A working theory

Following up on this post on Yens Pedersen's rather surprising internal poll results, let's see if we can piece together exactly how those might have come about.

I noted in comments on the previous post that the one robo-call which I received from Pedersen didn't seem to be associated with the poll in question. But in retrospect, that may have been a matter of how the call was organized - in essence starting with what amounted to an audio ad for Pedersen which resulted in my classifying the call as a recorded message rather than a poll, before going into an informal-sounding "who do you support?" question (all delivered by Pedersen personally).

If that's indeed how the call was structured, it could explain the otherwise dubious poll results in a hurry.

To start with, there would be a significant selection bias in who would still be listening to the call by the time the poll came along. Anybody who wasn't interested in listening to Pedersen's personal message (whether due to a dislike for robo-calls generally, or an intention to support another candidate) would have hung up long before there was any hint of a poll, while those still on the line would be those who were most interested in what Pedersen had to say. And that possibility would be consistent with what looks on its face to be a large non-response rate (16,000 calls to 2,000-odd respondents).

And of course it couldn't hurt Pedersen's numbers to have the poll question presented immediately after his own appeal for support, while the other candidates were presented in name only. Which might further explain why the support numbers look so far out of line with expectations.

I'll be looking to confirm with the Pedersen camp that this is roughly how the call was set up. But assuming it's right, let's note what the results do and don't say.

One would be hard-pressed to say that a call which gives precedence to one candidate's message can be taken as an accurate reflection of the relative support for the four candidates involved. So the percentage numbers highlighted by Pedersen don't figure to carry much force.

That said, whatever the methodological problems with the poll as a means of measuring public opinion generally, there still looks to be some value to Pedersen in the raw polling data.

For a candidate whose greatest weakness has been a lack of demonstrated public support, it has to be a plus if 2,000-plus respondents listened through his personal appeal and close to 500 indicated their intention to support him at the end. Even if that number doesn't figure to put Pedersen in contention personally, it surely signals that he may have enough voters in his camp to help swing the results at the June convention.

Moreover, it's entirely possible that a call which seems to have served at least two purposes may have served another as well. After all, to the extent Pedersen is now able to link his own supporters and the undecideds in the poll to the numbers called, he would figure to have a highly useful support list for use for the rest of the campaign.

So while there are some apparent problems with the poll as a survey of public opinion, it may still serve a couple of useful purposes for Pedersen's campaign. And we'll hopefully know before long whether or not that's the case.