Frank Graves, president of Ekos, sent this note on what ITQ readers will already know is a fairly poll-intensive day. Graves argues for his methodology and promises more poll-related fun over the next few days.

Hi Paul

I think the real value and implications of this poll may be missed. It has a huge sample. As you know it also includes cell phones( i believe you received a call from our robot). [True story — pw] One unreported finding is that those who don’t include the growing cell only population (nearly one in four households in the United States now and Canada is trailing but following that pattern) will have increasing difficulties covering key parts of the population . Liberal supporters, for example, are higher in the cell only segment. This trend to cell only households is also much more pronounced in Quebec adn amongst 18 to 40 year olds of moderate and lower socioeconomic status.

What is unusual, and we think valuable about the poll is the ability to drill down and get a very detailed demographic and regional profile of the eligible voter population. Because of the very large sample sizes we can also examine daily trend without resorting to the 3 day moving average approach. This allows us to see some interesting shifts and test some key hypotheses. For example, we saw the Liberals opening up a five point lead since the announcement of the 50B dollar deficit. It’s possible some other factor explains this highly significant shift but it’s highly unlikely. It is also notable that the Tory “attack” ads don’t appear to be bearing any immediate fruit for them.

There are a range of analyses we will be rolling out in the coming days. For example we will be analysing where voters have migrated since the last election adn who is doing well with first time voters.

The economy is clearly a crucial factor but different parts of the economy are having diverse effects in different segments of the population. For instance, It appears that a sense that the economy may be turning the corner is overall working mildly in favour of the Prime Minister and the CPC. Although Canadians are not highly optimistic a sense that things look better than they did three months ago is growing amongst older more affluent voters, who had defected from the Tories but were starting to return . Alarm over the deficit numbers may be extinguishing this recovery for the conservatives. This mild recovery of confidence in the economy is not being felt by younger voters who are more tuned into labour markets (which continue to deteriorate) as opposed to the equity markets that affluent older voters are following. These contradictory forces are crucial to understanding the shorter and medium term prospects for the parties adn they should factor into election timing questions.

In a shameless plug for the new suite of polling methodologies we are developing I note that the in depth analysis of diverse demographic segments is beyond the reach of most on-line polls which are restricted to self selected group which opts in to the “panel” . These on line approaches also excludes a very large number of households which don’t have broadband internet access. The approaches we are trying to bring together are based on the conviction that all members of the population should be able to appear in one’s sample and that the only method for recruitment is through a random invitation to participate. Problems with rising rates of refusals are no reason to abandon probability sampling . We will be linking this work with more detailed diagnostic work using our hybrid online-phone panel which allows in depth questioning of a more lengthy issues outside the reach of an IVR (robo-poll). The problems aren’t the online method of contact; it can work very well. It is the problem of assuring random probability sampling while using those methods. In our view, all polling should be based on random, probability sampling that includes the entire eligible population.

Cheers

Frank