see, again, i’m starting to wonder if you need to interpret the undecideds differently when measuring over ivr – like, literally, rather than throwaway as following the same distribution. at 7.7% undecided, scaling down puts the conservatives at 32.4 absolute response rate – which is consistent with the other numbers, although it’s also almost in the margin of error. even 33-34% doesn’t strike me as inconsistent, it’s the consistently lower liberal numbers that are weird and, regionally, that seems to be in bc and quebec. i’m even willing to accept that nanos may be overweighting the liberals in bc. but, what was the sordid story about underweighting the liberals in bc, again? aha. but, then if you scale them down you get to 28.5, which you can the scale back up to the mid 30s (and there’s even a bit room to give to the ndp).

i know this is being very creative and everything, but it does make a lot of sense if you consider that actually talking to somebody is going to get a different undecided response than a telephone system. i mean, yeah: i’m just playing with the numbers. but it adds up. and it makes sense.

that said, it might not be generalizable, it might just be the specific nature of this election, where support for the incumbents is so deeply entrenched, and support for the two opposition parties is so uncommitted.

it’s an idea, to consider, anyways.

although i’m curious about what an “invalid response” is. 11% is a lot. please shoot me an email if you decide to respond in this space, so i can come back and see it.

www.ekospolitics.com/index.php/2015/10/stable-but-narrowing-conservative-lead-as-ndp-in-a-holding-pattern-well-back-of-leaders/