Stand and be counted

On a sweltering Sunday in the humidity of the Hong Kong summer, thousands of people descend on Hennessy Road, a major thoroughfare in the city’s Wan Chai district. The crowd move slowly, ebbing westward towards the government headquarters in Admiralty. The people carry handwritten banners, chant and hold smartphones aloft, trying to capture the size of the throng they’re marching with.

Demonstrations are a regular feature of Hong Kong life and they are seldom about one thing: universal suffrage, freedom of speech and a call to return to British rule all happen to be in the mix today, on the anniversary of the city being handed back to the Chinese on 1 July 1997.

As in most countries, the turnout of any demonstration is a marker of public sentiment, which makes accurately counting heads an important job. Perched on a footbridge overlooking the road, Edward Tai, a senior data analyst at the University of Hong Kong’s Public Opinion Programme, peers down at the moving mass of people. His eyes dart from head-to-head, silver tally counter in hand, furiously clicking to record the number of protesters passing through a checkpoint. Tai, a practised crowd-counter with 11 years of experience speaks as he clicks: “It’s not difficult to tally up the number of participants here, as they move slowly,” he says. “It’s only difficult in wet weather when people are holding umbrellas; then you have to look harder to see how many people are under them.”

Tai and his team of 12 counters have a finely tuned system to get an accurate result. Each team member is assigned a different traffic lane, they count the number of protesters who move past a predetermined point – in this case a set of traffic lights – for two minutes before swapping with someone else so that they can record the results and take a quick break. Then, after the event, the team take a telephone survey of 6,000 Hong Kongers using a maths formula to estimate the real number of marchers.

But the University of Hong Kong are not the only crowd-counters at work. The pro-democracy Civil Human Rights Front (chrf), who are responsible for organising the rally, and the pro-government Police Force also have crowd-counters deployed to gauge the size of the protest. While the chrf has a tendency to exaggerate the number of people who attend rallies, the police have a motive to play down the scale of the gatherings and their associated discontent.

In the case of this protest, it is no different. Newspaper headlines the following day focus on two figures: 50,000 from the Civil Human Rights Front and 9,800 from the police. Meanwhile the University of Hong Kong team count approximately 30,000 – almost exactly in the middle. According to Tai, there is a statistical reason why the numbers from the chrf and police are out of sync: they are counting different crowds. The chrf is documenting the total number of participants throughout the course of the march while the police count the maximum number at a specific checkpoint at any one time. “We are not comparing apples to apples,” says Tai.

It is conceivable that manual crowd counting might vanish in the near future. Facial recognition technology exists that would make light and accurate work of counting the number of participants in a protest, but until such software comes down in price, Tai’s thumb is destined to go on clicking: “What’s most important to us is to have a reliable methodology,” he says, resetting his counter to zero.