There are a million ways to measure financial risk, and if one wants to calculate global risk, it requires a truly heroic set of assumptions. But it’s much easier to calculate global market risk, and that is what I set myself to do below.
So what is the difference between global risk and global market risk? Market risk captures risk in equity markets, while risk generally is everything. Since that everything includes illiquid and nontransparent loans and hopelessly complex derivatives and all the other myriad parts of the financial system, it’s a hopeless task to calculate global financial risk.
Market risk, with its emphasis on equity markets, is much easier.
Start with national market risk and wield the riskometer, the device used to measure market risk. I’m using the six most common, each with their own pros and cons.
Having obtained the national market risk, I then take the GDP weighted global average for the G20 member countries, and that gives me global market risk. It is a number between zero and 100, with zero being no chance of loss and 100 all the money in all the world stock markets is fully wiped out.
On the day I am writing this, global market risk or GMR is 3.1 quite low compared to the three most dramatic events in recent history, the 8.9 on 27 October 1987, 9.3 at the height of the global crisis in 2008 (31 October 2008) and 9.1 when the Covid 19 crisis was at its worst on 17 March 2020.
There are two bits of technical details worth mentioning, which actual riskometers are used, and how GMR compares to the VIX.
Take the six most commonly used riskometers, run them on a country’s main stock market index, and take the average.
I calculate these daily to measure how much money I might lose over the next day if I hold $100 (or 100 units of each country’s currency) in the index. Technically, it is a 1% VaR, and before anybody complains, a 1% ES would give almost exactly the same result, but be less accurate.
So, it is a number between zero and 100, with 100 being a total loss on all major stock markets globally, and zero a perfect safety.
The six riskometers have all their own pros and cons, and as I have discussed them elsewhere, I will not repeat that here. They generally focus either on time dependence (heteroscedasticity) or fat tails. Only one does both. That means that each has its own strengths and weakness.
|Method||Focus on time dependence||Focus on tails|
I can provide more colour by seeing how each of riskometers reacted to a particular stress scenario.
Start with the biggest one day stock market crash in human history in October 1987.
It clearly shows how the riskometers that focus on time dependence shoot up, while others barely budge. The one riskometer that does both fat tails and time dependence reacts most strongly. These results are entirely in line with every other study on riskometers.
Then take the global crisis in 2008, which started in July 2007. We see market risk slowly Increasing until it shoots up in September 2008, slowly falling back.
Here we see the less time sensitive riskometers react much more strongly than in 1987, the reason is that the stress events were much longer lasting.
And finally, the Covid crisis, which is more like 1987 than in 2008. The more time-dependent estimators react strongly, and the others barely move.
The most obvious alternative to the GMR estimate is the VIX, the most commonly used global risk measure. It is so popular it even got its own thriller, Robert Harris’s The Fear Index. Many academic papers tell us how the VIX affects capital flows and asset allocations throughout the world.
The VIX isn’t a very good measure of global market risk even though many people use it that way. It is the one month volatility of the main US stock index, the S&P 500. As such, it is both dependent on the US markets and even worse, on volatility being a good measure of risk.
Well, it isn’t, because financial returns have fat tails.
Because the VIX is a one month volatility, and the typical trading month has 21 days, I take my GMR index and multiply it by the square root of 21 sice the typical number of trading days in a month. The numbers look similar, but certainly not identical.
If one only looks at the US, they are more similar still,
Focussing on 2020 and three methods, and the VIX I get
The VIX is pretty much identical to the GARCH riskometer estimate in the crises, but declining a bit more slowly.
The fat tailed and time-dependent t-GARCH reacts much more aggressively.
But what accounts for the differences? Will do that later.
© All rights reserved, Jon Danielsson, 2021