Steve Singiser of Daily Kos points out something interesting. While the Democrats have been doing worse lately on generic preference polls (e.g., “Do you prefer a Republican or Democratic House/Senate?”), many of the most hotly contested Senate races have been moving modestly in their direction.
AK-Sen: Last five avg—Begich +0.8; Previous five avg—Begich +0.4
AR-Sen: Last five avg—Cotton +1.6; Previous five avg—Cotton +0.2
CO-Sen: Last five avg—Udall +2.8; Previous five avg—Udall +0.6
IA-Sen: Last five avg—Braley +1.0; Previous five avg—Braley +0.8
LA-Sen: Last five avg—Landrieu +0.2; Previous five avg—Cassidy +1.0
MI-Sen: Last five avg—Peters +5.0; Previous five avg—Peters +4.0
NC-Sen: Last five avg—Hagan +2.2; Previous five avg—Hagan +1.2
It’s hard to be certain what explains these contradictory trends. A shift from registered voter polls to likely voter polls should shift races in the Republicans’ direction, but that doesn’t seem to be happening in the polls of actual races, which could mask even greater recent gains for Team Blue. One possibility is that the DSCC and the individual candidates are doing a great job, while areas of the country without competitive Senate races are seeing the Dems slip. In this scenario, the generic polls really have little bearing on the Senate races.
I don’t know the answers but I will say that the recent dump of YouGov polls is shockingly negative for the Democrats and includes many outliers that are driving down the polling averages for Democratic candidates. I’d like to see someone take a crack at explaining the methodology that is producing those results.
To test the validity of a poll, simply check its internals on racial composition.
If it says that fewer non-white voters are going to the polls than in 2010, throw it out.
If it says that black voters are not lining up behind the Democrat according to historical trends (>80-90% no matter who the candidate is), throw it out.
Bye bye, YouGov polls. Heck, middle aged and older black folks might be the only damn Democratic constituency to even turn out at all in six weeks. If a poll is undercounting them, it’s systematically (and I believe deliberately at this point, based on 2012) biased for Republicans.
Hagan will win. Udall will win. Braley will win. Begich will win. Nunn and Pryor should be considered favorites given their opponents’ inability to open even a hint of daylight. Landrieu is in trouble in a runoff. Only Grimes is doomed, given her inexplicable choice to run just as hard against Obama as McConnell. The best possible starting hand for the GOP will net them a whopping two or three senate seats. Democrats will almost run the table on governors races. That’s how bad Republicans suck.
I was going to reference this site in regards to Kentucky, but I see that in Illinois they have Rauner in a seven point landslide over Quinn. No way, unless everyone in Chicago boycotts the polls.
I don’t think Rauner is going to win. I don’t see enthusiasm for him in Republicans. He might win, but by a squeaker.
Quinn only has himself to blame for the shape he’s in. What does he expect when you shit on your main constituencies?
Democrats savaging the Unions. Shaking my head.
Sounds about right, and about what my projections say, too.
I had Grimes as neck and neck with McConnell back in July and early August, but if I had to tip the scale I’d have given it to McConnell. Now I’d say Grimes will lose for sure.
Ask again in October if she can turn it around, but trend at the moment is doing the opposite of what I wanted to reverse my previous position.
Not sure I’d put Nunn as the favorite, though. How is Carter polling?
In 2008 when KY voters preferred McCain 57% over Obama 41%, McConnell won his re-election 53% to 47%. The polling about this time in 2008 had McConnell within 2 to 4 points of his final number. Lunsford’s numbers were within 1 to 17 points of his actual performance indicating that the undecideds broke mostly in his favor.
Current polling has McConnell about where he was in 2008 and Grimes numbers are better than Lunsford’s were. While there’s still room and time for her to beat McConnell, not seeing anything that would suggest to me that she can close the deal.
They are scheduled for a debate on October 13th. McConnell only has to perform well enough for a draw. Grimes needs a commanding performance to get enough of a bounce to win the election. Or a solid and better performance than McConnell and an excellent TV ad campaign in the last three weeks. Will she have enough money to purchase the TV ad slots?
My impression is that the amount of partisan polling is way up. I think the monied elite has noticed that polling has a strong influence on the media, especially after several years of good results for aggregators like Silver and Wang. So now I think a lot of polling is being done to influence donations and reporting rather than to find out what people really think. For this reason, I suspect there’s a lot of deliberately bad methodologies intended to skew the polls. The era of poll aggregators may also be coming to an end, because GIGO.
There is truth to that. Part of the reason Rasputin – I mean, Rasmussen – ran an avalanche of on-line polls in 2012 was to overload the aggregators with pro-right wing results. It worked on the aggregators that simply averaged every poll.
If Colorado votes for the GOP for either Governor or Senator I will be more depressed than usual. Both of these guys are as extreme as they come, but unfortunately I’m not seeing much coverage showing that. I am seeing a lot of ads from the Reich Wing Billionaires tying the incumbents to Obama, including Obamacare, and yet not seeing much pointing out how many Coloradoans are benefiting from ACA (we have an excellent health care exchange).
The Governor candidate, to illustrate the extreme thinking, literally is proposing seizing the national parks and BLM lands to manage as a state and selling off the assets. I have to think that if half of their positions were known it would be a 60-40 landslide for the incumbents.
Maybe a touch of at least our guy is better than gonorrhea?
People don’t vote for “Congress”. They vote for specific members of Congress. Traditionally everyone hates “Congress” but loves their long-term incumbent.
Republicans in red states are going to break somewhat toward Democrats because they have actually experienced Obamacare. They are slowly going to identify their preference as the election approaches. They still hate the Democrats in Congress, except for the Republican nutcake in their state races they hate worse. Tom Tillis is running against Pres. Obama on women’s internet craft channel ads. Isn’t that very interesting? And it might get a backlash of “How stupid do you think I am?”
The GOP contempt for women in general is cutting over against their ability to motivate on cultural issues.
The whole notion of risk-managed nationwide campaign in which some races are more important than others for the national and state establishments is hurting candidates who are perceived by the establishment as not comptetitive–which becomes a self-fulfilling prophecy.
Also hurting is the failure to take the issue of Koch money and what it is doing head on to the voters. The establishment loves its cozy income prospects in its post-political career.
If Democrats win or even hold their own this year, it will be because local activists pulled together GOTV vote campaigns that mobilized voters who rarely turn out in midterms. Hagan is benefiting from the Moral Mondary activism raising a number of registered voters to likely voters in polls. Cotton is benefitting from Koch-sponsored media saturation bombing of Arkansas–in September. And Kochs are turning on a Sagebrush Rebellion campaign to privative federal land in the Colorado campaign against Udall. How have national Democrats responded?
Who and how does YouGov poll in order to get so many races polled so quickly? Who finances YouGov polls?
I’m still of the opinion that if the polling is predictive of this election, it shows that the political culture is not changing sufficiently from previous elections to stave off disaster. What is different this midterm is that are major grassroots organized efforts to change the political culture and undercut the power of money in campaigns.
screens are just bad. It was a problem in 2012 and I think it will be a problem in this midterm. They also have a problem with getting too many old and white voters in these polls especially YouGuv.
I know I’m a pessimistic guy, but it feels like there is a lot of ‘unskewing’ going on in this thread.
Part of Statistical Analysis is “interrogating the numbers.” One step is to look at the Sample and see how it compares to the Statistical Population. If, say, men are 65% of the Sample* then there is a screw loose. In a real world example, white voter percentage of total voters is dropping roughly 1% every two years. If a poll is constructed using the percentages from 2010 for their voter screens it is almost certain to over-sample white voters.
These are Real World fundamentals having nothing to do with “unskewing,” as I understand the term.
* Exaggerated for illustrative purposes
I was so looking forward to the results of Kerry v. Bush, based on my reading of leftblogs.
Lieberman v Lamont
There’s oddities in the polling reports. Example, GOP-biased Rasmussen released a poll yesterday (no link) reporting Hagan leading by a couple of points in likely voters but the internals showed her winning women by 21 points among registered voters. Granted there is a fall-off of voter turnout in mid-terms but their top line result can only happen if no men vote for Hagan, something I find unlikely.
A second point: national polling of Voting Intention is worthless for predicting state results. The statistical populations don’t match; a Voter’s Intention in South Carolina is predicatively null for the North Carolina, Colorado, Louisiana, and Michigan races.
I don’t know what is happening. It could be malicious. It could be unconscious bias. It could be incompetence. It could even be a combination of all three.
The polls are probably as good and bad as they always are.
Hagen was only slightly vulnerable before the GOP selected her opponent. Hasn’t been vulnerable since then.
Not to be dismissed are the Kansas polls for Gov, Sen, and SOS All three GOP could go down in Koch territory. Probably large pots of money ready to be spent in the closing days of the election in support of the GOP candidates. But maybe, just maybe, Kansans have had enough of RWNJ governance.
One minor quibble about the way polls are reported. Polls that don’t report margin-of-error are essentially worthless because you have no idea of the band of randomness in the poll.
As for moving averages, IMO, polls that have distinct trends are in races in which no one is doing anything strategically. Moving averages are great for tracking trends in climate or unemployment in which there can be very little assumption of agency. Campaigns however are so focused on poll numbers and moving the needle that a consistent trend show someone in the campaign office asleep at the switch. And observation YMMV.
Are any election poll results issued that don’t include the MOE and the confidence level (most often 95%)? Most pollsters also release a drill down on the sample populations because that’s where a sample bias would more easily be detected. MOE is goes down as the sample size increases (regression towards the mean of the true population).
What the average Jane blogger seems not to grasp fully is how to read the polling reports. And how to incorporate polling trends and aggregations from various pollsters.
The Loras College Poll September 2014 (Iowa Senate race) is about as good as it gets. Still, after including likely and leaners for the two candidates, 14.2% are undecided. Braley’s at 45.3% (+-2.82) and Ernst at 40.5% (+-2.82). On respondent shared values and interests between the two candidates the respondents are less confident and there’s less separation between Braley and Ernst. On “heard of/favorable,” almost no difference between the two. On “heard of/unfavorable: Ernst 39.5% and Braley 33.7%.
The September results differ significantly from the June results that had Ernst at 48% (+-4) and Braley at 41.7% (+-4). Ernst’s favorables were higher and her negatives were lower in June. Braley’s favorables and unfavorables both increased but is positive.
The June poll sample was half the size of the September sample. (Hence the larger MOE) But more significant is the difference in party registration/preference between the two polls: June: 39.3/36/27.4 (R/D/I) September: 30.2/32.6/33.8. So, what is the true population by political party ID in Iowa?
In 2010 it was 29/34/37. No wonder Ernst had such good numbers in the June poll!
Looks like the undecideds would have to split 70+% in favor of Ernst for her to pull this out. Given her negative favorable:unfavorable ratio, that’s highly unlikely. Braley is getting close to closing the deal.