The Washington PostDemocracy Dies in Darkness

No, it’s not a push poll

Republicans are crying foul over a Public Policy Polling survey that finds Senate Minority Leader Mitch McConnell (R-Ky.) and Kentucky Secretary of State Alison Lundergan Grimes tied in a hypothetical 2014 Senate matchup.

Two things are true: 1. McConnell is probably not as vulnerable as this poll might suggest. 2. it's not a push poll, as some Republicans have alleged.

The survey was conducted on behalf of Senate Majority PAC, a super PAC that supports Democrats, by PPP, a Democratic automated firm. Republicans complain that the questions were biased towards Democrats. National Republican Senatorial Committee spokesman Brad Dayspring labeled it a "push poll," linking approvingly to a conservative blog breakdown of the questions. He and other Republicans are using the hashtag #pppquestions to tweet out accusations disguised as queries.

Push polls are phone calls that are disguised as surveys but are really meant to affect the outcome of a race by disseminating negative information (true or untrue) about a candidate.

Many legitimate polls -- including some of those done by candidate's own campaigns -- test negative messages within the context of divining the proper strategy and messaging to adopt in a race. This poll contained several criticisms of McConnell; you can read all the questions here.

The American Association for Public Opinion Research explains the distinction between the two kinds of polls on its website.

“Negative or disturbing information about a candidate does not automatically make a survey a push poll,” AAPOR President Nancy Mathiowetz said in a 2007 statement.  “Message testing, when campaigns test the effectiveness of possible messages about opponents and even themselves, is very different; and it is a legitimate form of surveying.”

Push polls go out to thousands of people and generally involve only negative questions, because the callers aren't interested in the responses, just in getting out the attack. Message tests typically go to between 500 and 1000 people -- a standard and statistically sound group of people. In the PPP survey, 556 Kentucky voters were polled.

"No sane person would think he could affect a Senate race by calling 500 people -- certainly not in a state with over 4 million people," said political analyst Stu Rothenberg, who has written repeatedly about the difference between message testing and push polling.

While 7,405 people were called to get the 556 who completed the survey, that still means only .25 percent of the state's registered voters heard even part of the poll.

Not all Republicans agreed that the survey was a push poll.

"Polls can and should include negative messages that you might want to use in a campaign," tweeted Logan Dobson, a staffer at a GOP polling firm. "It's good to test how effective they are." However, he argues that the head-to-head numbers may be skewed by the negative questions, even if they come after the straight ballot test (as they did here), because supporters of a candidate are more likely to hang up after they hear an attack line. PPP pollster Tom Jensen says the firm weights its sample to account for that possibility.

Pointing to a dictionary definition of a push poll as "a seemingly unbiased telephone survey that is actually conducted by supporters of a particular candidate and disseminates negative information about an opponent," Dayspring argued that the PPP survey "meets only 95 percent of the criteria."

He added, "the lack of a Democratic candidate reveals more about the state of the play in Kentucky than any poll asking obviously loaded questions."

His point: Lundergan Grimes hasn't actually gotten in the race. No Democrats have. A more salient attack on this poll might be that it's designed more as a tool to convince her to run than anything else. While non-partisan polls also suggest McConnell's numbers are weak, Democrats are struggling to field a candidate in the increasingly conservative state and the senator is well-prepared for a challenge.