On November 6, all 435 seats in the House of Representatives and 35 in the Senate will be up for grabs. So will 39 state and territorial governorships, as well as plenty of state legislative chambers, not to mention local elected offices. For much of the year, markets appeared pretty unconcerned about the upcoming U.S. midterm elections. Lately, however, concerns over political risks have been growing. This has contributed to the weakness we have seen in global equity markets, and other risky assets. How justified are such fears? What may be the plausible outcomes and their market consequences? And, based on the data, how likely is it that the market could be caught by surprise this time around? In this note, we tackle each of these questions in turn. In conclusion, we give you our own best guess as to how things might turn out come November 6.
Superficially, you might think that there is not all that much to worry about. Looking back to all U.S. midterm elections since 1934, history suggests a fairly clear pattern of equity-market weakness in the months leading up the November elections. Typically, the S&P 500 swiftly recovers in the year that follows.
As usual, however, historic correlations tend to offer a less-than-perfect guide for forecasting purposes. First, we are not talking about all that many data points. That's a general problem when it comes to drawing any inferences from political events to developments in financial markets and the broader economy. A second reason is that the average hides plenty of variation from one midterm election to the next. To take the two most extreme examples, the S&P gained 43% in the 12 months following the 1934 elections, but lost 5% after the 1938 elections. Talking about the 1930s, there is a third concern and it happens to be the most relevant for our purposes. We would argue that in the present context, the historic correlation is likely to turn out downright spurious.
Think about why correlations have usually held in the past. Typically, the president's party tends to lose Congressional elections during the midterms between presidential elections. And, in general, equity markets in the United States and elsewhere tend to quite like a divided government. If neither of the two big parties is fully in control, it typically lacks the ability to implement big changes unilaterally.
Of course, there are exceptions to these two related rules. In 1934, two years into the first term of President Franklin D. Roosevelt (a Democrat), Democrats made unusual and large gains in the midterms, winning them a two thirds majority in the Senate. Conversely, Democrats saw sweeping losses in 1938, effectively marking the end of Roosevelt's domestic policy reforms. In between, there was a recession, starting in mid-1937. 80 years on, economic historians are still debating the merits of some of President Roosevelt's policies.
The point of the story? We still cannot really know what exactly the longer-term economic market impact would have been, had Democrats unexpectedly bucked the trend again in 1938. You never get to re-run an election as a controlled experiment. That hints at the dangers of glib assessments of the potential market impact of electoral outcomes. Including, as it happens, in the next section of our own report.
Among the theoretically possible outcomes three are plausible. We roughly rank them in descending order of probability based on our own assessment of publicly available and commonly used forecasting methodologies. The assumed market impact reflects the current assessments by experts across our platform, which happen to be fairly closely aligned with current conventional thinking on the likely market impact.
In terms of concrete policy-making, we believe the impact is likely to be fairly limited. Democrats would be able to block most new major legislation, in particular any further attempts to dismantle the Affordable Care Act (also known as ObamaCare), let alone entitlement reform more broadly. However, reversing past tax and spending initiatives would be unlikely.. Political pressures on the White House would intensify, with Congressional investigations into various allegations. Impeachment hearings would be a distinct possibility. More immediately, there would probably be some erosion in President Trump's political clout.
For markets, the most obvious question should be how voter rejection might impact the administration. Might the White House take on a more measured tone in global trade conflicts and perhaps try to de-escalate geopolitical tensions? Or conversely lash out even more harshly at foreign foes, real and perceived? Based on the developments of the past two years, we unfortunately fear that the latter is more likely, especially on trade. Mostly, though, we would probably see purely political concerns to recede somewhat, with investors focusing more on other issues, such as the future course of monetary policy.
The immediate policy-making implications would probably be similar to Democrats just taking the House. A majority in the Senate, no matter how narrow, would likely enable them to stall judicial nominations and hamper additional deregulatory measures, for example by blocking cabinet appointments. Erosion of President Trump's political clout would be more severe and discussion of impeachment more likely. However, any such measures could risk backlash, similar to what happened after the 1994 Congressional takeover by Republicans. In the longer term, there would also be a risk of U.S. politics becoming even more contentious. On a more positive note, a chastised White House might embrace a more bipartisan approach to governing. One area to watch would be infrastructure spending.
A lot would also depend on what happens in other races, further down the ballot. For example, losing a string of governorships in swing states such as Florida, Nevada and Wisconsin would hurt Republican morale and party cohesion. High-profile losses in even redder states, such as Georgia, Ohio and perhaps even Kansas and Oklahoma could make matters worse. All these races are surprisingly close, which also augurs well for Democratic prospects in state legislatures. In turn, these state-level races will largely determine Congressional maps following the 2020 United States Census. This would strengthen Democratic prospects of holding on to their majority in the House, after the next round of redistricting in 2022. (Partly because of Republican control of many states and the gerrymandering of both parties, the current map of House districts is very unfavorable to Democrats. According to most estimates, Democrats need to win the popular vote by about 5% to 7 % to win a majority in the House of Representatives.)
Unlike the other scenarios, a fizzling of Democratic support could have immediate policy-making implications. Most likely, we would see additional legislative initiatives, notably an attempt to make recent tax cuts "permanent" and to index taxable capital gains to inflation. Republicans may most likely also try to, once again, repeal the Affordable Care Act. They might even attempt to tackle welfare and entitlement reform more broadly, by cutting funding for Medicare, Social Security and Medicaid. However, narrow majorities in Congress would make significant changes to entitlements much more difficult. (The 2020 elections would loom large over any proposals to cut funding for Medicare, Social Security and Medicaid or perhaps trying to reform them by assigning a larger role to the private sector. Any attempt to privatize Medicare, Social Security and Medicaid would almost certainly require at least 60 votes in the Senate. Passing such proposals looks highly improbable, not least as these programs remain very popular, even among Republican voters and Senators.)
In the short term, U.S. equity markets could perhaps benefit from renewed hopes on taxes and deregulation. In the longer term, the risk of the U.S. economy overheating would increase. For the world economy and global financial markets, the rise of President Trump's political clout could increase the risks of further trade escalations. It may also contribute to the further rise of U.S. Treasury yields, and a worsening fiscal outlook, making the task of the U.S. Federal Reserve (Fed) of engineering a soft economic landing even trickier.
As in 2016, a lot depends on how investors and policy-makers react to the results. Indeed, over the past two years we have seen on a number of occasions that predicting electoral surprises is sometimes the easy part. More often than not, the harder challenge is foreseeing what markets perceive to be surprising and how markets react to real or perceived surprises.
In our view very likely. Indeed, we do not share the confidence of many market observers and professional forecasters in the high probability of the fashionable base-case assessment - Democrats retaking the House and Republicans holding the Senate. Nor do we believe that the market consequences would necessarily be benign. Instead, we think that this base case is less likely to materialize than standard forecasting models suggest. And even if it does, the market implications could still be quite significant.
Partly, that reflects our experience in analyzing elections since 2016 and the resulting market fall-out of political surprises. The dirty little secret of forecasters like ourselves is that while it is not all that hard to predict what voters might do and what the longer-term economic consequences might be, the same cannot be said for market reactions in the weeks, months and years that follow. When it comes to politics, it turns out that financial markets are far from perfect in discounting upcoming risks. To take just one recent example, it was statistically quite plausible for Italy's populist Five Star Movement and the right-wing Lega to win a majority in one or both houses ahead of this year's elections on March 4. As we wrote at the time , "On our reading of the admittedly ambiguous data, we believe the anti-establishment Five Star Movement may only be a normal polling error away from being able to block the formation of any realistic, broadly centrist, pro-European government."
Six months on, market participants are still trying to come to terms with the impact the new, populist Italian government is having in trying to push the limits of the Eurozone's fiscal deficit rules. In part, this is because a lot depends on what policy-makers make of an electoral mandate and what constraints others impose on them. That, in turn, partly depends on the specifics of what voters actually decided on Election Day. Glib assessments of likely scenarios can at best offer a usual starting point.
To see what that might mean for the midterm elections, let's just focus on our Scenario 2.a. of Democrats retaking the House and Republicans holding the Senate. U.S. elections are quite unusual in that you have a wealth of historical data, plenty of high-quality pollsters and a vibrant community of data-driven analysts and commentators. A data-driven, probabilistic approach to predicting and analyzing electoral outcomes was pioneered by Nate Silver and https://fivethirtyeight.com/. Quite a few others have embraced a similar approach. That is one of the reasons covering U.S. elections is both easier – because we have to do less number crunching ourselves – and more fun – because there are plenty of knowledgeable commentators to learn from.
We have some minor quibbles with the FiveThirtyEight probability assessments, which we will turn to in a minute. But even taking it at face value, it hints at plenty of underlying uncertainty for investors to fret about. That is especially true for the Senate. According to the model, there is currently a 5-in-6 chance of Democrats winning a majority in the House, but only a 1-in-5 chance of gaining control of the Senate. Moreover, the range of plausible outcomes with an 80% chance of materializing is somewhere between plus four net gains for Republicans (giving Republicans a 55-to-45 seat majority) and plus two net gains for Democrats (i.e. the narrowest of majorities of 51 Democrats versus 49 Republicans in the Senate). Conversely, the model suggests a less-than-10% chance each of Republicans winning 55 or more seats, and of Democrats winning 52 or more seats.
In order to think about the potential market implications of Scenario 2.a, let's zoom in on an even narrower range: Republicans holding somewhere between 50 and 54 seats, while losing their House majority. Depending on how closely elections in the House and the Senate turn out to be correlated, there is something like a 3-in-5 chance of falling into that range, according to the FiveThirtyEight model. That 3-in-5 chance, however, hides a pretty wide range of political realities, to which investors are likely to react very differently, come November 7.
To take the edges of the range, imagine that Republicans gain 54 seats, i.e. succeed in holding on to all their current seats and defeating 3 vulnerable Democrats. Heidi Heitkamp in North Dakota, Claire McCaskill in Missouri, and, in one of the night's genuine upsets, Robert Menendez in New Jersey, say. Other incumbent Democratic Senators, including Joe Donnelly in Indiana, Bill Nelson in Florida, Jon Tester in Montana and Tina Smith in Minnesota, survive, but by far narrower margins than expected. Democrats win a very narrow majority in the House. Down-ballot, their wave fizzles. In particular, Republicans manage to hold on to most of the vulnerable looking governors' mansions, including Wisconsin, Georgia and Ohio. President Trump promptly takes credit and pledges to pursue his "America First" agenda with renewed vigor, especially on trade. Bond investors become increasingly concerned over rising deficits (as a result of making tax cuts "permanent"), without any meaningful spending cuts. After some initial relief that the vote is over, equity markets take fright. Presidential attacks on the Fed continue, prompting market fears that the Fed might tighten faster than expected, simply to defend its credibility.
Contrast that with a scenario in which Democrats make a net gain of just one Senate seat (Arizona, say), exceed expectations in the House and run the table in down ballot races, including some high-profile governorships. President Trump immediately disowns any responsibility. In the weeks and months that follow, his clout continues to diminish. With the Senate evenly and Vice President Mike Pence acting as the occasional tie-breaker for judicial nominations, legislative progress stalls and government shutdowns loom. A growing number of Republican Senators start to call for bi-partisan compromises, if necessary in defiance of the president. In particular, a group of moderate, pro-market Republicans led by former presidential candidate and newly elected Utah Senator Mitt Romney, make it clear that attacks on the Fed and the World Trade Organization need to stop. In private conversations with the White House (which are inevitably leaked), they make their support of judicial nominations conditional on an end to trade tensions. Markets react with modest relief. Investors mostly focus on other issues, such as the future course of monetary policy. Markets increasingly start to see Trump as a lame duck unlikely to get much done in the remainder of his term.
Such thought experiments highlight that a lot can still go wrong – or right – after November 7. Moreover, the Senate forecast appears quite a bit less certain than many think. By now, you have probably read – perhaps ad nauseam – that the 2018 Senate map favors Republicans. That is both true, and potentially quite trivial for forecasting purposes.
To see why, remember how in 2016 the Electoral College was supposed to provide Hillary Clinton insurmountable advantages? That was correct in the limited sense that using the 2012 presidential-election results as your starting point, Clinton appeared to have plenty of routes for a majority. Her floor in the Electoral College also appeared to be quite high. By contrast, there were only a comparatively limited number of paths to victory for Trump. Statistically speaking, something quite analogous might turn out to be true for the 2018 Senate map. It all boils down to how closely key races will turn out to be correlated. Call it the Trump effect 2.0, only with the respective roles of the two parties potentially reversed.
So, the next time someone mentions how unfavorable the 2018 Senate map is for Democrats, think back to 2016, or even better 2012. That's when all current incumbent Democratic Senators last ran, apart from Tina Smith in Minnesota. They all won, in an environment far less favorable for Democrats than the current one, according to national polling throughout the 2018 campaign so far. To be sure, the race is likely to be nationalized far more than in 2012, when local issues and candidate personality played a larger role than in is likely to be the case this time around. However, it is far from clear that this will necessarily hurt the Democrats' Senate prospects.
Here is another sign of how uncertain Senate forecasts still are. Forecasting models for the Senate have swung quite sharply in recent weeks. Usually, that happened in response to a handful of surveys, often states such as North Dakota that are both rarely polled – and difficult to poll. (As a result, pollsters in such states have historically tended to get it wrong more often, and by wider margins than you commonly see in U.S. polling.) In itself, this is by no means intended to criticize the models. Instead, it suggests that the models are working reasonably well in trying to extrapolate as much information as possible from relatively little and potentially dubious data.
Indeed, we think it is more likely than usual that even polling averages of high-quality surveys might be off somewhat nationally. Needless to say that could have an even bigger impact on the probabilities of upsets in the Senate races everyone is watching than in the House. This is not, we hasten to add, because there was anything particularly wrong with national U.S. polling figures in 2016. U.S. pollsters suffer from the same challenges, notably declining response rates, as their peers encounter around the world. They generally manage to cope rather better, in part by having gotten quite sophisticated in how they weight responses. This tends to work surprisingly well – as long as the recent past offers reliable pointers as to what the electorate in an upcoming vote might look like demographically, in terms of its geographic split and partisan lean. We believe these conditions are less likely to be met this time around, mainly because of how both parties have changed since 2016.
Statistically, this suggests polling errors of unknown magnitude and direction. In other words, we think that unexpected outcomes are a touch more likely than models based on polling and correlations with historically reliable indicators would suggest. This could mean Republicans will be able to hold on to the House. Or it could mean even larger gains by Democrats, including winning the Senate. In such instances, we have found it quite useful in the past to consider analogous electoral dynamics in other countries and measures of voter engagement, as well as the polling data.
Just as President Trump is different from previous Republican presidents, the same might well turn out to be true of the electorate at this year's midterms. Demographically, Trump's core base of supporters looks less like yesteryear's reliably Republican midterm electorate and more like typical protest party voters long familiar in Western Europe. If so, there are a couple of reasons to expect that polls might either overstate or understate their electoral strength. On the one hand, pollsters have an unfortunate tendency to overcompensate for past failings, whether real or perceived. Typically, this leads polling to somewhat underestimate protest parties when they first surge – and to overestimate their staying power in subsequent elections. On the other hand, there can sometimes be a bit of a bandwagon effect spanning electoral cycles. Once a protest party comes into power and starts to deliver things its core supports long wanted, this can strongly motivate its supporters to mobilize, often late in campaign. Word-of-mouth can prove especially effective in such situations – and difficult to see hints of in the polling data until the very end.
The same also tends to happen on the opposing side, though usually with more advance warning in the polls and other data. Indeed, there has been plenty of evidence of grass-root activity likely to benefit Democrats up and down the ballot (from Women's Marches and teachers protesting low pay in several Southern states to students organizing in favor of gun restrictions). State and special elections since 2016, including the Alabama Senate and the Virginia's governor's races last year, also offer pointers that polling might, if anything, understate the size of the Democratic wave in 2018. This would be consistent with fundraising data, which has been extremely strong for Democrats across a wide range of low-profile races – and pretty dismal for Republicans. Add other measures of voter engagement, and it would be quite surprising even to us if Republicans held on to the House. Take it all together, and we would be somewhat less surprised, if Democrats gain another seat or two, in the Senate, as well as larger than expected gains in the House and lower down-ballot. Both scenarios, though, are hardly unthinkable. Risk-averse investors might be well advised to consider them carefully instead of simply assuming that political risks will swiftly fade again, once the voting is done.
Sources: Bloomberg Finance L.P., DWS Investment GmbH as of 10/19/18
* Daily S&P 500 performance, indexed to midterm-election date dating back to 1934.
Sources: Bloomberg Finance L.P., DWS Investment GmbH as of 10/19/18
1. In the U.S. context, "divided government" means that the two houses of Congress are held by different parties
2. Nevertheless, Bill Clinton was able to secure a second term in 1996.
3. Needless to say that the forecasts might have moved by the time your read this. You can follow the FiveThirtyEight Senate forecast here: https://projects.fivethirtyeight.com/2018-midterm-election-forecast/senate/?ex_cid=midterms-header. Their House forecast is available here: https://projects.fivethirtyeight.com/2018-midterm-election-forecast/house/?ex_cid=rrpromo. And their recently added Governors forecast here: https://projects.fivethirtyeight.com/2018-midterm-election-forecast/governor/
4. For description of the FiveThirtyEight Senate model, see: https://fivethirtyeight.com/methodology/how-fivethirtyeights-house-and-senate-models-work. The latest results: https://projects.fivethirtyeight.com/2018-midterm-election-forecast/senate/?ex_cid=rrpromo. A new, alternative modelling approach by the conservative Weekly Standard addresses some of the methodological issues slightly differently: https://www.weeklystandard.com/david-byler/introducing-the-tws-swingseat-model. It comes to similar results: https://www.weeklystandard.com/swingseat-senate-election-model.
5. For good overview of what really happened in 2016, see Nate Silver, "The Real Story Of 2016: What reporters — and lots of data geeks, too — missed about the election, and what they’re still getting wrong." Jan. 19, 2017, https://fivethirtyeight.com/features/the-real-story-of-2016/
6. For a nice illustration of how this works in practice, in a difficult to poll state like Minnesota, see Nate Cohn, "Did Minnesota’s Eighth District Really Swing by Almost 20 Points? An apparent drastic shift toward the Republican candidate highlights the challenges of polling generally, and of polling in certain states particularly." Oct. 15, 2018, New York Times, The Upshoot, https://nyti.ms/2QRDdoT
7. For the latest figures and their potential implications, see Nate Silver, "Election Update: Democrats’ Unprecedented Fundraising Edge Is Scary For Republicans … And Our Model", Oct. 19, 2018, https://fivethirtyeight.com/features/election-update-the-democrats-unprecedented-fundraising-edge-is-scary-for-republicans-and-for-our-model/
8. In the house, gerrymandered districts would probably begin to backfire on Republicans, if the Democratic wave turned out to be even larger than expected. For an illustration of what might happen, see Nathaniel Rakich, "A Big Blue Wave Could Overwhelm The GOP’s Advantage In The House", Oct. 9, 2018, https://fivethirtyeight.com/features/a-big-blue-wave-could-overwhelm-the-gops-advantage-in-the-house/