Campbell R. Harvey is a Professor of Finance at the Fuqua School of Business, Duke University and a Research Associate of the National Bureau of Economic Research in Cambridge, Massachusetts. He served as President of the American Finance Association in 2016. He has also published over 150 scholarly articles on topics spanning investment finance, emerging markets, corporate finance, behavioral finance, financial econometrics and computer science.
For this week’s episode (our longest to date), we get together with the legendary Professor Campbell R. Harvey and take a deep dive into a diverse range of topics that draw on his incredible breadth of knowledge and extensive research. Campbell is the Professor of International Business at Duke University and is also a Research Associate at the National Bureau of Economic Research. In 2016 he served as the President of the American Finance Association, and from 2006 to 2012 he occupied the incredibly demanding role of Editor for the Journal of Finance. One of his earliest achievements was identifying the inverted yield curve’s ability to predict a recession, a highly regarded metric that is near-ubiquitous in its implementation. For the first half of our conversation, we focus on his research in areas like skewness and emerging economies. We cover specific topics like the factor zoo, why it’s problematic, and how Campbell, along with his student Yan Lui, found through their research that approximately half of the published empirical research in finance at the time was, in fact, false. We also unpack his most downloaded paper entitled The Golden Dilemma and get into the intricacies of why gold is an unreliable inflation hedge. For the latter half of our conversation, we hear about Campbell’s latest book DeFi and the Future of Finance along with his most recent research. Discover how Campbell first became interested in the topic several years ago and decided to put together a course for his students. We also delve into the rise of decentralized finance (DeFi) and how we can expect it to shape global finance, trading, and the future of the internet. Join us today for this essential episode on everything from the pitfalls of academia, to emerging markets, to Bitcoin, and much more!
Key Points From This Episode:
Introducing this week’s guest Professor Campbell Harvey. [00:02:46]
How Campbell’s research brought him to Chicago's Ph.D. program. [00:03:55]
How Campbell identified that an inverted yield curve had preceded the past four recessions and could be a reliable economic predictor. [00:07:03]
Hear about Campbell’s research on skewness, as opposed to simply mean and variance, which is often the focus of portfolio theory. [00:11:40]
Why it’s surprising that skewness is still largely disregarded in favor of mean and variance. [00:16:42]
Why mean and variance are insufficient for measuring risk when comparing a concentrated portfolio with a more diversified portfolio. [00:20:45]
Some of the special considerations that Campbell prioritizes when assessing emerging markets in context and managing an overall portfolio.[00:22:04]
Observations on the cost of capital being higher before integration and liberalization. [00:25:11]
The implications that Campbell’s research on emerging markets has on asset allocation. [00:26:51]
Dynamic asset allocation, Campbell’s research in emerging markets, and how those lessons can be applied when investing in emerging markets at a time when the cost of capital is high. [00:30:04]
The factor zoo, why it’s problematic, and how it is caused by data mining. [00:32:26]
How Campbell and his student Yan Lui estimated that half of the published empirical research in finance was false and how this has occurred in other industries due to data mining. [00:33:02]
How economic incentives from the investment industry inform research. [00:39:38]
The important distinction between academic research and practitioner research, and asset management. [00:44:15]
The extent to which asset management research could be considered to be more reliable than academic research. [00:47:23]
Some of the mistakes that investors make when they pursue these factor premiums that have been identified [00:49:29]
Machine learning and its impact on investment decisions for retail and institutional investors. [00:56:06]
Whether the benefits of potential alpha from machine learning will be passed on to investors or remain within a firm as their scale increases. [01:00:22]
Campbell’s research on traditional active management within the context of a firm’s ability to continue delivering alpha in the future, and how that incrementally decreases as their asset base increases. [01:06:23]
The arguments in favor of allocating gold to a portfolio, especially at times of higher inflation, and whether it holds up to scrutiny. [01:09:54]
How technological changes can affect the real expected return. [01:16:51]
Why gold can be a valuable asset in diversifying your portfolio. [01:17:22]
How Campbell became interested in DeFi, cryptocurrency, and blockchain technology. [01:19:19]
How digitized finance cuts out the inefficiency of having a middle person and fosters inclusion and financial democracy. [01:26:35]
Dr. Harvey’s thoughts on how cryptocurrencies facilitate criminal and fraudulent activity. [01:31:07]
How DeFi could disrupt traditional asset management and how to prepare for those changes. [01:36:43]
How to invest in DeFi even though it’s decentralized. [01:38:33]
How companies can increase their revenue by using cryptocurrencies in their transactions. [01:43:24]
Why the current wave of FinTech will be replaced by DeFi. [01:44:58]
Why it’s important to have a very diverse portfolio when investing in DeFi. [01:51:22]
How DeFi will allow people to monetize their content and disrupt the money that Google and Facebook make from their users’ data. [01:52:22]
Some of the risks of DeFi and investing in cryptocurrencies. [01:55:27]
How Campbell defines success: positively impacting the world. [02:00:17]
Read The Transcript:
Some of the earliest research was done while you were at York University here in Canada. That research was so good that it ended up bringing you to Chicago's PhD program, which I think is just amazing. Can you tell us that story and what the research was?
It's actually completely unexpected on my part. I do my first year of my master's and get a summer internship. I get dropped into a really interesting job. I'm put into the corporate development team for Faulconbridge copper, at the time, the largest copper producer in the world. They gave me a pretty simple job, at least to describe it, but a really complex job if you think about it. They wanted me to develop a model to forecast GDP growth. It's pretty good reason to do that for them. Copper is very procyclical and it's very costly to shut down a mine or to start a construction of a new mine.
They really value any information about what will happen to the economy in the future. Me, as you know, first year master's student, that's my job for this giant corporation. I thought I looked around and I saw that there were all these econometric services that you could contract for very expensive using very complex models. My intuition was, "Let's do something simple." When I look at financial assets because financial assets actually have information about the future in their prices today. If you think, for example of a stock. Well, the price is the discounted value of the cash flows. Those cash flows are strongly dependent upon the state of the economy.
That was my first stop looking at stocks. I read some of the academic literature and then quickly discovered that stocks were going to be unreliable. At the time, there was some joke that the stock market had predicted 11 of the past five recession. I started to think about, "Well, why does it fail?' The intuition was pretty powerful. That number one, there's no fixed maturity for a stock. You don't know how far those cash flows actually go out. Number two, the dividend, which is the cash flow to the investor is variable. It's might be zero for a number of years. Then third, the risk that you actually discount with changes through time.
That is basically enough to produce a lot of noise, a lot of false signals. My next stop was to think about the bond market. The bond market, some of these disadvantages for stocks disappeared. If you think about a government bond has got a fixed maturity, it's got a coupon that you know in advance. Let's say you're going to get a coupon for 10 years on a semi-annual basis. Then the third aspect, the risk, well, it is a government treasury bond. It seemed to check all the boxes. I did some preliminary research and showed one thing that was important was to get rid of expected inflation. That's why I looked at a spread or a yield curve, long-term money short-term.
I saw that this have remarkable ability to forecast GDP growth. I put this together and I'm about to present it when the entire corporate finance group got laid off. I thought I'd done good work and they didn't see it. We were gone and basically a cost reduction sort of thing. I naively, looking back, wrote to the board of the directors to complain. "I'm a student, how can you lay me off after five weeks of work." That didn't go well, obviously, but I had some extra time in the summer. I finished this project and I thought it was interesting work. Then, I went back to York and showed one of the professors that I was doing some work for the project.
They were very flexible. They said, "Well, there's a number of courses you need to take, but we're going to put them into just a single course. Your job is to work on this paper because it's really interesting." This was again, really surprising to me. I worked on the paper and then presented it. Then the feedback I got from the faculty is you need to apply for a PhD. PhD, I wasn't even thinking of anything like that. I didn't really know what to do. They gave some recommendations. I applied to various programs. I didn't know in real time, but in hindsight, my application was the ideal application because what I did is I included the research paper in the application.
The people evaluating the applications said, "This person actually is capable of doing some research. This is rough, but at least there's an idea here." I got various offers and decided to go to the University of Chicago. That's basically how I got started. At Chicago, I had a bit of an advantage of course, because I showed up with my paper. I started working on that immediately and it turned out okay. I've got to admit that during my dissertation defense faculty were skeptical. I showed that inverted yield curves preceded four of the past four recessions. I said, "Well, a four and a four that can be luck. "They were impressed. There were a couple of things that impress them.
Number one, that I got the double dip recession in the 1980s and other forecasters didn't get that. Number two, that the model was based upon sound economic theory. It wasn't just a data mining expedition. You can discount the luck in there. Number three, the cost of the forecast to go to these services, to get your forecast, it was extremely expensive. Whereas my forecast was the cost of a Wall Street Journal, which at the time 25 cents. They passed me. We're still talking about the dissertation. Look, this is the way it works in science, as you know, you publish a paper and there are two things that could happen. Number one is the good scenario where the effect gets weaker, it doesn't work four out of four, it gets weaker. Then number two is the bad scenario. It completely goes away and that's possible also. It could just be a lucky finding. For me, we've had four recessions since I published my dissertation and you'll curve inverted before each one of them with no false signals yet. That's why we're still talking about it.
It is amazing. It's such a well-known metric today. Very cool. It's very cool to hear you tell that story. Some of your other research that's very highly-regarded is the work that you've done on skewness as opposed to just mean and variants, which is what often gets discussed in portfolio theory. Can you talk about those findings?
One of the most influential papers for me was Harry Markowitz's 1952 landmark Journal of Finance portfolio theory paper. I've read it many times and everybody should read this paper because it's foundational. A careful read reveals two key assumptions, that Harry, and I call him Harry, because I actually know him and he's a coauthor.
He's very clear on these assumptions. Number one is that you exactly know the inputs. We're trying to develop the optimal portfolio that's got the highest expected return for some level of volatility. In doing that, he assumes that you exactly know the expect of returns, the volatilities, the correlations. There's no uncertainty whatsoever. That's the reason that if you alter just by a small amount, the expected returns, for example, for a particular asset, it could completely change the portfolio weighing. That's number one, that uncertainty is not really incorporated into that model. You need to allow for uncertainty in the real world. Real world implementations are much different than the textbook implementations.
I make it a point of showing my students how to implement these models like they do in the real world, rather than the theoretical. There's something else in the paper that's in a footnote on page 92. I remember the page very well. Basically, what he says is, " Well, this framework is only optimal if the investor has no preference for skewness." It's right there. That means that the investor only cares the risk that the investor cares about is variants. We know intuitively that people care about other things. The downside is really disliked and it's disliked more than the upside. There's plenty of experimental evidence that shows that there's downside version.
Also, the case that the universe of assets that we look at are not normally distributed. It would be unusual to find that. You put those two things together, you've got the assets are not normally distributed. You need other parameters to describe the distribution. Plus investors have a clear preference for skew. They like positive skew, the big upside and they dislike negative skew. That led me and I'm certainly not the first person to actually talk about some of these higher moments. What I did is I implemented this in an extension of the capital asset pricing model. That model basically says then again a famous paper by William Sharp in 1964, that the risk of an asset is its contribution to the volatility of the portfolio.
If there's an asset that increases the volatility, that increases the risk, then its expected return needs to be really high to compensate for putting that into your portfolio. If it is an asset that reduces the volatility, a hedging asset, then you expect the return could be lower, negative even because it's providing that hedge. I'm thinking, "Well, there's got to be an analogous situation for skew." We've got some portfolio that has got some volatility, but it also has some skew. The intuition is identical in my model that if an asset, when you add it to the portfolio adds to the downside risk, then that asset needs to have a high expected return to compensate for taking on that downside risk. My paper which was published in 2000 in the Journal of Finance Conditional Skewness in Asset Pricing Tests is pretty well cited paper, and it actually shows how to do the so-called efficient frontier, where we trade off expected return and volatility in a three- dimensional way. Where we put skew directly into the equation.
Do you think that's being used enough in practice? We still hear about mean and variants a ton and not as much about skewness.
It's remarkable to me that it's been so many years and we still make the mistake of operating in a mean and variants world. People will take a look at different investment strategies and pick the one with the highest Sharpe ratio. Well, let's take a step back. The high Sharpe ratio might be a strategy that has got a giant downside. It's got negative skew. Then the lower Sharpe ratio maybe has the positive skew. The variation in the Sharpe ratio is purely explained by the risks. This is really important because often people will claim that their trading strategies got some alpha. Then you ask them, "Well, how are you measuring that alpha?"
Well, we're using the capital asset pricing models. Well, that's only appropriate if you only care about mean and variants. You could have a giant downside and that's generating on average higher returns, but as purely a risk. Indeed, I've got an exam question that I've used and hopefully my students don't listen to your podcast.
It is a true story that a student gets me on the phone, an alumnus. Graduated a number of years ago and wanting to use some of my materials, was seeking permission. Then, I said, "What do you do?" He said, "Well, we're incredibly successful. We're close to a billion dollars AUM and a strategy is really simple. All we do is we buy the S&P 500, nothing fancy. Then, we consistently right out of the money put options. Then, we've done two to 300 basis points of alpha every year. The fund is growing. I have had many students, I thought this might be one of my students. I said, "You didn't take my course in asset management, did you?" He said, "No, I didn't take it." That was a mistake. I said, "Well, I know you didn't take it because you actually have zero alpha. What you're doing is just increasing the risk of the portfolio. What you think is alpha is just a compensation for taking risk. At some point, there's going to be a draw down in the market and you're going to greatly underperform."
It is an exam question that I give now to students actually it was a useful conversation because I can tell the story. It's all about skew. There's an interesting, this is not my paper, paper out there that might surprise people in terms of the finding. That is that if you look at all of the stocks that were ever listed in the U.S. from 1926, and look at their lifetime returns just measured the lifetime returns, that 56% of them don't beat the Treasury Bill. Basically, 99% of the value that's been created is just like a handful of stocks. Let's say 100 out of 25,000. What's the story there? It's skewness. A bet on the market, you're looking for that giant upside and most firms just don't get that.
I have a follow-up question on this. The skewness and the best and binder paper that you just mentioned is something that we've talked a lot about on this podcast and skewness as well. If we're comparing a concentrated portfolio, we'll talk about factors later.
I'm getting ahead of myself, but I want to ask while we're talking about skewness, if we're comparing a concentrated portfolio to a more diversified portfolio, you'd expect more skewness from the concentrated portfolio are mean and variants inappropriate or insufficient measures of risk to compare those two portfolios?
Yes. People say, "Well, you shouldn't have a concentrated portfolio because you're not diversifying." People don't understand the concept of diversification. This is again, remarkable to me that there's a few things that people in finance think that everybody understands.
One of those concepts is the personification. It really depends upon your preferences and you need those preferences to be over, not just variants. If it is variants, then it doesn't make any sense not to diversify because you get rid of a lot of risk in diversification. However, if you've got a preference for skew, then you might have to take what appears to be an undiversified portfolio to get that profile that you want in terms of the upside. Those criticizing the concentrated portfolio just don't understand the concept of diversification in the broader sense. Where you're looking at diversification, not just in terms of variants, but in terms of higher moments also.
You've done a lot of work in emerging markets. Can you talk about the special considerations in assessing emerging markets in the context of managing overall portfolio?
I was very fortunate as a junior faculty member to get my hands on a database that was being developed by the World Bank and the International Finance Corporation. I had the data two years before everybody else did, and it was a great opportunity for me. Indeed, I basically shifted my research. I basically dropped everything. I thought this was an opportunity for me to make an impact in what I considered a very important sector of the world economy emerging markets. In dealing with these markets, there's many special considerations.
The first thing that you might think of doing is just apply the standard tools of finance, like the capital asset pricing model. Indeed, I did that just to see what would happen. The problem with that is what we've been talking about already that these assets just don't look like assets in developed markets. There could be giant downside risks, for example. It doesn't fit really nicely into our framework. Then on top of that, there are a number of barriers to investing in these markets. There might be restrictions on foreign investors investing in these markets. There could be restrictions on domestic investors in an emerging market, having a portfolio diversified outside of that market.
These were all the challenges that I was dealing with. It was an ideal atmosphere to apply a model that had a preference for skew. Then, one of the things that I did with a coauthor is to actually try to model the evolution of these markets in terms of becoming more integrated into world markets. If we think of, let's say the U.S. and Canada as an example, these are integrated markets. You assess a project in Canada and a project in the U.S., the risk profile would suggest the same expected return. An identical project in Zimbabwe, it doesn't have the same expected return because that market is largely segmented. There's so many barriers. That was part of the challenge in dealing with these markets. It wasn't just that they were different. It was that they were evolving. In order to have a portfolio strategy that needed to be taken into account. Just to look to the past, to calibrate, what you're going to do is going to be misleading if there are some structural changes happening in these markets.
The other thing that I did, and this is a paper that did get a lot of attention was I showed what happens as a result of a liberalization in emerging markets. For example, in emerging market might take down the restrictions for foreign investors coming in and buying equities. That turned into a large scale study. It was a lot of work to put the data together on these various liberalizations. What we found was a remarkable increase in real economic growth as a result of a liberalization. It makes sense that if you think of the economics or closed economy, you open up your equity market to foreign investors.
The foreign investors want a piece of it because they want to diversify portfolio. Prices go up, expected returns go down. The cost of capital for corporations within that market goes down. The cost of capital goes down, investment goes up, investment goes up, employment goes up, GDP goes up. This was the thing that we were documenting. We were finance researchers competing against all of these very distinguished researchers in development economics that hadn't looked at this angle. Our claim was that this was a huge chunk of extra growth that was low-hanging fruit for many of these economies. Again, the paper got a lot of attention in terms of just the size of the effect was so large.
Are there implications from your research on asset allocation?
Yes. My research is really well-founded and asset allocation is where I began my research. We've already talked about asset allocation in terms of skew, that the skew needs to be taken into account, but it's more than that. One of the early ideas that I had published a paper with Wayne Ferson in the Journal of Political Economy, one of the top econ journals.
The title of the paper is called The Variation of Economic Risk Premiums. If I was writing that paper today would have a different title and the title would be Factor Timing. That's what we were after back then. The risk premia changes through time. The intuition, you could just look at the stock market as a whole, as an example.
If you're in bad times like a recession and the stock market is really low, to get people to invest in the market, that's pretty tough. You don't have a lot of extra capital around. To get people to invest the risk premium or the expected return is really high. If you're at a period in the business cycle, good times, stock market is really high. Then lots of money flowing in, then the expected returns are lower. The risk premia for the market is very countercyclical. I documented this in a number of different papers, but there is a direct implication for asset allocation. In asset allocation, you need to take these economic dynamics into account that the expected returns are not, again, you just can't take the average over the last 75 years.
You need to take into account the current conditions. The state of the economy today in building your model for asset allocation, which has critical inputs. The expected returns are changing through time. The volatilities, the correlations, all of these are changing through time, as well as the skewness. My paper in skewness is called Conditional Skewness in Asset Pricing Tests.
Conditional means condition upon the economic information that's available today. Again, there's a there's asset allocation. A lot of it is naive where people just use past averages. When do you use a past average, like the past 10 years average return of our volatility or correlation, that is the same thing as ignoring all of the current information. You're saying that information is completely irrelevant. What I've done in asset allocation in my research is to incorporate this information about the state of the economy, so that you get something that is dynamic through time, and depends upon business conditions as it should.
I know one of the questions our listeners are going to have as they're listening to what you're just saying, we just talked about emerging markets and the cost of capital being higher before integration, before liberalization. When the cost of capital is high in an emerging market, is that a reason to over-allocate if someone has the risk capacity to do so, and they're seeking higher expected returns?
This is really important point. It directly links to this dynamic asset allocation and my research in emerging markets. I told you the story about the dynamics as the country becomes more integrated. Often, that doesn't happen overnight. It's a series of measures. Often you can see that, so bills are introduced.
They have to go through committee and process, and it might take two years before it actually happens. At that point, when the investors come in, it drives the prices up. To get in early, it's a trade. It's a.
So to get in early, it's a trade. It's a trade that's based upon a foundation that is very clean, in terms of the economics. So, if a country will become more open, that is generally, a good thing. Prices are, on average, not every single time, but on average, this is a way to make a higher expected return. And then once it's fully integrated, then it could be useful in terms of diversification still, but this extra amount, the so-called liberalization premium, well that's already gone.
Right. Historically, do emerging markets become more developed? Does that cost of capital tend to drop at some point?
Yes. And indeed, the market as it progresses becomes a developed market. And that's basically the story of emerging markets, they're emerging for a certain period of time and then become developed. So, it's like the International Finance Corporation's goal is not to be needed in the future, when there'll be no emerging markets, everything will be developed.
Interesting. Earlier you mentioned factors, so some people call the current environment the Factor Zoo. Can you describe what that Zoo is and why it might be problematic?
Yeah. The Factor Zoo was coined by John Cochrane in his presidential address to the American Finance Association. I've got a paper that I published in 2016, and it's called, .... and the Cross-Section of Expected Returns. Basically, when I was editor of the Journal of Finance, I noticed that there were so many papers that were submitted with in the Cross-Section of Expected Returns, so then propose a factor. So a factor one in the Cross-Section of Expected Returns.
Factor two, in 2014, I decided after I was retired from editor to actually just catalog all of these papers that were basically trying to do the same thing, to try to find something, to explain the cross-section. And when I say explain the cross-section, that means that, well, if the exposure to this factor is high, then you've got a high expected return. And if it's low, low expected return. So I thought, well, maybe we'll find that the complete database of all the papers I'd handle about 8,000 or 9,000 papers. So a lot of papers. So I had some students going through and one student in particular PhD student of mine, Yan Liu who's at Purdue University now. I remember Yan coming to my office and I'm expecting, he's got a list of 25, he comes to my office with 200.
And this is just the Journal of Finance. And we're looking at each other and instantly we realized that, oh, this is a problem, a big problem. And let me describe the reason it's a problem, that if you've got researchers that are out there looking for some factor that explains the cross-section of expected returns, and given that there are millions of possibilities out there, if you work hard enough, you're going to find something that appears to work historically. And that is basically you find the stuff by luck has nothing to do with economics. So, as a result, you need to alter your approach to take into account, how many things have actually been tried? So, if I try 100 factors that are random, so not even real, likely five of them will appear to work just by chance. So, you need to take that into account. So, if I found 10 rather than five, then well, maybe there's something there. But if I find five, well, that's just what you would expect given just by random chance. So what Yan and I realized was that we'd been making a mistake in finance in terms of our approach to discovery of these different factors that we needed to take into account what was already done historically. So, if basically you discover factor 302, that's not an independent discovery, you need to take into account that 301 have already been discovered. And what we did in this paper was we went back historically and adjusted the threshold for declaring something significant. And when we got to the present day, we said that the usual tests, we usually look for Two Sigma or two standard deviations from zero for a 95% confidence.
Well, that is only that if there's a single factor test, once you have multiple tests, then that threshold increases, that has to increase, or you're going to have a large number of false discoveries. So, what we did is we need to jump that to three, so not two. And then I also made the claim that it was a major conference in finance and looking back, it's funny at the time it wasn't, but I made the claim at this conference that half of the published empirical research in finance was false. And that wasn't very popular thing to say in front of 200 finance economics. I tried to basically say, well, I had applies to my research too, by the way. And it was like an outrageous thing to say and people's reaction, well, it just can't be true. It can't be right. It can't be right.
And then a couple of years later, maybe Cam Harvey has got a point. And then a couple of years later, we already knew that he didn't say anything new. You know how it goes, you can't win here. But it turns out that our field is late to realizing the problems that arise out of the data mining, the stuff that can happen by luck. Other fields have already discovered this. Indeed, many years before I published my work, the medical field realized this, that over half of the research studies in medicine, I started with that between epidemiology were false. And another classic example is a genome association studies. And the problem there is worse than in finance. So think of you've got a 20, 000 genes. So you look for an association between a gene and some disease.
You're going to find some association just by chance and then to make things worse. What about a sequence of genes? So, think of a combination of two, and you've got like 20,000x19,999, the number of possibilities just explodes. So, they realized in that field at some point that it wasn't more than half. It was like 90% of the research that they had published was false. That people at top universities that had been given tenure or a chair where every single one of their publications was false. So, you think the Three-Sigma is bad in finance and that field is five. So again, this is an insight and I've spent a lot of time thinking about this, and it is in my opinion, much more serious to have half the discoveries in medicine to be false, because it has a matter of life and death. In finance, it's a matter of, is there Alpha or no Alpha? That can be painful for people, but nevertheless, it's not usually a situation of life or death, but it's strongly related to the incentives.
And this is one of the things that embarrassed that didn't figure this out earlier, especially given I'm a graduate at the University of Chicago and incentives very important in shape the research process and the publication process.
So, to expand on that if you could, the question I have is how much of this might be exacerbated by economic incentives from the investment industry?
Okay. So let me expand on this a little bit as to what I mean. So in the vast majority of schools, so probably 95% of the schools in the world, a single publication in The Journal of Finance, get you a job for life.
Wow.
It is remarkable. So, at a top school like Duke University or University of Chicago, they don't count the numbers. They look at the paper and really carefully judge, whether they believe it's going to have impact in the future before doling out a job for life. So, the other aspect is more subtle. So, the journals are ranked basically by something that's known as an impact factor. And the impact factor is the number of citations from other journals to the journal in question. So, if people are citing work that's published in the journalist finance a lot, then The Journal of Finance gets a very high impact factor and it's does have the highest impact factor in finance.
So, the other aspect is that it's well-known that research papers that have a positive result. And what I mean by that is you propose a hypothesis and it actually is supported by the data. So you propose a factor and you actually show that that factor explains the Cross-Section of Expected Returns. That's a positive result. A negative result is your proposed a factor doesn't work. Okay. So the positive results, they get cited. And the negative results don't get cited. So if you're a researcher thinking about, "Well, my only hope of getting published in a top journal is to find a result that's significant." Because if it's insignificant, then the chances of me being published are greatly diminished. So what happens? You put all this together and then essentially you've created the foundation for a massive data mining expedition, or people are out there data mining like crazy to try to find that result.
And I've called this p-hacking. And that just refers to getting the maximum possible of significance C-level. And you think about data mining, what is it? You're sweeping a vast array of different factors. You're trying different transformations of the factors. You might try different data samples. You might exclude certain periods like the global financial crisis. You might transform the variable by various methods. You might delete some outliers that are causing a lack of significance. You might try a different estimation method that has a better chance of getting significance. These are the tools of the p-hacker. And unfortunately, when the paper is submitted, often you don't know all this stuff that's gone on in the background. And for an editor is really hard to figure out what actually happened in the research process. So, these incentives basically cause a paper flow that is very select, that is biased in terms of the significance.
Indeed, if you look in economics and finance, more than 90% of the papers that are published have empirical work that supports the hypothesis. And you think about finance in general, it's really hard to find that trading strategy that beats the market to put it in a very simple term. So, it does make any sense that 90% are successful. Research is really hard. So this, again, we need to differentiate again, the 95% or 97% of schools, just count the papers versus the other schools that actually read the papers very carefully. If I went on a data mining expedition and got lucky at The Journal of Finance, my peers would read my paper and say, "Well, there's just a data mining expedition. It's not going to hold up." And if it is data mining, what happens is that in real-time after the publication, the effect goes away.
So, these incentives are really important. And I want to draw the distinction between academic research and practitioner research in asset management. So practitioners, they might actually do research for publication, but most of the research is not done for publication. It might be shown to a client, but it's not for public consumption. And if you think of the incentives, the incentives are a lot different. So, if you do this data mining expedition and develop a great backtest and present it to the client, the client puts money on it, then you're going to disappoint the client. So the return and live trading is going to be lower, or even be negative compared to the backtest. So, what does that mean? Well, that means two things. Number one, any performance fee doesn't exist. And number two, you lose reputation as your client redeems. So, the incentives in research in asset management are much more aligned to deliver what I call repeatable performance.
So, most practitioners, I'm not saying all practitioners, they know that if they pick the best backtest, that is exactly the wrong thing to do because the best backtest strategy is the one that's most likely overfed and most likely to disappoint. So, what you do instead is to choose the backtest that you think is the most robust, that will have the best chance of actually performing close in real-time compared to the backtest. And then when you show the clients above, we said that the sharp was 0.6, in the backtest we got 0.5, it's pretty close. And the client says, "Yeah, that is close. Congratulations. And we're going to up the allocation."
So, that's the incentive. And indeed the fundamental difference here is that in the practice of finance in asset management, there is no equivalent of academic tenure. So you don't perform, you're gone, you're redeemed. So, I do think that again, it's all incentives, you just need to take that into a can. And again, I'm talking on average, I've seen plenty of research by practitioners that is incredibly self-serving and incredibly naive in terms of the analysis, but I'm just saying overall, the incentives differ and that needs to be taken into account.
That was all absolutely fascinating. It kind of implies that, like you said, on average, you just based on the incentives. It doesn't even matter on average, based on the incentives as it management research could be more reliable than academic research. Is that fair to say, or is that overstepping?
Again, we need to be careful as to what we're talking about in research. So I'm talking about the academic publications, and a lot of what I'm talking about is not what a practitioner might publish in The Journal Portfolio Management or something like that. That it's the internal research that they do and present to a client saying, "This is what we've done. Will you allocate to that?" And the incentives are just aligned to do a good job there. Yeah. So I would say in terms of that research comparison, that is much more likely that the research coming from practitioners is repeatable. The academic research, most of it doesn't even take transactions costs into account. There's soon to be zero. So you could never get away with that in asset management research from a practical point of view. So, it's a convenient assumption that I know it's difficult to deal with, but the asset managers actually have a history of trading cost and they can work that into their calculations.
And it's really important to actually do that. So, a lot of the stuff that is published with no transactions cost, overstates the significance. So think of, for example, cross-sectional momentum strategies, the turnover is massive. So, even with equities that are cheap to trade, you could be talking over a thousand basis points a year, just in transactions cost, in implementing a naive cross-sectional momentum strategy, and what I mean by naive is what's actually published in the academic journal. Again, you don't want to disappoint your client, and reputation is really important and fleeting if you make a mistake.
We've talked about factors, this idea that you can pursue higher expected returns systematically based on research. And you just gave us a whole bunch of amazing nuance on that. What are some of the mistakes that investors make when they go to actually pursue these factor premiums that have been identified?
Well, the most basic mistake, and I know everybody listening knows what I'm going to say. And the basic mistake is buying high and selling low. So we know that investors in general are basically momentum investors that you see something that has gone up in value so you buy it. And then when it drops in value, you sell it. So it's the complete opposite of the strategy that I teach my students that you buy low and sell high. That is one basic mistake. And that's a systematic mistake that is made quite often. Another mistake is that people don't take the Factor Zoo into account. So they see various different strategies. They see something that looks good and they go for that. And that's a keen to choosing the best backtest. And it's related to the first issue of buying high, they just pick the one that does the best and that's the one that's most likely to be overfed.
Number three, often you just look at the numbers rather than actually thinking about the economics of the strategy. So, a lot of the silver fitting you can get around by having somebody explain the economic foundation of the actual strategy. There's a paper out there that looks at millions of random combinations of balance sheet items, and income statement items to form all of these factors. And some of them look amazing for Sigma significance. But if you actually look at some of these factors and try to tell a story, it's really hard to do. So for example, one of them in this paper is a version of earnings and then minus an adjusted earnings, which okay there could be some story there, but the key variable is the denominator where they divide by estimated rental payments four years out.
So you tell the story for that one. What about three years, what about maybe that works two or five years out. There's no economic foundation, whatsoever. Yet, it works in quotations and it works because it just lucky. So, that's another mistake that people make, that they don't actually consider the economics of the strategy or they're fooled by some explanation that's been cooked up after somebody discovers it. So the idea is you've got some economic intuition to start with. You sketch out the variables you want to look at, and then you look at them empirically, that's the right way to do science. The wrong way in finance at least is you just do the data mining, you find something, and then you concoct a story that's consistent with what you found. That is the sort of thing that leads to disappointing performance, how to sample.
So, these are some of the mistakes and the list is a long list. And did we started off with something on the list that I gave the example that people are looking at various strategies with different sharp ratios and picking the highest sharp ratio, not realizing that that high sharp ratio of strategy has got a crash probability in it. So something like cross-sectional momentum does really well on average, but oh, there's a crash or you're going to lose 60% very quickly. So, that it needs to be taken into account also. So, these are common mistakes. Indeed, I've got a paper on blunders that investors make in factor investing. A lot of these mistakes could be avoided. And indeed, especially if you're an institutional investor where you've got the luxury of actually quizzing the people that are presenting an idea to you, the retail investor has no chance to do this.
And that's why it's better for retail investors just to invest in a broadly diversified portfolio, but the institutional investor, your major pension fund or something like that, they can go and ask the questions. For example, you're presented with the strategy and I've actually been in this position where I was in a previous job, a scientific advisor to a major U.S. pension plan for alternative investments. So, I would go in to these hedge funds and to scientific due diligence. And they would present the research pretty thoroughly, but I would have some standard questions. And one thing I probably shouldn't even say this on the podcast, hopefully that they're not listening, sorry to be giving you a negative business here, but what I would do is basically say, "Oh, this is really interesting, be really positive." And then just like subtly asked the question, "Oh, well this is related to X. Did you look at X?"
And then they say, "Oh, yeah, we looked at X, and X didn't work." Okay. So X didn't work, but it isn't reported anywhere that X didn't work. What about Y, what about Z? What about hundreds of other things that could have been tried that they're not reporting? So, to me, when I hear an answer like that is like seeing the cockroach in the hallway knowing that there's a dozen behind the wall. Though it is a red flag. This is a company that, again, it could be that the incentives are that you get a big reward for getting the allocation and the performance is secondary. Maybe the person is incented in the wrong way, but this is just an example of the due diligence that institutional investor can do, but a retail investor, pretty well impossible.
So, is machine learning going to make this Factor Zoo better or worse?
I've got a paper that is related to this. And I actually did some early work and deep learning, published the paper to 20 years ago. And I came to the realization that for factor work, which is largely done on a monthly basis, there's not enough data. So, machine learning needs an enormous amount of data. So, one thing does very important in machine learning is to have enough data where you can do what's called cross-validation. So you hold out a large amount of data and then validate what you found. And this cross-validation is really important because the natural thing in machine learning is over fit, right? Basically, you're not guided by the economics necessarily. You're trying to find something that fits, but you need to have it fit, not just the data that you're looking at.
So machine learning demands very large amount of data. And we just don't have that, for example, in looking at various different factors. So, it might be that you could use machine learning to develop a trading strategy. And that might be not looking at monthly stock returns, but looking at data, I don't know. Satellite data estimating number of cars in Walmart parking lots, and that's a large scale, a project or a weather project or something like that. So, it's possible that you can use a giant amount of data to come up with a trading strategy. But I think to use these tools to focus on, again, balance sheet and income statement, expectations, monthly data, just not enough data there. And it's unfortunate that you get a lot of people that do a one or a one and a half year master's program in finance.
They get a machine learning course. And then all of a sudden they're experts in machine learning at some company. And really all they know are the few examples that they did in their course. And machine learning is a vast field. I've got a graphic in one of my presentations. I scroll through hundreds of different approaches to machine learning. And you need more than a course or two from a master's program to do this right. And you need more than one person. And also you need a room of GPUs. So not many firms can actually do this. So, I actually believe that machine learning is very important to the future of systematic trading strategies, but not a lot of firms can pull this off. You need to have scale. So, in my kind of vision of asset management in the future, there will be a shakeout that the smaller firms just cannot afford to acquire a machine learning team.
These PhDs in machine learning, they don't just have interest from financially oriented firms, but every company wants an expert, and is definitely the case for Google and Facebook and Uber. These graduates are in demand and a very large way. There are a lot of barriers to actually setting up an operation within your company to actually do this successfully. So, to be clear here, I believe that machine learning is very important for the future of finance, but you need to understand the various different techniques. You need to understand when to apply it and when not to apply it in terms of the particular situations. And the key thing is you need lots of data. If you don't have a lot of the data, then you're just wasting your time. You're just going to set yourself up for failure.
You made me think of one of the other papers that I wasn't going to ask about in this line of questioning, but we have to go there now. If there's Alpha to be had from machine learning, if a firm figures out how to do it right, will that Alpha be passed on to investors or are they going to absorb all of the benefits of their Alpha as their scale increases?
This is a deep question. Let's think about this premium just in general. So, how do you generate a premium? So, one way to generate it is from risk. So, let's say you're benchmarked in a naive way what the capital asset pricing model. So you take on some extra risk. I give an example of adding some option writing that looks like Alpha when it really isn't. So, you can get an extra return just for taking on extra risk. So, that's really important to separate out the risk aspect, but it's also possible that it might be much more subtle. So, the option I'm running was you're just taking extra skew risk, but it might be much more subtle in a factor, a strategy that there is a source of extra return there. There's also the possibility that there's mispricing. So, it's naive to think that markets are perfectly efficient.
So my chair is Eugene Fama and he certainly doesn't believe that markets are perfectly efficient. Maybe the U.S. market is the most efficient market in the world, but it just doesn't make any sense that prices are at the true prices all the time. So, the second aspect here is that you might discover a strategy that actually takes into account the perhaps small amount of inefficiency in the market. So, think about value strategy. So value strategy...
So is, so think about value strategy. So value strategy has been around for a long time since Graham and Dodd, right, in 1930s and you can actually tell a story that value is related to risk and indeed value measures directly map into expected returns minus growth in a very simple framework, like a Gordon growth model, but you can also think of value as an indicator of mispricing. So the stock is cheap because it's basically undervalued. So think of a strategy of cap weighting so with capitalization weighting, you're asking for trouble because even if there's a small amount of mispricing, maybe it's only 200 basis points on average, which is really small and we never observed the true. So we're just guessing. So let's say it's two, 300 basis points on average. So if you cap weight, then by construction, you will overweight overvalued stocks and you will underweight undervalued stocks.
So, you can think a value is kind of capitalizing upon that where you're not you using, let's say necessarily a capitalization weight to actually do this. So I think that that is really important also, and this extra return that you're getting can decrease through time as markets become more efficient, but it's pretty unlikely to go away, right? Because it's pretty hard to imagine that the equities are priced correctly, all the time, even in the future. So that's the sort of measure that's got some longevity to it. Whereas another factor you might discover has a pretty good backtest, but once people kind of realize that this is a mispricing, they actually pile into the factor and it effectively drives the factor price up to such an extent that whatever was there goes away.
So again, when you actually, if you're doing factor investing, you need to think through all of this, and what ideally you want is a portfolio that's got factor exposure where you're exposed to factors that you don't think are fleeting, that have actually got a long runway to them, and it's okay to have exposure to some factors that might be temporary but understand that they're temporary.
So maybe you've got five years a life and you need to be willing after the five years to ditch that factor. The other thing that you need to take into account is the luck aspect that you might have a factor that's got a very good economic foundation that hasn't done well and do you want to exclude it from your portfolio or dump it from your portfolio if you already have it? Well, again, go back to the economics, figure out, does it make sense? Is it possible in the future that, that reverses? And often the answer is yes and we see this all the time, investors making this mistake. It's kind of the same thing as the buy high, sell low reinvest in the factors that have done really well. But those are the factors that are likely the most overpriced and then you dump the factors that have done poorly, values a good example of doing poorly over 13 years, everybody reducing their value exposure exactly at the wrong time. Again, it's remarkable to me that these basic lessons are not learned.
You gave the example of everybody piling into a factor and it going away. The paper that I was thinking about was as opposed to factors, looking at active management, where you empirically found that scale as an active manager, accumulates more assets because maybe they've had good performance, their ability to continue delivering alpha in the future, incrementally decreases as their asset base increases. So I guess a similar thing could happen with a factor fund where it mops out a bunch of assets and then the factor goes away. But I found your research on traditional active management within this context be fascinating.
Yeah and so the paper you're referring to, I think is really interesting. I'm biased, of course, from my own research, but it's kind of makes sense that scale is important and certainly we're not the first to suggest decreasing returns to scale. The Birkin green obviously is the landmark paper. But to think about this is really important that there is a certain amount of alpha out there and as you take on more assets, it's just like spread thinner. So one thing I like to look for, especially in the hedge fund space, is you've got an asset manager with various different funds that they offer, a hedge fund has got various different things that they offer. How many of them are closed to new investment? So that I think just a bit, they got a good idea. They've got a discipline, they've got a certain amount of capacity to protect the investors that are there and then they close to new investment.
So I think that's really important. So the study that you referred to, we don't look at hedge funds, we look at mutual funds because we got all the holdings and stuff like that, and it, what we find is that the, what we call decreasing returns to scale, so meaning that more assets come in, then the returns actually decrease the alpha decreases, that is not the same for every situation. So you can actually mitigate this problem and you can mitigate it in a number of novel ways. So one way is this moving from a solo manager to team management. So if you've got a solo manager and all of these sponsor coming in, there's a limited number of ideas that that solar manager has. So if you actually bring in somebody else, then you've got more chance of actually avoiding this lower return as you grow in terms of assets, under management and then it's even more subtle that if you bring in a replica of yourself, then there's no real effect there.
What you want is a diversified set of ideas. So you bring in somebody as a co-manager, that's got a different background in terms of experience, then that also reduces this degradation in performance and then the last thing that we actually look at is we look at systematic investment. So a systematic investment processes driven by an algorithm, and it doesn't suffer from the problem of a limited number of ideas. So when we look at scale for systematic algorithm-based investment, we don't see this drop-off that you would see with discretionary investment. So all of that suggests that the investor needs to take that into account and it's not the same for everybody, that different sort of approaches can mitigate this decreasing returns to scale.
So given all that's going on in the world these days and talk of inflation, etcetera, there's a lot more chatter about gold. So do the typical arguments in favor of allocating gold to portfolio, hold up to scrutiny?
So I do have this paper with Claude Erb called the golden dilemma, and maybe you could link it on your website and teamed it is my most downloaded paper. Yeah, and I started that research with Claude, the first thing I did was a lit review because gold, I'd done work with Claude on commodities in general, but not on gold and I found that almost all the academic work was on the gold standard, not gold as an investment. So our paper looks at gold and it's got obviously a very long history and what we found is that over history, gold has held its value and we document that in a number of different ways and the most colorful way that we document it is that we went back 2000 years to Roman times and Romans kept great records and we found how much they paid their soldiers. So we looked at what a Roman Centurion got paid in gold, and the coins still exists, so you can actually figure out how much gold is in those coins and then we took those coins, the amount of gold in those coins and figured out, well, what's it worth today? And what it's worth is very close to what the U.S. Army captain would make. So what does that say? So that means over a couple of millennia, gold holds its value, and we have many other examples and analysis that suggests that gold basically, another way of thinking of that is that the real return on gold is zero. So it basically keeps up with inflation. The conclusion of our paper, which might seem a bit of a paradox, is that gold is an unreliable inflation hedge. So I've just told you that gold is a very reliable inflation hedge over the past 2,000 years.
How is it unreliable? And it's unreliable because my investment horizon is not 2,000 years. So we find that given the volatility of gold, that it can underperform inflation for long periods and outperform for long periods. So you need centuries of investment horizon to actually realize this inflation hedging ability. So gold is volatile. So gold is as volatile as the S&P 500. Inflation is not volatile. So if you look at inflation rates, let's say over five or 10 years, the volatility is trivial. Even taking the inflation in the early 1980s into account over longer periods, just not volatile at all and then, whereas gold, even over these longer periods, it is volatile. So it's unreliable as an inflation hedge.
I've got a more recent paper looking at various different assets, over eight periods of inflation surges in the US over 95 years and if you look at the paper carefully, what we do is we detail period by period. So we just don't present the average and actually gold looks okay. But then if you look in detail, you see why it's one of those surges in the early 1980s where gold in 1979, just skyrocketed in value. So it's a lesson, I think, for reading a paper like this. If somebody has just presented the average, well, the average might look good, but the average might be driven by single observation and that's exactly the case here. So gold is remarkably misunderstood. I remember I was at a reception in London and I was talking to somebody, but then I overheard somebody else bragging about their gold holdings and that it was such a great inflation hedge.
So I turned around and I said to this person, "I'm Cam Harvey, I heard you talking about inflation, hedging, and gold, have you read my paper, the golden dilemma"? And the person looked me in the eye and said, "Yes, I know that paper quite well". And I said, "Yvonne, I'm kind of curious as to how after reading my paper, you can claim that gold is a good inflation hedge". He said, "Well, professor Harvey, you didn't really hear the beginning part of the conversation and you don't know what I do. So I actually manage a family trust and our investment horizon is about 400 years and we've been around for 800". So it was perfect, I shook the person's hands and said, "Thank you this is exactly so they were looking at it correctly. So given they a 400-year horizon, fine, they should have some gold, of course, and this is important, it's also addressed in my paper, gold like any commodity is subject to technological change.
If you look at the real return on most commodities, that real return is negative over a long horizon and it's kind of intuitive, you think about agriculturals 100 years ago, we didn't have the technology that we've got today in terms of the weather data, the fertilizers, the pesticides, the genetically modified seeds, the harvesting ability to scale, all this stuff didn't exist, and the prices have gone down pretty dramatically. So the real return is negative. The same thing could happen to gold and indeed gold historically did have a technological change when the new world was discovered, the price of gold plummeted when the supply more than doubled. So the technological change you need to worry about is basically harvesting some of the near-earth asteroids that are just loaded with gold, and eventually, that's just going to happen. So even this person with the 400-year horizon for the family legacy trust in England, they need to be aware of technological change, that value of gold could be severely disrupted.
Wow, so is that kind of a real return of zero in the best-case scenario, but if you account for technological change, it could have a negative, real expected return?
That's exactly right. So if you look at other commodities, they've got a negative, real return, gold is at a zero, but it is subject to a caveat that is true for all research and finance that this is the backtest. This is what we've seen, and there's always the possibility of a structural change.
Wow. Are there any reasons then that somebody would allocate to gold or any many good reasons, I guess, is the question that somebody would allocate to gold in the portfolio?
Oh yeah, definitely. Gold is a real asset, a diversified portfolio, which you want, has got obviously equity, fixed income, but it's got more than that. So it's got real assets. So commodities are part of that. So you wouldn't just invest in gold, but gold as part of a diversified commodity investment, real estate, things like that are just natural to be in a portfolio and gold is part of that. So I'm not negative on including gold in your portfolio. What I am negative on is people allocating a large amount, some story of some pension allocating 5% to gold and actually borrowing to do it, some Ohio pension plan, thinking and the motivation was to protect against unexpected inflation. Yeah, no, exactly, and the horizons not 400 years. Yeah, I wish I was at the investment board when that was pitched. I would've had something to say about it.
So I don't have a problem with gold. We do have this framework for thinking about valuation of gold. So if you really believe the real return is zero, you're going to actually calculate what the real price of gold is and the real present gold is much higher than the historical average. So, think of it this way, if we were at that zero rate of return, the gold price would be around a thousand dollars. So it is expensive, but gold is not the only thing that's expensive today. It's a real dilemma for investors.
You mentioned technology, and clearly, from this conversation, we have an incredible sense of your breadth of research research that you've done over the years and what I think we can call traditional finance, but you've also have incredible interest in defy and crypto, including your recently released book. What got you interested in this? I'm so curious.
So it happened after my stint as editor of the journal of finance. So I was basically not teaching for seven years. When you edit a journal like that, it's a full-time job because it's like 1500 papers a year. That's just, yeah, that's more than a full-time job. These are not easy reading. These are difficult to do and all the stuff that I talked about, trying to detective if there's data mining, is really challenging. So I went back to my advanced asset management course and decided to redo it. Hadn't taught it in seven years and I wanted a fresh start. I didn't want to be one of these professors that go in with slides that are like 10 years old. Everybody's seen that experienced that, and I was never impressed, so fresh syllabus. So I have a module on foreign exchange, so I'm thinking, well, I do the dollar, Euro, yen, dot, dot, dot.
What about this new thing, Bitcoin? So I decided I would add that as something that was kind of leading-edge, that thought it was just another currency. So I read the paper by Satoshi Nakamoto, and this paper is not published in a top journal, it's just on the internet, and I'm reading this paper thinking, oh, this is a big paper. This is a big paper. This is in the category of the Markowitz paper, the sharp paper, this is a big paper, and then I started trying to understand the mechanics in much greater detail. I don't have a degree in computer science. I've actually published in a top computer science journal, but not in this particular area. So this was a lot of work for me in terms of preparing a lecture on cryptocurrency.
My course has 12 two-hour lectures, and one of them was going to be on cryptocurrency and I found that this single lecture was taking me more time than the sum of the other 11, and I didn't mind it. So it was the more I read, the more interesting the material got, and the more interested I was. So I reached out for help and it was hard to find help because it was so new, this area and our relied upon some local resources to help me. But I was very, very nervous, indeed. In asset management, given I've published a number of papers in top journals that I'm pretty comfortable with almost any question that my students would ask and I can answer most of them quickly, and if I can't, then I say, "I not really sure, but I can find out" and I can rely upon my network of experts in the actual field of asset management to help me out.
For this lecture in crypto, I didn't have that network and I thought with substantial probability, there could be some students in the class that know much more than I do. So it was probably, I was more nervous over this single lecture than any other lecture in my career. So I go into the classroom and there were actually extra seats at the back that were filled because people had heard that I was doing this, even though they weren't in the course. I do the lecture and we've all been through this before that and, this is typical at Duke, that as soon as it comes to the hour or the lecture is over, everyone just gets up and leaves. Right? And if you try to tell people you're going to be late or need two more minutes, then they're just not receptive to that, whatsoever. So I really try to finish on the hour and immediately people get up and leave.
So I finished the lecture and nobody gets up. They're just sitting there and I'm thinking, oh, I made a mistake that maybe I finished five minutes earlier, or 10 minutes early, or an hour, I was just so nervous. I'm looking around, no, it is time and people just sitting there and I'm thinking, oh, okay, I must have really bombed. That this must've been a disaster of a massive scale because people just sitting there and kind of staring and then people came to the front and said, this was a transformational experience for them, that this was a lecture they will remember the rest of their lives. Some said it was the best one they ever had in, not just their master's program, but undergrad also and then they said, "This should be more than a lecture. It should be a course" and that's how I started in this space. It's a difficult space to deal in because it's just transforming so rapidly. So that lecture of seven years ago, as almost nothing in it that I do today. Indeed, the course I taught this year in 2021 was 85% different than the course that I taught in 2020. So you need to be really interested in this space to do that. The course this year was based upon kind of the framework of my book, define the future of finance and we look at infrastructure, but we look at some of the primitives. We do a deep dive on the leading companies in the space and then we look at all the risks that this technology faces. But it was a journey to get here and I see something different. I see something really transformational to finance. The problems that are solved are remarkable. You think, actually in my course, I show a photo of one of the first Western union wire transfers from 1873. It's for $300.
And then the fee to do it is $9, 3%. So 150 years later, it's 3%. It's, so little has changed over the past hundred or 150 years. You've got digitization, but the same level of inefficiency exists. I transferred some money into euros to Europe the other week and my bank said, "Oh, you're such a good customer, we're going to waive the fee" I said, "Fine" and then they quoted me a rate. I said, "Okay, well that's 2.5% off the market, so it's like a credit card" and the person doing the mechanics had no idea what I was talking about, so, no this is the break. So inefficiency is obviously a big deal, but decentralized finance is about trading with your peers. So there's no middle person you're trading with an algorithm. So an algorithm kind of matches you with your peers.
So you send money, for example, you send by the algorithm to your peer. There's no spread, there's no middle person. So if you're doing exchange, you're buying one thing for another, you send some token to the algorithm, and the algorithm sends you the other token that you want to buy. Everything is wide open. You can see the code, you can see the liquidity, and again, the cost is really low, and importantly it is available 24/7. The algorithm doesn't care if you're buying or selling and it's open to anybody. So it is really about a financial democracy. So everybody is just a peer. It's not like the banker and the retail client. So there's no labels like that. It's about peers and the issue of opacity in our current system that you deal with the financial institution. We don't know their health. Maybe the regulators got an idea.
You just kind of delegate to the regulator and the regulator has got a dubious history of actually carefully monitoring in the U.S., at least, these institutions. So it's completely open. So again, these algorithms are open source, if you can do better. So the speed of change is remarkable because the algorithm is available. You might have an idea to improve upon it. You just take the code that already exists and then you bolt on your idea, and you're in business. You'd be in business in one day. So you get rid of this thick layer of middle people. You basically make this much more inclusive and I believe it's important, this inclusion is really important.
I mentioned in my book that 1.7 billion people in the world are unbanked and probably more are underbanked and the way that I described the under banking is the following, you're a small entrepreneur, you've got a great idea. It's got a 24% projected return on investment. You go to your bank, so your bank and you want a loan and the bank says, "well, you're too small. We don't want to deal with it. But what we'll do is increase the amount you can borrow on your credit card" and it's remarkable how much entrepreneurial financing happens on a credit card.
So, you know the story, right? So you have 24% projected rate of return, 24% interest rate on your credit card and the project is never pursued and these are exactly the types of projects that need to be pursued. The U.S. is stuck in this 2% growth zone. Europe is worse, maybe 1%, Japan is at zero, and I think one major factor here is the financial frictions, that the projects that should be financed or not being financed. So the larger firms are financed and they actually don't have the growth opportunities that these smaller firms have.
So you want to see 5% or 6% or 7% growth, then you need to change the financial system, need to make it more democratic that you get the financing if you've got a good idea and you do it directly from your peers and basically cutting out this middle layer. So it means that Boren rates go down and savings rates go up and it kind of makes sense because you're not paying for the brick and mortar, the middle layer, and all this stuff. So the book is about not a renovation of the financial system, but it's about a rebuild from the bottom up and we are so early into this, it's less than 1% and this is exciting to me and it's kind of consistent with my teaching philosophy, that I want to give my students a glimpse of the future. It might not be accurate, but it gets them to think about the future so they can make better decisions. I tell my students that I want them to be disruptors, not disruptees.
How much should we worry about sketchy activity or fraud with respect to crypto?
Let me answer this in two ways. Number one, if you wanted to criminal activity, then Bitcoin or aetherium is probably your last stop and the reason is that every single transaction is posted to a ledger that's available to anybody to see and is there forever, it's immutable. So while you can set up an address that appears to be anonymous, to receive some ransom from a ransomware attack, you've got to move it at some point and when you move it to basically cash out, you can be caught and when you're caught, the justice is swift. Exhibit A is here's your address in this ledger that's immutable, guilty.
Okay. So you want to do something anonymous, use what the most anonymous technology cash, and most criminal activity is done with cash. So on the criminal side, this is a bad technology. There are some anonymous coins, but most of the press kind of mentions the ransomware, and Bitcoin, and things like that. The second aspect is people basically just taking advantage of this technology and people wanting to get in. So given this as a young technology, there will be multiple situations where people try to take advantage of potential investors, and as many schemes that are possible, many have been documented. I get asked all the time to be advisors, to some sort of token and all ask questions and almost always decline.
And, I'll ask questions and almost always decline. There was one that was interesting for me, in that they called me, wanted me to be academic advisor. Set up a meeting with the CEO and I went along with it. The meeting started out, the CEO wasn't there. They went through their pitch, didn't make any sense to me. The CEO shows up.
Basically, what they were doing was they were going to issue a token and you would buy it, and then you could use it later to pay for digital advertising. I'm thinking, "Well, why would you do that? That's like pre-paying. Who knows if the token's even going to be around? Why don't I just pay when you have to pay? Why would you pre-pay?" I tried to figure out their business model. "Why are you doing this, why are you issuing this token?" Finally, they admitted that, "Well, because this is a way to raise money." It's not like, "Oh, this is a way to raise money to fund this great idea that we've got," this is a way to raise money.
Obviously, I just brushed them off but one of my students did a little research on the CEO. It turned out that he had been found guilty of domestic abuse on 100 different counts and he was awaiting sentencing. Now, it makes more sense to me that there's zero chance that if he went to a bank to get a loan that this venture would be funded. The only chance of getting it funded is to go into this new space and take advantage of all the buzz about all these tokens.
Again, you need to be careful here. This is, again, not without risk. Maybe it's obvious to you, that if something's without risk there's no upside to it. You might as well just invest treasury bills. There is risk here and this is a complicated space. The book, a lot of the motivation for the book, is to access the millions of people that are interested in finance or work in finance, don't really understand this space. So one of the motivations for my book is that there are millions of people that work in finance, or just interested in finance, and they might not realize what's happening in terms of the structural disruption. Maybe they know a little bit about crypto, they read about Bitcoin going up and down, they read about Elon Musk tweeting about Dogecoin. And, my book isn't about Bitcoin or Dogecoin, it's about a potential future. And this helps people, I think, with what they're doing.
I've got a big picture question that relates DeFi back to asset management. A lot of the people listening to our conversation right now are invested in maybe index funds, cap weighted, maybe they've got some over valued companies in there, but maybe they've got a bit of a factored tilt. But either way, they're diversified stock market investors. For those types of people listening, if there's a DeFi revolution coming that's going to disrupt traditional finance, do they need to be buying crypto or investing in DeFi? You can't invest in DeFi companies, it's all decentralized. So what do you do? Is the total stock market going to benefit from the DeFi revolution or do we have to be doing something special to take advantage of it?
Just last week, I had a conversation with a major pension fund. They were thinking of putting on some DeFi exposure. It's not just DeFi within this disruption, it's generally linked to blockchain technology so there's other applications of blockchain.
They said to me, "Well, we don't really have any exposure to DeFi right now." I said, "That is false." And they said, "Well, what do you mean? We are not invested in any of these tokens or companies." I said, "You've got exposure because the traditional companies that you're invested in had this negative exposure to DeFi, that DeFi can actually potentially put them out of business. You are totally exposed to this risk and you need to look at it that way. What is the risk of potentially disrupting many of the stocks in your portfolio right now, with this technology? And then, you need to realize that risk exists." And then, the other part of this is do I invest in some of these companies directly, to diversify my portfolio? So right now, given that they're not invested in DeFi, they're undiversified on the downside.
I've got a question that's maybe going to sound naïve. I said, as part of my previous question, how would you invest in DeFi if it's decentralized, but you were just talking about companies that exist that you could potentially invest in. How does that work? If total stock markets undiversified with respect to DeFi and somebody wants to add more, what is a company that you could invest in that's in the DeFi space?
How do invest in DeFi, that's the question. There's basically a number of different ways to do this.
Number one, you could, and often institutional investors do this, they invest in a venture capital fund that specializes in DeFi. So leading funds might be A16z crypto or Paradigm, funds like that. They actually do the work of actually doing the individual investments.
If you're at a scale where you're going to do this on your own, there's other possibilities. One is directly investing in the equity of a company that is developing product to be deployed in the DeFi space, so that's a possibility. The second possibility is to invest in a governance token that is linked to a particular protocol. I mentioned a system, for example, where there's an algorithm that you can send some token to and get another token back. There is a governance mechanism for that algorithm that fine tunes the parameters, so you can get exposure investing in a token like that and it gives you voting rights. There is another way of doing it, where you actually invest in the platform token. The platform token doesn't give you a voting right, but if the platform is successful, it will go up in value. That's a possibility, also.
There's also this idea of yield farming where ... Remember I mentioned that, given that there's no fixed cost, brick-and-mortar sort of stuff, that the savings rates can be higher. Another possibility is just to invest as an alternative to investing in bonds or certificates of deposit to get a rate of return that's much more reasonable, like four or five percent, compared to close to zero today.
There's many different ways to actually do this. The last way, when I said platform token, that could be Ether or Bitcoin. Some people just buy directly. There's various different ways to do this and it is a little difficult to get your head around. Because when you think of a regular equity, you buy that and get some rights to the cashflows, the residual cashflows, dividends, but you also get the voting right, whereas in decentralized finance that does separate it usually. It is different, but there's many different ways to get exposure.
Again, if you're an institutional investor, often you're using a VC to do this. But just the average retail investor, this is really easy to do. You set up a wallet and you can have a small portfolio where you experiment in investing in some of the coins in some of these DeFi protocols. You might have $500 and maybe you put that to work. In doing that, you learn about this space.
In my course, I put up a word cloud that's got about 80 different words on it. The students have no idea how these words apply to this space. Some of the words are understandable, like minting, we know what minting means. But, what does that mean in DeFi?
Right.
Or, slashing. Slashing is the opposite of minting so you're actually decreasing a supply.
It is, I think, a good exercise to actually ... You can read about the stuff, you can read my book, but it's much different to actually play in the space. In my course, I do the lectures but my students also run portfolios where they actually do it and invest in these protocols in a small way, just to get the experience with that. It's way different. Once you actually go through those steps, you really understand it.
This is a subtle point, also. It is a complicated space and if you invest some time, it means that you've got a huge advantage over other investors. It's a really good investment of your time.
And then, the second point is equally is important. I mentioned that we're 1% in so this is exactly the time you want to get in. You don't want to get in when we're 98% of the way.
Could all this technology not benefit public companies in a broadly diversified portfolio?
Yes. Many companies, basically, this will be really important for them because ... Well, if you think just in general, you think of the retailer that accepts the credit card, they lose the 3%. But even a company like Amazon, I had my students do the math, just suppose half of their revenue comes from credit card, and probably more does. And then, just do the calculation. 3% times 50% of Amazon's revenue, and then take the present value of that, that's just one year. Can this technology benefit these firms? Oh, yeah. Just in the most basic ways.
Even the banks realize that their days are numbered and they need to embrace some of this stuff. JP Morgan has got a stable coin, and they realize that they need to reduce the cost for their customers or their customers are going to leave. Think about what's happening in the banking sphere, you've got fintech knocking on their door, neo banks and things like that, that are just much easier to use. Better user experience, all around. They're being attacked by DeFi, they're being attacked by the fintech and they're trying to embrace some of this, they're trying to reduce costs.
But, a lot of the fintech is basically using the same structure. Think of Apple Pay, it's great. I use it all the time. It's way more secure than a credit card. But, the credit card company is more than willing to give .5% to Apple and they can take 2.5. It's a great trade off because it's much more secure to use the Apple Pay. But, it's using the same structure, it's the same 3%. Some of the other fintech is just using the basic infrastructure.
You need to be careful, again, thinking about the future. One of my guest speakers, a very prominent person in the DeFi space, described the current wave of fintech "putting lipstick on a pig." Effectively, what he was saying is yeah, it's great, it reduces costs but it's fleeting. The current wave of fintech will be replaced by decentralized finance. I think it's a credible story. The banks are very powerful, they will resist, but I think they know the writing's on the wall.
Various different countries, the US banking system is very concentrated, other countries much more concentrated, like Canada. Again, if you think about it, the cost is really clear, you're giving up growth. The growth opportunities that I described, that these small projects not being financed, or the borrowing rate, if it is too high, then a lot of stuff just isn't pursued. So what we need to do is reduce those frictions and to do it in a way that's efficient. And indeed, this is all peer-to-peer.
I remember when my grandfather passed away, going through his will, we discovered that he actually held a mortgage. He actually funded a mortgage, so he held the note. That was a peer-to-peer transaction that somebody basically, he lent money to them for a house. That was pretty undiversified, that was a major part of his portfolio. But, think of the possibility of something like that, but there could be thousands of people that are participating in that. Again, you're diversified across mortgages all around the world potentially and it's all peer-to-peer.
It's, to me, interesting that market exchange started out with a barter method thousands of years ago. That was the first peer-to-peer. There's a possibility that we actually move to a more efficient barter in the future. My vision is that there are billions of different tokens. Your wallet has got whatever you want to hold. I might have some US dollar token, I might have some gold, I move have some token that's based upon equities, like a token that is linked to the price of IBM stock. And then, I go to pay for something and I'm at my grocery store, I want to pay in gold but they don't want gold but they'll take something else, no problem. Totally seamless that I basically go to an algorithm and trade out of the gold into what the grocery store actually wants. And indeed, this is an existential threat to central banks, also.
Their reason to exist becomes marginalized. We see this already. A country like Venezuela, where it's in hyper inflation, just a gross mismanagement of their fiscal and monetary policies. If you're rich in Venezuela, then it's no big deal. You've got a bank account in Miami, you're protected, you're in US dollars. Who cares if there's 700% inflation in Venezuela? But, the vast majority of the population really gets hammered by this inflation.
What's happening is the vast majority can't afford to have a bank account in Miami, but they have a smart phone. That smart phone becomes their bank because they hold US dollar token, like USDC. Effectively, they have dis-intermediated the banks, number one, and the central bank, because who wants to even use the local currency when you can just use these tokens.
Again, the central banks will try to issue their central bank digital currencies, they'll do that. I think people will resist because it's now clear you want the government to see every single transaction that you do, but I really believe that they're too late to the game. The horse has left the barn.
You mentioned USDC, so that's a stable coin linked to the value of the US dollar. Does the importance of a currency like that, like the USD, does that persist if DeFi takes over?
I guess what I'm saying is it doesn't have to. My hypothetical wallet had USDC in it but it also had gold.
Oh, okay.
It also had stocks. It doesn't have to have USDC.
I understand.
It's just one of the possibilities. Look, the US dollar will stick around for a long time, as other central currencies will. I just believe that the money supply is out of control already. They have no control over the real supply of money because money is not just the US dollar. It's a much broader concept now. If I can pay for things with my IBM shares or my gold, then it's just the different world.
Okay, I've got another big picture question. In 1990, if someone could have told you, and I think there were analysts that were telling people that eCommerce and the internet was going to be the future, and there was going to be a big revolution and they were right, 30 years later. But ex ante, if you were there at the time in 1990, the chances of getting it right in terms of investing in the right companies were pretty low. I mean, talk about skewness. Is it going to be similar with DeFi or with crypto?
Yes, except it happens a lot faster. You can be successful and then immediately unsuccessful. We call it forking, in DeFi. The pace is much more rapid. It's important to have a diversified portfolio so you just can bet on one name.
DeFi actually, you mentioned the internet, and when the world wide web was created in the late 1980s, it was assumed that there would be digital currency. And actually, Mosaic and the browsers afterward, actually incorporated features to allow for that but it never happened. There was a huge amount of research in the 1980s about digital currency and it went nowhere. It went nowhere for a very simple reason. That, just like a movie, or a book or an image, you can make a perfect copy so it went nowhere, until the Satoshi Nakamoto paper in 2008.
Basically, the way I view DeFi is that it completes the internet so we can have an internet of value. What I mean by that, if you're paying for something on the internet, you have to load in your credit card and it's really unlikely you can pay for something really small, like five cents, with your credit card. It's really clunky. Even worse, to put your bank account in. It's difficult to be paid, also. This is going to allow value so it'd be very easy to sell stuff or to buy stuff on the internet. The internet changes so that it's really easy. If you're a blogger, or you have a podcast, you can actually have a small fee for it and maybe the one that's live is more expensive than the one that was done a few weeks ago. You can monetize your content.
This fundamentally disrupts companies like Google and Facebook. Google and Facebook, basically they take your information and sell it. I tell my students that Google probably makes $10,000 a year off each of them. Google will claim, "Well, we make $100 a year per user," but my students are exactly the market that advertisers want to pay for. So think of a different world where you're actually paid directly. Somebody wants to get to you, then you're paid in a seamless way. That somebody wants to get to send an email to get to the top of your inbox, there's a price for that and you harvest that directly.
Indeed, and this is a speculation on my part, I think part of the reason that Facebook is so interested in developing its own crypto is to solve this problem where they can actually pay their users directly for the content that they produce in a very simple way. Amazon, it's simply to save the money from the credit card swipe, but Facebook understands that this model is not sustainable and the same thing for Google.
If Facebook's users are able to monetize their own content and Facebook's being dis-intermediated and then they're no longer the central hub, where does that leave Facebook?
Well, that's obvious. Again, this trend of decentralization is a broader thing than just finance. There are many ideas, in terms of decentralized internet, decentralized social media. This is just the tip of the iceberg. Decentralized finance is just the low hanging fruit.
Okay, I've got one more question before our final, final question. We just talked about a lot of the benefits and potential for decentralized finance and cryptocurrencies. What are the risks? How could this go wrong?
It's a huge amount of risk. The last chapter of my book goes through various different risk factors and my view on whether they'll be mitigated.
For example, these algorithms that I've described are called smart contracts. I've also mentioned totally open source, so anybody can see it. Think of a hacker, so a hacker trying to get into Target. That's hard to do. So you need to break through, and then you need to figure out millions of lines of code to get what you want. Well, this is just wide open and it's a new attack factor. That's a risk, that these algorithms could be flawed. There could be a logic error, there could an economic exploit, that's a type of risk.
A major risk is scaling. That right now, for example, Ethereum maybe can do 15 to 20 transactions per second, where Visa can do 75,000. There's a lot of stuff going in, in the space, to bring the transactions per second to something equivalent to Visa, and then it must go higher. In my world, you potentially can do billions of transactions per second, in the future. That will likely be mitigated.
Let me talk about environmental risk. Bitcoin uses as much energy as the country of Argentina and much of it from coal, so you can argue that's environmentally reckless. Ethereum, which is the backbone of DeFi, they're switching next year to a different algorithm. They use the same one as Bitcoin right now, that's very energy intensive, but they're switching so I'm not too worried about the environmental risk.
There's issues in terms of custody, that the user experience is not as friendly as it should be. Many institutional investors have avoided the space because they don't know what to do with their private keys. But now, we've got custodians like Fidelity or Coinbase Pro, so that'll be mitigated.
And of course, there's the regulatory issue. The regulators, I think they realized that the sort of story that I told, they embrace. That if we reduce financial frictions, that's a good thing and that's good for economic growth. They also realize, if they're too harsh on the regulation, then the innovation either doesn't occur or goes offshore. It's a balancing act because we also want to protect the users from being exploited. The regulators need to do this really difficult balancing. And what is important is that this technology, given that it's deployed to a blockchain, which means that the same algorithm is running on tens of thousands of computers, all over the world, that even if you block out the US, the algorithm still runs. Regulatory risk definitely exists but it is a new world so it is how do you sue an algorithm? It's just there.
Right.
It could be there forever. There's no CEO, there's no board of directors, there's no head office. It's just an algorithm that people are using. Yeah, maybe you can block it in one country, but then people get their VPN going. It's really difficult to control.
Again, we need to be careful. I think that some level of regulation is a good thing, in terms of the exploitation of uninformed users, but too much regulation will kill this innovation. It's the last thing that we want because it is a path to economic growth and I think that that needs to be paramount.
You think about our situation, in the US. That we've racked up so much debt, and that debt has to be paid off and there's three ways to pay it off. Number one is to raise taxes and that's toxic for growth, so you're shooting yourself in the foot doing that. Number two is to inflate so just print the money to pay off all the debt. It's pretty simple but a disaster, in terms of inflation, and inflation is just like a tax and it's bad for economic growth. The third way is just to grow, so the more you grow, the more tax revenue comes in and you pay down the debt. Right now, we're not growing because of these financial frictions and other factors. But, if we can reduce the frictions we have a path.
So our final question, Cam, to cap off a pretty incredible conversation. How do you define success in your life?
That's really easy for me and it's really what I've been talking about. The reason I do my job as an academic, to impact the world in a positive way. Whether it's a better approach to asset allocation and portfolio management, or whether I can play a small role in the transformation of finance in general to bring real financial democracy to the world, that's what energizes me and that's why I'm in this job. It's more about the impact. Impact through my teaching, impact through my writing and I impact through talking to you.
Well, this has been a truly amazing conversation. I mentioned earlier, I finished your book this weekend, it blew me out of the water and I have to go back and learn a lot. But, this has been a great time together. Thank you.
Book From Today’s Episode:
DeFi and the Future of Finance — https://amzn.to/3EL5x63
Links From Today’s Episode:
Rational Reminder on iTunes — https://itunes.apple.com/ca/podcast/the-rational-reminder-podcast/id1426530582.
Rational Reminder Website — https://rationalreminder.ca/
Shop Merch — https://shop.rationalreminder.ca/
Join the Community — https://community.rationalreminder.ca/
Follow us on Twitter — https://twitter.com/RationalRemind
Follow us on Instagram — @rationalreminder
Benjamin on Twitter — https://twitter.com/benjaminwfelix
Cameron on Twitter — https://twitter.com/CameronPassmore
'The Variation of Economic Risk Premiums' — https://www.jstor.org/stable/2937686
'Conditional Skewness in Asset Pricing Tests' — Conditional Skewness in Asset Pricing Tests on JSTOR
'… and the Cross-Section of Expected Returns' — https://faculty.fuqua.duke.edu/~charvey/Research/Published_Papers/P118_and_the_cross.PDF
'The Golden Dilemma' — https://www.nber.org/system/files/working_papers/w18706/w18706.pdf