William Essex: Let's get started. The magazine is called Automated Trader, which gives you a pretty big clue as to who's reading it. What can you tell our readers about risk?
Gerald Ashley: I'm certainly not against low-latency trading, but there is a philosophical divide that your readers might want to think about in terms of their models. There is potentially a dangerous assumption that risk is a puzzle you can solve. I really don't think it's that straightforward.
William Essex: Good to know you're not against us, Gerald. Tell me more about that dangerous assumption.
Gerald Ashley: If you think it through, what is happening with algorithmic trading, whether it's delivering more efficient execution, or creating buy and sell signals, or whatever, you're basically saying that you can create a model that is a very fine approximation of the real world, and the real world going forward, based on what has gone before. What else is there to work on? There is nothing else to work on than what went before.
William Essex: There's no 'reference future' available, obviously, or at least, not a reliable one.
Gerald Ashley: That model may be built upon a number of dangerous assumptions, in fact. One problem I have is risk management. There are not many people doing risk management. There are a lot of people doing risk measurement. We've got large international banks, with thousands and thousands of people in risk management, and what have they managed to do? They've managed to get through about a trillion dollars of losses. What have these guys been doing?
William Essex: We'll print the word 'guys' rather than the word you've just used in that question. How do you answer it?
Gerald Ashley: They haven't been managing what's been going on. They've been creating incredibly complicated scoreboards. So we've got a scoreboard that is phenomenally complex and undoubtedly very clever, but it's not the same to be managing the risk, as measuring the risk.
William Essex: Those scoreboards are all historic, aren't they?
Gerald Ashley: And this all ties into one really simple topic, which must be at the heart of what a lot of these guys are doing, which is liquidity. Because all these models make - not necessarily strong assumptions, but some sort of assumption about the liquidity and structure of the market they're dealing in.
William Essex: But you can't do anything without making at least some level of assumption.
Gerald Ashley: Maybe to give you a side-light on all of this, if you recall 1998, with the Russian default and the LTCM crisis, something very interesting happened there. We had markets that had never correlated in the past, starting to correlate going forwards. That was really driven by VaR [Value at Risk; see also page 62]. If you own an asset, and it rises or falls dramatically, it has a volatility impact on your risk management. So lots of people owned Russian bonds, the Russians defaulted, those bonds went from par to zero.
William Essex: Mmm. Something of a volatility impact.
Gerald Ashley: Imagine you were long Russian bonds. Your model says you have to reduce your risk on your overall portfolio, because the VaR has gone up, because the portfolio's volatility has gone up. As a result, you have to sell other instruments. In 1998, we saw the dollar fall, gold fall, because people were getting back to cash. That's an example of where risk models can overtake the marketplace.
William Essex: We know that now, but your point is, people didn't know it then, and there might be other things we know now. That sounds like a reasonably safe assumption.
Gerald Ashley: To pull this back to the centre, my angle is that of course modelling risk is sensible, but it's about managing it rather than just measuring it. There are all sorts of human aspects to how we treat risk that are not easily stuck into a model. There's more to the world than the perfect black box.
William Essex: Okay, so what can we do about risk?
Gerald Ashley: It's a problem you shouldn't try to solve. It's not a matter of just hiring more and more PhDs in maths or natural science or physics. A lot of this has to do with just looking at long-term investment and business cycles.
William Essex: That's a kind of modelling …
Gerald Ashley: And using common sense and experience. The interesting thing that has just happened is that models have made a claim to be more efficient.
William Essex: Efficient as distinct from effective?
Gerald Ashley: If you look back over the past ten or fifteen years, we've seen an explosion in computing power and a dramatic narrowing of profit margins that is compensated for by huge turnover and increased levels of business. That's all predicated on the idea that these models are going to be any good.
William Essex: Or rather, very good to the point of infallibility? Let me ask you this question, finally. To a readership of automated traders, system traders, algo traders, for whom second-guessing the machine is typically a no-no; for that readership, in this time of flash crashes, volcanoes and heavy-duty regulation - what's your message?
Gerald Ashley: The message I have is: come one step back. Don't look at risk in quite such a granular manner. Try and look at it more holistically and look at it from a more judgemental point of view. I think there are loads of things in technology that are really interesting, and really clever, but haven't we just fine-tuned things to an almost guaranteed level of failure?
William Essex: I think I'll take that as a punchline, rather than a question. Gerald Ashley, thank you for such a thought-provoking conversation.
Gerald Ashley is an adviser, writer and speaker on business risk and decision making. He has over thirty years' experience in international finance, having worked for Baring Brothers in London and Hong Kong, and the Bank for International Settlements in Basel, Switzerland. He is now Managing Director of St Mawgan & Co, a London-based consultancy specialising in strategy consulting, risk management and decision making in finance, business and risk-taking. He is a Visiting Fellow at Newcastle Business School.
*Gerald Ashley may be contacted via his publishers, at www.harriman-house.com. Financial Speculation: trading financial biases and behaviour is published in the UK at £24.99.