Over at SNP Tactical Voting Jeff is cautioning his readers not to give too much credence to the latest UK-wide Mori poll placing the Tories a full twenty points ahead of Labour. He makes the observation that among the entire electorate the Tory lead is a mere seven points, and the dramatic headline figure is generated by only including those respondents who report they are 'certain' to vote at the next general election. In some ways he's got a point - different pollsters use different filters and weightings to arrive at their headline figures, so the numbers from different companies should never be treated as directly comparable.
Where I think Jeff is on slightly shakier ground is in implying that the Mori figures lack credibility on their own merit. After all, Mori apply the 'certain to vote' filter for a very good reason - it's because they feel on the basis of past evidence that it's the figure most likely to prove accurate. They may or may not be right about this (other pollsters apply a lower threshold of 'likely' or 'very likely' to vote) but they're certainly not doing it in an attempt to generate artificially dramatic results. For one thing, it wouldn't be in their own best interests do so - a polling company lives or dies largely by its reputation for accuracy.
Incidentally, to give another example of what a dramatic effect Mori's filter can have on the result, in the 'northern region' sub-sample the SNP are on a very healthy 10% of the vote among the entire electorate - but among those 'certain to vote' it falls to a pathetic 4%! Given the low numbers involved, I think we can probably put that discrepancy down to sampling issues.