Enrich Your Future 24: Why Smart People Do Dumb Things

Andrew, fellow risk takers, this is your worst podcast host Andrew Stotz from a Stotz Academy, continuing my discussion with Larry swedroe, who for three decades was a head of Research at Buckingham wealth partners. You can learn more about his story in episode 645, Larry stands out because he bridges both the academic research world and practical investing. Today we're diving into a chapter from his recent book, enrich your future, the keys to successful investing. And that chapter is chapter 24 why? And Larry, come on, why do smart people do dumb things? Take it away? Larry,

Larry Swedroe 00:36
yeah, it's a really challenging question, and we all know people, and probably including ourselves, who have done some dumb things. I recently had, you know my best friend, he had a fix something on the ceiling that was right out the top of the staircase, and the dummy put gets on a step ladder, climbs up and slips, goes down the stairs, crashing, miracle. He only broke his shoulder, didn't snap his neck and die. And you want, you know the guys was a multi millionaire. He could have hired somebody for 100 bucks to fix the thing, but no, he has to get on a step ladder at the stop, top of a lot, you know, the staircase. You know, we do dumb things, right, especially probably for us men, when we were teenagers, and, you know, probably took risks we shouldn't take. So we all know that this to be the case, but when it comes to more important issues, maybe, and when there's science and knowledge, hopefully we can avoid them. For example, I think certainly in the US, especially, the amount of people who are smoking today has gone way down because of the warnings that say from the Surgeon General, smoking is hazardous to your death. You're likely to get cancer. Not everybody stops smoking, but most of the people, clearly, the vast majority, do that. What's really interesting about this analogy is, when it comes to investing, the SEC does provide some guidance and on all advertisements for mutual funds, if you're an actively managed fund, of course, it has to carry the warning that past performance is not a predictor of future performance. Now you would think that investors would heed that warning, but all of the evidence shows that investors money flows into the funds that have had the most recent best returns and flows out of the funds with the recent poor returns. One psychologist said the problem is it's not descriptive enough, and what the SEC warning should say more explicitly is specifically pass out performance has no predictive value as to future outperformance that might make a difference. Maybe, I'm not sure

Andrew Stotz 03:12
that, yeah, and I didn't when I read what you wrote about it, and then I looked at the research and their recommendation, I was like, I don't know if that's tough enough, you know, and you can't, but you can't say it. Past performance does will the past performance will not repeat, you know, because it

Larry Swedroe 03:29
good, right? Absolutely, you know. But here's the thing about we talk about smart people doing dumb things. Here you have the average retail investor. You could excuse maybe their behavior because they're ignorant. I don't mean it in a pejorative sense, like being dumb. I mean, I'm intelligent, at least, I think so. I graduated from at the top of my class in one of the better MBA programs in the United States, but I'm totally ignore about nuclear physics, and my wife and three daughters tell me women is another subject, right? So unless you get an MBA in finance today, you probably haven't taken a single course in capital markets theory. So where do you get your advice from Barron's and CNN and they're going to count, you know, active managers, because that's, you know, they need your attention to sell their ads and all that stuff. Hm, okay, but you would think the big pension plans in the United States will hire world class consultants who certainly have the knowledge about the academic research, you would think that the people are on the board and charged with Mount should be at least aware of the academic research, just doing their due diligence to do their jobs. And yet, here's what the evidence shows on every study done on the. Performance of pension plans. They hire consultants. They measure performance based on, like, typically, three year period, some cases, maybe five, when all the evidence says that's way too short a period. It's noise irrelevant. But they do it every three years or so. And the managers they hire to replace the under performers, they go on, on average, to underperform, and the managers they fired go on to outperform, which means they would have been far better off doing nothing, let alone incurring all the trading costs that are implied when manager A comes in to replace B and they don't like their stocks they're holding. So you got a lot of trading going on, and yet they keep repeating this. Now I've asked, I've had to present at pension plans trying to get their business by using more systematic, transparent and replicable funds like or similar to index funds, okay? And I asked them. I said, How has that worked out? And why do you keep hiring new managers and telling you that something must have gone wrong in your process because it didn't work? So I ask you, tell me what you're doing differently this time in choosing the managers that will prevent you from making the same mistake. And you know what the answer I've gotten every single case is, nothing. Never once got an answer. Why they think they're going to get some different outcome repeating the same dumb behavior. It's amazing. So there's really no good explanation except human stupidity. And there, there's only two things that are infinite, the universe and man's capacity for stupidity.

Andrew Stotz 07:03
Now you know in this, in this chapter, you highlight some great research that talks about what's happening with returns. And I mean, it's so perfectly clear when analyzed. You know about the top performer, the prior top performers, versus the prior worst performers. And if now one of the questions I had, but I thought we should go through that just a little bit so someone understands that gap between it like you talk about the CAPM alpha on page 145

Larry Swedroe 07:38
and go ahead. Go right ahead. And so

Andrew Stotz 07:40
let's, let's just look at this what, what you've what this talked about is. So let's say the average benchmark adjusted return for the median strategy beat that of the winner strategy by 1.32 percentage points, and the loser strategy, meaning buying the losing funds, outperform the median strategy, buying the average fund, let's say, by about one percentage point. And so the loser strategy, buying the loser or the underperformers, prior losers, the prior losers, outperform the winner strategy by 2.28 percentage points. And what you talk about is, okay, well, maybe that's just has to do with the volatility that they're exposed to. But no, when you do a sharp ratio and try to bring in the volatility, you find that the ratio of the median strategy was 0.42 versus 0.25 for the winner strategy, while the loser, prior loser strategy produced a sharp ratio of point four eight higher, meaning better than the other two. So the investors

Larry Swedroe 08:49
a simple explanation that people say, How could that be? Winners strategies tend to be ones that have performed obviously well in the recent past. So valuations have gone up, meaning future expected returns are now lower. And loser strategies, you know, the reverse is true. That's why value stocks have outperformed over the long term. That's really a prior loser strategy. And the same thing is true, by the way, in commodities, you know, Cliff, sorry, AQR runs a strategy based on long term returns to commodities, and they look at five year cycles. They buy the five year loses and short the five year winners, you know, because what happens is, right, your returns are poor. That means commodity prices were down. So what happens? The mines get shut down, capacity shrinks, and then you have a problem. So prices go up, and the reverse is true. And when prices go up, new mines open up, capacity increases, and you get a reversal. So you know the same logic applies when you get out. Performance, whether it's individual stocks or an asset class or any particular strategy that's chasing recent performance. Yeah, and

Andrew Stotz 10:10
I think that when you're talking about like a capital intensive industry, like mining as an example, or oil production or something like that, you just have this natural long term cycle before they can deploy those assets that they decide, okay, we're going to expand. It could be five years before there's any revenue coming out of that expansion. And so it's just a natural, like cycle to get but what I wanted to ask when I was

Larry Swedroe 10:34
agricultural prices, right? Yep,

Andrew Stotz 10:37
yeah. So like, coffee prices went through the roof, and that hurt our coffee business, as we were buying coffee, raw, raw coffee, green coffee, that was, you know, just going up in price constantly. But in theory, that should drive the farmers to be much more aggressive at making sure that they're getting the most out of their current plants and planning more. But it doesn't take three it takes three to five years before a sampling becomes a bean producing and so we end up three to five years from now, we could have a glut

Larry Swedroe 11:07
of coffee. It's possible. Now, of course, climate and other things can impact that.

Andrew Stotz 11:13
Yeah. Now let me ask you a question, because carhartt's model brought in a fourth factor to the three factor model, originally of fama French, and that model was that that that factor was momentum, and he showed that there was value, there was there was a performatory, yeah, yeah. Explanatory power. So how does more and momentum is ultimately buying current winners. So how does that jive here? With the opposite of saying, you know, buying current under performers? Well, that well, you

Larry Swedroe 11:55
have to remember very importantly, momentum number one is a very short term tool, okay? It is short term positive momentum. And long term version to me, you get a reversal, okay, because things can't grow to the sky. So that's why you get some persistence in, like one year, right? Because things are going up. So maybe another on average, momentum works. Call it four to five, six months, right? Sometimes it works a year or longer, sometimes only a few months. But on average, it's somewhere around that. So often a fund that won one year might win one more year, but generally, then over the longer term, you get that reversion because prices get too high. So the problem with that is it looks great on paper, but when you have very short term factors like momentum, what happens to your turnover? Because to capture it, you have to trade frequently, so you easily can have over 100% turnover momentum strategies, which especially and momentum is most powerful in the illiquid stocks like smaller caps, and there your trading costs are much higher. So momentum can be very tricky to implement, and it may not survive transaction. You have to know how to manage your trading course well, or I would not try to run a momentum strategy. And I'm

Andrew Stotz 13:30
assuming that the funds or the ETFs that are done by, I don't know dimension or AQR, that are trying to take advantage to some extent of momentum, they just have such a competitive advantage in the trading costs. Well,

Larry Swedroe 13:43
let me give you an example of how dimensional uses it that actually reduces turnover. All right, which sounds contrary to our prior discussion. So dimensional used to ignore momentum because farmers said it's BS, it's now, it doesn't exist. It's phony. It's in the charts. And then finally, I think Ken French convinced them. The data was so overwhelming they shouldn't ignore it. Okay? But what they did do is this, let's say DFA would buy the stocks in the bottom 20% of P E ratios to keep it now a stock. How do most stocks get to be value? They were once growth stocks, and the stocks do poorly and they're going down once it got to that bottom 20% then DFA would buy it. Okay, okay,

Andrew Stotz 14:40
so just clarify what you're talking about is kind of a D rating where a stock share price is falling down because the fundamentals have slowed down, or something like that. So the market's no longer valuing that stock at, let's say, 25 times. It's now all of a sudden, D rated to be 15 or 10 times, 12

Larry Swedroe 14:58
or whatever the break. Point is, let's say 12 is the break point to buy. So now it goes down to a, you know, a PE below 12 and DFA prior to, I think 2003 would have bought it. They changed that to say, well, it will go on our eligible list, but it's suspended. We won't buy it until the negative momentum stops. So what does that do to your turnover? It lowers it because you delay the trade on the other side, how to value stocks deliver outperformance. They become growth stocks. Their PES rise as their profitability tends to revert to the mean, okay? And they get a turnaround. So the PE may have been eight, and now it's 12 and a half, and now they would sell it. David says, No, we won't sell it. We'll put it in our eligible to sell this, but we won't make it a priority to sell unless the momentum term is negative, so it'll be on the list, but it's not the highest priority in our algorithm to sell it. So that delays the sale, so they have a buy and even a hold range. Well, we'll we won't buy it. We won't sell it when it gets above 12, till it gets above 14, and then we'll sell it. So created these buy in bold ranges. So that's incorporating momentum without specifically trading on a signal to buy or sell. So that's one way you could do it.

Andrew Stotz 16:35
It's like creating buffer zones or places where you don't act but you it's, it's, getting closer to act interesting, right? So what

Larry Swedroe 16:42
that did value historically, if you look at any value fund, they're going to have a negative exposure to momentum if they don't screen for it at all, because that's how you get to be valued. So you have a negative, right, right? And, you know? So the problem is, what that was? So not the problem. When DFA did that, instead of having, let's say, a minus point one exposure to momentum, it went to like zero or plus point oh, five. So a very small amount of positive exposure. And that, if you think there's a momentum premium, for argument's sake, let's say you think it's 4% Well, a point one exposure gives you 40 basis points a year.

Andrew Stotz 17:29
And why? Why would they base it upon when, when the falling momentum stops, versus when the falling momentum stops and rising momentum returns?

Larry Swedroe 17:41
Well, soon as the momentum stops being negative, they say it's not a factor anymore. So that would become eligible the bar,

Andrew Stotz 17:49
okay? And I think that, how would you summarize the key lesson from this as a individual or as a professional, as you think about, you know how I'm allocating what is the key lesson from this chapter? From your perspective,

Larry Swedroe 18:07
first of all, rule number one is follow the empirical research. The empirical research says that past performance tells you nothing virtually about future performance. The best predictor of future performance are factor loadings and the expense ratio. So those are the two things you want to be looking at. And then you could add in turnover, obviously, as well. Those are the three things that you should look at, the construction rules used to create the fund. And then do they do patient trading to slow down the turnover and not pay away. You know, liquidity premiums there, you don't want to be a liquidity taker. That means you're taking the offer price and hitting the bid. You want to be on the other side where you when you have to sell, DFA doesn't go in and hit a bid. They'll put an offer in, maybe in between the bid and the offer, and they hope it gets taken. And if it doesn't, they got 100 stocks they're doing that for and they don't care which one gets taken. They don't know which one will do better or not. So they use these algorithmic trading programs, and that cuts down trading costs. And

Andrew Stotz 19:20
the last thing before we end, I just, you know, I want to highlight a little bonus here the article that you wrote about artificial intelligence and the risk of harking. Could you just tell us the general conclusion you got from that? Like, what were you? What were you? Was it? What you expected? And I'm curious, like, what your thoughts on about because there's some people that think, Oh, now that we got AI. I mean, I just, I just code something, and then I outperform

Larry Swedroe 19:47
right? Well, here's we've always known there's a problem with data mining, and soon as you got big computers and much better databases, the problem became massive. It, because when I was in college and studying, you know, Markowitz and portfolio theory, when if you had a theory, first you had to have a theory, then you would take your punch cards, believe it or not, to the data center, hand them in and pray you coded it right. Because if you made one error, it threw everything off, and then maybe three days later, because it was so expensive to run, you'd get your data back and hope it was right. So you couldn't test like 50 different theories, right? You ought to have some logical reason to believe that this correlation existed, because there was causation there. Okay, once you got big data, you could run all kinds of data mining, like a famous experiment on the United Nations Economic database, and they found that the best predictor of the S and p5 100 was butter production in Bangladesh. Now that's torturing the data until it confesses it has no meaning, right? Because there is no causation there. So now with artificial intelligence, what you have is the ability to use these large language models and much more massive databases you could not have used before because the computers weren't fast enough and you didn't have the large language model capability. So you could tell the computer to go find something and it will deliver but you have no idea about two things. Number one, is there any logical reason to believe that correlation will persist in the future. Is there some hypothesis that was created before the fact, not after the fact? Okay, I configured right? But here's the other thing that most people aren't aware. You have to know how you train the model, because you could have this look ahead bias. You're training it on the data that includes the data you're training it on. So what you should do is, let's say you're looking at the period 1929 through 2024 you should train it maybe on the data from 1929 to 2000 that's your in sample. See if we're and then run it on the out of sample. Post 2000 so you don't know unless they specifically tell you how they train the model. That's another risk. So you can get this stuff without any hypothesis. You know that's the problem. It's really a dangerous thing. And in the paper I wrote there, these are top level economists, they trained a large language model to look at like 200 factors and write peer reviewed academic paper qualities on all these factors, and it would cite the citations. And sometimes, by the way, it's stated phony stuff. It's amazing. They're not always accurate. So it can be a real problem. So if, if a

Andrew Stotz 23:15
young person comes to you and they say, Why do I have to learn all this stuff? AI is going to produce performance for me in the future. And I'm going to be a, I can be a hedge fund manager, and I've got, I've got a better AI model than the other guy, and therefore AI is going to replace fund managers, and these models are, you know, going to deliver? How would you say? Would you say, Yeah,

Larry Swedroe 23:42
I would say, from everything that I've read, that AI is a tremendously powerful tool that only works well when you have human beings overseeing understanding the processes, the construction of the models, how it builds portfolios, and all of the research that I've read so far has said that when you if you run just aa models and look at performance, or you run it combining a models with human intelligence, the human intelligence with the models does better than either human intelligence alone or the models alone, and

Andrew Stotz 24:24
I can outperform, I think I've highlighted so I can outperform. Now

Larry Swedroe 24:28
you may be able to outperform if you can figure it out before everybody else. That's the problem, because everybody, once you discover something, you know it's not like Renaissance technology is the only firm out there doing it. You got all these other big hedge funds in Apollo and AQR, they're spending 10s of millions of dollars, if not more, to try to figure this out. And once somebody figures it out, they may be leaving sought their own funds. And replicate. So the advantages are going to tend to be very short lasting. In my opinion, you'll have to keep out running it to gain any real advantage. But I don't mean it won't disappear. We've seen the hedge funds like Renaissance do well for you. But here's another indication of the capacity problem you read almost every month now, some big hedge fund that was highly successful is returning billions, if not 10s of billions, to investors so they can manage only their own money now, because they can't manage more than that, because the capacity won't allow for and I think that's a result of the problem we've just discussed, yeah,

Andrew Stotz 25:46
and it reminds me of the book Future hype, which was all about teaching us that that advancements, you know, are so like, if I think about how long it took for people to catch on to what, let's say, what's the medallion fund, the, the one you just mentioned, the the Simmons guy,

Larry Swedroe 26:14
yeah, Renaissance, Renaissance, sorry.

Andrew Stotz 26:17
So you know it took, it took the world many years to catch on what Ray Dalio and Renaissance and these guys were doing. But nowadays, if AIS now my AI is going to compete with your AI, and all of a sudden, the gap of time for inefficiencies to survive could possibly just get tiny and shorter and shorter and shorter,

Larry Swedroe 26:39
and a lot of it could even be legislated away. I don't know why the SEC, for example, doesn't put a stop to the high frequency traders by forcing them to stop these ghost bids. So in other words, it should force them to say, if you put in a bid or an offer, it must be outstanding for two seconds. But they ghost them, and they put and trying to out Fox everybody else. And so you see phony bids, and it tries to be misleading, but they could put a stop to all of that immediately, and that would cause a lot of their profits even to disappear.

Andrew Stotz 27:15
Well, the other thing which we see in the space of startups, particularly here in Asia, is that for years, it was free money. You know, any startup could get funding, and all of a sudden. And the same thing with AI models and all the funding they're getting and all that. But at some point, you know, I gave a speech last week where I talked about how a typical AI search is 10 times more energy consuming than a Google search, and at some point that's going to come that price is going to come home 100 times. The number I saw was 10 times

Larry Swedroe 27:46
I thought I read. That's even bigger than that.

Andrew Stotz 27:49
Yeah, my study that I had looked at may have been, you know, limited, but I wouldn't doubt it if it's 100 times, but let's just say it's massively more energy intensive than a Google search as an example, I

Larry Swedroe 28:02
don't want to we're short of energy capacity. Certainly in the United States, we're setting records now. I mean, we may not have enough heat or electricity for Texas and Florida because of a cold snap there. And imagine when you're adding, I mean, Amazon alone is spending 80 billion in the next year to build data centers. Where's the power going to come from, and where's the water going to come from? We're short water to cool all of these computers. We're short water, and they're building these things in the desert where land is cheap but there's no water,

Andrew Stotz 28:40
huh? Maybe it makes sense to buy a big piece of land in the northern hemisphere that's really cold and surrounded by water. Yeah? Maybe one last thing I was just going to say,

Larry Swedroe 28:54
maybe that's called Greenland. Maybe that's why Trump's trying to Yeah, that's

Andrew Stotz 28:58
where all our data centers go. I want to go back to 1967 I was two years old. My father graduated in his PhD in organic chemistry in 1965 and a movie came out in 1967 called the graduate. And in that movie, the young man, or the young man, was getting advice, yeah, and he got advice, and he said, what's the future? And he said, plastics, yeah, and my dad was going to work for DuPont, and he sold plastics his whole career, and he rode that way. And now, ladies and gentlemen, if you're listening in the future is power. And on that. I want to thank you, Larry, for this great discussion, and I look forward to our next chapter, and that is chapter 25 battles are won before they are fought, which is going to be a fun one, because we're going to talk about sunsuit and the Art of War, where we kick off. So I'm looking forward to that one. So. And for listeners out there who want to keep up with all Larry's doing, including that piece that he did on AI and all of that, just follow him on X, on Twitter at Larry swedro. And also, you can find him posting those things on LinkedIn. This is your worst podcast host. Andrew is not saying, ladies and gentlemen, I will see you on the upside. You.