Monday, May 24, 2010

Fun With Numbers: Liberal Vote

Last summer, I spent a good deal of time playing around on my election spreadsheet - if you scroll way down on the sidebar, you can see some of the posts under the "Fun With Numbers" heading.

But the one thing I never got around to looking at was the link between riding demographics and the vote. Let's face it - certain types of ridings certainly seem like they "should" be Liberal, while others feel like they "should" be Conservative.

To explore this, I took the mean vote for the Liberals over the past three elections as the response and looked at 25 Census variables as possible predictors. Everything from single parent families, to resident mobility, to immigrants. I did some transformations where necessary, tossed in some region variables and population density, and threw it all at the regression wall to see what stuck.

The point of this wasn't necessarily to find what types of people vote Liberal. After all, states with a large percentage of African Americans tend to vote Republican (not because black voters flock to the Republicans). The purpose was to find the types of ridings that tend to vote Liberal.

The final model I came up with used 8 variables and some of the results aren't too surprising - the Liberals perform better in well educated ridings with more first generation immigrants, and worse in the west.

What's interesting (well, to me) are some of the variables that don't make the model - namely, the "Atlantic Canada" and "Quebec" dummy variables. In other words, once you control for demographic variables, Liberal support is largely consistent east of Manitoba. Population density also didn't make it into the model, but only because there are better demographics out there to identify the urban/rural split.

All in all, the model accounts for 72% of the variance in the vote- not bad when you consider all the unknowns that impact voting behaviour.

So, technical jargon aside, what use is this to us? Well, for starters, we can see where the Liberals have "overperformed" in recent years:

1. Wascana (+32%) - King Ralph's seat
2. Yukon (+22%) - I don't really expect northern seats to fit nicely into a model like this.
3. Winnipeg South Centre (+21%) - This riding has a strong Liberal history, since Lloyd Axworthy took it in '88.
4. Edmonton Centre (+17%) - We saw what happened when landslide Annie left last election.
5. Sainte Boniface (+17%)
6. Brome Missisquoi (+16%)
7. Cardigan (+15%) - The Liberals do better than expected across all PEI ridings
8. Esquimalt Juan De Fuca (+15%) - Clearly, Keith Martin brought some Conservative voters with him when he crossed the floor.
9. Shefford (+14%)
10. Vancouver Quadra (+14%) - The bad by election isn't included in this sample...

Wascana stands out like a grain elevator on the Saskatchewan skyline. Over the last three elections, the Liberals have over performed their expected vote by 32 percentage points in this riding - this is really a testament to Goodale and speaks to just how difficult it will be to hold the riding when he retires.

A quick scan of the rest of the list shows that ridings with long time and popular Liberal MPs usually do over perform their expected results - the ridings of Scott Brison, Maurizio Bevilacqua, and Dominic LeBlanc, for example, are all in the +7% to +12% range. Now, this doesn't mean the grits are necessarily doing better there because of those MPs...but that's certainly a possibility.

On the other side of things, where can the Liberals grow?

Well, they under perform the most in Central Nova...although that may have something to do with them not running a candidate there last election. After that, it looks like there's the most potential for long term growth in Windsor West, Outremont, Laurier-Sainte-Marie, Jonquière-Alma, Carleton-Mississippi Mills, Rosemont-La Petite-Patrie, St. John's East, Roberval-Lac Saint Jean, Sackville-Eastern Shore, and Hamilton Centre.

And since this blog is still called Calgary Grit, I'd be remiss if I didn't look at Alberta. There, the news isn't great - there isn't a single riding the Liberals are projected to be within 20 points of winning. So it's not just the myth of the NEP that's killing the Liberals in Alberta; once you correct for the Liberals' poor performance across western Canada, the demographics really aren't there for them.

There is, however, a bright spot. The regression models gives the NDP and Greens a lot of votes in some of Alberta's downtown city ridings - as a result, the projected CPC vote in Edmonton Centre, Edmonton East, Edmonton Strathcona, Calgary Centre, and Calgary North Centre hovers in the 40% range. We've seen what a strong NDP campaign in Edmonton Strathcona can do and we've seen what a strong Liberal campaign in Edmonton Centre can do.

A concerted, long-term focus on these ridings, capped with a good candidate and a great campaign, and any of these seats could eventually fall - the votes are there, the voters just need to be given a reason to vote the way they "should".


Post a Comment

<< Home