Tuesday, October 6, 2015

i think the mainstreet poll shed a little light into this, if you check the full results. this is the mainstreet poll:

conservative: 32
liberal: 24
ndp: 20
undecided: 15

that puts the conservatives in the range that nanos is putting them in (i have no faith in the internet polls), but the liberals and ndp lower. now, if you take into consideration that there is broad consistency that the conservatives are getting almost none of the undecideds, what it really seems to suggest is that that "promiscuous progressive" vote is coming up split in the ivr polls.

i've worked as a telephone interviewer, and i can tell you that it is common to get somebody on the phone, have them say they're undecided, and then have them say "well, i'll probably vote for ...", in which case you have to record that. you can't do that on the ivr. and, yeah, the internet polls may be splitting the difference due to something like this, too.

it may be that the split the ivr is picking up is more accurate in determining how many people are literally undecided, but less accurate in picking up where they're leaning. and, so, the better option may be to take them both seriously.

if you take 10 of that decided and give it to the liberals, and take 5 and give it to the ndp, it's pretty close to the nanos poll. and, you can justify that by pointing out that the mainstreet poll has a much larger sample size.

i wish nanos called more people. but, i do prefer his methodology because that discussion is far better in getting an accurate response. at least ivr is still random, so i'll take it over internet. but, the key may be in interpreting the results literally. and, it may be saying that, in this election, the conservative vote is decided and the ndp/liberal vote isn't.

www.cbc.ca/news/politics/canada-election-2015-grenier-polls-oct6-1.3258683

Mint-Berry Crunch
I guess you forgot that Nanos was most accurate last election.

jessica murray
no, i agree that nanos is the benchmark - not just from last election, but over the last several elections. but, that doesn't mean one can discard the other polling. what you want to do is find a way to make them consistent with each other.

what i'm saying is that nanos may be picking up people that are *leaning* liberal right now, but that are in fact still undecided.

that would mean that the ivr polls are overweighting the conservatives *because* they are properly measuring the undecideds. if you're merely leaning liberal right now, there's no contradiction in telling nanos that you'll probably vote liberal and telling mainstreet that you're undecided.

there's also still two weeks to go. that's half the length of a normal election campaign.

Conned No More
Those 30% are die-hard radical populists. http://www.4data.ca/ottawa/archive/harper/harper-noble-lie.html

jessica murray
i'm not going to read that right now. i've probably read it before. but, yeah. the conservatives don't go under 30% very often. and they've never received less than 27.

Mint-Berry Crunch
Harper thanks you for this predominantly Conservative attitude.

jessica murray
oh, get real. i'm a libertarian socialist.

rather, i might suggest that ignoring the evidence you don't like is rather conservative, in nature.

it's not hard to make the data consistent. live interviewers can get a better handle on where people are leaning, whereas ivr gets a better handle on absolute categories.

so, you can read it consistently by looking at the ivr data as core support and the live phone data as leaning support.

that would mean that nanos is probably the most accurate as to who people would vote for today. but, what the ivr polling is really suggesting is that liberal-leaning voters haven't completely written off the ndp, as of yet

or if you want to give it a liberal spin, think of it like this: it suggests the liberal ceiling right now is somewhere around 40.