Monday, November 5, 2012

Flawed polling hides Romney landslide

On Nov. 4, The Columbus Dispatch published the results of a poll that seemed to indicate that Ohio voters had swung, perhaps decisively, towards Barack Obama in the presidential campaign. The poll indicated that likely voters in Ohio favor Obama over Mitt Romney by a 50-48 margin.

When the details of the poll are examined, however, the results don’t look nearly as positive for the president. The final paragraph of the article notes that the sample has a 2.2 percentage point margin of error. This means the race is still statistically tied, but another detail should be even more disturbing for Democrats.

The big problem is that the sample is described as being 40 percent Democrat, 36 percent Republican and 21 percent independent. To put those numbers in perspective, CNN Ohio exit polls from 2008 show that in that Democratic landslide year 39 percent of Ohioans considered themselves Democrats, 31 percent Republicans, and 30 percent independents. It is extremely unlikely that the number of Democrats has increased by a percentage point over the 2008 election.

Last week, Gallup released a survey of the 2012 electorate that indicated that the demographics of voters this year resemble 2008 in almost every respect except one: party affiliation. In 2008, 39 percent of Americans considered themselves Democrats and 54 percent leaned Democratic. Twenty-nine percent were Republicans and 42 percent leaned Republican. This is almost identical to the Ohio exit poll data as well as the results of the national popular vote.

In 2012, 35 percent identify as Democrats (46 percent lean Democratic) versus 36 percent who identify as Republican (49 percent lean Republican). That is an 11 point swing in the party identification toward the Republicans over four years. When leaners are included, the swing is 15 percent toward the Republicans. It is extremely unlikely that the percentage of Democrats in a swing state like Ohio would increase while Democrats became less common nationally.

The flawed polling sample might occur for a number of reasons. It is possible that Democrats were more likely to answer the pollster’s calls. Democratic households might be easier to reach for the survey. The pollster may have tried to duplicate the mix of the 2008 electorate on the assumption that it had not changed. It is even possible that the pollster was sympathetic to the Democrats.

In reality, this year’s blend of parties in the electorate will probably more closely resemble that of 2010 than 2008. According to New York Times exit poll data from 2010, 36 percent of voters identified as Democrats and 36 percent as Republicans. Twenty-seven percent were independent.

That year Ohioans elected Republicans as governor, senator, attorney general, secretary of state and treasurer. They also elected Republicans to 13 of 18 seats in Ohio’s delegation to the House of Representatives. Josh Mandel, who received 54 percent of the vote in the election for state treasurer according to the N.Y. Times, is running for Ohio’s other senate seat this year.

When an accurate picture of the electorate is considered, the Republicans should win in a landslide in Ohio on Tuesday. In 2008, President Obama won Ohio by five points with an electorate made up of 39 percent Democrats. In 2012, a 40 percent Democratic sample only puts him in the margin of error. If the sample is adjusted to the true mix of parties by reducing the Democratic sample by five percentage points, Mitt Romney should win the state handily. Josh Mandel, who trailed Sen. Sherrod Brown by 45 to 51 percent in the Dispatch poll, may win as well.

Not all polls publish party affiliation data for their respondents, but a Michigan poll by FMWB-Fox 2 News from Nov. 2 reflects a similar imbalance. This poll reports 43 percent Democratic and 35 percent Republican respondents. In 2008, a mix of 41 percent Democrats and 29 percent Republicans voted for Obama by a 57-41 percent margin according to CNN. This year’s poll favors Mitt Romney by less than a percentage point, however. This means that Mitt Romney will probably win Michigan.

A recent Public Policy poll of Wisconsin and Pennsylvania may also overstate Democrats. The Wisconsin poll showed 35 percent Democrats and 32 percent Republicans yet the GOP has repeatedly won important electoral battles there since 2010. The Pennsylvania poll split between Democrats and Republicans by a staggering 48-38 percent. Yet the results of the presidential question found Obama leading by only three points in Wisconsin and six points in Pennsylvania. When the swing in party affiliation is considered, both states might well fall into the Romney column.

2008 was a banner year for Democrats while 2010 was a Republican landslide in most states. Many of this year’s polls show Barack Obama losing or in a dead heat with Mitt Romney while using a partisan mix based on 2008 even though the 2012 electorate is much more like 2010. This means that when the ultimate poll is taken this Tuesday, the results should be a Republican landslide that was expected by very few.

Originally published on

1 comment:

Anonymous said...

Flawed writing and flawed logic hides Thornton lunacy