Princeton Election Consortium

Innovations in democracy since 2004

Oct 25: Biden 360 EV (D+5.6% from toss-up), Senate 52 D, 48 R (D+4.2%), House control D+5.0%
Moneyball states: President ME-2 NV NC, Senate AK MT IA, Legislatures KS TX NC

Roy Moore as an ultimate test of the power of partisan loyalty

December 12th, 2017, 9:11pm by Sam Wang

Today’s Alabama special election to replace Senator Luther Strange (R) is of obvious interest for two reasons. First and foremost, since the Senate is now 52 R, 48 D/I. After tonight, it will either stay the same, or become 51 R, 49 D/I. This would adversely affect the legislative ability of an already-dysfunctional Republican Congress.

Second is the emotionally wrenching nature of the race. The Republican candidate, Roy Moore, has attained worldwide fame…for bad reasons. A judge who has been ejected from office for violating his oath, he is also credibly accused of acting on his sexual attraction to teenage girls, and of molesting a 14-year-old (that’s 8th grade). He runs against Democrat Doug Jones, a rock-ribbed Alabaman who is known for prosecuting KKK members who killed little girls in a bombing. Despite all this, Moore has been slightly favored to win. No matter who wins, the closeness of this race tells us something useful about the current national landscape.

Basically, partisanship overrides other factors in the current national political environment. As I’ve written before, voters these days don’t change their minds during a campaign, and they vote straight party ticket. We can use Roy Moore – and before him, Donald Trump – to measure the loyalty of Republican voters to their party. Think of these two specimens as useful extreme cases, which tell us just how entrenched voters have become.

Consider 2016. Trump lost the popular vote to Hillary Clinton by 2 percentage points, despite the fact that House Republican candidates beat Democratic candidates by an average of 1.1 percentage point. That 3-point difference demonstrates just how few voters were put off by the most disruptive Republican candidate since Barry Goldwater. If that difference arose from GOP voters switching to minor-party candidates, it means that as many as 3% of voters were Republicans who couldn’t vote for Trump. Since Republicans were about 48% of the popular vote, that means about 6% of otherwise-Republicans nationally couldn’t vote for Trump – and 94% stuck with him. Even if we only count the approximately 38% of survey respondents who approve of his performance, that means 79% of Republican voters currently support him.

Now let us turn to Alabama, a deep-red state. In three statewide races in which no sexual predator was representing the Republican party, we have the following results:

2014 Governor: Robert Bentley (R) 64%, Parker Griffith (D) 36%.
2012 President: Mitt Romney (R) 61%, Barack Obama (D) 39%.
2010 Governor: Robert Bentley (R) 58%, Ron Sparks (D) 42%.

The median of those three races is R+22%. The last six polls for today’s race give a median of Moore +6%. That’s a swing of 16 points, which would be consistent with 8% of Alabama voters being turned off by a child molester and switching to the Democrat. (It could also be more, if the difference comes from depressed GOP enthusiasm.) If 61% of Alabama voters are Republican-preferring, it means that only 13% of Alabama Republicans will take a Democrat over a molester – and 87% of them are still okay with Moore. Even if Jones pulls out a win, this percentage only drops to around 82% of Republicans.

Obviously, turnout affects these calculations. For example, if Jones wins by getting higher turnout in the African-American community, that would suggest that GOP voters are loyal, but too demoralized to vote. That would be a surprise in a special election, and worth watching out for.

As Nancy Pelosi has said, Donald Trump’s success proves that to Republican voters, “any mammal will do.” Counting Roy Moore as a mammal, today’s election fits with that idea.

Update, Wednesday morning 8:00AM: Jones did win, by 2 percentage points. Which means we can quantify the Alabama swing as 24% toward Democrats. It should be noted that over 15% of this swing occurred before the Washington Post bombshell. A 15-point swing is consistent with other special elections this year. Yesterday’s election could only be made possible by Trump’s deep unpopularity.

Exit polls showed that 91% of Republicans voted for Moore. The discrepancy between that and 82% suggests that turnout played a major role – see my Twitter feed for analytics from others quantifying this.

Tags: 2016 Election · Senate

20 Comments so far ↓

  • Jack Rems

    The NYT live results are kind of thrilling right now; Moore leads but the projection is 90% Jones; must be based on where the votes are coming from.

    • Matthew McIrvin

      The way it shook out was analogous to what we often see in Virginia: heavily Republican rural areas reported in early, the heavily Democratic cities late, so it was possible to predict that Jones was going to win even while Moore was still leading.

    • LondonYoung

      And this is a big problem with trust in elections. Republicans who are unaware of this fact are often left with the feeling that election officials cheated them. As a bit of public service, it might help if every citizen was required to serve at a polling station on one election day early in their lives …

    • Matthew McIrvin

      Democrats tend to be convinced that elections were rigged whenever they do worse in the count than in exit polling (which often happens).

  • bks

    Finally, I can breathe again.

  • Pat L

    Hooray! A glimmer of hope from Alabama. Let’s keep moving in this direction. Thanks for all your work and information on gerrymandering.

  • Some Body

    The one other factor to take into account for 2016 is switches in party ID. Some Republicans have been sufficiently turned off by Trump to leave the GOP altogether, perhaps even switch to Dem (so, not just when voting for the mammal in question, but down-ballot as well). We know anecdotally some such people exist (and, say, write columns in media outlets), though I don’t have good ideas for measuring how many. Your method for 2016 seems to miss them. Correct me if I’m wong.

    • Sam Wang

      I don’t think that’s right, since I am using actual vote totals to define “Republican” and “Democrat.” You are thinking of survey data.

      However, this approach does leave out differential turnout, which is why I wrote about that.

    • Some Body

      Sure. But if you’re trying to measure the strength of GOP partisanship, I’d think people who switched from straight-ticket GOP to straight-ticket Dem *because of Trump* should be in the same category with those 6% who only bucked Trump himself. Election results definitely don’t give us a count of these people, though. And why exactly they switched is subjective.

  • bks

    Is it true that Moore won 6 of Alabama’s 7 congressional districts (because gerrymandering)?

    • Sam Wang

      See the analytics at It’s a little ambiguous, since under the Voting Rights Act I believe there had to be an ability-to-elect district for blacks (and therefore Democrats). The question is whether it was overpacked. Probably – the incumbent appears to have been unopposed.

      In a 64-36 state, one might reasonably expect a delegation to split 5-2. That would suggest one seat was stolen by redistricting. You can see the nature of the offense by inspecting the one Democratic district: those protuberances reaching out to the NE and SE. No prizes for guessing what’s in them! It seems to be a gerrymander on a par with Maryland.

    • LondonYoung

      I think part of the problem is a “bug” in the voting rights act and SCOTUS interpretations. Cracking a potential majority minority district is a no-no, however those same african american votes could be districted to elect either two white democrats or one AA democrat. The wording of the VRA leaves both open and depending on what J. Kennedy had for breakfast the court could go either way. The protuberances of AL-7 represent the GOP preference for the second districting option.

    • Sam Wang

      Until not so long ago, the preferred approach was to build majority-minority districts, the option in AL-7. However, I believe that district is super-packed, even more than needed to achieve majority-minority.

      Because of creeping polarization, standards have shifted toward ability-to-elect districts, which are less packed under current conditions of partisanship. This concept seems less prone to overpacking, and is easily brought into conformance with some proposed standards for partisan gerrymandering. Partisan gerrymandering standards might make it hard to build super-packed districts, unless a GOP district is also made super-packed…a bipartisan gerrymander.

    • LondonYoung

      AL-7 might not be so super-packed. Note that AL-3 is over 32% AA, yet it is R+16.

  • Andy

    Really great story and graphic in today’s WaPo: How Doug Jones lost in nearly every congressional district but still won the state. Amazing graphic that shows how the state was carved up to concentrate Af Am voting power.

    • Sam Wang

      The “Alabama is severely gerrymandered” meme is problematic. Turnout was different between districts. If you just average the vote share across districts, then Moore wins by about 0.2 percentage point. Also, the black community was far more lopsided in its vote than the white community. I believe (not sure) that this effect was likely to be larger in this week’s election than in the November 2016 election.

    • LondonYoung

      There is a bit of history which we need to remember – when the VRA was first written and litigated, racists still called the democratic party their home. So for african americans a white democrat didn’t do them any good – the feds considered it a plus when an african voter could be in an african district.

      To this day, white democrats will not usually vote for the descendants of southern slaves. The only dark skinned senator is a republican … We still have a long way to go.

  • Matthew McIrvin

    The generic House ballot polling is annoying, because live phone polls are most showing gigantic gaps like D+14, D+15 (more than enough to power through national gerrymandering), while Internet polls show D+6, D+7 (right on the edge to maybe eke out a bare majority).

    The question of which to believe is key. Live phone polls might have more social-desirability bias. On the other hand, special-election results suggest a big enough swing from previous cycles that the higher numbers might be the ones to trust.

  • LondonYoung

    Random question – Maryland has enough african americans that they could easily elect three of the eight reps. But districts 2 and 5 have obviously been cracked to siphon off AA votes and create an extra white dem at the cost of losing one AA seat. (They could even have created four AA seats.) Why is this not a slam dunk VRA case?

Leave a Comment