The press release that Suffolk University put out yesterday regarding its new poll wasn’t exactly inaccurate. But it was incomplete in significant ways – and it led to erroneous reporting in most of the newspapers and wire services that covered it.
The big omission in the release was that it contained no breakdown of who, exactly, was being polled for the question regarding the Democratic primary for Governor. The only information in the release was that the poll consisted of 600 registered voters who had voted in the 2002 general election, and that the margin of error was +/- 4.0%. And most news outlets just reported it that way without further inquiry – that is, they said that the Suffolk poll said Gabrieli 32, Patrick 24, Reilly 20, based on 600 voters, at +/- 4% (not every news outlet gave both the 600 voters figure and the MOE, but each of the ones I’ve linked gave at least one). That gives an impression of a wide Gabrieli lead that is at the very edge of being outside the margin of error, and that is based on a substantially larger sample than the Survey USA poll, released the same day, which showed Patrick with a 4-point lead and Gabrieli tied with Reilly for second based on 429 “likely Democratic primary voters.”
But that impression, widely reported as it was, was not correct. In fact, as we reported shortly after the press release came out (and before the poll’s crosstabs were made public), Suffolk’s question regarding preference in the Democratic primary was asked of only 369 respondents – those who were either registered Democrats, or who were independents who “leaned” Democratic. It was those 369 voters who gave the widely-reported 32-24-20 breakdown, not the 600. And the actual margin of error for that question was +/- 5.1%, meaning that Gabrieli’s lead over Patrick, while certainly real, was well within the poll’s margin of error. Moreover, of those 369 respondents, only 214 had actually voted in the 2002 primary. I don’t know exactly how one determines who a “likely Democratic primary voter” is, but Suffolk surely didn’t talk to more than 369 of them, and maybe talked only to 214.
All of this information was available in the 300+ pages of details that Suffolk released much later in the day. But by putting out the release beforehand, Suffolk guaranteed that the news reports would be based on the release rather than the actual poll data. And since the release was incomplete in important respects, many of the news reports were inaccurate.
Moral: polling is complex, and the people who write press releases may not be fully versed in its intricacies. So look for the details when these things are reported, and don’t believe everything you read.
wish trad media outlets did this kind of work.
Hey, maybe they will if we keep doggin’ ’em!
<
p>
My latest pet peeve is about the “statistical dead heat” thing — I can feel a couple of letters to the editor coming on…
and that’s primary day. I’ll worry about the rest later.
<
p>
I’m a one foot in front of the other type gal.
I’ll pull the lever for any shmuck with a D after his name to keep out Healey.
<
p>
The primary doesn’t mean much — it’s like having two decisions: do you want to starve or eat a steak, and do you want salt with that?
There are significant differences between the candidates and the style in which they’re likely to govern. Furthermore, they all say they support lots of the same things… but how great their support for those specific things differs widely. For example, I don’t trust for one second that Reilly will strongly defend equality in marriage. In fact, with the massive ballot fraud, he had his chance and blew it.
Ryan, I realize we’re mainly Democrats here, but look at the spectrum in front of us. Not even including the Greens and Libertarians.
<
p>
Spend a day or two on the Mass GOP website if you want to see significant differences.