17 January 2020 — 0424 mst
Montana’s women march tomorrow —
will data journalists get accurate crowd counts?
Tomorrow, Montana’s progressive women march and rally in Helena, Bozeman, and I think, in Missoula, Hamilton, Whitefish, and perhaps elsewhere. Organizers in Kalispell and Missoula are busing marchers to Helena instead of holding local rallies.
In 2017 the reported count for the rally in Helena was 10,000, which I thought then, and still think, was too high. Now Rep. Mary Ann Dunwell claims the number actually was 15,000, but offers no evidence to support her claim.
It's an old story. Crowd inflation occurs because organizers have optimistic eyes, seldom if ever count with scientific rigor, and predisposed to error on the high side. That’s not opinion. That’s fact.
FiveThirtyEight’s Nate Silver found that local officials’ crowd size estimates of the massive 2017 women’s marches were markedly lower than the estimates of the organizers of the events.
As FiveThirtyEight did for the tea party protests in April 2009 and for the Occupy Wall Street protests in 2011, we sought to collect credible estimates of crowd sizes at the Women’s Marches based on local news accounts. (You can find a complete accounting of our estimates and sources here.) We wanted to avoid estimates given by march participants or organizers, since these often exaggerate attendance compared with estimates by public officials such as local police and fire departments. In St. Louis, for example, police estimated the crowd at 13,000 participants, while a march organizer said 20,000 people had come.
Overall, we found 11 cities where there were separate estimates of crowd sizes given by organizers and local officials. They followed a remarkably consistent pattern: In all cases, the estimate by local officials was 50 to 70 percent as high as the one given by march organizers. Or put another way, the estimates produced by organizers probably exaggerated crowd sizes by 40 percent to 100 percent, depending on the city.
In late January, 2017, I wrote “Applying [Silver’s] 40 percent discount to the march in Helena, whose organizers pegged at 10,000, yields a turnout of 6,000.” I immediately caught hell for not believing women.
The next day I posted three paragraphs on counting crowds.
The Seattle Times explains how to estimate crowd size using grade school arithmetic. Yesterday, ruffled some feathers when, referring to Nate Silver’s work, I said I was discounting the organizers’ estimate of the turnout for the women’s march in Helena by 40 percent. Was I accusing the organizers of lying? No. I was observing that the organizers of rallies and marches often let their enthusiasm get the better of their judgment, leading them to claim crowd sizes not supported by independent analysis. That Trump loses credibility by lying about size should caution his critics’ not to inflate their own numbers about size.
That independent analysis could come from the news media. It could also come from universities and colleges; in Montana, from the University of Montana and Montana State University. I suggest that the political science, journalism, and GIS, departments collaborate, teaching students the science and art of counting crowds, and send teams of students to count the crowds at events such as parades and rallies.
I’m sure the regents and the administrators of the universities would welcome the opportunity to endorse, with foot-stompin’ excitement, the educational and public policy benefits of providing solid estimates of crowd size, especially of controversial political events. There’s more to university math than Second Down and 20.
A year later, determined to get an accurate count of the women’s rally in Kalispell, I made a panoramic image of the rally at its peak. After dividing the image into sectors and counting heads in each sector, I concluded that the crowd numbered 450–500.
My methodology was sound, my conclusion solid, and I expected my numbers would be accepted, and accepted with gratitude and enthusiasm.
Hoo, boy, was I wrong. A number of women, none of whom had employed rigorous counting methods, disputed my count, arguing I had missed people behind trees and so forth. They preferred their own unscientific eyeball estimates that yielded considerably higher numbers.
Inaccurate crowd counts are not harmless. They deprive the public of accurate information about the size of marchs, rallys, and protests. And they cheat the organizers of subsequent events. If a crowd of 5,000 is reported as 10,000, and the crowd at the same event the following year is accurately reported at 7,500, the organizers of the second event do not get credit for assembling 50 percent more people. Worse, they’ll also be criticized for not doing as well as their predecessors.
Accurate counts of large counts require analyzing overhead imagery
Ideally, a drone with a high resolution video camera could produce imagery suitable for counting the crowd. Whether private citizens could legally fly a drone that close to the state capitol could be a problem, although I suspect that by working with law enforcement journalists could acquire the imagery. Otherwise, I would put a Go Pro on a 20-foot pole to photograph the crowd. There are other ways to obtain the imagery, and I leave it to Montana’s data journalists to devise ways to obtain the data they need to produce accurate crowd counts.