Dogs and Demographics
From the Toronto Star:
The major trend suggested by what may be the first-ever mapping of Toronto, by popularity of dog breeds, using dog license data, is that some stereotypes awkwardly hold. There seem to be working-, middle-, and upper-crust dogs.
We’ve posted here before that, in much of today’s society, modern dog breeds have developed a distressingly close relation to brand names. Now, according to the Star, demographic data indicates that they really may be more like status symbols than we’d like to believe.
Predictably, the idea that most readers will get from this survey are pretty much what it appears the Star expected – that pitbulls are more popular in the less desirable parts of town and poodles are popular in tony neighborhoods. What I got out of it was the bad taste I find in my mouth when someone publishes a haphazard imitation of statistics as science.
If one goes to the AKC website and looks up the 2006 registration numbers (I couldn’t find the 2007 data, but 2006 should illustrate my point just as well) you will see at a glance that the population of dogs within each individual breed varies over four orders of magnitude; from Labrador Retrievers with over 100,000 individuals registered to English Foxhounds with only 11. The data do not appear to be normally distributed and unfortunately I lack the time and software to crunch them right now. But - I was able to find a study published in The Royal Society Biology Letters in 2004, in which Herzog et al discuss the distribution of modern dog breeds in detail. They note that:
…the number of new registrations within each breed obeys a steeply descending distribution with a long extended ‘tail’, such that while most breeds include a small number of new purebred puppies registered each year, a few breeds have a very large number of new registrants. When the relative frequencies of dogs in each breed are plotted as a function of their expected number (…) the data clearly form a power-law probability distribution…
Herzog et al’s study, which used population genetic models to study the biological relevance of how cultural change affects dog breed popularity (and btw, verified for me that it is not a Gaussian distribution), came up with a much different explanation for the distribution of dog breeds. In a nutshell, they believe that we’re a bunch of copycats. To whit:
A simple model of random copying among individuals, similar to the population genetic model of random drift, can predict the variability in the popularity of cultural variants. Here, we show that random drift also explains a biologically relevant cultural phenomenon—changes in the distributions of popularity of dog breeds in the United States in each of the past 50 years.
…A remarkably useful way to study cultural change is to assume that individuals, confronted with many different choices, simply copy other individuals rather than make ‘optimal’ or ‘rational’ decisions. In general, this process relates to a classic evolutionary phenomenon called random genetic drift, for which quantitative models originally developed in population genetics have proven useful in identifying mechanisms of cultural change.
Proving, once again, that gangbangers don’t get pitbulls and rottweilers because they’re inherently vicious breeds. They get them because other gangbangers have them. And yuppies get chichi rare breed dogs because nobody else has them. If we outlaw pitbulls and rottweilers, gangbangers will just latch on to different breeds and incorporate them into their culture. And – when pitties and rotties then become sufficiently rare and exotic, they’ll be the new yuppie dogs de jour.