The Bridge Ep. 11: Expensive Isn’t a Bubble

The Bridge, EP 11, AQR, Cliff Asness – Transcript

Sonali Basak (00:00:00 -> 00:00:02)
Part of this conversation should be how you define quant.

Cliff Asness (00:00:03 -> 00:00:03)
It’s all changed.

Sonali Basak (00:00:04 -> 00:00:22)
We’re about to talk to Cliff Asness, who was really a pioneer in the quantitative investment world. He runs AQR capital management that has more than 240 billion in assets under management. And we’re going to talk about the markets. Are we in a bubble? Are we not in a bubble? And what happens if things go south?

Cliff Asness (00:00:22 -> 00:00:27)
Saying there are bubbles and saying you can make money from bubbles are not the same thing.

Sonali Basak (00:00:45 -> 00:01:26)
Welcome to the latest episode of the Bridge by I Capital. I’m Sonali Basak. I’m the Chief Investment Strategist at iCapital, and today I am joined by Cliff Asness. He is the co-founder of AQR Capital Management, 240 billion in assets under management. And many people know you and AQR as pioneers of the quantitative investing world. We’ll talk about the future of quant and the age of ai, but first I want to address one topic that is really, I don’t winna say plaguing investors, but okay. But a real problem. And it’s the diversification problem. In a world where stocks and bonds have become more correlated, how do you diversify? Do you need to diversify?

Cliff Asness (00:01:26 -> 00:03:28)
So well, first of all, pioneering is a very polite and nice way, and I appreciate it of saying old . So, um, but it’s just true. Um, yeah, we were early 1990s quants. So, we’ve been, we’ve been doing this a while. We have younger, smarter people to do the modern stuff. So not to worry, diversification, the way we approach it is our job for many years, and we do this in different forms. Sometimes it’s equitized, sometimes it’s standalone, uncorrelated to the market. But we’re trying to create returns that are not very correlated to the market. You always want to add that to a portfolio. If stocks and bonds are negatively correlated, we’re greedy in the famous efficient frontier where you have risk here and return here, you always want to move up to the left. So, if you find something that’s not very correlated, that on average makes money, you want to add that is the need a little bit greater when stocks and bonds are giving you maybe a little bit less diversification than normal? I actually think people exaggerate this. A typical 60 40 portfolio of stocks and bonds. I don’t, have you ever met someone who actually owns 60 40, right? In perfect amounts, ? Yes. It’s this, um, it’s this, uh, Shibboleth. We all have to, we all have to use, I think we get sued if we use 50 50, right? Um, we have to say 60 40, but a 60 40 portfolio stocks are about three times more volatile than bonds. Bonds in their normal form. You can buy a a 30 year zero. I’m not talking about that. But if you buy, say, the 10-year treasury, you’re mostly diluting equity returns. Not diversifying the direction of bonds. Doesn’t matter. It matters, but not quite as much. So, the correlation change is bad, but it’s not quite as extreme though on that, between that and another thing, the fact that we think the stock market’s fairly expensive versus history, we’re not market timers. We’ve written a lot at timing The market based on pure valuation is, I, I’ve, I’ve tried to do this. It’s a suicidal quest. Well,

Sonali Basak (00:03:28 -> 00:03:33)
I think also, I’ve been talking to you now about whether we’re in a bubble or not for the last three or four years.

Cliff Asness (00:03:35 -> 00:04:10)
I have screamed bubble twice in my career. And this is a hard thing for a a, I was a disciple of Gene Fama, the Efficient Market hypothesis. He, he co-chaired my dissertation committee. Um, but even Gene will tell you, markets aren’t perfectly efficient. He doesn’t like the word bubble. Uh, but I, I do think I’ve lived through two in my, in my career, um, the, uh, the.com or tech bubble at the end of the nineties, uh, peaked in March of 2000. I don’t want to talk about how young you were at the time, but, but I was in hell.

Sonali Basak (00:04:10 -> 00:04:21)
Well, it was so funny, we were talking about oldness and stuff. You know, you were just on a podcast talking about how you were doing the Rolling Stones tour. Yeah, no, this is the Aris tour. Okay. We’re going to make you Taylor Swift. We’re talking about the new stuff here. .

Cliff Asness (00:04:21 -> 00:04:43)
I think I, I, I think I’d rather be Taylor Swift than the 82-year-old, uh, Mick Jagger, 80 I don’t, I saw him at Jazz Fest like two years ago. I think he had turned 80. It’s freakish how good he’s,

Sonali:
you can keep going.

Cliff:
He’s jumping around the stage. I’m like, I’d be whining about my back and holding things. And I’m 23 years young. Uh, uh,

Sonali:
but you have

Cliff:
not a quant topic.

Sonali Basak (00:04:43 -> 00:04:53)
You’re not 80, but you’ve seen the late nineties. And you know, it’s interesting, I’ve heard you talk about this, where you’ve made it through a currency crisis at that time, but then you ran right into the dot com bubble.

Cliff Asness (00:04:53 -> 00:06:40)
Yeah. Um, in August of, of 98, Russia defaulted on its debt. Uh, certainly emerging market currencies were in were crazy. But the biggest thing was in August of, uh, of that year, stock market dropped about 20%. And I often describe this as the crash that nobody remembers. Um, people remember like October 19th of ‘87 when the market fell in the low twenties in a day. And the reason I think people don’t remember it is two things came back fairly fast and people remember paying that’s long lasting more than short second, it just didn’t happen in a day. It was fairly steady over the month. And you never get that newspaper headline. You know,

Sonali:
it’s black Monday,

Cliff:
The Tuesday after Black Monday.

Sonali:
Exactly.

Cliff:
Um, but that first month we were up and, um, we were just starting to get invested and we’re, and we were a little bitter ’cause we’re like, we would’ve been up a lot if we were fully invested, whatever. But still, we claim to be market neutral. First month of our existence, the market drops like 20% and we’re up. So, we’re high fiving. It’s like proof of concept. Um, I learned a, a lesson that I think I’ve been good about since then. Never high five in this business. You get to high five once when you retire and you’re like all in safe assets and you’re not doing this anymore. You can high five once. And even that can be a little dangerous ’cause the market gods are listening ’cause the next 18 months were the blow off top of the.com bubble. Um, and that was, you know, round trip. I think we made a lot of money. Um, it wasn’t, uh, a black swan. It wasn’t as geeks would say, a 10 standard deviation event. It was a two standard deviation event. But you don’t want to have a two standard deviation event in your first year and a half in, in business. You never want to have one.

Sonali Basak (00:06:40 -> 00:06:55)
So, you’ve learned a lot then over the last few decades.

Cliff:
Yes.

Sonali:
Right. Um, this idea of market neutral, you, you started this conversation about saying that you want to create uncorrelated returns. How do you be uncorrelated in this market?

Cliff Asness (00:06:55 -> 00:08:12)
, the, um, uh, let’s only talk about individual stocks. Uh, we can talk about macro also. Um, it’s just simpler to talk about one set of what we do. The simple way, um, which not everyone does, is if you want to be uncorrelated, start out by being long and short, a very similar amount of stocks. Um, if there’s a big difference in the so-called betas of the long and the short portfolio, you may want to apply a hedge ratio. Um, but if they’re similar betas, if you’re long, a thousand stocks and short a thousand stocks, which is not atypical for us, uh, a typical portfolio for us is global. It is, again, don’t hold me to a thousand. There are times it may be a little more, a little less, but it’s long. A thousand stocks short, a thousand stocks, mainly balanced by industry and country. Not entirely, but mainly. So, one way to be neutral is to, to, to actually take the actions to neutralize the market part. And in fact, a bunch of other parts, the country part, the industry part, it doesn’t make it riskless. There are times where you choose what you think are all the right stocks, condu, country and industry neutral and, and the wrong stocks win. Uh, but we do think it wins on average. And we do think it makes it a very, a good diversifier. ’cause you’re not as buffeted by all these other things.

Sonali Basak (00:08:13 -> 00:08:29)
Maybe you can bring us behind when you’re long and when you’re short. You and I have talked about this a lot. I think a lot of people from the outside look at at this world and say, oh, well, when hedge funds are shorting, they’re making a targeted short. But you’re not, um, doing that kind of single stock analysis from where you sit. So, what are the qualities?

Cliff Asness (00:08:29 -> 00:09:49)
Yeah. Um, quants always non-intuitive to to, to people. The qualities we would look for in a long and a short are relatively symmetric. Not always, but basically the shorts are bad on what the longs are, are are good on. Um, I won’t take you through everything. We do some things that are more proprietary today in the world of, of alternative data, the world of, of natural language processing and some of the AI tools. Um, AQR has always published and talk about things, and we still do, um, I think more than others, but there we’ve never talked about. If you, if you find something that you think is actually relevant, unique alpha, even AQR will not write about it. So, I will talk about some of the more publicly known things, but they’re still a big part of what we do. The academic literature and the kind of quant Wall Street literature. Um, over time it has really grown. When we started AQR, you were looking for really two things. And this was, it sounds really simplistic, uh, but it was also the state of the art of the time you were looking for low, multiple good momentum stocks. Um, those were the two major findings in academia. The third was that small stocks outperformed. We never believed in that one particularly. Um, but yeah, uh, low, multiple, so-called value stocks, which I think, I think that’s a misnaming by the way,

Sonali Basak (00:09:49 -> 00:09:51)
Right? We can get into value versus growth, for sure

Cliff Asness (00:09:51 -> 00:10:00)
Um, But, but, but even if it’s misnamed and it doesn’t capture all of value, I do think on average paying less is a good thing. And particularly paired with momentum.

Sonali Basak (00:10:00 -> 00:10:07)
Have you seen that meme going around on the internet where people are throwing out the Benjamin Graham book? You’ve never seen this?

Cliff Asness (00:10:07 -> 00:10:09)
No, but it’s an idiotic meme. .

Sonali Basak (00:10:10 -> 00:10:11)
I think a lot of people,

Cliff Asness (00:10:11 -> 00:10:17)
People, people have thrown out that book to their detriment quite a few times. Um, I don’t know how long it’ll take, but, uh,

Sonali Basak (00:10:17 -> 00:10:18)
People have lost the art of value.

Cliff Asness (00:10:18 -> 00:12:02)
Yeah. Well, part of the evolution of quant is ironically getting more like, like Benjamin Graham, um, people will quote him and, and, and, and the, and the Graham and Dod book and his other book, uh, and they’ll make it sound like he’s the deep value, cigar butt destroyed company, but selling for pennies. And it was never that, it was always closer to what Buffet and Munger talk about. Um, uh, and, and Buffet and Munger, they, they are really good about this. Um, sadly, Charlie has passed, but, but I don’t want to refer in the wrong tense. If you watch those two, I go, I’ll go on like deep YouTube rabbit holes where I’ll watch, watch these guys. Um, Buffet would quote, would, would credit Munger with changing him from kind of a cigar butt looking for super cheap to looking for good companies at a good price, not at a dirt-cheap price and quant. Uh, we had to learn our lessons slowly. Um, ’cause a lot of quant, you think of the quote factors that have been added to the quant universe. It was just, just valuation multiples not full value and momentum in, in 1990. Um, profitability measured all kinds of different ways. Risk, looking for low beta or low vol, low earnings variability. Um, those two things explain more of Warren Buffett’s returns than value does.

Sonali:
Hmm.

Cliff:
Um, we wrote a paper on this. Three of my colleagues wrote it. Uh, some, I didn’t make it as a co-author and it’s one of our best papers.

Sonali:
Is it the insurance one?

Cliff:
No, they talk about his insurance float in there. That is part of it. ’cause part of what he, he did was, uh, had a pretty ingenious way to finance, right, what he did.

Sonali Basak (00:12:02 -> 00:12:04)
Genius cost of leverage. Yeah.

Cliff Asness (00:12:04 -> 00:14:09)
Yeah. Uh, and everyone says that kind of dismissively like, ah, that’s all he did. And I’m like, yeah, he did it. Nobody else thought of it. And he did it for like 40 years. Um, you get credit for that.

Sonali:
Yeah.

Cliff:
You don’t get to just go, I, you know, you know, it’s like I can go, well, Jim Simon’s at Renaissance Medallion and all he did was, you know, he was early to short term. I’m like, yeah, cliff, you could have done that, and you didn’t. So, I, I wouldn’t say that a a about a, about Rena, people are too dismissive. Um, but the paper said, we know he is not a quant, but let’s pretend we have a return series. And we’ve done this for other famous investors. There’s a series on our, our website saying, if you use the quant factors simply to explain, if you have a style to what you do, even if you’re picking individual stocks and owning a concentrated portfolio, if you’re looking for reasonable multiples and profitable companies with low volatilities and low betas, you end up pretty correlated to the quant versions in a given year. If your particular picks do great or terrible, you can deviate. But and we found, we explained a lot of his returns. This is very rough, but call it 40% looking for more profitable, uh, companies, 40% looking for lower-risk companies and about 20% valuation. So, valuation, he does, you think of them all at once, but you can also think of it as a tie breaker, not a deep value person, but a, am I paying a reasonable price value? Gets a say, but, or multiples get a say I, by the way, um, you’ve probably heard me say this before, but I don’t think the quant should have called low multiples. The value factor, it’s not. Value is more holistic. The Graham and Dodd people, uh, Warren Buffet are right. Um, you can have a high multiple company that is value if you think it’s going to grow enough, if you think it’s current profits or are gigantic and it’s going to grow from, from here. Um, low multiple still works ’cause on average people just go too far. Um, but I, I think of quant, a big part of quant evolution. The, the less fancy, less modern part, just evolving the factor set has been getting more and more like a full holistic investment process.

Sonali Basak (00:14:10 -> 00:14:45)
So, translate that into today’s market because what we’re talking about is how to find different versions of value. Would, if you’ll give me that, because if you look at the market, a lot of people are talking about how, you know, 10 companies making up 40% of the indexes right now, or, you know, the s and p 500 at least, um, more so right when it comes to some of the other indexes. So, but if you split it up, actually, it gets so much more industry across industries, across size of companies. And so, you could argue that we are kind of in this era of pretty extreme dispersion. So, what does that mean for somebody like you?

Cliff Asness (00:14:45 -> 00:15:19)
Sure. First, in most of what we do, the concentration of the index does not bug us very much. Because if you have a thousand stocks long and a thousand stocks short around the world, equal weight is going too far, we have weighting schemes, but it, but it’s closer than, than say, market cap weighting. What seven stocks are is close to irrelevant. We will bet on those stocks. I generally won’t know if you ask me, which would be very embarrassing if you did right now, what’s your position on Nvidia? I I would say we probably have one . Uh, but we’re quant.

Sonali Basak (00:15:19 -> 00:15:20)
You don’t know If it’s long or short or

Cliff Asness (00:15:20 -> 00:16:33)
What. I have a thousand long and a thousand shorts. Right? It’s, I I’m deeply invested in the characteristics that, that the long portfolio looks better than the short on all the things we care about. And the hedging, that we’re not taking giant, like we said, market or, or country bets. But the concentration in a liquid alt, for instance, does not bother us in a long only beat the benchmark portfolio, which we also do run concentration matters in the following way. Negative views are hard to express in a long only, all you could do is not own something. Um, if you have a negative view on Nvidia, you can express that. ’cause it’s such a big weight in the index that for instance, if you don’t own it, that’s a fairly big bet. If you have a negative view on the, even the hundredth biggest stock in the S&P500 is still a pretty big company. Its weight is small and you can’t move the dial very much. And when the market gets concentrated, that effect gets bigger because all the small ones are even smaller. Um, so it does constrain the, the, the traditional law only manager. Uh, I don’t think it’s damning. Uh, it’s just, uh, makes it a little harder. I don’t think it has hardly any effect at all. Um,

Sonali Basak (00:16:34 -> 00:16:35)
Is it better for you in some ways?

Cliff Asness (00:16:36 -> 00:18:47)
It, it could be better or worse. I was about to add. It has no effect at all mechanically. Um, and, and that’s the first thing you worry about. Uh, can you make the bets you want to make? If it’s indicative of a certain market environment, if it says, this is getting back to your question, is it a bubble, which you still haven’t answered ? Um, if, if it’s indicative of a bubble or if it comes along with a bubblish attitude, um, that can affect what works and doesn’t work. So, it might be coincident with things. If you tell me the market got super concentrated and, and people threw value out the window and the only thing that worked was momentum, I think that’s part of the story. So, it can say something about the market, but it has not bugged us to date. Um, I will brag for a second. Our versions of value since COVID, I have been pretty good. That’s not genius on our part. I think other quants might tell you a similar story. Um, if you look around the world, uh, value is held up a lot better outside of the US than in the US and we’re fully global, which amounts to being about half the US these days because the US market cap is so big, but still half is, is a big change. Um, being industry neutral, we’re not sitting there short tech, we’re long and short within industries and if there’s a tech crash, like today’s a fairly ugly day for, for tech. I’m going to guess I haven’t even looked at our versions of value will be trailing the more traditional ones today. ’cause we’re not really taking that bet. We think it’s a higher risk adjusted return not to take that bet. We think comparing tech to, I’ll make up an extreme to textiles on valuation multiples is a little crazy. There are industries that will grow for a long time different than each other and should sell at different multiples. So, uh, all considered it’s, it’s been a pretty good time for us. If you look at stuff like the Russell 1000 growth versus value, look, when you’re making an index, you got to create something simple, representative, investible, I’m not criticizing, but that takes huge industry bets and is cap weighted. So, if you get into a very concentrated market after that, the things you’re concentrated in will drive a lot. So that

Sonali Basak (00:18:47 -> 00:18:50)
And a lot of those companies aren’t, aren’t profitable either. Right?

Cliff Asness (00:18:50 -> 00:20:00)
And, and you get into a big, short, the Mag seven short AI, uh, bet in value versus growth in the traditional indices. And that is something, again, they’ll have their day and if they have their day, I hope we’ll make money. But they’ll, they’ll make more money for a while. But we think those things are really hard to bet on in a risk adjusted sense. And, um, so we think those indices are really interesting to know what’s going on in the market, but they don’t give a good indication of what systematic players are doing. Sometimes I would say at the peak of major bubbles, remember I said there were two and I only got to one. One peaked in March of 2000, the other peaked in say October of 2020. After, after COVID finally started to, to to fade. In both of those, we had a fairly concentrated market that was so crazy that it leaked over into even very diversified portfolios. There was no place, if you cared about price at all. A simple model like, you know, price to earnings. A complex model where you build a whole DCF and you’re, if you cared about price at all, you were miserable for the last six to 12-months before the peak. And then you enjoyed multiple years of happiness.

Sonali Basak (00:20:01 -> 00:20:06)
So, if we were in two bubbles, in your estimation, in 2020 and 2000, right?

Cliff Asness (00:20:06 -> 00:20:06)
Yeah, those were the peaks.

Sonali Basak (00:20:06 -> 00:20:10)
Those were the peaks. Um, why are we not in a bubble today?

Cliff Asness (00:20:10 -> 00:21:17)
Um, that’s, that’s a harder one. A I have a high standard before saying bubble. Uh, there’s enough Gene Fama rattling around in me that I, uh, I will disappoint him by saying that there are bubbles. I still have enough of him where I don’t see him everywhere. I think our industry’s a little guilty of, of saying, oh, that looks expensive. It’s a bubble. A bubble to me has to be beyond the pale prices. Um, I, I’ve often used this a subjective assumption. You might come up with a different, you might have a different verdict on a particular bubble than I do. But the framework of, can I come up with assumptions that aren’t ridiculous? With ridiculous, again, being a somewhat subjective word that could justify these prices. And at the peak of the dot com bubble, I wrote a whole piece called Bubble Logic where I tried to do it for the NASDAQ 100. And I picked on one stock; it was Cisco systems only. ’cause it looked very much like the NASDAQ 100 and it’s just more fun for people. Um, and I started, I built the DCF and I said, what kind of growth assumptions do we need?

Sonali Basak (00:21:17 -> 00:21:20)
How often is a quant doing a discounted cash flow?

Cliff Asness (00:21:20 -> 00:21:21)
Um, Well,

Sonali Basak (00:21:21 -> 00:21:21)
I feel like that’s not Like that’s not a common

Cliff Asness (00:21:21 -> 00:21:25)
I did study that and teach it when I was a TA. Um

Sonali Basak (00:21:25 -> 00:21:27)
But It’s normally not the math you’re doing.

Cliff Asness (00:21:27 -> 00:22:03)
No, we’re not on an individual company. But again, in a more holistic process where you’re looking at it co it doesn’t come out that dissimilar. Uh, the things that you would look at in A DCF, you would say, what is the price today? What’s the growth? And what do I discount? What’s a fair discount rate? And if it’s riskier, I want to a higher return and is it a good bargain now? And all those things are part of quant models. So modern quant models look more, not entirely, but look more like a very ver diversified version. It’s a grand bet that a DCF works. It’s not a bet that a DCF will get any one company. Right?

Sonali Basak (00:22:03 -> 00:22:06)
So then back to the bubble question. Yeah. I’m going to hold you to this.

Cliff Asness (00:22:06 -> 00:22:06)
Yeah, why is it not a bubble.

Sonali Basak (00:22:06 -> 00:22:08)
Why is it not bubble? Why is it not a bubble?

Cliff Asness (00:22:09 -> 00:24:32)
OK, I don’t think we’re there yet. Um, for one thing, some of it’s just mechanical. We measure, and we started doing this in 1999 during the, um, later stages, but not quite the peak of the dot com bubble. We built something, and I’m quite proud of this ’cause I think we were the first to write about it, um, that people have come to call the value spread. I don’t remember if we called it that in the first paper. Sometimes we’re always guilty of looking at history and using today’s terminology and feeling like we were calling it that then. Uh, but the value spread said this. And believe it or not, all the academic work at that time and all the published work that we had seen, again, I always feel bad somebody maybe did it somewhere and we are not giving them credit, but everything we had seen sorted stocks on these measures and then said, do the good ones beat the bad ones? And they didn’t ask, well, how big are the differences? If you sort your stocks on price to free cash flow, price to sales, pick your favorite on whatever you like. Um, you will always find some that look cheap and some that look expensive. That, that, otherwise your spreadsheet is broken, right? Uh, unless the price to sales of every stock is suddenly the same, if you sort them, you’re going to have a bottom, uh, third bottom decile, whatever you want to do and a top. But the difference between the expensive and the cheap is not going to be constant through time. So, we did really simple things. We took like the one third of stocks that were expensive on various different scales, divided that by the one third of stocks that were cheap. Um, so the price to sales of the expensive divided by the price to sales of the cheap and we showed how that varied through time. And for a few of the measures. I only remember price to book actually, which we don’t give a lot of weight to today, but it’s the only one I remember historically for about 50 years, if you did that calculation, it varied between about three and six times. So, when, when, when there was big disparities in how the market was valuing things, the expensive was about six times more expensive. When they were narrow, it was about three times. Never got to no disparity ’cause again, you’re sorting the stocks. And we found that over that prior 40, 50 years, yeah, it was better to own value, um, on, not in the next day, but called the next couple of years when the disparity was bigger. At the peak of the tech bubble, we saw that thing hit 12 or 13. So it was one of these, I hope this comes out on camera. I’m drawing a very sophisticated diagram on my hand. . It was three to six. Three to six, very well behaved.

Sonali Basak (00:24:33 -> 00:24:35)
Then it’s shot up to four times that.

Cliff Asness (00:24:35 -> 00:25:04)
And, and that’s when you sit there and go, all right, it may be a quant, but let me, let me roll out my MBA skills and say, are there any numbers that could justify 13 times the spread? And the answer we came up with was, no, that doesn’t make it easy. Betting against the bubble is famously difficult. If anyone gets it to the day, I’ll be shocked. If you’re six to 12 months early and you hold on, you’re going to make a lot of money round trip. But it’s not a pleasant six to 12,

Sonali Basak (00:25:04 -> 00:25:09)
Right, that was the story of 2025. Everyone was so worried about a bubble that too many people didn’t actually invest.

Cliff Asness (00:25:09 -> 00:25:14)
Saying there are bubbles and saying you can make money from bubbles are not the same thing.

Sonali Basak (00:25:14 -> 00:25:17)
So where are we now? You were talking about that disparity.

Cliff Asness (00:25:17 -> 00:25:57)
I’m leading up to it last I checked globally that disparity, which hit a new hundredth percentile during COVID surpassed the dot com, at least the way we measure it.

Sonali:
Mm-hmm .

Cliff:
So, and everyone can measure it slightly differently at least. But most people I think found fairly similar results, had hit a new hundredth percentile. The widest ever. So that 12,13 was eclipsed Uhhuh. Um, I would jokingly call this the hundred 25th percentile. ’cause it was about 25% more than the prior one. It’s a math geek joke ’cause that there’s no such thing. It just makes it the new hundredth percentile. But I found it funny. We’re about 75th percentile now on our measures. Um, if

Sonali Basak (00:25:57 -> 00:25:58)
On the new a hundred percent Yeah,

Cliff Asness (00:25:58 -> 00:26:45)
, uh, yeah. Uh, if, if we bet on industries it, it would look more extreme

Sonali:
mm-hmm .

Cliff:
Uh, but we don’t love that bet. We think that bet is, uh, you know, the growth differences in the risk differences can be very long tailed there. And that could justify some valuation differences. So, we don’t trust it quite as much. But the way we and a lot of other quants do it, we find things look wider than normal. I don’t think I should scream bubble at the 75th percentile. For me, uh, 75th percentile makes it may be a better time than normal tone some of this in your portfolio. Interesting. But it’s not screamingly crazy. I can’t look at this disparity the way we measure today in what’s most important to us.

Sonali:
Mm-hmm .

Cliff:
And say, this is crazy. And I have said it at least twice in the past.

Sonali Basak (00:26:45 -> 00:26:50)
I was talking to you in 2020, I remember this. You were screaming

Cliff:
Oh, repeatedly.

Sonali:
People were definitely screaming

Cliff Asness (00:26:50 -> 00:27:50)
Screaming. Um, at the near the end of 2019, I wrote a piece where we’ve said market and, and timing factors like value is an investing sin.

Sonali:
Mm-hmm .

Cliff:
And we’ve always, we said it in a joking manner, but we’re serious. We say we recommend you sin a little, um, at true never before extremes.

Sonali:
Mm-hmm .

Cliff:
You can have a little bit more, uh, enough that you think you can hold it if it gets worse. ’cause it’s always no matter what you Yeah. Your pretensions, it’s probably always going to get a little worse before it gets better. So, we wrote a piece saying it’s time for a venial value sin. A we got a little Catholic there, you know, the more minor sin, not a cardinal sin. Little sin into value. And then I wrote a piece in early COVID saying, no sin has ever been punished this violently and this quickly. You know, things got a little pretty crazy before COVID that spread was about its tech bubble peak before COVID. So, quants like me who want to blame COVID, that’s not fully fair. It was, it was there. COVID sent it to the stratosphere. That’s where we got the 125th.

Sonali Basak (00:27:50 -> 00:27:51)
After the Fed stepped in and all that jazz

Cliff Asness (00:27:51 -> 00:27:59)
All of it. And you remember, there’s, you were only supposed to own Peloton. You everything else was a, that’s a bit of an exaggeration, but, you know,

Sonali Basak (00:27:59 -> 00:28:01)
No but every, a lot of people were owning Peloton then. I, I remember this,

Cliff Asness (00:28:02 -> 00:28:33)
We will shout bubble a lot of things make me nervous. I’m enough of a grumpy old man to look at, uh, single day options to look at meme stocks. Um, uh, it’s no longer there. But to look at things like, uh, MicroStrategy selling for three times it’s worth in Bitcoin and, and, and looking at my nephews and hopefully they’re not listening. I’m just using them as an example. Who I’m fully convinced. I won’t say my nephews, I’m sorry guys. Sonali Basak (00:28:33 -> 00:28:33)
What are they doing?

Cliff Asness (00:28:33 -> 00:30:17)
But, you know, 23-year-old males seem to have both Robinhood and FanDuel’s on their phone. Yeah. And I’m entirely unsure they can tell the difference, um, that, you know, stocks on average go up, FanDuel’s on average you lose. Yeah. If you trade too much in either one, you tend to lose. Um, but I think they’re just, uh, but all that, that’s me being a curmudgeon. If it’s not showing up in the numbers and we’re not seeing numbers where I can go, well, this is attractive and I like having this in my portfolio, but I wouldn’t call it a bubble. There are other softer signs, um, Owen Lamont of, uh, Acadian, um, they, I hope it, I will actually name competitors I like, um, at times. And I, I think some people at my firm are like, why are you giving them free press? ’cause he’s great. Um, he’s a former academic, he’s a strategist there. Um, he and, uh, has a whole list of kind of criteria as to what makes a bubble. And one of the major ones on the, on the list is issuance in, in the other major bubbles we’ve seen, we’ve seen pretty massive issuance just starting to ramp that up. Obviously, the SpaceX IPO was a pretty big fricking deal. Yeah. Prior to that, we had seen a dearth of IPOs for a number of years. So, we didn’t see, uh, another sign of a bubble we did not see is all these companies saying, yeah, we got to sell at these prices. Companies will never say we’re selling because our stock is too expensive. You will hear that zero times. Um, I won’t say, is there one, have you heard that? Is there one honest man treasurer who says that you’ll, it’s

Sonali Basak (00:30:17 -> 00:30:29)
A hard thing to say. And you know, it’s been a long time. You know, the, the, the true and tried like Golden day investor used to look at the Wall Street Journal and look at the insiders drop sales too. Uh, I think that a practice has kind of gone away. Yeah. Yeah.

Cliff Asness (00:30:30 -> 00:30:39)
So, they come up with some other reason. Um, but you do see near the peaks of bubbles, firms selling a lot of their own shares. And we may, I look sometimes

Sonali Basak (00:30:39 -> 00:30:40)
Something to keep an eye on,

Cliff Asness (00:30:40 -> 00:31:14)
Keep an eye on, um, could happen, doesn’t mean it will happen. Things that have happened in the past don’t necessarily just repeat. Um, but for me to start shouting bubble, at least in what I do.

Sonali:
Mm-hmm.

Cliff:
Uh, again, if there’s something else, someone else, uh, does, the overall price of the stock market is very high versus history. It’s not.com levels Shiller CAPE, but it’s, it’s getting there. I’m not really talking about that. I’m talking about what quants do.

Sonali:
Yeah.

Cliff:
Um, we see an attractive market for what we do, but we do not, I’m not screaming bubble, so I like screaming bubble too. I want to, I just

Sonali Basak (00:31:14 -> 00:31:19)
Once, once in a while, twice now. I hope I’ll be there for the third time, and we’ll figure out that how to navigate it together

Cliff Asness (00:31:19 -> 00:31:30)
I actually hope it never happens again. Um, I, I’m joking about liking it because I, I have said who the gods would drive crazy, they make first to identify evaluation bubble, but don’t, because you lose a bunch of money on the way healthy.

Sonali Basak (00:31:31 -> 00:31:33)
But don’t…Isn’t there something healthy about a retrenchment? Yeah.

Cliff Asness (00:31:33 -> 00:31:34)
Uh, for one thing,

Sonali Basak (00:31:34 -> 00:31:36)
Not a crash, but a retrenchment.

Cliff Asness (00:31:36 -> 00:31:46)
Yeah. For one thing, a lot of strategies that have worked for many years that I think will still work, would not work if they didn’t have these periodic tough times.

Sonali:
It doesn’t create entry points, right

Cliff:
It kind of flushes out the market

Sonali Basak (00:31:46 -> 00:31:49)
You need a new entry point for investors to get in.

Cliff Asness (00:31:49 -> 00:32:50)
And, and it can’t be so easy that everybody does it. And it, if, if something is, you know, a two sharp ratio, never lo, that’s geek for you know, super good almost never loses money. Um, if it’s at all publicly known, if it’s secret, that’s one thing. But if it’s at all publicly known, it doesn’t last very long at all. If something is a 0.3 sharp ratio, that’s a little less than the stock market’s been long-term. So, it’s still actually pretty good. It has some horrible short-term periods and just when everyone thinks, I want this in my portfolio forever, and they’re right actually to think that. It’ll go through one of those periods and flush it out. Um, so there is, uh, for people in my business, be careful what you wish for. If you didn’t have these tough periods, what you do might get arbitraged down to, to nothing. Doesn’t mean that I’m not throwing things around my office during the tough periods. I don’t enjoy them whatsoever. But I do have these moments of clarity where I go, I wouldn’t get to do what I did for a living if it were easy.

Sonali Basak (00:32:50 -> 00:32:50)
Keeps you humble.

Cliff Asness (00:32:51 -> 00:33:54)
Oh, and, and I need that desperately. Um, markets are a very humbling. Warren Buffett, when we studied him, he had horrific three-year periods, both relative and absolute tend to be different periods. Um, the market, uh, tells you it’s hard, tt’s, it’s not impossible to beat. I’m not a perfect efficient marketer. It is hard to beat. And even if you have a way to win over the long term. Um, the only people, uh, again, I’ll use Jim Simon’s, unfortunately he’s passed too, but Jim Simon’s in the Renaissance medallion fund, investors should be careful what they wish for because you find one of those, they tend to kick you out and only run their own money. Um, my favorite question, I don’t get this too much anymore, but every once in a while, over the last 20 years, I, I get a question, is the Medallion Fund better than you guys? And I’d go, oh, hell yes, . But I think we’re very good and make your portfolio better and we will actually take your money and they won’t. So how is that relevant and….

Sonali Basak (00:33:55 -> 00:34:33)
Well, it’s incredible, right? Because I, I, the next thing I really wanted to ask you about is, you know, back in the day to get into a hedge fund, you really couldn’t, right. It was just the way hedge funds work. There were these super-secret elite pools of capital. And AQR really has been on the cutting edge of making strategies more available. Um, now you see an entire investment industry trying to make different types of investment strategies, whether it’s private markets or hedge funds, more available to more investors. What happens to, um, the investments themselves as they become more available? Do you, do you think about that? Do you worry about that at all?

Cliff Asness (00:34:33 -> 00:36:46)
Oh, yeah. Um, enough capital can ruin any investment strategy. Um, there’s no strategy that can just add capital infinitum without affecting what it, it does. On net, the value spread is only one factor. Um, but we look at a lot of things like that to go, has the net compressed, the expected returns going forward ’cause that’s what you see. If everyone tries to buy the market, the market goes up, the price of the market’s more expensive against fundamentals. And we would forecast lower kind of five-to-10-year returns going forward. The same could happen to us, I would say on net. We haven’t seen it yet. Um, again, it’s not a perfect measure, but 75th percentile is not a market, um, where this is all gone. Um, there are a fair amount of people in the liquid alt space who do some similar things to us, but there are a fair amount of people investing in the stuff we don’t like. I I, this sounds really obnoxious, but the retail world, I think we’re usually on the other side of them. I think they’re usually on the faddish stuff and that’s become very popular. So yes, some parts of the market can grow, but it’s the net that matters. And when you look at the actual numbers on the stocks we trade, we’re not seeing diminished opportunities. That could happen one day if capital just kept flowing, uh, and the, and every strategy has the seeds of its own, if not destruction, very painful periods, um, when, when that unravels, I have a running joke at AQR where I pick the youngest person in the room and I go, that hasn’t happened yet, but if it does happen, we’re probably going to have a great 10, 15 years. Because how does something get arbitraged away? It’s people adding capital to what you do, buying what you long, selling what you’re short and driving up the longs and down the shorts and, and you can benefit from those flows for many years. And then I get to retire where people think I’m way better than I am ’cause I was the lucky beneficiary of these flows. And I look at the now young person then 15 years older who’s now taken over AQR and I go, you know, the good news is you’ve won the AQR Game of Thrones. You’re, you’re in charge. The bad news is I used up all the return forever could happen. Um,

Sonali Basak (00:36:46 -> 00:36:57)
One day I want to show this to your successor. Uh, you know, the reason I asked this also is because I thought you made a really interesting comment that I’ve never heard you talk about before in a recent podcast about how

Cliff Asness (00:36:57 -> 00:37:00)
Well, that’s rare because I usually say the same things over and over again. So that’s nice of you.

Sonali Basak (00:37:00 -> 00:37:15)
This was different. This, you talked about the other hedge funds in the industry and the, the risk of crowding. Um, I think this has become a big thing in the minds of a lot of hedge fund managers these days in the current market we live in. Can you explain where your fear comes from?

Cliff Asness (00:37:16 -> 00:41:41)
First, I separate crowding, uh, and this is too binary and there’s all points in between. But to just think about it, it makes it simpler. Two kinds of crowding. One we’ve been explicitly discussing a strategy with too much capital in it is crunching down the differences between what it’s long and what it’s short, um, and can lower the expected return. That’s a more long-term thing. That’s like saying the market is, the stock market’s very expensive. We don’t know what that means for the next year. We do think the next 10 years you should use at least a more modest forecast. So, it’s more than that, crowding can lead to that. Um, crowding is more typically talked about in the short term. The risk of a very bad short-term event. And that certainly exists. Uh, I know you know it well. The mother of all of them was the August of ’07., it was called the quant crisis for about 10 years. And then somehow people switched to
Sonali:
quake.

Cliff:
Yeah, I said, I I remember it as the quant crisis. Um, and at some point, people said, you know, we, we need some alliteration here. Quant quake is better. Um, this was all of about six days. Um, for us it was very survivable. It was painful. Um, it was, uh, just like a stock market crash is painful for a long only investor. Um, but it was no risk of ruin going on with us, it was just, and we looked and we said, well, uh, my friend, the value spread again, went from the 50th percentile at the beginning to the 95th over six days. That is a gigantic move in six days, but it also makes you think we’re going to be fine. Um, we did, uh, in those kind of things we do a little bit of non-quant stuff like talking to every dealer trying to figure out who has taken their positions off yet. Um, in some of these,

Sonali:
the sleeves roll up it we

Cliff:
just ’cause for quants doesn’t mean we, we think all human intelligence is not useful, particularly for something like this. Um, we actually added a little bit to what we did at almost the exact bottom of that. Um, because we heard Goldman Sachs had had injected cash into their very, very suffering long/short equity fund. And we thought they were the last kind of, uh, weak hand at the time to, to de-lever. You can de-lever by literally trading your positions down or by adding more cash. Um, and Goldman partners made out like bandits of course everything…

Sonali:
you’re alma mater.

Cliff:
Everything came back. I think they did that for the clients. Uh, not, not for, because they would make up, but they did do very well on it. Um, those kind of things. There’s not a strategy in the world, particularly one that more than two people know about that doesn’t have a bad left tail. And the bigger and the more popular, the worse the left tail because you don’t control your own act. Uh, the, you control your own actions, but you don’t control the actions of people very similar to you. So, if someone very similar to you and we’ll all argue we’re unique and ours are so much better. Uh, so this is a correlation, not the same model. But if someone’s similar to you and overlapping with you, panics and sell, you’re going to suffer.
Sonali:
Do you worry about another quant quake?

Cliff:
Yeah, I do. Um, my plan in that is to do what I lasted in the other quant quake along with the other AQR partners and team, keep monitoring to make sure that we’re nowhere near, uh, a true, you know, uh, event, uh, you know, a margin call. We’ve never, never had a margin call. And if so, and we think the strategy still sound weather it. We’ve tried a million ways and again, you can, I’m sure you can get someone to come on here and say, they’ve built the system to forecast this. We’ve never come up with a way that, that we think can really move the dial of saying it’s so crowded, you want to get out now. It has the same timing problem of, of saying it’s a bubble you want to get out now. It can go on for quite a long time just because the conditions are there. Doesn’t mean it happens to get out now is to get out of a good strategy long term because you’re scared. I wish I could tell people, yeah, not only do we make money on average over the long term, but we get you out for the, for the crowding months. Doesn’t work that way. Um, but we’ve gotten really good at analytics and knowing where we are. Um, and the worst one ever was extremely survivable event. Very unpleasant while it’s occurring. Um, we’ve seen many ones since then. Last year we saw a few. I think one was earlier in the year, maybe one was October. I could, I could be off on the marks.

Sonali Basak (00:41:41 -> 00:41:42)
Liberation Day you thinking,

Cliff Asness (00:41:42 -> 00:43:31)
um, we saw some craziness?

Sonali:
Yeah.

Cliff:
Um, sometimes it is hard to tell whether it’s a de-leveraging among similar people, which I think was what the quant quake was,

Sonali:
right

Cliff:
Or if it’s just retail buying very expensive low-quality stocks, which quants tend to like and sometimes it’s a mix of both. Um, but by and large good risk control where you’ve looked at the worst cases in the past, you’ve said, what if it’s substantially worse than that? That should hurt, right? If you say it’s the worst case for my strategy and it’s worse than that- should hurt. But if it’s survivable, you know, if you own equities, you should know that they can go down 20% in a day. And in the GFC they were down about 60% peak to trough. A lot of good strategies that make money long term have uh, uh, particularly a short-term, uh, left tail that can make them still good strategies. One thing particularly important is when that left tail tends to hit. The two bubbles I’ve talked about in particular, quant as we define, and a lot of other quants too, made money round trip. But when it lost money, everything else was going straight up in a bubble. And when it made more than all of it back was when people needed it the most. And that’s not our goal. And those funds are goal is to be market neutral. We were accidental heroes on that. We didn’t want to be so counter cyclical, but both times it’s happened it has looked that way. So, it’s not just, do you have a left tail? It’s both short and long-term left tails, it’s when they occur. One thing I’m fond of pointing out, and this is held up for the other mini ones in the quant quake, peak to trough the S&P was pretty much unchanged. The world did not notice the travail, the six-day travails of a bunch of quant geeks.

Sonali Basak (00:43:31 -> 00:43:33)
And a lot of your peers were out of business by the end of that.

Cliff Asness (00:43:34 -> 00:44:07)
Um, some of the more levered ones, I think most common was, uh, multi-strat funds that really didn’t understand quant. That added quant in a six, seven-year bull market for quant. Um, if it’s not really what you do, and this is a self-serving commercial for, for dedicated quants, I admit if it’s not what you really do, you’ll be fairly quick to abandon it for bad reasons. Uh, and I’d be the same if I did a strategy that was unfamiliar with to me just cause it had worked for a while. I don’t deeply understand it. And when it doesn’t work,

Sonali:
yeah.

Cliff:
I’m like, uh, this is…

Sonali Basak (00:44:07 -> 00:44:15)
So, so I understand you perfectly here. What is the AQR the Cliff Asness strategy for the next quant quake? What’s the playbook?

Cliff Asness (00:44:15 -> 00:45:56)
Um, the, the playbook is not to take too much risk going into it. And we take decently less leverage than we used to. We used to run, um, some things we, we labeled high vol, literally in the name of the fund was high vol. So, we were pretty happy with that disclosure. Um, and, and it was meant to act this way. Uh, but we did say – alright, we survived this one, we’re not going to run those again. We are a little more circumspect. So, if you saw the full quant quake again, I think it would be less painful to our firm. Still not be fun. If you saw one and a half times it, which would be really extreme, I still think it’s, it’s, it would be as bad as the last one. I’m making these up. These are guesses. Quants don’t like single events. What we want to see is a thousand quant quakes. So, we can tell you this is the histogram of how we do in quant quakes. Um, the humans don’t want to see a thousand disasters. The statisticians want to see a thousand disasters.

Sonali:
Yeah.

Cliff:
Because then we can, you know, maybe figure out what’s actually going on. Uh, but I, I think the quant quake, mostly what we do is prepare the portfolios, try to create things. We do more things that are more proprietary now. Uh, both, uh, alternative data and some of the, the ML stuff we think looks more different than our peers. They might be very good too. I’m, I’m not saying we’re the only good ones, but I think the degrees of freedom in those things have led to, in the short term, maybe we may end up looking all alike in 10 years, but in the short term, it’s led to some more diversity. Uh, we rely a, a decent amount less on the original factors. The ones I’m allowed to talk about- are less of our portfolio. The ones I’m not allowed to talk about are somewhat more.

Sonali Basak (00:45:56 -> 00:46:06)
I was going to say the last question I have for you,

Cliff:
Please

Sonali:
is on that alternative data, what can you tell us? Gimme some secret sauce. I was a former journalist. I need to scoop .

Cliff Asness (00:46:06 -> 00:46:34)
Well, first of all, no . Um, I, I can describe what we do in alternative data. Um, alternative data is exactly what it sounds like. It’s, it’s databases that didn’t exist before that somebody often proprietary new firms sometimes designed just to create this database and to sell to people like AQR. Sweat equity. Uh, if, if if sweating behind a computer, screen counts as if you’re sweating that’s

Sonali Basak (00:46:34 -> 00:46:35)
Depends on the day in the market…

Cliff Asness (00:46:35 -> 00:47:16)
Yeah. You, you probably have other issues. Um, not that I haven’t sweated while looking it a while looking at a, a computer screen. So, it, the classic one about the only one I am willing to talk about, um, is credit card receivables. It was one of the venerable first ones. And I, if, if people are willing to talk about this, it means it’s probably largely played out at this point and probably they’re not betting a lot on it.

Sonali:
Sure.

Cliff:
But people-built databases that were essentially, this is legal public data. You do have to make sure of that. This can’t be an excuse to have data. You’re not supposed to, to have. And we do. You actually have to spend some time checking on that to make sure there’s not some sneaky data in there that’s making it look good.

Sonali Basak (00:47:16 -> 00:47:17)
The collection of the data is the hardest part.

Cliff Asness (00:47:17 -> 00:51:04)
Well, you have things, um, and like if you’re polling experts, if those experts are using data, they’re not supposed to use, you shouldn’t be using it. Um, so you do have to go through some steps. But stuff like the credit card receivable data was just on the web. Um, and somebody put it together, collated it, and it was a pretty darn good strategy. The lifecycle of alt data tends to be fairly high risk adjusted return when you’re early to it. Um, again, this example, um, pretty much only applies to retail stocks, right? Uh, credit cards are not that helpful for Nvidia. You know, nobody walks in and says, gimme a million of your best chips Jensen, and here’s my Visa card. Um, but for retailers, knowing who in the last three months, I’m making that up, this is not precise. Who in the last three months has seen an increase in credit card usage and who has seen a decrease? You can imagine if you have that data a little faster than other people, that’s an advantage. But information edges that are about speed, I think are the most arbitrageable out there. Things about valuation and risk where you’re up against huge behavioral biases that people want to believe in a thousand stocks that are expensive will keep going forever. Enough capital can arbitrage anyone anything away. But that’s harder. I’m a little faster than everyone else is not going to last that long. So, when alt data often not always, uh, you tend to see very high risk adjust to return when you start doing it. And then a steady decrease. It may not go to zero. You may still want it in your model at a smaller weight, or it may really go to, to zero. The world has figured this out.
I had one funny interview where, um, somebody was, uh, the, the interviewer was asking me about all data and were…was not as nice as you, they were pushing for a real example,

Sonali:
Right, give me more .

Cliff:
And, and I said, I’d love to give you an example, but our, our heads of, of stock selection have asked me not to because it is proprietary. And unlike some of the other things we do that we think are fairly hard to arb away, here we’d be helping our own destruction by by doing that. When he wrote the piece, very nice guy, a smart guy. But when he wrote the piece up, he didn’t say that I was asked not to speak about it, he said, my team won’t tell me what we’re doing. Which, um, has a whole different tone to it -that has a don’t worry old man, we got alt data. You don’t need to worry about it. That is not the case. They in fact will tell me. Uh, but um, that world continues to evolve. We’ve probably thrown out more from alt data than we have from the core factor stuff over our whole history. Um, any measuring it when you alter a core factor and measure it a little better, have you thrown it out, I don’t know but there’s certainly a lot more dynamic change a few years later alt data has to be refreshed and replaced with new sources. I can’t tell you how many years or decades that will go on for. I will tell you, so far so good. We’ve been able to find new sources. Um, alt data, by the way, is ultimately a form of fundamental momentum. Forever we’ve looked at both price and fundamental mo. We love valuation, we love profitability, but we also like good things to be happening. Simple measures of fundamental momentum that quants have looked at since maybe the eighties, at least the nineties, are earnings revisions and surprises where they’ve tended to continue. If you have good earnings, there’s often more to come. Alt data is just a way to try to be as quick as you can on, on those things. So, in some ways it’s brand new, and it is in a lot of ways these didn’t exist before, but in some ways, it’s just additional ways to measure a theme and something we’ve believed in pretty much forever.

Sonali Basak (00:51:05 -> 00:51:08)
Okay. Truly last question, um, on AI.

Cliff Asness (00:51:08 -> 00:51:09)
I don’t believe you anymore, but that’s the ok

Sonali Basak (00:51:09 -> 00:51:26)
I know ’cause we went into the data thing and usually people want to go into the AI thing.

Cliff:
My time is your time,

Sonali:
But the data, the data part is, is really interesting as it pertains to quant finance and particular. But on AI, you once said that you had AI taking over parts of your job. How much can it take over really?

Cliff Asness (00:51:27 -> 00:58:00)
Well, nobody knows at this point. I can tell you kind of where we are today, we’ve always prided ourselves. Uh, and you can’t really do this mathematically, but can conceptually being 50/50, does it work in the past, including add a sample, live trading being better than a back test, but both counting and does it make sense to us? Do we understand every drop of what it’s doing and do we get the intuition? Um, and the reason you do this part is if you only do what’s worked in the past, you’re just mining the data, and you tend to grossly overfit. And you know the classic, uh, examples, if you study a thousand independent things, one will look one out of a thousand wonderful. Is it really one out of a thousand wonderful? If you find the person in one Powerball, are they good at Powerball or do they just get lucky? You know, barring cheating, I think they probably just got lucky. Um, so you worry about overfitting. One of the lessons of ML and my colleague Brian Kelly, um, and, and other people like Laura Serbin at AQR works a lot on this. Uh, but I’m quoting a lot of Brian Kelly on this. He, he is Yale professor and an AQR partner. Great deal you got Brian; you got two full-time jobs that pay you for doing 90% overlapping work. I just want you to know I’m onto you. Um, but he’s brilliant on this and most of the things I say about AI, I’m paraphrasing what he says so I want to give him credit. One of the things he says, and, and this is a math concept, it’s not an easy one to convey here so, I’ll just tell you the fact under fitting is a problem on a par with overfitting. We’ve always been worried about overfitting that we’re going to use these statistical techniques to find random patterns that aren’t real, that won’t repeat and bet on them to our detriment. Under fitting is there’s actual true complication out there and you’re not picking it up because you are using overly simple models. For instance, we historically have used, when I talk about valuation and most of the bets are within an industry,

Sonali:
Mm-hmm

Cliff:
aside from a few industries, we recognize things like finance and REITs need some special stuff. But for most industries we’ve used essentially the same composite valuation indicators, because yeah, you can say which ones work best in all of 50 industries or 10 sectors, but are they really better? They’re highly correlated measures, price divided by some fundamental, as that fundamental differs, they’re still correlated. So, you’re really going to tell me this one’s better, this one’s worse. We’ve also used very simple functions where sort the stocks in an industry, by again, I’ll use price to sales, even though we use many measures and average them. Sort them on price to sales. You go long, the cheapest short, the most expensive, and it’s linear in between the weights. You add all these up; you have many other factors to come up with a total. Why is it linear? It’s not, it’s linear to avoid overfitting that if we let models just go crazy and do whatever they want, it’s going to catch every little wiggle and it’s going to overdo it. But there’s no conceptual reason why it’s linear. Some factors might work in the tails, call it the top decile and the bottom decile, but be kind of flat, not give you a lot of information. Some might work on the short side but not provide a lot of information once you get into the reasonable, uh, area. One thing that modern ML is good, is better, at than traditional statistics. And it’s still statistics by the way. They’re just new, wonderful tools we didn’t have before. One thing, it is better, we believe, at balancing overfitting and under fitting. Penalizing, um, endogenously to creating your model, penalizing overfitting while allowing it to do some and some reasonable amount. Um, and so things like that, we’re still looking at valuation, but ML is building us a subtler valuation, um, uh, model that we think adds to our process. We haven’t completely turned ourselves over to it. Um, and again, my my phrase earlier about turning ourselves over, I want to defend myself. So, this is half a joke, but half true. I often say I slowed us down a little on, on moving to ML.

Sonali:
Really?

Cliff:
Because you got to give up some intuition. When you start using ML, there is usually a stage where you go, I’m not quite sure what it’s doing, but it works. And that’s something we kind of prided ourselves on not doing forever. That you understood each step

Sonali:
Because you want to know.

Cliff:
So, I, I think it is the old man at the firm’s job when you had something that’s worked for 20, 25 years and you want to have a philosophical change, even if it’s a modest one. Again, this is not turning ourselves over, it’s going from 50/50 to two thirds, one third, something like, like that. But any change to a philosophy you’ve stuck with and talked about should be done slowly and carefully so, I don’t apologize for it, though I do think I probably cost us a little money by going slow. And one, there was a, we were in a meeting and I think it was a fairly junior person, I was whining basically about this loss of intuition and the, the junior person said something like, if it was all super intuitive and obvious, what do you think the machine learning is doing that’s special? Um, if there’s not a part that’s at least hard. And uh, one of the big uses we use for ML is, uh, called natural language processing. It’s an ML function where you take textural data

Sonali:
mm-hmm .

Cliff:
and you say, is this good or bad news? Fundamental momentum to us. Um, the way quants did this for a trillion years was, get the text to the call, we had these digitized 30 years ago and add up good words and phrases and bad words and phrases with numerical values and come up with a total. So, at its absolute simplest, and it’s not this simple, increasing is a plus one. Obviously, the problem is if the sentence was -embezzlement is increasing- plus one was not your best call. Quant can survive that. If it’s looks stupid 47% of the time but looks a genius 53 and put a little bit of it in the model, we can live with that. NLP- natural language processing- is better than we are at processing that call, representing that call. This gets very geeky, has a very long set of numbers.

Sonali:
Well now you’re telling me where you trust the technology and the data better than yourself and you’re telling me where you trust yourself more than the data. That’s what you’re saying.

Cliff Asness (00:58:06 -> 00:59:56)
Well, the example here, just to finish it out, is there are steps I can’t tell you like what NLP actually does is represent each earnings call as, uh what’s called a vector of a whole bunch of numbers. And then do empirics on those numbers to go which ones tend to be good returns, which ones tend to be bad returns. The intuition at the end, you create a strategy that is actually fairly decently correlated with other ways to measure fundamental momentum. We just think it’s better, or at least very diversifying, adding a very new aspect to it. The correlation gives us great comfort because it means we are measuring what we’re trying to measure. But if you ask me this, that vector of numbers that represents each earnings call, what does number, what does the seventh number mean? I’m going to be hard pressed to figure that out for you. And if I turn to a far younger, far smarter AQR employee who’s fully up on these things, they’re going to be hard pressed. Maybe they can torture it and figure out one for me. It, it, you are giving up a little bit. That’s kind of the stage for NLP where I think of that, uh, if it wasn’t doing something you didn’t fully understand what’s it doing, but we, I’m almost sure we wouldn’t do it if it didn’t come out to a strategy that the end active acted very intuitively like we think it’s supposed to. When fundamental momentum suffers, measured simpler, natural language processing doesn’t have its best time. We think it does better in both environments, good and bad fundamental, we think it’s a higher Sharpe, but it looks like the original. That gives us comfort. But you do have to again, do that uncomfortable thing and say, all right, I’m going to give it, I’m, I’m, I’m going to avert my eyes and skip this step. Um, which I, I think is part and parcel of using ML just took me a bit to get used to it.

Sonali Basak (00:59:57 -> 01:00:04)
Cliff, thank you for joining us. That is Cliff Asness. He is the co-founder of AQR Capital Management. And you have been watching The Bridge by iCapital.

END