Thursday, March 14, 2013

Something old, something new: the Conservative slide in the polls

In this week's piece for The Globe and Mail, I take a closer look at the polling trends. They are certainly negative for the Conservatives, but as I show in my article it is nothing they haven't seen before. Sort of - while the mid-mandate slide is something they experienced after the 2006 and 2008 elections, this one might be the worst yet.


I love scatter plot charts as they are probably the best way to represent just how noisy the data can be in polls. But they also show how (to borrow from Nate Silver), there is often a signal in the noise.

A simple linear trend line gives one indication of where each party's support is headed, but it is worthwhile to look at where they are meandering as well. The chart above shows support at the national level, and a few things are immediately clear: the Conservatives are sliding and the Liberals are gaining. That Conservative slide is especially consistent, as their polling has been rather tight. Liberal polling is far more erratic (likely due to the leadership race), but it is clearly improving. The New Democrats have been steady, for the most part, aside from the bump after Thomas Mulcair took over.

Plotting the data like this shows how noisy it is: various polls were placing the NDP and Liberals second or third during the NDP leadership race, while they placed the NDP and Conservatives first or second after Mulcair's victory. Now, they are putting the parties in a single mass.

And this is with large samples of 1,000 people or more. The effect of smaller regional samples becomes quickly apparent when we plot the data in the same way at the regional level.

You can see this quite clearly in British Columbia. The Liberals and Greens have been relatively consistent, but even so their results vary wildly. The Conservatives and NDP have varied even more, but as the race is almost entirely between them you can see just how much the two parties overlap (you can also see a few individual Liberal results squeaking into the top of the graph). The Greens, Liberals, and NDP have been relatively steady overall, but a negative trend is clear for the Tories.

In Alberta, the Conservative lead is obvious but the negative trend is also present. Among the opposition parties, the Liberals are clearly gaining while the NDP are on a slight decline. It isn't much, but it is interesting to see how the NDP was clearly ahead of the Liberals until the fall of 2012, at which point their poll results start to get mixed up.

The Prairies are a bit of a mess, but the order of the parties is easy to see. The Conservative slide is less pronounced here, but you can see that the NDP appears to be heading south while the Liberals are gaining. You can also see that after Mulcair's victory the NDP was flirting with first place in the region, but have not done so for several months.

In Ontario, the Conservatives have been on a negative trend since the election, with their polling dropping from 40%+ to the mid-30s. The New Democrats and Liberals are jumbled together, though, with the only clear difference having been in the middle of 2012. The trend for the NDP is a positive one, but it is hard to see it in the data.

Five parties means a jumbled chart for Quebec, but even so you can easily see where the parties are. You can also see how the NDP's fortunes have ebbed and waned. You see good numbers for them after the election, sinking numbers during the leadership race, soaring numbers after Mulcair's win, and a slow decline since. It varies greatly with the trend line, but you can see the overall slide.

The mass of results for the other parties is a little more difficult to see through, but the trend lines help. For one, the Conservative results are hidden behind those of the Bloc Québécois and Liberals for most of the chart, before they slip below them more recently. That says all you need to know about that. The Bloc's numbers have been pretty steady, while the Liberals are heading northwards. You can see that the Liberals flirted a little with the NDP before Mulcair became leader, and is now doing so again (and with stronger results).

And then we come to Atlantic Canada, which understandably has the noisiest data. Sample sizes are usually the smallest in Atlantic Canada and the Prairies, but unlike in the Prairies the race here is a three-way one. That puts the parties in a bit of a tangled mess. Nevertheless, you can see the trends if you try: the Conservatives started out with the best results, then it was the NDP that moved ahead, and now it is the Liberals. The trend lines are pretty stark, with the NDP steady, the Conservatives falling steeply and the Liberals gaining. 

It is an interesting way to look at the polls. The Conservative slide is present in each region, the Liberals are gaining in most of the country, and the New Democrats are trending upwards in some parts of Canada and downwards in others. Averaging the polls out as I do on this site is one way to make some sense of it, but plotting the data this way shows how there is a different way to make sense of it. It demonstrates that the general trends are visible, even if the polls can appear to be all over the place. Sometimes, though, they leave little doubt as to what is going on.

25 comments:

  1. The trend lines... do they represent a mean between the various poll results? In other words, how are the trend lines plotted from the poll results?

    And very interesting, by the way.

    ReplyDelete
    Replies
    1. It is the line of best fit for the data, using a straight line. I wanted to keep it simple - overall, in which direction have the parties been headed since the election?

      Delete
    2. so how do you remove any subjectivity in the orientation of the line? What is the quantitative nature of the trend line, if you see what I mean?

      Delete
    3. Chimurenga, there is nothing subjective about the definition of the trend line -- there is a precise mathematical formula (look up "linear regression" in any basic statistics textbook if you want to no more; the comments section of a blog is not the place to get into it). The basic idea is you want the line that fits the aggregate of the data points as closely as possible.

      What is debatable, is whether a linear model is the best one to use, or whether the dataset is the most useful base for regression. For example, if something changed at time X, and you use data from both before and after X, then you may get a very different line from either pre-X or post-X. Since the pre-X and post-X data might show somewhat different trends, then a single line chosen to fit both as closely as possible, will not fit either of them that well. Then you might argue that the best way to extrapolate into the future would be to ignore everything before X.

      Delete
    4. Thanks MGK. It was the point in your second paragraph to which I was alluding re : subjectivity. One's impression of a party's progress (or regress) depends a lot on the selection of initial and terminal points...

      Delete
  2. It looks like you've got another byelection to add Eric. :)

    ReplyDelete
  3. Very interesting to see. In most place the Conservatives are in a ongoing decline, the NDP is often fairly steady and the Liberals are trending up. It is often said that the Liberals need to win over NDP votes to grow their share as the Conservative voting bloc is solid. This suggests that maybe the increasing support for th Liberals may be coming from the Conservatives.

    ReplyDelete
    Replies
    1. J.S. Woodworth15 March, 2013 14:42

      The NDP decline in Quebec is much steeper!

      Delete
  4. Charles Harrison14 March, 2013 17:13

    Peter Penashue just stepped down!

    ReplyDelete
  5. The somewhat amusing thing about all this is that out here in Calgary (I am formerly from Moncton), the ultra-right leaning Calgary Sun had a column last week (I forget who from) predicting that the Conservatives were set to win the next 20 to 40 elections!!!! And that the Canadian political landscape was becoming more conservative than ever!!! Obviously, they certainly haven't been looking at the same graphs!!!

    ReplyDelete
    Replies
    1. There were a few in-depth surveys conducted last year of Canadians' political orientations (i.e. conservative, middle-of-the-road, progressive) and how they've changed over the past 10 to 20 years (or more). The findings were that attitudes have been polarising, with a marked greater increase in progressive vs. conservative attitudes. In the past few years, for the first time in the era covered, the long-standing plurality for middle-of-the-road (or centrist) attitudes has dropped below either conservative or progressive. In the face of a Conservative Party that managed to achieve a majority government with less than 40%, it's clear that the Conservatives are very close to the edge... with little reason to believe they could move upward.

      Delete
  6. What goes up must come down. The CPC have been in power for seven years and their performance is average at best. Their best case scenario is to win one more majority before they are defeated.

    The CPC know they need to win a majority in 2015 or they would not be able to govern. A coalition or alliance between the LPC and NDP would not be frowned upon in 2015. It was in 2008, due to the leadership of Dion, and the huge role that the BQ played.

    The CPC is unlikely to gain that many seats in the next election. They peaked in the West and Ontario, and are going nowhere in Quebec.
    The biggest battleground that the CPC need to defend is Ontario. The CPC managed to win many suburban Ontario ridings from the LPC in 2011, but the Trudeau-led LPC will be aggressively targeting these ridings again. BC will also be interesting, but a lot would depend on upcoming premier Dix's performance in the next two years.

    Stephen Harper knows that he is not invincible. Our best PMs MacDonald, Laurier, King, Trudeau all lost elections after time in office.

    ReplyDelete
    Replies
    1. Anon: March 14: 20:13,

      It's Macdonald not MacDonald. This is an important difference as Macdonald denotes descent from a clan chief.

      Secondly, blame for the fiasco of December 2008 (and I think Brian Topp makes this clear in his book) lies squarely with the late Jack Layton and Mr. Topp. Neither understood how government formation worked much less the conventions surrounding the Royal Prerogatives. It was foolhardy not to mention unbecoming of a parliamentarian to assume an alliance of opposition parties who had not defeated the Government much less gained the confidence of the House, to cajole the Crown to grant them a commission of government or, assuming opposition parties hold a right to access and advise the Crown. Dion deserves some blames for being naive but, I think it clear from what I have read the idea began in the office of the NDP. leader.

      Delete
    2. Frankly, the blame lies with the media, that coddled the mistaken notion that a coalition was somehow both undemocratic and unprecedented.

      Delete
    3. chirumenga,

      The way in which Layton, Dion and Duceppe went about the coalition was unconstitutional, ultra vires, undemocratic and unprecedented in Commonwealth legal history.

      I will not get into the details on all my points. However, it is important to remember a government can only be dismissed in two ways; unilaterally by the Crown (often because they fail to garner supply) or through a vote in the Commons. Since, the Harper ministry was not defeated on a motion of no-confidence the request of Her Excellency to remove Mr. Harper and appoint Mr. Dion as prime minister was undemocratic because no vote had occured.

      It was unconstitutional because Layton et al schemed to give advise to the Crown for which they were not entitled to give, making their actions ultra vires.

      Finally it was unprecedented because dismissals in the Westminster systems, while extremely rare, must have legitimate and legally sound reasoning. The fact the opposition parties did not like a piece of legislation is not sufficient.

      Had they defeated Harpoer in the Commons all the poibnts I have raised except the second (the ability to give advice to the Crown) would be mute.

      Delete
    4. "or through a vote in the Commons"

      Harper prorogued parliament before the vote could happen.

      There was nothing undemocratic or unconstitutional about what the three of them tried to do. Ignatieff just pulled the plug on it before parliament returned.

      Delete
    5. Penashue and Harper's defense of this scandal and the continuing elections scams that are surfacing will sink Harper as his own Adscam. No coalition will be necessary.

      Delete
    6. The coalition was perfectly legal and Harper perpetuated the lie to get away with contempt.

      Delete
    7. Anon March 15: 20:06,

      You are incorrect there was no motion of contempt before the House in Dec. 2008. The Government fell on a contempt motion in 2011. I believe you have confused the two.

      chirumenga et al,

      Harper was well within his right to ask for a prorogation. He was an elected PM who had recently won a confidence vote (the Throne speech). As such he advised the GG. A vote did not happen but, that is irrelevant: the GG must act based upon the facts at the time: Harper had won a confidence vote, was not defeated in the Commons, the opposition had not won a vote nor the election, nor had they defeated the Government in the Commons. As per responsible government the GG was REQUIRED to follow Harper's advice. Nor was there a compelling reason to dismiss the PM such as; criminal activity or the inability to guarantee supply.

      Much like the Denison rule the GG agreed to a procedure that would not limit the Commons' ability to decide the matter in the future, hence a prorogation of a reasonable period not a dissolution. In hindsight this turned out to be the correct decision as the opposition who claimed a right to form government were unable to stay cohesive long enough to defeat the Tories in January. If they could not work out a plan to defeat the Government how could they have possibly been prepared to govern for the next 18 months?

      As I mentioned above Layton et al wanted a change in Government before a vote. They wanted the GG to refuse prime ministerial advice. To refuse advice is equivalent to dismissal. Since, no vote came about, nor could have come about, the actions of the opposition were undemocratic. They wanted to become government without so much as winning a vote in the Commons much less an election. Such actions are undemocratic, likely also ultra vires and unconstitutional since, there was no compelling reason to refuse advice or dismiss the PM.

      Some may ask why then did Lord Byng refuse advice and appoint Meighen? The situations are similar but, not identical; MacKenzie King had lost a vote on a censure motion and it was debateable whether he retained the confidence of the House. Clearly Lord Byng thought he did not and revoked his commission.

      The Crown does have the power to unilaterally dismiss the Government but, needs reasons for doing so such as; the inability to guarantee supply as in Australia in 1975 or, losing a vote on a censure motion as occurred in 1926. In 2008 there was no compelling reason to dismiss Harper. The question of supply had not been put forth and the Government held confidence in the Commons. What the opposition asked: remove the Government without following proper constitutional procedures is ultra vires!

      The opposition acted in an unconstitutional manner by demanding the Crown listen to them. Such unsolicited advice is ultra vires and unconstitutional as it contravenes responsible government. Only ministers may offer the Crown advice.

      It all boils down to the incompetence of Jack Layton who thought the making of a coalition was enough to win the Crown's confidence. It is not. A vote to defeat the Government as well as a vote of confidence to confirm the new Government are required for the confidence of the Crown. There are limited cases where only one vote may be necessary or when the Crown dismisses a PM but, in such circumstances the new PM serves on an interim caretaker basis until an election is called; this was the case both in Australia in 1975 and Canada in 1926.

      Delete
    8. I would like to remind people that the Ministry remains in office until it resigns or loses a vote. As per this convention the premier remains premier even if he or she loses an election and retains fewer seats in the House than another party. If once the House is called after an election the premier loses a vote he or she must resign or advise a dissolution.

      It should also be remembered that the Crown not the legislature appoints a premier. The Legislature merely confirms or disapproves of the appointment. Dippers may not like this attribute of our constitution.

      Delete
    9. Anon is correct and Derek is incorrect. A coalition government is never "illegal." It might be impossible or unlikely to happen, but never is a coalition illegal.

      Delete
    10. Anon March 18 15:04,

      Please re-read my posts. I do not write a coalition is illegal. I write the way Jack Layton et al went about their coalition was ultra vires and unconstitutional because they did not follow constitutional procedures.

      I stand by my statements and ultimately am confident in their accuracy since, a prorogation was granted. If you stood by your statements you would post under your name or at least a pseudonym!

      The NDP wants to have a different set of rules for themselves. We see this time and again whether it be the questionable actions of Adran Dix while in the premier's office, Layton not wanting to follow constitutional procedures in Dec. 2008, or Mulcair disregarding the rule of law with his Unity Bill. Dippers assume they follow a different set of rules.

      Most importantly, I try and explain my reasoning you have not! You have not engaged in debate much less presented a counter argument. I have no problem with others criticising my writing but, if you are to write that it is "incorrect" common courtesy would deserve an explanation.

      Delete
  7. Somebody tell Ibbitson and Bricker.

    ReplyDelete
  8. Robocalls, in and out fraud, Penashue, the list is long and growing. This is pulling down both Harper personally and his party. If this trend of scandals continues the CPC is finished. This is what is shaping the polls we see recently.

    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.