COVID-19 forced central banks to cut rates to record lows and delve deeper into quantitative easing, and most have maintained these extraordinary loose settings despite growing signs of economic recovery.
This environment is ripe for debt-fuelled asset price booms. At some point, however, the imbalances will unwind—usually abruptly—which can cause a credit crunch or, worse still, a financial crisis.
Perhaps it is human nature, but when the good times roll it can often be difficult to foresee a crisis. To quote the late Professor Rudiger Dornbusch, a “crisis takes a much longer time coming than you think, and then it happens much faster than you would have thought”.
Accurately forecasting the timing of financial crises is likely to remain an elusive goal, but the vast literature shows that it is possible to build early warning systems that can help assess the build-up of risks and provide a useful wake-up call.
To assess the risk of domestic credit crunches and financial crises, Nomura built a model we call Cassandra, named after the Trojan priestess of Apollo in Greek mythology. Cassandra was cursed to utter true prophecies, but never to be believed; she foresaw the destruction of Troy, and warned the Trojans against accepting the Greeks’ gift of a giant wooden horse, in which Greek warriors were hiding. Similarly, our early warning model Cassandra warns and signals in advance of the economic Trojans’ fate.
To capture the build-up processes that, in the boom phase, sow the seeds of the subsequent financial crisis, we calculate deviations (or “gaps”) from long-run trends for five early warning indicators (EWIs): the ratio of private credit to GDP, the debt service ratio, real equity prices, real property prices and the real effective exchange rate.
Exhibit 1: Methodology
Next, we assign thresholds to each EWI which, when breached, flash a signal of a crisis occurring within the next 12 quarters. The optimal threshold values are selected to minimise the noise-to-signal ratio – in other words, to minimise the risk of having too many false alarms (noise) and missing too many crises (signal). Importantly, we could further cut down the noise-to-signal ratio by combining some of the EWIs into joint indicators.
We combine the predictive power of our six parsimonious indicators by weighting them by the inverse of their noise-to-signal ratios, to produce our composite index, Cassandra.
What we learn
Cassandra is designed so that whenever the index exceeds 100 it should be interpreted as a warning that the country is vulnerable to a financial crisis within the next 12 quarters.
Setting a threshold of 100, Cassandra correctly signalled two-thirds of the past 53 crises in our sample of 40 countries since the early 1990s. Using data up to the first quarter of 2021, or the latest available data, Cassandra is currently warning that six economies—US, Japan, Germany, Taiwan, Sweden and Netherlands—appear vulnerable (with scores above 100) to domestic credit crunches and financial crises over the next 12 quarters.
Exhibit 2: Latest Cassandra scores
On the other hand, over half of the 40 countries have Cassandra scores of zero, indicating very low risk of systemic credit crunches and financial crises.
Interest rate shock
However, Cassandra’s latest results are positively influenced by very low interest rates that probably will not last. Rising and lower quality corporate debt was already a cause of concern in many countries prior to the outbreak of COVID-19, and there is little doubt that the extraordinary global monetary and fiscal policy response and regulatory forbearance to the COVID-19 crisis enhanced the ability of firms and households to service (or defer) their debt obligations. This allowed most to avoid the spectre of default and bankruptcy, but has come at the expense of a further increase in private sector debt levels in most countries and a deterioration in credit quality.
According to the OECD’s latest economic outlook report, around 30% of the non-financial corporate debt stock currently rated by Standard and Poor’s sits in entities rated as “speculative”, and 40% in entities with only a BBB rating, the lowest rating in the investment grade category (about $5 trillion and $8 trillion respectively).
It is only with generous policy support that some unviable firms have been kept alive. Many of these, if left to fend for themselves, will have insufficient operating revenue to meet their interest payments. This also underscores the heightened risk of financial vulnerabilities in the event of an interest rate shock.
To this end, we stress test Cassandra to an interest rate shock. We do this by manually switching ‘on’ the signal for the debt service ratio in all 40 countries.
In the face of an interest rate shock, the economies that would appear vulnerable (scores above 100) to financial crises are the same six economies as above, but there is now a sizable group of another five economies—France, Hungary, Romania, New Zealand and Portugal—with scores not too far from the 100 threshold.
Incorporating climate change risk
Catalysed by the COVID-19 crisis, there is increasing recognition among government officials—and to a lesser extent, investors—that dealing with climate change is a source of physical and transition risks for the financial sector that will most likely have significant implications for financial stability.
Climate change physical risks involve extreme weather events—storms, floods, wildfires, landslides and droughts—that can cause large economic and capital losses, not to mention loss of life. The frequency and severity of climatic hazards is increasing in line with global warming, as is the economic cost. For example, since the 1980s the number of registered weather-related loss events has tripled; and inflation-adjusted insurance losses from these events have increased from an annual average of around $10 billion in the 1980s to around $50 billion over the past decade.
The financial sector is exposed to climate change physical risks through underwriting activity (insurers), lending activity (mostly banks) and market value of affected securities (all financial firms). The other kind of risk is the transition risk to a greener economy which includes ‘dirty’ assets becoming stranded; reputational damage to companies, industries and even governments; and market-driven financial distress of polluters.
One example is energy companies. If government policies were to change in line with the Paris Agreement, then two-thirds of the world’s known fossil fuel reserves could not be burned. This could lead to changes in the value of investments held by banks and insurance companies in sectors like coal, oil and gas.
Standard asset pricing theory suggests that investors should demand a premium for holding assets exposed to climate change risk. But, because the nature of the risk is new and long term—similar to financial cycles—markets may not price future climate change risk correctly. Market mispricing of climate change risk today heightens the danger of a sharp, sudden re-pricing of this risk at some point in the future, which could trigger financial crises.
There is not strong evidence that financial risks related to climate change are fully reflected in asset valuations. For example, the IMF, in a major empirical study last year, concluded that aggregate equity valuations as of 2019 did not reflect climate change risk; thus, equity investors may be paying insufficient attention to climate variables. The Network for Greening the Financial System, a group of 83 central banks and financial supervisors, warn that research on the impact of climate change risks on the financial system is still in its infancy and is one of the most urgent and prominent issues, given the paucity of data and given the contagion and feedback effects from a re-pricing of climate-related financial risks could be potentially large.
However, quantifying climate change risk is challenging. We have taken a metadata approach, relying on the consensus of data sources by aggregating the standardised scores of eight different climate change risk indicators, which themselves are composites of various data. According to the analysis, Nordic countries display the lowest climate change risk. At the other end of the spectrum, India, South Africa and the Philippines appear most at risk with index readings of over 250. In fact, the top 13 with the highest risk are all emerging market economies.
Finally, we combine climate change risk and an interest rate shock to Cassandra. Curiously, once Cassandra is augmented for climate change risk and an interest rate shock, the number of countries vulnerable to financial crises drops from six to five (Sweden falls below 100). However, the list of countries with scores between 70-100 lengthens to 10, including seven emerging market economies.
Exhibit 3: Cassandra adjusted for climate change risk and an interest rate shock
In the original Cassandra, the average score of the 40 countries was 35. When we augmented Cassandra with an interest rate shock or climate change risk, the average score rose to 52 and 57, respectively. The original Cassandra, when combined with an interest rate shock and climate change risk, raises the average score to 69.
A word of caution
Finally, it is important to note that no early warning system is fully reliable, and it would be foolish to make any exaggerated claims. The nature of financial crises can change over time and there are many variables that quantitatively are too difficult to measure, such as political systems, regulatory buffers and policy responses. Cassandra is also a very long-run model, with a warning horizon of the next 12 quarters. All that said, the benefit of Cassandra is its objectivity in letting the data speak.