The first thing a lot of people want to do today after seeing the results of the special election in Pennsylvania’s Eighteenth Congressional District last night is to use the result to justify their preexisting theories for why Hillary Clinton lost the election in 2016. I’d be wary of listening to these folks. It’s better to test your theories than to bend everything to fit them.
Let’s start by looking at the county-by-county results. The following chart comes from Daily Kos Elections and provides the Clinton/Trump numbers along with what Lamb would need to last night to win.
| COUNTY | % OF 2016 VOTE | 2016 RESULTS (D/R) | WHAT LAMB NEEDS TO WIN |
|---|---|---|---|
| DISTRICTWIDE | 100 | 39/58 | 49/48 |
| ALLEGHENY | 43 | 46/50 | 56/40 |
| WESTMORELAND | 33 | 31/65 | 41/55 |
| WASHINGTON | 22 | 35/61 | 45/51 |
| GREENE | 2 | 27/70 | 37/60 |
The actual results were:
DISTRICTWIDE: 49.86/49.54
ALLEGHENY: 57/42
WESTMORELAND: 42/57
WASHINGTON: 46/53
GREENE: 42/58
The third party vote was lower than the Daily Kos folks expected. The Libertarian Drew Miller received 1,379 votes, or just 0.6 percent of the total. As a result, Saccone exceeded his benchmark percentage for victory in three of the four counties but still lost. This was because Conor Lamb did better than his benchmarks in all four counties, including the most populous, Allegheny County.
According to a snap exit poll taken by Public Policy Polling, forty-two percent of the electorate last night reported having voted for Hillary Clinton in 2016. She actually received only thirty-nine percent in this district, so there was a three percent improvement in turnout of Clinton voters. Lamb, however won 49.86 percent of the vote, so there was an additional 7.86 percent of non-Clinton voters in his tally. Some of these people did not vote in 2016. Some voted for a third-party candidate. Most of them voted for Trump.
This is just data from a snap poll and not necessarily perfectly accurate, but it roughly describes a victory for Lamb based on thirty percent Democratic enthusiasm/Republican depression and seventy percent persuasion. Remember this when you see people argue that no Trump voters were convinced to cast a vote for Lamb.
The only voter I talked to yesterday who voted one party for president in '16 and another in PA-18 yesterday went Clinton-Saccone. All others voted for same party or hadn't voted in '16.
— Jonathan Allen (@jonallendc) March 14, 2018
If I'm reading this right, 51% of PA-18 voters on Tuesday voted for Trump. 49.6% of them voted for Saccone. That means that fewer than 2% of the voters on Tuesday were switchers. The majority of the 20 point flip was therefore mobilization. https://t.co/qm8jiOHySE
— Amanda Marcotte (@AmandaMarcotte) March 14, 2018
Another indicator is from the actual election results we looked at above. If we look at the net (+/-) numbers for the benchmarks in each of the four counties, we can see an outlier.
ALLEGHENY: Actual: 57/42 Benchmark: 56/42 Net: Lamb +1
WESTMORELAND: Actual: 42/57 Benchmark: 41/55 Net: Saccone +1
WASHINGTON: Actual: 46/53 Benchmark: 45/51 Net: Saccone +1
GREENE: Actual: 42/58 Benchmark: 37/60 Net: Lamb +7
Greene County is the most rural, least populous (by a wide margin) and most Trumpy of the four counties. But this is where Lamb over-performed the most. It’s also a county that gave fifty percent of its votes to Barack Obama in 2008. John McCain carried the county by a mere sixty votes. Hillary Clinton got twenty-nine percent of the vote there (and only twenty-seven in the 18th District portions of it).
This is strong evidence that a lot of Obama/Trump voters came home last night to the Democratic Party in Greene County. Given the closeness of the election, this may have been the decisive factor.
I’ve never discounted racism and misogyny as key factors in Trump’s victory in Pennsylvania, but I have argued that these voters were not irretrievably lost to the Democrats. I’ve argued for both moral and practical reasons that their votes and support needed to be courted. Lamb tested out my theory in this campaign, and he didn’t do it by pandering or catering to their prejudices and bigotry. This, too, is what I recommend and said was possible.
This election has been overhyped, to put it mildly. Conor Lamb will probably seek a different district when he runs for reelection in the fall because this one will be too Republican for him to win. But Lamb showed how to dramatically improve Democratic performance with rural voters without selling out liberal values on key social issues. And this is the key to the Democratic Party reclaiming its place as the dominant party in this country on both the state and the national level.
Wait a minute:
Doesn’t that mean more non-Hillary voters turned out last night than in 2016?
If all of last night’s voters who also were Hillary voters (voters meeting both those conditions) were only 39% of last night’s turnout, but those same people were 41% of 2016’s turnout, then there was more turnout last night of people who voted for Trump, right? The people who stayed home last night aren’t counted by definition.
What? No.
The electorate in 2016 gave Clinton 39% of their votes.
Last night, the electorate was made up by 42% people who voted for Clinton two years ago.
In other words, the Clinton voters made up a bigger piece of the pie this time around. That’s a result of differential turnout, a combination of enthusiasm, depression and ground game.
But there’s additional almost eight percent of Lamb’s vote that came from people who did not vote for Clinton two years ago. Some may have voted for Stein or Johnson, and some may not have voted through apathy or because they weren’t then eligible.
Most of the eight percent, though, came from people who voted for Trump.
The same number of people — Clinton voters — are a bigger piece of the pie because the rest of the pie got smaller.
There’s no such thing as a “Clinton voter” who stayed home in 2016. The 39% and the 42% have to be the same number of people, because they’re the same people: they have to vote in both elections or they aren’t part of that statistic. So it’s the remainder of the voting public that’s getting smaller.
what? no.
while you’re right that the pie got smaller and therefore a given number of voters would make up a higher percentage (or bigger slice), that’s completely irrelevant.
not every Clinton voter participated in both elections.
what’s relevant is that Clinton voters grew as a percentage, not that they grew as a number (they didn’t).
what happens when your side has a better ground game than they did the last time around? you do a better job of getting out your vote and your percentage of the electorate increases. same things happens with no ground game if your side is just more motivated (relative to the other side) than before.
this is what shows up in Clinton going from 39% to 42% of the electorate.
the remainder is what concerns us here. Who were they? They weren’t primarily third-party voters nor were they non-voters. Statistically, there will always be more non-voters in a election with much lower overall turnout.
We’re talking about Hillary Clinton voters, right? Then, yes, the ones we’re discussing participated in both elections.
If they didn’t vote for Hillary in 2016, they’re not Clinton voters — they’re Clinton supporters. (It’s conceivable that some voted for Clinton in the primary and then sat the general out, but I can’t imagine any significant number of people doing this; it makes no sense. Sanders supporters, yes; Clinton supporters, no.)
So every “Clinton voter” voted for Clinton in 2016. (Otherwise, again, we’re just contradicting ourselves.) That’s the 39%.
And we’re not counting anyone who if they didn’t vote last night, because they’re not part of the snap poll. There’s no way to get that figure, even if such people exist: “I voted for Clinton in 2016, but I didn’t vote tonight so you’re not talking to me right now; I’m not here.”
The two percentages work out to the same number of people; the same voters.
Where is the evidence for this?
The evidence, per Marcotte’s reference, is that the large majority of the non-Clinton pro-Lamb voters were also non-Trump voters – nearly 75%. So the story isn’t Trump voters coming back to vote Dem, it’s third party and non-voters showing up to vote Dem, as per the usual interpretation.
I’m interested in hearing from Michael McDonald, who like Booman was focused on persuasion. He points to differentials between straight party ticket voters and Lamb solo. Allegheny Co had it at 57.2 Lamb, 54.97 straight Dem. Greene 41.5 Lamb and 40.3 straight.
He will only tell you “What a Fool Believes”
were there any other candidates on the ballot yesterday. I don’t think voting party ticket means anything when there’s only one candidate.
Here’s his tweets:
link
Ends with:
I’d have to see the ballots or the touchscreen or whatever, because I think he’s making something out of nothing. Basically, it’s whether you check the first box or the second, when they mean the exact same thing. And the first box is the party box.
what? no.
you’ve just been shown a chart demonstrating that 172 precincts that voted for Trump wound up voting for Lamb. There weren’t enough third party votes in first election for the drop-off to swing but a tiny handful of precincts.
that’s one piece of evidence.
another is that the most Trump-Friendly county of the four was the only with the biggest swing.
another is that despite better turnout in relatively Dem-Friendly Allegheny County (Clinton actually lost the 18th District portion of it), the upswing in turnout and depressed turnout in stronger Trump territory can only explain 30% of the swing.
Taken all together, that’s massive evidence of both suburban Trump voters leaving in the GOP in droves but the phenomenon being even stronger among rural (deplorable) voters.
you have to try incredibly hard not to see this swing in an election that just went from 19 points Republican to dead-even.
Where are you getting this?
19 point overall swing (let’s make it 20).
need to explain 10 point shift.
Clinton voters increased from 39-42 for a three percent increase. This is ground game, higher motivation relative to Republicans, also seen in Allegheny getting higher turnout than, say, Westmoreland.
Where do the other seven points come from?
They are not Clinton voters by definition.
It’s a lower turnout election than the general, so we can’t find a wealth of people who didn’t vote last time but did vote this time. They are, of course, some of these people, including new registrants.
Most of the seven percent came from people who either voter third-party instead of for Clinton or Trump but now voted for Lamb or who voted for Trump and switched over to Lamb.
There’s a little pool of Stein voters, but there’s also a pool of Johnson voters. Some fraction belongs to these folks, and they wash out to some degree.
In any case 70% of the surge is from non-Clinton voters who cast a ballot for Lamb. Most of them are Trump voters.
Again, how can “Clinton voters” increase from 39% to 42% if they only voted for Clinton once, the first time? What makes a Lamb voter a “Clinton voter” if not the fact of having voted for Clinton in 2016?
“I didn’t vote for Clinton in 2016, but I turned out for the special election last night. Call me a ‘Clinton voter’!”
You can’t increase the number of participating or increasing “Clinton voters” in a year when Clinton isn’t on any ballot. All that can happen is that people who previously voted for Clinton and came back are now a larger percentage of the total participating voters because other numbers changed.
Why is this perplexing to you?
The 42% number comes from the Exit Poll. Forty-two percent of people who voted yesterday reported that they had also voted for Clinton in 2016. That means 58% did not. It also means that the electorate yesterday had 3% more Clinton voters than the electorate in did in 2016.
Is this too hard to understand?
But Lamb did much better than 42%. He got almost 50 percent. He did not get from 42% to 50% by turning out more Clinton voters. He got there by winning over additional eight percent of the electorate who DID NOT VOTE for Hillary Clinton.
This would not have been enough without the increased percentage of Clinton voters in the electorate, so turnout efforts mattered. The ground game was important. Republican apathy also mattered. But it accounts for just 3% of the 10% boost we’re looking for.
Who could these other people be?
Well, there was no Green Party candidate on the ballot this time, so Lamb presumably got the lion’s share of their votes. There was a libertarian on the ballot, but he only got 0.6% of the vote. In any case, Lamb would be lucky to get much more than half of their votes. Third party voters can only explain a small fraction of the improvement.
Who else is there?
Well, people who didn’t vote at all in 2016, either because they were too young, hadn’t registered, didn’t care or had to take their dog to the vet that day. Whatever. But, actually, there were many fewer people overall voting in this election than in 2016 and it’s unusual for people to vote in Special Elections when they aren’t in the habit of voting in presidential elections. So, this is a very small pool of people.
What’s left is Trump voters. And that’s who made up most of Lamb’s non-Clinton votes.
3% more Clinton voters, proportionately, means that the percentage of other voters went down; it’s fewer voters in total.
The number of Clinton voters was set in stone in November 2016 — it cannot increase or decrease, ever. It’s the only time she was ever on a ballot; by definition every “Clinton voter” turned out that day and voted for her.
If fewer of those people turned out last night than on that day, sure, that’s possible. They stayed home. But the number can’t go up — you can’t conjure “Clinton voters” out of thin air if they weren’t at the polls in November 2016. If the percentage goes up, it’s going down elsewhere.
I am going to kill myself now.
I re-read the thread and I apologize. I misunderstood several of your posts.
Apologies in general; I’m having a terrible day and am irritable and distracted.
No, no, Martin, don’t do that! Even I got it, and I need a calculator to add single digits.
At least I think I got it; see my attempt to explain what you meant.
you did okay, but there were way fewer Clinton voters at the polls yesterday than in that district in 2016. Not the same number. Not even close to the same number.
the rest of your explanation is accurate. What Jordan said almost gave me a stroke.
He was arguing that the 39% of Clinton voters from 2016 could never go up as a percentage of a different sized electorate, or if they somehow could (which they couldn’t) it could only be because non-Clinton voters didn’t show up. Unless, of course, some Clinton voters didn’t show up.
I had a bottle of aspirin I was going to gobble down, but he apologized and said he’s having a bad day so maybe I won’t have to set myself on fire and shoot myself in the head.
Jordan must have been having a reeeeeaaaalllly bad day because he’s normally a very sensible person. No wonder you were so frustrated.
Maybe you could just set the bottle of aspirin on fire? Or shoot it.
I think the disconnect is that you’re treating the percentages Booman is comparing as representing actual numbers: 2016 = 39 voters, 2018 = 42 voters, in essence. Booman is saying that 2016 = 39 percent of the total XXXX pool of people who voted; 2018 = 42 percent of a smaller pool of voters; thus roughly the same total number of Clinton voters per se, but they constituted a larger share of the total pie.
Happy Pi Day, by the way.
C’mon Jordan, you know better than this. Say 100 people voted in 2016 and 50 of them voted for Clinton. So the percentage was 50%. Now comes 2018 and you have only 60 people voting overall because it’s a special election and not even the November mid-term election and only 25 of those same Clinton voters vote this time (i.e. 42% of those same “Clinton voters”). That means about half of the 2016 Clinton voters didn’t show up (at least in the exit poll), which means the difference taking Lamb up to a tiny fraction under 50% and the win had to have come from 2016 “Trump” voters, at least a portion of which, as we know from Booman’s and others’ previous election results were not traditional GOP voters but probably disaffected conservative Democrats or so-called “independents”. I doubt that Booman is saying that traditional GOP voters suddenly went for Lamb (possibly a few suburban, middle class GOP voters disgusted by Trump and the GOP Congress but probably not many of these).
From a tweet: https://pbs.twimg.com/media/DYOaIWHXcAAjJRb.jpg
PA-18 Precinct Loyalty
HRC->LAMB 122
HRC->SACCONE 0
TRUMP->SACCONE 291
TRUMP->LAMB 172
A crude back of the envelope calculation says that 172 out of 291+172 = 463 == 37% of precincts switched from voting for Trump to voting from Lamb in aggregate.
Out of a total of 585 precincts, where ~80% had voted for Trump in 2016, that is quite a shift. And the data at this level of granularity seems to indicate that many Trump voters did vote for Lamb!
Does it? Based on what?
Well, in addition to avenging white resentment, Conservative voters also want elected officials that give them the stuff FoxNews and AM radio tells them day in and day out that liberals taking from them and giving to brown people.
From Reagan’s “Shining, White’s Only, City on the Hill” to W’s “Companionate Conservatism” in 2016 Trump told them they’d get it all, it’d be cheap and great, and the “undeserving” would be punished for daring to take it in the first place. Hillary made the mistake of trying to talk to these voters like adults.
Trump won the argument then promptly betrayed them. Likely the more self aware Trump voters are the ones who jumped over to Lamb.
Also, if Lamb indeed provided your argument, how is it not going to create blowback and another crop of disillusioned voters if he immediately jets to a more friendly district in a few months? Arguments about the political practicality of such a move are going to fall on deaf ears and Lamb/Democrats will just appear to have abandoned them.
Actually the 18th in its current form ceases to exist after this term is up. If I understand correctly, Lamb is intending to run in the newly formed 17th district, which is supposed to contain much of the territory that led to his victory (e.g., Allegheny County) and his opponent is expected to run in the newly created 14th district. Not sure what the makeup of the new 18th will look like or who will run in it. I’m sort of an outsider looking in when it comes to Pennsylvania. That said, it was a stunning victory, given that it had been well over a decade since any Democratic candidate had won in an 18th that we would recognize as the current one.