Thursday, June 2, 2011

Canada's Partisan Voter Index

If you don't know what PVI is, that's OK - you can be excused for it. It's not something widely discussed up here, since it's an American political tool.

The PVI, or Partisan Voter Index, is a creation of American stats whiz Charlie Cook. The Index rates all congressional districts based on their average variation in partisan leanings, based on how those districts vote during presidential elections.

For example, if a district voted for a Republican presidential candidate with 55% of the vote, while nationwide that candidate got 50% of the vote, the Index rates the district as 5% more Republican than the rest of the nation.

This is an oversimplification, but you can look at the results here. Cook's PVI is widely recognized and respected because it paints an interesting picture for trends; if this district is leans more Republican, it is more likely to vote Republican in congressional elections. The exceptions tend to be if it's a low PVI score, or if there is a popular local candidate. In this way as well, it's interesting to see which congresspeople buck trends in districts the other party has a higher PVI score in. It's also very predictive of how some of them may vote in the House, as this unrelated article by 538 shows.

So why doesn't Canada have something like this? The obvious answer is that we don't have a presidential system, so it can't work the same way. However, it doesn't mean we can't still do it. Which is why I present the below chart:


You may be wondering what you're looking at here.

These are Manitoba's 14 ridings, arranged alphabetically, with their current MP's names, their partisan affiliation, and then my version of the Canadian Partisan Voter Index - though should change the name.

How I've figured this out is pretty simple. Taking the average difference over four elections between a party's results in a specific riding as compared to that party's province-wide results, you get the numbers you see (with rounding).

So over four elections, the voters of Brandon--Souris have favoured their Conservative candidate by 11 percentage points more than voters as a whole in Manitoba have voted for the Conservatives.

If you look at the actual results for Brandon--Souris, this seems right; in 2004, votes in Manitobans gave 39.1% of their votes to the Cons, while voters in Brandon--Souris gave 51.7% of their votes to the Cons. In 2006, the split was 42.8% province-wide, and 54.4% in Brandon--Souris. 2008 was 48.8% to 57.1%, and 2011 was 53.5% to 63.7%.

So if in 2015 the Conservatives get 40% of the vote in Manitoba, expect them to get somewhere around 51% of the vote in Brandon--Souris, barring any major changes (like a split!).

An interesting case would be to look at Winnipeg North, which the PVI gives the New Democrats a 26-point advantage, though Kevin Lamoureux, the Liberal MP, has won the riding anyways.

Or, another interesting one is Charleswood--St. James--Assiniboia. Why are the Liberals here giving a PVI of 6-points, but in 2011 they ended up third with only 18%, to the Conservative Party's near-60% of the vote?

The interesting fact behind this is that in Charleswood, the Liberals end up outperforming their provincial numbers by at least 2 points. It drifts closer to 6% because of the candidacies in 2004 and 2006, which were much closer races where the Liberals outperformed their provincial by 10 points. Hence, it's demonstrated that the Liberals do have serious strength in the riding, but given their low scores in the province, they're unable to capitalize on it; and given that the Conservatives have a PVI of +5 points in Charleswood, and their sky-high numbers regionally, they waltz in unopposed. However, for purposes of simplicity, the party with the highest PVI gets first spot.

(That being said, there is an argument that one should rate the older votes lower than the newer ones, which would bring down the Liberal PVI - possible, but let's keep it simple for now.)

I think this is a fair enough ways of doing things. After all, if a party is bucking its regional trend, that must be a good sign that this riding is one that's in their favour, right?

I would do this for all 308 ridings if I could, but that would take an inordinate amount of time - all summer, at least. Might be a project worth pursuing, but we'll see. I wanted feedback first, so tell me - what do you think? Can the Canadian PVI work?

(Fyi, if it does, I probably need to change the name; pretty sure Charlie Cook doesn't like people stealing his names for his creations, even from a small-time blogger from Canada - you know how these people are)

1 comment:

  1. This is interesting. They say it's not important to find out that you're losing or winning, but the reasons why. (This may fall apart in Quebec with the NDP's surprising rise...) This might be fun to fiddle around with....

    ReplyDelete