Fulbright in Portugal

Steve Doig in Lisbon August-December 2010

Lisbon skyline

Head-counting in Lisbon

I got a chance to do journalism here in Lisbon today. The Frente Comum de Sindicatos (Joint Trade Union Front) had organized a rally of government workers for Saturday afternoon. The rally was to protest planned cutbacks in salaries and benefits of public employees as part of austerity moves designed to bring Portugal’s budget into better balance. Not surprisingly, public employees complain that they are being asked to make the most sacrifices, and so the rally was designed as a show of solidarity in advance of a countrywide general strike planned for Nov. 24.

In my class at UNL on Thursday, we discussed crowd counting and particularly the obligation of journalists to do independent estimates of the size of political rallies. The problem with such rallies, of course, is that the organizers and the opponents often make wildly inflated or deflated claims about the size of the crowd. Knowing that the rally would be happening today, we decided to see if we could produce such an independent count of the size of the crowd. We did: The answer is below.

To do this estimate, student volunteers stationed themselves along the march route, which was down the 1.1 kilometer length of Avenida de Liberdade. It went from Marques de Pombal plaza where the marchers gathered in advance to Restauradores plaza for the rally at the end. The students gathered information about how many marchers were passing during various short time periods, like 30 seconds, and also how long it took the whole march to pass by. (See results below.)

At Restauradores, my UNL colleague Prof. Antonio Granado talked his way into an excellent vantage point above the plaza, a high window in a building overlooking the whole square. In advance, I had measured the area of the plaza in square meters, information that tells us how many people possibly could fit in the square under various densities.

The march began about 3:30, and the last marchers reached the plaza about 5:15. They marched VERY slowly. When the crowd finally gathered, Antonio took some detailed photos showing the extent and density of the protesters who gathered to hear the speeches from the stage that had been erected in advance.

Between Antonio’s photos and the ground-level observations from myself and the students, I estimated that perhaps 8,000 to 10,000 union members had participated in the march, and that about 5,000 of them had gathered in the Restauradores plaza to hear the speeches.

This estimate is well below the claims that some organizers were making about how many would take part — as many as 100,000 or more. That’s the problem with wildly over-estimating attendance in advance; reality-based counts can make it seem like an event was a failure.

In reality, the area in which the rally took place at Restauradores is about 5,000 square feet in size. A loose crowd in which everyone is arm’s length from those around them, a crowd in which it’s possible to move around without squeezing past others, needs about one square meter per person. As Antonio’s pictures and my ground-level observations show, this was a loose crowd. Therefore, the actual gathered crowd was no more than about 5,000 protesters. The estimate of perhaps 10,000 in the march makes the quite-generous assumption that as many as half of the marchers didn’t stay for the rally.

It will be interesting to see if some of the news media outlets in Lisbon and around Portugal pay attention to our estimate or just accept the inflated claims that will be made.

UPDATE: For those who insist our estimate of no more than 10,000 is way too low, consider this. One of my students, Eudora Ribeira , took very careful counts of the marchers during 30-second periods as they passed by her position on Ave. Liberdade. Her counts of marchers passing in those 30-second periods were 51, 40, 75, 52, 45, 20, 70, 52, 58, 70, 51, 49, 46, 28, 17 and 25. The average of those counts is 94 marchers passing per minute. The front of the march reached her at 16:02 and the last marchers passed her at 17:05, a total of 63 minutes. If you multiply 94 marchers per minute by 63, you get a total of 5,922 marchers, or a rounded figure of approximately 6,000. Another student on Ave. Liberdade, Rita Oliveira, also used the same passing-marchers method, and came up with a march estimate that was somewhat lower. By either method we used — counting marchers or measuring the rally crowd area — you can see the crowd could not have been larger than 10,000.

UPDATE: The newspaper Publico posted a long story from the LUSA wire service. Television also is interested. I was interviewed by the SIC (video here at the 8:00-minute mark) and TVI (video here at the 14:30 mark) news stations. And TV station RTP posted a brief version of the LUSA story about our count.

16 Responses to “Head-counting in Lisbon”

  1. November 7th, 2010 at 3:50 am

    Celso Miranda says:

    Both of you (and the volunteer students), made an excellent job. But there is one thing that must be taken into account. How many of you were needed to this kind of head-counting?

  2. November 7th, 2010 at 9:02 am

    Steve Doig says:

    This type of head-counting can be done with just one person, if necessary. The important element for accuracy is a good photo of the crowd taken from above. With that, it is relatively simple to see how far the crowd extends and how dense it is, in terms of square meters per person. Along with the images taken from above by Prof. Granado, I used Google Earth to zoom in on Restauradores and measure the total area of the plaza.

  3. November 7th, 2010 at 12:33 pm

    João Gonçalves says:

    I have to disagree with your counting procedure.
    The average of that count is indeed 94 marchers per minute, for a better statistical meaning also please disclose at least the standard deviation, so that readers can apreciate a “measure of dispersion”. For this dataset you get a SD of 35 people, which is a really high figure. It does not have for sure any statistical significance, therefore you cannot extrapolate.
    On a related subject, if you are doing journalism, expected to be fully independent, please do not use capital letters when trying to highlight more meaning “marched VERY slowly”.
    Of course this is not a journal, its your own webpage, but Im sure that nevertheless you want to be taken seriously…
    Best regards and my appologies for any grammar mystake,
    João Mesquita

  4. November 7th, 2010 at 12:51 pm

    Steve Doig says:

    Thanks for your comment, João. You’re right that the numbers my student gathered are imperfect measures. The high standard deviation is a product of the variable density of the marchers; it was not like ranks of soldiers marching by. (You are also right that if I was writing this for a newspaper I wouldn’t use capital letters.)

    But even if you take the very highest count — that would be 150 marchers passing during a minute — and apply it across all 63 minutes of the march you would get a count of 9,450. That’s still below — far below — the claims that 100,000 participated in the rally. The 100,000 simply isn’t credible; that’s a bigger crowd than can fit into Benfica’s stadium.

    My point is not to dismiss the obvious passion of the protesters. Instead, I argue that it always is necessary for news media to supply an independent estimate of crowds at all political rallies. The same phenomenon of wildly-inflated claims of crowd size occurs in the United States; journalists should not just let such claims stand without being verified independently. If the news media regularly did such estimates, rally organizers would stop making unsupportable claims.

  5. November 7th, 2010 at 1:33 pm

    José says:

    I found your work very interesting and valuable, specially because it’s a theme that usually ends in worthless discussion.
    But its really a pitty that i couldnt share it on facebook.
    Please correct that, you’ll find it a lot easier than counting mobs 🙂

  6. November 7th, 2010 at 2:09 pm

    Carlos Meneses Oliveira says:

    João Mesquita presumes that the marchers were evenly distributed as they went down the avenue. So he thinks that the measuring method failed as the SD is too high.

    The testemony of the students is that they were not uniformly distributed but very heterogenous distributed.

    The counting method is somehow fragile as is dependend on very short periods of time, but was confirmed by three other accounts.

    The result is a much better evidence than the newspapers method.

    A bigger mistake than having a not very strong method is descarting the best evidence as being as good as any other claim.

    Sorry for the bad english

  7. November 7th, 2010 at 3:44 pm

    Steve Doig says:

    Thanks for your comment. (And please don’t apologize for your English, which is VASTLY better than my Portuguese!)

  8. November 7th, 2010 at 6:08 pm

    Filipa Faria says:
  9. November 7th, 2010 at 8:21 pm

    Jose Simoes says:

    “Her counts of marchers passing in those 30-second periods were 51, 40, 75, 52, 45, 20, 70, 52, 58, 70, 51, 49, 46, 28, 17 and 25. The average of those counts is 94 marchers passing per minute”

    Without disclosing the protocol for counting (written in advance, and including the instruments used), the places of the observers – and the raw data this makes no sense at all.

    Counting heads and watching the clock at the same time is not easy without proper instruments and good lines of conduct.

  10. November 7th, 2010 at 9:16 pm

    Steve Doig says:

    Not easy, perhaps, but not impossible. For instance, I did several 30-second counts of the marchers going by using my iPhone as a timer. I agree that counting people marching past is not exact, but they were going by slowly enough that it was possible to get pretty good counts.

    Another way to look at the claim of 100,000 marchers is to consider how many would have to march past every 30 seconds for all of them to finish in 63 minutes. The answer is 794 people every 30 seconds. It may not be possible to say that exactly 58 people passed in a given 30-second period, but it should be clear that the person who counts didn’t miss more than 700 additional people walking by.

  11. November 8th, 2010 at 12:57 am

    Edwin van Raaij says:


    I was searching on google for ways to count people in rallies and your system seems very solid. Great work and excellent detail presented here.

    The Hague, Netherlands

  12. November 8th, 2010 at 11:22 am

    João Caetano says:

    Some people are loosing the scope here.

    The difference between Steve’s estimate (10k) and the organizers (100k) is a factor of 10.

    Steve’s method looks pretty solid to prove that the organizers number is well over estimated.

    Much better statistics, atomic-clock devices for timing, or laser-based devices for measuring the length would only allow you to say that the organizers number was over estimated by a factor of (let’s say) 9.3865, instead of by a factor of 10.

  13. November 20th, 2010 at 7:37 pm

    drcursor says:

    I saw you today at the anti-NATO demonstration, and I confirmed what I already thought… your method os anything but scientific… you just we’re in a place where it would be impossible to see all the way to the other side of the street, and counting average numbers without any scientific ou mathematical base…

  14. November 21st, 2010 at 3:11 pm

    Steve Doig says:

    Well, I disagree. As you must have observed, I was standing right on the east side of the main trafficway of Avenida da Liberdade within arm’s-length of the marchers, and I could easily see across the 20 meters to the other side of the street. See this blog entry for a full description of the math.

  15. September 16th, 2012 at 5:10 am

    António says:

    Has any of you made a count on today’s rallie? – 15th September 2012. Ps: congrats on your work. Seems solid.

  16. December 5th, 2012 at 8:48 pm

    Steve Doig says:

    No, alas. I’m not in Lisbon now, though I look forward to returning again soon.

Leave a Reply