Wednesday, May 12, 2010
Salmon Migration - Untangling the spin wheel of "crowd estimates"
I could not fathom how the Times Colonist daily paper could have seen "nearly 1,000" people at Alexandra Morton's salmon migration rally in Victoria last Saturday, when I had seen many thousands. I really needed to understand.
So when last Monday, Times Colonist editor Stephanie Coombs kindly responded to my complaint letter about the event's coverage, I seized the opportunity.
"Reporter Katie DeRosa, who attended the rally," Coombs had written in her response to me, "spoke to two different Victoria police officers, as well as two rally helpers, who all estimated about 1,000 people."
Strange. That same day, I had received a message from Rafe Mair who quoted Global News reporter Holly Adams saying that "I spoke with Police outside the Legislature and they estimated just over 4,000 people, and that was just before 5:00."
The Globe and Mail had reported 4,000 people as well. So it appeared as if police officers had been mischievously telling 4,000+ to Global News and the Globe and Mail, and "nearly 1,000" to the Times Colonist.
So I wrote back to Stephanie Coombs asking her if she would care to comment about the Global News reporter's account of 4,000.
Her response: "We have spoken again with Victoria police today, and their official report on the rally indicates a crowd estimate of 1,000 to 2,000 at the legislature."
One to two thousand? But that was already double the number that her paper had initially reported. What was going on here? I figured that if I was going to understand those numbers, I'd better go to the source.
So on Tuesday morning, I called the Victoria Police Department's main line and asked to talk to someone about their attendance estimates for Saturday's rally. The switchboard operator connected me to Kathy Jorgensen from Operational Planning. Let me get back to you on that one, she kindly said when I explained the purpose of my call.
About an hour later, she left a message on my voicemail saying: "Our police estimated the count at maximum 1,500 once it got down to the Legislature Building."
Yet a different estimate. It was the fourth one I had received already. I called her right back. How did the police department go about determining that number, I asked, what's the methodology used? We just ask police officers who were there to give us their estimate; it's a casual count, we don't use a specific method, she explained.
I told her about what the Global News reporter had said - that police officers who were there had told her 4,000. I don't know who told her that, she responded, so I cannot comment.
Do you use photos of the crowd to help you refine your estimates? I ventured to ask. No, was the answer. Then she became a little nervous and told me: I don't know where you are going with this, so you need to call Sgt. Hamilton who is our media person.
Which I did right away. But he never returned my calls, so I was left spinning my wheels about Ms. Jorgensen's responses. No specific methodology to count the crowds... But why not? What's wrong with introducing a little bit of objectivity in estimating a number which is so critical to so many different stakeholders? With modern technology and a bit of planning and brain power, you would think that something could be done.
So I decided to give it a try.
What we saw.
I downloaded the above sample picture (courtesy Don Staniford) which I found among hundreds on the web. I took a snapshot of the BC legislature's lawn in Google Earth and imported it into Google Sketchup. I then plotted a polygon representing the approximate location of the crowd according to the picture. Some distinct features allow you to situate the crowd in the picture fairly easily, such as the fountain, the statue of the Queen, the trees, the flagpole, the stairs where the photographer was standing, etc. Here is what it looked like:
What we saw (continued).
Google Sketchup calculated the area of my polygon: approximately 6,000 square meters. The lawn itself is a 100 by 100 square, or about 10,000 square meters. My polygon therefore occupied about 60% of the lawn.
I then proceeded to estimate how many people could be standing in that polygon according to the picture. Densities vary: people closer to the stairs are clearly shoulder to shoulder, while people towards the statue were able to sit in the grass. Your typical "cocktail party" average density is about 0.5 square meter per person. People close to the stairs were probably using less than that, while people in the back were using more.
I took a very conservative guess: I assumed - which is very unrealistic, based on what the photo shows - that each person used 2 square meters on an exclusive basis. That's a rectangle of one meter by two meters with no one else but its sole occupier on it. Measure that at home, and you will realize that it's a very, very conservative assumption indeed. I also assumed that not a single person was standing to the left or the right of the frame of the photo, and I further assumed that the columns of people still moving toward the lawn in the photo's far background were actually not going to the rally.
In spite of that, I still found that approximately 3,000 people were occupying my polygon. Once you add more realistic estimations that other people must have been standing outside of the picture, that some people in the far background are actually going to the rally, etc. you easily find yourself in that 4,000+ range which was given to Global News on that day by several on-site police officers.
Let's continue my little experiment. Let's assume that the Times Colonist got it right and that "almost 1,000" people attended the rally. At 2 square meters per person, that's a 2,000 square-meter polygon. Here is what the Times Colonist "saw" happening on the lawn of the Legislature last Saturday:
What the Times Colonist "saw".
Notice how the Times Colonist's polygon does not even extend to the fountain, which was in actuality covered with people. Does it look to you like they got their numbers right? Well yes, me too.
Now just to be clear - I am not claiming to have discovered a new "methodology"! My point is simply that some methodology would not have hurt. If I was able to hack those estimates in a couple of hours on my home computer using some free software and a publicly available picture, imagine what a trained staff could do with sophisticated software and pictures that were taken with that purpose in mind.
After spending the past three days chasing phantoms in spinland, my head was hurting a little. But then suddenly today, out of nowhere, someone - finally! - made some sense. Sgt. Matt Waterman from Victoria PD Operational Planning returned one of my many calls. He told me that the Department's official estimate was 1,500 to 2,000 people (yes, a fifth different estimate!), and he confirmed what Kathy Jorgensen had already told me - that the Police Department does not use any particular methodology to come up with the number. We just guessed it, he said. Well sure, we could have used pictures and fancy methodologies to come up with a number, he explained quite candidly, but we had no reason to do that! We didn't know it was important. After all, our job is not to count, but to escort and protect people.
I explained to him that it was actually very important to many people, starting with the 5,000 or so people who were at the Legislature, because media used those Police estimates and presented them as reliable hard numbers, rather than the subjective wild guesses that they really were. I told him how the Times Colonist's editor had pointed the finger to the Police Department as soon as I started asking her some hard questions. He was clearly not pleased to hear that. Well, lesson learned! he commented. I will recommend that the Police refrain from making any more such estimates in the future.
Halleluia, brother! I am totally with you on that one. Stop counting, just focus on your job of protecting. Let other spinmeisters trip all over themselves with ridiculously low-balled estimates.
So, two lessons learned here, as the good Sargent Waterman would have said:
1. The Victoria Police crowd estimates for Saturday's event are worthless, according to people working in that very Department.
2. Times Colonist, you have hereby been put on notice. Next time you publish absurdly inaccurate crowd estimates that get people mad at you, don't run to the Police for cover because they will kick your butt.