Head counts at protests and other large gatherings can vary wildly, but it’s the police evaluation that is seen as authoritative. How do they calculate it?
How do you assess the size of a crowd? Absent ticketing or turnstiles, it’s a question that has vexed authorities, organisers and media – just about everyone except Donald Trump – since time immemorial.
The subject reared its blurry heads most recently in New Zealand during the Hīkoi mō te Tiriti, first as the crowds marched through Auckland and, especially, as protesters demonstrated their opposition to policies affecting Māori on the streets of Wellington and outside parliament. Some asked whether it might just be the most populated New Zealand protest of all time.
There were around 100,000 there, reckoned Rawiri Waititi. More like 22,000, said Winston Peters. “A hell of a lot more than 35,000,” said Shane Jones. Media estimates ranged from 17,000 to 100k. Online sleuths rattled out their appraisals like Wall Street traders. A Ngāpuhi data nerd in Canada whose aunty was massaging tired marchers in Wainuiomata studied drone footage and concluded 50,000 at least.
At times such as these, the counts that tend to get repeated into calcified form come from the constabulary. They have no formal responsibility to stick a number on the tin, but their calculations are of critical importance – in managing resources, forestalling any dangerous overcrowding and so on. The official police estimate just before 1pm on that Tuesday in November was 35,000. Four hours later, that was revised up to 42,000.
How, then, do they arrive at that number, with its polish of authority? Is there some fancy, cutting-edge tech they deploy in the cause? A crowd-size manual? A superintendent with special responsibility for estimate-ops? I filed an official information request to find out.
The response is in, and if you were expecting something involving elaborate algorithmic mesh computations, you’re going to be disappointed. Turns out they do it pretty much the way any of us might. “The technique to estimate crowds when the event is occurring is usually by taking screenshots of streamed video from a wide range of sources and counting from that,” wrote inspector Marc Paynter in response. “This includes publicly available video and photos.”
There isn’t even a policy. “There are no documents or written guidance or methodology on estimating crowd size used by police,” he said. “Usually crowd size is estimated by aspects such as the capacity of a stadium, previous events of a similar nature, number of entries sold by the organisers etc. Even the weather forecast will impact the turnout to many events.”
A request for any correspondence relating to the estimation of the numbers attending the hīkoi was rejected on the grounds that it “would be likely to prejudice the security of New Zealand” or cause “prejudice to measures protecting the health or safety of members of the public”.
Paynter did note that an estimation of likely attendance had been made in advance of the hīkoi, given the need to “determine resources as part of the planning process”. That figure, arrived at in late September, was 15,000-25,000 – “based on previous events”, he said. “The Auckland Harbour Bridge attendance was used as an indicator of crowd size to predict the Wellington crowd size.”
There was no mention in the police response to a “formula” described in remarks attributed to a police spokesperson in an RNZ report from December. They’d said then: “A formula of 1.5 people per square metre resulted in an estimated crowd size of nearly 42,000 people.”
That approach – of counting the crowd density in a small area and extrapolating out – is often described as the Jacobs’ Method, and dates from the 1960s. It’s not without its uses, but it’s a crude measure, and limited in its efficacy.
“I was expecting,” said Ruggiero Lovreglio, a professor at Massey University and expert on crowd size measurement, when I shared the police response with him, “a bit more data-driven solutions.” He pointed me to a letter he and a colleague had published in Science magazine, which lays out why it is so important to reliably estimate crowd sizes. Getting the number right matters because it is “a matter of safety as we need to avoid crowd crushes”, he said.
The science and the technology is constantly evolving, said Lovreglio. “Today, AI applications in computer vision can provide great support in estimating crowd size in real-time.”
He added: “This can today be achieved by using traditional CCTV videos or even drone footage. In recent months, we have been identifying what algorithms are available to study crowd movement, and we are assessing their accuracy as part of my Rutherford Discovery research. A lot needs to be done to bring the latest research findings on this topic to the public domain so they can be used to have safer events with crowds.”