The Science Behind the March for Science Crowd Estimates
The March for Science spanned the globe with events on every continent. Marches took place at the North Pole, the tip of South Africa, in the shadow of Brandenburg Gate, and along the edge of the Great Barrier Reef.
We’ve spent the last couple of weeks crunching the numbers and using some crowd science techniques to make an official estimate of the total number of marchers around the world.
We’re excited to announce that the 600 marches around the world amassed an unprecedented total of approximately 1.07 million marchers -- making the March for Science the largest global science event in history. We're also incredibly grateful to the team of approximately 10,000 volunteers and organizers made these events possible.
Our estimates include some of the biggest marches of the day:
- Washington, DC: 100,000 estimated marchers
- Boston, MA: 70,000 estimated marchers
- Chicago, IL: 60,000 estimated marchers
- Los Angeles, CA: 50,000 estimated marchers
- San Francisco, CA: 50,000 estimated marchers
As well as some smaller, but no less mighty marches in the U.S. and around the globe:
- 11,000 estimated marchers in Berlin, Germany
- 500 estimated marchers in São Paulo, Brazil
- 255 estimated marchers in Huntington, WV
- 90 estimated marchers in Skagway, AK
- 84 estimated marchers in Kampala, Uganda
And though we didn’t include them in our total, we can’t forget to mention the March of the Penguins for Science at the Monterey Bay Aquarium, which included 5 feathered marchers and garnered more than 3 million views on Facebook.
Check out a map of the Marches and their sizes below:
So, now you might be asking: how did we arrive at these numbers? And are we sure that our total crowd count isn’t an inflated estimation -- a problem that often plagues many large events?
Crowd estimation is often an inexact science with wildly different methods leading to very different results. With 600 marches, ensuring that every march used the same protocol and counting method was impossible. As we amassed the data, each march shared how they arrived at their final number and how confident they were in the number. A few basic methodologies emerged, depending on the march size.
The History of Crowd Estimation Science
Modern crowd estimation science traces back to University of California, Berkeley journalism professor Herbert Jacobs. During the protests on campus in the 1960s, he took photos from above through a window at an adjacent building. He was able to hand count the number of people in a small known area. Using that density, the total area, and the percentage of total area occupied, he arrived at a total attendance.
This basic technique of density x total area is still the most commonly used method for estimating crowd size of large events today. While modern computer visualization techniques have improved on this counting method, this basic idea of taking a photo from a slightly elevated level (400-800 feet is ideal) and calculating the density is still used for most large events in public spaces.
74% of marches with an estimated crowd size of 500 or less used a simple hand count to determine attendance. The remaining set either used some combination of registration/sign-in for estimates or utilized fire marshal estimates. Of those using a hand count, 53% used a system that involved multiple people counting the crowd independent of each other. Most marches that used this method reported an average number from those counters.
Hand counts are prone to some error, especially in the context of a moving crowd as is the case of marches like these. The process of averaging of counts makes it very difficult to assess the potential error rate. A review of crowd science literature found very little information on the error of the hand counts method.
The total reported attendance of those using a hand count was approximately 60,000. We estimate that even with a high error rate (>25%), the total impact on the final attendance number would likely be 1-2%.
Police and Fire Estimates
Many police and fire estimates are based on the Jacobs method. The NYPD uses a modified version, estimating a number based on the density and number of barricades used. In practice, many police and fire departments declined to provide estimates to local organizers citing issues in estimation or a departmental policy. Where police and fire estimates were available, these numbers were used, unless a more reliable form of data was available.
52 marches reported data via police and fire estimates, representing an aggregate attendance of 151,000. While police estimate protocols in each of these cities cannot be verified, we can assume police in these towns use similar procedures for estimating attendance event to event. This gives us contextual confidence in the reported attendance numbers.
Density Calculation using Known Landmarks
Most of the marches with more than 5,000 attendees reported data using photo or video evidence. For example, Indianapolis had a conservative estimate of 10,000 attendees. At the beginning of their event, pictures of the square were taken with reference landmarks visible. Using this and other pictures, the organizers estimated a fairly consistent density across the square. The organizers then drew an outline of the area using MapChecking.com and combined that with state crowd density still data from Professor Keith Still, a crowd estimation scientist. They settled on a likely density of 2 and 3 people per m2 resulting in an attendance calculation of just under 10,000. Given that more people arrived after this picture was taken and there were people at the periphery beyond the borders of area, this estimate is likely conservative.
Aerial Imagery and Time Lapses
Additionally, some marches used time lapse footage and drone footage to help estimate the the total number of marchers. More than 100 marches used a combination of drone footage and aerial photography to estimate their crowd size, including 95% of the marches with attendance >10,000.
In San Francisco, a combination of aerial footage was used with the below time lapse to estimate the density of marchers over a known area. Density was sampled at multiple times to account for any potential gradient as the march proceeded. Combining this with the drone aerial footage and the known area occupied, an overall estimate more than 50,000 was calculated. Given limitations with some of the aerial footage, the organizers reduced the reported estimate to 50,000.
The D.C. March for Science
Unfortunately in the case of the D.C. March for Science, official estimates from the National Park Service were not available and cloud cover over DC prevented the use of overhead or satellite imagery to gauge crowd size. The D.C. event began with a rally on the grounds of the Washington Monument. We know that with the stage and production setup, these grounds can hold an estimated 100,000 people. Views from the stage show significant crowd density in this area. Coupled with organizers' accounts of the large numbers of attendees who remained outside of the security perimeter of the Washington Monument grounds, the D.C. March for Science team is confident in an estimate of at least 100,000 attendees.
Independent of the final numbers, the support and enthusiasm for the March for Science around the globe was overwhelming. We are confident in our assessment that more than 1 million people participated in the March and took every effort to ensure that this estimate was informed by sound scientific practices for evaluating crowd sizes.
Be sure to stay tuned to learn more about the insights we gained from March RSVPs as well as the sustained action we saw during the post-March Week of Action!