In less than three weeks, the 2018 election will head to Arizona. Republican Rep. Trent Franks resigned late last year amid a scandal involving money, staffers and surrogacy (it’s a bizarre story), triggering a special election in Arizona’s 8th District. In a normal year, this district would be safe—Donald Trump won it by 21.1 points in 2016.
But this isn’t a normal year. The president is historically unpopular, and less than a month ago, Democrat Conor Lamb eked out a win in a district that Trump won by 19.6 points a special election called after a Republican congressman resigned in disgrace. So it’s worth asking whether Arizona’s 8th District will flip and how much the results tell us about the November midterms.
Why AZ-08 Might Be an Easier GOP Hold Than Other Districts
The first step in understanding Arizona’s 8th District is understanding its context. The 8th sits in the middle of Maricopa County.
Maricopa County is home to much of the Phoenix metro area, and the majority of Arizona’s voters. In every presidential election since 1988, between 58 and 60 percent of the votes for major parties were cast in Maricopa County. That’s a lot of votes—Trump got more votes from Maricopa County than he did from the entire state of West Virginia, and Clinton got more votes there than she did in Maine.
These are precinct-level results for every presidential election since 2016 (colors are based on the two-party vote share). Over time, Democrats have gained strength in the center of the city, while Republicans have fortified their position in the outlying areas, leading to something of a county-wide stalemate. George W. Bush won about 55 percent and 57 percent of the two-party vote in 2000 and 2004 respectively. John McCain (who was likely able to offset some of his national disadvantage with a home state advantage) won 55 percent of the two-party vote, as did Mitt Romney in 2012. But in 2016, the gap narrowed and Trump won only 51.5 percent of the two-party vote.
But this change didn’t hit the 8th District nearly as hard. According to Daily Kos Elections data, Trump won 61 percent of the two-party vote in 2016 after Romney won 62.6 percent of the vote there. Clinton won a percentage of the vote nearly equal to Obama there (Obama got 36.9 percent of the vote in 2012, Clinton got 37 percent), and the losses mostly came from Trump failing to match Romney’s percentage of the vote.
That’s good news for the GOP.
There’s some evidence that Democrats have performed better in special elections in areas where Trump improved on Romney’s performance (though the strength of the relationship varies depending on how you set up the regression and there’s real noise there). Maybe more importantly, registered Republicans outnumber registered Democrats in the district—something that wasn’t true of Pennsylvania’s 18th, where Lamb likely reconverted some registered Democrats who had been voting like Republicans. So it’s not hard to imagine a somewhat old, historically Republican district like this one sticking with the GOP despite a national environment that favors Democrats.
It’s also not clear that Democrats have found a Lamb-quality candidate in this race. Candidate quality is a slippery thing, and the “good candidate” or “bad candidate” label often gets deployed as a post-hoc explanation for a disappointing loss. But it’s clear that Lamb was a good fit for his district. He’s a veteran from a local political family and often passed on criticizing the president much while attempting to strike a moderate tone on social and cultural issues. The Democratic candidate in this race, Hiral Tipireneni, hasn’t commented on whether she’ll support Nancy Pelosi. Tipireneni may end up being as talented as Lamb or Alabama Sen. Doug Jones (who won a special election to that seat in December 2017), but it’ll take more on the ground reporting to be sure of that.
The demographics, candidates, district, and national environment add up to a race where Republican Debbie Lesko is favored. Sabato’s Crystal Ball and Inside Elections rate the race as “Likely Republican” and Cook Political Report says it’s “Solid Republican.”
So How Should I interpret the Eventual Result?
The best way to interpret special election results is to compare them to some appropriate baseline (e.g., previous presidential election results, possibly corrected for overall popular vote) and average them together. That’s the best method because these races don’t always point in the same direction and are often individually weird (note that part of this section is based on a chain of emails between me and Sean Trende of RealClearPolitics.)
The level of Democratic overperformance in House races is all over the map.
In most races, Democrats strongly outperformed Clinton. Lamb, for example, outperformed Clinton by about 20 points. In Montana, South Carolina’s 5th District, and Kansas’s 4th District, the GOP candidate fell short of Trump’s margins by double digits, but those underperformances ranged from about 14 to roughly 21 points.
Those results are reasonably consistent with each other. But in Georgia’s 6th District and Utah’s 3rd District, Republican candidates outperformed Trump’s 2016 margin. And it’s tough to know what to do with the results in California’s 34th District since two Democrats made it to the runoff.
House specials then, on average, show a Democratic advantage, but the range of outcomes is large and the total number of data points is low—meaning that it’s not a great idea to extrapolate too much from any one election.
More broadly, individual Trump-era elections have been idiosyncratic, and it’s possible to reasonably object to using many of them to project midterm results.
Gubernatorial elections often (though not always) pivot on state rather than national issues, so a reasonable person could pass on using results from off-year gubernatorial races in Virginia and New Jersey.
Special state legislative elections have been low-turnout affairs where high Democratic enthusiasm sometimes created a turnout advantage that would be unrealistic in the November midterms. And, at the risk of pointing out the obvious, governors and state legislators typically work on state issues rather than national issues. So maybe it’s only appropriate to use Senate and House special elections to predict midterm House and Senate results.
But the only special U.S. Senate election of the Trump era—Alabama—featured Roy Moore, a GOP candidate who was credibly accused of having improper sexual contact with teenagers while he was in his 30s. Other Republicans might have a tougher-than-usual election due to Trump’s unpopularity, but Moore’s unique problems may make this data point imperfect for projecting what’ll happen in November.
And some of the previously mentioned House specials have real issues. It’s not clear whether Trump (who was uniquely unpopular among Mormons) or Mitt Romney (who was extremely popular with the group) provides a fair baseline to judge John Curtis’s performance in Utah’s 3rd District. In Montana, Republican Greg Gianforte might have lost some votes by body slamming a reporter right before the election. The GOP’s underperformance in Kansas was likely a result of Trump’s unpopularity, but it’s hard to gauge exactly how much of the blame should be shifted onto unpopular then-Gov. Sam Brownback.
The point here isn’t to pare the data down to nearly nothing. It’s to show that the individual data points are imperfect and uncooperative, and that it’s dangerous to extrapolate too much from any one of them.
So If Democrats see a strong result in Arizona’s 8th District, the GOP shouldn’t drastically change its midterm forecasts. And if Lesko performs better than other Trump-era GOP candidates, Republicans should keep the Champagne corked. In both cases, both parties should compare the result to a decent baseline, throw it into an average with other results and see what the result looks like. That method of aggregation will help quiet the data down and provide us with another updated measurement of how the public feels about both parties in the Trump era.
