Wednesday, March 23, 2011

Examining EKOS Canadian Federal Voter Intent 10 March Data:: Should the Conservatives Be Worried?

EKOS Federal Voting Intent, January—March 2011
14 March 2011
Notes:: I've been sitting on this post for over a week, due to recent political developments and the tabling of the budget. It's a bit long, but I decided to post it anyway. A new EKOS poll is due tomorrow, so I may update the analyses in a new blog post, if there are major shifts. On 8 April 2011, I revised some of the methodology on second choices. On 28 April 2011, I similarly updated the section on loyalty to revise the methodology. See notes below.

The last Ekos poll results is full of interesting bits of information and it will be interesting to see how this plays out over the next few weeks as the Conservatives and the opposition parties dance around a spring election. The Conservatives have plenty to be concerned about, given Oda, Kenney, the "in-and-out" scandal, and the fact that the high-profile Cummins, Strahl, and Day aren't seeking reelection. While the Conservatives might have a plan {or a chess-like trap regarding the optics of a spring election}, I feel these EKOS numbers should give them more sleepless nights.

First, the overall numbers are fairly stable for the Conservatives, Liberals, and the NDP {see above chart}. The overall numbers are close to the last EKOS poll before the 2008 election, with the Conservatives, Liberals, and NDP at 34.8%, 26.4%, and 19.4%, respectively. What should concern the Conservatives are the highly volatile numbers in Atlantic Canada and seat-rich Ontario. The 2008 election cycle saw high opinion poll volatility in Atlantic Canada and Ontario, along with a general decline in the CPC numbers between early September and mid-October {-3.2% nationwide, -5.2% in Ontario, and -8.3% in Atlantic Canada}. The loss of high profile CPC incumbents in BC {albeit in safe ridings}, where the HST has angered local voters, is likely to make CPC efforts in that province more difficult. While 2011 isn't 2008, I'm not sure the Conservative marketing machinery is up to snuff. I've seen plenty of CPC ads since January, but if I were in their war room, I'd be disappointed in the numbers. Stephen Harper is a leader with popularity problems, but nevertheless the leader of a relatively popular party. The Conservatives really needed to pad the margin, given the high-stakes game they're playing. They haven't done that.

I've revised the analyses in order to make them clearer. There are key differences from my original post that looked at the percentages shifting from 2008 to 2011, without accounting for relative support, i.e., the voting intent percentage for the EKOS survey reported on 10 March 2011. I also used degree centrality to examine loyalty shifts.

Second, there's the issue of voter loyalty patterns over time. This addresses the issue of how the composition of Parliament might change in 2011. Canadians tend to be more promiscuous with their party loyalty, particularly when compared to their US counterparts. Looking at the "diagonal" on the above table, shows how well the parties are retaining voters from 2008. The Conservatives and the Bloc appear to be doing well, given that 77.3% and 73.8% of those who voted for them in 2008 intend to vote the same way now, respectively. The Liberals and NDP aren't doing as good a job, with 63.8% and 61.6%, respectively. The Greens are far behind with 51.8% of their 2008 voters intending on voting Green now.

Moreover, it appears as if the Conservatives are doing a good job of attracting those who voted for the other parties in 2008. They are attracting sizeable percentages from each of the parties. In terms of loyalty, I wanted to see patterns of which parties are sharing voters between 2008 and 2011. Care should be taken with the use of the following statistical analyses {but not the diagrams}, as there are some methodological caveats, but the statistics are fine if viewed as descriptive.

This graphic {from Netdraw 2.104} shows the pattern of 2008 voting and 2011 intent::

2008 Voting & 2011 Intent from EKOS 10 March 2011 Poll
Nodes represent federal parties. Inbound links (figures closest to each node) represent percent of voters who voted for another party who now support that party. The size of the nodes represents the percent of those who voted for that party in 2008 who intend on voting the same in 2011.
The above diagram depicts a core-periphery based upon 2008-2011 loyalties and is a graphical representation of preference structures. The Liberals share promiscuous voters with the other major federal parties with more defections than converts. The Liberals have lost 9.5% of the support of the electorate to the other parties. On the other end of the spectrum, the Bloc shares very few voters with other parties and are retaining 6.5% of the total, which shows defection from their 9.97% 2008 result. The Conservatives are siphoning off 2008 voters from other parties, while relatively few who voted for them in 2008 are defecting {27.2% of the sample}.

Generally speaking, I think this is good news for the Conservatives. The pattern with the Liberals, NDP, and Greens represent a fragmentation of the centre and the left. The danger for the Conservatives? So, if a whiff of impropriety starts to taint the Conservatives, they may see their support evaporate, as voters go back to their 2008 choices. This is a non-trivial 8.1% of the sample. They will be able to count on the hardcore Conservatives, but if they start to lose support in volatile areas like Ontario and Atlantic Canada, they could find themselves coming up short in many ridings.

Shifts from 2008
I plugged the above table into a social network analysis programme, UCInet v. 6.328, and looked at degree centrality.  Degree centrality assesses the pattern of loyalty, in terms of converts and defectors. Here's the output from the EKOS loyalty numbers::

Degree centrality for EKOS data released 10 March,
OutDegree are defections and InDegree are converts from respondents' 2008 vote.
Both are percentages of the entire sample.  Example: LPC InDegree means 9.171% of the sample who voted for another party were converted to vote Liberal in late February/early March 2011
Comparing InDegree to OutDegree shows that the Liberals, Greens, and Bloc have more converts than have defectors, while the NDP and Conservatives have more defectors.

Second Choices
Note {8 April 2011}:: I've updated this blog to note some refinement of my methodology. Specifically, the multiplicative coreness statistics takes into account the relative support of the parties, which are 35.2% CPC, 27.8% LPC, 14.9% NDP, 10.1 GPC, 8.8% BQ, and 3.1% Other. I've used percentages and considered "no second choice" percentages to be second choices for a given party. This captures a loyalty effect.

Finally, analyzing second choices can further shed light on how voters might shift their preferences during a campaign. The last time I looked at EKOS second choices was a year ago and the pattern is similar. About one in five of Liberals and Conservatives have the other party as a second choice. About 23% of Greens and NDP supporters have the other party as a second choice. About one in three Liberals and NDP supporters have the other party as a second choice. Conservative and Bloc supporters are the most likely not to have a second choice, approaching the 50% mark. I decided to look at multiplicative coreness to see the pattern for party preferences for first and second choices. Multiplicative coreness is a measure of the extent to which a party shares voter intent with other parties, which may not be readily evident in the tables. One uses this to determine core-periphery. In this instance, I was interested in seeing if there was a core political preference patterns {using percentages of all voters}. Using percentages as opposed to raw numbers provides more diagnostic value, since absolute magnitude would drive the solution and the nuances of a multi-party system would be lost. Percentages of all voters allows a quick gauge of how a party is faring, given the entire electorate. For example, it's one thing to say that 5% of those intending on voting Liberal have the NDP as a second choice versus 5% of the entire sample who intend on voting Liberal have the NDP as a second choice. The latter makes comparisons easier. The statistical output is fairly straightforward, showing how a party is close to or correlated with a core preference pattern. The closer to 1 means a party is increasingly correlated with a core preference pattern. The closer to 0 means a party is on the periphery. {A conservative test is the core is greater than the average corenness statistic plus the standard deviation, while the  periphery is less than that calculation}.

The Conservatives represent a core with their numbers of supporters. The NDP and Liberals have the most connection to other parties, representing an amalgam of opposition parties. This speaks how they are sharing ideological space. This is also evident in a Netdraw graph::

Second choice patterns from EKOS 10 March 2011 Poll
Nodes represent federal parties. Inbound links (figures closest to each node) represent percent support of voters who support another party for first choice who support the focal party. The size of the nodes represents the percent of federal vote intent. The thickness of the lines represents the second choice relationships between the parties.

Several things stand out. The NDP and Greens are sharing ideological space, as are the Liberals and the NDP and the Liberals and the Conservatives. Also, the strength of the NDP in Québec as a Bloc supporter's second choice is interesting, but more analysis is required to see if it can translate into seats. This 2010 blog post on BluntObjects does a good job of going into what might come of the rise of the NDP in the polls. Additionally, here's an abacusdata blog post shedding light on NDP election transfers.

The fragmented centre-left vote is nothing new, as is the fact that the Liberals and Conservatives are sharing ideological space. The Liberals know they need something dramatic to get voters back. The real issue is can the Conservatives hold onto the inroads they made into other parties' supporters in light of controversy and scandal? Given that almost 21% of Conservative supporters see the Liberals as a second choice, that there's volatility in Atlantic Canada and Ontario, and 3 high profile Conservative candidates in BC aren't seeking reelection should be a cause for worry. While it's too early to predict what the campaign issues are, it makes sense that the opposition parties will focus on fairness and Conservative improprieties and relentlessly hammer those points home.

Strategically, I feel the Conservatives are sending mixed messages. On the one hand, they're trying to increase support by new Canadians, yet launched a new campaign [français] with a tough on illegal immigration stance. The problem with this rhetoric is that it can serve as a reminder to those going through the realities of the legal immigration process. The tough on immigration stance appeals to their base of older white males, but they ostensibly already have these votes locked up. {Note:: A CARP survey points to possible Conservative weakness with respect to older voters consistent with the CARP demographics}.

Conservatives looking at this could see an opportunity for divide and conquer in ridings where the Liberals, NDP, and Greens are splitting the vote. Their ads aren't doing much to foster this agenda, since they're attacking the Liberals, Bloc, and NDP. The Conservatives and Liberals will be battling it out in a tug-of-war over voters who can go either way in Ontario and Atlantic Canada. Places like Egmont, West Nova, Kitchener-Waterloo, Kitchener Centre, London West, Mississauga—Erindale, Oak Ridges—Markham, not to mention Durham {Bev Oda} and Simcoe—Grey {Helena Guergis}. 

The implications for the NDP and the Liberals battling it out isn't clear cut and I feel much of how this plays out in an election may have to do with how the Conservatives fare during the election campaign. The NDP's numbers are soft and it looks like they might be vulnerable in Ontario, where they did well in 2008. If the Conservatives start to falter in the polls, the NDP will gain due to some voters feeling more comfortable about voting NDP over a Liberal strategic vote. Again, about a third of both Liberal and NDP intended voters consider the other party as a second choice, which demonstrates sharing significant ideological space. I think a huge X-factor for both the Liberals and the NDP is voter turnout, but the biggest factor will be how the Conservatives weather the next few weeks. In this climate, I think there's a tipping point for the entire election. Albeit not earth-shattering news, I think that 30% Conservative voting intent in Ontario will signal impending disaster for the Conservatives.

Twitterversion:: [blog] In-depth analysis of EKOS 11 March poll data. Should the Conservatives be worried or are they in the catbird seat? @Prof_K

No comments: