ARP report card: Assessing diversification and performance
In the first blog in this series, we introduced ARPs and a few of our key philosophies when incorporating them into multi-asset portfolios. In this instalment, we want to assess how ARPs have performed and what we have learned from our years of integrating them into our strategies.
Before we address performance directly, it’s worth covering what to expect when considering ARP strategies. Given their quantitative nature, it’s not only easy to produce a back-test of a modelled strategy but in fact to iterate a model specifically to produce a favourable back-test.
This ‘data mining’ can create overfitted and overly optimistic results with apparently outstanding risk and return characteristics (think Sharpe ratios above 2). Academic research, which we have replicated and updated, shows that we should expect a haircut of 75-100% from the expected Sharpe ratio when turning a back-tested ARP live.
That means most – if not all – of an ARP’s return advantage over risk-free assets can be lost. We will discuss the importance of avoiding data mining in a later blog. In reality, we are not seeking strategies with such high Sharpe ratios; rather, we believe that combining a number of diversified strategies with reasonable risk/return expectations can create an overall package that is attractive to multi-asset investors.
So what have we, and other investors in ARPs, experienced over the past five years? Well, even allowing for such drastic Sharpe ratio haircuts relative to back-tests, performance has on the whole been below expectations. At the industry level, ARP strategies have on average delivered negative returns through this time.
However, just as it would be a mistake to trust excellent back-tests, we believe it is too early to reject these strategies as unsuccessful solely on the basis of a five-year period. This has been a challenging period for ARPs, but we conduct regular, sophisticated statistical testing on whether the realised track records and live data for our ARP strategies are consistent with the back-tests including haircuts. This robust process suggests that the returns we have experienced recently are within the normal range of expectations, albeit at the low end.
Indeed, our ARP strategies have provided positive returns in 14 of the 25 quarters since their launch and have delivered flat returns overall, beating the majority of similar strategies out there. Nonetheless, we certainly seek improvement here. The main negative contributions have come from a small handful of strategies. We believe that timing exposure to different risk premia is difficult, so we do not seek to increase or decrease the size of exposures in portfolios frequently, except to manage overall risks in portfolios.
However, we do think it’s important to look out for structural breaks in markets that may make the continued performance of some strategies more challenging. One such case arose for currency carry and momentum strategies.
As interest rates around the world have moved closer to the effective lower bound and differentials have narrowed, the potential for carry trades has been marginalised. Increasingly, policymakers are also turning to managing their currencies in order to manage their economic competitiveness, either with verbal or physical interventions in FX markets.
This means that any currency momentum is quickly squashed by policymakers, so in turn currency momentum strategies have become challenged. We therefore stopped following those signals five years ago, a decision we believe has been vindicated as the struggling strategies would have continued to deliver losses over the subsequent years.
We also replaced slow-moving momentum strategies with a much more responsive, defensive strategy in the midst of the COVID-19 crisis as we did not have faith in their ability to keep up with such rapidly evolving market dynamics.
The diversification effects of ARPs, on the other hand, have been more in line with our expectations and have in our view improved investor outcomes. On average, the strategies have had no loading on equity or duration risk, and showed close to zero average correlation to other return streams in multi-asset portfolios – namely, market risk and tactical asset allocation. In periods of volatility, such as the fourth quarter of 2018 and March 2020, ARPs offered relative tranquillity.
We believe ARPs can bring genuine diversification to multi-asset portfolios. The returns they generate are not only different from traditional market risk premia; their systematic nature also differentiates them from a people-led discretionary macro investment process.
This is demonstrated by their low correlation of -0.2 with our tactical trades. This has been especially important in periods of drawdown, when ARPs have been positive or at least remained flat. This return profile is arguably the key reason for us to keep using them.
That said, we do believe it is important that ARP research is embedded in a team that can understand and interpret the macroeconomic environment in order to determine when market dynamics may change and when certain risk premia may fail to deliver. My colleague Willem will cover this in more detail in the next blog.
Overall, when ARPs are integrated into a broader portfolio, we believe risks need to be understood holistically. ARPs are therefore best considered not in isolation, but in our view as an element within a portfolio of complementary risks and returns.
Past performance is not a guide to the future. The value of an investment and any income taken from it is not guaranteed and can go down as well as up; you may not get back the amount you originally invested.
 Source: LGIM internal analysis, data to 31 March 2021