Callimachus Miltiades
IT IS is not every day that I agree with Kevin Bonham, but as Kevin has accurately pointed out — the situation is now ridiculous where there are such disparate estimates in the number of marchers taking part in political protests in Tasmania. (March against the mill, comments) Likewise, as Kevin also points out, this seems to be an all-too common phenomena. Why?
I am not a member of the Wilderness Society or any political party, but I was present at the Launceston march. I was equally gobsmacked by the ridiculous crowd estimates — so I decided to waste my own time (all 30 minutes) in applying some basic math to the problem (I expect the Examiner has done the same).
Whose head count?
The common reliance on physical “head count” methods is known to be both highly inaccurate and biased (usually towards those whom control the media). The classic example of this problem is given by the controversial Million Man March in Washington on 16 October 1995. It has been the custom in the US for the National Parks Police to issue official crowd estimates for such events. For the MMM event in 1995, the National Parks Police estimated a crowd of 400,000, although the organisers released an estimate of 1.5-2 million (as well as a law suit against the National Parks Police for inaccuracy).
In a 2003 study conducted by Prof. Farouk El-Baz of Boston University, a series of satellite images and a team of statisticians was used to estimated the MMM crowd at about 870,000 (somewhere between the two camps). El-Baz concluded that it is not possible to obtain reliable crowd estimates from “head counts” on the ground. Not surprisingly the US Congress have officially banned the National Parks Police from providing crowd estimates.
The science of marching
The logistics and science of marching is well established and has been for thousands of years. Most military officers, and quite a few senior police, can accurately determine the number of people in a body of marchers from the known dimensions of troop movements — a company or regiment marching in a designated formation will occupy a fairly constant amount of ground. Whilst this cannot be strictly applied to a civilian protest march, the principles are the same.
In the case of Launceston on Saturday, there are a number of known constants and some assumptions on reasonable grounds that can be effectively utilised in a deterministic model to produce a probable outcome. This is a statistically sound approach which is utilised in various industries for accurate estimations.
The Launie lunacy of estimation
Firstly we know that the total length of the Launceston march route stretched from the corner of Cameron and St Johns Streets to the bandstand in City Park, a known distance of 1050 metres. However, as I had to run from near the end of the parade to the near the front to find some missing friends, I was in a unique position to note that the parade had well and truly entered City Park at about the same time as the tail had left Cameron street. Hence the total continuous length of the parade was at least 770 metres.
Secondly we can see from Oosting’s photo (March against the mill, comments) that the crowd was rather tightly packed owing to frequent stops-and-starts at intersections. Despite this I have generously estimated that the average distance between marchers was 1 metre with an average person “thickness” of 0.4 metres. This means that, in single-file, about 550 people could have occupied a single file from start to finish of the parade. However the photo also shows at least one (possibly two) prams in the first 10m or so of the parade depth, suggesting that there may have been between 77 and 115 prams in total, each occupying a length of about 0.7 metres. There was also an 8m tilt-bed truck in the middle, so I should shave 2% off the total number of people per file. I have decided to be generous to the Examiner and have shaved off 10%, giving a total number of 495 people per file.
Finally we need to estimate the width of the parade. We can see from Mr. Oosting’s photo that the width of the parade is at least 16-a-breast and possibly more. If we assume that the width could have swelled to about 20-a-breast and declined to about 10-a-breast at the rear, then 16-a-breast seems a reasonable middle value. Certainly, where I was standing towards the end of the parade the width was about 10-12 a-breast.
It is possible to plot the logical variations on a log-probability graph to determine a realistic outcome. In the table below is a list of likely probabilistic outcomes based on the known data. Clearly the most likely outcome (P50) is 7,920 people, with a 10% chance of 10,800 people on the high-side and a 10% chance of 4,840 people on the low-side. It is standard to utilise the P50 value as the most likely outcome. I have subsequently noted that the Wilderness Society estimated 8,000 people which is close to the P50 probability. Of particular note is that the Examiner estimate of 2000-3000 people is in the range of a 5% probability.
Statistics, damn lies and spin
I know that most lay-people reading this will see yet another bewildering set of numbers to form a new basis for argument. I also know that a tiny group of people with knowledge of the power of statistics will see a somewhat different story. This story will centre on why seemingly respectable newsagencies such as the ABC have become embroiled in such a farcical affair as deliberately falsifying crowd estimates. The 2000-3000 number promulgated by a number of agencies is so statistically far-fetched that one must question the origin the validity of the number and how it has become to be promulgated as an authourative value.
Critics have suggested that there was a deliberate plan to promulgate inaccurate figures on the protest and to falsely focus the ‘angle”of the story on a “Greens-Wilderness Society” fringe event. Unfortunately the real stats seem to suggest that, on the balance of probabilities, the critics may have a valid case.
