Ipsos exit poll of 2011 election (Canada)

[King of Kensington]

I've been going through it recently (joined LISPOP) - curious to know how good/reliable people think the Ipsos Weighting Factor is.  Ironically when I e-mailed Darrell Bricker - who has said a good polling company worth their salt can always tell you the detailsabout their weighting methodology - he just responded that it is "just standard demographics" (presumably, age, gender, region, etc.)

Anyway, been plowing through it and will try to give some demographic analysis of various voting groups besides the standard "chart" we've seen in the papers.

[Hatman 🍁]

looking forward

[DL]

This is the same Ipsos whose Ontario polling was wildly off last month. Caveat emptor.

[King of Kensington]

So you're saying their weights are garbage? 

[Hatman 🍁]

Exit polling is going to be far more reliable than using an opt in panel.  Ipsos' exit poll of the BC election was more accurate than any in-writ polls.

[King of Kensington]

The reason why we didn't hear about the "Sikh vote" is because it was a very small sample, just 45 respondents.  Skews on the younger side, but that's not surprising they're a young group.  It is dominated, not surprisingly, by Ontario and BC and the numbers aren't out of whack with the proportion of the Sikh populations in those provinces.  

Unweighted it went 44% NDP, 27% Liberal, 24% Conservative.  Too small to really isolate further.

Being a small sample: more suggestive than representative.

[Hatman 🍁]

Not surprising it would go NDP though.

[DL]

Exit polling is going to be far more reliable than using an opt in panel.  Ipsos' exit poll of the BC election was more accurate than any in-writ polls.

Keep in mind that Ipsos' "exit poll" was not an exit poll the way we think of it from US and UK exit polls. They simply asked everyone from their opt-in panel how they voted. It's still interesting to know how members of the Ipsos opt in panel voted, but as we learned in Ontario, you can have a perfectly demographically weighted online panel that just has way too many Conservatives in it....though to be fair the final polls by Ipsos in the 2011 federal election performed well, not like their wildly inaccurate BC and Ontario polling programs.

[Hatman 🍁]

Oh, I wasn't aware of that. I suppose if they weighted by the election results, there would be some value in the data.

[DL]

Oh, I wasn't aware of that. I suppose if they weighted by the election results, there would be some value in the data.

Yes actually - that could work

[King of Kensington]

Hindus (unweighted) went 35% Conservative, 34% Liberal, 25% NDP.  Of course there is no way to split Indian Hindus and Sri Lankan Hindus though the former group is larger (and the latter more concentrated in a few ridings).

[Filuwaúrdjan]

Well we already know how Tamils voted without needing to check sketchy exit poll figures...