Value at Risk (VaR) is used as our main risk measure, but how good is it?
Can we just rely on the output when it is based on a range of assumptions and underlying relationships, which can and do change?
When we are told that returns are expected to be within a certain range with a 95 per cent probability we accept that at face value, and when something goes wrong, well it was just one of those nasty one-in-20 events. We might discuss the possibility of risks having a ‘fat tail’ but we can’t quantify it so it just gets ignored. And that’s the point, it gets forgotten.
But what if the one-in-20 events actually occur more frequently? How should that change our attitude to risk? What if so-called one-in-20 events occur almost every year?
I have looked at a 20 year period, 1987 to 2007, from the equity crash of ’87 that was “so unlikely, given standard statistical models, that it caused the entire basis of quant finance into question”. A 20 year period when, perhaps, we might expect one or maybe two one in 20 events to occur. I have selected events that either did, or had the potential to affect global markets and at the time were referred to as exceptional, once in a generation, once in a lifetime etc.
From the financial crash of the 80s to the 'dot-com' bubble
In 1987 we had “the crash of ‘87” and in 2007/2008 the financial crisis. Twenty years apart as it happens, so there we go. So let’s look at the years in between.
Starting with October 1987, memorable globally for an equity market crash, and domestically for the storm that demolished so many trees. There was indeed a fierce bear market starting in mid-October; in the US the S&P 500 lost 40 per cent in three days. The UK suffered a similar though slightly more muted fate, equities only lost 25 per cdnt that month.
Less than two years later followed the Tiananmen Square massacre of June 1989. Reports of the number of deaths vary, the Chinese authorities admitted to 241, in a letter released years later the British Ambassador put the number at around 10,000. It didn’t help global market sentiment.
In November that year the Berlin Wall, dividing Berlin between east and west, came down leading to the reunification of Germany.
In December 1989 the Japanese equity market peaked with the Nikkei index at 38,916, it represented more than half of global equity market cap. Over the next two years it declined by two thirds, two decades later in 2009 it bottomed at around 7,000 – a decline of over 80 per cent. Over 33 years later following a strong bull run it is still 15 per cent below its 1989 peak.
In September of 1991 the USSR was dissolved, Mikhail Gorbachev resigned as president of the Soviet Union, leaving Boris Yeltsin as president of the newly independent Russian state.
‘Black Wednesday’ in September 1992 saw the UK's Bank of England base rate raised from 10 per cent to 12 per cent and then again to 15 per cent, the UK was ejected from the Exchange Rate Mechanism (ERM), as Italy had been two days before.
In 1998 emerging market currencies were in turmoil and Russia defaulted on its debt and devalued the rouble. It was the first time this had happened since 1917, clearly an infrequent event, though it did default again last year.
The dot-com bubble, when the NASDAQ index rose by 400 per cent, followed by a decline of 78 per cenrt is clearly worth a mention. Tech stocks up four times on average, then they lost all of their gains, and then some more.
The 'Black Swan', banking crisis and global pandemic
The Enron scandal, where an enormous company, $75bn in market cap, just imploded after “accountancy irregularities”, to put it politely, were disclosed. Companies go bust all the time, what made this one memorable, apart from the scale, was that it brought down Arthur Andersen, one of the big five accounting firms, there is now a ‘big four’.
The boxing day tsunami of 2004 was unbelievably tragic. 230,00 people, almost a quarter of a million, dead in fourteen countries as a result of a 9.1 magnitude earthquake off the coast of Northern Sumatra in Indonesia. Though appalling in human terms, it had little effect on the financial markets, unlike the much smaller magnitude 6.9 quake that hit Kobe, Japan in January 1995. It’s effect on the Japanese equity market was enough to trigger the downfall of Barings, Britain’s oldest merchant bank.
If I have been selective in choosing 1978 – 2007, just look at the following 20 year period. A global banking crisis, a European bond crisis four years later, a global pandemic, a war in Europe, another tech boom and bust, gilt market volatility unprecedented in over 300 years, the list goes on. And we are only three quarters of the way through this 20 year period.
Perhaps one in twenty events do occur more frequently than that in the real world. Is VaR a good measure?
In his excellent book The Black Swan, Nassim Taleb argued that VaR “ignores 2,500 years of experience in favour of untested models” and “is charlatanism because it claims to estimate the risk of rare events, which is impossible”
Also not to be forgotten, the VaR number does not represent the amount at risk, it represents the smallest amount of risk in that range of outputs. In the model it is not the amount that might be lost, it is the minimum amount that might be lost in that one in twenty event and importantly, statistics calculated following a period of low volatility may understate the potential for risk events to occur and the magnitude of those events.
So when given a VaR by advisers we should note that the number is spuriously accurate, dependant on assumptions, gives absolutely no indication of the maximum potential loss, and will vary according to volatility in the preceding period. Apart from that, it may be a good measure.
Clive Gilchrist is a trustee executive at BESTrustees
