We have long been fans of the Good Judgement Project, an initiative led by Philip Tetlock and Barbara Mellers at the University of Pennsylvania that has regularly beaten even intelligence analysts with access to classified data when it comes to forecasting geopolitical events.
As a team, we have always been very interested in the topic of forecasting and you may of course think it’s our bread and butter, but we recognise it is not an easy task. We spend considerable time running scenarios for political risk events, macro outcomes and tail risks, as good forecasters are obliged to consider different scenarios. Our team motto is ‘prepare, don’t predict’. We have learnt to quantify our forecast likelihoods, and to update those likelihoods frequently and incrementally, with new information or simply as time passes.
The Project’s methods teach us how to become better forecasters ourselves, and this year they have released some time-series predictions from their network of ‘Superforecasters’ on a number of COVID-19 related economic topics. These have helped us navigate the shifting timeline probabilities for vaccines and the likelihoods of fiscal stimulus in the US.
The US election obviously poses our next challenge. It’s interesting to note the gap between the Superforecaster probabilities of a Democrat win, at 83% at the beginning of this week, and the betting odds that continue to suggest a somewhat closer race.
Fireworks on the fourth or fifth of November?
Despite the wide gap in polls and a high apparent likelihood of a Democrat victory, we believe this race will feel tight until the night of the election. Due to an increase in postal voting and tensions between the parties, we see a high probability that no party will concede the election until sometime after the vote on 3 November. We believe the chances of an official result by 4 November are below 50%.
Markets are also showing an elevated risk of a contested election, with expected volatility for November unusually elevated, even for an election year. But overall, we believe investors will price a likely winner well before either candidate concedes the election.
So what can we do to prepare for the big day? The events around President Trump’s COVID-19 diagnosis earlier this month demonstrate our process in action. We didn’t know exactly how the president’s health would impact the election, but it introduced even more volatility into the race. Per our scenario modelling, this meant the distribution of potential outcomes fattened, with the more extreme outcomes becoming more likely, one of those being the possibility of a Democrat clean sweep of the White House and Congress.
That, in our view, is the most negative scenario for US Treasuries. Market pricing earlier in October confirmed the sensitivity of the US Treasury markets to the kind of stimulus a unified Democrat executive and legislature could enact. We see more fiscal stimulus driving yields up through two channels: greater issuance for the market to digest, and firmer near-term growth prospects.
We had already been looking for an opportunity to reduce duration risk ahead of the election, so the market’s knee-jerk reaction to the news of Trump’s positive COVID-19 test – as perceived safe haven yields jumped higher – presented us with the chance to move our US Treasury and overall duration outlook to become more negative.
The main risk to this is the prospect of stalling economic growth in the fourth quarter if fiscal stimulus fails to flow. However, we struggle to believe that that will be as important as news on potential vaccines and the election.