With the special Senate election behind us it’s a good time to recap. I’m not getting into messaging or strategy here: this is a pure numbers post: where people turned out and where they didn’t. This builds on Brent’s post, and one of the first things I noticed was what he points out: that Markey did very well, outperforming Elizabeth Warren by quite a bit, in many upscale MetroWest suburbs (82% vs 74% in Brookline, 75% vs 65% in Lexington, 75% vs. 67% in Newton, 59% vs 50% in Wellesley, 63% vs. 56% in Wayland, 67% vs. 58% in Acton, 55% vs. 48% in Boxboro). It is important to note that a number of these towns are not in his House district. Brent offers the hypothesis that Markey’s message (anti-guns and pro-choice) resonated more here than Warren’s economic populism. That’s entirely possible, particularly in a town like Wellesley. Other possibilities I might suggest are that voters in these towns were less impressed with Gomez than with Scott Brown, or that it had something to do with the composition of the electorate in this, our lowest-turnout Senate election in decades, compared to November 2012, our highest-turnout Senate election ever.
About that low turnout: When I posted on May 30, warning about the dangers of low turnout for Democrats, I was right and I was wrong. I was right that Democrats suffer when urban turnout is low (more on that another day soon). I was very, very wrong that Markey couldn’t win a low-turnout election handily. Turnout for this one was far lower than I ever imagined. In recent years, there has been a remarkable consistency in the number of votes cast in Presidential vs. non-Presidential elections: about 3.1 million votes for a Presidential contest, about 2.2 million for a non-Presidential statewide contest. Remember this table?
Go back a bit more and the numbers remain consistent with this trend:
- 2006 Senate: 2.17 million
- 2002 Senate: 2.22 million
- 1994 Senate: 2.18 million
- 1990 Senate: 2.42 million
- 1982 Senate: 2.05 million
- 2004 President (Kerry): 2.91 million
- 2000 Senate:2.73 million
- 1996 Senate: 2.56 million
- 1998 Senate: 2.60 million
- 1984 Senate: 2.53 million
Well, compared to everything we’ve seen in the past three decades the June 25, 2013 special broke through the floor, accelerated through the cellar and lodged in a sub-basement of its own making: Only 1.17 million votes were cast, about as many as Scott Brown alone received in the January 2010 special. The turnout rate was about half that of 2010’s special election, and barely a third of November 2012’s record showing.
Reasons aplenty have been offered: strange time of year, oppressive heat, election fatigue, boring candidates, lack of press coverage, obsession with the Bruins and the Bulger trial, not to mention the unfolding Aaron Hernandez saga. I won’t belabor them here. Instead, I’d like to look at the good, the bad, and the ugly. Since we’re looking backward, I’ll take them in reverse order.
The statewide turnout. 1.17 million votes. 27%. However you look at it, it’s not good. It’s a sad thing for American democracy, even if it’s not in this instance a bad outcome for the Democracy, capital-D.
A few days before the election, Bill Galvin predicted “record-low” turnout of 1.6 million. At the time I thought he was being a Debbie Downer, because he wins no prizes for predicting turnout correctly and should not say such things, but also because I believed his prediction was unduly pessimistic.When an updated report came out saying absentee ballot requests were lagging behind January 2010 far more than previously reported, I thought he might be on to something. Now that the real numbers are in, I see I was totally wrong about his prediction being too low. If anything, Galvin looks like the Pollyanna of voting prognosticators. I therefore half-apologize. (The part where he should get of the public prediction business and do nothing but encourage people to vote still stands).
No more blowouts? Instead, a continuing trend of relative Republican strength. Ted Kennedy and John Kerry (admittedly facing minor league opposition most of the time) routinely used to be re-elected with 2/3 of the vote. Kennedy’s big scare came from Mitt Romney in 1994, a big GOP year, and he still won 59-41. Kerry’s big scare came two years later, in 1996, from Bill Weld, and he rode the spectre of Jesse Helms to a 52-45 win over the sitting governor, the strongest GOP candidate for Senate in Massachusetts (aside from Scott Brown, who did, after all, win) since the 1970s. Otherwise, their races were a cakewalk. Evidence: this map from Kerry’s (65-31) 2008 win over Jeff Beatty (2008 was, I admit, a great year for Democrats, but this was far from an unusual result for Kerry).
Those days appear to be gone. We’ve now had three post-Kennedy/Kerry elections for U.S. Senate in Massachusetts, and the Republican scores have been 52% (Brown in 2010), 46% (Brown in 2012) and 45% (Gomez in 2013). The good news here is that we Democrats have won two of those three elections by decent margins and now control both seats. The bad news is that a pattern is emerging whereby the GOP floor in a statewide election might be 45% and the map looks very consistent, with huge swaths of Massachusetts where Democrats are not particularly competitive in a statewide race. The South Shore is a sea of red, the North Shore only marginally better.
See how similar these maps, from the 2010, 2012, and 2013 Senate races plus the 2010 Governor race, look? (All maps from boston.com)
Incumbency might change things for Warren and Markey, but it seems like we’ve entered a world where Democrats should win here, but also should count on a win no more than about 10 points.
Losing Blue-Collar Voters. It really does not seem to matter if the GOP candidate is a novice with no substance whatsoever ike Gabriel Gomez. The Democratic message is not connecting with a large number of Massachusetts voters, particularly white blue collar voters. Like Coakley and Warren before him, he did poorly in many blue-collar towns (losing in Woburn, which he’s represented for ages; Saugus, which borders his hometown of Malden; Rockland, where he did worse than Warren; losing 60-40 in Billerica and Wilmington). Even though, as Brent points out, Warren did better in blue-collar towns than Markey did, she still didn’t do all that well: Scott Brown got 61% in Wilmington, 55% in Woburn, 57% in Saugus, 60% in Billerica, 54% in Weymouth.
This trend may have started over social issues several decades ago, but it seems clear to me right now (from the numbers and personal experience) that many white blue-collar voters in Massachusetts have an antipathy to Democrats on every issue across the board. Perhaps thanks to the mischief of the Herald, Eagle-Tribune, and Sun. Perhaps because they see Democrats getting 3/4 of the vote in Newton and Lexington.
Hampden County. This area has been the exception to the solid blue wall west of the Quabbin, and it’s starting to get to be a trend. Markey got killed in town after town there, in many cases far worse than Warren before him, as Brent’s map illustrates (most of Hampden County a steel blue, showing a big Markey dropoff from Warren’s scores). He took 66% in Springfield, 61% in Holyoke, and lost every other town in the county. As a result, he lost the county overall by 2,358 votes (48-52). Senator Warren, benefiting from higher turnout in those two cities and her higher percentage elsewhere (44% vs 39% for Markey in the rest of the county), took Hampden 55-45 in 2012.
Obviously Massachusetts Democrats can win statewide while getting killed in much of Hampden County, but it might be worth looking into the reasons why we’re doing so badly there (Kerry, in 2008, didn’t run as far behind his statewide totals in these towns.)
Well, of course, we won. And we won by ten points. So that’s the most important good news.
Had you told me a month ago that only 1.17 million votes would be cast in the election, I’d have said Markey was toast. As June wore on and I gave it more thought, it occurred to me that January 2010 was unique because GOP voters were super-energized and many Democrats were not. I started to think that, in a pure base election with each side equally (un)motivated, the numbers had to favor Democrats. It seems I was right. Democrats should take significant comfort in the fact that in a high-wattage, high-turnout 2012 election we were able to defeat a very popular incumbent by almost 8 points with a first-time candidate, and then we came back and won by a slightly larger margin, running a 37-year Member of Congress against a first-time candidate, in the lowest turnout Senate election in Massachusetts since practically forever.
How’d we do it? Ground game. It sounds counterintuitive to laud the Democratic ground game when turnout was so abysmal (and it was significantly lower than November 2012 and January 2010 tunout in all of the Commonwealth’s 351 cities and towns), but the drop in votes cast generally was much lower in Markey-friendly towns than in Gomez-friendly towns. Let’s visit some crude statistical analysis to see how this worked.
In my May 30 post, I focused on the importance of urban turnout to Democrats. So how we’d do? Short answer: Compared to 2010, the urban turnout was pretty good (not that it was higher than in 2010 — it wasn’t — but it fell off less than it did elsewhere). Compared to 2012, it was pretty lousy. More on that coming in another post soon; for now let’s look statewide.
I divided the 351 cities and towns into three groups, depending on how the number of votes cast last Tuesday compared to the number cast in the January 2010 special election. Statewide, the number of votes cast on June 25 was only 52.11% of the number cast on January 19, 2010, and I used that figure as my point of departure.
- Group 1 consists of 87 towns where turnout dropped by a smaller amount relative to January 2010. These are the towns where the number of the votes cast was 57.11%, or higher, than the number cast in January 2010. That is, these towns had June 25 turnout (I am using votes cast as a proxy for turnout percentage in the absence of good numbers re: registered voters per town), relative to the Brown-Coakley election, at least 5 percentage points higher than statewide average.
- Group 2 consistes of 172 towns in the middle, where the number of votes cast was within 5 percentage points (in either direction) of the statewide average relative to January 2010. In other words, about the same dropoff as the state had as a whole.
- Group 3 consistes of 92 towns where turnout dropped by a larger amount relative to January 2010. In these towns the number of votes cast was 47.11%, or lower, than the number cast in January 2010.
Here’s what I found:
Group 1: In Group 1, where we kept the bottom from totally falling out on turnout, Markey won 68 of 87 towns, and many of them by a lot. He took 11 of the 87 towns with at least 80%, 38 of the 87 with at least 70%, 52 of the 87 with at least 60%, and 62 of the 87 with at least 53%. All told, Markey got 70% in the 87 towns in Group 1, up from 63% for Martha Coakley and 67% for Elizabeth Warren. In those towns there were 349,550 votes cast, representing just under 30% of the statewide total. The same towns represented exactly 25% of the statewide total in January 2010, so their relative importance rose by nearly 20% (from 25.01% to 29.83% of the statewide vote total) compared to 2010; in November those towns were 26% of the statewide total. So he did better in those towns, and those towns counted for more, which is a winning formula.
Importantly, Markey cleaned up in the biggest towns where turnout remained (relatively) high. He won 19 of the top 20 municipalities, in terms of votes cast, among the 87 in Group 1 (North Attleboro being the exception, Boston being the most important). The margins were impressive: Markey won 15 of the 19 with more than 60%, 12 of the 19 with more than 70% of the vote, and all of the top ten with at least 75% of the vote except Falmouth, where his victory was narrow. He also outperformed Elizabeth Warren, in terms of percentage, in all of them but Northampton (79.5% vs. 81.6% for Warren). As a result, within those 20 towns (which themselves represented almost a quarter of the electorate), Markey won 74% with a cushion of 71,000 votes. Pretty nice.
Even better: Markey’s top 11 towns (by percentage of the vote) were in Group 1, and 31 of his top 34 towns. In all of those towns Markey got 73% or higher. In many of these super-Markey towns the drop in turnout was far, far lower than state average (e.g. Northampton down only 10% from January 2010, and Cambridge and Somerville down 35%, compared to a 47.89% statewide drop). That’s what I call identifying where your voters live and getting them out as much as possible in a cycle marked by apathy.
Group 2: These 172 towns, where turnout compared to January 2010 was about as much lower as it was statewide, were pretty evenly split. Interestingly, the electorate’s composition in Group 2 during this super-low-turnout affair was more like 2012’s than 2010’s:
- In 2012, when they represented 49.1% of the votes, these towns gave Warren a narrow 50.2% win.
- In 2013, they represented 49.6% of the vote and Markey won 50.3% of the votes in those towns.
- In 2010, by contrast, they represented an almost identical 49.5% of the vote, but Martha Coakley got only 43.7%.
This suggests that some people in those towns switched from Brown in 2010 to Warren in 2012 and Markey in 2013, or (my guess) turnout was disproportionately high among liberals in 2012 and disproportionately low among conservatives in 2013. Either way, it balanced out and Markey matched, or narrowly surpassed, Warren’s performance in this cluster of towns.
As is clear from Markey’s winning more votes than Gomez in Group 2 despite winning far fewer individual towns, Markey tended to win the larger towns in this group (e.g. Worcester, Quincy, Framingham, Lowell, Lynn, Brockton) and win by larger margins. He also won most of the towns at the top of the group, where turnout was at 52-57% of 2010 levels, and fewer in the bottom half, where turnout was at 47-52% of 2010 levels. Thus the trend of Markey doing well in the the towns that had less dropoff in turnout continues.
The graph below shows the trend. Each blue mark represents one of Massachusetts’s 351 towns. Markey won the towns above the thick black line, Gomez those below it. The red vertical line represents the statewide average of 52.11% of the number of votes cast in January 2010. The orange lines to either side of it represent 5 percentage points away from that mean in each direction. Thus Group 1 towns are to the right of the right-hand orange line, Group 3 towns to the left of the left-hand orange line, and Group 2 towns between the two orange lines. As you can see, there are many towns in the bottom left corner of the cluster: strong Gomez towns were turnout, relative to the 2010 special, was far lower than state average. The line of best fit, heading from there to the top right, shows that Markey did very well in places where turnout (relative to January 2010) remained higher. It doesn’t hurt that the many of the marks at the top right are high-vote towns like, most importantly, Boston (ID’ed in black). In general, the top right was (1) higher Markey percentage; (2) higher population; (3) less dropoff in turnout from January 2010 relative to the rest of the state. Again, a good formula.
Group 3: The group containing the final 92 towns, where turnout was even farther below 2010 levels, was much more Gomez-friendly: he took 78 of the 92 for an overall score of almost 56% in these towns, with Markey getting 43.7%. But they counted for less than in the past. In 2010, these towns represented 25.3% of the electorate. In 2012, they represented 24.8%. This time they only represented 20.7%. It thus mattered less that Markey underperformed Warren here (43.7% vs. 45.9%; Coakley received only 38.2%) because the relative importance of these towns to the statewide race fell significantly. Compared to the 2010 special, the influence of conservatives within these towns fell sharply as well.
To be sure, there were important Democratic cities in this group (New Bedford and Fall River come to mind, each at under a quarter of 2012 turnout and under 40 percent of 2010 turnout). But most of Markey’s towns in Group 3 were small Western Mass. towns whose individual impact on a statewide election is small. Other towns where the number of votes cast was under 40% of those cast in 2010 include conservative bastions like Wrentham, Attleboro, Carver, Blackstone, Bellingham. When I issued my turnout warning a month ago, it was predicated on the assumption that conservatives almost always show up to vote. This time they didn’t.
One reason I’ve heard for low conservative turnout is that Gomez’s attempt to present himself as a moderate turned off the GOP base. To the extent that’s true, it may bode well for Democratic candidates going forward. As a commenter on Daily Kos noted last week, astutely I believe, in recent years the floor has been rising for Massachusetts Republicans, but the ceiling has been falling as well. Just by being Republicans, and particularly if they must work to hold the social conservative base, Republican candidates will be associated with policies that a majority of Bay State voters oppose.
And perhaps Massachusetts Democrats do best in elections with either very high or ridiculously low turnout, but have a challenge where turnout is low, but not abysmally so. You might liken it to the Patriots in recent years: like the Democrats, they tend to do better in high-scoring (analagous to high-turnout) contests. You’d expect them to have trouble in a low-scoring game. But maybe — if it’s a situation like a blizzard — their exceptional offensive prowess would make them more likely to get the touchdown needed to win 7-3.
Still, there is reason for concern. We can’t count on low conservative turnout all the time, nor can we continue to win if cities like Lawrence or New Bedford continue to show such pitiful voting numbers. We have to crank it up – even more – and make sure our supporters come to the polls.