“In a contest defined by record turnout for both candidates, the outcome of the race hinged on who was able to attract the most new votes — and where. A blue wave in the majority-white suburbs of Pennsylvania and Georgia was pivotal to Mr Biden’s win in the electoral college. In Arizona, both majority-white and majority-Latino areas in Phoenix and Tucson contributed tens of thousands of net votes to Mr Biden, overturning a swing towards Mr Trump in majority-white rural areas.”
Financial Times, November 15, 2020
It is fair to say that many Democrats are disappointed about the Senate results in Maine and Biden losing the second Congressional District.* However, as the map above from The Washington Post illustrates, Biden won the state in a landslide. Both Democratic Congressional candidates won their races. Yet, Democratic Senate candidate Sara Gideon lost by eight points (see map below). The margin is surprising, but not the outcome - even though most Democrats believed she would win (based on a misunderstanding of state of the polling in race, in my opinion). Many want to know why Gideon lost when Biden won by ten points.
Did Gideon underperform in Maine? There are two ways of looking at that question. First, did Gideon underperform the polling? Second, did Gideon underperform Biden? The answer to the second question is an unequivocal yes. It is why this happened that is the more interesting question. But, first let’s deal with the former question: did she underperform the polls?
If we compare the polling in the Senate and Presidential races in Maine, there is no question Gideon did worse than Biden. Biden’s margin was consistent with the polling, but Gideon’s was way off. But this is more about misunderstanding polling and how that leads to false expectations. In this case, expectations were unjustifiably high. While it is true that Collins only led in one of 22 polls conducted for that race since June 1, the typical margin was only a couple of points. This is why most observers had Maine as a toss-up Senate race until the very end. What is surprising is the margin by which Gideon lost (just under eight points). However, when you add in progressive independent Lisa Savage’s votes the race was closer (50.6-45). Unlike what might be true in other states, Savage did not take votes away from Gideon and cost her the election. This is because (1) Maine has ranked choice voting, (2) one would assume that most of Savage’s voters chose Gideon as the second choice, and (3) Collins won over 50% among first choice votes (which means all other candidates together got less than 50%).
As of this writing, GOP Sen. Susan Collins won 219,000 votes to Democrat Sara Gideon’s 129,000 in ME-02. Trump beat Biden there 180,000 to 156,000. Democratic Rep. Jared Golden was reelected to Congress there by 174,000 to 150,000. There were two other candidates in the Congressional race. It appears that about 48,000 more votes were cast for Senate than for president (or at least for the two major party candidates – Jorgensen and Hawkins got 2.7% of the statewide vote). In ME-01, Gideon got 211,000 to 196,000 for Collins. Biden got 263,000 and Trump 162,000. Democratic Rep. Chellie Pingree got 211,000 to 139,000 for her opponent.
Biden outpaced Gideon by 52,000 votes in ME-01 and 51,000 votes in ME-02. That’s 103,000 voters that presumably voted for Biden and Collins. That’s 13 percent of the entire electorate. These voters are the ones who reelected Collins.
Compared to the top of the ticket, Gideon did underperform. From the exit polling, it looks like about 15-18% of Biden voters voted for Collins rather than the 13% it appears from a glance at the election returns.** Considering that we do have polling error associated with exit polls, the number that ticket-split is likely in the 13-15% range. Collins tied in the suburbs whereas Biden won them 57-41. (This is exit polling data.) Biden tied in the rural areas, whereas Collins won them big (59-36). The exit polling split the state into four areas: Portland Area; Southern Maine; Bangor/Auburn; Downeast/Upcountry. Biden won the first three regions. Gideon only won Portland Area. Women went for Biden 61-38. Collins won women 49-47.
So, it looks like Gideon's big problem was that too many suburban Biden voters split their ticket – or bulleted Biden. Biden won statewide by a big margin. That makes sense when you consider he won a lot of rural votes. That margin is an 18 point swing for Gideon in the wrong direction. If Gideon could not win with Biden taking a 10 point win, it’s hard to see how she could have ever won.
Challenges with polling are not just with the mechanisms of polling itself, but in public perception. The public demands exactitude from polling – which will never happen (polling is not a population count, but only an estimation of that count). The media does not do enough to educate the public on what polling means and how probability works. Instead, it has click-bait headlines that entice readers and viewers to think that when Biden has 52% in a poll it means 52% of voters support him, rather than there is a 90-95% chance that Biden has between 49% and 55% of voters supporting him. And this only holds true for the period the poll was in the field.
In Maine, the polling missed on Collins’ strength. Why? Is it because Republicans did not respond to polling? There’s no reason to assume the partisan non-response problem would not be operative in Maine as it appears to have been in other states, although that could be the case. However, if that was true shouldn’t the polling have greatly overstated Biden’s strength in the state? In fact, the final two A-rated polls showed Biden winning the state by +11 and +13, which is well within a normal polling error of the final result - unlike the polling on the Senate race. Partisan non-response simply does not explain why it appears that over 100,000 voters in Maine appear to have voted Biden/Collins.
Here's something that resides mostly in the expectations of the public, not in the actual polling: Collins’ reelection in Maine being a surprise. The polling was very close in that race. If we applied the 2016 polling error (3 points) or a normal polling error (4 points) to almost all of the polling in the Maine Senate race, Collins would have been in the lead in many of those polls. In fact, in the final weekend Collins would have been leading by one or two points. Her actual margin was much larger – almost eight points, well outside a normal polling error. However, the reason most people assumed Gideon would win is because Collins was behind in almost all 22 polls of the race since June 1. If a normal polling error were applied, then it would have appeared that the race was, at best, a toss-up. In fact, many observers thought this race was a toss-up.
It appears that the working hypo I articulated in last Friday’s newsletter is operative also in Maine, but with some differing emphases. For one thing, there is such a small share of the statewide electorate that is nonwhite in Maine that the issue of decreased Black share of the electorate seen in other states is probably not applicable here. However, the other three concepts are operative: (1) there was a nontrivial amount of suburban ticket-splitting (this appears stronger in Maine than in other states), (2) Biden picked up rural voters that did not otherwise vote Democratic, and (3) women voted stronger for Biden than down ballot races.
In the past few days have talked with several folks who phone banked in Maine for the coordinated campaign. I hear two things: (1) Mainers did not like being called by people from outside the state and some voters seem to think voting for Collins was a way to show “outsiders” that the election was for Mainers to decide*** and (2) regardless of who was calling, Mainers were very frustrated with the number of calls they were getting by the coordinated campaign. Golden was not part of the coordinated campaign. In what may have been a very smart move, Golden apparently ran a separate campaign from Biden and Gideon. Maine’s second CD is GOP-leaning, as we can see from the presidential and Senate results this year. However, they voted for Golden in 2018 and reelected him this year with almost 54% of the vote.
Absent future scandal, the first reelection is typically the hardest electoral barrier for a Congressional incumbent. That Golden did so well in a district that Biden lost by a clear margin (and where Gideon was wiped out by over 25 points) suggests that Golden is very responsive to constituent matters, he’s seen as an independent Democrat, and his campaign ran on a message that was more convincing than the coordinated campaign’s to the voters in his district. If I had to guess, I would think Golden probably did not focus on an anti-Trump message while the coordinated campaign did. This could have been a problem in other down ballot races in other states too. If anyone has direct familiarity with the election in ME-02, I would be interested in hearing about it.
* Maine is one of two states that apportion Electoral Votes by Congressional District.
** We know that Biden got at least 103,000 more votes than Gideon, which represents about 13% of the electorate. However, we do not know which voters ticket-split or whether there was a significant amount of bulleting, or whether there was ticket-splitting with minor party candidates. We do know that about 48,000 more votes were cast in the presidential race than in the Senate race. Since the exit polling tells us that roughly that amount did split their ticket Biden/Collins, it is fair at this point to assume almost all or more of that 103,000 pulled for Biden and Collins.
** This is far from the first time I have heard of voters – any voters, anywhere – saying that they were going to vote one way to show outsiders, or activists, or because of frequent calling/mailing. The irony of the matter is that – if they are being honest – then they are voting against their interests just to make a self-defeating point. However, in my experience I have found that many people who make this argument were never going to vote your way in the first place. I did a lot of textbanking during this election – no phones, just texting this year – and no matter what the state a certain number of people complained about getting texts and a subset of them (again, in every state) said they would vote against Democrats just because they were annoyed with the texts. Where were those numbers high in the states I texted? Texas, North Carolina, Ohio, Missouri, and Florida. All states Trump won. In Michigan, Wisconsin, Arizona, Pennsylvania, and Nevada very few people said they would vote differently just because they were annoyed by texts (but plenty still complained about getting texts). Biden won those states. (This is just my personal experience.)