A crisis is a sudden degradation: it becomes systemic when it affects an entire system – which system? The financial system, for a start.
A financial crisis has a lot in common with a bomb. Each has:
- A detonator: for example subprime loans,
- Instant amplification mechanisms: securitizations, structured products, collaterals, derivatives, and other pro-cyclical mechanisms like rating agencies’s and central banks’ ratings.
It also shares some common features with an infectious disease:
- Initial sufferers: certain subprime borrowers for instance,
- Contagion mechanisms: dissemination of the risks through various market and banking canals.
In today’s financial system, can we find such amplification or contagion mechanisms?
Yes, partly the same as in 2007, with a few new threats on top. Such as speculative products with an internal leverage (CFD, Contracts For Differences). Or some new structured products bundling other financial products, real, or worse, synthetic (both called ETF, Exchange Traded Funds).
Or more generally “passive” or index-based asset management (a new type of herding behaviour) which is set to account for the majority of asset management in the near future (already 38% in the United States, half ETF and half index-based, and the trend is on the rise in Europe as well, due to the MiFID II directive imposing transparency on management fees, which are smaller for passively managed funds).
Or what about inter-bank flows, another contagion mechanism, which will gear up again once Quantitative Easing comes to an end in 2019, and even more so when ECB’s medium term loans to banks come due in 2021.
Where are the detonators, the self-ignoring sufferers?
Nothing obvious here, otherwise one could defuse the bomb or prevent the epidemic.
Some fear China, others high leverage, be it sovereign or private, or both. What about Mr Trump, a detonator in his own right (but how to defuse him?), or real estate once more, but where exactly? Or Brexit, or Italy, or… but you see the problem…
The difficulty is that we are not dealing with an “objective” detonator or disease, but a subjective one. Are Greece or Italy’s debts, or nowadays Italian or Greek banks’ debts, sustainable? Never mind the rational answer, what matters is the subjective opinion of the “markets”, that of the traders, of their animal spirits, as Keynes put it.
If traders don’t buy into Italy’s debt any longer, will the disease spread to other states suddenly considered “too leveraged” , like France, for instance?
Animal spirits, I tell you. In short, nobody knows.
Financial mechanisms are chaotic in the mathematical sense, that is to say deterministic but with steep amplification mechanisms (“the famous butterfly effect”), which makes them difficult to forecast, even more so than meteorological events, which obey to well known physical laws, computed by our most powerful computers: they can see hurricanes coming.
But we didn’t find formulas for animal spirits, yet.
What’s more, we haven’t even really tried, at least in official circles.
Still worse, economists, including those in the control towers of central banks, are using DSGE models, coming from the general Equilibrium Theory (GE, General Equilibrium, but crises are ruptures in an equilibrium), not even deterministic (S for Stochastic: nobody would dare making just ONE prediction).
In short, exactly the contrary of what we would need in order to model a crisis efficiently.
Of course their models do move (D for Dynamic), but preferably in a linear mode, and so, by construction, they are exempt from any mechanism that would amplify the beginning of a crisis; indeed, quite to the contrary, when such a deviation is looming in a DSGE model, it is supposed to be countered by a “back to equilibrium” intrinsic trend, a trend not seen in the real world….
Nevertheless, some central banks economists did try and model, however imperfectly, early warning signs of a crisis; and we should welcome the ECB Task Force on Systemic Liquidity’s work, and its Single Supervisory Mechanism’s decision to implement its LiST test in 2019. The LiST test is a stress test for banks exposed to liquidity shocks, knowing that the 2007/08 crisis did propagate any more through bank liquidity shortages than it did through proper solvency problems.
Let’s go, dear modellers, still one step further, please do include non-linear financial functions in your models, please simulate asset prices prone to sudden crashes, as happens in real life. If you do, who knows, you might even be able to forecast the next crisis, or even better… to prevent it from happening?
Meanwhile, please use the model of Paul de Grauwe et Yuemei Ji, which proposes to emulate animal spirits, simulating waves of optimism and pessimism, which appear to be even stronger in countries tied by a monetary union…
Or, perhaps, take inspiration from epidemiology modelling, which is able to compute, for each infectious disease, the minimal vaccination rate that will avoid the epidemic spread?
 In particular Jan Hannes Lang, Tuomas A. Peltonen, Peter Sarlin, A framework for early-warning
modeling with an application to banks, working papers of the ECB, N° 2182 / October 2018 ; but they only work on banks’ loss functions, not on liquidity risks. And they test their model only on publicly available data, whereas they are in a position to have access to the data obtained by the Single Supervisory Mechanism. However their model seems to be powerful enough to help this supervisor, albeit frequently giving results already known or self-evident.
 Systemic liquidity concept, measurement and macroprudential instruments, Occasional papers N° 214 / October 2018: “The case studies showed that substantial expert knowledge is needed to explain changes in the indicators for the banking sector.”. In particular, liquidity in international banking groups proves to be especially stealthy. Moreover, the scarcity of data for the non-banking sector has to be completed. The dashboard being proposed in this report should be completed in many ways, but it will take time and hard work.
 Animal Spirits and the International Transmission of Business Cycles, Cesifo working paper n°5810, March 2016. But of course, being very sensitive to small variations, this model is also prone to difficult calibration problems…