Tuesday, December 15, 2009

New Angus Reid

Angus Reid provides us with our first non-EKOS national poll in 3 weeks. Compared with the last two EKOS polls:

- The Tories are doing better in BC, which usually happens when comparing Angus Reid polls to EKOS polls. Also, their Québec result is similar to the one in the latest EKOS poll, which is 5 points lower than in previous polling. Copenhagen?

- The Liberals are doing better in Atlantic Canada, Québec (picking up votes from the Tory drop?) and Alberta (useless).

- The Bloc is also up.

- No significant change for the NDP.

- The Greens are way down, but again that's probably due to the different polling methods of EKOS and Angus Reid. The only two pollsters that tend to produce reliable results for the Greens are Nanos and Angus, and both have them at just 6% nationally in their latest poll.

In terms of seats, the Liberals retake the seat they lost to the NDP in the last update, and the Conservative drop in Québec makes them lose two seats to the Bloc. Thus:

CON - 145
LIB - 72
BQ - 53
NDP - 38

2 comments:

Anonymous said...

The notion that the Nanos and Angus results are more reliable measures of "true" Green support is incorrect and shows a (commmon) misunderstanding of the nature of polling between elections. The parameter which we are interested in estimating is the incidence of all eligble voters who currently prefer the Green Party . To compare this to the last election and conclude that because the lower figures must be "right" confuses the issues of predicting some future hypothetical election with the problem of modelling current support levels amongst the overall population of eligible voters.
Clearly the Green Party support is higher amongst eligible voters than it has been amongst the smaller population of voters who will actually vote. This problem of the gap between eligible and likely voters becomes more meaningful as we approach an actual election . It may , in fact make sense at that point to report two separate estimates (likely and eligble voters). At this point, however, it makes no sense to do so and it is ill advised to discount those who would select the Green Party when given a choice on a ballot (if they actually showed up). By not prompting with GP one will achieve a lower estimate but this will bias the estimate of how all eligible voters are choosing . A correction for likely voters will make the proper adjustment as an election looms . In the interim I have no doubt that the higher estimates are a better measure of the overall popularity of the Green Party.
We are not interested in predicting a hypothetical future election now and the lower estimates are indeed the biased indicators of the current intention of the entire eligible population. It is , however, important to note the caveat that current GP supporters are hsitorically much less likely to actually vote than supporters of other parties. It remains to be seen if the attrition from eligible to actual voters will be of the same magnitude in the next election as it was in the past one.

Election Watcher said...

I have no doubt that the eligible vs. likely voter effect explains *some* of the discrepancy with the high Green numbers by some polling firms, and actual election results. However, I think that it would be a mistake to attribute all of the difference to that effect. For example, frustrated voters may say "Green" to a pollster, even though they don't actually support the Green Party, and wouldn't pick the Green candidate even if they actually show up to vote. Such protest responses to pollsters are much more likely when "Green" (or any other minor party) is explicitly given as an option.

So I would agree with the assertion that the true level of Green support is probably somewhere between the ~6% given by Nanos, and the ~10% given by EKOS and others. However, Nanos would produce results closer to actual vote tallies. You might say that this occurs by luck (they get lower numbers because they don't prompt, while the actual reason is likely vs. eligible voters), but it doesn't negate the fact that the Nanos numbers tend to be better predictors of election results. Your point that all this might change if Green voters become more motivated is well taken, but I believe that this is unlikely to happen on a large scale.