The benchmarks that are supposed to measure performance and create alignment with end investors are working against asset managers, and the industry must find new ways to demonstrate value before it’s too late, according to MFS.
The $341 billion AustralianSuper has taken a significant stake in US-based DataBank to catch the AI and digitalisation waves that pension funds all around the world are trying to surf.
Cliff Asness: Markets are getting crazier – and more rewarding
Markets have become less efficient, and social media is to blame.
Or so says Cliff Asness, founder of hedge fund AQR, who recently penned a paper for The Journal of Portfolio Management titled “The Less-Efficient Market Hypothesis”. The CliffsNotes version of his thesis is this: with more clueless punters on social media piling into beaten-up stocks like GameStop, and fewer old school stockpickers doing price discovery, things have gotten badly out of whack.
“It’s two sided; the people who are supposed to be thinking about value are fewer, and the people who are influenced by these crazed mob situations are greater,” Asness tells ISN.
The assessment doesn’t come with “tables and tables of t-statistics”. And more indexing, alongside lower rates, could also have had an effect. But if he’s right – and Asness concedes that being precise about the magnitude of the change is very difficult – markets are going to get crazier, and stay that way for longer.
The implications of that are also two-sided: it will be harder to make money, but there will be more money to make.
“It’s virtually a tautology that if you’re a rational investor and, on average, right, and the masses are wrong, more inefficiency should lead to a more profitable long-term strategy,” Asness says. “If people are going to make bigger errors, and your job is to take the other side of those errors, if you can stick with it, which is the rub, it should be more profitable long-term. But that rub is very real.
“That strikes me as fair – more profitable but harder to do, because the storms you’ll have to weather will be more severe and, as important, last longer… Having experienced bad periods of anything from a day up to two and a half years, I can tell you the length matters at least as much as the severity. Having to go back to the same investors, even for yourself, having to convince yourself of the same story; your confidence can be tested.”
It’s not for no reason then that more and more investors are piling into private market assets, where the effects of market volatility are somewhat less visible – though no less impactful.
“By no means do I deny the psychic and therefore somewhat real utility of volatility laundering,” Asness says. “I never claim I don’t understand why people do it even though it doesn’t help some people. Understating risk works really well over short horizons where things bounce back. If we have something we haven’t seen in a long time – a long-term bear market – that gets much harder to do.”
“It’s the old Warren Buffet quip about only when the tide goes out do you find who’s been swimming naked. You will find that if you’re using a 4 per cent vol on your private assets and a 16 per cent vol on your public assets you’re somewhat over-allocated and surprised at those times.”
“Machine learning is not going to triple our risk-adjusted returns… And there are some problems I don’t think it will ever improve upon – what is the true equity risk premium?”
Asness isn’t a prophet of doom, and thinks that markets will eventually adjust to the effects of social media. Until then, AQR is adjusting by doing more with machine learning (ML) and using more alternative data sources, both of which produce strategies that have “a fairly different flavour” to the traditional long-term investing that will be periodically disrupted in crazier markets.
Cloak and dagger
What alternative data does AQR use? Asness jokes that Andrea Frazzini, the firm’s head of global stock selection, won’t even provide him with that information. There’s a good reason: unlike classic valuation strategies, which are “very resilient” to being arbitraged away, the usefulness of alternative data tends to decay quickly when it finds its way into the hands of too many investors.
It typically comes from vendors that only sell to a small number of clients – sometimes just one. The classic example, which “everybody knows about”, is credit card receivables; of the rest, all Asness will say is that it’s unlikely to be interesting to those outside the asset management industry, it’s typically shorter-term in nature, and that some of it is expensive enough that purchasing it can erode any alpha that might be earned from using it.
“It’s a weird dance to decide if you’re going to buy that data. Quants want to test everything but they won’t give it all to you. It’s a far more bespoke activity, and we’re far more tense about it being arbitraged away. Without sharing details, there are sources that we’ve already stopped trading after making money from for a few years.”
“The main challenge in traditional investing is sticking with a good process. In alternative data, the main challenge is continually finding the next one. Some of the places we buy it from, part of the agreement is we don’t disclose the data… it’s suddenly a cloak and dagger world.”
Somewhat less cloak and dagger is AQR’s use of ML. But while there’s been enormous advances in computing power, Asness cautions against getting carried away with the artificial intelligence hype, which is already producing what he calls “ML-washing” – where one per cent of the investment process can be described as ML but it’s suddenly all over the fund branding.
“(ML) is not going to triple our risk-adjusted returns… And there are some problems I don’t think it will ever improve upon – what is the true equity risk premium?”
But AQR is now pretty heavily invested in it, even though Asness reckons he slowed the firm’s efforts down by about a year.
“Ex ante, I think it is the role of the old man at the firm who’s been doing this a long time to say ‘you want to change things in big ways? Let’s go slow on that’. In that sense I don’t think I was doing the wrong thing. But AQR has always pushed and bragged about being quants but not going purely where the data leads. If something works historically, but has no economic or even commonsense story for why it should work, we either don’t believe that or believe that considerably less… we push that to avoid overfitting.”
The use of ML doesn’t completely change that approach; AQR still puts the most credence in the economic story, but less than it has in the past. The point of using ML is to let the data speak, and if you could intuitively understand everything it said “you wouldn’t need it”.
“For me it was a little bit hard – you do have to be willing to let go of some things you’ve cherished for a long time,” Asness says. “I just got more and more convinced, not that you want to bet the ranch on it, not that you want to completely stop doing the old school stuff that will be crazier but better than the past, but I’ve gotten convinced by the evidence.”
