It’s all down to the original model brief: to predict the potential losses if at 4.15pm the powers that be decide to close out – or stay another day in the game. The poker-like way of thinking is simple and, in a gentlemen’s game, effective: an instant decision, the cards are laid on the table, and it’s a clear loss or a win. The players move on to the next round. There is no cheating and no guns are drawn – the SEC holds the big one. If a player is broke, he simply leaves the table. In the VaR world, at the end of the evening, everyone still playing totals their wins and losses, and goes home.
How about playing a different game? Players join by paying an upfront fee, and expect regular payments while they are still in the game, plus a big dollop of cash at the end of the evening. If the food gave them the runs, an earthquake strikes or their broker rings with a margin call, the player has to leave the table immediately, with whatever the bank is prepared to pay at that time. Depending on how many players have joined and left since our player joined, the bank may be able to pay all or only part of what was due at the end of the evening. This is not a VaR world.
So who should like VaR then, if they were explaining their wife how they play with family money? Depends on the game they play; below there’s a possible look at several well known names in the industry:
|Firm||% of profit from short-term trading|
Is VaR a risk measure suitable for all, or even many? Who wouldn’t like to be the dashing, all-conquering player who breezes in, wins, and breezes out into the night? And who likes to tell his mates he’s playing James Bond, only to be found to work as a bank clerk?
They don’t have a good name in finance these days, being blamed for all sorts of issues – but primarily for not being able to predict crises. Sure, mis-selling financial products, from “plain” mortgages to “sophisticated” hedge funds doesn’t help – but the “black boxes” are what concerns me most.
In the historic scheme of things, the world has benefited greatly from models’ predictive abilities; we cannot unlearn modelling, but just have to do it properly: put money only in models that have proven themselves over time and are widely accepted in the industry, feed the models with clean, reliable data, and read the model results within their accepted limitations.
I see several issues with risk modelling today:
- The current fad to see “black swans” everywhere, when in fact they are historically – naturally – quite rare. Those who believe in uninformed statistics will always be surprised when they see more “black swans” than usual, even in the forests around Chernobil; only those who consider their environment will be able to make the causal connection.
- Commercial pressures can impair the scientific rigour of the modelling process. Today’s finance takes models from worlds that work in different time scales than its own: a geological model that could lead to a whole mountain being blown up will be tested until ready – unlike its financial application, which has to be used when the next big deal comes.
- How much profit can be honestly traced to a new model, that nobody else has? In large financial centres like London or New York, a model cannot remain confidential past the next Friday night – but that often contributes to more innovation.
- There are private and public agencies entrusted by the public to spot the gamblers and stop them from passing themselves as informed investors, but as the pace of innovation accelerates in an entrepreneurial environment, how can such agencies differentiate for sure between gambling and superior knowledge?
Transparency allows all stakeholders – competitors, regulators, customers, academics – see each model and test it on their own. If it passes all the tests, we all get a tool we can trust; if a single person manages to break it, then perhaps we avoided catastrophe. Also, publishing the methodology attracts specialists from other fields than finance, which raises the general level of the debate. The RiskMetrics project has already tackled market risk this way, but challenges have expanded since to credit default, liquidity, ALM, insurance and operational risks.
With this idea in mind I have started this blog, which aims to stimulate open discussions about all models that may be used in finance. A Wikipedia for financial risk models? I’d love to think so, but am happy to start with a blog.
Next posting I should start publishing the first iteration of the first model – with due respect to the RiskMetrics project, I’ll write about market Value-at-Risk.
It took me a year to start this blog, after pondering whether I should start with a forum instead, etc… I’m so pleased that I finally made it. Next, I should start writing what this is about. Next post.