By Hamlin Lovell, NordicInvestor
Alternative Risk Premia is mainly systematic but it involves many active choices that result in a wide variety of approaches. Nordic Investor interviewed Veritas Pension and Insurance’s new CIO, Kari Vatanen, to discuss some of the variables investors can research when comparing risk premia managers and strategies.
Vatanen’s investment philosophy sees space for ARP in the context of a diverse range of approaches. He expects a balanced portfolio could contain strategies ranging from purely passive index tracker type management to discretionary management, and from traditional high fee systematic strategies to lower cost alternative risk premia strategies. For instance, “It is hard to beat the S&P 500 using either discretionary or systematic factors, so it may be better to take a passive approach and cut costs”, he points out.
Bond Market Bias
Vatanen previously spent nine years researching and implementing systematic risk premia strategies in house at Varma Insurance, rather than allocating to external managers for this strategy. Vatanen builds his own ARP strategies partly because the space shows a wide spread of exposures and performance. “Some ARP strategies have a high tail risk dependency and bond market bias”, he points out. This is not surprising: after a more than 30-year bond bull market, a high bond weighting makes strategy back-tests look good for many tail risk strategies that have lots of long bond exposure. Going forward, in a world of negative interest rates, we may question whether bonds are likely to provide a useful portfolio hedge. Recently, in March 2020, government bonds have sometimes been selling off in line with equities, for at least two possible reasons. Investors need to raise cash from anything they can sell, and may not be able to absorb the volume of issuance arising from increased government borrowing to address the Coronavirus pandemic.
But there is no straightforward solution here: “reducing bond beta tends to increase equity beta, and vice versa. It is very hard to avoid all equity and bond beta. So it is better to just accept that there is some bond and equity tail dependency”, he says.
ARP does not have to exclusively invest in equities and bonds of course. It can also be exposed to currencies and commodities. Vatanen argues that smaller markets, including commodities, can be richer sources of inefficiencies, but this must be traded off against their lower liquidity. He thinks it is worthwhile trading the most liquid 20 commodity and energy markets.
But even within the same markets, there are different ways to define factors. “The most important thing is how the factors are defined”, he says. “Some managers use a traditional academic definition, but markets evolve over time. For instance, they may define equity value as being price to book value and this worked okay for tangible assets in old economy industries (countries). Now that some technology and internet companies have no tangible assets, book value is irrelevant, and metrics such as market capitalization or earnings may be better”.
Types of Risk Premia
Vatanen also distinguishes between fundamental, behavioural and structural risk premia: “fundamental premia include a risk-based explanation. This might be high tail risk correlations with equity markets or risky assets. Behavioural premia are mainly based on investors’ behavioural anomalies, which are not strictly rational. Trend following is an example. Structural inefficiencies are caused by seasonal flows or imbalances, and are more temporary inefficiencies, and as such are not as robust as fundamental or behavioural premia”.
The allocation to the different “families” of premia will vary according to the investor mandate. “Fundamental premia are the most persistent, even when markets are changing. Behavioural premia may or may not work. Structural premia need to be more actively and tactically traded”, he has seen.
There is a portfolio diversification argument for maintaining exposure to all three types of premia. “They diversify each other a bit. For instance, fundamental premia have higher tail dependencies and structural premia have weaker tail dependencies”.
The portfolio construction methodology for ARP strategies is another active choice. Vatanen argues that, “19th century, linear statistics may not be appropriate because in real life dependencies are rarely linear. Traditional statistics do not consider these non-linear relationships”.
Even so, Vatanen respects that fact that traditional statistics is robust. “Machine learning also has its own pitfalls, of over-fitting and over-optimising, which can result relationships being identified that are essentially just statistical “noise”.
Therefore, machine learning techniques have to be used very carefully, on a “supervised” basis: “we define a framework really early on, based on clear rules, and then use machine learning techniques to find dependencies in it”.
One use of machine learning is to arrive at more sophisticated factor definitions: “for instance, some factor definitions, such as the quality factor, can be improved by using a cross sectional dataset”, he points out.
One of the big debates in ARP and factor investing is whether to time factors. Here, AQR’s Cliff Asness famously argues it is difficult to do so while Research Affiliates’ Rob Arnott is an advocate of factor timing. Vatanen is in the skeptical camp: “it is hard enough to time between equities and bonds. Factors are much more complicated”, he says.
Timing apart, trade execution can be important in terms of controlling transaction costs. “We find some bank ARP products could be trading at the end of the day at the closing price, which can create a crowding effect which means they do not obtain good prices”, he says. Vatanen is mainly trading in listed markets and not OTC markets that can entail additional complexity in ARP strategies.
Trading cannot necessarily be completely automated and may require human input. Though ARP is perceived to be a fully systematic strategy, some sorts of premia may require some degree of discretionary input. “For instance, VIX and VSTOXX trading is half way between systematic and discretionary. It may require more tactical trading”, he says.
Fees are hotly debated in ARP and a wide range of fee levels and structures exist. Some of them are not much lower than traditional hedge fund fees, while others are very low flat fees with no performance fee.
“It depends on the complexity of the strategy. If it is easily replicated, then the fee is only really for the convenience of outsourcing. For more complex strategies, or portfolios that cannot be managed in house, it is easier to justify higher fees”, says Vatanen.
The average ARP strategy was down about 10% in 2020 as of mid-March, according to the SG Multi Asset Risk Premia Index. Is ARP in danger of becoming a victim of its own success? If too much money chases the same inefficiencies that may disappear. But the jury is still out, according to Vatanen: “it is possible that inflows have already diluted returns in some areas of ARP where factors have become overcrowded, but we cannot be sure – only time will tell”.