Thursday, June 26, 2014

Some observations about the Ontario 2014 election

There had been a sufficient number of close polls during the election that the result, a Liberal majority, came as a greater surprise than perhaps it should have. In my final blog post I uncharacteristically did not offer a seat forecast. This reflected my own lack confidence in the outcome of an election where polls consistently suggested high levels of dissatisfaction with the state of the province and the incumbent government.

Nevertheless I did make a prediction in an email to my brother-in-law that turned out to be prescient.  I averaged the last Ekos and Forum Polls (ignoring the Ekos likely voter model), entered the data in my electoral model, and turned up a seat count of Liberal-59, PC-33 and NDP-15. When the actual vote shares are entered in my model the seat totals are identical to the actual results, albeit with 10 individual errors that offset to give the accurate total. More confidence in what I produced would have been the order of the day.

What determined the Outcome
The outcome actually was determined early. It is clear that Hudak made an enormous strategic miscalculation to promise to eliminate 100,000 public sector jobs - a fact unequivocally confirmed by PC candidates post-election who said the pledge came as a complete surprise to them. Paul Wells in commenting on the results offered this contrast to how Stephen Harper dealt with prospect of federal layoffs likely to come from his spending cuts.
Hudak ... decided his target audience was people who think eliminating public sector jobs is always excellent. Compare and contrast: During the 2011 federal election, I worked hard to get a succession of federal Conservatives — Jim Flaherty, John Baird — to give me any indication of the scale of public sector job cuts the Harper government had in mind. Baird pledged, with a straight face, to protect the National Capital Region’s bureaucrats from the kind of ravages the Chr├ętien-Martin Liberals had inflicted in the 1990s.
The outcome of this ideologically polarized election will no doubt reinforce the federal Tories' penchant for circumspection and secrecy. But they are likely feeling chills down their spine. This election made clear that voters rejected the Hudak PC program, which was identical in its essentials, if not how it was communicated, to the Harper Conservatives.

In mid-campaign economist Paul Boothe, a former senior federal finance official and Saskatchewan finance deputy offered the following summary of the three parties fiscal platforms (the whole thing is worth reading).
The Liberals and NDP propose to maintain Ontario’s already relatively low program spending at approximately the current level.  In contrast, the Conservatives propose to lower program spending substantially to finance a large corporate tax cut now and lower personal taxes once the budget is balanced. 
Overall voters preferred a Liberal program based on a relatively austere budget offering slightly slower deficit reduction, which nevertheless included a variety of progressive measures such as a wage increase for home care workers, and an income tax increase for those receiving higher incomes. The NDP offered a partially similar, if somewhat incoherent, program that appears to have been based on identifying popular bits and pieces from polls and focus groups. In terms of their major fiscal/economic positioning, they embraced the Liberal timetable on deficit reduction, implying that they would have required their own austerity program.

Numerically, what gave the Liberals a majority was the 7.4% gap between them and the second place PCs, up from a 0.8% advantage in 2011 that left the Liberals just short of a majority at the time. The NDP performed relatively well, particularly in southwestern Ontario where the economy has particularly suffered during the recession following the 2008 financial crisis. However, they did lose three seats in Toronto largely, I would argue, due to the relatively small 'c' conservative character of their platform and its accompanying campaign and rhetoric. An "exit poll" conducted by the Globe and Mail confirmed that the NDP had some appeal this time for PC voters:
Jarring though it may be with long-time followers of Ontario politics, the NDP seems in particular to have made headway with PC supporters. Twenty-eight per cent of respondents who voted for the Tories said their view of the NDP had become more favourable during the campaign, next to 20 per cent who said it had become less favourable. The NDP also split with the Liberals for second-choice support among PCs. 
Remember these are voters who embraced Hudak and his extreme program of austerity and tax-cutting who are saying this. The conservative nature of the NDP's pitch appears to have had political benefits in the southwest and costs in Toronto.

The Polls 
A polling experiment conducted by the Toronto Star during this campaign explains well the polling issues generated by the campaign (I think the print and broadcast media have overstated the degree of polling inaccuracy). The most striking aspect of the campaign polls was the wide variation in reported support for the NDP.  The experiment found that IVR (Interactive Voice Response, essentially a talking computer that makes phone calls) tends to understate NDP support while online polls overstate it. This is certainly consistent with the results of the polls released during the campaign.

I measured polling accuracy in two ways: I compared the election vote shares to the reported poll shares for the three major parties only and then a second calculation that includes the Greens and Other. I convert the errors to absolute values and sum them.The justification for doing two calculations is that it is the strongest parties that determine outcomes in a first past the post system. The errors on Greens and Other are interesting from a polling perspective but matter little in determining seat outcomes.

I have ignored the likely voter reports, which were estimates that were not as accurate as the basic eligible voter polls. I think it likely that if someone is willing and patient enough to answer a poll on politics, it is likely that individual is a voter and no further effort should be necessary. The performance is worth noting. (Note negative values indicate the poll underestimated the actual vote, positive that the performance was over-estimated.)



The online pollsters Angus Reid and Abacus performed well as did IVR pollster Ekos and phone pollster Oraclepoll.  Phone pollster Nanos had a poll that concluded on May 26 that could be deemed the most accurate, but it finished up two and a half weeks before voting day so I have excluded it here. Ekos was closest on the Liberals and PCs and Oraclepoll closest on the NDP.  All pollsters had at least one error outside their reported margin of error and most had two.

The polls despite the errors did perform better than the impressions left by the media (and I certainly agree with Paul Adams' assessment of the value of the polls during elections). The media convinced themselves that a Liberal majority was impossible. While most of the predictions ahead of time were for a minority the major prognosticators had the Liberals in the high 40s or above. At those levels just a slight shift in the polls is all that is necessary to produce a majority.

TV Debates
Part of the explanation for all the surprise was that Kathleen Wynne was perceived to have lost the TV debates. Most commentary I think misses the point about a TV debate. It can affect public assessment of the capacity of the participants. This matters more in the U.S. where voters are asked to make a direct choice between individuals. In Canada, however, debates are mainly about issues. Where they can matter is raising the importance or salience of a given issue. In this case the key moments were about the gas plants scandal, which already had a very high profile and no new information about it came out. As a vote consideration it had already been largely factored into how voters felt about what mattered, which turned out to be not gas plants, but rather the Hudak plan for governance. The Liberals had the most money for this campaign, and no doubt bombarded the airways in the final week with ads like this one, effectively overcoming what they might have lost during the debate.

As a more general observation I would say that while scandals matter to newspapers and the electronic media for sales, they matter a great deal less to the public when it comes to casting a vote. There are a couple of reasons for this. One is a cynicism that reflects a generalized "they all do it" belief. But the other is that most scandals simply don't matter in material ways to the average voter. Thus we have had two Liberal governments in Quebec and Ontario recently re-elected in spite of the various scandals laid at their door, and one in BC a year ago.

Strategic or Tactical Voting
There are two questions one can pose about tactical voting. First, did it occur, and, if so, did it matter. The point of those who engaged in this kind of voting was to defeat Hudak. At a macro level it is clear that strategic/ tactical voting did not matter simply because the PCs were so decisively beaten. When you finish seven points behind your opponent you are always going to lose.

Was there identifiable strategic/tactical voting? I have not crunched the numbers for all the ridings but took a close look at several that changed hands using both the raw numbers and my projections to see how well parties performed in an individual constituency compared to the overall trend.  I examined Burlington, Cambridge, Durham, Halton, Newmarket-Aurora, which all switched from the PCs to the Liberals and Oshawa, which went NDP. Tactical voting appears to have added to the winner's margin in all cases except Durham but all would have been won on the overall trend regardless. Nominally it appears that it may have been decisive in Oshawa, but the numbers are close enough in the calculation that one can't be sure. Assessing this kind of voting behaviour is a difficult exercise in my view (we cannot read voters' minds after all) and simple calculations should be viewed with caution.

Looking Ahead
I doubt that either tactical voting or scandals will be decisive in the 2015 federal election. I would add that whether or not Stephen Harper wins next year will likely have little to do with the scandals such as the Senate expenses affair (for which he is properly being blamed). It will more likely be issues that matter - especially the role of government. The more the federal opposition parties can identify damage done to public services and jobs by Stephen Harper, and there has been plenty, for example, the damage to the federal government's core public administration I wrote about in February, the stronger will be their case for replacing Harper in 2015. Mike Duffy is likely to be a footnote when we look back on the 2015 federal election.

Ontario's greatest challenge is restarting its slow-growing economy (not repairing its finances as is commonly assumed). Average growth has been very slow over the past decade because of the financial crisis and subsequent economic downturn. Ontario has been 10th out of 10 in Canada because it is not resource rich,  and manufacturing has suffered because of the downturn and the overvalued Canadian dollar.
Slow economic growth in Canada's industrial and financial heartland should be seen primarily as a federal government failure. However, if the Bank of Canada encourages a lower Canadian dollar (we'll see), and the American economy continues to gain strength albeit slowly, the fiscal challenges facing the Wynne regime will be diminished. It was the "Roaring Nineties" in the United States, as Nobel Prize-winning economist Joseph Stiglitz dubbed them, that allowed the federal Liberals to balance their budget quickly while economic growth continued. And average voters care more about jobs than balance sheets.

Wednesday, June 11, 2014

Ontario Election: Final Assessment


Ontario's 2014 provincial election is tomorrow and has been characterized by an unusually high degree of negativity. After 11 years and various scandals there are considerable negatives attached to the incumbent Wynne government. However, the opposition PCs have behaved in some respects as if they don't want to win. Their promise to cut 100,000 jobs was a major blunder and making an elementary error in arithmetic is unforgivable as well as a perfect way to demonstrate the party's unfitness to govern. I have the impression (as do others such as Lispop's Barry Kay) that a more careful, moderate campaign might have had Mr. Hudak's party sailing into office. Instead he appears to have maximized the small 'c' conservative character of his platform. NDP leader Andrea Horwath could not articulate a clear rationale for opposing Kathleen Wynne's budget and has suffered an unprecedented degree of internal party dissent (for an opposition party), particularly in Toronto. It is rare for an opposition party to do such a poor job of appealing to its own base.

The closing polls so far appear to be headed slightly in the direction of the Liberals. However, that doesn't mean they are headed for a majority.  The following blogs plus the Laurier Institute for the Study of Public Opinion and Policy all provide estimates of how many seats each party is expected to win in the election for the 107 members of the Ontario legislature. Note: this is probably not their final estimate.


All these sites in my view are quite capable and know the art of forecasting seats well.

Strategic / Tactical Voting

Strategic voting has emerged in Canadian politics in the past 15 years. It is a consequence of the federal and provincial Conservative and Progressive Conservative parties moving to the right opening up a large ideological gap between them and the NDP and Liberals on the Centre-Left. I think it more precise to use the British term 'tactical voting' as it is simply a decision by individual voters to opt for a second choice to prevent a third party from winning. In Canada this typically means NDP voters casting a ballot for the Liberals to block Conservatives (as a consequence partisan New Democrats hate it).

It first bloomed in a major way in the 1999 Ontario provincial election where it succeeded in contributing to several Liberal victories and one NDP win. In this election the Wynne Liberal strategy has as a key element a pitch to NDP voters to prevent a Hudak victory. The last Forum research poll estimated that 37% of 2011 NDP voters have moved to the Liberals this time, indicating that it is likely to be partially successful. There are groups online such as http://www.one-big-campaign.blogspot.ca/ that advocate tactical voting although it is doubtful that many voters will even become aware of the existence of this or other such groups let alone be influenced by them. Nonetheless many will cast a tactical ballot.

The size of the NDP vote will have a critical impact on the overall shape of the election and here is where polls disagree. Support for the NDP has been reported at various levels between 30% and 17% and overall there has been a significant difference between online and phone polls (including IVR) in both early and late campaign polls - an average of five or six points.



My observation is that online polls have overstated NDP support in the past. Regardless, the real level of support matters. As an illustration, if we assume that the most recent Ekos likely voter model numbers (L - 41.1, PC - 33.2, NDP - 17.1) are accurate it would easily produce a Liberal majority. (While Ekos' likely voter model adds to Liberal support, others such as Angus Reid do the opposite boosting the PCs at the expense of the Liberals.) The key is that the 17.1% for the NDP would represent a large drop from their 22.7% in 2011 and the direct loss of seats to the Liberals. It is third parties who create minority legislatures if the race between the two leading parties is close. The chances of a majority for either party rises in direct relation to the increasing or declining strength of the third party.

Based on the 2011 election we can say the Liberal vote is the most efficient.  That voting pattern, if something like it prevails again, would permit the Liberals to be as much as two points behind the PCs but still be ahead in seats. The Ekos likely voter model seems improbable to me. It is my impression that likely voters models should boost the Conservative share. We will soon see. Overall the polling points to the Liberals having the most seats.

A Final Word on Polls
No one pollster is the most accurate in each election.  I have been tracking this for some time.  The most accurate pollster in the 1999 Ontario election was Ekos.  In 2003 and 2007 it was Harris-Decima (known simply as Decima in 2003) while in 2011 it was Forum Research with Abacus close behind. In reviewing these old spreadsheets I made one amusing discovery: in 2003 the Lick's Hamburger Poll was slightly closer to the election result than one of the commercial polling firms.