Briefly, Some Inflammatory Numbers

It's done!

We've compiled a whole Excel file full of voting data and demographics alongside for each of the 107 Ontario ridings. We've already gone over how much this was a challenge, particularly for the eleven Northern Ontario ridings. As a result we're splitting up our reporting based on the whole of Ontario, the North, and the Rest of Ontario (ROO). If someone can come up with a better name for that, we're open to suggestions.

We thought we'd try it out by opening up another Excel worksheet and running some basic correlations just to see if it all made sense. Since the Ontario NDP, which managed to impose a minority government on the liberals due to its showing in 2011, will be looked at closely for the next few days, we thought we'd start with looking at how how it gained support.

These are correlations, of course, and we can't read too much into them. And the usual environmental fallacy caveats apply. But faced with an apparent absence of extensive exit polling in Canada, this is what we've got to deal with.

The Digits

Below is a graph based on the correlations we found based on Chinese residents, number of rented dwellings, and people "not in the labour force" in the most recent census (2001 for the North, and 2006 for the ROO), measured against raw vote totals (not "percentages of valid votes").

According to this admittedly rough first sketch, the NDP managed to increase its support from the Chinese-Canadian community. Or to put it more accurately, they managed to decrease the negative correlation between Chinese-Canadian presence in a riding and total NDP votes.  The biggest jump was in the north, with more marginal gains down south. 

Click to Enlarge

Why did I pick Chinese-Canadians? Well, I met my friend during a recent campaign, so I wanted to amuse him later with some statistics about "his people", as he might put it bemusedly. It was the first demographic listed alphabetically, and I just wanted a visible minority group from the census.  

Perhaps more interesting is what's going on with "# of rented dwellings", which we'll use as a proxy for "renters". Unlike the Chinese community, the NDP started out with a positive correlation for rented dwellings in 2007 everywhere, that is, except in the north. But in 2011, they then lost some renters mostly in the Rest of Ontario, while in the North they actually made quite a big leap (though still ending up slightly negative). 

The "Not in the Labour Force" numbers were interesting as well. In 2007, Ontario-wide, the NDP actually had a negative correlation for this group. Only in the Rest of Ontario was there a slight correlation in favour of the NDP. The North, however, did not see the NEET flocking to the NDP by any means (-0.587).

In 2011, a time of greater labour market difficulties, the NDP improved quite a bit on their score with those "Not in the Labour Force", with a marked jump in the North (going to -0.101). The Rest of Ontario saw also saw a large move toward the NDP, but one not nearly as dramatic. Province-wide, the party went from a negative to positive correlation.

Where to Go From Here

In the next few days, we'll be looking at some of the most significant correlations to party vote. We will also begin drawing up some rudimentary election projections and discussing some flaws with some of the existing models out there.

The Challenges of Creating Data for Ontario

It's not easy getting the numbers you need. Any fool can go into Excel and start typing stuff, of course, but how do you get methodologically sound data to use in your projections and other number-crunching pursuits?

Several years back, Canada (or at least certain politics and public-policy circles) had a big public debate on research methodology. It was a reassuring reminder of the elevated state of our public discourse (we had to justify knowing fewer things, cf the country to the south of us), but the result - a sabotaged 2011 census - was very disappointing indeed. For that reason I've completely dismissed using the data from that survey here and on Québec Projections.

Also disappointing is the fact that Elections Ontario, as far as I can tell, doesn't publish reports on provincial electoral ridings, as does the DGEQ (Québec elections officials helpfully provided me with an Excel file of their socio-economic reports upon request, which they've since made public). I've bypassed this problem for Ontario by digging into the Federal Electoral District profiles, which use the more robust (but dated) 2006 Census data. Since Ontario's current provincial riding boundaries are more or less the same, I'm using the federal data as a substitute.

The Northern Ontario ridings, however, have kept their old boundaries from the 1996 Federal Representation Order. Or, alternatively, the 2005 Representation Act (Ontario) incorporated the 1996 boundaries by reference, with minor modifications. So I've gone back and looked at the old profiles from the 2001 Census, using the 1996 federal boundaries. I've found the corresponding fields and calculated new ones, and I'm plugging them into the data from the newer profiles. 

Unfortunately, that leaves some gaps. We don't have religion data for 2006 and beyond, which was interestingly left out of the 2006 Census. Going backwards, we're missing the interesting data on sources of income, industrial categories and so on for 2001.

You don't have to be familiar with Geocities, but it helps

 Doesn't that bring into question the changes that have taken place since 2001 and 2006? Won't these changes make the data less valid? Maybe, is the answer to that. The 2000s did see a lot of immigration to Ontario. We're hoping that the fact that most of the data came from 2006, and that voting data came from 2007 and 2011, will keep our analyses relevant overall. The US, after all, uses a much less detailed census, and only every 10 years.

For Northern Ontario, the outmigration, or the stability of the population since 2001 that led to their losing a riding (if it hadn't been for an legislative intervention) probably mean that Northern Ontario has become an even more exaggerated version of whatever patterns make up Northern Ontario. Not really knowing too well myself, at this point let's pretend I can guess that Northern Ontario is older, whiter, more working class, and more francophone than the rest of Ontario (or ROO -- today's candidate for worst acronym). If immigrants and young people have been leaving / not moving to the North since then, the tendencies described by the lagging-behind data will probably be rather conservative.

We can, of course, also run analyses for data that exclude Northern Ontario.

The next step is to see how much the federal riding boundaries have changed in Northern Ontario from 1996 (i.e. the boundaries today in the North) to 2003 (the most recent FED profiles available for that geographic territory).

It's not a perfect solution, but with more fluidity in the definition of ridings likely to come in the next few years, as well as the provinces' distrust of Federal data, we might see Elections Ontario decide to publish some data for the benefit of parties, the media, academics, as well as the all-blogging, and tweeting Peanut Gallery. Or we can build them ourselves, census tract by census tract.