Most potential severe economic scenarios remain just that: a plausible but theoretical set of numbers that capture dire events that fail to materialise. And as such there is a risk that when the worst happens, it is not the worst that was planned for.
Recessions have been rare (outside the pandemic which is perhaps best viewed as a special case) since the Global Financial Crisis. In the era of using IFRS 9 for calculation of Expected Credit Loss, risks, including supply shocks as the world exited lockdown, energy price spikes as Russia invaded Ukraine and latterly the threat of tariffs have all been modelled by economists. And economies have largely been resilient. Credit losses have been muted.
Now the world, and particularly GCC countries, have to address the question of which scenario we are actually in, and how will it actually unfold?
Viewed from the GCC countries, the crisis has now arrived, and a downside scenario is no longer a distant prospect. How should we react?
Ideally, we would just take the downside scenario (or stress test results if we are really worried) and use those answers. But there are several problems with that. How close is our downside scenario to the events that are unfolding? If the answer is ‘not very’ then the first step is to revisit it and revise it. After the event, we might ask why we did not have the correct scenario, just as was hopefully done after the inflationary shocks in 2023 which most economists did not see coming.
Do we trust the Economic Response Models (ERM) that links the economy to default rates? Real life crises tend to highlight deficiencies in models. Not many models have been calibrated during periods where inflation and interest rates rose for example. The shocks to oil trade and tourism are new. The business will know what is happening to its customers well before it has the GDP figures that drive the model. Bottom-up approaches may be a better guide than econometric models at this stage.
It is easy to mis-specify ERMs. These models are at best a useful guide for informed debate and not a substitute for it. Economic data are a version of the truth based on surveys. Models are often overfitted or rely overly on statistical selection criteria and miss the structural story.
Anecdotal evidence suggests that some ECL models used in GCC countries are predicting a fall in ECL in the coming quarters. And the reason is that oil prices are a key driver of default. Of course, higher oil prices are usually good for the GCC economies.
Our demand-driven scenarios tell us prices will rise when global growth is strong. But may turn out that the model does not work properly when that high price is a supply problem. The model is conditioned on price, but it should be conditioned on revenue. And that will tumble if export volumes collapse.
Our models should work better when it comes to GDP. The link between defaults and output is hopefully well established. But taking the UAE as an example, the economy is very different to the one of just a decade ago. Oil still represents a big share of GDP and government revenue. Models that reflect the past may not always be a reliable guide of the structure of the economy has shifted.
First, are we in the severe scenario? Is there a set of early warning indicators and triggers to help make this judgment. This is vital input into the decision about how much weight to allocate to each of the scenarios.
Second, as the severe scenario looms, firms may need to use PMAs to deal with the exact nature of the shock. If early signs are that the hit to tourism will be severe and extended, and the firm is heavily exposed, then a model that uses aggregate GDP as a driver will not capture the sectoral impact properly. It was likely built around an implicit view that all sectors would suffer in tandem.
Third and finally, when we are in the stress, models built using a top-down macroeconomic assessment will only get us so far. Rules of thumb about the sectoral impact and the number of customers that will be affected by loss rates may be a better approach. And knowing and monitoring the biggest counterparties may also be an option.
Approaches that look at distributions of past shocks are a useful benchmark to assess severity. But this has led to situations in the past where risks have been missed. Even as inflation rose in 2023, the possibility that it could climb much faster was often discounted. Basing scenarios around a plausible narrative first, while still being consistent with the distribution, can help firms understand the emerging risk and manage it better.
But when the downturn comes, it will not be exactly like the one we envisaged. At that moment, macroeconomic projections take second place to what is happening on the ground. Knowing where you are in terms of exposures to different sectors and having a good structure to gather and disseminate management information is vital.
You will know what is happening to your book long before much of the economic data that drive your models is published.
