Chanting Points Memo: Camouflaging The Battleground
By Mitch Berg
The Strib “Minnesota Poll” is doing what it’s paid to do: create a pro-DFL bandwagon effect, and suppress GOP voter turnout. It’s calling Minnesota at Obama with 48% and Romney with 40%.
But the poll uses the same absurd D41/R28 breakdown that the Marriage and Voter ID polls. This polling would have you believe that while in 2008, with a messianic media darling running against an unpopular two-term candidate (McCain was irrevant) and the war the DFL had a six point advantage in partisan turnout (D39 R33), this year, mirabile dictu, we have a 13 point Democrat advantage in this state?
If you use turnout numbers from somewhere in between 2008 and 2010 – say, D36 R34 – and multiply the changes by the percent of each party that the poll itself says plan on voting for their candidate (93% of Democrats plan to vote for Obama, vs 96% of Republicans), then you wind up lopping off roughly .3% of Obama’s numbers, and adding a whopping 5.8% to Romney’s.
That makes the real split 47.7% Obama, 45.8% Romney.
Question – especially for you libs in the audience: In what way is a widely (one might say “lavishly”) publicized poll using a partisan split that this state hasn’t seen since Watergate to be interepreted as anything other than an elaborate voter-suppression scam?
September 24th, 2012 at 2:31 pm
Nate Silver as 538 aggregates all other MN polls and comes up with mostly the same result as the Strib.
FiveThirtyEight Projections Dem Rep Margin
Polling average 50.3 41.7 Obama +8.6
Adjusted polling average 49.6 42.8 Obama +6.8
State fundamentals 51.2 42.0 Obama +9.2
Now-cast 50.3 42.5 Obama +7.8
Projected vote share ±4.9 53.2 45.2 Obama +8.0
Chance of winning 95% 5%
And if you think he is biased, his 2010 predictions were Senate 51.7 Ds (+.7 for Ds) and House 232 Rs (-10 seats for Rs)
September 24th, 2012 at 2:47 pm
“The Strib “Minnesota Poll” is doing what it’s paid to do: create a pro-DFL bandwagon effect, and suppress GOP voter turnout.”
Maybe they’re trying to keep down voter fraud.
September 24th, 2012 at 2:50 pm
Silver is fine and dandy, but he has a key clinker in his methodology; he aggregates polls, weights them (partly by recency, partly by some secret sauce known only to him) and generates his results.
The problem in 2010 (and I haven’t looked at his figures for this year yet), when he predicted Mark Dayton by a six point margin, was that he gave the Strib and HHH polls a *lot* of weight on the MN Poll and relatively little on some of the better polls. He gave an awful lot of weight to the Strib’s foetid polling – and it showed.
And, I suspect, it still does.
September 24th, 2012 at 3:29 pm
Here is his 2010 prediction for MN. Strib was middle of the pack for accuracy and was not his top weighted poll. Looks like it was tied for 4th in weight. In any case, the result was in the margin of error.
http://elections.nytimes.com/2010/forecasts/governor/minnesota
September 24th, 2012 at 3:34 pm
The margin of error was eight points.
Please.
September 24th, 2012 at 5:42 pm
OK, first of all, MPR and Humphrey were not “middle of the pack” according to the data Rick presents. Rather, their results were tied with St. Cloud State’s for the most off. Sorry, Rick, but you seriously need to learn to read a chart.
Moreover, note as well that the margin of error in the MPR document is calculated–per proportion test results–at +/- 3.6%, far less than the difference between the actual results and the predicted results. In short, there is something wrong with the MPR methodology that is statistically significant.
What’s wrong with it? I’m not sure, but notice here that the Minnesota Poll spread of 10 points dates back to July, while the posited Dayton lead over Emmer really shows up in August. One would suggest that it’s not just a leading indicator, but is rather a leading agitator.
I would suggest that if ethics ruled at the Strib and MPR, the polls would either disclose and fix their methodology, or they ought to be shut down.
September 24th, 2012 at 7:59 pm
The results of the 2010 gubernatorial race were 43.6% Dayton, 43.2% Emmer.
The Strib called it 41% Dayton, 34% Emmer, a spread of 7%.
The Strib poll claimed a margin of error of 3.9%
http://www.startribune.com/newsgraphics/105573478.html?elr=KArks8c7PaP3iUHc3E7_ec7PaP3iUiacyKUbPi87EK_g:D_GD7EaDh_0c:aD:aUr
Survey USA, Rasmussen, and PPP showed the spread was within 2 pts. Every other survey dramatically underestimated Emmer’s percentage.
September 25th, 2012 at 7:44 am
Don’t blame Democrats for polls to suppress conservative White voters in Minnesota. The New Black Panthers are understaffed here.
.
September 25th, 2012 at 11:47 am
Bubbasan, you need to learn to read. I said the “Strib” poll was in the middle of the pac for accuracy, not the “MPR and Humphrey “.
September 25th, 2012 at 2:58 pm
OK, I confused two polls–but both of them are not middle of the pack, but rather in error in a statistically significant way. Sorry, the methodology on the Strib is off just as certainly as are MPR/Humphrey and St. Cloud.
September 25th, 2012 at 4:39 pm
Still, a good suggestion, Rick. It should wouldn’t hurt anything if bubba were to learn how to read.
September 25th, 2012 at 4:40 pm
“Sure” wouldn’t.
Angryclown has sacked his intern for that.
September 26th, 2012 at 10:58 am
bubbasan, Mitch claimed that the Strib was a uniquely bad pro-dem poll. The 2010 polls show that the Strib was in the middle of the pack for accurately predicitng the final result. Pointing out other polls are worse only undermines Mitch’s point.
If the Strib is trying to depress the GOP why does their poll give Emmer better numbers than Survey USA, St. Cloud, and MPR/HHH which all had a 10 point plus lead for Dayton. If Mitch’s thesis was correct the Strib should have had a 15 % lead for Emmer.
September 26th, 2012 at 11:16 am
Rick,
You’ve got it completely wrong.
For starters, as I pointed out in the series I wrote in 2010, the HHH and SCSU polls are worse than the Strib Poll. Their numbers are even more skewed, and more prone to “distortion” in close races. The HHH’s final pre-election poll in 2010 was even more ludicrous than the Strib’s.
The “Strib” is only “in the middle of the pack” because the other ones are even less well-done, or more in the bag for the DFL.
September 26th, 2012 at 11:44 am
Mitch, fair enough. Did not read your 2010 polling review. Your problem is now that it is far easier to imagine one rouge polling organization leaning pro-dem. If your claim is that nearly all polling organizations conspire to promote Dem candidates in MN, then you have entered tinfoil hat territory.
October 22nd, 2012 at 8:29 am
[…] We saw some sketchy evidence of this in the last Star/Tribune “Minnesota” poll, of course; if you rolled the turnout model from its’ absurd +13 Democrat split even to 2008 turnout numbers, Romney was within two points. […]
October 26th, 2012 at 11:54 am
[…] latest Star/Tribune poll showed Obama with an eight point lead, it took a D+13 sample to do it. If you roll those numbers back to even 2008 Democrat turnout levels, the race is within 3-4 […]