+/- 4 percentage points or so. That means Markey’s lead is well outside margin of error (at a 95% confidence level), which is great news. However, the trickiest thing about a primary poll may be how you model likely voters. We’ve got to turn out those “likely to vote” Markey supporters!
Aside: Lynch favorability among women, although below 50%, is still higher than I would have expected. In fact, there’s just a few points difference between men and women voters in this survey, which I’d assume to be within margin of error. Not much of a gender gap.
shows a far smaller gender gap than Lynch’s own poll, leaked to David Bernstein. In that one Lynch was down only 5.6 points, but losing by 14 among women and big time among registered Dems and
This one, scrolling through, shows Markey leading by a lot among older voters but Lynch ahead handily with voters 45 and under. Not this voter! It also shows Lynch far ahead among Latino voters (there might have been 50 or 60 sampled), who were very likely to (mistakenly) believe Lynch has been endorsed by the major environmental orgs and is the better environmental candidate. No idea what’s up with that, but my Puerto Rican wife will not like it.
to the one PPP released in early April. In that one Markey led 49-32, here it’s 50-36. WNEU/MassInc.’s polls have both been more favorable to Lynch, with more undecideds, but that can be attributed to their methodology.
beyond gender, where already the margin of error would be pushing +/- 6 percentage points assuming roughly half the sample size per group. Once you start splitting into three groups and margin of error is 7 or more, hard to see how the slices have too much meaning.
Latinos were 4 percent of the poll, so about 25 people. That’s pretty much useless as a sample. It just struck me that the disparity was huge (Lynch up 30-40 points on every question, while trailing in the other demographics). Makes you think they polled 15 people at a Lynch campaign office.
The current results are (rank candidates 1-5, where 1 is good):
Dan Winslow(R) 2.2
Ed Markey (D) 4.3
Gabriel Gomez (R) 3.1
Mike Sullivan (R) 2.4
Stephen Lynch (D) 3.1
It seems to me that there’s no harm in BMGers wandering over and voting their preferences.
the Democratic Primary? Or is this a poll only of the people they contacted who said they were voting Democratic next Tuesday?
I would have to think it’s the latter, although the way the first question is worded it looks like everyone is voting in the Democratic primary.
Polls in low turnout elections like this one are limited in value unless they have some real good way of determining who the likely voters are.
They ask questions to determine if they’re likely to vote in the Dem primary. First, they ask if they plan to vote in that primary. Sometimes they ask if they’ve voted in various recent primaries and elections. The rest of the poll is supposed to be only those who passed the likely voter screen.
This likely voter screen may not be perfect, but I have a lot more confidence in this poll than the recent WNEU/MassInc. poll. That one polled everyone so they could ask about the general election too. As a result, they had only 270 voters surveyed for the Democratic primary. Too small. And some of them may not really be likely Dem primary voters, because they were asked to choose which one they were voting in from among Dem, GOP, or none. That kind of question, in my view, leads people who would not pass the PPP screen to pick a primary rather than say “none.”
I was thinking of a different poll in my earlier comment. Looks like they just asked “Do you plan to vote in the Democratic primary, the Republican primary, or no primary at all?” This question has the same problems, in my view, that the WNEU question has. It overstates truly likely Dem primary voters.
But it’s not that 100% of the people they contacted said they’d vote in the Dem primary. As you said, it’s just that those were the only people they continued to question.
Commissioned by League of Conservation Voters, who endorsed Markey. Most people knew that, but there was that sliver who said the major environmental groups had endorsed Lynch. As my grandfather would have said, not bloody likely.
This poll doesn’t break down Dems and unenrolleds who say they’re voting in the Dem primary. In other polls, including Lynch’s leaked internal polls, Markey is doing much better among registered Dems than among unenrolled voters. I’d imagine that trend continued here, so Lynch’s numbers among the registered Dems in this poll would probably be worse.
But If I’m Lynch, I might take solace from that low rating. It could suggest that this poll over-sampled Markey-friendly, anti-Lynch people. Lynch’s favorables, even among Dems, have been better in most other polls.
This poll was also in the field from April 23-25. It might be that voters who saw his Swift Boat debate performance or got his robocall are now pissed at him.
that showed him down by less than 6 points. Huffington Post reports that Lynch’s usual pollster says it did not do that poll.
Yesterday I noted that it seemed like a fishy robo-poll. David quite rightly pointed out that some of the pollsters with the best track records, like PPP, do robo-polling. What I meant is that this poll had some elements that suggest dubious methodology. So far Lynch’s campaign hasn’t released details on that.
There were 1,750 respondents or so. For MOE, bigger is usually better. But done right a poll that large would take time and a lot of money. That large a sample, especially for a primary candidate at a financial disadvantage, suggests something’s off. Either it was a one-day poll (disfavored because one day’s news cycle could skew things) with few safeguards (e.g. not much screening, insufficient live confirmation with random respondents), or it went on for over a week. That’s also disfavored because a lot can change in a week. In this case, a long time in the field (BTW, the WNEU poll was 8 days in the field) would mean the poll was running the whole week Massachusetts was focused on events at the marathon.
+/- 4 percentage points or so. That means Markey’s lead is well outside margin of error (at a 95% confidence level), which is great news. However, the trickiest thing about a primary poll may be how you model likely voters. We’ve got to turn out those “likely to vote” Markey supporters!
Aside: Lynch favorability among women, although below 50%, is still higher than I would have expected. In fact, there’s just a few points difference between men and women voters in this survey, which I’d assume to be within margin of error. Not much of a gender gap.
shows a far smaller gender gap than Lynch’s own poll, leaked to David Bernstein. In that one Lynch was down only 5.6 points, but losing by 14 among women and big time among registered Dems and
This one, scrolling through, shows Markey leading by a lot among older voters but Lynch ahead handily with voters 45 and under. Not this voter! It also shows Lynch far ahead among Latino voters (there might have been 50 or 60 sampled), who were very likely to (mistakenly) believe Lynch has been endorsed by the major environmental orgs and is the better environmental candidate. No idea what’s up with that, but my Puerto Rican wife will not like it.
Or is the polling in this race really erratic?
if they are at all reasonably competitive.
to the one PPP released in early April. In that one Markey led 49-32, here it’s 50-36. WNEU/MassInc.’s polls have both been more favorable to Lynch, with more undecideds, but that can be attributed to their methodology.
their “internal polls,” they are desperate. It means their internal polls tell them they are losing.
they want to spin the news, or they want a fund raising push, or lots of things.
Internal polls tend to be biased in favor of the candidate, but the motives for releasing them vary…
on desperate.
beyond gender, where already the margin of error would be pushing +/- 6 percentage points assuming roughly half the sample size per group. Once you start splitting into three groups and margin of error is 7 or more, hard to see how the slices have too much meaning.
Latinos were 4 percent of the poll, so about 25 people. That’s pretty much useless as a sample. It just struck me that the disparity was huge (Lynch up 30-40 points on every question, while trailing in the other demographics). Makes you think they polled 15 people at a Lynch campaign office.
If you’re interested in ranking the candidates
http://www.redmassgroup.com/diary/16836/survey-rank-all-5-masen-candidates-how-will-you-vote
scientific as RMG gets, except when they deny global warming.
The current results are (rank candidates 1-5, where 1 is good):
Dan Winslow(R) 2.2
Ed Markey (D) 4.3
Gabriel Gomez (R) 3.1
Mike Sullivan (R) 2.4
Stephen Lynch (D) 3.1
It seems to me that there’s no harm in BMGers wandering over and voting their preferences.
the Democratic Primary? Or is this a poll only of the people they contacted who said they were voting Democratic next Tuesday?
I would have to think it’s the latter, although the way the first question is worded it looks like everyone is voting in the Democratic primary.
Polls in low turnout elections like this one are limited in value unless they have some real good way of determining who the likely voters are.
They ask questions to determine if they’re likely to vote in the Dem primary. First, they ask if they plan to vote in that primary. Sometimes they ask if they’ve voted in various recent primaries and elections. The rest of the poll is supposed to be only those who passed the likely voter screen.
This likely voter screen may not be perfect, but I have a lot more confidence in this poll than the recent WNEU/MassInc. poll. That one polled everyone so they could ask about the general election too. As a result, they had only 270 voters surveyed for the Democratic primary. Too small. And some of them may not really be likely Dem primary voters, because they were asked to choose which one they were voting in from among Dem, GOP, or none. That kind of question, in my view, leads people who would not pass the PPP screen to pick a primary rather than say “none.”
I was thinking of a different poll in my earlier comment. Looks like they just asked “Do you plan to vote in the Democratic primary, the Republican primary, or no primary at all?” This question has the same problems, in my view, that the WNEU question has. It overstates truly likely Dem primary voters.
But it’s not that 100% of the people they contacted said they’d vote in the Dem primary. As you said, it’s just that those were the only people they continued to question.
Odd poll — I’m guessing it was commissioned by an environmental pressure group measuring who publicized its endorsemenet was?
Commissioned by League of Conservation Voters, who endorsed Markey. Most people knew that, but there was that sliver who said the major environmental groups had endorsed Lynch. As my grandfather would have said, not bloody likely.
For me is that only 50% of Dems have a positive opinion of Lynch. Yikes.
This poll doesn’t break down Dems and unenrolleds who say they’re voting in the Dem primary. In other polls, including Lynch’s leaked internal polls, Markey is doing much better among registered Dems than among unenrolled voters. I’d imagine that trend continued here, so Lynch’s numbers among the registered Dems in this poll would probably be worse.
But If I’m Lynch, I might take solace from that low rating. It could suggest that this poll over-sampled Markey-friendly, anti-Lynch people. Lynch’s favorables, even among Dems, have been better in most other polls.
This poll was also in the field from April 23-25. It might be that voters who saw his Swift Boat debate performance or got his robocall are now pissed at him.
that showed him down by less than 6 points. Huffington Post reports that Lynch’s usual pollster says it did not do that poll.
Yesterday I noted that it seemed like a fishy robo-poll. David quite rightly pointed out that some of the pollsters with the best track records, like PPP, do robo-polling. What I meant is that this poll had some elements that suggest dubious methodology. So far Lynch’s campaign hasn’t released details on that.
There were 1,750 respondents or so. For MOE, bigger is usually better. But done right a poll that large would take time and a lot of money. That large a sample, especially for a primary candidate at a financial disadvantage, suggests something’s off. Either it was a one-day poll (disfavored because one day’s news cycle could skew things) with few safeguards (e.g. not much screening, insufficient live confirmation with random respondents), or it went on for over a week. That’s also disfavored because a lot can change in a week. In this case, a long time in the field (BTW, the WNEU poll was 8 days in the field) would mean the poll was running the whole week Massachusetts was focused on events at the marathon.