AI is ‘still just statistics’

Good morning. The takeaway from Federal Reserve chair Jay Powell’s two congressional testimonies this week was a definitive shrug. Powell the hawk warned that “ongoing progress toward our 2 per cent inflation objective is not assured”. Powell the dove said we’re “not far” from a good enough inflation record to cut rates “so that we don’t drive the economy into recession”. Markets ignored it, and so will we. Email us: robert.armstrong@ft.com and ethan.wu@ft.com.

Friday interview: Cliff Asness

Three decades ago, Cliff Asness was a PhD student under Eugene Fama, often considered the originator of the efficient markets hypothesis. Long before quantitative investing was a multibillion-dollar investing complex, Asness wrote his dissertation on what would today be called “equity factors” — the measurable attributes of stocks that account for their returns. Think of value (cheap stocks beat expensive) or quality (strong companies beat weak ones) or size (small beats large). Today, factor investing — and quant investing more broadly — is big business, with lessons for the non-quants among us.

In the interview below, Asness talks quant investing, private equity, pod shops and why he thinks artificial intelligence is less than revolutionary for finance.

Unhedged: The academic background you came from has produced to major offspring, factor investing and passive investing. Passive attracts more assets all the time. Has that made your job easier or harder?

Cliff Asness: I get this question all the time, and don’t have a good answer, because nobody does. We all know we don’t know what happens if everyone tries to go passive. We had Jack Bogle [of Vanguard, father of passive investing] on a podcast. I asked him: what fraction of the market can be passive? He said 75 per cent. I go, Jack, where do you get that? And he goes, I made it up.

Unhedged: But as more of the market goes in that direction, do the factors that your strategy depends on become more or less pronounced?

Asness: Again, the final answer will be I don’t know. I always frame the question as: who’s moving to passive? On a very simplistic model, some people are good at investing and some are bad; some supply alpha and some consume it. So what fraction of the [alpha suppliers and consumers] have moved to passive? If it’s exactly proportional, I don’t think the situation’s changed very much.

I look at indirect evidence. I’ve beaten to death this idea of examining the spread between valuation multiples on cheap and expensive stocks through time. In 2000 we saw the widest spread ever, and then in late 2020, we saw even wider than that. It makes me think that the market has gotten somewhat less efficient in my career. So, on net, I lean into the idea that more smart than dumb investors have moved to passive. Because passive is actually a pretty smart thing.

I’m sure you saw David Einhorn’s [comment that] the market is so inefficient you can’t make money. I agree that maybe things are a little less tethered to reality today. But I do find it odd to say: “My job forever has been identifying errors, and now errors are so big I can’t make money.” It can change the timeframe and how much pain you have to go through. But if your job is taking the other side, if you can stick with it — obviously a rather large if — it should be more lucrative.

Unhedged: Are these longer periods of mispricing why value has struggled so much for so long?

Asness: It’s certainly a big part of it. I will brag that while our versions of value had a terrible 2018-20, since then they have actually been way better than the broad versions of value. We don’t take industry bets, and we don’t just use book-to-price. Also, global value has been stronger than in the US; people tend to over-focus on the US.

More relevant, there’s a piece I wrote called The Long Run is Lying to You, showing how any strategy is susceptible to valuation changes. If your portfolio has three- to five-year holding periods, as a value strategy will often have, and then there’s a steady richening or cheapening of your strategy — your longs getting more expensive, your shorts getting cheaper, or vice versa — it matters a lot for your realised return. This piece tries to explain what fraction of [value’s] recent returns come from this richening and cheapening. The answer: most of it. Even adjusted for this, it did somewhat worse than history, but by a relatively modest amount. Most of it is just being out of favour on a long horizon.

The problem is you have no other choice; no one knows the future. So you allocate what you think is the right amount of risk to things, because the secret is the whole stock market is just as susceptible. Maybe the most interesting example is US versus non-US developed markets. Famously, the US has crushed everyone [in the past 15 years]. During the 15 years prior to that it was: why invest in the US?

It tells you something that the stories can change so much. The US was cheaper than the world in 1990. Now the US is far more expensive than the world. Almost all of the US’s victory was from richening. You can argue if it’s justified, but you tend not to get a repeat — another 30-year relative tripling of the valuation ratio. I tell any US investor with some international diversification: you’re doing the right thing. It’s just the timescales these things work on.

Unhedged: Let’s talk about private markets. You could argue that private asset managers — the “volatility launderers”, to use your term — make it easier to be rational in the face of long-term mispricing, by making it impossible to overreact to every market swing.

Asness: It’s a fair point. I have great sympathy for the notion that stock markets are not perfectly efficient. The bouncing around is probably a fair amount of noise. What I don’t get — and a fair amount of my whining about it is professional jealousy — is why they’re permitted not to care about the bounces, and we aren’t. We absolutely know what is happening to their portfolios.

There’s a disconnect here, and I’m perfectly fine with the fix coming from either direction. Either they can mark things to market, or you can just not look at my portfolio for seven years. Your chance of being happy is extremely high, if you only look every seven years.

Let’s just assume that [not marking to market] makes people better investors, and they make more money long-term because they stick with it. Illiquidity was once, in the 1980s and 1990s, viewed as a bug, and you got a premium for it. If it has become an acknowledged feature, you should pay for that. It’s not a free lunch. You’re taking this smooth ride and you’re giving up, say, 3 per cent a year.

Unhedged: Given all this, how do you explain the success of the private equity business?

Asness: I think a lot of them are rather brilliant investors. Their ability to do things that a quant can only dream about — to go in and actually make the company better — is real. But if you started an industry in the 1980s, and then had a 40-year bull market in equities, and you really are 1.2x leveraged in equities . . . and never had the same drawdowns as other people, at least in recorded space, it’s an impressive package that has delivered.

Unhedged: And the cost of leverage was, until recently, continually falling.

Asness: Interest rates falling for 40 years is this massive fact we all have to deal with. The multiples on stocks in general have gone up. I’m not a paranoid “end the Fed” guy, but do I think 10 years of highly suppressed interest rates may have contributed to some irrationality? In the case of private equity, I think they definitely had a tailwind.

Unhedged: One last argument in favour of private assets. The worst things that happen in finance happen to you all at once, over like three days. Just being able to do nothing for those three days is powerful. You don’t get a call on your assets, nobody asks for margin and you ride out the storm.

Asness: I don’t disagree with a word of that. But my worries still apply, especially if that means you end up taking more equity risk than you would have otherwise. There are also things that aren’t three days. My other worry, again, is that the illiquidity premium might be much smaller or even negative now.

[People make the argument that] we all have to bury our heads in the sand because it’s good for us. I wish they’d spend half their energy on that and half their energy on telling people, if somehow you could overcome [your urge to sell at the bottom], you could actually do a little bit better. Maybe more than a little bit better.

Unhedged: Changing topics now, as a student of Fama’s, how is it possible that Jim Simons exists? [Simons is founder of Renaissance Technologies, whose Medallion fund has beaten the market by wide margins for decades.]

Asness: Amount of dollars matters a lot. Fama himself does not tell you markets are perfectly efficient. I took his class and then TA’d it [worked as a teaching assistant] for two years. Every year he’d look at the class and go: “Markets are almost assuredly not perfectly efficient,” and you’d get a gasp. Only at the University of Chicago would you get a gasp! Of course he’s right. The efficient market hypothesis is an extreme hypothesis. He thinks markets are a lot closer to efficient than I probably do; I think they’re closer to efficient than the average active manager.

The fact that Jim Simons has thrown out all the clients [from Medallion, which only manages Renaissance employees’ money] is important. I’m not downplaying the fact that Jim can take $2bn or $3bn out of the market with a high degree of certainty every year. To be clear, I would trade with him. But it is a drop in the bucket in terms of market efficiency.

Value working, if you believe it’s behavioural, creates far more dollars of alpha. It’s simply a 1.5 Sharpe ratio [it generates 1.5 units of excess return per unit of volatility], not a 3.5 Sharpe ratio. So it doesn’t feel like an inefficiency. If Jim could take $200bn out every year with a 3 Sharpe ratio, that would be a lot harder to stomach. But when it comes to market efficiency writ large, dollars are what counts.

With factor investing, I think we can add alpha on $200bn. I don’t think Jim can, with his Medallion strategy.

Unhedged: We’ve been interested in reading about the rise of multi-strategy hedge funds, or “pod shops”. It seems like their clients care about returns, but they care more about the lack of correlation to markets. Is that true of you guys?

Asness: I admire many of those shops. It’s a different way to skin a cat. And say somebody had told me a decade or two ago that we’re going to pay people a lot, and if they have a mid-sized drawdown, we’re going to fire them. I would’ve said that’s not going to work. Every strategy has drawdowns. Clearly I would’ve been wrong! What that says is the people who have been great at this — the Millenniums and Citadels of the world — have to be providing some alpha in choosing who to run the pods. 

Compared to the pod shops, the bad news is our strategies don’t have a 3 Sharpe ratio; the good news is we have very high capacity [for investor funds]. The multi-strats are a harder to scale business than a factor business. I wouldn’t short one of these firms, but it’s going to be harder for them going forward.

Unhedged: What do you make of tail risk hedging shops?

Asness: You’re trying to get me to fight with Taleb [Nassim Taleb, an adviser to Universa, a tail risk fund]. I used to say to him: long vol [a position that profits when volatility rises] shouldn’t get paid. Maybe the hedging properties are good to have. But it’s insurance — you should pay for insurance, you shouldn’t get paid for it. Nassim would say, well, there is a fair amount of alpha to how you construct the options strategies. And I will not impugn another man’s alpha. But that doesn’t mean I believe it all the time.

Unhedged: You’ve argued against those who claim the optimal portfolio for the long-term investor is 100 per cent equities. But what about Warren Buffett? He has 100 per cent equities and he’s the third-richest guy in the world.

Asness: First, at no point did I ever say 100 per cent equities won’t be a good long-term portfolio. I say that a diversified portfolio that achieves a higher risk-adjusted return by applying mild leverage will be better. So if Warren Buffett had 90 per cent in equities and added 60 per cent bonds [summing up to greater than 100 per cent because of leverage], I believe he would’ve done better.

Unhedged: As journalists, we’re very worried about AI replacing us. How about you?

Asness: We don’t think AI, at least in our field, is as revolutionary as others do. It’s still just statistics. It’s still a whole bunch of data going in and a forecast coming out. Some of the key things we’ve talked about here I don’t think AI will help with at all: what is the premium for high-quality versus low-quality stocks? Value versus growth? We don’t have a big data problem there; we have a small data problem. If we had 8bn years of stationary comparable markets, we could answer these questions with any kind of statistics.

We use AI mainly with natural language processing. For years, quants have looked for momentum not just in price, but in fundamentals. This was done by analysing the text of corporate statements to look for positives. The old way to do it was with a big table of keywords. So “increasing” gets plus one point, and so on. You can see the flaw: if it’s “huge losses have been increasing”, whoops. Natural language processing has made that way better.

Unhedged: So the innovation is that you’re using fundamental momentum to supplement price momentum?

Asness: I think “supplement” might understate it. We do fundamental momentum as a standalone factor, for each company. If you parse each company’s statements, is it net good or net bad? Most news gets incorporated into the stock price, but not all of it.

This is really a standalone signal. When I talk about the families of factors, one family is fundamental momentum. We’re using it as almost an equal partner to price momentum. Fundamentals aren’t better, but they’re as good as price, and not perfectly correlated.

You can also do fundamental momentum at an asset class level, measuring trends in economic data that impact prices. This preserves a very important property: many people who invest in trend-following are looking for positive convexity. They’re looking for something that tends to do particularly well when the world has a really crappy period.

Price momentum will by definition get a sharp inflection point wrong. For example, price momentum would’ve shed long positions after March 2020, and then gotten whipsawed. Fundamental momentum does a bit better on that score. Conversely, if a price trend just keeps going, but it’s going to stop because the fundamentals have started to deteriorate, fundamentals will help you.

We still like price momentum among the four major asset classes — stocks, bonds, currencies and commodities. But now we give about half our weight to fundamental momentum, too. Ten years ago, we gave all our weight to price momentum. That’s a gigantic change, and it’s the simplest thing in the world.

Unhedged: The core concept seems brutally simple. How much can you really change about the trend-following process?

Asness: It does sound childishly simple, but in trend-following we’ve just been doing a lot more things, more esoteric securities. The best example is alternative commodities such as milk, coal or my personal favourite, Malaysian palm oil. It turns out if you do a vast number of them, even if they’re all lower capacity, you have decent overall capacity.

The second thing is more structured trades. One simple example is trading the trend of the shape of a yield curve, not just the direction of a bond market. That is a little trickier to do, since some parts of the curve are more volatile than others. But we do think trend-following applies nicely.

You can also do trend-following in the classic equity factors, like value, quality or size, which tend to trend on a one- to 12-month basis. The beautiful part about factor trends for a trend-following strategy is that they’re already attempting to be market-neutral.

If 10 years ago you said one of our bigger improvements would be in simple trend-following, I would have been shocked. But I think we’ve made it much better.

One good read

The wisdom of crowds of robots.

FT Unhedged podcast

Can’t get enough of Unhedged? Listen to our new podcast, hosted by Ethan Wu and Katie Martin, for a 15-minute dive into the latest markets news and financial headlines, twice a week. Catch up on past editions of the newsletter here.