is Allison Puccioni. She works now for James Limited. But before she went to work for them, she was in the U.S. military, and she was doing satellite imagery analysis. And so you can get, you know, a very good image from several thousand of feet in the sky. But it doesn't last very long, because the satellite is going over. So you can't really follow the changes across time. But it does a very good job of that.
Prof. DOIG: There were beautiful images of the inauguration, there was a GeoEye satellite going over about 45 minutes before. And it took a very you know, there was a fascinating picture of the crowd. You could really see that, the (unintelligible) density across there because of it.
FLATOW: Mm-hmm. Well, we'd think that some company would go in the business of crowd measuring.
Prof. McPHAIL: There is one.
(Soundbite of laughter)
FLATOW: I'm always far behind.
Prof. McPHAIL: (Unintelligible) is visually doing it.
FLATOW: Yeah. And they're using would you agree, accurate measurements - making accurate measurements of the crowd?
Prof. DOIG: We got...
Prof. McPHAIL: They're doing their best, I think.
Prof. DOIG: Can you discuss, just a bit, the panoramic digital imagery that's attached to the weather balloon that you use and tell the people a little bit about how high that's tethered above the ground on which the gathering is assembled?
Prof. McPHAIL: I'd tell you what I know about it. My problem is, I'm actually in Portugal and I've never actually seen the actual equipment. So I only know what I've been told. But what normally they do, is they have a balloon. They have equipment that hangs from it. My understanding is that it's going up around 700 feet.
Prof. McPHAIL: D.C. is a difficult area to get aerial images because there are not very many tall buildings around to get good overhead shots and the airspace is very restricted. So the balloon approach is actually a clever solution to that problem. They have to get permits and so on.
Prof. DOIG: But the photographs, they're really very useful.
Prof. McPHAIL: But that's...
FLATOW: Should the Park Service start giving out those numbers again?
Prof. McPHAIL: Yes. Absolutely.
(Soundbite of laughter)
Prof. DOIG: That would be good. I'm sure they're doing I would hope they are continuing to do that internally. You just need...
FLATOW: They have to.
Prof. DOIG: ...need that kind of thing for planning.
Prof. McPHAIL: For budgetary reasons, for planning, for personnel requests and things of that sort. I think there's another flawed way of getting an elevated vantage of point. The Washington Monument is pretty high in the sky. And while you get some foreshortened perspective, at least looking down, you can see the length of the mall and you can determine how many of the panels of the mall are occupied. And if you had always had photographs from the top of the Washington Monument -the Post could do it - and a reported on the ground, walking through the gathering to make estimates of the density, it seems to me that you could have a standardized way of doing that. Having said that - and I have nothing at all against the private enterprise that Kurt Westergaard has developed it seems to me that the First Amendment Rights of assembly and speech really merit an independent, noncommercial estimation that's systematic.
I think that people understand how well that can be done, that you could have a way of showing the magnitude of support, for or against, the various kinds of issues that people assembled to protest or to push forward. So I would like to see the park service do it again. And, certainly, the technology has developed that they, perhaps, could do it better than what was possible before.
FLATOW: Mm-hmm. You agree, Steve?
Prof. DOIG: I agree.
Prof. DOIG: Sure. I think it's good (technical difficulties). I mean, the problem with rallies like this is the size of it becomes a totem, I guess...
Prof. DOIG: ...of how popular the event was...
Prof. DOIG: ...and it's incumbent on the press to provide some sort of independent answer to that question.
FLATOW: If you could get the perfect picture, overhead, no shadows, confusion, face recognition, something, what would be your margin of error? What would you be shooting for?
Prof. McPHAIL: It's - if you're talking margin of error, you're really dealing with sampling. And the best you could say if you were trying, well, there's another approach that can be done. You could grid it off and use something like the Mechanical Turk, which is the Amazon thing where you parcel out a large job into small pieces and basically have a whole bunch of people pay small amounts of money to try and actually count the ebbs in...
Prof. McPHAIL: ...a, let's say, a random selection of squares and then turn that into an estimate. Then you could have a real margin of error. In reality, the kinds of estimates that, for instance, I'm doing with the, you know, with the images, I make a guess and then I basically, doing a margin of error, I'm saying, well, it's probably within, you know, 10 or 15 percent of...
Prof. DOIG: Yup. Right.
Prof. McPHAIL: ...whatever of I'm guessing.
FLATOW: All right. Thank you, gentlemen, for taking time to be with us. I want to thank Clark McPhail, who's professor emeritus of sociology, University of Illinois at Urbana-Champaign. And Stephen Doig, Knight chair in journalism at the Cronkite School of Journalism at Arizona State University in Tempe. Thank you for being with us today.
Prof. DOIG: Thanks. Enjoy (unintelligible).
Prof. McPHAIL: My pleasure.
FLATOW: You're welcome. This is SCIENCE FRIDAY from NPR.
NPR transcripts are created on a rush deadline by Verb8tm, Inc., an NPR contractor, and produced using a proprietary transcription process developed with NPR. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.