Fulbright in Portugal

Steve Doig in Lisbon August-December 2010

Lisbon skyline

Counting heads

It’s not all porto and pastéis de nata for me here in Portugal. I spent yesterday doing journalism.

I was contacted a couple of days earlier by Curt Westergard, whose Airphotoslive.com company uses cameras on tethered balloons to produce high-resolution aerial photos. He had been hired by CBS News to get images of the crowd that gathered Saturday for the Glenn Beck “Restoring Honor” Tea Party rally at the Lincoln Memorial in Washington. CBS also wanted a credible estimate of the size of the crowd. That’s where I came in.

I started doing crowd estimates back in my Miami Herald days, for local events such as a visit by Pope John Paul II or the annual Calle Ocho street festival. After the 2008 presidential election, I wrote a piece about crowd counting for MSNBC.com, and that led to news media requests for an estimate of the crowd for President Obama’s inauguration. I used a satellite image taken over the National Mall on January 20, 2009, to estimate the crowd there at about 800,000.

Crowd counting, particularly of political events, always is controversial. The organizers of the event inevitably hype their crowd estimate — often grossly — to demonstrate the popularity of their cause, and opponents inevitably underestimate to fit their own agenda. Because of the wild pre-inauguration predictions of how many would attend in person — up to 5 million! — my reality-based estimate was ignored by many left-wing commentators and embraced by those on the right.

Naturally, I expected more of the same about my Beck rally estimate. To calculate it, I used Airphotoslive.com’s very striking images (MUCH larger than the thumbnail posted here)

Airphotoslive.com shot for CBS News

to make density estimates across different zones of the crowd; a variety of ground-level images from news photographers and attendees who posted their photos on Flickr; and Google Earth to measure the square footage of the different zones. Yes, I included the crowd areas under the trees; the full-size Airphotolive images were detailed enough to discern the edges of the crowd even there.

My estimate is that about 80,000 people were at the rally. Ryan Shuler, an Airphotoslive image analyst, used the same images and a different grid-density method to produce an estimate of 87,000. Considering the error margins around our separately-calculated estimates, they are statistically identical. CBS went with the 87,000 figure, which I certainly can accept.

Now the fun begins in the blogosphere. NBC News, the New York Times, and other large media outlets that didn’t attempt a scientific estimate uncritically accepted Beck’s claim of “300,000 to 500,000”. (At least Rep. Michelle Bachmann’s laughable claim of 1 million isn’t being treated seriously.)

The CBS News estimate immediately was vilified by conservative bloggers, and often rabidly-virulent comments from readers are being posted on news stories that mention the CBS estimate. I won’t post more links, but you can find plenty by Googling “beck rally attendance” and similar search terms.

The frothing underscores the problem with hyped predictions of crowd size. Organizers and supporters are forced to insist loudly that the actual crowd met or exceeded their expectations, for fear that the realistic estimate will be painted as a disappointment. The time-honored way to dismiss scientific estimates that don’t reflect the pre-event hype is to claim political bias on the part of those doing the estimate. I am amused to see that those who embraced my Obama inauguration estimate as soberly realistic are now attacking the Beck rally estimate, produced using exactly the same methods, as deliberately biased.

I expect that kind of behavior from partisans on both sides. I am disappointed, though, by the many responsible news organizations that failed to produce their own independent estimates and instead reported only ungrounded hype. Their readers and viewers deserve better journalism than that.

[DOIG afternote: I welcome reasoned comment and questions. Don’t bother sending ideological rants from any direction; I trash those. Also, it may take a while for your comment to be read and approved. I’m in Portugal five hours ahead of DC, also the flood of hits on the site has slowed it way down.]

[UPDATE: Those interested in this topic might also read this followup note.]

75 Responses to “Counting heads”

  1. August 31st, 2010 at 12:20 am

    Lois says:

    I would really like to see the photos that were taken at about 12:30 and not the ones taken at 9:30!!!!!the line for the metros were 1 hour long. Metro said they had NEVER seen anything like it. lots of people got there late you could see the crowd thickening at the Washington monument lawn all threw the event. there was a plane flying over at the end of the rally those are the ones that we will probably never see.

    [DOIG reply: The images I used for the estimate were taken right at noon. I don’t know why you think 9:30.]

  2. August 31st, 2010 at 12:31 am

    Lois says:

    I just noticed that the WWII memorial and Washington monument are not in this photo???

    [DOIG reply: That’s because the balloon was tethered about halfway between the Lincoln Monument and the memorial. Airphotoslive also took images back towards the east. Don’t worry; the combined images show the whole crowd.]

  3. August 31st, 2010 at 1:08 am

    Dennis says:

    10 sqft is way too conservative for a tight crowd..

    Even the park service used 2.5 sqft back in the day…which to me is mosh pit like:)

    I used several ariel pics and local shots, and used google earth maps myself, and while I agree there isn’t even enough sqft including the pool itself to support 1 million, let alone 500,000, the crowd was well pack 800 feet out on the south side of pool and in that area alone I estimated 75,000 people…

    Another simple way that I probably shouldn’t share but will is to take a known football stadium, since sitting in a seat easily equates to the normal shoulder room in a crowd, and see how much sqft the total capacity of the seats in the stadium are…
    I used this method to get 875,000 for Obama inauguration, which seems to be on par with everyone else..

  4. August 31st, 2010 at 4:21 am

    Thomas Ray says:

    Gee, looking at all the pics, the crowd looked like a typical gathering of football fans during a Tennessee home game vs. the Gators! No politics here… Go Gators!

  5. August 31st, 2010 at 7:07 am

    Steve Doig says:

    Nice article about the crowd counting controversy by Charlie Petis in the Knight Science Journalism Tracker hosted by MIT. http://ksjtracker.mit.edu/2010/08/30/a-journalism-professor-asks-and-answers-how-many-tea-partiers-on-the-mall-on-saturday-try-80000-whyd-we-read-300000-plus/

  6. August 31st, 2010 at 7:59 am

    Wyn Williams says:

    Amazing the amount of guess-estimates made in the media, good to see so many experts….

    Im not an expert but assume you have a good idea of what you are doing and are a professional, to me that means that although of course your figure is not exact and there may have been more (or less) people then you estimate to suggest that a professional using good images and a proven method could under-estimate the crowd by three or four times is obviously ridiculous.

    Its good to get an open opinion on the attendance with methods used and information on how you did it stated, thankyou.

  7. August 31st, 2010 at 11:29 am

    JTOv says:

    Gotta agree with some of the comments about the Beck rally not being a Tea Party rally. I don’t like Beck or Palin. I personally find them frightening, but I would consider myself exceptionally sympathetic to the Tea Party. Somebody likened it to the Promise Keepers’ rally about a decade ago, which seems like a fair enough comparison.

    Anyway, thanks for explaining your methodology and replying to the comments above. Very interesting, by any measure.

  8. August 31st, 2010 at 4:28 pm

    Jeff says:

    Hi GreginIowa:

    leftcoastrebel.com posted this link to many pictures at ground level for the event


    In these pictures there are easily 3 feet by 3 feet for every person in every picture shown here. Much more in some of them. That means at least 9 square feet per person. Doig’s estimate of 10 square feet per person doesn’t seem to far off in my opinion.

  9. August 31st, 2010 at 4:46 pm

    Craig Aikin says:

    Good job. I am a local (northern VA) and I am surprised that nobody has thought about the fact that it was a beautiful day in D.C., the last big week end of the summer before school starts, several different events occurring at the same time or the number of tourists that would be in the area as a matter of fact. Has anyone taken a head count on a weekend when no events are happening? Just curious.

  10. August 31st, 2010 at 4:48 pm

    Dnanderson says:

    Sorry I just looked it up- 510,000 far and above any day this month (Google Washington metro ridership)

    [DOIG reply: Actually, about 180,000 above the normal ridership for the day. Even if you assume that all of the extras went to the rally, those who rode in pretty assuredly also rode out when it was over. So that’s about 90,000 extra people — not nearly enough to support the claims of >300k.]

  11. August 31st, 2010 at 5:02 pm

    PHLippel says:

    For some independent data, one can ask about how the people got there. Metro reported about 180,000 excess trips for the entire day, relative to an ordinary summer Saturday. There was another rally, and a crowded baseball game that evening, tha may have contributed to the excess. Some people may have gotten there by other means, but Metro was the suggested route, and I haven’t seen any reports of major traffic jams.

    180K trips is pretty reasonable for 70k to 90K attendees(two trips per person minimum, many will have taken additional rides some time during the day). And it’s pretty hard to square 180K metro trips with 500K attendees.

  12. August 31st, 2010 at 6:03 pm

    Carole Joyce says:

    I can attest to the high density of the crowd. We were staked out near the fence around the Reflecting Pool. My cousin was affected by the heat and laid down. My daughter
    then had to stand over her body.

    We refrained from drinking much because the distance to the potties-not far- was a laborious trip. You had to search for every spot to take a step and I feared falling on sleeping babies or starting a domino like collapse of the multitude.

    We decided going back to our original spot was too arduous and ended up far away where we could no longer hear, but had some shade and breathing space.

    [DOIG reply: Yes, I can certainly believe what you are describing in that area around the Pool; I could see that in the images. But the fact that you were able to walk, albeit carefully, in that area shows that it was about the 10 sq ft per person density that I used there. If the mosh-pit densities that some are insisting on in fact were there, then your cousin wouldn’t have been able to lie down.]

  13. August 31st, 2010 at 9:24 pm

    Tsu Dho Nimh says:

    Interesting … I worked backwards, from the area occupied by the crowd and the figures for typical crowd densities that Professor Jacobs worked out. To help you visualize this, imagine the crowd densities in a 2-car garage of 400 square feet. A loose crowd is 40 people, a dense crowd is 89 people, and a mosh pit crowd is 160 people in the garage.

    Given the acreage of the area the crowd was occupying (as seen in Beck’s releases photos, flickr, etc), I calculated possible attendance numbers: “If the entire 38-acre rally area was completely full of people standing at a uniform density, loose crowd density means 165,500 people (4356 x 38) were there. A dense crowd would be 367,840 people (9,680 x 38). Jammed into mosh-pit togetherness, you could squeeze 662,112 people into the rally area (17,424 x 38).

    If Bachman’s million people had been there, each attendee would have had 1.7 square feet of standing room. That’s the equivalent of stuffing 235 people into a 2-car garage, which common sense will tell you isn’t going to work.”


  14. August 31st, 2010 at 9:45 pm

    Tsu Dho Nimh says:

    OOPS: Forgot the footage estimates: “Loose crowd needs 10 square feet per person. A more tightly packed crowd fills 4.5 square feet per person. A truly scary mob of mosh-pit density would get about 2.5 square feet per person.”

    That’s with people standing up. “Festival” or “Picnic” seating takes more (how much more would be a good study for someone’s thesis), and lawn chairs would also take more.

  15. September 1st, 2010 at 1:35 am

    NewDaddyDave says:

    Guys, this company may be right. I don’t really care, the mass of humanity was huge be it 80,000 or 200,000 or more.

    But if we are to take this company seriously they should post ALL the full size images captured and let other experts evaluate the time of day the shots were taken, crowd density, etc.

    Heck this could be the best advertisement this company could EVER get. Or… if frauds, they would be exposed.

  16. September 1st, 2010 at 1:43 am

    Jeff Bell says:

    One way to estimate crowd size is to compare it to other large crowd events that document their seating capacity. Having lived in Indianapolis IN and attended several Indy 500 races over my 50 years (plus being at the 8/28 event) my sense was the event attendance was equal to or greater than an Indy 500 race. IMS has fixed seating at 257,000 with most grandstands between 25-45 rows deep. Infield seating raises capacity to 400,000. At 8/28 it was nearly impossible to move through the crowds in the grass as most were shoulder to shoulder vying for the best clear view of the stage or the jumbotrons. At an Indy 500 you have plenty of room in between rows to enter and exit your seat. See my graphic at http://www.askjeffbell.com/images/HonorAttendance4.jpg for a rough comparison of the 2 venues and attendance capacity.

  17. September 1st, 2010 at 1:53 am

    NewDaddyDave says:

    We did go to the rally. We got there about 11:20, we wound up sitting about 10 yards in front of the Washington Memorial. I saw the balloon going up to take photos twice over the field off to the side. But I can’t remember the exact time it went up.

    When we tried to go about 100 yards or so in front of us it was way too tight for comfort, kids playing accidentially hitting my wallet (pet peeve of mine), bumping into others just trying to move around, I took my family back up close to the monument where we had pleanty of room to spread a blanket, and actually feel a bit of a breeze.

    I hope the publish the full size images, and the claims will easily be verified or proven false. If the company is credibile they could offer their images to the public. Heck for the attendees like myself I’ld buy high res images of my family in attendance of this event. He could probably make millions on just that alone if he sold them for $50 for a poster size image for a given location and time.

    Attendees and people who wish they could have attended would buy these full size images.

  18. September 1st, 2010 at 5:33 am

    Mekhong Kurt says:

    This was fascinating, especially the explanation of your methodology. I have no expertise in estimating crowds from aerial photos, but I did work in security for several years, and on occasion was involved in trying to get some sort of rough, inside-a-reasonable-range guesstimate of crowd size, and your estimate seems far more reasonable than one much higher (or much lower, had there been any of those). I know your estimate must disappoint a great many actual attendees (as opposed to tourists, people there for other reasons, etc.), but even Beck’s estimate of 300k-500k is way out of reasonable range, while the 1 million is downright wild (and laughable).

    Regarding the objections to calling this event a “Tea Party” one, well, then someone needs to tell those I’ve read in scores of comment threads who self-identify as Tea Partiers and insist that virtually *everyone* there was a Tea Partier. I must have seen at least 180-200 such claims just so far. And there’s the fact that Beck and Palin (especially Palin) are closely linked with the Tea Party, so it’s reasonable to assume a sizable number Tea Partiers would show up I strongly doubt those claims — I imagine some fair number of onlookers were there to gawk, not to support Beck and/or Palin.

    In any case, I enjoyed this piece and most of the comments.

  19. September 1st, 2010 at 6:18 am

    Fulbright in Portugal » Blog Archive » Crowd counting (wrapup) says:

    […] About me « Counting heads […]

  20. September 22nd, 2010 at 2:58 pm

    Ken J says:

    I hope you also are called on to estimate the crowd size at the upcoming rally by Stewart and Colbert. The comparison of the two different crowds will certainly fuel a lot of discussions in the future.

  21. October 30th, 2010 at 6:44 pm

    Patty Jones says:

    I, too, hope you will count heads for the Rally to Restore Sanity, happening right now in Washington D.C.

  22. October 31st, 2010 at 5:47 pm

    Brian says:

    I was debating how many people were at glen becks rally with one of my family members who had actually gone to the event. When I showed her this website, she told me that the picture shown here must have been taken very early in the day. She actually went to the event and stood in the upper left area of this picture. She remembers it being crowded but this picture shows a sparse area. Do you happen to know what time the pictures were taken that you used for your estimates?

  23. November 1st, 2010 at 6:03 pm

    Steve Doig says:

    As noted in my posting, the shots were taken right around noon, at the height of the rally.

  24. November 2nd, 2010 at 4:19 am

    Jay says:

    I just looked at photos from the Beck Rally at http://www.leftcoastrebel.com/2010/08/photos-aerial-pictures-of-glenn-beck.html to compare with my experince from the midst of the Stewart/Colbert Rally.

    We were much more densely packed than the people in the Beck crowd photos were. I can see space between people here; some folks are standing with their arms akimbo; I see people seated in camp chairs or with their gear arrayed around them on the ground. There was no way we could have done any of this at the Oct. 30th Rally. I literally could NOT move in any direction as we were shoulder to shoulder where I was standing – mid-Mall at 7th Street. I never felt like I was about to be crushed, but was glad that, as knocked about as I was, I never fell beneath all those feet. (Don’t get me wrong; the knocking about was from the domino effect of people trying to move when they couldn’t – not from discourtesy.)

    So the density of the crowd was quite tight down the center of the Mall. I also know that, due to the crowd pressure, a number of people left the Rally after Stewart & Colbert initially spoke and likewise, due to complexity of transporting so many people on buses, a bunch didn’t make it to Washington until well after 1P. (For those who could move), it was a crowd in flux.

    Even though the US Park Police no longer releases “official” crowd numbers, surely they must still be calculating them as a practical management tool?

  25. November 2nd, 2010 at 10:34 pm

    deathbymedia says:

    Thank you professor for your thorough calculus. For the last few years, it seems like people wish to go with their gut feeling to estimate anything and everything. Perhaps it is time to bring back scientific methodology back in to the equation.

Leave a Reply