Friday, April 12, 2013

CONVENTIONAL WISDOM CHECK -Leadership Bounces

As both federal opposition parties head toward convention weekends, pundits are rightly discussing the significant bounce that Justin Trudeau's looming victory has given the Liberal Party of Canada. By any measure, his impact has been impressive. The Liberals, previously in third place and surrounded by carrion eaters in the eyes of some observers, have rebounded to take the lead in most credible surveys.

This blog post will be a bit different than many of the ones on this site. Exercising some measure of discipline, I will offer no comments on whether Trudeau's support is a good or bad thing. Instead, I am going to indulge my inner academic geek and try to offer a statistical look at the durability of leadership bounces. I've always been taken by Paul Wells' advice on assessing conventional wisdom -- that when everyone in Ottawa knows something it is usually wrong. Something I'm going to start doing in this space is taking some of these easily-accepted truisms and seeing if they hold up under the numbers. I don't fancy myself a Nate Silver, of course, but just as I can enjoy playing basketball and not be Kevin Garnett, this may prove fun for my fellow geeks.

What I've done is take a look at the polling bounce for leaders chosen after Pierre Trudeau beat Joe Clark in 1980. Happily, Environics has a polling archive housed at Queen's University that allowed me to take the polls immediately preceding and following a new leader's convention win.

Here, in descending order, are the new leader bounces:

John Turner, Liberal (1984) +7.7%
Stéphane Dion, Liberal (2006) +6.7%
Michael Ignatieff, Liberal (2008*) +6.4%
Stockwell Day, Canadian Alliance (2000) +4.7%
Paul Martin, Liberal (2003) +2.8%
Audrey McLaughlin, NDP (1989) +2.4%
Kim Campbell, Progressive Conservative (1993) +2.1%
Brian Mulroney, Progressive Conservative (1983) +2.1%
Alexa McDonough, NDP (1995) 0%
Stephen Harper, Conservative (2004) -0.4%
Jack Layton, NDP (2003) -1.0%
Peter McKay, Progressive Conservative (2003) -1.9%
Jean Chrétien, Liberal (1990) -1.9%

First, a few notes on the data. The bounces here likely seem small. Because Environics provided the best archive where I could actually compare the same polling firm and method, I've accepted some of their constraints. One of them is that these are numbers without the undecided and leaning factored in. The reason polls generally add up to 100% is because the top line numbers involve some polling science of reallocating those who don't answer. Environics numbers in this archive are their raw percentages of who actually gave a party preference, so the numbers are smaller. These bounces likely got reported as larger ones on the top lines.

As well, using a consistent "closest to the date" rule may understate some of the bounces. If I used a less controlled method of searching for the largest reported polling bounce, both Campbell and Martin would do better, as would Day to some degree. Both Campbell and Martin had some polls show a Trudeauesque bounce of 15 points or more. As well, because they were both inevitable and ubiquitous front runners for a while, their bounce built even bigger over a few polls. However, this would be a less consistent comparison, and moving those two leaders up wouldn't change what you likely saw as most notable about these numbers. But if you want to anecdotally recall Campbell and Martin as even more successful than this, go ahead.

Finally, a couple of technical points. I had a choice for Ignatieff, whether to use the date Bob Rae and Dominic Leblanc withdrew and he became interim leader, or the date where a convention ratified this choice. I chose the former, since that was when he moved into the Opposition Leader's chair and was publicly seen as starting his leadership. Also, I could not fix a date for Jean Charest's more organic ascension to head of the shell-shocked, two-person PC rump (he likely was seen as leader by the public the next day), and thus excluded him.

So, what do we notice here?

Obviously, the most striking thing is that the size of the convention bump provides us with a list where the most successful leaders seem clustered at the bottom, and some of the great cautionary tales of politics seem to be at the top. Chrétien, Mulroney and Harper are the only guys who have kept winning, and Layton had a historic breakthrough. Even McDonough saw a modest improvement in NDP fortunes, and we never got to evaluate McKay. The guys at the top imploded.

I recall that at the New Brunswick Liberal convention one disappointed onlooker, given the consolation that people wanted a new, untainted face, replied "yes, because heaven knows the Liberal Party has never done well with an older, experienced guy who was in the cabinet of an unpopular government". And, indeed, Jean Chrétien was all those things and paid for it in his post-convention bounce. Yet his admonition to the "Nervous Nellies" in the party fretting about the lack of instant gratification paid off quite nicely with three straight majorities. It does seem clear that the standing of a leader right after the convention is a poor predictor of electoral success.

I would caution against the logical trap of trying to explain that certain bounces can, in hindsight, be "explained". Yes, Dion had a charming underdog win driving him. Yes, Chrétien had the Meech Lake debate raging when he won and took a beating for his past. Yes, Harper suffered from the post-convention loss of more progressive Stronach supporters. And yes, Martin took the sponsorship fallout right in the proverbial shorts. But this exactly the point -- the standing both at the convention and on Election Day are influenced by events and how the candidate handled them. The cold numbers simply tell us how the early predictive verdict panned out.

What you choose to make of this is an open debate. There are two plausible explanations that I can suggest, and I cannot prove or disprove either by the numbers.

The first suggestion is that electoral bounces are simply unrelated to eventual success. In this explanation, the real predictors of electoral success are more fundamental predictors -- a party's brand popularity, organization, fundraising and policy clarity/appeal all matter. An appealing face probably does help, but if they don't have the fundamentals then they are eventually gong to be weighed down. This explanation would ask you to accept that the larger bumps for unsuccessful leaders are simply because their party's weak fundamentals gave them more room to improve (Campbell and Turner followed very unpopular incumbents, Day took over a new party, etc).

The second plausible explanation would be the more intriguing one. If convention bounces are actually a negative predictor of electoral success, it would suggest that party electors make choices on the wrong factors. In this scenario, you would believe that the things that make leadership candidates appealing in the short-term are actually the wrong criteria.

Advocates of this conclusion might cite that the factors that produce leadership enthusiasm often involve voters and skills that are least helpful to winning elections. The qualities a leader needs to create a bounce must be ones that cause at least a short-term change in how the party is seen. By definition, this will favour "new" faces who will be least associated with the party brand, have the fewest unpopular decisions on their record, and are thus the most "different". This may mean that skills the Chretiens and Harpers and Layton's acquired along with the baggage of unpopular decisions and election losses go unnoticed because they are not different enough to quickly create a change in voters' views of the party. Yet these very skills -- a grasp of policy, an intuitive sense of how to avoid trouble, a knowledge of the election terrain -- serve them well in the long run.

It may also be that polling booms exist because new faces move the most volatile voters. When Colin Powell was flirting with a 1996 presidential run and had a 70% approval rating, one Clinton strategist remarked "I can take him down 20 points in one day. Ask him whether he's pro-choice or pro-life". The point is well-taken. It may be that new faces hold their appeal because they are least like a politician, and this appeal to swing voters moves party elites to chase this appeal. Yet once elected, a new "not-a-politician" will immediately begin doing things that look, well, like a politician. The Harpers and Chretiens may not attract these more volatile voters immediately, but the support they have is more durable because it has been tested by actual decisions and past mistakes. And the swing voters will simply choose the guy who campaigns best when it counts. In this telling, parties were lucky that a cabinet minister in a rejected government, a failed Toronto mayoral candidate, and an uncharismatic former MP had the organization to win as unlovable favourites, because otherwise they may have made the mistake they made chasing fool's gold in more open races.

Now, lest I be accused here of some wish fulfillment, let me make something clear. The numbers don't lie, but they also don't predict with absolute certainty. Just because something hasn't happened before doesn't mean it won't. Teams down three-games-to-zip in the playoffs don't win, unless you're the 2004 Red Sox. Insurgent presidential candidates don't beat establishment favourites, unless you're 2008 Barack Obama. And just because exciting new faces haven't proven to be as solid a bet as veterans grizzled by controversy and defeat doesn't mean they'll never win. It may be that someone out there combines the sizzle of Stock Day with the substance needed to win. And as anyone who ever watched my wife win the weekly football pool by choosing uniform colours knows (yep, she did), sometimes bad bets pay off.

In fact, if I'd had polling data from 1968, I'd have found a leader who had style and who turned out to make a pretty big impact on the country. A fellow called Trudeau. I could compare his readiness to the new guy's, but I'd break my promise at the top of the post to avoid opinion.

Lets just say that maybe, this time, everyone knows something. They just don't know it for the reasons they think they know it.