The Ephemeric

Two weeks ago The Ephemeric spoke about the upcoming United States Congressional Midterm elections and made an argument for why participation is so important. Now with just a week of the campaign remaining, I wanted to switch things up a bit and turn our attention to election day itself. What can we expect, and what does it all mean going forward?




Introduction
So what will be the story of the 2018 elections? Well the expert opinion has been pretty unequivocal: they are expecting a blue wave. Donald Trump is historically unpopular, and the expectation is that voters will overwhelmingly back his opposition, potentially handing the Democratic Party Congressional majorities for the first time since 2010. Indeed polls have suggested an overwhelming Democratic advantage ranging from the high single digits to mid double digits. By comparison, Obama's landslide 2008 victory which resulted in a historic filibuster-proof Congressional majority was won with a margin of only 7%. Polls today suggest a considerably larger margin of victory in 2018, so surely the Democrats should sweep the elections?

As much as I would like this to be true, I'm going to pour some cold water on that consensus. As you will see, the Democrats are still very much a long shot to take back Congress despite their strong edge in the polls. As contradictory as this might seem, the US electoral system is set up (by design) so that the party which receives the most votes does not necessarily win the most seats, as we will explore in greater detail.

Today I will be taking an in depth look at the House and Senate separately. But before we get into that let's quickly go through the golden rules of midterm elections, which apply to both the House and Senate elections and are vital towards understanding the dynamics at play this year.

1. Midterm elections generally show lower voter turnout than Presidential elections. Last week I briefly touched on the reasons why this might be the case, but in essence it's because people just care less. Rightly or wrongly people just see the President as a more important and more glamorous role. Everyone knows who he is, he's a celebrity. By comparison, very few Americans can even name their Congressional representative.

2. Low voter turnout invariably favours the Republicans. This one might seem less obvious, but statistically it is undeniably true. We could spend an entire article discussing the reasons why this might be the case, but most experts will agree that it comes down to something quite obvious. In a low turnout year where people are less motivated to vote, the most likely people to still show up and vote are a) those who care the most (ie those who are most switched on to the latest political happenings, spend more time watching cable news, etc) and b) those with the easiest opportunity to vote (ie those with more free time on a Tuesday). So who watches a lot of cable news and is less likely to have no work or other commitments on a Tuesday? The elderly, and low-education voters, two voting blocs who have very heavily backed Republicans in recent years.

3. Midterm election turnout almost always favours the party in opposition. This is another fact that is very clear in the data, the party which holds the White House almost always comes off worse in the midterms. It makes perfect sense really. As the Republicans have so ably shown, fear and anger is a great motivator to vote, and the party out of power is invariably the angrier - although you can understand now why Trump and his accomplices have been so keen to dial up the anger and hate among their base.

So from the above dicta, you can see why the Republicans have generally done very well at the last few midterms under Obama, despite losing both Presidential elections quite handily. The demographics of a midterm election clearly favoured the Republican Party, as did the fact that they were the opposition party at the time. This time, however, things are slightly different. They still have the demographic advantage, but they are no longer the opposition. With these two factors at odds, you might expect that in an ordinary year the polls would reflect a stalemate, or even a slight Democratic advantage. But that's not what we're seeing in the polls at all.

This year there is the additional X-factor of Trump. As we have said, Trump is historically unpopular this year, and that is serving to increase the Democratic advantage considerably, to the point where many states have been reporting turnout estimates comparable to those of Presidential elections. This is both remarkable and potentially devastating for Republicans.

With this additional factor, the advantage is very clearly leaning in the Democrats' favour. And yet, as we will see in the following analysis, even a significant lead in the polls is no guarantee of electoral success this year.

Now let's give you the summary verdict that you came here for: The Ephemeric predicts the Democrats to retake the House of Representatives, and the Republicans to hold the Senate.


House of Representatives Verdict: Democratic Majority

Current House Map: Republicans - 235, Democrats - 193.
Predicted House Map: Republicans - 205, Democrats - 230.
Approximate Net Change: Democrats gain 30 seats.

The above map is based on data from a variety of aggregators, including Pollster, Fivethirtyeight, and analysts including Sabato and Cook Political, and shows the expected House map. The rest is pretty self explanatory: dark blue represents safe Democrat wins, light blue leans Democrat, grey is toss up. Meanwhile on the other side, light red to dark red represents lean to likely Republicans.

So first thing is first. You are probably looking at that map and thinking, "Wow that's a lot of red, and here I thought the Democrats were having a good year!" A reasonable thing to think, but look closer and you will see that most of that red belongs to just a few very large seats, covering vast tracts of rural land where very few people actually live. If you zoom in to the denser population centres you will see dozens of much tinier, but far more populated blue districts.

In actual fact, this represents a very bad map for Republicans. How bad? Well based on the above data, even if Democrats only win the districts highlighted in blue that already lean in their favour, and lose all of the toss-ups, they will still win the House. And of course, the size of their popular vote margin is so significant that you would probably expect them to win a lot of the toss-up and even a few of the lean Republican seats too.

So the polls are pretty unambiguous, the Democrats have the advantage, and a pretty big one. At the time of writing, the most recent average shows Democrats with a 9% margin, and there are several polls showing leads in the double digits. It can not be overstated just how astonishing a margin this is. For comparison, Obama's landslide victory in 2008 was with a 7% popular vote margin, his comfortable re-election in 2012 came with just a 5% margin. American politics is very evenly matched and elections are typically won by just a low single digit margin. A high single digit or even double digit margin is rare, and in an ordinary year would suggest a historic blowout.

Unfortunately, such a historic blowout is highly unlikely despite the significant polling lead. This is due to a little thing known as gerrymandering.

Gerrymandering, in a nutshell, is the process through which partisan actors draw up the borders of Congressional districts in such a way that voters more likely to vote for your opponent are pooled into as few districts as possible, while your voters are spread into as many districts as possible whilst still maintaining a lead in those districts. This process effectively allows you to increase the number of seats you win, even if you don't increase the number of votes you get. It results in some bizarre and shockingly manipulative district boundaries. For lack of a better description, it's a legal way of rigging a democratic election. For further clarity on how this trick can be used to manipulate election results, please see the diagram below.



So how bad is the gerrymandering problem? Well if that North Carolina example above doesn't make it abundantly clear, it's pretty damn bad. It's bad enough that even with the Democrats' 5% winning margin in 2012, and a 2% winning margin in 2016, they still only won a minority in the House. It's bad enough that analysts expect that Democrats would need a winning margin of at least 7%, itself a rare and historic margin, just to have a shot at a majority. So as you can see, with the current expected winning margin of about 9%, Democrats are still only just favoured to win a House majority.

Now here's the silver lining for Democrats. As you may have surmised from the above examples, gerrymandering can be used to take a region that leans Democratic and create a lot more Republican leaning districts out of it. The trade-off is that while there will be a lot of red districts, by their nature a huge number of these will have only a slender Republican lean, since the process requires their voters to be spread out over multiple districts. This means that, while the threshold for a Democratic majority is higher, once this threshold has been met suddenly a huge number of seats becomes winnable. So to put this into numbers, while the difference between the Democrats winning by 2% or 7% might be pretty small, maybe 10 seats, the gains become exponential after that, potentially an additional 10 seats for each percentage point. So while the current data suggests a gain of maybe 30 seats, if that data underestimates Democrats by even just 2-3%, we could be looking at staggering gains of 60-70.

So as you can see, this is a challenging election to forecast. The unique combination of extreme gerrymandering and Democratic enthusiasm means that even a change in polls of a percent or two can potentially be enough to change a dozen seats. In my view, the key takeaway is that despite the hype, the House is far from a slam dunk for Democrats. Thanks to gerrymandering there is a significant threshold for them to cross before they really become competitive.

Based on the current data, Democrats will most likely win the House, but it probably won't be the historic majority that the polling or punditry suggests. They need 25 seats for a majority, and my expectation is that barring a polling error, they will probably only get a little more than that amount, with an approximate gain of 30 seats.


Senate Verdict: Republican Majority


Current Senate Map: Republicans - 51, Democrats - 49.
Predicted House Map: Republicans - 52, Democrats - 48.
Approximate Net Change: Republicans gain 1 seat.
Key states to watch: AZ, FL, IN, MO, ND, NV, TN, TX, WV

So if the House paints an optimistic, yet cautious picture. The Senate is a very different story, and despite the Democrats' significant national advantage, I actually expect them to have trouble maintaining even their current status in the Senate, and could potentially lose seats.

How can that be possible when they have such an advantage this year? It's an issue with term lengths. Whereas every single seat in the House is up for election every cycle, Senators get to serve 6 year terms, meaning that only a selection of seats are actually up for election each cycle. 6 years ago was 2012, a good year for Democrats in which Obama won re-election, and the Democrats increased their then Senate majority by 4 seats. This means that Democrats will be defending a lot of seats this year.

So how bad is it? It's very bad. Democrats will be defending 26 seats this year, compared to just 9 seats being defended by Republicans. Even worse, Democrats will be defending a number of seats in deep red states like Missouri, West Virginia and North Dakota, states which reliably voted Republican even in Obama landslide elections. So not only do Democrats need to defend these extremely difficult seats, they then need to try and pick up at least 2 Republican held seats, which are also exclusively in deep red states this year.

It is a daunting Senate map for Democrats this year. In a normal year where the polls are roughly even, you would expect Republicans to dominate this map, maybe pick up 4-5 seats. But again, this is not a normal year, it is one in which the Democrats have a significant nationwide advantage. The fact that we are even talking about the Senate elections this year as competitive is an impressive achievement for the Democrats, and shows how successful they have been at expanding their map.

The media narrative has been one of a Democratic wave this year, so it's important to set appropriate expectations. Democrats should have no business being competitive on this map. They should have no business winning elections in West Virginia and Indiana in 2018, especially in a midterm year where the voter demographics inherently favour their opponents already. So while 2018 may very well be a strong Democratic year with a significant national advantage, the chances of them taking a Senate majority based on this map are exceedingly slim, and even maintaining the status quo would be a remarkable victory.

Of the 35 seats up for election this year, there are 9 key races that will determine the outcome. These can broadly be divided into three categories: the "pure toss-ups" which are so close that they can't be called one way or the other, the "lean" states which are highly competitive but clearly leaning in one direction, and the "likely" states which are still considered winnable but can be predicted with relative confidence. These states are categorised as follows, with the current seat holder indicated in parentheses:

• Pure toss-ups: Missouri (D), Nevada (R), Tennessee (R) 
• Lean states: Arizona (R), Florida (D), Texas (R)
• Likely states: Indiana (D), North Dakota (D), West Virginia (D)

So let's look at these in reverse order, starting with the likely states.

Likely states
Indiana, North Dakota, and West Virginia: these are all deep red states held by Democrats. In a normal year you would expect these to be easy pick ups for Republicans, especially in a midterm year, and yet Donnelly and Manchin appear to hold quite comfortable and consistent leads in the polls in Indiana and West Virginia respectfully.

The same is not true of Heidi Heitkamp in North Dakota. Now North Dakota is a weird state. Very Republican, but also very small and therefore much more locally focused. This means that when it comes to state and local elections, partisanship has historically not mattered as much as personal brand, and Heidi's brand is so strong that even her opponent needs to start speeches with, "Look, I like Heidi, who doesn't like Heidi?"

On top of this, the polling out of North Dakota has been very poor, with only two polls in the past month and a half, both done by the same pollster, and a partisan (Republican operated) pollster at that. Being a partisan pollster doesn't necessarily mean you should ignore the result, but it does suggest you take it with a pinch of salt. Given the uncertainty, I was hesitant to put North Dakota in this category, but have ultimately done so based on a convergence of factors: the older (higher quality) polls showed a small Republican lead, and since that time polls in other red states have, if anything, shown movement away from the Democrats rather than towards them. Leaked internal polls also seem consistent with the notion of a moderate Republican advantage. Add to this the fact that we have seen a marked increase in the nationalisation of other small-state elections this year, and it all points to an outcome where a Republican victory seems very credible, and the most likely outcome here.

Lean states
Arizona is looking like something of a gift for Democrats. This is the election to fill the seat vacated by retiring Republican, and consistent thorn in Trump's side, Jeff Flake. It's a Republican-held seat in a pretty red state that has consistently voted Republican for years, and yet the polling has been very strong for Democratic challenger Kyrsten Sinema. This is actually not as surprising as it seems. Trump is pretty unpopular in Arizona, despite its reputation as a red state, and Hillary actually came surprisingly close to winning there in 2016. Given the swing towards the Democrats this year, a moderate advantage for Sinema is more or less what we would expect to see here.

Florida is a state that until recently was considered to be a bit of a toss-up, and even a good pick-up opportunity for Republicans. Bill Nelson is your classic establishment Democrat after all, his challenger Rick Scott is a (strangely) popular former Governor of the state, and Trump's support in-state proved surprisingly robust during 2016. Now it looks like the race is starting to get away from them. Quite why this is the case is open to interpretation; perhaps accusations of corruption against Rick Scott are finally gaining traction, or perhaps it's a knock-on effect of Andrew Gillum's wildly enthusiastic run for Governor firing up the base. In any case the polls of late do seem to show races in this swingiest of swing states getting more and more comfy for Democrats.

Texas is an interesting one. Typically considered the deepest of deep red states, Texas polls have been surprisingly close, with many showing the Democratic challenger Beto O'Rourke within the margin of error, particularly in polls of registered (rather than likely) voters. Texas has long been discussed as an increasingly purple state due to its high Latino population, as well the increasing influence of its young and educated urban centres (both of which heavily vote Democrat). I'm not sure we're seeing any particular evidence of this yet, and rather the closeness of this race appears to be down to the candidates themselves. Ted Cruz is notoriously unpopular, both at home and in Washington, whereas Beto's campaign seems to have given us the next big political superstar, smashing fundraising records and earning comparisons to JFK and Obama. Ted Cruz remains very much favourite in the polls, but here's the thing: polls of all voters are actually showing Beto in the lead or tied, it's only the likely voter models that give Ted Cruz the advantage. In other words, this comes down to turnout and enthusiasm. If Beto can turn his hype into actual votes, he still has a shot.

Pure toss-ups
Missouri is the quintessential toss-up state this year. A Democratic incumbent in a traditionally red state, who in an ordinary year should have no chance of winning. And yet polls have been showing McCaskill slightly ahead or tied with challenger Hawley. This is a difficult one to call. With the race essentially tied, my instinct would be to give the advantage to the Republican due to the partisan lean of the state and the inherent turnout advantage in a midterm election. But at the same time, McCaskill has been in this position before, in much less favourable political years, and still managed to win. For one reason or another, she typically does outperform her polls and grind out the result, so if on election day the polls show her as tied or narrowly leading then I suspect she will again. Ultimately though this race could go either way.

Nevada is a peculiar state. Traditionally red, though it did vote twice for Obama, and even went for Hillary during the 2016 election. One would therefore expect that in a year which has swung decisively towards the Democrats from 2016, they would be favoured to win. But this is still a red leaning state, and the Republican has an incumbency advantage. The result is a race where polls show the contest essentially tied, with perhaps a slight Republican lean. It is worth pointing out that in 2016 polls did show a similar Republican advantage, but then voted comfortably for Hillary, and the same happened for former Senator Harry Reid in 2010. This is a state with a track record of underestimating the Democratic lean in polls, and so despite the Republican polling lead, I consider this state very much a toss-up.

Lastly, Tennessee. It might seem strange to have a deep red state like Tennessee listed as a toss-up, but that's undeniably what the polls have been showing. In Phil Bredesen, Democrats have a very popular former Governor of the state who has never lost an election in Tennessee. But this is still a deep red state, and Blackburn has sensibly sought to nationalise the race in order to make Bredesen pay for his Democratic affiliation. Whether it will work is anyone's guess. Bredesen has led in most polls this cycle, including the most recent poll from SSRS, but following the Kavanaugh confirmation there were some polls showing quite decisive Republican leads. As with North Dakota, the problem here is that there have been very few polls in recent weeks, meaning that any prediction necessarily entails a lot of guesswork. Nevertheless, the Bredesen lead of recent high quality polls, coupled with the local dynamics of this popular candidate, lead me to believe that he has at least as much a shot as Blackburn here. This is a true toss-up.

Conclusion
Now if you were to look at each of these states individually, and tot up the results of my predictions above, you would get to a stalemate 51-49 Republican advantage. And yet I have not predicted this result, but a +1 gain for Republicans. The reason for this is simple. Senate polls are often wrong, particularly in states where there is very little polling (see Tennessee and North Dakota in particular). The fact is that there is a greater than likely chance that some of the current state predictions will be wrong, and so the question becomes which side has the greatest potential upside in the event of polls being off.

The crucial point here is that there are a lot of states where Democrats look likely to win by very slim margins. This means that if we assume a polling error in line with the historical average, an error in favour of the Democrats would not likely flip any additional seats towards them, whereas such an error to the Republicans' favour would probably flip at least 1 that we have currently predicted as a Democrat win. In making this forecast, I have attempted to predict not just what I think will happen in each key race, but where I think the likely polling errors could be. All put together, it depicts an election where the advantage lies with the Republicans, but probably less than it should be considering the favourable map this year.

So based on current data, and taking into account historic polling errors and the current expected margins, our expectation is that the Republicans are most likely to gain seats in the Senate. Given the closeness of many of these races, there is a lot of room for error in this prediction. We could feasibly end up with anything from +4 Republicans to +2 Democrats. But on the balance, +1 Republican seems most likely.

So there it is. An expected Democratic gain of around 30 seats should give them a House majority, while a favourable Senate map means Republicans are looking far more comfortable there, with an expected gain of 1 seat, and a retained Majority. What does this mean going forward? We'll come back to that after the election, and give you a preview of what the remaining two years of Trump's term are likely to hold.