A just-released Suffolk Poll commissioned by the Herald shows a Walsh surge, with a 46 – 43 lead over John Connolly. This is a surge, based upon an apples-to-apples comparison from an earlier Suffolk poll, which showed a seven-point Connolly lead.
The race is a dead heat within the poll’s margin of error (4.2%).
The Walsh surge is primarily due to increased support by black, Latino, and other votors of color.
On the at-large City Council Races:
…newcomer Michelle Wu is leading the field of eight candidates with 20 percent of the vote, according to the poll of very likely voters. Incumbent Ayanna Pressley is in second place with 16 percent, trailed by former mayoral candidate Michael Flaherty with 14 percent of the vote.
Incumbent City Council President Stephen Murphy is holding down a shaky fourth place with 10 percent, followed by Jeffrey Michael Ross at eight percent, and Jack Kelly and Annisa Essaibi George with seven percent each. The top four vote getters win the at-large council seats.
The full marginals and tables haven’t been released by Suffolk yet. I’ll post links when they become available.
I’m not entirely convinced, but this race does seem tight.
If this poll is right (and who knows if it is), Walsh should be in good shape. I think Connolly needs a safe lead outside the MOE in order to win. Walsh’s voters are more likely to vote and better at GOTV.
But was foiled by software. This poll is significant for a roughly 10 point swing in Walsh’s favor. I’m starting to think Walsh might win. Still we’re talking about very likely voters–Connolly needs to turn out the somewhat less likely.
…hence any analysis is premature.
That said, and for what it’s worth, I made a few calls to people that are good gauges of sentiment within their respective neighborhoods. At present there are indications of a pro-Walsh populist surge. Also, for what it’s worth, Walsh’s field is coalescing…
Not really a synonym for “within the margin or error.”
of a margin of error greater than the Walsh lead in this particular poll, but we can (I hope) agree to disagree on the matter of my word choice.
What it interesting is the indication of a surge, based upon Walsh’s numbers in areas where he was previously vulnerable:
However:
It ain’t over ’til it’s over.
(usually with a 95% confidence internal)–it’s worth noting, but it’s not a tie. Just a measure of uncertainty.
To be just a little technical: Visualize, on the same graph, the two normal curves representing the probabilities for the two candidates. Where they overlap is the margin of error.
Unless the two curves are very close that area is going to be small, representing the small probability of a reversal.
To be a lot less technical: Where would you rather your candidate be, up by 7 or down by 7? If it were a dead heat you would not care.
I care. I care.
The margin of error is not telling you how close the results are to the truth. It is a function of the sample size and only accounts for errors due to random sampling variation.
It tells you nothing about other sources of error including bad assumptions about what constitutes a likely voter, omission of recently registered voters, inclusion of voters who no longer live in the city, failing to reach representative sample of people who cannot be reached by phone, people lying to pollsters, etc.
I also don’t understand your visualization. If both candidates were polling at 50% then wouldn’t their curves overlap 100%? Am I missing something?
it would be a dead heat.
If the curves are identical then the intersection would be 100%. What does that have to do with MOE? What are the curves supposed to be anyway? What are the x and y axes?
And one more time, what does any of this have to do with margin of error? As I said, it is essentially a measure of the poll’s sample size. It has nothing to do with how far apart the two candidates are in the results. It does not tell you how likely one or the other candidate is likely to win.
to be precise.
I should have said that half the area represents the probability that the race could go the other way, given the confidence interval (and based on the sample).
Though I think that’s clear from context.
If the curves don’t overlap by much, it’s not much of a shot and not much of a horse race.
But you still haven’t said what these two curves actually represent. Is it probability of getting x votes? If so, I agree that the percent of the overlapping region would represent the amount of uncertainty you have over who is going to win. However, polling doesn’t actually give you these curves, nor are they actually normal curves in any case both because the possible number of votes is bounded at both ends and more importantly because the two variables are not independent of one another.
The “margin of error” is the estimated difference between whatever result is reported in the measure and the “true” value.
For typical random samples and a 95% confidence level, it is about 0.98/sqrt(n), where n is the sample size.
As the sample size increases, the moe decreases (much more gradually, though).
Suffolk now has the data on its webpage.
The marginals are here.
The tables are here.
I think the variability of turn out makes firm conclusions difficult but a 10 point swing is potent evidence of momentum. I’m keeping tabs on LGBT opinion leaders in the downtown neighborhoods and detect no “populist” trend here. But the LGBT vote is not always in lockstep with other minority communities. This race just got even more interesting–I’ve decided last minute to do GOTV work for Connolly.
There are divisions between and among those communities that makes this race fascinating from a field wonk perspective.
That said, if Marty Walsh manages to close the gender gap in various black communities (which obscures various organizational, inter-ethnic and intra-neighborhood differences), he has it in the bag.
In the same sense, it will be interesting to see how much broad neighborhood culture and economic issues affect choices among LGBT voters. I suspect that Dorchester will trump the South End to Walsh’s benefit through sheer weight of numbers in the absence of LGBT-specific differences between the two candidates (Connolly spin notwithstanding).
Michelle Wu!
shows Connolly ahead by 7 points. I’ve never heard of this polling outfit but in a race this vote it’s hard to get a fix on the electorate using any likely voter screen. I wish we had Nate Silver to sort things out, but I’ll bet he’s still fuming about the Red Sox.
and some outside group that supports him has him up 7? Sorry, not buying it.
on Tuesday by all means share your insights.
Oh come on now. Every independent poll released in the last couple of weeks, AND both candidates’ internals, save only the Globe’s which was carried out by the lousy-track-record UNH Survey Center, either shows a tie or Walsh with a modest lead, and also show that the race is trending to Walsh. Now, suddenly, Dems for “Education Reform” which desperately wants JCon to win has him ahead, and not just ahead but almost outside the margin of error?
Nope, I don’t buy it. Nate Silver would be guffawing at this DFER poll.
Since David Bernstein was shown the full toplines, there is obviously no propriety information in the poll: the pollster should release the full results to the public.
For what it’s worth The current HuffingtonPost/Pollster average includes the Anzalone/Liszt/Grove poll done for the Connolly-allied Democrats For Education Reform cited by Tudor586.
HuffPo/Pollster’s aggregate has Walsh at 44.7% and Connolly at 43.2%.
From today’s Huffington Post/ Pollster:
so let’s do it the old-fashioned way: http://www.bostonmagazine.com/news/blog/2013/11/01/outside-group-poll-connolly-ahead/
Because if you are, it is only about a 3 or 4 step process.
Walsh and Connolly should both follow the leadership of the Red Sox. Here’s my advice to the two candidates. Grow a beard! If one of them grows a beard really fast, I bet that he would win.