Steve Danby and myself are the co-managers of the fund. We've both been quant analysts for the best part of 20 years. Steve has spent nearly all of that time at Henderson, I've spent most of that time on the broking side, where I finished up running the European quant team at JPMorgan Chase. I left there three years ago, and came to Henderson with the idea that I was going to work with Steve to build hedge funds. He'd been given some time by his line managers to start thinking about the product. With that in mind, he came to me and put the proposal to me.
Because we both have quant backgrounds, but from different sides of the fence, we came together with wide experience of not only seeing what kind of quant techniques other managers and other funds were using, but also what many of the pitfalls were. When we came together to build our model, it was inevitable it was going to include a number of the factors that other quant-based funds have. If you look at some of the other quant-based factor models, you will find that they have some kind of value model in there, some kind of momentum model, and numerous other factors.
There is nothing particularly unique or clever in having a value and momentum-based approach, given that value and momentum as a strategy in the UK alone has been very well documented for over 20 years. It is not particularly unusual having those factors in your models. The reason a lot of people have them is because they work pretty well most of the time.The thing we think we bring to the table which is slightly different is that we have approached our models from the point of view of: "These factors work well most of the time, but not all of the time; therefore, what can we do during the periods when they're not working in the conventional sense to try and drive them a little bit harder? Can we create the circumstances where these factors work more often than other funds tend to be able to achieve." To do that, we had to have an appreciation of factors (or styles as we generally call them), what they are really trying to do, what they really mean.
If you think about an investor who invests on the basis of a value strategy only, all that they're really is doing is ranking all companies on some sort of value measure. They're picking their longs from the good value group, and their shorts from bad value. Something about this style discriminates between the stocks at either extreme, which typically perform differently and there is an assumption made that the good value extreme will do better over time. Over a very long time horizon, most back-tests show that that is indeed the case. This is how value investing has got such a broad-based appeal, over time it typically works. But if you were to draw a chart of the return of this strategy over a very long period of time, sadly it is not a straight line. There are periods where it doesn't work.
In the past, the response of most funds to a period where their styles were not working was generally to cut the weighting on those styles. If you had a multi-factor model, including value, revisions and momentum, if good value was doing badly, you could reduce your weighting on value. One example was during the end of 1999 and into 2000. If you were buying very good value stocks at that time you would probably have had quite a difficult six months, depending on what your value measure was like of course. The response would have been to reduce the weight on value, possibly to zero. If you do that, by definition all your style can contribute at that point is a flat line.
We look at things slightly differently: with any style all you really have the right to expect is a good discriminator between the two extremes. It is possible for either extreme to outperform the other. We can isolate style regimes, for example where typically good value is beating bad, or where bad value beats good. This might not be the prevalent style regime, but there are times where it happens, the most obvious example being the tech bubble, where bad value was beating good value.
We believe this is applicable to other styles, for example a risk model. There are periods when the market wants to be very pro risk, and clearly there are periods when it wants to be risk averse. If you can try to identify which of these regimes we're currently in, then you have a chance to iron out your downside. By accepting that these styles can operate in a series of different regimes, if we can pick the regime the style is operating in, we can make the style work more often than in traditional models.
We don't believe that our revision score is superior to any other revisions model, but what we do think is different is that we can make a revisions strategy perform more strongly than other models in this regime analysis.
There are overlaps. In the period until mid-May, stocks with extremely good earnings estimate revisions had been performing sufficiently strongly to become very strong in terms of price momentum, and also were typically fairly high risk, and so those three styles had become correlated. They don't all have the same correlations over time, but price momentum is looking at whether the price of a stock is going up or down particularly strongly. If any of the factors other than momentum are working for long enough, they become correlated with momentum.
I think it's appropriate to spend a lot of time on R&D. There is a phenomenon known as model risk: you build your model, it starts to work extremely well, but then the results begin to deteriorate. It isn't necessarily because it's a bad model, but because all of these things ought to be regarded as dynamic. We think of the process in that way.
If we can find new styles that can add value to the system, then they would certainly be included after an appropriate period of testing and analysis. We are actively engaged in trying to make the model still more flexible to cope with all circumstances, and trying to learn from what has happened in the more recent past.
For example, what was appropriate after the whole May/June experience was to look to see whether there was anything unusual that had happened to any of our factors that could have been subsequently viewed as a warning sign that we'd missed. There really wasn't anything like that, but had there been, then it may well have been a change we'd have implemented in the system.
We believe the approach is portable by geography. The basic investment philosophy is applicable to all markets: you can find a series of factors that would typically work, with regime analysis you can make them work more effectively, and therefore can build a fund from that.
If you are looking at Japan, you might believe that the market dynamics are different from Europe in some respects. It is quite clear that Japan has a more value orientation than Europe, and earnings estimate revisions have typically done less well than in Europe. Japan is typically more volatile. There may be other factors that don't work that well in Europe that may work in Japan. What is very likely is that there will be some different factors in the Japanese model compared with the European model, but that doesn't mean the whole notion of analysing styles for the regime that they're in would be done differently.
One of the roles we have had at Henderson is an advisory function to long-only and more latterly to long/short managers from different geographic regions to advise them on style and portfolio positioning. We did that across a whole series of geographic regions, including Japan, Europe, US, and UK, and we therefore go into this with some experience of Japan, and understanding how Japan is different.
It's simply that there are more companies to choose from. One of the issues that many hedge funds face is the liquidity of the markets in which they operate . You want to have a big universe to choose from, because, in common with many other quant funds, we want to have a reasonably large number of positions. If we want 200 positions, we can't be dealing with a market that only has 150 liquid stocks.
There's nothing that we've liked or not liked. We haven't gone as far as studying the US to any great degree with our current methodologies. It was logical for us to start with Europe, partly because that is where we have most of our stock-specific knowledge, and partly because it's where we are based, and partly because it's one of the geographic areas where Henderson has demonstrable expertise.
It wasn't a clear cut decision to research Japan before the US, but we believe that we could produce a very good product for Japan. We believe we could also produce a very good product for the US, but we can't do both at the same time.
The fund is managed according to a series of rules. We try to keep our historic volatility between 8-12%, and the gross exposure is managed with that in mind. We do regard that as a target we should try to hit under all circumstances: we're actively interested in keeping that level of risk up. One of the advantages of quant funds in general, is if you're reasonably open with the people who are investing in your product, and you govern yourself according to a series of rules, and you stick to your rules, then it means that your investors know what they have bought.