Une nouvelle étude de Robeco confirme que les biais comportementaux des investisseurs existent depuis l'époque du Far West américain. Guido Baltussen et Pim van Vliet parlent de la tâche gigantesque que représente la constitution d'une base de données remontant aux années 1860, et de ce que cela signifie pour les investisseurs. Écoutez cette conversation.
Erika van der Merwe (EM): A major study from Robeco seems to have breathed new life into the famous Fama-French factors. The researchers tested for the existence of equity factor premiums, using huge volumes of manually collected companies’ data for the US, that go all the way back to the 1860s. Their findings confirm that modern investors’ behavioral biases have been around since the days of the American Wild West.
Guido Baltussen and Pim van Vliet are two of the three authors of the paper. Guido heads Robeco's factor investing strategies. Pim is co-head of quant equities at Robeco, and they've joined us for a discussion on the significance of this new study. Welcome, gentlemen. Really good to have you here.
Let's set the scene on this research that you've done. There are two critical aspects to this research – firstly the findings on the robustness of equity factor premiums over time, but secondly also the fact that you and your team did the hard slog in terms of putting together this very important database. Let’s start there: why was it so important to collect the data and build this new database?
Guido Baltussen (GB): Okay, so why do we believe this data set to be very important? Fama-French, they first started with about 30 years of data. And since then, a lot of academics have looked at those data sets and they further extended that [backward], roughly to the sixties and sometimes even to the twenties. But still, a lot of that data has been analyzed. Over that period, we have seen many developments in the US in terms of economic development, economic growth, a major war. However, the period before is also very fascinating. The period that we now extended with, gives you 60 years of additional data, that's the average period used in most of those studies. And it's also the period when really the US became the leading economic power in the world, when stock markets were very important, when they financed a lot of disruptive technologies. For example, electric cars were quite upcoming in that period, but also utilities, telecommunications, etc. So it's a very fascinating period where we can learn a lot about how markets behave, how investors behave, and especially what do factors tell us. Now, as an addition to that, we also know that finance research and also the research that we do, happens on the same kind of data by millions of tests. And that's a worry because that's what we call p-hacking or the factor zoo, that many things have been identified. But it's very important to offer those things and to study those findings that are very robust and also for us will be there going forward for our clients and that's where out-of-sample testing also really helps.
Pim van Vliet (PV): If I might add, so factors what’s a good example is value. I think we have done a podcast on value before. Robeco offers value-tilted products, both quant and non-quant, and a couple of years ago there was this question: Is there a value premium at all? Why would value outperform growth? It was an existential question, and the reason was that in the last decade, the last 15 years, the value premium basically was not there. The last year this has changed dramatically, by the way. But when we did our studies, this was in the midst of this question: is there a value premium? And by uncovering this new dataset, we could test, is there a value premium. Now, we tested more premiums, but the short answer to this question is ‘yes’. Also in this old pre-sample data, there is a value premium and that's very important for investors today who might be thinking ‘should I have a value tilt to my portfolio?’. So that's why we believe this study is extremely relevant.
EM: So we’ve got both of you, Guido having spoken first and Pim second, just for those who aren't in the room and who can’t see who spoke. So Guido, you really spoke about the importance of this data set and the way I understand the terminology, is that needing independent data to really confirm the modern views or today's views on factor premiums. And Pim, you are emphasizing the findings from this research. But let's look a little deeper at those findings. You've entitled this working paper ‘The cross-section of stock returns before 1926 and beyond’. So looking beyond that, you've spoken about the value factor. What else were your findings from the data?
PV: Maybe to kick off and Guido please add. So you mentioned the Fama-French factors. So these are originally size and value. We tested those. Size is also criticized: is there a size premium or not. So we could also test that. And we find that there is no standalone size premium, but size works really strong in combination with other factors. Fama-French do not include a low vol factor or momentum, so that is a shortcoming of their factor model. But these are proven factors which are added to the Fama-French model, which we also use at Robeco. So we also tested those. So low vol means ‘do low risk stocks outperform high risk stocks’ and momentum means ‘you buy the winners’, and then you wonder whether that continues or not. So these two premiums, and we also found them to be working in this sample. So value works, low-risk works, momentum works and size does not. And the latter is important because for size we were the first to have market-cap data because that was crucial, and this is the first study who has this data to be able to test also to make a difference between small caps and big caps. So some factors cannot be used, but we find that a common set of factors which are used a lot by practitioners and investors, which is low vol, value, momentum that also works in that sample. And that's very comforting. That you shouldn't make quant investing too complicated, and that a robust set of a few factors basically explain the cross-section of stock returns. The cross-section is an academic word, it means you can describe the behavior of stocks over time. So whether they go up or down, they often move together. So small caps behave as small caps together, and value stocks behave in the same way. And that's how you can basically get the whole stock market, so not just the market index, but all those stocks, which are a couple of thousands in our database, you can describe the behavior. That’s one, so that's for the risk, but also for the return, that's the premium. And our study is really opening up this data before 1926. And this data will also be more publicly available so that also other researchers can use it also to falsify our results. But our results make sense, we are pretty confident with all the data cleaning that it's robust.
EM: So let's talk a bit more about the data, the database, the data cleaning, you used all the terms, p-hacking, etc. So basically what you're saying was, with the limitations that you might have had in building this new database, there really is no opportunity or scope to be torturing the data. It is what it is, it’s really quite simplistic. So just for the background experience, Guido. What is it that you had to do to the data? What were your limitations, for instance, in the kinds of factors that you could test for? And what proxies, for instance, did you need to use given that not all of it was available?
GB: Yeah, it was a lot of blood, sweat and tears as we like to say in Dutch. What we did, this is a project that started about five years ago in collaboration with the Erasmus University, also based here in Rotterdam, the Netherlands. We hired a lot of students, student assistants, also internships here at Robeco, who we gave the assignment to dive into the data, hand type data into our databases and cross-check that. Some of those actually, Bart van Vliet who is also now a PhD student and working here at Robeco at our quant team, did most of the work. He likes to make the joke that he started without glasses and ended up with glasses. I think that's also a bit how it happened because he had to look in all those historical archives, look into what were the stock prices, what were the dividends, wat was that cross-section of that hundreds or thousands of stocks that traded at that time, and especially also what were the market capitalizations? Because you do see there were a lot of stocks also at those times, but like nowadays, some are less important for investors than other ones. Apple is way more important than a small stock listed here locally at the small-cap exchange in the Netherlands. So you do need to account for that.
PV: So the basic added value was, there is stock return data available in the global financial database, which is a database which extends to CRSP database, which starts in 1926. The problem with that, if you take that off the shelf, is that you don't have market-cap values. And that means that you don't know whether it was a large cap or a small cap. And especially what we see with data also in the twenties and also in this century, that usually errors and pricing errors are bigger in small stocks. Sometimes the reporting is not right. So a very, very effective way of getting a clean database is knowing whether it's a large-cap stock or whether it's a small-cap stock. And also when you want to test investment strategies and you want to know whether it's investible or not, this data piece is crucial, and that's what we added. And it was Bart van Vliet and the students and the interns. And that was really going into the journals, making a photocopy, then hand typing it over. So it's really the Chinese army approach, it's lots of labor, and this gives us then a unique edge.
EM: Really historic. Looking at the quite extensive media coverage already for this paper, one of the statements in a very well-put-together graphic was that the value effect has existed for longer than we knew. So elaborate on that.
PV: Effective premiums seem to be an eternal part of markets, where markets are. What makes us human, is that we can trade. You never saw a dog trading a bone with another dog, that's something really unique to humans. Like even making music is something which animals can do. Trading is really a human feature, makes us unique. And when humans trade, we have some biases. For example, overconfidence is the mother of all, some say. And also when you do a trade, it's always a win-win, because if you buy something, the other one sells. And that means you have to be confident that you gain something by doing this. So what's really beautiful about our database is that the human behavior, which goes in cycles, because in your lifetime you learn, you have experiences and there's also behavioral finance showing us what you experience determines your preferences about risk and return. And that means generations. So if you have 150 years of data, basically you have 7-8 generations, which is nice, but still you would love to have hundreds of them. So having 7-8 really gives you the opportunity to look at these long-term premiums and see whether they exist. The existential question: is it not just luck that maybe cheap stocks did well in 20 years, but maybe it was just coincidence? And that's the impact of this study. And that's why it's also picked up in the media, is that this really shows that this is sort of a foundation of markets and also that it's not going away or something. And the fact that it's not going away, that's also what we can confirm in practice because we set up strategies to profit from those factors. And it's not easy. It's not easy.
EM: What’s not easy about it?
PV: It's difficult to harvest premiums because also we, quants, are humans. You wouldn't say, but maybe quants even have more emotions. And that's why they stick to rules and strategies because it's sort of a protection for rationality. So the thing with harvesting factor premiums, what makes it difficult, is that for example with the value premium, it didn't show up for ten years. And then you start to doubt. And what research also shows is that people throw in the towel exactly at the wrong moment. So that means that you give up once you should pursue. And when I was doing my PhD thesis, Guido referred to it, I was looking at factors going back 50 years or something like that. And I was like, wow, this is really cool. You can make alpha money. It’s so simple. You just buy cheap stocks, low risk, that’s it. So I spend my PhD to check if it was really that simple. Long story short, yes, you can make money. But then when I entered the industry, then it became more difficult because you have benchmarks, peers, career risk, your insecurity, and short-term reporting. A year is extremely long in the industry. Well, in a back-test, it’s just one observation, you know, so that’s why I say it’s difficult. But Guido, maybe you think it’s easy to harvest factor premiums or do you say no?
GB: I think I fully agree with that. That’s what we see in practice. You need a lot of skill, but also patience and overcoming your own biases to harvest factor premiums. And we believe, actually, that's also one of the key reasons why they're there and will remain there, because it's not easy. It's not easy money. That's also the interesting thing, I think, about this deep sample. The investors that encountered the markets in the 1900s, those were more or less our great-grandfathers. They did not think that differently from how we think nowadays.
EM: And was that a surprise for you?
GB: Not at all. But that's for a lot of people. They think like, OK the last ten years it's very interesting. We already often encounter the last 20 years, people were quite different then. A hundred years ago, probably they still lived in caves, something like that. I'm exaggerating a bit, but that's sometimes the impression people have. But these were just our great-grandparents. They had similar brain powers that they thought and similar manners. They were also very smart. I like the analogy of nowadays people talk a lot about high-frequency traders; and high-frequency traders they compete for speed. And from that profit, they try to make the shortest cable between the stock exchanges in New Jersey, Chicago, New York to gain some nanoseconds or even quicker. In the 1900s, the same game was there, but then it was not via the computer but via the telegraph and then high-frequency traders, jobbers called at that time, they competed for speed by following the shortest telegraph cable between Boston and New York stock exchanges to gain some seconds. So to speak, yes it’s a bit quicker, we cannot even notice it, that quick. But the same kind of behavior incentives investing for making money is there.
EM: So you're saying people are the same. Those cognitive biases are fairly consistent over time. Time for a tempo change, time for our quiz. I know you both love puzzles, you’re also quite competitive. So let's put you to the test. We're going to play you some audio clips. So let's see which one of you can identify the voice first. But it's not that difficult, there is a clue: either or both of you follow each one of these people on Twitter. So we’re gonna find out why you follow them on Twitter, why these people are important. So first clip. Audio fragment: Quant doesn't mean you're a Vulcan. Quant does mean you run the process like you're a Vulcan. And that creates some internal tension. If you are subject to emotion, you've got to fight that. That's something I haven't done much very well in the last few years, but I've done that.
PV: That’s Cliff, I think. But he wasn't very angry in this part.
EM: He was very calm.
EM: It was Pim who got it right. That was Cliff Asness, indeed, speaking to Bloomberg. And why do you follow him? Why is he important?
PV: He's a fellow quant. He has written a lot about momentum. So Robeco wrote lots about low vol. He is also speaking out. He uses Twitter to also show his latest research, sometimes also politics. So that's what I follow. And he's an entertainer. Sometimes he also picks a battle. Quant fights are really nice to watch.
EM: Does anyone understand them outside of the inner circle?
PV: Maybe not. But still, then it's fun to watch.
EM: And he said about being a Vulcan and that would be a Star Trek reference, right? I'm not a Star Trek fundi, but he's talking about quants being unemotional. And we did touch on this earlier that even though as human beings, you are not unemotional, of course. But that's why you have models.
PV: I dare even to say that on average quants are not less emotional, maybe even a bit more. Also, Cliff Asness is a good example of that. I'm also not the most rational, but models are really a good way to tame your spirits and taking a rational approach, because we believe that that's giving you long-term alpha and also this study.
EM: So talking about rationality, here is the second voice. Guido, here's your chance to catch up. Audio fragment: As to how rational I am, I'm probably the last person to ask because one of the deepest kinds of irrationality that is baked into us, is that we all think we're perfectly rational and that only the other guy is irrational, sometimes called the biased bias, namely everyone else is biased. I'm not.
GB: I have no clue.
EM: He has American accent, but I think he is based in the UK. Rationality, research on rationality and human thinking, aspects of language, mind and human nature. It was Steven Pinker being interviewed by Freddie Sayer on Unheard.
EM: I can't quite recall which one of you follows him, but if you can let us know why he's relevant for your thinking and your research.
GB: I leave this one to you.
PV: Steven Pinker, he voices this rationalism and also optimism. Matt Ridley is something related, I follow him more, like him more. Steven Pinker has a bit more the common Western beliefs about progress and enlightenment. So that's why I follow him, to see what he's up to. And also because I know others follow him. I think the link is with rationalism, optimism and enlightenment. That's also typically a clean sort of quant approach. But they're not two markets but more two societies.
EM: Right. Ready for the next clip? Audio fragment: We did another article called ‘The best way to add yield to your portfolio’. And we kind of demonstrate, we walk through that, how much do you make per hour or how much time do you spend on investments per year? Here’s how much alpha you have to generate.
EM: Meb Faber, who was interviewed on Real Vision Finance. So Guido, who's Meb Faber and why is he important?
GB: Meb Faber is a quant, based in the US, who has done a lot of interesting studies. I think he actually has one of the most downloaded papers on the Social Science Research Network (SSRN), where a lot of quants and also academics post their work. So that it can be shared in the cloud, in the air, and everyone can access it. He has done some great work there and also in quant space on for example, quant investing, trend following investing, factor premiums. PV: Yeah. Really cool guy. He’s also doing a podcast, a couple of hundred now, it's running and he makes quant investing, he talks about it in a very casual way. He has very interesting guests on his shows, so I can recommend his podcasts.
EM: Excellent. Last one. Audio fragment: My odds are good. I'm on a winning streak. Everybody in this place wants to get in on the action. How can I lose, right? Now this is a classic error. In basketball, It's called the hot-hand fallacy. A player makes a bunch of shots in a row…. People think whatever's happening now is going to continue to happen.
PV/GB: Dick Taylor
GB: This is Richard A. Taylor, that's his official name. Dick Taylor is a professor in Economics at the University of Chicago. Nobel laureate a few years ago and essentially the founder of behavioral finance. I also had the honor to meet him and work with him about ten years ago, when we worked on studying behavior in large game shows, for example, but also other contexts. And he's really the founder of behavioral finance and that inspired a lot of the things we look at, like value and momentum. He was also one of the first to discover those.
PV: Interestingly, he's also at Chicago University, which is always seen as the rational market, free market. So that's the behavioral finance economics professor is there, was really a paradigm shift, I don't know when he joined, it was about 20 years ago or something. And also his Nobel Prize is a testimony to that.
EM: In 2017. So for a bonus point, do you know where that quote was from, that audio quote?
GB: I know it's about a particular study he did, or something he studied quite a bit on the hot-hand fallacy, especially with his fellow psychologist at the time. But where it's from?
PV: From the movie? He appeared in a movie, The Big Short. That's where he explained the game.
EM: Well done guys. If we were to take your perspectives from your research and we look to where we are now and where we might be headed, so much is happening right now as we are recording this podcast. What is your, what are your conclusions, your assessment of what era we might be in right now?
GB: I think a key conclusion is factor premiums are very persistent and you can expect them to remain there and be there the next years ahead or the next centuries, actually. But it requires patience to harvest them. On the other hand, we also know that the last 20, 30 years have been quite special. Low inflation, the favorable economic growth, etc. We have already seen some shocks that might well be, that is also a bit more than normal, also with higher inflation, etc. We know factor premiums are pretty persistent across those cycles, so that also gives a lot of comfort to us. Factor premiums is something that you can structurally embed in portfolios to make your portfolio more stable, increase the returns that you make on them a bit.
PV: I think to add, investing is not a test of IQ, it's a test of character and what we think now with lots of data coming up and that we become much smarter, but do we become wiser? That's the question. Also, when we did our studies, we noticed this overconfidence. We know that everybody thinks they're above average. This applies to individuals in a cross-section, but it also applies to generations through time. So often people read only books from their own time. They only read about technology from their own time. So we are a bit overconfident that we are top of civilization. And in history this was always the case. The Greek thought they were on top. The Romans thought that, and the Etruscan culture thought that. Everybody always thinks now this is different, this is unique. And I think this study makes us a bit humble, if you look at the sophistication of our ancestors and that we should not be too arrogant, that this time we are so much smarter or better. And that's basically also confirming that this test of character, that it's difficult to make money on the markets – acknowledging that – and that a factor-based approach might be helpful to give you some extra returns, but you need to be patient. And we hope that this research helps to build character. So a bit of knowledge, of course, but also confirmation and belief. Because we see, to be a good investor, it's really about those traits and not so much about having the best data or the latest data.
EM: Indeed. Pim van Vliet, Guido Baltussen, thank you so much for your time and your insights. I enjoyed our conversation.
PV/GB: Thank you
Les informations présentes sur ce site Web sont destinées exclusivement aux professionnels. Un investisseur professionnel est : un investisseur qui, à titre professionnel, dispose d'assez de connaissances et d'une expertise et d'une expérience suffisantes pour pouvoir évaluer de manière adéquate les risques financiers liés aux décisions d'investissement prises par lui-même.
Les visiteurs de ce site Web doivent être conscients du fait qu’ils sont eux-mêmes tenus de respecter toutes les lois et règlements en vigueur dans leur pays.
En cliquant sur J'accepte, vous confirmez que vous êtes un investisseur professionnel. Si vous cliquez sur Je n'accepte pas, vous êtes orienté vers la partie réservée aux particuliers.