The eagerly anticipated day is just about here: the special election in Pennsylvania’s 18th district, where Democratic candidate Conor Lamb faces off against Republican state Rep. Rick Saccone. Despite the district’s dark-red recent history—Donald Trump won the district by about 20 points, and Republican Rep. Tim Murphy (whose resignation triggered this election) won his last two elections without any opposition at all—this is looking very competitive. Lamb has had small leads in three of the last four public polls here, and Republicans have taken to preemptively disparaging their candidate in a bit of expectations management. The race is likely to come down to few points one way or the other, though, and the vote count should be a nail-biter all the way through.
On key election nights, one of the tools that Daily Kos Elections offers is benchmarks for the most important counties that show the numbers that the Democratic candidate needs to hit in order to eke out the barest-possible win. The way we do this is to start with the 2016 presidential election results, both in each county and districtwide, as a baseline, and from there, adjust each county’s numbers so they add up to a Democratic win districtwide. This helps you gauge whether your preferred candidate is on track to win, even when one county (which might have a lean very different from the rest of the district) is disproportionately reporting its results early. As we’ll explain below, because Libertarian Drew Miller is on the ballot as well, we’re modeling a 49-48 Lamb win rather a 51-49 Lamb win.
The first column gives the name of the county (there are parts of only four counties in the district, so we don’t need to leave anything off the table this time), and the second column shows what percentage of the total vote share came from the county portion; in this case, Allegheny County makes up the plurality of the district, though keep in mind that none of Allegheny’s largest city, Pittsburgh, is found within the 18th. The third column shows each partial county’s 2016 results, and the fourth column shows the adjusted number that Lamb would need to hit in each one in order to get to 49 percent districtwide.
||% OF 2016 VOTE
||2016 RESULTS (D/R)
||WHAT LAMB NEEDS TO WIN
In 2016, Donald Trump beat Hillary Clinton by a 58-39 margin in the 18th, so getting to that 49-48 win means adding 10 to each county’s Clinton results and subtracting 10 from each county’s Trump results. As we said above, we’re modeling a narrow Lamb win where he takes less than 50 percent of the vote because Miller will prevent Saccone and Lamb from securing a combined 100 percent, and we’re estimating that the Libertarian will take 3 percent. But it’s certainly possible that Miller doesn’t do even that well. For instance, only 1 percent of the respondents to Monday’s Monmouth poll selected “other” as their choice. Also, this generally isn’t an area that’s amenable to voting third party, as seen that only 3 percent of the 2016 presidential vote went for a third party candidate in the 18th, compared with 6 percent nationally.
This approach certainly isn’t the only way to calculate benchmarks for the PA-18 electorate, though. When I wrote a similar post for the Alabama Senate special election in December (which, by the way, was eerily correct; the model was based on a two-point win for Doug Jones, which is precisely what happened, and was correct within a point or two in most counties), I got various questions asking why. Instead of the 2016 race, I wasn’t using a more obscure race where the results were closer (in that case, the 2012 Supreme Court race that Roy Moore only narrowly won, presaging the 2017 election).
The answer was that doing so doesn't really change the benchmarks much. County-level results tend to stay in the same position, relative to each other, from one race to the next, as they float up and down based on the overall tide, and you just have to use a much smaller adjustment to reach a very similar result.
Handily, though, we have a recent statewide election located very far down the ballot where we can test that out: the 2016 state auditor’s race, where Democratic incumbent Eugene DePasquale ran well ahead of Clinton in the 18th en route to a statewide victory, losing the entire district by only 0.6 percent. According to analyst Miles Coleman, DePasquale hit 56 percent in Allegheny County, 44 percent in Westmoreland, 47 in Washington, and 43 in Greene. If we say we don’t need to adjust at all for DePasquale’s results (since they were close to a tie), we can see those are pretty close to my projected 56 in Allegheny, 41 in Westmoreland, and 45 in Washington. (Greene is a little further apart, but, again, it’s only 2 percent of the district’s total votes, and not really worth obsessing over.)
You might be wondering why the Allegheny County portion of the district had little falloff between Clinton and DePasquale, while the less-populous, more-outlying counties saw Clinton underperform her ticketmates. The answer has a lot to do with the demographics of different portions of the district! If you just looked at this district on the map and thought “well, it’s the blue-collar counties that surround Pittsburgh, so it’s all a bunch of white union guys who were Obama/Trump voters,” you’re missing a big part of the district. For one thing, it wasn’t even a big Obama/Trump district in the first place; Obama got 41 percent here to Romney’s 58 percent, so it wasn’t even a big falloff from 2012 to 2016, districtwide, compared with Appalachia or rural parts of the Upper Midwest.
But the demographic difference that I’m pointing to is that Allegheny County’s portion of PA-18 is, for the most part, affluent suburbs. In this portion of the district, Clinton actually got a higher percentage than Obama did in 2012 (Obama got 45 percent in the Allegheny County portion of the 18th in 2012), consistent with the movement toward Clinton that we saw in a number of well-to-do suburbs around the country, and that helped Clinton run even with DePasquale.
For instance, compare educational attainment in the different county portions. Allegheny County’s PA-18 portion is 45.3 percent college-educated or better among people 25 or older, much higher than the national average. Washington and Westmoreland clock in at 30 and 28 percent, much closer to the national average, and Greene is down at 17 percent. You can see the same thing if you look at median household income: Allegheny’s PA-18 portion is about $69,000, well above the national average. The other counties range between $54,000 and $62,000, closer to the national average of $57,000. (One way in which Allegheny and the collar counties are demographically similar, though, is how white they are; the non-Hispanic white population share is over 90 percent throughout the district.)
The other, more blue-collar counties in the district are what you’d call stereotypically “historically Democratic”—as seen in, say, the 1980s, when candidates like Walter Mondale and Michael Dukakis won here even while getting blown out nationally. But you can still see the remnants here on Tuesday; despite being red at the presidential level, they still have significant Democratic voter registration advantages, and you have a lot of ticket-splitting in favor of local Democratic candidates even when they’re unwilling to vote Democratic at the top of the ticket, mostly for cultural reasons. These are counties where they’d vote happily for, say, DePasquale while voting against Clinton … and Lamb’s best hope of winning the 18th is that the residents of these counties see him the way they saw DePasquale, instead of the way they saw Clinton.