Here are some fun facts for you.
In 2012, President Obama got almost four million more votes than George W. Bush received in his successful 2004 reelection. Mitt Romney got about 800,000 fewer votes than John McCain had won four years earlier.
After all the 2012 votes were tallied, President Obama won the popular vote by 3.8%, which was quite a bit more than when Steve Singiser took a first shot at comparing the likely voter (LV) and registered voter (RV) screens and discovered that the registered voter screens were more accurate.
- Gallup Tracking, 11/5/2012: Romney +1 (LV); Obama +3 (RV)
- Ipsos/Reuters Tracking, 11/5/2012: Obama +2 (LV); Obama +4 (RV)
- CNN/Opinion Research, 11/4/2012: Obama +0 (LV); Obama +2 (RV)
- Fox News, 10/31/2012: Obama +0 (LV); Obama +1 (RV)
- YouGov, 10/31/2012: Obama +1 (LV); Obama +2 (RV)
- CBS/New York Times, 10/30/2012: Obama +1 (LV); Obama +5 (RV)
The most accurate prediction came from Ipsos/Reuters’ registered voter poll. All of the likely voter screens were far too optimistic for Romney.
You should keep this in mind when you look at the latest CNN/ORC poll (pdf) out of Arkansas. This poll has Tom Cotton beating Mark Pryor 49%-47% among likely voters but losing 38%-47% among registered voters.
That’s a huge Republican-tilting screen. That doesn’t mean that it is necessarily wrong, but the recent track record for these types of screens speaks for itself. Singiser looked at the 2004 and 2008 polling, as well, and found that registered voter polls fared better overall in those elections cycles, too. However, those were presidential years. We know that Democratic voter drop off is often pronounced in midterm elections. Still, if polling outfits can’t figure out who is actually going to vote in presidential cycles, why should we trust them to figure it out in midterms?
Finally, I think it is impossible to research, but I have no doubt that President Obama’s final numbers were harmed by the persistently wrong skew of the likely voter polls which gave the impression that the election was closer than it was. It prevented demoralization on the Republican side which kept the volunteers active, kept the donations flowing, and colored how the media portrayed the contest to the public. I still believe there was a time right around the time of the first debate when Romney and the Republicans were on the verge of collapse and that a real blowout was possible. Obama’s poor performance in that first debate stopped Romney’s downward trajectory, but the race was never really close. It only looked kind of close because of those likely voters screens that turned out to be completely wrong.
So, I still think the aggregate of polls is very accurate, but you have to drill down into individual polls to know what you are looking at. A poll that showed Mark Pryor with a huge lead among registered voters is being reported as a narrow deficit among likely voters. And that helps Tom Cotton raise money, avoid questions about why his campaign is floundering, and maintain the enthusiasm of his base.
Is it true that registered voters in Arkansas overwhelmingly support their Democratic senator but that he is going to be narrowly defeated anyway?
Could be. But, I wouldn’t bet on it.
All that is needed to make Arkansas a slam dunk is more Obama Executive Orders of policies conservative democrats oppose.
Why oh why doesn’t Obama see this?
.
I think Democrats still possess a significant advantage in party registration in Arkansas. If the LV model ends up being close to accurate then it may just be a reflection of the trend away from Democrats in that state.
For a long time we’ve all understood and accepted that among “Likely Voters” GOP candidates do better than among “Registered Voters”. It’s just one of those things, you know? Something “everybody knows” – like how Democrats spend way to much and only Republican presidents balance budgets. You know.
Only, we don’t know. Certainly the readers here know that the common perception of the GOP being better at balancing budgets is actually a complete lie. Similarly, the reasons for the LV screens favoring the GOP versus the RV screens never really have been subject to rigorous scrutiny. The pollsters suggest something about whites and the better off being more likely to vote, but if that were true you could test it with statistical correlations in different areas – but they don’t line up.
I think the root of the problem is that the pollsters can, to some degree, measure the electorates intent to vote but have no way to measure just how many resistance barriers have been set up to allow different people to vote or to permit their vote to be counted. Somehow the GOP has successfully put a lot of barriers in the way but they vary from place to place making it hard to measure – so the pollsters are left with inventing LV screens to try to approximate.
Sometimes, like in Florida 2000, the LV polls underestimate just how effective the GOP vote-stopping schemes are. Other times, like in Colorado 2012 they overestimate (Hispanics in Colorado were determined to jump through whatever hoops the GOP Secretary of State put in front of them – and local Democrats were determined to get every last provisional ballot – which were predominately Democratic voters – counted).
In short, the democratic system sux and will be until the GOP can be shunted aside long enough to enact a fair, uniform, national voting system.
An increase the minimum wage initiative is on the AR ballot. That will affect turnout in all the Democrats favor.
Arkansas minimum wage hike plan makes fall ballot
Good.
I’m not sure why every Democrat doesn’t run on it.
Gallup’s (LV) truly sucks — off by 4.8 points. (MOE is generally smaller than that, but Gallup generally has a GOP bias.)
But CNN’s (LV) was only somewhat better in being off by 3.8 points and its (RV) was still short 1.8 points. So we’re to accept an 11 point difference in the AR Senate race between LV and RV? Did Rove construct the LV screen?
“…kept the donations flowing..”
Always, always follow the money. A whole lot of folks made a hell of a lot of money off of Romney’s doomed campaign. Never underestimate the levels of grift in political campaigning.
In the past forty years which three POTUS candidates received the largest percentage of the AR vote?
Scottish independence: Vote ‘will go to the wire’.
Referendum on September 18th.
Why?
If the referendum passes, Cameron’s going to have a more difficult time maintaining his hard line against the Ukrainian separatists.
I do wonder how far right England will go without the Scottish block of lefties.
Thanks for this, Booman. Great post and question.
Adding: Arkansas has a minimum wage referendum on the ballot in November (to raise it from $6.25 to $7.50 on Jan. 1, 2015, $8.00 in 2016 and $8.50 in 2017). That’s a 36% increase in a little over two years.
How, if at all, are pollsters factoring that question into their LV screens? (And if they’re not, why not?)
See how easy it is for the Wall Street media to put their fingers on the scale at the margins. And Democratic candidates still are paying them big campaign media ad buy bucks and subsidizing this. Neat trick that money in politics accomplishes.
It would take some data work, but couldn’t you compare LV/RV screens from midterm elections in competitive races?
Midterms are simply a different animal, but I’d like to think that OFA finally realizes that they made a huge error in 2010 by not fully mobilizing.
Today’s NBC/Marist poll shows Pryor getting thumped in LV and losing RV, and is hemorraghing badly.
Good new is that same pollster has Udall opening up lead outside of MOE in CO in both LV and RV numbers.
The obvious question isn’t answered in the post or comments: What is “likely?”
As I understand it, it varies from pollster to pollster, but uses things like past voting history or enthusiasm.
Likely voter polling screens were developed to filter registered voters who said they were going to vote and then didn’t. Any such screen must be based on historic data. As long as the historic data reflects the actual voting population the poll (prediction) will have a high confidence factor and be accurate within the MOE.
However.
When the voting population is changing, as I would argue it is, the pollster’s Likely Voter screen’s algorithms will cause a drift, as we saw in 2012 when Gallup had white voters at 78% of the total vote when they were actually ~72%. Thus under dynamism polling Registered Voters gives a better sample.
Apologies in advance, but as a numbers person, I always cringe when political junkies learn the wrong things from data.
First of all:
You subtracted the wrong columns. Obama beat Kerry by nearly 4 million. In 2012 he only be Bush’s 2004 total by about 800 thousand. 800 thousand might still seem pretty good, but we’re a growing nation with a growing voting population. For the last 60 years, the tally for the winner has beaten the tally for the winner of 8 years prior by an average of 5 million votes.
Second, laugh all you want at Romney for losing 800 thousand McCain voters, but they didn’t switch to Obama, they quit voting. Obama lost nearly 7 million of his own voters that same year. Voter participation had dropped by 7.4 million from 2008. Most of those were Democrats.
Good article by Sam Wang Senate Democrats are outperforming expectations. Comments are also worth reading. Wang and Linzer also weigh in a few times in the comments section.
Note the comments on the 2004 election that Wang said was a particularly difficult election to forecast. The polls that year slightly favored Democrats right up until election eve. Yet, certain “fundamentals” had favored the GOP by a slight margin for most of the year and the trend lines didn’t indicate that those “fundamentals” weren’t operative, but when they kicked in with the last polls, projecting all of the Senate races was easy.
It is amazing to see such things playing their role in american presidential elections. Upcoming election will answer many questions which need answers now.
url=Business Loans in Florida[/url]