Every day the Asset Allocation team has a morning meeting, in which our economists and strategists debate the latest news and expectations for the day with our portfolio managers. I have to admit that, increasingly, we discuss the same thing many people do in offices around the world: politics!
Macro investors have scrambled to become political experts, and we are no exception. The time we spend in our research meetings debating the intentions of politicians, the latest tweets, and recent geopolitical machinations has skyrocketed.
This is not idle speculation or punditry. Our economic roadmap view is now unusually binary, with upside and downside scenarios competing for prominence. Our recession probability stands at 35% – a high for this cycle – and never in the past few years have we been so dependent on the outcome of one particular set of political negotiations (around trade wars) to determine our outlook.
Moreover, we are 12 months away from a US election which brings a serious risk of a sharp swing to the left in the US. We expect investors’ and the media’s focus on this to increase in the run up to the Democratic primaries in February and early March 2020.
Our New Political Paradigm theme is an amalgamation of all the work we have done in this area. It covers Brexit, Trump’s new doctrine, and the rise of populism, grounding them all in concepts like the Thucydides trap identified by historians.
One of the challenges we face in this type of research is how to avoid behavioural biases when looking at political trends. So-called ‘recency bias’ – the phenomenon of overemphasising recent developments and extrapolating too much – is very common in political research.
It is easy to become extremely bearish due to recent political events or just after a crisis, but there are two risks in doing so:
1. How do we make sure we calibrate this bearishness against historical events and risks? It might look bad in the moment, but how do you know whether it’s any worse than in the past?
2. How do we make sure we stay ahead of the curve rather than pricing in risks once they have occurred and risk premia have spiked already? It’s tempting to become bearish when everyone else is panicking as well, but isn’t that the worst moment to do so?
To mitigate these behavioural traits, we use a scorecard or checklist approach. The benefit of this approach is well documented, as set out by Atul Gawande in The Checklist Manifesto, and allows you to consistently review the same information, become more forward looking, and document signposts.
What are the factors we look at in this political risk scorecard? Like all the scorecards we use, it has a mixture of qualitative and quantitative factors. There are three useful external indicators we use: the Geopolitical Risk index, the Global Economic Policy Uncertainty index, and the World Uncertainty index.
All three go back a while, which is important for calibration reasons, and all use word searches in press articles to build their scoring. In their Geopolitical Risk index, for example, Dario Caldara and Matteo Iacoviello of the Federal Reserve Board have constructed an automated monitor of 11 top international newspapers; it dates back to January 1985 and captures the incidence of words related to geopolitical tensions each month.
Their index thus quantifies the proportion of news stories highlighting geopolitical risk each month to give us a longer-term perspective, and it does broadly correspond to variations in these risks through time. Its zenith occurred when the US invaded Iraq in 2003, for instance, with other spikes evident around the September 2001 terror attacks and the Ukraine crisis in 2014.
So what is our political risk scorecard telling us today? Well, all three underlying indices are now reasonably high and – more importantly – they are rising.
This is especially worrying as we are not looking for a peak in this indicator (that’s when the geopolitical risk has happened and we are too late to anticipate it), we are looking for the build-up of political tension that makes a future risk event more likely.