Thursday, July 26, 2012

The Mulcair bump

A look at the last seven months of polling shows a clear trend where the New Democrats experienced a big bump in support when Thomas Mulcair became leader. But is that bump starting to flatten?

The charts below plot every national poll that has been released since the beginning of 2012, each dated by their median field date. The trend line shows how the polls have shifted over that time.
What stands out most in this national chart is that the New Democrats made a big leap in support after Mulcair's leadership win. Polls between January and March showed the Liberals and NDP jostling for second place, with the New Democrats placing ahead of the Liberals in most polls.

Things shifted dramatically after Mulcair became leader of the NDP, as the Liberals slipped away to around 20% while the New Democrats moved into a tie with the Conservatives and then moved ahead in June. The latest Nanos survey puts them on a significant downward trend, but the poll could be an outlier.

What is interesting to note in this chart is that the Conservatives have been generally steady, but also steadily declining. Whereas the NDP trend line is a bit of an 'S' and the Liberal line meanders a little, the Conservatives are almost on a straight (but modest) downward trajectory.
That has also been the case in Ontario, where the Conservatives have slowly slipped from around 40% to closer to 35% in the all-important province.

The New Democrats were firmly in third back in January and February, but moved into a tie as the NDP leadership convention approached. After Mulcair won, the New Democrats moved decisively ahead of the Liberals. They were also moving into the Conservative range, placing ahead of them in a few polls, but the last two surveys have sent them heading south.
The NDP's bump in support in Quebec was most pronounced, and the fortunes of the Liberals and Bloc Québécois have shifted accordingly.

At the beginning of the year, the New Democrats had a modest lead over the other parties, who were jumbled together in second place. In February and March, however, the Bloc moved into a sort of tie with the NDP while the Liberals also saw a gain in support.

Mulcair's win changed all of that as the party stormed ahead in the polls, sending the Liberals down into a tie with the Conservatives for third place. The Bloc saw a modest drop in support, but have not been in a position to challenge the NDP for top spot in the province since Mulcair took over.

Of note, though, is that the New Democrats are experiencing a bit of a bounce-back. They have been scoring in the low 40s and high 30s since mid-June, whereas their polling numbers were in the low-to-mid-40s in April and May. Though the last survey from Nanos puts the Liberals on an upward trajectory, without it the Bloc, Tories, and Liberals have been generally flat.

These charts point to a bit of a plateau for the New Democrats, who made large gains after their March 24 leadership convention. A plateau is not a horrible thing for the NDP, as they need to solidify their support over the next few years, but it does indicate that the Conservatives and New Democrats might be at a tipping point, with the Tories near their floor and the NDP at their ceiling. Something could force either party through these barriers, or they could each return to more familiar ground. The result of the Liberal leadership race may provide that spark.

UPDATE: Due to popular demand, here are the same charts with a smoother trendline that is not so dependent on the latest Nanos poll. While it doesn't show the same sort of S-pattern for the New Democrats, the plateauing is still apparent. According to the numbers, however, the trendlines in the above charts are a better fit to the data.

These sorts of scatter plots do a good job of showing the variation from poll-to-poll. They also do a good job of showing how, despite these variations, patterns are still clearly discernible. Polls can paint with a very wide brush, but they are far from resembling something by Jackson Pollock.

UPDATE 2: There has been a lot of misunderstanding with this post. The graphic included in the original update calculated the trendline using two points instead of three, it did not remove the Nanos poll. Below, you'll see the same national chart as the one at the top of the page, but without Nanos. You'll still see that the NDP trendline is still the same sort of S-line.

44 comments:

  1. Would be nice to see a sensitivity analysis without that last poll. It might indicate a trend but it might also be an outlier. Seems to be pulling pretty hard on that line. What are you using to calculate those lines?

    ReplyDelete
    Replies
    1. I'd rather see it without all Nanos polls actually. Because their previous polls have higher CPC/LPC numbers and lower NDP/GPC numbers I believe?

      Removing only one Nanos poll and leaving the rest would create the false impression of a downward trend for the CPC and Liberals. What I'm more interested in seeing is whether or not the various pollsters who disagree on the relative numbers for each party still agree on the trend.

      Delete
    2. Agreed. I'm eagerly awaiting the next results from other pollsters, since prior to that last Nanos poll no one else had found the NDP below 32% or the Liberals above 23%.

      Dom

      Delete
  2. I would think the summer lull has an effect as well - and a disproportionately negative effect on the NDP. The return to Politics in the fall will put the NDP back in the spotlight, with the potential for further movement upward, assuming the party manages its actions/image favourably.

    ReplyDelete
  3. I would guess that the graph is heavily influenced solely by the results of Nanos' polls, both pre-Mulcairmania (where Nanos gave the Libs the best numbers, and the NDP the worst) and also the latest one.

    ReplyDelete
  4. Are just using a cubic trend in the excel charts for these? there is a real danger of poor parametrization with that method. You may want to do a centered rolling average manually (2 leads, 2 lags and current or some such combination) to extract the obvious non-linearities in those trends. You will have endpoint problems but curve fitting with quadratics and cubics can be no better.

    ReplyDelete
  5. Seems like a piece-wise linear (pre and post Mulcair) would fit the data better then the spline you're using. Your curve is being exaggerated by the last Nanos poll.

    ReplyDelete
  6. Told you.

    I've been tracking poll numbers with a polynomial best fit line for years. This was always a bump, and it was always going away. That it not to say that NDP numbers won't ultimately stabilise above where they were before Mulcair took the helm, and that it not to say that there hasn't been genuine weakness in the CPC numbers, but the big NDP climb we saw beginning toward the end of the leadership race was always a temporary effect.

    ReplyDelete
    Replies
    1. That's operating on the assumption that the Nanos poll is not an outlier. An awful lot of triumphalism based on a single data point.

      Delete
    2. The trend was evident long before the Nanos poll.

      Delete
    3. If you actually look at Eric's charts without the Nanos poll, what is evident is a plateauing, not a decline. The latest Nanos numbers dramatically alter the pattern.

      Delete
    4. The Nanos poll is included in the bottom charts. The polynomial trendline has an order of two instead of three.

      For the NDP in Quebec, to choose one example, using the Order 3 polynomial trendline is a far better fit than an Order 2. The same goes for the federal numbers, believe it or not, based on the R-squared value.

      Delete
    5. Does that still have the older Nanos polls in it? Because that would make things look like more of an upward trend for the NDP and Greens and more of a downward trends for the LPC and CPC...

      Delete
  7. Trudeau could easily lead the Liberals to a second place finish.

    Liberal party is NOT dead yet, especially because so many NDP seats depend on just Quebec.

    ReplyDelete
    Replies
    1. No disrespect to the Liberals, but perhaps you ought to be a little more cautious about carrying on this exact same mindset that got the party in so much trouble to begin with: "If we can just find the perfect leader, we'll instantly soar back to glory!"

      First Dion was supposed to be "The One", then Ignatieff, then Rae, and now evidently Trudeau.

      Dom

      Delete
    2. I agree with you Dom. If the Liberals believe their answer is in Trudeau, they're in for a shocker. Yes, they will get a bump in the polls, but the whole messiah concept needs to be thrown out of the window if they want to overtake the NDP again. Interesting how the tables have turned. RIP Mr. Layton!

      Delete
  8. A word of caution about summertime polls. They are usually incumbant friendly as the government gets to set the agenda, and opposition parties have to struggle for ink and video time. I'll wait for fall and the reopening of parliament, when folks start to pay attention before I'll make a judgement about trends. Eric, what do your polling histories telll you about summertime polls?

    JKennethY

    ReplyDelete
    Replies
    1. I wrote this for the Globe last summer:

      http://www.theglobeandmail.com/news/politics/summertime-and-government-living-is-easy-not-necessarily/article2093158/

      Delete
  9. I'm trying to think back to what percentage the Conservatives got in Ontario in the 2011 federal election. Wasn't it 44% or 45%? 32% is a real drop.

    ReplyDelete
    Replies
    1. Elections Canada has a website! It was 44%. What is important is not so much the numbers but, the trend-that shows both the Dippers and Tories in decline.

      Delete
    2. "Dippers ... in decline." Ha. In your dreams!

      Delete
    3. Not my dreams the trend lines of the linear regression! Both the NDP and Conservatives are on a downward slope only the Liberals are on a slight upswing.

      The NDP trend line is so steep at this rate they will have fallen to pre-Jack levels by Christmas!

      You're lucky the next election is 3 years away, however, denial will only exaccerbate the pattern.

      JB

      Delete
  10. Hello Eric!

    So your data has only proven to me that my earlier assumptions (which I do not believe were posted on this site) that Nanos has leaned more and more to the Liberals since being hired to work for CTV news. I remember watching Nanos on CPAC and his polls always seemed very accurate. But ever since he was hired by CTV news, he continuously has the Liberals more ahead than any other pollster. This also includes some of the polls during the last election.

    Ironically, you see a dramatic change (in my opinion) when you removed the Nanos poll(s) from the equation. My question to you is a good one, does Nanos call the same people when conducting his polls? Especially when it comes to Quebec, it seems like he finds these niches of Liberal support in the province that other pollsters do not come across. This makes me wonder if he is in fact just calling the same households, cell phones etc... to see if those individuals have changed their minds. I'm sorry I raise this, it is only from personal discussions I've had with others.

    As for Trudeau being the saviour of the Liberal party: I would still be considered to be part of a generation of voters that do not vote. The youth in this country often feel like they are not being taken seriously. Trudeau is not someone I personally take seriously. I honestly put him in the same class as Ben Mulroney. I think it would be a big mistake for the Liberals to vote him in as leader. Amongst my group of friends the only thing he is recognized for is his boxing match and his relation to his father. The youth that I have had discussions with seem as interested in Justin Trudeau as they are in Ben Mulroney. Perhaps a serious poll should be conducted on how the youth really feel about the Liberal Party or Justin Trudeau.

    ReplyDelete
  11. Let's face some facts here.

    a) The majority of the population detests Harper and that number is growing !!

    b) The Liberals are currently a "failed" force. So the vote shift won't go there !!

    c) The NDP is the only viable alternate nationally. It's NOT rocket science !!

    ReplyDelete
    Replies
    1. Except that the evidence above suggests your point B and C are factually incorrect. The evidence for a shift to the Liberals may be inconclusive, but at the very least the data rules out a Conservative or Liberal to NDP shift...

      Delete
    2. Can you honestly see anyone that is conservative that voted CPC last election switching to the NDP? That is one place the Liberal party will get votes, the other may be some of the shedding of NDP votes now that Jack Layton is gone, especially if the Liberals elect a charismatic leader (like Layton).

      Delete
    3. Jolo: I wouldn't underestimate the NDP even without Layton. Mulcair is a very talented politician and if you think he'll let the Liberals call the shots regardless of who they choose you're gravely mistaken.

      Delete
    4. Peter,

      Usually when people state facts they provide evidence to substantiate their claim. You have failed to provide evidence that: "a) The majority of the population detests (sic) Harper and that number is growing"! Subsequent points (b, c) are not facts at all but, opinion or commentary!

      Polling firms do not ask whether you hate or detest A,B,C they query whether one has a very favourable, favourable somewhat unfavourable opinion of A,B,C, or ask : do you strongly approve, approve, slightly disapprove of the job A,B,C, is doing? Or they ask some variation thereof. One can not equate disapproval or favourability with detestation or hatred. To do so is to misunderstand the data or to mould the data to fit one's own beliefs. I challenge you to produce a scientific poll of 500+ respondents where the vast majority of Canadians detest or hate Harper.

      I doubt most Canadians detest or hate anyone. In fact the vast majority of Canadians are mature and intelligent enough to disagree without becoming spiteful or hateful. It is disturbed individuals-unable to recognise detail or nuance- who insist on turning a complex situation into a black and white choice resorting to spite instead of reasoned critique and debate. As the old adage states; insults are the last resort of the out argued.

      Language such as you have used poisons the political environment and justifies dirty tricks, fraud, slander and other illegal activity in our political and electoral processes. Cooperation becomes a dirty word and polarization follows.

      As for points B and C thus far the data does not support your hypotheses. While you have failed to provide any evidence to justify your remarks.

      Delete
    5. Your opinions, not mine.

      Plus if you look at the national scene CPC support is slowly dropping in the polls. I think my remarks where justified.

      Delete
    6. Absolutely not justified nor backed by evidence even circumstancial! They are opinions you misportray as "fact".

      You write the vast majority "detest Harper" -not CPC support is slowly eroding! At best your writing is disingenious, at present it border on illegality. I would draw your attention to S. 298(1) of the Criminal Code:

      -298. (1) A defamatory libel is matter published, without lawful justification or excuse, that is likely to injure the reputation of any person by exposing him to hatred, contempt or ridicule, or that is designed to insult the person of or concerning whom it is published.-

      I would ask you choose your words more carefully in the future.

      Delete
    7. Did you notice the title of this particular section Peter II ??

      The Mulcair Bump which turns out to also be a CPC slide and a Liberal sort of mini-bump.

      So your PMO bilge is just that !!

      Delete
    8. Exactly Peter not; "Poll(s) demonstrates Canadians detests (sic) Harper"!

      The only filth comes from your keystrokes- you are not fooling anyone except perhaps yourself.

      Did you notice the title of S. 297-316 of the Criminal Code Peter??

      Defamatory Libel.

      Delete
  12. I disagree Ryan as the number give you the lie !!

    ReplyDelete
    Replies
    1. This "there's only two options" crap was wrong when the Liberals tried to pull it, and it's wrong now too. It didn't work for us so don't expect it to work for you either.

      Delete
    2. Not to mention the poll numbers indicate Ontario still has far more Tory-Liberal seat battles than NDP-LIberal or NDP-Tory battles. If you live in Ontario and want a new government clearly the best option is the Liberal party to defeat Conservative incumbents.

      Delete
    3. Frankly I think people should just vote their conscience, regardless of where they live. It's a failure of our electoral system that you can't do that and have your vote counted fairly.

      Delete
  13. Eric,

    This post looks very intersting although I have not yet had time to analyse your results in detail.

    I know you are busy but, I would be interested to see comparisons with other leaders; Turner and Mulroney in the 80's (although the timeline for Turner is much shorter), Jack Layton, Chretien in 1990 even provincial examples such as Christy Clark, Wall, Dunderdale and especially Dexter in Nova Scotia since, Dexter and Mulcair charted and are charting new territory.

    Cheers,


    To those commenting on Nanos,
    While I acknowledge some polling firms may have institutional biases I do not think one can pick and choose which polls to use. Either one includes all or one uses only Ekos polls or AR polls for data. Excluding polls are just as likely to skew data as including biased polls. Hopefully aggregating all results limits the amount of bias.

    ReplyDelete
  14. I am by no means a political scientist or a statistician, but the updated graph seems much more plausible when the Nanos data is given less weight.

    With that said, my layman's interpretation of this data is that all party's have settled at a low or high "plateau" for now and will remain give or take the same for the next while.

    At this point I think things are more or less settled. What will be interesting to watch is if there are any big shake ups---a scandal, a big gaffe, blow back from some policy or another, some kind of disaster of one sort or another, and how this shakes things up.

    If something Walkerton-esque happens, the Cons could see their numbers plummet and the question is who picks up the gains. Similarly, if Mulcair gets into another "Dutch Disease" kerfuffle (in interests of full disclosure, I'm with him on that and think media coverage has taken him out of context) that could benefit the Liberals. On the other hand, the Liberals are basically fighting for survival and therefore you can't underestimate them.

    So all I take from this is that we're at an interesting crossroads.

    ReplyDelete
    Replies
    1. A couple of issues are still out there and will probably get more coverage in the next couple of years.

      First is Harper's "control" over his ministers and essentially his refusal to sit down with the Provincial Premiers and negotiate on several key issues such as health care.

      Plus the obvious "resources at all costs" thrust isn't going to fly well in today's world.

      The numbers indicate the fall from grace by the CPC. Now the real question is what will supplant them ??

      Delete
  15. Not to be a tremendous jerk, but the fact that the trends are so dramatically different with the exclusion/inclusion of just one poll kind of shows how worthless the trending analysis is.

    You should really stay away from Excel...

    ReplyDelete
    Replies
    1. Where did I say that the last set of charts excluded the Nanos poll?

      Delete
  16. This 308 post should be renamed 'The Nanos Outlier and the Dangers of Extrapolating Trends from It'.

    How do you choose which polls justify undertaking a similarly sensitive trend analysis sharply affected by that particular poll? For example, the largest poll in months of Quebec federal intentions by Leger (1,000) showed the NDP at 52%. Wouldn't this have been a more reliable indicator (at the time) than the latest Nanos poll numbers in Quebec which were headlined here? Or Forum's last Ontario poll of 1,072 people, also the largest there in months, showing the NDP at a high of 35% in Ontario?

    Devil's advocacy aside, I actually think this is not the way to go. I will stick to your excellent monthly average graphs to discern trends, as difficult as even that is.

    ReplyDelete
  17. I find reaction to this post utterly bizarre. Nanos is just one of the 40+ polls included in this analysis. It is not the focus, nor was it the reason for writing this post. It has as much importance as the other 40+ polls included in the chart. I even mention on several occasions how the Nanos poll might be an outlier. Without Nanos, the trendline would look only slightly different. I've added another graphic to show that.

    I get that the NDP trendline lines up with the last Nanos poll on the national numbers, so it makes it seem as if the entire trendline was dragged down to it, but if you look at the trendlines in the other graphs and for the other parties you'll see that the trendline does not lead inexorably to Nanos's latest poll. It is just one of the 40+.

    I said to myself: wouldn't it be neat to put all the polls in a scatter plot and see what comes out of it? It shows all the variation in the numbers but there is still a pattern in it.

    Instead, the post has been interpreted as: look how Nanos shows the NDP losing ground, I'm going to focus on that.

    Sheesh.

    ReplyDelete
    Replies
    1. When your posts don't feed the narratives we like we try to grasp at straws.

      Delete

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.