Thursday, October 14, 2010

Federal vs. Provincial Results in New Brunswick, Part Deux

Yesterday, Alice Funke of the Pundits' Guide wrote an article pointing out a few of the concerns she had with my piece yesterday. In that piece, I compared the federal and provincial election results in New Brunswick since 1978, hoping to find some sort of trend that could tell us what to expect in the next federal election.

But before responding to her piece, I think it is necessary to talk a little bit about the philosophy behind each of our two sites. The Pundits' Guide is, bar none, the best political reference for Canadian politics available online. Alice does an incredible job compiling all of the data and her site makes it possible to analyze that data in any number of ways. She also posts on timely topics with great insight, and keeps us up-to-date with riding nominations, by-elections, and financial data for the political parties.

While The Pundits' Guide is about historical election results and the most current information about Canadian politics, ThreeHundredEight.com is about what's next. I cover polls and make projections, and now and then I look at tangential issues from a numbers point-of-view. When I look at those issues or reach back into history, as I did yesterday, it is usually as an attempt to find out what we can about what will happen in the future. ThreeHundredEight.com is about the hypothetical, but based on the best data available.

While our two sites are related by subject matter, our content is very different and our approach is, I'd say, different as well. That is reflected in what I focused on in my post about New Brunswick's electoral history and what Alice did in her critique of it.

So, now that the preamble is out of the way, here is my friendly rebuttal.

First, one of Alice's points of contention was that I used vote-share rather than actual votes. In other words, instead of focusing on how a party got 50 out of 100 votes in one election and 25 out of 50 votes in another, I've focused on the fact that in both cases the result is 50%.

It's a fair point. Using actual votes would have given me a more precise result in terms of comparison, especially when speaking about individual voting behaviour. But when I change the chart to reflect actual votes rather than vote share, it doesn't really tell a different story. Here are the two charts for comparison:
To my eyes, these charts are virtually the same, and in fact using actual votes makes the provincial and federal results mirror each other more closely. I will, however, admit that I should have gone this route as it is more indicative of what has happened in New Brunswick over the last 32 years.

Second, Alice did not think it was proper to combine the results of the Confederation of Regions with those of the Progressive Conservatives at the provincial level, and the results of the Canadian Alliance/Reform with those of the Progressive Conservatives at the federal level.

She's correct that combining the results of these parties simplifies matters and tells only half of the story. But that was my intent. My post was not meant as an appraisal of New Brunswick's voting history, it was meant as a means of looking into the future based on what has happened in the past. Does the performance of the Confederation of Regions or the Reform Party tell us anything today? It doesn't, other than that a populist party can gain votes in New Brunswick. Unless a new populist party appears at the federal level, there is nothing to gain from separating these results from those of the Conservatives.

They do, after all, draw from the same pool of voters. As Alice points out, the Confederation of Regions did take votes from both the Liberals and Progressive Conservatives in 1991, but that is more likely because the 1987 election was an aberration. In that election, New Brunswickers ousted a long-time PC premier who was mired in scandal. Those PC voters who voted in Liberal in 1987 are almost certainly the same voters who cast their ballots for the CoR in 1991. If anything, it is probably unwise to read too much into the Liberal performance in 1987, which actually works out fine as the federal performance in 1988 was unimpressive by comparison.

Third, Alice felt I was misguided to not include the Greens in my analysis. Again, this was because of the purpose of my article. I only wanted to look at the Conservatives, Liberals, and NDP because they are the only parties in play in New Brunswick. And since there was no provincial Green Party before this 2010 election, there was nothing to compare the federal Green results to.

Fourth, and this is a minor one, was my use of the word "correlation". She interpreted it in the statistical sense, while I only meant it in the linguistic sense. The statistical definition of "correlation" is fair more restrictive than the English definition of the word.

Fifth, she took issue with my statement that the same people were voting either Liberal or Conservative (in all its forms) election after election. I'll submit that it was a simplification, but virtually every Liberal gain is matched by a Conservative loss, and vice versa. This is visible in both the vote share and raw vote charts. While a few voters are certainly switching their votes from NDP or Green to the Liberals or the Conservatives, these are undoubtedly a small minority.

Sixth, and this is perhaps the most important, is her disagreement with my conclusions. I concluded that, based on the trends visible in the charts, the Conservatives should gain in the next federal election while the Liberals should lose. Alice called these predictions but I didn't mean them as such. I was taking the next logical step from my original conclusion that provincial and federal voting behaviour in New Brunswick generally follow the same pattern. I was only forming a hypothesis based on what I saw as a trend in the chart.

The purpose of my post was to compare the voting behaviour of Liberals, Conservatives (all branches), and New Democrats at the provincial and federal levels. I've discovered in other provinces that these don't always match, and sometimes they differ by a significant amount. It doesn't seem to be that way in New Brunswick. It goes without saying, however, that I would not base my projections of future electoral results in New Brunswick on this little exercise.

3 comments:

  1. Hi again Éric.

    I'll correct you on one point. I believe the correct metric to use is votes expressed as a percent of the electorate, and also expressing non-voters as another "party" in the same way.

    I would have used that metric myself, but then I could not have mirrored the timeframe you selected, because the number of electors for New Brunswick only for the 1979, 1980 and 1984 general elections does not seem to be available online. Thus the number of votes was merely second-best from my perspective. But even there, the raw votes methodology shows the federal Conservatives as being flat between 2006 and 2008, while vote-share shows their line moving up quite sharply. Given that the most recent results are probably also the most influential in determining trends, that's a key difference.

    I guess we'll have to agree to disagree on what the emergence and disappearance of ideologically based small parties in the past and present tell us about the fortunes of the two old parties going forward.

    But it's the issue of saying an election doesn't count because it's an "aberration" that always leaves me wanting more from a seat prediction approach. A half to two-thirds of seats in Canada will likely not change hands in the next election, and they can be easily predicted. What's more interesting to me are what variables could be used to determine the potential for swings and which way the swings would go.

    As to your fifth point, if you had limited your scope to the provincial realm only, even making the decision to include CoR and PANB with the PCs, then your conclusion would have been correct. It was broadening that out to include the federal scene that I think was less safe. The federal and provincial Liberals do mirror one another. The federal NDP shows up about 10 points higher than the usual provincial NDP result, at least since 1997. So obviously some provincial PCs are voting NDP federally (a simplification perhaps, but the Occam's razor explanation).

    Pulling trends from charts is hard when there aren't many data points, which is why I focused on your decisions about which data points to include. We could run some regressions, but I don't know if the proper fit is linear, curvilinear, or given what we know about political cycles, maybe it should look more like a Sine Curve.

    In any event, let's continue to keep each other sharp. It certainly perks up an otherwise grey fall day, doesn't it.

    Cheers, and keep on trucking!

    ReplyDelete
  2. My error, I think, was reading too much into the results. Perhaps I should have limited myself to a comparison, and left the conclusions to the individual reader.

    I will be doubly careful next time, thanks!

    ReplyDelete
  3. You see, I'm not so sure about your "conclusions" either, Eric, though it would help explain high Conservative totals in Atlantic Canada.

    Yet, I don't know if you could claim that the federal Liberals would go down lower because NB Liberals did, especially considering that you can claim the NB Liberals followed the path of Dion's Liberals in 2008's election. And the likelihood is that Alward's government is going to stumble, and the Liberals will rise in the polls again eventually, if not sooner rather than later, if Alward falls short. If the provincial vote in polls rises, will not the federal vote as well?

    ReplyDelete

COMMENT MODERATION POLICY - Please be respectful when commenting. If choosing to remain anonymous, please sign your comment with some sort of pseudonym to avoid confusion. Please do not use any derogatory terms for fellow commenters, parties, or politicians. Inflammatory and overly partisan comments will not be posted. PLEASE KEEP DISCUSSION ON TOPIC.