The most accurate polls in the 2012 election have the race tied.
I’m really intrigued by the wide variance in polling results for this year’s election.
The most publicized polls – usually associated with MSM or university sponsors — have Hillary up by 5 or 6 points.
That feels about right to me.
But, there are a couple of outliers that have Trump either tied or marginally ahead.
Those polls are usually dismissed by the MSM as “unscientific” since they might not use “pure” random sampling or might not have a live call center person asking the questions or use some form of rolling sample technique (vs. a fresh start with each poll).
Rather than looking at methodologies – all of which have issues – I thought I’d look at past performance ….
Nate Silver is the polling darling of the left – largely because he boldly predicted Obama’s win in 2008 – early, often and loud.
His FiveThirtyEight group did an analysis of pollster accuracy in the 2012 election.
Results are revealing …
For starters, note that 4 of the top 7 pollsters – based on actual 2012 performance — use the much blasphemed internet polling methodology.
Pollsters using the RoboCall method (“press 1 for Hillary, 2 for Trump”) fared in the middle of the pack.
Traditional phone call surveys were sprinkled throughout the standings … and landed in 5 of the bottom 6 spots.
Also note that a couple of the most often reported “scientific” polls – Quinnipiac, Marist, Washington Post / ABC – are in the middle of the pack
Two of the current “outlier” polls – IBD and LA Times (nee. Rand Corp poll) were #1 and #4 in accuracy.
Let’s dig a bit deeper …
The most reported polling result is probably the RealClearPolitics (RCP) poll-of-polls … the average across the half-dozen or so polls that RCP deems to be the most credible.
In 2012, the RCP average just before the election – the purple line below — pegged Obama’s lead at about 1 point.
Obama ended up winning by about 4 points.
That’s a pretty big miss.
The poll that came the closest according to RCP was the Rand Poll … the aqua line above.
Rand had Obama as a 3 point favorite … within a point of the final tally.
The RAND poll was rebranded as the USC-LA Times poll.
Recapping, the IBD poll was most accurate in 2012 – according to left-leaning Nate Silver’s 538 group.
The LA Times (LAT) poll was # 4 on Silver’s list and outperformed the RealClearPolitics average in 2012.
So, what are LAT and IBD reporting these days – post the 3rd debate and the “will he accept the results” dust-up ?
In weekend polling, LAT has the head-to-head race a dead-heat.
What about IBD?
In a 4-way race (which LAT doesn’t survey), IBD scores the race a dead heat.
FYI: IBD had Trump up by a point before the debate … all survey responses are post-debate, but pre-Gettysburg.
Summing up: the polls which were #1 and # 4 in accuracy in 2012 have the race as a dead heat.
But, the polls are under (or never reported) … pushed aside for polls that have Clinton up by 5 to 10 points.
This will be fun to watch.
Both the Washington Post poll (Clinton up 12) … and the IBD poll (dead heat) … can’t be right.
Given the extreme poll results, one of these pollsters is going to end up with serious egg on their face.