Liberals and Conservatives said their rallies drew thousands. We decided to count

Political campaigns may be significantly off-base when it comes to the number of people they say are present at campaign rallies across the country, a CBC News investigation shows.

Both the Conservatives and Liberals have held major rallies since the beginning of the federal election, with the Conservatives in particular emphasizing the size of their events.

CBC News's visual investigations team hand-counted people in images taken at four rallies across the country — Conservative rallies in Surrey, B.C., and near Edmonton, and Liberal events in Toronto and Richmond, B.C.

The investigation, including expert analysis by crowd scientists, shows that the actual number of people attending those rallies was likely far lower than the counts the campaigns published on social media or reported to journalists at the events.

The Conservatives claimed that the Edmonton-area rally was attended by 15,000 people, for example. CBC News counted roughly 1,558 in one panorama photo shared by the campaign. Even assuming that a large number of attendees were missed due to perspective or obstruction, 15,000 is "not possible," according to G. Keith Still, an expert in crowd dynamics and safety, who is a visiting professor of crowd science at the University of Suffolk.

CBC News asked Still to analyze the events. His hand-counts were similar to our team's for all four events. Still noted that hand-counts carry a margin of error of five per cent.

For comparison, Winnipeg's NHL arena, the Canada Life Centre, has a maximum capacity of 15,321.

Similarly, the Liberals said their rally in Richmond had  2,000 people in the main room. CBC News's count, corroborated by Still, puts the number closer to 800. The manual count, as with the other rallies, was done by dividing the images into grids and tallying the heads visible.

Crowd sizes are often difficult to estimate, and hand-counts are not perfect. CBC News's count could be lower than the real number of people in the room, as attendees may not have been counted for a variety of reasons: They were behind the camera or otherwise not in the shot, they were obscured by a sign, another person or an object, or the perspective of the image may have meant they were hidden. To address these limitations, campaign photos of the rallies were compared with videos and images, from social media, showing other perspectives so that the count included most of those present. 

Manually counting heads is the most reliable way of determining crowd size, according to Still. 

"But for that you need a good high angle, good clear lines of sight, reasonably good lighting, and then a lot of time to actually count the heads," he said.

Chronologically, the first event examined was the Conservative rally on March 27 in Surrey. The party claimed 5,500 attendees.

CBC News counted approximately 1,522 in a picture posted by the campaign.

The photograph is in many ways ideal for counting: it's high definition and captures much of the area from a high angle. But some attendees were on the venue's mezzanine (from which the photograph appears to have been taken) and were not counted, while some could have been missed in the corners of the room.