Swing batta swing

Needless to say, the range of potential outcomes in the Alberta election (along with the continued flow of news battering the Prentice PCs as they try to regain some type of footing) has made for a fascinating campaign. But it's worth pointing out that single polls and seat projections may miss important parts of the picture - meaning that the actual state of the race is far less certain than it might appear at first glance.

Take for example this important explanation of Election Almanac's methodology:

Election Almanac uses a proportional swing model for its projections based on the latest election poll results. For example, if a poll gives a party a greater (multiplied by a factor greater than 1) share of the vote than they received in the last election, the projector assumes the party’s vote has gone up by the same proportion in every seat. Conversely, if a poll gives a party a smaller (multiplied by a factor less than 1) share of the vote than they received in the last election, the projector assumes the party’s vote has gone down by the same proportion in every seat. No projection model is 100% accurate. 

Now, a presumption that all seats will see identical swings between parties might make for a useful initial assumption. But the last Alberta election was based on a radically different set of party strategies and goals.

In 2012, the Redford PCs veered left in order to paint themselves as a relatively moderate alternative to a perceived Wildrose government-in-waiting - so most of the province's seats were decided by the allocation of a two-way split among 78% of Alberta voters. Meanwhile, the NDP's primary focus was on holding onto beachheads rather than expanding its support throughout the province.

Today, we can say with some certainty that the situation has changed. Rather than fighting primarily in the few most friendly Edmonton ridings, the NDP is looking to sweep the city, while the PCs are turning to the right and chasing a larger share of right-wing voters to try to stop the orange wave. And the rest of the campaign will determine which seats actually are the swing ridings - with surprises likely to pop up along the way.

For all the news in the headline poll numbers, though, there's relatively little means for the public to figure out exactly how those will translate into seats. And that goes doubly in a three-way race where (for example) ThreeHundredEight's seat projection sees the PCs' seat count potentially ranging from 5 to 31 based on a swing of under four points in popular support.

In theory, it might be possible to try to use more sophisticated means to generate seat projections - for example, by also comparing regional and seat-based polling to the 2011 results to test the assumption of a uniform swing. But increased complexity is no guarantee of greater accuracy, especially when subtle shifts (particular in the last-minute choices of undecided voters) can swamp the effect of any further adjustment.

Meanwhile, it's also worth keeping on eye on the parties' plans, since they'll likely have the best information as to which seats are close. But the parties themselves face a range of goals which may include maximizing votes, maximizing seat totals, and maximizing the possibility of a plurality or majority government. And I'm particularly curious as to whether the PCAA will bet heavily on a high-variance strategy, preferring to exhaust every hope of maintaining hegemony over Alberta politics rather than making any substantial effort to rebuild from the opposition benches.

To summarize, we should be hesitant to draw overly precise conclusions from the poll results generated so far: while it seems safe to say the order in party support is currently NDP 1, Wildrose 2 and PC 3, the noise far outweighs the signal when it comes to projecting the seats which will ultimately determine who forms government. And so no party should be resting on its laurels as the campaign draws to a close.

On uncertain measurements

Patrick has assembled an interesting response to Eric Grenier's work in quantifying the effect of endorsements in the NDP leadership race. But while I can understand the instinct to try to put together a measurement system for endorsements as a whole, I'd think it's worth being careful about simply adding up numbers, rather than looking at them in context to see which might most plausibly have the greatest effect on the course of the campaign.

Let's start with the seemingly noncontroversial statement that any candidate will have natural constituencies of supporters.

Of course, endorsements from within those anticipated support groups are certainly better than a lack thereof. But it isn't particularly news that Thomas Mulcair has a leg up among Quebec MPs, Peggy Nash within the labour movement, Paul Dewar within the current Manitoba NDP and Brian Topp among Saskatchewan's recent NDP legislators.

As a result, endorsements fitting the expected pattern don't figure to do much to change our baseline expectations for a campaign. And that goes doubly when they essentially duplicate existing endorsements of a particular type. (No offence to Lorne Calvert, but the cachet involved in Topp's claiming the support of a past Saskatchewan premier was substantially reduced when Roy Romanow had already provided exactly that.)

Coversely, a few endorsements loom as particularly significant in shaping the outcome of the NDP's leadership race in ways which may be missed by efforts to categorize by number without considering how they've challenged existing perceptions of candidates.

Most obviously, Brian Topp's ability to point to endorsements from Ed Broadbent and Francoise Boivin gave his campaign instant credibility from the start of the leadership race. And he's combined that initial endorsement from two high-profile political veterans who might not have been expected to support an outsider with an effective media strategy to get himself labeled as a top contender in a contest where his lack of history as an elected official might otherwise have been seen far more of an obstacle.

As for other endorsements with potentially massive impacts on the trajectory of the campaign which might not be captured in any metric, I'd point to:
- Pierre Ducasse's support for Peggy Nash, allowing a candidate who started the campaign without an obvious Quebec base to benefit from Ducasse's role in getting the NDP's progress started in the province that delivered its electoral breakthrough. For those eager to develop metrics, that might signal a need to account for "authoring major party documents" (or perhaps "name drops" based on Nash's frequent references to Ducasse in debates), but I'd consider Ducasse more an example of the difficulty developing meaningful objective criteria in the first place.
- Libby Davies' endorsement of Topp, which may have substantially affected the composition of the race to the extent it split urban B.C. support which Peter Julian may have needed to see a viable path to the leadership.
- Linda Duncan's support for Dewar and Don Davies' for Mulcair, both of which added regional balance from highly-respected members of the NDP's existing caucus to campaigns which otherwise lacked a great deal of visible support west of Manitoba and Ontario respectively.
- And most recently, the UFCW's recent nod to Thomas Mulcair - which signals that the front-runner is moving beyond the endorsements of individual union leaders to bring substantial parts of the labour movement onside.

Once a candidate has locked in a few endorsements of a particular type, though, I'd expect additions to the list to result in diminishing returns - meaning that I'd be skeptical of assigning equal weight to later examples. About the only exception to the greater potential impact of early endorsements is then to be found in rare cases which again arise only based on a candidate's particular circumstances: for example, a campaign that's facing a scandal or otherwise perceived to be on the downswing may be able to generate a new sense of momentum with a major endorsement or two regardless of whether it involves a new category of supporter.

All of which is to say that while it's great to see some effort to measure the state of the NDP's race through endorsement counts, they (like fund-raising totals and polling based on non-members' preferences) figure to show only a small part of the picture. And so we shouldn't be too quick to use them as a proxy for the state of the campaign.

Friday Morning Links

Assorted content to end your week.

- The NDP follows up on the Tony Clement G8 scandal by pointing out the connection between his pork-barrelling and the 2008 federal election (which, let's not forget, was called at the Cons' behest):

The NDP is accusing federal Conservative cabinet minister Tony Clement of using a controversial, $50-million G8 legacy fund to buy re-election, prompting a heated denial from the government.

Municipal documents obtained by the New Democrats show Clement met with local mayors and councillors in the midst of the 2008 election campaign. They discussed how to identify projects that could be eligible for the legacy funding.

Twelve days after that meeting, a local news outlet reported that Clement had posted video endorsements from "local townspeople, mayors and council members" on his campaign website.

"It gave him a major advantage over the other candidates," New Democrat MP Charlie Angus said in an interview Thursday.

"I think the question has to be asked: Was this a $50 million price of an election?"

- But then, the Cons have obviously lost touch with any sense that public money should be used for anything other than to fulfill their political whims - as Kate Heartfield points out when it comes to their dumb-on-crime policy.

- Meanwhile, the Calgary Herald slams the Cons for eliminating any actual judgment and discretion from the criminal justice system:

The Canadian Bar Association passed a number of worthy recommendations at its recent annual conference that Justice Minister Rob Nicholson should take time to consider.

They include a measure asking that a "safety valve" be brought in with mandatory sentences, which would give judges an ability to deviate from the legislation in rare circumstances where they feel the sentence would cause an injustice.
...
(O)ne size fits all does not work in sentencing. We just have to look to the U.S. for numerous examples of what can happen in the extreme. Under California's three-strikes law, a man with a record of two felony convictions for burglary, was sentenced to prison for 25 years to life, after getting caught at a pro shop trying to steal three golf clubs. The U.S. Supreme Court upheld that ruling. In another case a man whose third offence was stealing Batman and Cinderella videos from a department store, was sentenced to a minimum of 50 years in prison.

These are not hardcore criminals and it is not in society's best interest to treat them as such.

- Finally, talk about this fall's Saskatchewan election is starting to heat up, if only in highly general terms so far. Both Three Hundred Eight and the Numbers Guy have posted overviews of what to expect - albeit with little discussion of the factors that figure to influence the race. And a new Saskatchewan General Election blog has also launched which should hopefully serve a useful aggregator for election news (as well as a reminder of the unreliability of non-random online polls).

Not yet satisfied

Eric highlights Environics' polling as to how happy supporters of various parties are with last month's federal election results. And the findings look to bode about as well as possible for the NDP's prospects of expanding its reach over the next few years:

Conservative supporters felt very positively about the election result - and why not. Their party won a majority. Fully 82% are quite happy.

But New Democratic voters aren't as pleased as you might expect. Their party made a historic breakthrough in Quebec, won the most seats and votes in their history, and are now the Official Opposition. But only 27% of their voters are happy with the results.

Though it is somewhat surprising it is as high as it is, 13% of Liberal voters and 10% of Bloc voters feel positively about what happened on May 2nd. Thirty percent of non-voters are also pleased with the result.

As for having negative feelings about the election result, only 2% of Conservative voters have some regrets. That is miniscule. Surprisingly, only 21% of non-voters feel the same way (36% are, understandably, indifferent).

Despite their historic outcome, fully 42% of NDP voters feel sad or fearful about the election results. And despite being reduced to third party status, only 54% of Liberal voters feel negatively. It is a majority, but you'd expect Liberals to be a little more upset, along the lines of the 73% of Bloc voters.

Now, it's important enough that a substantial number of NDP voters were hoping for more rather than expressing satisfaction with the party's position. After it's surely easier to build a movement when supporters are concerned about the direction of the country and motivated to change it, rather than seeing reason to get complacent.

But the level of satisfaction with the election result among the opposition parties is especially significant when paired with the NDP's post-election polling boost. In effect, it looks like a number of the Lib voters who already looked like promising NDP targets have already made the jump - and are apparently finding a landing pad which fits their own attitudes about the election.

[Edit: fixed quote.]

On breakdowns

While Alice is working on debunking the all-too-familiar refrain about one-time political standings dictate future strategy and results, Eric is taking the concept to new depths. So let's take a quick look at why it makes roughly zero sense to focus in on the demographic breakdown of the ridings represented by a party (based on a one-time electoral result) as determinative of its support and direction.

In effect, the assumption behind Eric's demographic profiling is that a party's raison d'etre is to represent all constituents within the ridings they've won, and none in the ridings they've lost.

But the supposed changes since May 2 themselves serve as an indication of just how far off base that assumption actually is. If one applies Eric's logic to the status quo as of a month ago, the NDP would have been considered a party whose geographic and demographic base was utterly inconsistent with any interest in winning votes in Quebec - based not just on its past vote shares, but on the theory that it would be too tied to placating its existing bases of support to make an effort to speak to the concerns of Quebeckers.

To the extent anybody wanted to make that case, it would seem to have been entirely debunked. But Eric's analysis takes it as a given that we should use Canada's 2011 election results as a starting point in applying a theory which is utterly inconsistent with those results coming to pass in the first place.

In reality, of course, a party has many different constituent groups to take into account, and the demographic breakdown of currently-held ridings figures to be at most a secondary consideration. In terms of direct influence, the more important figures are bound to be the MPs themselves, along with the party activists and supporters who take a direct interest in the party's message. And in terms of strategy designed to maximize future electoral outcomes, it would be utter lunacy for a party to focus more on the most-distant tranche of voters in ridings it already holds (who are included in Eric's numbers) than on swing demographics which might actually result in winning additional seats in the future (who are excluded).

Fortunately, I doubt that the NDP (or any other party) will decide to follow the suggestion to represent only the demographics in the ridings they now represent, rather than looking at a more complete picture as to who might be able to elect MPs in the future. But it's worth pointing out the flaws in Eric's analysis for the benefit of outsiders too - lest we otherwise see many more opportunities like the NDP's Quebec surge needlessly brushed off by the commentariat.

Well said

Alice chimes in with the definitive post on how to handle between-election polling:

(S)top reading "the polls", stop obsessing about "the polls", and for goodness sake stop writing that "the polls say this will happen or that will happen", because they don't say any such thing. And take amateur seat predictions having a zero-general-election track record with as big a boulder of salt as Joan Bryden's sources advise taking the polls. "The polls" didn't predict the Conservatives would win Montmagny–l'Islet–Kamouraska–Rivière-du-Loup, or that the Liberals would win Winnipeg North either. Campaigns matter, and outsiders don't know what ridings political parties are going to target or send resources to yet, though we can certainly draw some conclusions now (and no, it's not based exclusively on whether they were "close" last time or not either).

On localization

It's fully predictable that a questionable narrative about Canadian politics turning into a two-way race will be blown out of proportion. But it's particularly worth keeping in mind how a focus on national polls alone might miss some of the most important steps taken by the NDP to get ready for the next federal election.

While the NDP closed the gap against its competitors in terms of national spending in 2008, it still had a ways to go in getting organized at the riding level to covert potential votes into actual ones. And so, as I've pointed out, the NDP has been engaged in a concerted effort to build capacity at the EDA level, providing incentive programs and support to make sure that local campaigns can do more to turn general support into riding-level success.

Like any choice, that strategy carries a price, with correspondingly less resources going into efforts which might do more for the NDP's national numbers. Which means that the focus on local development may go a long way toward explaining the fact that the NDP has mostly stayed at the low to middle end of its usual polling range over the past couple of years.

What the choice to focus on local development also means, however, is that a seat projection model which relies on national polling while missing any increased ability to translate that broad support into votes is bound to err on the low side in projecting the NDP's results. And while it's difficult to know exactly how the two factors will play out, I'd see at least as much upside as downside in combining strong candidates and improved local organization with a marginally lower top-line number outside the election period.

The full picture

Eric nicely summarizes how much public funding Canada's political parties actually receive. And it's clear from the full view that the Cons' main goal in trying to eliminate per-vote funding is to exacerbate the inequalities in a system that already favours them:

Combining the per-vote subsidy, the tax credits on donations, and the reimbursements for the 2008 national and local campaigns, the Conservatives would have received about $42.2-million from public sources in 2009, the last year for which complete financial data is available. The Liberals received $28.1-million, the New Democrats $19.8-million, the Bloc Québécois-$8.2 million, and the Greens $4.3-million.

With each party getting $2 per vote from taxpayers, it is a level playing field. But including all sources of public funding makes it far less equitable. In 2009, the Greens would have benefited from about $4.59 in public funding for every vote cast for them in the 2008 election. That amount is $5.91 for the Bloc, $7.75 for the Liberals, and $7.87 for the New Democrats. Because of their strong fundraising organization and the large amount of money spent in the 2008 election, as well as the per-vote subsidy, the Conservatives benefited most from public funding in 2009, at $8.11 per vote.

On blockage

Eric is right to note that most scenarios involving party disappearances or mergers would result in relatively little change in Canada's political stalemate. But he also confirms that there's one what-if which would make a world of difference:

The Bloc was born out of the failure of the Meech Lake Accord. Had the accord succeeded, it is plausible to assume the Bloc Québécois would never have existed.

In such a case, the social democratic NDP would take the place of the social democratic Bloc Québécois in the province, winning 30 per cent of the vote in the next election. The Liberals, at 29 per cent, would not be far behind while the Conservatives would take 24 per cent of the vote. This would result in the NDP winning 34 seats in the province, most of them coming in francophone Quebec outside of the two main cities. The Liberals would win 23 seats, mostly in and around Montreal and in the Gaspé, while the Conservatives would take 17 seats, concentrated around Quebec City and the Saguenay region.

Using current projections for the rest of the country, this change in Quebec would boost the Tories to 146 seats nationally, still short of a majority government. The Liberals would win 104 seats. With more than half of their caucus from Quebec, the NDP would win 57 seats. It would give the Liberals and New Democrats the possibility of forming a majority coalition.

Of course, many a commentator has looked silly for predicting the end of the Bloc over the years. But it's worth noting that the Bloc looks to be the one opposition party whose share of votes and seats would lead to an almost certain change in government if distributed along current voting lines - and wondering whether a strong message about the need for progressives to work together inside and outside Quebec (regardless of the party structures involved) might serve to get things moving in that direction.

Races may be closer than they appear

Eric Grenier's by-election projections certainly make for an interesting read, and it's a plus to see at least some commentary based on voting patterns rather than conjecture. But with two of the three seats projected to be won by 40+ points, I have to wonder whether the nature of a by-election will cause problems for a model developed for general elections.

Without having crunched any numbers myself, it seems that by-elections generally tend to produce closer races - which makes sense as the parties throw everything they have into one riding, eliminating the gap in expenditures and narrowing the difference in manpower necessary for a candidate to run away from the pack. And that would lead me to expect both Dauphin-Swan River-Marquette and Winnipeg North to produce far closer races than projected, even if the projected winners sound about right absent some major change over the course of the campaign.

On oversimplification

Like most Canadian political junkies, I'm a big fan of the work of Eric at Three Hundred Eight in compiling poll results and modeling the possible electoral outcomes. But I'll use one of Eric's recent posts to highlight one of the main problems that arises where analysts oversimplify the effect of a particular seat distribution rather than acknowledging the place of the House of Commons as a whole.

Here's how Eric analyzes a "best-case scenario" (i.e. highest possible seat count) for the Liberals which would result in party standings of 127 Lib, 111 Con, 51 Bloc, 19 NDP:

As for the Liberals, their best case scenario is the thing the party has been hoping for since being defeated in 2006: a return to government.
...
It's generally the same size of caucus that Paul Martin won in 2004, but the problem is that the Liberals would need to rely on the support of either the Bloc Quebecois or the Conservatives to get legislation passed - something Martin had to deal with as well.

Of course, the problem with that analysis is that it leaves out the question of whether the Libs on their own would have any ability to gain the confidence of the House in the first place. After all, the Cons would formally stay in power following an election until they resigned or lost a confidence vote - and given the measures Stephen Harper has taken to stay in power in the past, would anybody be surprised if he either tries to proclaim that his party should remain in government (with Lib or Bloc support) for the sake of stability, or simply shuts down Ottawa as long as he can in hope of a do-over?

So no, a push to the top of the party standings wouldn't guarantee the Libs a place in government. But then, nor would the Cons be assured of staying in power in a distribution of 148 Con, 87 Lib, 50 Bloc and 23 NDP as Eric suggests:

The Conservatives win 76 seats in the West and North, 55 in Ontario, nine in Atlantic Canada, and eight in Quebec. It ensures that their minority government survives for a few more sessions, but probably tests the patience of the Conservative Party for their three-time minority leader.

Once again, this analysis avoids the question of whether the Cons would be able to secure enough support from another party to actually hold on for "a few more sessions". And as some of us have pointed out, Harper's "our majority or their majority" rhetoric may only be laying the groundwork for a coalition to replace his government to the extent democratic processes are allowed to run their course.

In fact, what Eric's analysis ultimately shows is that even under the most wildly optimistic scenarios based on current polling, no party figures to be in a position to govern without somebody else's support (except to the extent the Cons decide not to face the House). And that means that it's absolutely essential to take into account the possible permutations among the second, third, fourth and potentially fifth parties, rather than assuming that the party with the most seats will be able to form government.