Portfolio Construction: Allocating risk, allocating time
Martin Steward talks to ex-FRR CIO Jean-Louis Nakamura about Lombard Odier's volatility-driven allocation and its time-horizon tactical allocation process
"The long-term is really only a succession of short terms," observes Jean-Louis Nakamura, executive vice-president and head of quantitative and overlay management at Lombard Odier Investment Managers. As the ex-CIO of the Fonds de Réserve des Retraites (FRR) and ex-CEO of France's Civil Service pension plan from 2002 to 2008, he knows that one of the biggest risks a long-term investor faces is of severe short-term losses creating an accounting headache at best, or, at worst, a funding hole too big to compound one's way out of. A robust strategic asset allocation is the best defence against that eventuality, but Nakamura maintains that the process underlying most pension funds' approach makes them more vulnerable to those short-term shocks.
A strategic asset allocation is, by default, an allocation of risk. But too often the focus is on the allocation of capital, based on an assumption that long-term average levels of risk will correspond to the levels of risk realised over the time horizon of the strategic allocation - that volatility is not, in itself, volatile, and its mean reversion will "iron out" risk over a three-to-five-year period. Nothing could be further from the truth, argues Nakamura.
"Implied volatility on the S&P500 for the past 25 years suggests that risk is clearly not constant over time, and you have successive periods of high levels of risk - Vix levels of 25-35, say - and low levels of 10-20," he says. "It is astonishing to see how regular these regimes are in terms of length - about four or five years."
An investor implementing a traditional three-year strategic asset allocation in 1996 did so based on long-term equity volatility of 20-25%, even though actual volatility was about 10-12% at the time, and within a year it had moved up to levels at 25-35% (with peaks at 50%-plus) which lasted for five years.
"The asset allocation with 60% in equities would stand for a very different level of risk in this new regime than it did when the strategic allocation was decided," observes Nakamura. "Very few institutional investors translate their asset allocations into risk-contribution allocation - but what is it that counts? The problem of 2008 wasn't simply that investors lost money. It was that they lost more money than was predicted even in their worst-case scenarios, because their asset allocation no longer stood for the level of risk they had in mind when they set it up."
If the initial decision is framed as a risk allocation (to both global risk and the risk of individual asset classes), the strategic management of that benchmark consists in rebalancing the portfolio not in terms of asset allocation weights, but in terms of risk contribution - maintaining the initial risk budget by buying assets whose volatilities are falling below their strategic assumptions (often long-term levels) and selling those whose volatilities are rising above those levels. This, in simple terms, is what Lombard Odier's volatility driven allocation process does (as well as volatility, the process also rebal ances against distortions in inter-asset class correlations).
If you accept that volatility exhibits momentum rather than mean reversion, especially in the short-to-medium term, then this kind of approach should help portfolios adapt to financial shocks. "The process will not capture very short-term crises like the 1987 crash or spring 2006 - it just isn't designed to react that quickly," says Nakamura. "Neither are we able to foresee the first shock in a higher-volatility regime. But our matrix is reactive enough to determine whether or not initial shocks are big enough to make it likely that they will be followed by shocks in the same direction. Although it is not designed to create outperformance, the nature of volatility regimes means that, next to buy-and-hold or traditional rebalancing, the process not only mitigates drawdowns but also enhances performance during periods of low volatility because it enables an investor to afford to increase exposure to ‘risky' assets."
Alongside this volatility-driven approach to beta, Lombard Odier implements a tactical alpha portfolio. The innovation here is to split the portfolio, not by asset class, region or style, but into three investment time horizons: long-term tactical (1-18 months); mid-term tactical (1-24 weeks); and short-term tactical (intraday to one week).
"Having different managers for different asset classes seems strange to us, because tactical management should look at all asset classes through the same lens, relative to one another," Nakamura reasons. "In addition, consider the beginning of 2006: a tactical manager would have been very confident in equities because of valuations, or macro, or technical analysis - but during the spring he faced a major drawdown. Does he stick to his scenario for the year and register downgraded performance for that period, or change his view at the bottom of the market? To some extent he's always the prisoner of that dilemma, and it's that which we try to avoid by slicing our risk budget by time horizon."
The process exploits the fact that different trading strategies enjoy their best efficacy over different time horizons. Long-only value-based strategies work, but only if you have long enough for fair value to be realised; technical analysis seems to add value in the very short term, "but we are sceptical that it works beyond about one week". And between those two sits the broad discretionary world of "macro analysis, central-bank watching, sentiment analysis".
The risk allocation between these three sub-portfolios is based on both their information ratio and correlation. When the system decides that these show the prevailing risk regime to be ‘normal', the sub-portfolios are almost equally weighted. When risk increases, the short-term sub-portfolio is underweighted (because trading costs climb and information ratios decline). The long-term, value-based tactical sub-portfolio is subordinated to a binary on/off switch: if the risk regime is deemed ‘abnormal', the portfolio is moved out of risk assets and into a mix of mainly short and long-term rates. In hyper-volatile regimes, therefore, it is possible for the mid-term, discretionary sub-portfolio to be the only one with a risk budget.
"The last time this happened was November 2007, and it stayed red right up until May 2009," recalls Nakamura. "So it obviously spared us the worst of the crash in 2008. It also missed a big part of the rebound in equity markets that started in March 2009, but of course that's not the objectve of this portfolio: the mid-term portfolio is the one designed to capture that kind of movement, and the mid-term portfolio did not miss it."
The key is that each of these sub-portfolios can have opposing positions: long equities for the long-term and to some extent short equities in the mid or short term, for different reasons, based on different tools; or positions that add directionality to the global portfolio if short, medium and long-term risks become aligned.
"It's a bit like having a steering wheel, a brake pedal and an accelerator pedal in a car," says Nakamura. "You know that there will be some corners that you will take by turning the wheel, hitting the brakes and stepping on the gas all at the same time."