An interesting paragraph from this week’s “Weekly Roundup” at State House News:
How would Patrick have fared against a Reilly-Gabrieli ticket? We’ll never know. We do know he hit the 50 percent mark, always impressive in a three-way primary – but much, much more than that, did it running as a true outsider, massively outspent, against the sitting attorney general and the party’s choice for lieutenant governor in the last election. And he appeared to make good on his vow to bring people back into political participation: turnout was higher than expected, a development pundits expected to help Gabrieli. Pundits expected wrong. And incidentally, the polls were right, as a look at the record will show they almost always are.
Right on all counts. The point about the polling is particularly interesting. Everyone always says they don’t believe the polls. But the fact is that, especially close to the election, they tend to be pretty good. Let’s just take a quick look at the late polls vs. the actual results:
| Actual | SurveyUSA/CBS4 (9/18) | Suffolk/Ch. 7 (9/18) | UNH/Globe (9/17) | SurveyUSA/CBS4 (9/12) | |
| Patrick | 50 | 46 | 37 | 46 | 45 |
| Gabrieli | 27 | 29 | 29 | 25 | 29 |
| Reilly | 23 | 22 | 21 | 18 | 21 |
Pretty good (except for the weirdly outlying Suffolk poll). In particular, I’d say that Survey USA’s strategy of using robocalls read by professional announcers is pretty well vindicated.
In addition, as has already been noted here, prognosticators (including the state’s chief elections officer) were wildly off-base in their gloomy turnout predictions. Galvin predicted 640,000. We had 915,000, 43% higher than the prediction. Lots of rules being broken in this race. Cool!
One of my former professors was interviewed in an article published on the day of the primary. When I read it, I wanted to slam my head against the table because of her stupidity. She claims to be a State Politics expert, but clearly her expertise didn’t really transfer when she moved to Massachusetts about two years ago – which is fine, but then you don’t go saying things like this:
<
p>
<
p>
It’s one thing to say you don’t trust polls in August, it’s another thing to say you don’t trust polls the week of the primary. Furthermore, when multiple polling companies are saying the same thing it becomes a little foolish to predict things are going to turn out radically different – especially when the guy in first also had the best GOTV movement.
<
p>
For her to suggest results that would be radically different, she should have backed it up! Maybe she felt that Deval’s GOTV operation was overrated? Maybe she thought that Tom Reilly’s inner-city contacts and endorsers would have had a bigger pull? I don’t know – she didn’t tell us.
<
p>
The sad thing is the bleeding didn’t stop:
<
p>
<
p>
That’s all fine and dandy, but that doesn’t change the fact that every polling company said he was down – by a lot. It would be one thing to say these things in August, completely different the day before the primary.
<
p>
For some strange reason, people try to slam me for being an academic. I always find that odd because, from what I’ve found, there are way too many true academics (you know, with actual PHds) who buy too much into conventional wisdom. It just seems lazy, when there’s new and better data available. There are new trends and they’re ignoring them. I’m not trying to attack academia either, especially when it’s just one professor who I happen to like, it just seems to me that sometimes they’re too insulated. Sometimes I think they sacrifice important – yet broad – knowledge, just so they can know some tiny, obscure thing better than anyone else. It’s great that they do that, if that’s what they want to do, but then don’t pretend to be an expert about all things State Politics when it’s pretty clear you just got schooled by a former student the very next day.
<
p>
<
p>
blah, sorry for the rant. It just really bothered me when I first read it and this gave me an opportunity to get it off my chest! LOL If, after this, I get one more crazy commenter on my website talking about how I’m just some random person stuck outside the ‘real world’ I may just have a meltdown – if this reply doesn’t qualify as one already. haha.
<
p>
I think everyone ran across this. Your professor wasn’t the only one. Glad she was wrong. I think quite a few people might be scratching their heads saying “huh?”.
to a “Talk of the Nation” show from Wednesday.
<
p>
Neal Conan (konan?) hosted a discussion with two “pundits” on the house and senate races across the country. One pundit was his usual “political junky” and the other was from “Cook Political Report”
<
p>
The Cook person, in particular, could not have been more favorable to the republicans – even when there was nothing good to say about a candidate’s chances, “He’s a great campaigner” or some other nonsense.
<
p>
Both pundits worked hard to say nice things about the rethugs and the only interesting race they discussed was Lieberman/Lamont.
<
p>
John Tester (MN) hardly came up, nor did Ford (TN) or Webb (VA)
<
p>
<
p>
My point – I think the pundits try really hard not to come across extreme in either direction for fear they won’t be asked back. So your professor friend was hedging her bets, tempering her response to avoid looking like a biased analyst.
<
p>
Face it it’s cool to be quoted by a newspaper – a real ego trip.
<
p>
However, times are different now. In the past, her comments would have been quickly forgotten. Not anymore. The internet has a great memory and we bloggers know how to use that memory…
I thought it might be a little better to normalize out the undecideds from the polls to compare to real numbers, and also looked at the dumb average of the four: the Suffolk and UNH polls seem to average to the same as the 2 SUSA polls, so it looks like a fairly consistent prediction overall. What I think might be most interesting is that despite all the warnings that polls tend to exaggerate support for black candidates, these polls overall underestimate Deval’s support, which is just another defiance of the convential wisdom this election has shown.
<
p>
CandidateActualSurveyUSA/CBS4 (9/18)Suffolk/Ch. 7 (9/18)UNH/Globe (9/17)SurveyUSA/CBS4 (9/12)Dumb Average
Patrick 50%47%43%52%47% 47%
Gabrieli27%30%33%28% 31%30%
Reilly23%23%24%20% 22%22%
I don’t think the polls underestimated Deval’s support, I think they measured it very accurately, and it probably really was in the mid to low 40’s in the week before the election. As I said in a comment on another post, I think he was the Edwards of this campaign (as well as the Dean) – the one with the hopeful positive feel that last minute voters, who make up their mind in the last couple of days or at the voting booth, break disproportionately for. Also, people tend to go for the winner, and by election day he seemed the likely winner, which would also have swayed undecided voters.
<
p>
What I think happened is that he got signficantly more than 50% among last-minute voters, which raised his result above the mid-40s support he had among voters who’d made up their minds in time for the polls.
<
p>
(Do we have exit polls? They might confirm or deny my assertion)
I think it is defintely likely well more than 50% of last minute voters broke for Patrick, and (though it’s become a cliche that usually is over relied on) the sampling of likely voters was less effective due to high turnout.
<
p>
I like that Dean-Edwards comparison đŸ™‚
In the 2004 primaries, Dean was the candidate who consistently got the most early support and committed activists and organizers; Edwards was the one who consistently got the largest share of last-minute voters. That pattern repeated state after state through February. Deval Patrick got both of those niches this week’s primary, and I think for similar reasons that Dean and Edwards had those niches in 2003/2004.
You can’t really vote “undecided” at the polls*, so while the polls rarely sum to 100, the election results do.
<
p>
Therefore, to fairly compare the two, it might make sense to divide by the summ of the poll results. This assumes that the undecideds will break in the same proportions as the decideds in the poll.
<
p>
SurveyUSA(9/18), 46+29+22 = 97
Patrick: 46/.97 = 47.4
Gabrieli: 29/.97 = 29.9
Reilly: 22/.97 = 22.6
<
p>
Suffolk(9/18), 37+29+21 = 87
Patrick: 37/.87 = 42.5
Gabrieli: 29/.87 = 33.3
Reilly: 21/.87 = 24.1
<
p>
UNH/Globe(9/17), 46+25+18 = 89
Patrick: 46/.89 = 51.7
Gabrieli: 25/.89 = 28.1
Reilly: 18/.89 = 20.2
<
p>
SurveyUSA(9/12), 45+29+21 = 95
Patrick: 45/.95 = 47.4
Gabrieli: 29/.95 = 30.5
Reilly: 21/.95 = 22.1
<
p> * Sure sure you can leave it blank. But, for competitive top-ticket races, people don’t. They may be undecided for the robocall, but not for the scantron machine.
especially, because the numbers of votes we’ve been looking at are only the numbers that were supporting one of the 3, the other and blanks aren’t included yet…
I don’t know if its because of the cookie issue with the changeover, but I didn’t see any comments when I made my post a full hour after yours.
<
p>
But, its nice to see our numbers work out!
But that wasn’t going to happen at 1 am! đŸ˜‰
Too. Much. Math.
<
p>
Make head explody.
<
p>
You number nerds are funny. đŸ™‚
My attempts to read into the LG numbers were wildly off, so I won’t attempt to analyze anything here. Then again, those polls were dealing with huge numbers of undecideds, so they were hard to work with anyway.
<
p>
What drove his surge over last couple weeks, from 35%-ish to 50%-ish, picking up almost every undecided….
<
p>
*DP ads finally on air?
*Gabs looking over DP shoulder during debate?
*Reilly stabilizing in second debate, enough to hold some voters that Gabs was vying for?
*BMG endorsementt?
<
p>
2. There were some pre-election cautionary murmurs on BMG that “racist voters” would tell pollsters they were supporting Patrick, and then vote for a white guy in the privacy of the election booth.
<
p>
Since the results ran in the other direction – Patrick outperformed the polls, not underperformed the polls like NC’s Harvey Gantt in 1990 versus Jesse Helms – can we make any positive pronouncements about the racial attitudes of the D’s and I’s voting in MA primary?
I think it was three factors.
<
p>
First and foremost, I think it was the GOTV effort by Patrick supporters.
<
p>
Secondly, Reilly was pretty obviously going down. Undecideds don’t tend to break for the clear loser. So they were going to break for Gabrieli or Patrick. Personally, I was expecting them to break for Gabrieli based mostly on name recognition — he’s been airing TV ads much much longer.
<
p>
However, I think Gabrieli got hurt by my third factor, which is Healey’s attack ad on him specifically. Since Gabrieli and Patrick were much closer than either was to Reilly, anything hurting Gabrieli would probably end up helping Patrick (also, see point two above).