ELECTION 2018: U.S. HOUSE OF REPRESENTATIVES FORECAST
Republicans 219, Democrats 216
The Biggest Electoral Surprise
In Two Years Ever
And all the reasons people will say they should have seen it coming
A week out from the midterms, the conventional wisdom that Democrats will regain the House majority has gone axiomatic. The party in the White House is always punished, after all, and Trump is simply too hot a mess to bear the brunt as well as his predecessors, much less to see his party outperform. Pollsters are consistently reporting a large Democratic advantage. And statisticians give Republicans’ House prospects only half the meager odds they allowed Trump on the eve of his election.
But if the surest bet in Washington since Clinton’s 2016 victory turns out to be wrong, and the GOP maintains unified control of Congress, both the punditry and financial markets will be caught even more off guard than they were two years ago.
We believe there’s a better than even chance that’s exactly what will happen. Before we get to why, let’s consider what happens next on Capitol Hill and on Wall Street if Republicans do keep a slim majority (and what happens if they don’t).
If Republicans keep the House…
Markets will roar. Not because Trump is a brilliant businessman or policymaker (evidence that he’s either is at best inconsistent). Not because trade wars or scandals are good for growth or stability. But because the policy trajectory that paved the way for 40% gains in equities over the last two years will be quite unexpectedly extended.
We’re likely to see another round of tax cuts (and the 2017 tax cuts made permanent); more regulatory relief for the financial services, health care, technology, and energy sectors; robust military spending; potential Obamacare repeal; and assorted pro-growth agenda items that leadership hasn’t bothered to seriously propose, because they too believe they’ll lose their gavels.
Assuming financial markets are pricing in the near certainty of a divided Congress, virtually none of this is currently reflected in asset prices, meaning a surprise next Tuesday could trigger a sudden and violent move higher, particularly in sectors that have done best over the last two years and/or stand to benefit from permanent tax reform and more regulatory rollback: health care, technology, energy, industrials, aerospace & defense, and financials.
If they don’t…
Should Democrats capture the House majority, we can expect to see most pro-growth items drop off the Congressional agenda, although this prospect can be presumed to be baked into current valuations. What’s not priced in (and likely won’t be for a while even after Democrats regain control, should they do so) is the potentially stratospheric policy uncertainty that would accompany a bonanza of investigations and impeachment proceedings.
For financial markets, Trump’s impeachment would feel a lot different from Clinton’s 20 years ago. For one thing, despite recently improved economic confidence, we’re a long way from the dizzily invulnerable sentiment of 1998-99, when Oval Office dalliances and the possibility of a lame duck Clinton being replaced by a lame duck Gore failed to rattle investors. Second, kneecapping Trump’s administration at the two-year mark (rather than the six-year mark) would render recent market-friendly policy gains quite vulnerable to unwinding. Third, while markets did well under Clinton, none of his policy achievements that helped usher in that growth (if we can say they did) were at risk of being halted or unwound, were he to have been removed from office. Fourth, conviction in the Senate (even under Republican control) and removal from office might be a more realistic prospect for Trump, given the remaining questions about what Mr. Mueller’s investigation might turn up, not to mention the – let’s say tenuous – degree to which Trump’s rhetoric, scandals, and general persona have engendered loyalty among Senate Republicans. Even with an extra seat or two added to the current razor-thin majority, it seems highly plausible that truly meaningful output from Mueller’s investigation could sway enough Republicans to convict. Fifth, there’s no reason to think impeachment efforts would necessarily stop with Trump. High-ranking Congressional Democrats have already explicitly targeted Vice President Pence and Justice Kavanaugh, and everyone from Jeff Sessions to Kanye West (the latter not technically impeachable) probably ought to brace for the same treatment. While a series of successful impeachments throughout the administration is exceedingly unlikely, a whole-hearted attempt is not out of the question. And that would be destabilizing in a way we haven’t seen and can’t reliably model. Sixth and finally, any impeachment proceedings against Trump, whether successful or not, would if nothing else achieve further fracturing of the Republican rank-and-file, meaning any pro-growth policy headway they’ve made or might hope to make in the next few cycles could effectively evaporate.
But good news: that’s probably not going to happen. And here’s why.
Blue Wave or Black Swan?
We’ll take it as a given that Democrats will fail to capture the Senate. The map is just too punishing to put the upper chamber in reach, at least this cycle. Instead, we expect Republicans to wind up with 1 or 2 (possibly more) extra seats. But is the House really equally out of reach for Republicans?
In a word, no. Our best guess is that Republicans will survive next week with a scant 3-seat margin (219 to the Democrats’ 216), with a national vote share of about 46.5%. We could be wrong of course, but we’re fairly confident this projection will prove closer to the actual result than those of oft-cited prognisticators.
Here, we’ll use Nate Silver’s 538 as a stand-in for electoral forecasters on the whole, since their projections are among the most comprehensive and quantitative (and they tend to track with the approximate sentiment of their peers). 538 gives Republicans a 15% chance of holding the House (as noted above, barely half the odds they gave Trump in 2016), with an expected haul of 202-205 seats.
That would amount to a swing of 35-38 seats, around 20% more than Democrats managed in 2006, against the aftermath of Hurricane Katrina, the height of Iraq War opposition, and the beginning of the housing crash. Indeed, such a swing would amount to a blue wave taller than any since Watergate.
That alone is reason enough to doubt the fait is 85% accompli, but here are a few more.
Reason 1: Prediction Markets
It’s worth considering how incentives drive the behavior and predictions of various election forecasters, among them pollsters, statisticians and other professional predicters, traditional financial markets, and political prediction markets.
Pollsters have some (imperfect) incentives to predict outcomes accurately (at least upon releasing their final polls, after all of the likely voter screens and final tweaks have been applied), but as we’ll discuss later, reliance on polls to signal electoral realities can lead to election night tears. And reliance on polls even a few days before an election can be significantly more misleading.
Dedicated political forecasters have a decent incentive to be correct, although media attention to specific forecasters tends mainly to reflect how right they were last time (or in some cases, how right they were the first time they made a name for themselves), which can lead to over-reliance on what might be more noise than signal. Out of 100 professional coin-flippers, three will turn up heads five times in a row, but beatifying those three and betting heavily that they’ll each flip heads again next time is plainly foolish. Press-fed forecasters may also be susceptible to a conscious or unconscious tendency to cater to the preferences of their media benefactors.
Traditional financial markets have very strong incentives to be correct in their forecasts, but a) the electoral signals embedded in asset prices are diluted by countless other factors, and b) market participants typically consume the same flawed inputs offered by pollsters, forecasters, and the congealed conventional wisdom the media derives therefrom.
Prediction markets, on the other hand (online exchanges where users can buy and sell the probabilities of electoral outcomes), offer pure electoral price signals, driven solely by expected gains and losses. Participants can still be swayed by the bad inputs noted above, but they do at least produce an isolated and purely financially-derived expression of the outcome they’re trying to measure.
The dominant political prediction market, PredictIt (by whom, in the interest of full disclosure, we are compensated as an outside consultant on market curation and settlement) does not give the edge to Republicans when it comes to keeping the House. But they show a much greater chance (ranging from 33-40% in recent days) than the commentariat.
Were it not for the consistently high confidence with which the pollsters and stat jockeys proclaim the House lost, we believe these market odds by now would likely have crested 50%. And if Republicans do emerge with a majority intact, you can expect next week’s pundits to be pointing at these prediction markets as the ones who got it – well, not right, but less wrong.
Reason 2: The Terrible Polls
While not terribly accurate, pollsters are relatively consistent in their inaccuracy. We make no assumptions about “skewed” polls, flawed turnout models, or politically motivated methodologies. We simply took a look at historical election outcomes versus generic ballot polling (gauging the candidate-independent national party preference) to see if it illustrates anything useful.
One week before the 2018 midterms, poll aggregator RealClearPolitics (“RCP”) showed a Democratic edge of 7.5 points. This still included a few (3 of 9) polls of “registered” voters, whereas the historical final aggregates tend to include only polls of “likely” voters (which typically skew more Republican than registered voters). This is one of the reasons that the generic ballot tends to shift toward Republicans over the final week of polling, by an average of 2.1 points over the last six cycles (and 3.1 points in 2016). If you take out the RV polls from the current aggregate, the edge drops to 6.8. Alternatively, if you assume a 2.1 point swing from the current 7.5 by next Tuesday, you get a Democratic edge of 5.4 points.
Further, while much has been made of the supposed “shy Trump voter” effect (Trump voters professing to support Clinton, to avoid divulging their sinful preferences even to strangers on the phone), we make no judgment about whether that factor led to polls underestimating Trump support, nor whether that factor remains. That said, for whatever reason, there is an observable historical outperformance by Republican congressional candidates, averaging 2.3% (actual national vote share vs. final polling aggregate) for all elections since 1994. The effect is even larger in midterm years, averaging 3.25% over the same period. What’s more, the “shy generic Republican voter” impact was larger than any “shy Trump voter” effect in 2016 (+1.7 points for congressional Republicans vs. +1.1 points for Trump).
For our projection, however, we ignore the shy voter impact and look solely at the final polling aggregate compared with actual seats won. Because we don’t yet have the final polling aggregate, we apply the average final-week Republican shift of 2.1 points to the current 7.5 to arrive at an estimated final Democratic edge of 5.4 points. (Note this is less than half the 11.5-point edge they enjoyed in 2006, which – it bears repeating – still didn’t produce a wave on the scale currently imagined by the forecasters.)
The charts below illustrate the Republican House majority (seats in excess/shy of 217) as a function of their final polling edge/deficit for every national election since 1994 (not cherry-picked because it was the year of the Republican Revolution, but because it’s the earliest year for which we could find comparable data) and just for midterms over the same time period. Given the disparity in polling accuracy between midterms and presidential years, and the likelihood that the peculiarities of presidential campaigns tend to significantly drive congressional outcomes, we focus on the latter.
It’s a small sample size, to be sure, but upon making the adjustments detailed above, the polling data as it stands now yields a best guess of 219 seats. Even if you discount the likelihood that RCP will see its typical final week shift, simply taking out the registered voter polls leaves you with a Democratic edge of 6.8 points, giving Republicans roughly 216 seats – one seat shy of a majority. Either way, it looks to be a jump ball, with the subtlest of advantages tilting toward Republicans.
Reason 3: It's still the economy, stupid.
Carville’s snowclone from 1992 remains evergreen. And if it truly is the economy, then Republicans should be optimistic, bordering on giddy.
To summarize the economic résumé of the 115th Congress and Trump’s first half-term:
Stocks are up 40%, even after the recent correction that followed a series of record highs.
GDP growth averaged more than 3% over the last four quarters (for the first time in three years). The last six quarters have each notched annualized growth of over 2% (for the first time in 13 years).
But the single economic statistic that appears most tightly correlated with Congressional elections is unemployment, with voters punishing the President’s party less amidst lower (and declining) unemployment rates. At 3.7%, unemployment has literally never been lower on the eve of a midterm. It’s more than two points below the long-term average and has declined a full point since year-end 2016.
Reason 4: Elections are correlated outcomes
Even if you don't buy the idea that Republicans are narrow favorites this year, it’s lunacy to think odds of their retaining the House are as low as 15%.
To stay above 217, Republicans need to hold all of the “leaning Republican” districts and virtually all of the nearly 20 true toss-ups. Returning to our professional coin-flippers, there’s a disheartening 0.0001% chance of flipping 20 heads in a row. But of course these toss-up races, unlike coin flips, are not independent events. Even if the chance of any one of them turning up R or D is indeed 50%, the odds of all or nearly all going the same way are not tremendously lower.
Just as bundling thousands of risky mortgages into a single security doesn’t effectively eliminate the risk they all go belly-up upon shifting macro circumstances or errors in assumptions that impact the components en masse, so too can all these races tip one way or another in the face of a single flawed assumption, a mild or unmeasured change in voter sentiment, or other broad-based fundamentals. If polls are off by a few points, if turnout models aren’t quite right, if enthusiasm gaps shrink, the entire battleground migrates and dozens of races jump from one column to the other.
Reason 5: The Partisan Wave Gap
If historical midterms represent an already limited number of data points, historical wave elections are even skimpier, being almost by definition highly infrequent events. Still, we can look back and make a few generalizations.
In the 2006 wave election, as noted, Democrats went into Election Day with an 11.5-point polling advantage. Bush’s approval rating was four points lower than Trump’s is now, Republicans were scrambling to either defend or distance themselves from the Iraq War, and it was generally agreed to be the worst environment for Republican candidates since Watergate.
Democrats gained 30 seats.
More of a swell than a tsunami, and more notable for the fact that it (briefly) changed the hands of the majority than for its size.
The last two Republican waves, by comparison, netted 54 and 63 seats (in 1994 and 2010, respectively). Those much larger swings came despite a lesser generic ballot lead in 2010 (9.4 points) and an outright deficit in 1994.
Are red tides inherently stronger than blue ones? The limited electoral data makes it hard to say with confidence, but the last time Democrats caught a wave as big as either of the two enjoyed by Republicans since 1994 was nearly 100 years ago, during the Harding administration, when they netted 77 seats (but not quite a majority). Even in 1930, a year into the Great Depression under President Hoover, Democrats swung only 52 seats, smaller than both modern Republican waves and one seat shy of a majority.
How confident are we that Republicans will prevail?
Not very. Could the Democrats triumph, coming off a 2-and-10 House record dating back to Clinton’s first term? Yes. Would it surprise us? Not at all. But suffice it to say drape-measuring is premature and ball-spiking is downright silly.
But whoever controls the House come January (and to pick on 538 one last time), neither the 35+ seat swing they peg as the expected scenario, nor the gargantuan 55+ seat swing they judge to be equally likely as Republicans holding the majority, seem realistic, in light of the latest polling, economic data, the judgment of disinterested speculators, and historical precedent.
Note: American Civics Exchange is not an investment advisory service. This material has been prepared for informational purposes only and is not intended to convey or constitute investment advice.