Articles and Analysis


The NY Times and "Push Polling"

Topics: Doug Schoen , Michael Bloomberg , Push Polls

I've lost count of how many times we've seen it happen. A campaign testing negative messages on an internal poll gets accused of conducting a "push poll." More often than not, the opposing campaign, bristling with outrage, reaches out to a reporter who plays along. Soon we have a story quoting in which the opponents spokesperson characterizes negative message testing as "one of the most discredited and dishonorable forms of negative campaigning." Never mind that virtually all campaigns test negatives messages. Never mind that the differences between message testing and so-called push "polls" are easily discovered via Google (say here, here or here). It happens over and over.

What's unusual about today's example is that it appears on the front page of the New York Times. It was based on calls received by "several people" in New York City who picked up the phone, were told that a survey was being conducted, but "were soon asked a series of questions featuring negative information about [Representative Anthony] Weiner," one of the potential opponents of Mayor Michael Bloomberg.

The story confirms that the poll was conducted by Bloomberg. It also provides the following description of the negative information included in the poll:

The issues highlighted in the telephone calls closely echoed negative stories that have appeared in city newspapers about Mr. Weiner, and that the congressman's aides have accused opposition researchers from the Bloomberg campaign of planting.


The questions, [a respondent] recalled, accused Mr. Weiner of taking contributions from lobbyists and of securing federal funding for an organization whose members had contributed to his campaign.

Shocking. Negative campaigning may be breaking out. In New York City.

Do those facts add up to a push poll? No. Politico's Ben Smith saves me the trouble of block-quoting myself:

The story suggests the poll is a "push poll," and doesn't raise the alternative -- that it was the kind of message-testing that, those who followed the presidential campaign will recall, was often confused with push polling.

The difference is explained at length here -- "push poll" is a pollster term of art, with a technical definition -- but the main distinction is that a push poll is designed to push a sharp, memorable, negative message, typically right before an election; a message testing poll is designed to test an arsenal of attacks to see which works best, for later mail, television, or other public attacks, not to persuade the call recipient of anything.

A so-called "push poll" is not a poll at all. It's a telemarketing call made under the guise of a survey. So if the pollster makes only a few hundred calls, asks 15 to 20 minutes worth of questions, includes demographic items, those are usually clues that the intent was to conduct legitimate research for the campaign. Of course, as I've written many times, the pollster isn't off the hook just because they are testing messages. They still have an ethical obligation to tell the truth and not abuse the respondent. Unfortunately, the Times makes no effort to examine the veracity of the attacks.

If on the other hand, if the Times had evidence that the calls were very short, included just a few questions and then the negatives attacks, and were being made to many thousands of New Yorkers, there might be more to this story.

Such an effort would be important given the sort of calling that Bloomberg's pollsters, Doug Schoen and Mike Berland conducted for the Mayor's first campaign in 2001, as described in Schoen's autobiography, The Power of the Vote (pp. 281-282):

Up until this point, our work had been highly sophisticated, but not unprecedented. However, our next step proved even more radical [...] The effort that was taking shape under Mike Berland and our associate, Bradley Honan, was something new. We were now proposing to move beyond sampling and extrapolating to something new -- a census of the city, a database with information on every voter, not a sample.

"This election is going to involve over a million voters," I said. "We don't have to sample, we can do it all." By purchasing consumer information and using phone banks, we could build a profile of every swing voter that combined demographic, voting history and consumer data. We could tag every voter as a member of one of the groups and develop a dialog with them that emphasized the issues they cared about most. Every piece of mail, every phone call, would be targeted precisely on them.

Bloomberg's answer was simple and direct. "Let's do it," he said.

The theory was simple and elegant but assembling the actual database was an enormously complex endeavor: after merging voting demographic and attitudinal information, it was not uncommon to have as many as two hundred and fifty variables per voter.

So they used "phone banks" eight years ago an attempt to collect "attitudinal information" on every voter in New York City. Did these voters know the information was being collected so that advertising could be "targeted precisely on them," or did they think they were taking part in a confidential survey? If the 2001 calls were billed to respondents as the latter, then it would have amounted to what market researchers call "sugging" - or "selling under the guise of research." Not push polling, to be sure, but not entirely ethical either.

Now it would be huge leap to assume that Bloomberg's campaign is once again conducting "census" of voters, and an even bigger leap to assume that negative message testing is a part of it. However, Smith reports that Schoen is once again polling for Bloomberg. If every voter in New York City is hearing negative arguments about Bloomberg's opponents under the guise of a poll, then the "push poll" label might fit. Also, as we have learned several times in recent years, mass political messaging campaigns are now commonly conducted under the false guise of less expensive, interactive-voice-response (IVR) surveys. Those efforts, while not a "poll" by any conventional standard, do collect the "data" provided by those who stay on the line to answer questions. So questions about the kinds of questions that Bloomberg's poll asked, and the length and mode of the survey are certainly worth asking.

However, all we know for certain in this case is that Bloomberg's pollster asked questions in a survey about negative charges that have appeared in newspaper stories. Those facts don't add up to a "push poll."



Interesting post. It would have been nice if you had also quoted exactly where New York Times says that this "message testing" is a "push poll".

I read the entire article, and didnt see the NYT take any position on whether this poll definitely comes under push polling. Simply because Ben Smith of the infamous POLITICO (the capitalization is how they want to be referred to as) says that the story "suggests" this, does not mean it actually does.



Responding to Jon- The issue isn't that NYT accused Bloomberg of push polling, but rather that they wrote a piece without at all acknowledging the difference (in substance and purpose) between message testing and push polling. The end of the article quotes a person who received the call accusing Bloomberg as trying to spread negative messages- by not addressing this statement with any context of or at least acknowledgment that they don't know whether it is a push poll, they leave their reader ignorant.


Post a comment

Please be patient while your comment posts - sometimes it takes a minute or two. To check your comment, please wait 60 seconds and click your browser's refresh button. Note that comments with three or more hyperlinks will be held for approval.