Did Marty Walsh Over-Perform in the Preliminary (And Will He Do It Again)?

Breaking down the school of thought that the mayoral candidate is better off than the polls suggest.

There is a school of thought among some Boston political insiders that says Marty Walsh is in better shape than polls suggest because his massive, super-dedicated army of volunteers will bring out more voters—just as they did in the preliminary election. So, if Connolly currently leads by some six to 10 points in the polls, it’s really a toss-up for who would win if the election were held tomorrow.

I don’t quite buy it.

To be clear: I think that Walsh’s volunteer operation is very real, incredibly impressive, and vitally important to his success. I believe they played a big role in his early rise and separation from the rest of the pack of candidates.

But did they really give Walsh a huge bonus showing that wasn’t picked up by polls?

Let’s go back and take a look at how the polls did in predicting the actual preliminary results. It’s important to remember that the polls didn’t have the votes adding up to 100 percent, because about a quarter of respondents were undecided. If we take those out from my averaging of the September polls (in effect, assume they break in the same proportions as the decided voters), we get the projected share of the vote, which is the first column below. The actual result of the September 24 election is the second column.


Clearly, Walsh outperformed his poll numbers, and Connolly underperformed—resulting in a substantial six-point net swing, and a ticket-topping for Walsh. If that’s the result of something not picked up by polls—something that we can expect to see repeat in the final election—that would indeed change the nature of the race.

But all of the polls included here were taken before Walsh ran any TV ads. He subsequently ran a $450,000 ad blitz—along with a substantial mail and online strategy, the heightened media attention of the final two weeks, and, yes, hundreds of door-knocking and phone-banking volunteers.

At the time of the polls, Walsh suffered a significant gap in voters forming an opinion of him, compared with Connolly and other top candidates. He made up an awful lot of that ground in those final 10 days or so. I tend to think that the bulk of his “over-performing” can be traced to that. That’s impressive, in terms of campaign strategy and execution, but it’s a fairly normal strong close—that would have been picked up if we had polling during that time.

It’s also clearly true, I think, that Walsh’s core base of support in Dorchester outperformed expectations (both in numbers and percentage for Walsh)—which I would suggest resulted from people who don’t look to pollsters like likely voters, showing up to vote for Walsh because of that eager volunteer army.

Connolly’s under-performance is a little more intriguing. I would suggest a few possible factors. One, I suspect, is the other side of the Dorchester turnout coin: the southwestern corridor of the city, where many of Connolly’s and Conley’s votes came from, came out strong—as expected—but not relative to the whole, given the strong Dorchester turnout. So, to the extent that accounts for Connolly’s drop, you can perhaps chalk that up to Walsh’s turnout machine as well.

But it doesn’t look to me like that accounts for much of Connolly’s under-performance. I think a fair amount came from some combination of his relatively softer support (compared to Walsh’s diehards), and the final-week reporting that Connolly appeared almost assured of a spot in the final. That is to say, a lot of people who were inclined to vote for Connolly might have tossed votes to Richie or Barros, or not even made the effort to make it to the polls at all, because they thought Connolly was safe.

So in sum, what I think is that most of the apparent six-point swing was A) Walsh’s normal campaign persuasion, primarily from his ad blitz; and B) soft performance from Connolly voters who felt their candidate was safe. The outside-the-polling effect of Walsh’s volunteer army may have been worth a net percentage point or maybe two. And, most of it was probably within his core Dorchester area, which does not mean it can be duplicated in a general election, where Walsh will need many times more than the 20,000 or so votes he got in September.

Maybe all of that is correct, maybe not. I don’t have real data to say either way.

But I think my explanation fits better with what the data does show, which is a significant but not huge gain in Walsh support (in favorability, and head-to-head against Connolly) from the early-September pre-prelim polls to the October post-prelim polls.

That’s the sort of gain I would expect to see from the ad blitz and media coverage, and that would pretty much explain most of his apparent over-performance. And there is no sign of any accompanying drop in Connolly support from those early-September polls, which suggest to me that his under-performance was not reflective of any actual decrease in his support.

Now, what all this does suggest is that Walsh has been converting voters, while Connolly has been fairly stagnant. That’s important—but also not entirely unexpected, given how far ahead Connolly was in familiarity and support at the start of the race.

I think that if the election were held tomorrow, Connolly would likely win by a comfortable margin. I also believe that the race is fluid enough that Walsh has plenty of time and opportunity to make up that ground—starting with the first televised debate, tonight.