About EKOS Politics

We launched this website in order to showcase our election research, and our suite of polling technologies including Probit and IVR. We will be updating this site frequently with new polls, analysis and insight into Canadian politics. EKOS's experience, knowledge and sophisticated research designs have contributed positively to many previous elections.

Other EKOS Products

In addition to current political analysis, EKOS also makes available to the public general research of interest, including research in evaluation, general public domain research, as well as a full history of EKOS press releases.

Media Inquires

For media inquires, please contact: Frank Graves President EKOS Research Associates t: 613.235-7215 [email protected]

ELECTION REPORT CARD

EKOS Once Again Top of the Class

[OTTAWA – October 27, 2008] At EKOS, we have now completed the first phase of our internal performance review of the 2008 election. We have some work to do, but are very satisfied with the results, in particular because we pioneered two new methodologies in this election with great success: Probit, which features a combined an internet and phone panel, and IVR, or Interactive Voice Recognition.

As we have been able to say consistently election after election, EKOS proved to be among the most accurate, if not the most accurate of polls available to the media and the public in the 2008 campaign.

EKOS’s daily tracking polls was the first to identify the softening of Conservative support to below-majority levels in the first week of the campaign, the temporary surge of Green Party support in the second week of the campaign, the growth of the Bloc Québécois at the expense of the Conservatives, and the bursting of the Liberal bubble that emerged after the leaders’ debates.

We were also able to come very close to projecting the final results of the election with our last poll and seat projection.

We were able to do all this because the IVR methodology enabled us to accumulate much larger sample sizes each night – often three times that of other companies. This allowed us to rely on shorter field dates – three days instead of four – making our tracking poll the most up-to-date available. The large sample sizes also gave us a more granular picture of individual regions and sub-regions in making our final seat projection.

EKOS’ seat projection, released on the Monday before election day, was the closest to the final result of any published. Our final tracking poll, released the same day, was the second-closest of all the polls, and the closest of the daily tracking polls that dominated coverage of this election.

We drew on a larger range of approaches to sounding public opinion in this election, because we were both concerned about the limitations of traditional methodologies and impressed by the success of innovators in the United States such as Rasmussen and SurveyUSA.

During the election, we used IVR for daily tracking and Probit for deeper soundings of public opinion on issues of policy, attitudes and leadership. For the first time, we conducted our work completely independently, from our own resources, and published our findings on our ekospolitics.com website.

In addition, we published frequent seat projections throughout the campaign, to illustrate possible outcomes based on our polling. We did so because we saw that other polling companies and media commentators were writing and speaking frequently about the Conservatives “nearing majority territory” or “falling out of majority territory”. These statements were either based on seat projection models that were not publicly disclosed, or else they were simply best guesses. While recognizing the limitations of seat projection models, we felt it was more transparent for the public to be allowed to see our work.

In reviewing our methodology in the wake of the 2008 election, we have identified a couple of emerging concerns: lessons for next time. With voter turnout now in the neighbourhood of sixty percent, it may be the time to start screening more aggressively for likelihood to vote among respondents to our polls.

Traditionally, most pollsters in Canada have surveyed “eligible voters”, assuming that those who actually turned out to vote would have similar preferences in similar proportions. Of course, that is not necessarily the case, especially in our system, where supporters of certain parties may consider their vote “wasted” if the candidate they support has little or no chance of winning a seat.

As part of our post-election analysis, we found that by imposing a likelihood-to-vote filter, we got results slightly closer to those at the ballot box, though not dramatically so.

All the polls (including our own) somewhat underestimated Conservative support and most of them also overestimated Green Party support. We believe that this was likely due in significant part to differential turnout: i.e., the effectiveness of the Conservatives in mobilizing their potential support and the relative failure of the Green Party to do so.

This was likely a function of both a more effective ground campaign by the Conservatives and the fact that they drew their support disproportionately from demographic groups with a higher propensity to vote, i.e., the older and the better-off. In the obverse, the Green Party had a relatively weak ground campaign, their supporters were disproportionately drawn from younger cohorts with notoriously low turnout, and their supporters may have felt their votes would be “wasted” anyway because of the unlikelihood of the party winning seats.

Another issue is the growing number of cellphone-only households. Our IVR methodology does not capture these, though the Probit does to a degree. In retrospect, it seems extremely unlikely that the small underestimate of Conservative support and overestimate of Green Party support can be explained in this way. If anything, from what we know of the demographics of cellphone-only households, they would skew away from the Conservatives and towards the Greens.

However, it is clear that over time, as the number of such households grows, the industry generally must come to terms with this phenomenon. In the United States, for example, where such households appear to be somewhat more common, a recent Pew study suggested that the inclusion of cellphone-only households produced a bump for Barack Obama of a percentage point or two.

We are in a period of change in terms of the use of technology, the availability of new methodologies, and of course the underlying political landscape. By continuing to innovate methodologically, scrutinize our own work and that of others, and equipping ourselves with a full suite of methodological options, EKOS intends to continue to be a leader in our field.

Party Standing

CPC

Liberal

NDP

BQ

Green

Ind.

Aggregate Error

Actual Results

143

77

37

49

0

2

EKOS

136

84

35

51

0

2

(18)

LISPOP

135

87

33

51

0

2

(24)

Democratic Space

126

92

36

52

0

2

(36)

Election Prediction Project

125

94

36

51

0

2

(38)

Note – All of the seat projections presented in this table were published on October 13, 2008.

Popular Vote

CPC

Liberal

NDP

BQ

Green

Aggregate Error

Actual Results

37.6

26.2

18.2

10.0

6.8

Angus Reid Strategies

(Oct. 11-13, n= 1039)

37

27

20

9

7

(4.4)

EKOS

(Oct. 11-13, n=2561)

34.8

26.4

19.4

9.8

9.6

(7.2)

Ipsos Reid

(Oct. 7-9, n=2000)

34

29

18

9

8

(8.8)

Harris Decima

(Oct. 9-12, n=1279)

34

25

19

11

9

(8.8)

Nanos Research

(Oct. 10-12, n=1502)

34.2

26.7

21.4

9.5

8.2

(9.0)

Strategic Counsel

(Oct. 10-11, n=2000)

34

28

19

9

10

(10.4)

Segma

(Oct. 5-9, n=1500)

34.6

22.9

20.5

10.3

10.2

(12.3)

Note – The figures presented in this table are from the last pre-ballot polls published by each organization.

Click here to download PDF: election-08-report-card-oct-271

Comments are closed.