Flawed Poll Models Underestimating Romney’s Lead

October 30, 2012
By
Share

Partisans on both sides have been making claims of a lead in the polls — the Republicans cite the national polls and the Democrats the state. I’m not a partisan or a Romney supporter, but I feel compelled to argue that the poll models are flawed and underestimating Romney’s lead.

As far as modeling in these polls, the example of Marist and Mason-Dixon in Florida is illustrative of skewing. Marist tends to report big Obama leads, and M-D has reported decent Romney results.

From the Tampa Bay Herald: “Mason-Dixon, which has been polling in Florida for 28 years, uses a survey sample based on people’s voter actual registration to match the electorate in Florida, while Marist uses a sample based on whether people say they consider themselves a Republican, Democrat or independent. About 20 percent of the likely voters in the Marist poll were Hispanic, while 13 percent were Hispanic in the Mason-Dixon poll, more in line with the Florida voting patterns.”

Nationally, Pew Research illustrates Obama’s problem, as does a new Washington Post poll. The first Pew chart shows Romney’s strong supporters exploding in October, once he showed up at the debates appearing to be a moderate.  Right now, each candidate has about the same level of strong supporters..

However, the badly flawed state polls don’t show that 5 percent fewer strong Obama supporters are more likely to vote than among Romney’s base.

Obama’s problem is even worse among leaning Democrats. This support, as I have predicted, hasn’t materialized. Only 62% of Democrats and Dem leaners were likely to vote or even registered to vote, and figure that hasn’t budged since September. That’s the real knockout blow for Obama.

Meanwhile, GOP and GOP leaners likely to vote have risen from 69% in September to 76% this month, which is a big 14% spread over Dems. This suggests that Dem leaners could be over counted in these polls relative to Republican leaners.

The Washington Post/ABC poll considers “partisan independents,” for which the gap is 8 percent Republican. Of Republican-leaning independents, 92 percent say they plan to support Romney, while 84 percent of Democratic-leaning independents are backing Obama.

The less partisan the voter, the worse outcome for Obama. When tracked for likely independent voters (including true independents), the Washington Post poll reveals an even more disastrous scenario for Obama. In the Post’s last three polls, Obama trailed Romney among independent voters by a range of 16 to 20 percent. That’s a striking reversal since 2008, when Obama won independent voters (who were 29 percent of the electorate) by 8 points over Sen. John McCain of Arizona. Another poll NPR gives Romney a 50-37 lead among independent voters definitely or probably going to vote.

Finally, Obama has lost considerable support among two groups of likely voters: whites and seniors. In 2008, Obama trailed among white voters by 12% over McCain. This election, it’s 20%. It’s even worse among voters over 65.  Obama trailed seniors 8 percent in 2008 and lags 19 percent today, according to the Pew poll. Obama has overwhelming support among black voters, but turnout is expected to drop to 59 percent this year, compared to the record breaking 65% in the last election. Among more unlikely voters ages 18-29, Obama has lost 13% of his margin since 2008, and can expect a much lower turnout to boot.

Purple Strategies (PS) illustrates some of the skewing issues with the state polls. It gives Ohio to Obama by 2 percent and Colorado to Obama by 1 percent. However, the PS independent-vote tally doesn’t square with the Post/ABC at all. PS labels 32 percent of Colorado voters as “independent” and surveyed a 42-42 split between Obama-Romney. They label 38 percent of Ohio voters as independent and gave those votes to Romney, 42-40. Yet, as mentioned before, the Post/ABC poll has consistently scored independents as favoring Romney, 16-20. PS used a 34(D)-33(R) sample in their registered-voter split, yet there are 837,732 active registered Republicans and 739,778 active registered Democrats in Colorado.

Further, even if the PS assessment is correct or even close, the likelihood of an Obama-leaning independent showing up to vote is much less likely  than a Romney-leaning independent. Among all registered voters, 69 percent of Republican-leaning independents say they are following the election closely, while just 49 percent of Democratic-leaning independents say the same. Among “pure” independents, 41 percent say they are closely following the election.

Democrat-voter registration is down in many key battleground states, and there has been a big increase in Independents — even more so than Republicans. In Ohio, about 7.9 million people are registered to vote in Ohio for the November election.  That’s down from about 8.2 million registered to vote in 2008. In Cuyahoga County alone (a Obama hotbed in 2008) , there are about 80,000 fewer registered voters than there were four years ago.

The Gravis poll in Iowa identifies its sample as 41 percent Democrats, 35 percent Republicans and only 24 percent Independent. The Dems and Reps surveyed said they were voting along party lines, while the Indies favored Romney 48-36, or 12 points. Because of the Dem-skew in the survey’s sample, Gravis gives Iowa to Obama by 4 points. Current Iowa voter registration figures show 35 percent are registered Independent, 33 percent GOP and 32 percent Democrat.

Even a more balanced national poll like IBD/TIPP, which gives Obama a 1.4% lead, shows a heavy 38-31 Dem skew in their sample that’s almost along the lines of the 2008 vote. The reason Romney is close is that the poll scores the Independent vote fairly high at 32 percent, although the 8-percent spread to Romney is lower than with other polls. The poll’s details: Sample size reflects 942 likely voters, who were identified from 1,091 registered voters with a party affiliation of 38 percent Democrat, 31 percent GOP and 32 percent Independent.

In sum, my election prediction is a 3-percent edge for Romney in the popular vote, and 301 to 237 lead in the electoral vote (see second to last chart) — and that’s giving Obama’s battleground-state ground-game strategy considerable credit for getting leaners into the voting booth. I project Pennsylvania will be close, with the outcome dependent on whether the 828,000 people without power get services restored before the election. Michigan may be closer than generally believed. Gasoline should be spiking by Election Day, which could psychologically effect voter outcome. Turn out nationally will be 6 percent below 2008 levels.

A Romney outcome would be perfectly consistent with Europe, where incumbents are booted out and replaced with new incumbents, who quickly become unpopular. Romney will have a huge mess on his hands.

This final chart shows national polls that are more accurately skewed and show a small Romney lead.

 

Leave a Reply

Russ Winter’s Actionable

Click here to access Russ Winter's Actionable service, for subscribers only.

To subscribe to Russ Winter's Actionable and get instant access, click here.


Monthly Archives