Apparently my previous post pricked some delicate sensibilities at VPR’s brand spankin’ new $10,000,000 Palace Of Genteel Broadcasting, because within a few hours this blog had received comments from VPR News Director John Dillon and Director of Digital Services Jonathan Butler, attempting to explain why their Castleton Polling Institute survey didn’t include the question foremost in political junkies’ minds: how are the primaries for governor and lieutenant governor shaping up?
Their explanations were earnest, extensive, and only partly convincing. I’ve still got problems and unanswered questions.
Starting with this. Nowhere in its poll-related online content, as far as I can tell, do they disclose the lack of direct, “who would you vote for?” questions on the key statewide races. Did it not occur to anyone in the P.O.G.B. that listeners might wonder about this singular omission?
Apparently not. Either that, or they were embarrassed about it and were hoping to slip it under the door while nobody was looking.
Well, on to the explanations. Which bore striking similarities, almost as though somebody had a meeting.
Dillon addressed the main issue thusly.
Unfortunately the one question that everyone wants to ask can’t be answered with reliable data, and we’d rather not distribute unreliable data to satisfy the curiosity. As you note, pollster Rich Clark explained on Vermont Edition that incredibly low voter turnout in our state primaries, coupled with a lack of data and exit polls from past primaries, means there is no reliable way to ask “who are you voting for?” without using a prohibitively enormous sample size.
How noble of VPR that it was unwilling to pander to our base desire for an assessment of the campaign. Butler said that VPR inquired about horse-race questions, but Clark advised against them:
Some organizations ignore this advice, we chose to follow it.
I don’t think he intended to come across as smug, but he kinda did.
It still leaves unanswered questions. How “prohibitively enormous” would the sample size have to be, to give us reasonably good answers? Was the decision not to go with a larger sample size based on cost, or was it logistically impossible?
(The July poll questioned 637 Vermonters. The February poll questioned 895. Why a drop of almost one-third in the survey size? Could you have gotten a better read on the “horse race” if you’d interviewed 895?)
Another thing. In the complete absence of polling by any other media entity, isn’t an iffy result better than nothing at all? For the sake of informing the public (or, as Dillon puts it, “satisfying our curiosity”)? Ask the dang question and slap disclaimers all over the results, if that makes you feel better.
Where is the line, to put it another way, between catering to your audience’s needs and pandering? Obviously VPR and I draw it in different places.
In a Twitter exchange with Butler, I’d wondered why the February VPR poll took a reading of the race for governor. If it’s intolerably vague in July, how was it any clearer in February? His answer missed my point.
In the VPR Poll in February 2016 we were able to ask about the races in the March primaries because the turnout in March is over 50%. Rich speculates that Town Meeting and related activities drive up turnout; if we just had primary voting that day with no Town Meeting, turnout might be much lower, compromising polling before the vote.
That’s not what I was asking. I was asking why the February poll included a question asking respondents who they wanted to see elected as governor. That’s slightly different than asking who they’d prefer in a primary, and I suppose Clark would say they were comparing Town Meeting Day apples with November apples, not August oranges.
There’s some truth in that, but it’s awfully Jesuitical.
Butler also got his history a little wrong.
The Poll just released today precedes the Vermont Primary on 8/9. Traditionally, turnout for these primaries is very low (if I’ve got this right it was under 10% of registered voters in 2014, according to VT Sec State http://bit.ly/2auY5Rk). If we poll c. 700 people, and less than 10% are likely to actually vote in the primary, it makes the poll results unreliable for predictive purposes.
(Is it a little surprising that VPR’s chief of Digital Services doesn’t know how to embed a link in a WordPress comment?)
He does have the 2014 figure right, but it’s irrelevant. There were no real contested primaries for the top two offices in 2014. It was Shumlin/Milne and Scott/Corren from the gitgo. A more accurate comparison would be 2010, when the Democrats had a five-way primary battle (the GOP had only Brian Dubie). That campaign attracted a turnout of 23 percent, in spite of the fact that only one of the two major parties had a contested primary. This year, they both do.
Having decided that a direct horse-race survey was beneath them, Clark and VPR settled on indirect measures of support: name recognition and favorability impressions. This is supposed to give us something of a read on the state of the race, I suppose.
But it’s subject to the same limitations, isn’t it? We have no idea if these respondents will be voting on August 9, so how valuable are their impressions of the candidates?
Precisely as valuable and relevant as a horse-race question would have been. It provides a fig leaf for the professionally demure Mr. Clark, but it isn’t any more helpful to VPR listeners.
So, sorry, but I’m not buying it. And I think this episode is one more example of the overly cloistered, insular culture at VPR.
Voter turnout in my town, based on early ballots, is shaping up to be about double the usual primary turnout – much more in line with Town Meeting Day. It wouldn’t have been very hard to find a critical mass of likely voters this year. Probably just more money than they wanted to spend overall.
In one account, Rich Clark referred to getting burned by the poll he took in the 2014 AG race, which showed Sorrell way out in front of Donovan. Despite all his disclaimers, he was criticized for putting out a faulty poll. Once burned, twice shy. I guess I can understand, but it doesn’t mean VPR had to go along with him. And they still haven’t said how big a sample size they would have needed.