Sen. Mary Landrieu (D-La.) is in the reelection race of her life, but two early polls in the past month disagree sharply on how much danger she's in. A closer look reveals why.
The latest was a siren: Landrieu's negative ratings rocketed 30 percentage points since 2012, and her reelection race lead had shrunk to a single percentage point -- 36 percent to 35 percent -- over top GOP challenger Rep. Bill Cassidy (La.). The poll was conducted by Southern Media and Opinion Research (SMOR), and it depicted a far different race than a New York Times Upshot/Kaiser Family Foundation poll last month. In that poll, Landrieu led Cassidy by 24 points (42 percent to 18 percent), with other Republican candidates trailing. If that poll result stuck, Landrieu would stand a reasonable chance of winning an outright majority in the general election and avoiding a runoff election against the leading Republican.*
The polls have numerous methodological differences that may account for the disparity. The SMOR poll sampled likely voters from registration lists (a sample that bodes better for Republicans and may be closer to the electorate), while NYT/Kaiser sampled the broader population of registered voters using Random Digit Dialing techniques (typically better for Democrats). See more on the methodological differences below.
But the polls also differed in what questions they asked respondents before the Senate vote question. The SMOR survey first asked whether respondents had favorable or unfavorable impressions of each Senate candidate, including descriptions of each candidate's residence and occupation. For example, the poll described Cassidy as "a Republican from Baton Rouge who is a medical doctor and U.S. Congressman." The NYT/Kaiser poll preceded election questions with items about the Affordable Care Act and health care -- which could have had their own impact on the results -- but did not provide any description of the candidates before asking the Senate question.
That small biographical cue may have made a difference among an electorate unfamiliar with Cassidy and other Republicans, as barely four in 10 likely voters had any impression of Cassidy in a February likely voter poll. Indeed, the biggest difference between polls was in Cassidy's support, with the Republican garnering 35 percent in the SMOR poll versus just 18 percent in NYT/Kaiser. The introductions did not appear to help candidates Paul Hollis ("state representative and businessman") or Rob Maness ("retired Air Force officer and currently works in the utility industry"), who received single-digit support in each poll.
In a phone interview, SMOR pollster Buster McKenzie said the poll took this approach "to give some kind of balanced description of each candidate." McKenzie also noted Cassidy's occupation as a doctor is not out of the ordinary for state officeholders. "I would expect over the duration of the campaign for that information to be made public."
There's no doubt voters will become more familiar with Cassidy and with the backgrounds of other Republicans as the election comes closer, though that information will come in a frenzied mix of advertising, debates and news coverage. The difficulty of replicating such a dynamic is the chief reason why pre-election pollsters tend to avoid providing biographical descriptions of candidates before asking vote intentions, or even basic questions about the candidates' individual popularity that will not appear on ballots. A review of Rutgers University polling in the 2013 New Jersey governor's race blamed inaccuracy on the survey's placement of candidate favorability and job performance questions before questions about voting.
The Southern Media and Opinion Research poll was conducted April 28 to 30, 2014, among 600 likely Louisiana voters interviewed on conventional (80 percent) and cellular phones (20 percent). Likely voters were defined as registered voters who cast ballots in at least two of the last five elections with statewide races or ballot initiatives, identified through voter registration files matched with telephone numbers. Results have a margin of sampling error of four percentage points.
The New York Times Upshot/Kaiser Family Foundation poll was conducted April 8 to 15 among 1,075 Louisiana adults on landline (49 percent) and cellular (51 percent) phones, including 946 self-identified registered voters. The sample was drawn by randomly dialing landline and cellular telephone exchanges within the state. Results among registered voters have a four-point margin of sampling error.
Full disclosure: The Washington Post regularly partners with the Kaiser Family Foundation on other polling projects but was not involved in the NYT/Kaiser poll.