[hurley_108]

Blue_Nose Blue_Nose:

Here's the last two election results along with the last poll results from SES/Nanos and Ipsos-Reid. I just summed the total error across the five parties for each pollster.

I didn't easily find poll results from 2000, so that's as far back as it goes.

Last election was certainly a good one for Nanos, but he's only marginally better than Ipsos-Reid in 2004.

0:

results.PNG

We can always trust Blue_Nose to be the voice of reason. SES (now Nanos) was within a tenth of a percent for the four major parties last time. The rest of them were well off, and most biased by about the same in the same direction for the Liberals (about 3 too low), though they pegged the Conservatives pretty well.

That is why I watch Nanos - because they nailed it last time. If this time they blow it, I won't watch them so close next time. But as you conservatives like to say, the only poll that matters is the election, so why do you care what one polling company says?

[Tricks]

Taking one poll as more correct than another regardless of who, or history, is just plain stupid. Take the central tendency of all of them is far better. Which is why the link Blue posted is one of the best to go by.

[Blue_Nose]

hurley_108 hurley_108:

That is why I watch Nanos - because they nailed it last time.

OK, but they were off by 8% in 2004, about as bad as Ipsos-Reid's results in 2006 - did something significant happen to their polling technique, or did they just get lucky with the numbers that year?

All I'm saying is that one year of good numbers doesn't suddenly make them the best, and from what I understand SES didn't poll at all in 2000, so those are the two results we can go by.

[hurley_108]

Blue_Nose Blue_Nose:

hurley_108 hurley_108:

That is why I watch Nanos - because they nailed it last time.

OK, but they were off by 8% in 2004, about as bad as Ipsos-Reid's results in 2006 - did something significant happen to their polling technique, or did they just get lucky with the numbers that year?

All I'm saying is that one year of good numbers doesn't suddenly make them the best, and from what I understand SES didn't poll at all in 2000, so those are the two results we can go by.

SES / Nanos is pretty new, but something big did change between 2004 and 2006, but which has not between 06 and now - the sample size doubled. SES in 04 polled 200 people per night, for a three-day rolling average out of 600. In 06 that was bumped to 400 per night, for a three-day rolling average out of 1200.

A poll from 2004.
A poll from 2006.
The poll from yesterday.

[Blue_Nose]

Updated results:

0:

partylinesnew.php.gif [ 13.34 KiB | Viewed 286 times ]

There's definitely something odd going on - looking across the entire record, the only major discrepancy in poll results has been between Liberals/Greens, Nanos/Harris-Decima. It happened in mid-September, and it's happening again now, with a period in between where results were similar.

[Toro]

DerbyX DerbyX:

Toro Toro:

DerbyX DerbyX:

You keep demanding that other people do your research.

YOU do it. You determine who is the most accurate or if you simply consider them all innaccurate then you shouldn't even be watching their trends for any information.

You made the claim. You back it up. Walk the talk.

I did. I provided the evidence to back up my claim.

One result - or two results - is not statistical evidence. Sorry.

[Blue_Nose]

Blue_Nose Blue_Nose:

Here's an update with this election's results to find out if Nanos is as amazing as their results from the last election supposedly indicated. I included the last polls performed by some others in the last days before the election for comparison.
(Source)

0:

results2008.PNG [ 10.32 KiB | Viewed 216 times ]

Short answer: Nope, looks like they got lucky.

[hurley_108]

Blue_Nose Blue_Nose:

Blue_Nose Blue_Nose:

Here's an update with this election's results to find out if Nanos is as amazing as their results from the last election supposedly indicated. I included the last polls performed by some others in the last days before the election for comparison.
(Source)

0:

results2008.PNG

Short answer: Nope, looks like they got lucky.

Apparently so. Thanks for the analysis, Blue.

[mtbr]

Angus Reid the online pollster is the new champion, so much for phoning teenagers on their cellphones nicky

[Toro]

Nanos consistently skewed the NDP higher than the other four polling firms during the final weekend.