When large-scale demonstrations first plunged Hong Kong into political crisis, the size of the crowds became a contentious issue, with police giving low numbers and protest organizers much higher estimates.
For Reuters, this was an opportunity to calculate our own estimate and provide an unbiased statistic. Collecting and analysing a solid dataset would be pivotal in the success of the piece.
After speaking to crowd science experts and working out the analytics, we embarked on one of our most ambitious and innovative data exercises yet.
This immersive presentation shows how we arrived at our final calculation.
The piece is the definition of public service journalism, delivering a valuable independent statistic at a time when the size of the crowds was being used as a political tool. The project resulted in an exclusive story and figure that was not matched in-house by any other news organisation.
The piece was shared widely on social media in Hong Kong and across the rest of the world.
After speaking to crowd science experts and working out the analytics, the next step was to study the three-km route of the next march step by step, spending hours to pin down locations where protesters would funnel through. These would be the places to monitor the flow rate of people, and thus calculate crowd size.
Armed with zoom lenses, DSLR cameras, tripods, GoPros, iPhones, extra batteries and folding fishing chairs, the teams manned their counting stations.
Bursts of HD video were recorded at specific times and later played back in Adobe After Effects in order to count the heads of people passing a line in the road.
As data was being calculated, the graphics team also edited video from other cameras to make striking timelapses and visualised the data we had collected in order to explain the entire process to the reader.
We felt it was extremely important to be transparent about how we arrived at this number.
The visuals, graphics and text were presented in an immersive experience.
What was the hardest part of this project?
There were a few difficulties to overcome. The first and most important issue was working out the science of crowd counting. This is extremely difficult when a moving march is being studied, rather than a static crowd. The narrow streets and concrete skyscrapers of Hong Kong add to the challenge. The team spoke to crowd science experts and worked on theory and testing before everyone decided we were well placed to execute the idea.
The next difficulty was identifying positions on the route that would be good for people flow AND had an elevated position we could set up and comfortably occupy for a full day. Early exploration was done in remotely in Google Earth and after narrowing locations down we physically went there to do some tests. There were some issues in checking out views from one of the city’s hotels so we had the bright idea of combing through traveller’s photos from the hotel on Trip Advisor and managed to find some shots out of the window of the room looking directly at the road we wanted. We booked the room ahead of the next big protest.
Actually counting the heads of people was an enormous task in itself. Luckily it wasn’t raining or too sunny so there were very few umbrellas in the crowd. Counting such a high volume of people, even in bursts, was extremely taxing on the team members but we knew we had to get this published as soon as we could after the march. Some counting was even done on the plane as some of the team who weren’t based in Hong Kong headed home.
What can others learn from this project?
There are often creative ways to tackle a lack of available data. This is a good example of that. These techniques can be challenging but extremely rewarding.