Tuesday, November 16, 2010

The value of star candidates

I was once told by a politician that the local race represents about 5% of the vote - everything else is influenced by the on-going campaign and historical voting patterns.

I thought over how I could test this hypothesis, but there is no way to conclusively look into this matter with the resources at hand. But one way to get a hint of it is by looking at the influence a star candidate has in a local election campaign, so several weeks ago, I asked my readers for some help on identifying some star candidates who ran for election in a given riding for the first time in 2006 or 2008.

Aside from other local issues and the on-going national or regional campaign, a star candidate can play a large role in making voters shift from one party to another. It can be easier to identify with an individual, particularly one you know from your neighbourhood or one that you've regularly seen on television or in the news, than a national party. It also makes it easier to vote for someone, rather than voting for a party's local candidate that you've never heard of or seen before.

I embarked on this little analysis wanting to compile a list of "star candidates" that first stood for election in a riding in 2006 or 2008. We can really only use the last three elections to tell us anything about what might happen in the next election, due to the merging of the Progressive Conservatives and Canadian Alliance. And since using the 2004 election would force us to look back into the 2000 campaign, I could only use the last two elections.

I would take these star candidates and see how they did compared to the previous election. If a star candidate increased his party's vote by 1,000 people, that may act as an indicator of the star candidate's pull. But in order for this analysis not to be too tainted by the on-going campaign, I also compared the change in vote from one election to the next election in each riding to the wider changes in each province. So, if a star candidate increased his party's vote share from 20% to 40%, but the party also doubled its vote provincially, I would assess the star candidate's pulling power as a factor of 1. In other words, no drawing power.

However, if the party only increased its vote share provincially by half while the star candidate doubled it in his or her riding, I would assess the star candidate's drawing power as a factor of 1.33 (actual vote (40%) divided by expected vote (30%) = drawing power (1.33)).

My method of choosing a star candidate was based on my own personal judgment, and I assembled a sample of 38 candidates. Star candidates included former provincial party leaders, provincial cabinet ministers, former federal politicians or ministers, actors, mayors, journalists, and well-known local notables and professionals.

My choice of star candidates is, obviously, the least objective part of this analysis. But, of course, my results are to be viewed not as gospel but as an indication of what a star candidate's drawing power can be, based on the parameters I've outlined.

Before getting into the results, here are a few selected "star candidates", and how they performed.

First, Maxime Bernier, who ran as a Conservative for the first time in 2006 in the riding of Beauce. I remember not knowing who he was, but that everyone in the Beauce knew who he was because of his father. Bernier took the party from 8,091 votes and 17.1% in 2004 to 36,915 votes and 67%. He increased the party's share of eligible voters from 9.9% to 45%, a huge accomplishment. But, the party did well in Quebec in general in 2006, and based on that provincial increase we would have expected the Conservatives to get 47.9% of the vote in Beauce. As Bernier received 67%, he gets a drawing power factor of 1.4.

But not every son-of-a-well-known-person does well. Justin Trudeau, in the Montreal riding of Papineau, did see his party's vote go from 16,785 in 2006 to 17,724 in 2008. That was 38.5% of the vote compared to 41.5% in 2008. But based on the Liberal Party's growth in Quebec as a whole in the 2008 election, Trudeau should have gotten 43.9% of the vote. So, his drawing power is rated at 0.95 for a negative drawing power.

The experience of Marc Garneau, however, demonstrates how an individual's "drawing power" can be inconsistent. In 2006, Garneau dropped his party's support in Vaudreuil-Soulanges from 38.8% to 28.4%. But he should've seen his support go down to 23.7%, giving him a drawing factor of 1.2. On the other hand, in Westmount-Ville Marie in 2008, he only increased his party's share of the vote from 45.7% to 46.5% (though that was a loss of about 800 voters), instead of the expected 52.1%, giving him a drawing factor of 0.89.

Here are the drawing factors of a few selected individuals:

Michael Fortier (CPC) - Vaudreuil-Soulanges - 1.42
Lawrence Cannon (CPC) - Pontiac - 0.54
Jim Flaherty (CPC) - Whitby-Oshawa - 1.09
Michael Ignatieff (LPC) - Etobicoke-Lakeshore - 0.98
Tony Clement (CPC) - Parry Sound-Muskoka - 1.00
Paul Summerville (NDP) - St. Paul's - 1.14
John Baird (CPC) - Ottawa West-Nepean - 0.99
Jean-Pierre Blackburn (CPC) - Jonquiere-Alma - 3.89

Now, to the overall results.

On average, star candidates for New Democrats averaged a drawing factor of 1.27. The result was 1.26 for the Conservatives and 1.27 for the Greens. The Liberals, unfortunately, broke the consistency with their 1.06.

If we dropped the best and worst results for each party, we got 1.27 for the Greens, 1.26 for the New Democrats, 1.14 for the Conservatives, and 1.04 for the Liberals.

Former politicians alone averaged a factor of 1.26.

So, there appears to be some degree of consistency. On average, a star candidate will actually help a party's standing.

In 2006, the average factor was 1.25, or 1.11 if we take out Blackburn's anomalous result. In 2006, it was 1.16.

Overall, that gives us an average factor of 1.21 with Blackburn, and 1.14 without.

The chart on the left gives an indication of the consistency and inconsistency of these results (blue are NDP, green are Conservative, red are Liberal). Each dash represents one candidate. As you can see, there are a few outliers at both ends of the scale, but most results are clustered between 0.9 and 1.3. If we were looking at all candidates from all parties, we'd expect a cluster centred around 1 rather than 1.2.

So what does this analysis tell us? Certainly, that there is no hard-and-fast rule for the drawing power of a star candidate. Some do better, and some do worse. But most do better, increasing their party's support by about 14% over and above what we would expect them to get. In general, they're worth about three to six percentage points, indicating that the "5% is the local race" rule might not be so inaccurate.

The next thing I need to mull over is whether to include this factor in the projection, and how. Should I use the 1.21 or the 1.14 that excludes Blackburn's result? Should I use a lesser portion of this factor to act as an indicator for probable outcomes but leaving a good deal up to the wider campaign? As I expand the projection into all 308 ridings and include a great deal of many different factors, each of which I will try to analyze in posts like this, I will have to ask myself these questions more and more.


  1. Maybe you should try and correlate this star value against the margin of victory instead/as well. Because I think you've underestimated Justin Trudeau's "star power" in a very competitive race against the Bloc. Without him running, the Liberals might have lost more of their support there and failed to win the seat at all.

    Also your math treats an increment from 50 to 51.5% the same as an increment from 33.5% to 35%, if I understand it correctly. Whereas the latter may be more significant an accomplishment if it meant gaining a seat.

    Food for thought anyway.

  2. Alice,

    Your margin of victory point is a very good one. It might become a problem, though, if there is another "star" candidate in a riding. Trudeau's is pretty clear cut, but if there is another riding where there is a popular Green candidate eating into the NDP's second place running while the Conservatives have a star candidate as the favourite, it gets a little murky.

    I was also considering going deeper than the provincial level by looking at how they performed compared to regional shifts.

    My math is looking at proportional change, so a shift from 33.5% to 35% would be the same as a shift from 50% to 52.2%. I agree it means a little more when it comes to winning a close race, I think looking at it the way I have mitigates this factor.

    The main goal of this exercise was to calculate a factor I could incorporate into a riding projection model. As I want to use uniform swing as the foundation of the model (and I know you have some issues with that), my analysis was geared towards that angle.

  3. 1.14 (or 14% in my terms) seems very reasonable, and I will credit you, if it's okay, and use this myself.

  4. Very interesting analysis, Eric. I think you did very well in the math department on this.

    Overall, however, I'd advise against putting it into any sort of projection system, simply because of the fact that "star candidates" are few, far between, and range across such a large scale and in such a large range of ridings, that it'd be difficult to nail down any sort of proper effect in a national or even regional projection. I could see it per specific riding, but only as a baseline.

    However, you could maybe do something similar with the party leadership. For instance, even though all the polls and projections back in 2008 pegged the Liberals sitting around 90 seats and 28%, they failed to realize the low-turnout effect of Dion's low leadership ratings, which brought them down to 77 seats and 26%.

    I'm fairly certain that using historical and statistical data, you could draw up a similar chart and data. Do you think its possible?

  5. Volkov, the effect of star candidates would only be applied for individual ridings. The model isn't quite at the riding level yet, but it will be eventually, and I'd like to use the results of this analysis at that time.

    As to your second question, it could be possible, if there is leadership polling data available for the 2006, 2004, and preferably earlier elections.

  6. I wonder if there wouldn't be value in doing this calculation the other way round: namely, look across Canada for ridings where the 2008, and 2006, results deviated significantly from what you'd expect looking only at historical patterns and national trends. I mean a riding where a uniform swing would expect Party X to get 10% but the party actually got, say, 25%. Having found such ridings, you could look for factors, particularly regarding the candidates in question, that might account for them. Other things you could look for might be residency in the riding, previous occupation - and of course money spent on the campaign.

    Additionally, I wonder how much 'star power' deviates from province to province. Are people in one province more dazzled by 'celebrity' than others?

  7. A "star" candidate may be a person who has shown great competence in some field. Or they may just be a person who's well-known. (Ahhnold, anyone?)

    So... what's the relationship between the "star factor" and simple incumbency? I doubt the answer is a soundbite.

  8. That will be the subject of a future post.

  9. For predictive purposes, I would say that there is a difference between "local" star candidates and "national" star candidates.
    Maxime Bernier and Jean-Pierre Blackburn were not "stars" nationally but were well-known locally and so got more votes there. Justin Trudeau and Marc Garneau were well-known nationally and therefore were "stars" for the Liberals' national campaign (though it would be hard to measure), but they had no particular connection to Vaudreiul, Westmount or Papineau, so it's not surprising that they didn't increase their party's vote totals by huge numbers in those ridings.

    Michael Fortier was a "national" star with little connection to Vaudreuil. The only reason he got more votes for the Conservatives is because there is a large angolphone population in Vaudreuil prepared to vote for whatever federalist candidate was perceived as having the best chances of beating the Bloc candidate.

  10. I guess we have a different definition of 'star candidate' on some of yours... there are regional 'stars' who, while famous in a community or even provincial jurisdiction, would only raise question marks outside these boundaries -- and i'd argue about fortier, blackburn, the London mayor for the Tories, etc. Then there are star candidates whose credentials transcend the region and/or politics, ie. Dryden, Thomas Steen, Garneau, Laroques and Cutler. I'd be interested in seeing how these kinds of 'stars' raised the bar. Fantino would definitely also fall in this category.


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.