Politics & Policy

Oversampling Dems

Party ID has shifted, but polls are stuck in ’08 methodology.

How important is it for polling organizations to include the correct proportions of self-identified Republicans and Democrats in their polls? The short answer is: Extremely.

In this hyper-partisan age, the partisan ratio can not only determine the poll’s top-line results, but also shape the ensuing media interpretation of what it all “means.” That interpretation defines the expectations game, which, in turn, affects fundraising, voter enthusiasm, and turnout.

According to a spate of recent polls conducted by Marist/NBC News, Democrats hold strong or even commanding leads in three key U.S. Senate races: Florida (up 4 points), Virginia (up 6), and Ohio (up 14). In three other swing states — Colorado, Nevada, and Iowa — recent Marist/NBC News polls found President Obama and Governor Romney locked in a dead heat.  

Are Republican prospects really that much rosier in these swing states out West and on the Plains than they are in the East and industrial Midwest? The discrepancy, it seems, may be due to the percentage of self-identified Democrats and Republicans in the survey samples.

In Marist’s survey of the Florida Senate race, for example, Democrats in the sample outnumbered Republicans by 8 percentage points (43 to 35 percent). In Ohio, the Democratic advantage was 9 points (37 to 28 percent); in Virginia, the spread was considerably smaller (31 to 29 percent). In contrast, two of the three polls that found the presidential race to be neck and neck sampled significantly more Republicans. Marist actually included more Republicans than Democrats in the Colorado and Iowa polls — a 4-point advantage in Colorado (35 to 31 percent) and a 1-point spread in Iowa (35 to 34 percent). In the Nevada survey, Democrats outnumbered Republicans by only 2 points, as in Virginia (40 to 38 percent).

So what partisan baseline should polling organizations use? After all, baselines can determine the shape and outcome of important Washington policy debates, whether the discussion is about budgets, taxes, or polls. Did the Gang of Six propose the largest tax increase or the largest tax cut in human history? The answer depends on the baseline one uses. Tell me your baseline, and you’ve told me everything I need to know about your analysis.

I have found that the most reliable baseline on party identification, both nationally and at the state level, comes from the extensive polling conducted by the Gallup Organization. Each year Gallup collapses the 350,000-plus interviews it has conducted throughout the year into one overall snapshot of party affiliation. In February, Gallup released the results of 353,492 such interviews conducted during 2011, including more than 1,000 in each of the 50 states and an enormous number in the states surveyed by Marist/NBC News: 18,090 people were interviewed in Florida, 13,172 in Ohio, 9,927 in Virginia, 7,105 in Colorado, 4,439 in Iowa, and 2,730 in Nevada. If volume counts, Gallup’s data mine of interviews is sheer gold.

In reviewing all this data, Gallup identified an important national trend:

In the last four years, the political leanings of Americans have increasingly moved toward the Republican Party after shifting decidedly Democratic between 2005 and 2008. In 2008, Democrats had one of the largest advantages in party affiliation they have had in the last 20 years. . . . Prior to that, the parties were more evenly balanced.

The net result of the movement is that the nation looks to be essentially even in terms of its party loyalties headed into a presidential election year. Clearly, President Obama faces a much less favorable environment as he seeks a second term in office than he did when he was elected president.

The partisan divide, Gallup found, has narrowed not only nationally, but in almost every state as well. Gallup’s 2008 surveys of state-level party identification found that the terrain in all the states Marist surveyed was decidedly pro-Obama and pro–Democratic Party. The Democratic advantage ranged from 9 percentage points (in Virginia and Florida) to 11 points (in Nevada and Colorado) to an overwhelming margin of 18 points (in Ohio and Iowa).

The intervening years, as the table above depicts, have not been kind to the president and his party. Whatever the cause of this shift, the two parties now are essentially at parity in all six of these states.

Here are the results:

Marist, however, uses Democratic to Republican ratios in two of its Senate polls — Florida (8-point Democratic edge) and Ohio (9-point advantage) — that are more in keeping with the state of play in 2008 or 2009 than that of today. Marist’s 2-point Democratic advantage in its Virginia Senate poll is closer to Gallup’s finding of a 1-point GOP lead, but it still bestows more of an advantage on the Democratic candidate than seems warranted.

The share of Republicans in Marist’s Iowa poll is 5 points more favorable than the Gallup finding, so that one Marist poll may overstate things in the GOP’s direction. But in both Colorado and Nevada, the partisan breakdown follows Gallup’s numbers almost exactly.

If all polling organizations were to use Gallup’s party-identification findings in their surveys, some Republicans would be added and some Democrats subtracted from the samples (except occasionally, as in the Iowa poll). The top-line political conclusions would no doubt tilt slightly more in the Republican direction.

The polling results on policy questions, meanwhile, would depict a nation that is more consistently right of center on everything from social issues such as abortion to economic issues relating to the size and scope of government. The journalists who report on these polls and the headlines accompanying their handiwork would highlight different narratives from the ones they do today. Most significant, there would be fewer Election Day “surprises” that prompt the disappointed to mutter darkly about special interests or rigged elections.

– Michael G. Franc is vice president of government studies at the Heritage Foundation.


The Latest