Crowd counting and cities: 5 things you need to know

When the streets of U.S. cities fill with people, how do we count them?

By Megan Barber@megcbarber Updated Sep 20, 2019, 11:53am EDT

An aerial view of Hamburg, Germany, with people crowded onto the streets amid buildings, a canal, and boats in the water. — The streets of Hamburg, Germany, fill with demonstrators from the Global Climate Strike on September 20, 2019. Big crowds are also expected in U.S. cities.

In cities around the world today, thousands of people will take to the streets as part of the Global Climate Strike, a protest to demand urgent action on climate change. But how many people are marching and attending the strikes? The answer is harder to pin down that you might think.

Estimates of crowd size often vary depending on who you ask. A study by FiveThirtyEight of the 2017 Women’s Marches found that in at least 11 cities around the country, organizers’ crowd-size figures topped those offered by local officials by 40 percent.

The contested nature of crowd estimates raises some questions: How do people count crowds? Why is it so difficult? And is there ever an accurate number? With the help of Dr. John McCarthy—a leading sociologist at Penn State University—we break down these answers and reveal five things you need to know about counting crowds in cities.

A crowd gathers outside the U.S. Capitol for President Dwight D. Eisenhower's inauguration on January 20, 1957, in Washington D.C.

Why is crowd counting so contentious?

How many people show up to public events and protests matters because people have a vested interest in the outcome. “There are lots of alternative sides who have stakes in what the counts look like,” McCarthy tells Curbed. “Demonstrators want to have higher counts and counter demonstrators want to have lower counts. And even police sometimes have a stake in the game.” Estimates from organizers are often higher than those provided by law enforcement or public officials.

As McCarthy and Dr. Clark McPhail of the University of Illinois point out in their article, “Who Counts and How: Estimating the Size of Protests,” not all large gatherings are disputed. Few disagree with the number of attendees at sports victory celebrations like the record-breaking Chicago Cubs World Series parade, where an estimated five million fans filled the streets. But by and large, these types of events lack the charged political climate that usually frames a contentious crowd count.

How many people attend an inauguration or a march is a visible manifestation of support or protest that can seem much more impactful than poll numbers or election results. Large crowds in the streets, in public parks, or at monuments demonstrate a common purpose and commitment, both to participants and to onlookers. They also disrupt the normal ebb and flow of cities—closing streets, rerouting traffic, and often requiring more buses and trains.

An aerial view—without crowds—looking west along the National Mall, Washington, D.C.. The sociologists McPhail and McCarthy argue that the National Mall can hold about 1,048,206 people at 2.5 square feet per person.

The National Mall: Ground zero for crowd counting

Washington, D.C.’s National Mall, managed by the National Park Service, is the area between the Lincoln Memorial and the United States Capitol, and sees an estimated 24 million visitors each year.

The National Mall has been a popular destination for marches and protests since a redesign in the 1920s and 1930s cleared trees and gardens to create open spaces that can accommodate large crowds. With so many protests happening on the mall, it has become what McCarthy calls the “ground zero” of crowd counting.

One of the most contested marches in U.S. history was the Million Man March of 1995 in Washington D.C. The National Park Service estimated the crowd at 400,000, far below what organizers had planned for and what they believed they had observed.

According to the Washington Post, event organizers called the figure defamation and threatened to sue, while Congress decided to forbid the Park Service from conducting future crowd estimates. Because of this conflict, the National Park Service still does not release official figures on crowd size, even though McCarthy points out that the NPS likely still makes the counts for their own purposes.

Farouk El-Baz, a researcher at the Boston University Remote Sensing lab, analyzed the crowds using aerial photographs at the Million Man March and concluded that between 669,000 and one million people participated in the march. McCarthy and McPhail disagree with El-Baz’s numbers, saying that the NPS figure is likely more accurate. But the Million Man March—and the steps the NPS and Congress took after the event—illustrate just how politically charged crowd counting can become.

The Million Man March on the National Mall.

Methodology of crowd counting

There is an accepted methodology for how to estimate the number of participants in marches and protests in cities.

McPhail and McCarthy argue that the actual practice of counting a stationary crowd is fairly simple: You need to know the square footage of the site (its “carrying capacity”), the percentage of the site occupied by participants, and the crowd density.

This technique is the brainchild of Herbert Jacobs, a journalism professor at the University of California, Berkeley, who devised the process by watching students protest on his campus against the Vietnam War in the 1960s.

Jacobs’ premise was simple: Area times density will yield more accurate results than biased guesses. Density differs whether people are standing (at about five square feet per person) or sitting (at about 10 square feet per person), and observers know that the fronts of crowds are always more dense than the rear of crowds.

Ideally, credible size estimates require observations from multiple vantage points in order to accurately assess density and account for differences amongst the crowd. But scientists can also use high-resolution photographs of crowds taken from known vantage points to try to calculate the density of specific areas.

For places like D.C.’s National Mall, officials know the square footage of every green space, sidewalk, and monument.

As Butch Street, a management analyst for the Public Use Statistics Office of the National Park Service, told the Washington City Paper in 2009, “We have Washington, D.C., completely gridded out, every inch of the Park Service properties,” he says, “and we know how many acres and square feet are in each area. Very simply, we get our aerial overflight pictures, we apply a [population] density level for each one of those areas, and if the proper density levels are applied in each area, we know now that it gets us pretty close. This has been worked out over time.”

Crowds—like this one in New York City on January 21, 2017 during the Women’s March—are more difficult to count if they are moving or in streets.

What makes crowd counting difficult?

People’s inherent bias and where they observe crowds in cities makes crowd counting more difficult than the rather simple calculations suggest.

“If you’re on the ground, it’s easy to overestimate,” says McCarthy. “Because if you’re in the middle of it, it looks like a million people.” Low-angle vantage points—like the steps of the U.S. Capitol building—make the spaces between the people disappear. This is likely why after his inauguration, President Donald Trump said in a speech, “I'm like, wait a minute. I made a speech. I looked out, the field was, it looked like a million, million and a half people.” Other estimates from the inauguration are far below the 1.5 million mark.

It’s also hard to count crowds if they are moving. Aerial photography can still help determine the carrying capacity of streets and make guesses at density, but it is always harder to count a group of mobile marchers in an urban area compared to a stationary crowd.

Even if crowd scientists use digital photography and the “gold standard” of crowd measurements, numbers may slightly underestimate total attendance by failing to count people who come early or arrive late.

An example from 2010 of the imaging and grids used by Digital Design & Imaging Service.

Is it possible to get an accurate number?

New technology is helping crowd scientists produce more accurate numbers.

Curt Westergard of Virginia-based Digital Design & Imaging Service uses an aerostat— a balloon carrying a high-definition camera—to take high-resolution aerial photos and then import them into a 3D modeling software. The software lays out a grid on places like the National Mall to make for easier counting and promises to generate “accurate, unbiased, and replicable” crowd counts.

Westergard’s company helped to count crowds at the 2010 “Restoring Honor Rally” hosted by Glenn Beck and Sarah Palin at the Lincoln Memorial in Washington D.C. and the “Rally to Restore Sanity and/or Fear” that was hosted by Stephen Colbert and Jon Stewart on the National Mall a few months later. The firm claims it can count the crowd to within 10 percent, and it observed about 87,000 people at the Beck rally.

Another way to ensure accurate numbers is to compare credible estimates with other known figures, like the passenger volume on buses and public transportation. In particular, metros and subways can provide helpful data, especially if users have to walk through a turnstile to access the train.

Overall, crowd scientists and sociologists like McCarthy believe that the accepted practice of crowd counting—using square footage, density, and the percentage of the site occupied by participants—can yield “moderately accurate” results. But with so much at stake with how crowd numbers are interpreted, the debate will likely continue.