Five Things We Learned from the AZ-08 Special Election

In yet another warning sign for the GOP, Republican Debbie Lesko scraped by with just a 5-point win in the special election to Arizona’s Eighth Congressional District on Tuesday, a seat that Donald Trump won by 21 points in 2016. Lesko’s win placates some GOP fears that the party would lose yet another seat to Democrats in this pre-November 2018 special election cycle.

However, that talking point misses the broader importance of special elections, and in this piece I’ll tell you why.

The bottom line from the election: it’s all about the margin. Democrats will do well in November so long as they win the national popular vote by about 6-7 percentage points.1 In special elections to Federal elected office so far, Democrats have beaten the partisan lean of the seat by an average of seventeen percentage points.

Swings in Federal Special Elections Since 2016
Seat Partisan Lean (%) Vote Margin (%) Dem. Swing (%)
California 34th D+70 D+87 +17
Kansas 4th R+29 R+6 +23
Montana At-Large R+21 R+6 +15
Georgia 6th R+9 R+4 +5
South Carolina 5th R+19 R+3 +16
Utah 3rd R+35 R+32 +3
Alabama U.S. Senate R+29 D+2 +31
Pennsylvania 18th R+22 D+0 +22
Arizona 8th R+25 R+5 +20
Average +17
There’s more information contained in the data, however (within the penumbras of the data, as SCOTUS Justice William Douglas might say), that tell us a lot more about special elections so far.

Along with confirming most of our prior beliefs about the 2018 midterms, here are five things we can learn from the special election to Arizona’s 8th Congressional District.

1. Democrats are Overperforming in High-Turnout Special Elections
What we thought we knew from the highly-anticipated PA-18 special election is very likely true. There, Democrat Conor Lamb swung the seat by 22 points from the district’s partisan lean (R+22) to eek out a win with a razor-thin margin. Turnout in PA-18 was about 67% of the most recent competitive election there in 2012. Turnout in AZ-08 — after everything is counted — will likely approach 60%.

The sixty percent turnout in Arizona’s 8th also rivals the 2014 turnout in the seat. In few words, this is great news for Democrats. When one of the prominent hypotheses for their performance early in the 2018 cycle was that they were simply turning out at higher rates than Republicans (this still could be true), this is not a valid explanation.

As FiveThirtyEight’s Nate Silver wrote after PA-18, there is no correlation between turnout in special elections and Democratic overperformance, at least in aggregate (IE: there could be a subtle relationship at the precinct level.


Democrats are winning in all sorts of elections. Perhaps that’s not because they keep running a bunch of good candidates under extraordinary electoral circumstances (more about this in section three), but rather, it’s possible that Americans just want to give Democrats more votes than they used to (this is how waves work).

2. Republicans are Switching Sides
The second thing to learn from AZ-08 helps explain the first: if Democrats aren’t winning because of differential turnout (or, not solely because of differential turnout), why are they? The only explanation is that Republicans are crossing over to vote for Democrats]( This is clear as day in the early voting numbers from Arizona’s 8th.

Here are the data: In the 155,000 early/absentee mail-in ballots cast in last night’s contest, Republicans ran a 21-point margin in party registration. One would assume (perhaps naively, as candidates from one party aren’t wed to that candidate) that this would give them a 21-point margin in actual ballots cast for either ticket. As I explained on my blog this assumption could go wrong for many reasons:

Early voting data are not “real results,” per se, despite what some analysts would have you believe, since partisanship does not equal vote choice. Though they are very correlated in modern America it is not a safe bet to assume all GOP ballots are for GOP candidates, and vice versa for Democratic voters and candidates. Such assumptions would have led us quite astray in the Texas primaries where Democrats cast more early votes than Republicans for the first time since 2010, but cast just 40% of total votes in the D or R primaries.

Indeed, the early vote did mislead. Debbie Lesko won these “R+21” early votes by just a 6-point margin, meaning there was enough persuasion of Republicans to Tipirneni’s side to move the needle fifteen points. That is certainly (or, at the very least, it ought to be) enough to make many Republican elected officials shake in their boots.

There is an extra point to be made here: even in a contest where 75% of ballots are cast early, our analysis of those results can often go wrong. Stick (though not exclusively) to the polls, folks; Emerson College pegged Lesko’s lead at 6 points. She won by 6.

3. Large Democrats Swings are Happening in “Generic” Contests
There’s a more grim point to be made for the GOP. As the New York Time’s Nate Cohn put it “there just aren’t any excuses” for bad Republican performance here. One can make the case that the other underperformance they’ve seen was because of extraordinary circumstances driven by candidate quality. Conor Lamb (D) was a “moderate,” some would argue, who could play up the ancestral Democratic roots in his district. Doug Jones ran against Roy Moore, a man alleged of sexual misconduct with a minor.

In Arizona’s 8th, these excuses do not pan out. Both candidates were well-funded (Lesko, in fact, received more financial help from donors than Tipirneni). Neither had a special appeal to a certain demographic of voters, like in PA-18. Both candidates were typical (normal, “generic”) candidates who should have been able to fit the partisanship of the district. Obviously, that didn’t happen.

In short, if a 20-point swing can happen in Arizona’s 8th district, it can happen (almost) anywhere.

4. Republican Districts are Still Swinging Harder than Democratic Ones
Another point that I have been making over the course of these special elections is that the Democratic swing has been more pronounced in Republican areas of the country than Democratic ones. We see this is the graph of special election swings based off a seat’s partisan lean:


The negative slope of the line means that the swing to Democrats decreases as a district gets more blue. This has been the pattern since the very first special election results started being reported in 2017.

The graph below from The Washington Post separates Lesko’s margin in AZ-08 out by Trump and Clinton districts. The same pattern from above is evident here.


If this holds true in November, our forecasts of Democratic vote margin in some competitive Republican-held districts could be underestimating the actual Democratic margins.

5. There’s Still a Lot We Don’t Know About November
Here’s where some Democrats might start to worry. Historically, the average swing in state and federal special elections have been fairly solid predictors of the November midterms — especially when combined with the generic congressional ballot. However, they aren’t perfect. If I run a predictive model with them (setting swing in votes from the previous congressional election equal to swing indicated by the average special election and the generic ballot), there is still a 3% margin of error wrapped around the estimate. This is shown in the graphic below:


While Democrats are certainly favorites to make gains in November, then, this is no sure bet. It’s useful to think of the possible outcomes in November as a continuous distribution of events, ranging from Republicans holding the House to Democrats winning it big. That is, after all, exactly how I approach this issue in the formal forecasting models I specify for the midterms. A wide range of outcomes are possible for the House elections (indeed, for most events).


In short, Democrats won big in last night’s special election to AZ-08, and they did so without even winning the seat. Though we do learn a lot from the 20-point swing from the partisan lean of Arizona’s 8th district to Lesko’s 5-point victory margin what we really find in the data are confirmations of our prior beliefs about the 2018 midterms.

A wave is coming, and so far, there’s no good evidence suggesting that it is fizzling. If anything, it’s only getting (slightly) larger). We’ll have even more evidence come August 7 in the special election to fill the vacancy in Ohio’s 12th Congressional District — a seat that, at a 14-point partisan lean (just over half of the lean of PA-18 and AZ-08), will almost certainly be very competitive for Democrats.


Democrats could win the House majority by more or less than the 6-7% threshold that is forecast for them. This is just the point in my simulations of the election in which they are expected to win. That doesn’t mean that a Democratic Speaker is impossible at lower margins or a sure thing at a higher margin. [return]