Episode 264: Pim van Vliet: The Volatility Effect, Revisited

28:22Which factor premiums did you look at? So we looked at the common factor premiums for which it's for some people clear that they are positive and significant. However, for others, they're not. So we looked at the main factor premiums, which is value, momentum, low risk, and also size. But for size, we know that there's debate whether it's a factor or not, because out of sample and post bonds, the results have been mixed, not very strong. The consensus view is that size is a catalyst factor. So other factors work better for small caps, but standalone, it's not that strong. So it's a good way to test.

29:03So is there a size effect in this market? We did have market caps. The answer was, again, not significant. So small, but not significant. So very much in line with the 20th century. Then momentum, strong one. There was already some evidence. Other papers looked at momentum, but then equal weighted because there was no market cap data. Strong equal weighted, also strong value weighted. Again, transaction costs matter also in the 19th century. good to mention is transaction costs were not that high as people think people think transaction costs were crazy back then and it's not our ancestors we should be a bit proud of them they were pretty sophisticated and knew how to deal with slower base of information but still pretty efficient because transaction costs in the 19th century were not much higher than the early 20th century they only came down really mid 20th century and with the optimization they went down also momentum now you have trading frictions but cross it's a strong premium value we found difficulty of defining that because book to price there were no standard accounting practices and that's what i mentioned before about because with quality it's even more difficult we couldn't test quality for dividend for value we took dividends as a proxy so if to price which is a value factor it's not perfect but comes close and we found a value premium but weaker than the momentum premium and also weaker than the low vol premium.

30:28So low vol, again, was a strong premium. So the risk-reterm relation was flat in that period. Not much inversion, but flat. And then if you then take the risk-adjusted approach, as I mentioned in the beginning, you see the factor coming out very strong. And also, it's a low turnover strategy. And so it's also easy to implement. So that's the factors we tested. And these were the findings. And I can tell you, when we first ran the tests, I was really a bit nervous about the outcomes because what would I do if low risk came out to not be working then? Because that would be a falsification, basically.

31:05You could still rationalize anything you find, of course, but that was tricky and more important than if low vol is doing good in the past quarter, yes or no, because that's more of a short-term markets movement, whereas, yeah, six years of data, although it is 19th century and it's not the same as today, of course, but it is six years of data and it says something about asset pricing. It's pretty cool to see the consistency with value momentum, low vol being relatively strong and size being not significant. Can you talk a little bit about why you think it's important to run out of sample tests like this?

31:43Yeah, it's very important. Like I said, for some people, fact agreements are a feature of markets. They're convinced, whereas others are not convinced yet. For example, we know that value had troubles in the US last decade. And, you know, low-fall is also in this strong rally bull market, which, as I said, that's the nature of low-fall, of course. So some are not convinced. And then also in academic circles, there's the p-hacking debate and that all the factors come out. Everybody has an incentive to publish T-stats above two. Could it be that it's data mining? Even out-of-sample is not enough because if you do an out-of-sample, you could say, hey, the factor was there, but it's arbitraged away.

32:22It's not a proof that it didn't exist. Whereas with pre-sample evidence, so going back, it's untouched and it's not influenced by the researchers. So that's why it's important to take a look at that. And that's why it should be known to everybody, to the researchers, and it should hopefully land in a nice journal and still a working paper. But we have high hopes, but also expectations that it will land and that people are aware of this finding because it's very important in the VACing debate. So our stance there is that, yeah, these common factors, which are easy to compute with just return data, so no degrees of freedom, no complexity, they are features of market.

33:02They are therefore not the result of data mining. That's still a very, very, very small chance, but this chance is pushed down close to zero because of this new 60-year of evidence. What did you learn from your sample about the economic explanations for the factor premiums? That's a bit more difficult because you want to disentangle different explanations. So what we do know that in the 19th century, there was hardly any delegated portfolio management. So investors were owners and apparently that's then not done. But we do know that this is not the only factor driving the low risk effect because otherwise we know it's not there and still there's a low risk anomaly.

33:43So that's a negative, negative finding. But low risk is caused by other effects. One of them is the lottery, SKU, payoff chasing, risk-seeking investors. Now, these are everywhere. They've been in the past. They come and go. Usually, they lose money, but they enter the markets every time. So that's one explanation. The second one could be envy. So even if you don't have a benchmark, you still want to outperform your rich tycoon neighbor who is also back then you had all those guys smoking cigars near post offices and you see the pictures pretty actively trading and i'm pretty sure that they were not buying the most boring stuff that they were buying back then you had these russian railroad stocks and it was i think it's a human baked in that we like risk if it moves quickly we're attracted to it like flies to a light bulb.

34:38And we know how it ends. If you take too much risk, it ends in tears. And that's basically what this history lesson tells us. Basically, these laws of nature, laws of human behavior, were also back there. That also means that if suppose everybody's, the benchmark starts to fade away, that doesn't mean that the low risk effect will be gone. So that's an implication of this deep historical research. Super interesting. You've got a paper, and this one did land in a big journal in the Journal of Financial Economics that looked at global factor premiums from 1800 to 2016. Can you talk about the difference between a global factor premium and the cross-sectional premiums that we were just discussing?

35:18Good to make this distinction. So when we talk about factor investing, usually we mean US stocks, but you can apply factor investing in international equity markets. You're still selecting stocks. You couldn't do this in Europe and Japan, emerging markets. In this paper, global factor premiums, we did it across markets. So we didn't look at individual securities. We looked at markets indices being stock market indices, and then comparing, for example, the US stock markets with the UK stock market with a German stock market index, and then applying factors to that. So you can then test time series momentum and cross-sectional momentum.

35:58So what we did is we took the big factors which were documented in the top eight journals in the past five years before we started our study. That is then time series momentum, carry, low risk, seasonal, and value. And we tested them across markets. And by doing that, you can then also move outside equities. You can also look at international bond markets, commodities, and currencies. So that gives you a whole matrix of six factors for markets. So then you get 24 factors or alternative risk premium that are called. Again, risk premium, that's not my word because I would call them factor premiums, which is a bit more agnostic.

36:37So 24. And then, yeah, going back to 1800. So we basically out-histored ourselves because with broad markets, you can go even further. And there we found positive results. So factor premiums seem to be a feature, a market feature. Also interesting that individually, they are not extreme. and usually they are basically all below one. And that's also good to be aware that it's gross, so you have implementation costs. That means they're not very visible, so you need to know what you're doing. And then for some investors, this could mean that you can reap profits from these gross factor premiums, which are probably a feature of markets.

37:16This is such fascinating information. Between these two papers, how confident do you think investors should be that factors are actually a real thing? How confident? Yeah, it's good to always have some doubt. You can be wrong. Predicting markets is one of the most difficult things. So it's all putting the odds in your favor. I think you can be pretty confident that the factors which are pretty easy and not complex, which have a solid economic rationale, that they will probably be around for the next decade. So if you have your whole investment portfolio, so you have your saving, you have insurance, you start investing, then I would certainly allocate some money to these factors.

37:56It's the same with the equity. The biggest one is the equity premium. It's also a factor. We understand it. It has a sharp of 0.4, 0.3, and we allocate to it. I have some doubts on this factor, like I have with all other factors, because they can disappear for decades. And I hardly meet any investor who doubts the equity premium, which makes me very doubtful. Thank you. By the way. So S &P 500, it's not like mana from heaven that you will get your 10 % per year. I don't know. But putting that aside, so I'm investing in equities. No worries. But it's always good to have some doubts. And then if you go to all the factors, keep it simple.

38:33So if you look at low risk value momentum and then the quality, you can be pretty confident that if you take these four in some way, you get a positive weight, that they will give you a risk-adjusted outperformance in the next. decades. Again, it's not a 100 % certainty, but you're putting the odds in your favor. And it would be also ignoring this could be a rational decision where you say, hey, I've got my insurance, I've got my savings, and I just have some money I want to play with. That's it. I don't invest. I only secure my money and I do speculation. In that sense, it can be rational where you say, I ignore it.

39:09I think those people hardly exist. You always have this middle part where you say, hey, I want to put money for my 401k or my kids' college or some long-term goals, then I would definitely be confident that these factor premiums exist and then think of a way of getting positive exposure to them. And that can be done in multiple ways. I agree that some level of doubt is important, but your pre-sample findings in those two papers are pretty incredible. I mean, the cross-sectional finding and then also the global premium finding, how far you went back and the fact that they were so consistent with the time periods after your pre-sample is kind of amazing.

39:47So you can say I'm more confident on multiple factors. The alpha for the next decade, I think the sharp ratio of the alphas of the four factors I mentioned, I'm pretty sure that that will be higher than the sharp ratio of one factor, which is the equity market. I've got a question on sharp ratio because you mentioned it a few times. So you mentioned that low risk is a negatively skewed strategy. Skewness doesn't work well with the sharp ratio. What are your thoughts on using the Sharpe ratio to evaluate a strategy like low risk that has a negative skew? Sharpe ratio has its pros and cons. It's at least better than the information ratio, as I said.

40:24It's closer to utility. Yeah, better is like to do a Sortino or get closer to what we call risk, which is losing money and the depth of loss. However, for practical reasons, I'm not that worried about skewness, that it's inappropriate. So I did my PhD on downside risk. And the conclusion is to predict downside risk. Past volatility is pretty good. To describe risk, it's better to move to more what risk actually is, which is losing your money, chance of losing, and the amount of money you're losing. So the lower partial moments. Cool. Yeah, I saw your couple of papers on downside risk, and I didn't work them into these questions, but I'm glad it came up in some way.

41:05You mentioned your paper on when factors dropped their shorts earlier. So when we talk about factor premiums, we're typically talking about long short portfolios. You obviously did that paper. Why do you think it's important to examine the long and the short legs separately? Yeah, it's important because they are different empirically. If things would be symmetric factors, then it would be nice. but we found that you can come to wrong conclusions even on things like quality being spanning low volatility audio or whatever around, especially given that most money is managed in a long-only setting.

41:41That's what you observe. So that's why it's very important that academics bridge this gap between theory and practice. So we published this in the Financial Analyst Journal, which is the journal who has the same aim of bridging the gap between academics and practitioners. It's a pretty cool finding. When I saw the first results, I discussed them with David, my co-author. I was like, wow, jaw-dropping for quant. So sometimes I tell my wife, I found a high R-squared, and then she's like, what's wrong with you? Well, I was really happy with my high R-squared. So that also happened with this. I was like, wow, this is amazing.

42:21then we decide to do something with it because it's relevant for investors and also some of our clients who are long only investors and they need to know this does the long leg or the short leg tend to offer better risk adjusted returns on a standalone basis you see some single factors do better on the short like momentum it's really great on the short but then when you start to combined factors, then the shorts don't add much to each other. And then on the long diversification starts to pay off. And this is before implementation, because in the short, you have higher implementation costs and even also risks, reputation risk, shorting risks, short squeezing risk, all those kinds of things.

43:05So if you take all that combined, you could say also to all the listeners that if you're in factor investing, long only multiple factors are pretty good. Well, the first five, there is something to gain, of course, on the short side. It is not that it's much higher and much bigger out there for factors. So for other short strategies, so some people say, okay, so shorting doesn't matter for individual stocks. We say no, it could add value, shorting individual stocks, but not if you combine quality, momentum, and risk. So it's a bit nuanced. People sometimes don't like nuance, but for factor investing, it's pretty clear that on the long side, also taking implementation issues into account, you're doing better on the long side with diversifying factors than on the short side.

43:50That's pretty fascinating and a bit counterintuitive. On the long side, factors are less correlated with each other. On the short side, they're more correlated? Yes. That's what we find. Interesting. What are the practical implications of those findings? The practical implementation is that if you're not shorting them, there's less urge or that you're leaving alpha on the table. So that's one implication. Another one could be that you say, hey, let's do long factors, short indices, then you basically get pretty easy, absolute returns without the hassle of short squeezes, reputational risks, and implementation costs.

44:29So these are just to name two of the implications of our findings. What concerns would you have with the Fama French five-factor model? A couple of them. So we wrote a paper, we had five. But let's just talk about it, what pops up. One of them is that farmer friends don't include the low-vol factor in their five. So they have five, and low-vol is missing. In fact, they kept one factor in, which is the market factor. So they stick to factor being rewards for risk. That's the Chicago paradigm. Anything with a return should have risk. So they don't include low-vol, low-beta as a factor. So that's one problem.

45:07Another one is that they keep sizing. It's a good explanatory factor, but it's not giving a premium, as we discussed. Momentum is clearly not in. It's lacking. So the momentum factor is one of the biggest ones. So together with low-volvementum, I would say, with value are the big three in that order, low-volvementum value. Yeah, in their five factors, two of them are lacking, which is low-volvementum. momentum. So in that sense, and still you have five, which is adding complexity. So it's a bit complex and not fully descriptive. So that's some of the issues I have. The other one is that it's long short in which the way it's constructed, it works pretty well.

45:49So they do the median size and then they do value weight. The problem is that if you're a small company, but then within In the smaller stocks, you're a big one. You have a big weight on the factory returns. And if you just switch sides, you become a large cap stock, but very small. You basically have zero weight in the factor definitions. So that is a flip-flopping thing most people don't know. It's also, on a 60-year basis, not a big problem. However, if you zoom in, it is sort of problematic. This is pretty random. So this can be done in a better way. But that's a bit nitty-gritty maybe, but still could distort results.

46:29Most important is it's five and two of the top three are lacking. That's low vol and momentum. How would you put low vol into an asset pricing model? How would it fit into another version of the former French model? If you try to do that, then basically lots of things collapse. For example, if you start modeling from a relative utility framework, where you say the representative agent as you have in the cup m is not concerned about mini variants but about access return and tracking error that's fine but then equilibrium doesn't exist because the equity premium goes away and your model breaks down so it's not general equilibrium it's pretty tough if you say hey let's assume let's forget about the benchmark let's say people are gamblers then you also have a problem so i had one paper on that where we said people are gambling for gains so So that's not even fully risk-seeking, but partially risk-seeking, which means they really love a positive school and then start modeling.

47:25Then the problem is that the market is very inefficient because if you like a good school, you shouldn't diversify because you throw away your upside at a much faster rate than you reduce your losses and your downside. So somebody who likes a positive school doesn't diversify. And that's also what you observe, that people who like a good school, they don't buy a thousand stocks, they buy like three or five. That's rational. However, to reach to an equilibrium model and have a cup M or any skewness cup M or any relative utility cup M, you need to have the outcome that the market is efficient. And for those examples I mentioned, it's not.

48:04This is basically the whole problem of behavioral finance, which is really good explaining individual behavior, but it's very difficult to bring this to macro and market level. That's basically the struggle our whole finance research is struggling with maybe a good reason that they didn't try and put it in the model because it would have blown up finance all together yeah so you can put it in so from an apt point of view so that's agnostic from utility you can throw in a full factor you can also create because then you do a long short and then you throw it at the right side but that's more statistical and very light assumptions and cochrane calls this the factor fishing license.

48:45Anything goes, just throw in long shorts, then you're really far away from what we call economics and first principles. That's what we did as well. So in the paper conservative formula, we create a conservative minus speculative factor long shorts, and it blows away many factors in a very elegant and simple way. And then we say by sticking to a couple of rules, so you don't have the problems I mentioned with Fama French 5, you only stick to thousand large caps. It's pretty simple and clear. And then you can get access to all those multiple factor premiums and you can explain others as well. Blows away in terms of explanatory power.

49:22Yeah. Wow. Super interesting. All right. We're going to move on a little bit from factor investing to a paper that you did on inflation regimes in the Financial Analyst Journal. You did this, another really incredible historical data expedition, I would call it, for asset class and factor premiums through inflation regimes historically. How did you define inflation regimes in that paper? We defined it as buckets. So we took 12 months rolling and then we looked at CPI above four, two to four, also deflation. Sometimes we forget, but that can also be a future scenario. So very data-driven, just observing it and then putting it in those four buckets.

50:05And how do asset class premiums vary through different inflation regimes? Basically, what's really bad for investors is high inflation. If you go to 4%, about 4%, hardly anywhere too high. Stocks do offer some protection, but bonds are really bad. So any 60-40 investor who had a great three decades in the US will really suffer. That's a bad one if you then add stagflation to it. So that means bad economic conditions, it gets even worse. So that's a scenario investors should be fearful for. How do factor premiums vary through inflation regimes? They hold up pretty well. So we find factor premiums to be persuasive, so very robust, whether it's deflation or inflation, factor premiums are there.

50:52We do see some variation. It seems that momentum and the risk tend to be a bit better in high inflation, but we do strict statistical tests, so it's at the edge. What's good is that the size of those premiums is also very stable, which means that if equity returns are really low on bonds and equities, you do get this 3 % premium, for example, which is in a low return environment, 3 % is more than when equities go up by 20%. So relatively, factor premiums become more important and a source of capital growth. And what economic environment did you find to be the worst for investors? The worst would be high inflation with that economic growth.

51:33So we looked at several measures for that. For example, recession or other measures of economic activity. And it's bad, especially if you adjust for inflation. That's also finding. So nominally, it doesn't look that bad. But in real numbers, it's really bad. And that's giving some cushion because also the last year, 2022, we had inflation going towards 10, markets going down almost 20. But still, most investors felt minus 20 and not minus 30. So in a way, behavioral biases help us a bit, feel less of the pain in a bad scenario. But however, the truth is, your wealth evaporates, you're purchasing power of that in this stagflationary regime.

52:16And you should be aware of that could happen and also try to take measures to limit those losses. And the factor premiums are still consistent in those brutal periods? Yes, they're very consistent also in those periods. So this cannot be an explanation. So also from a rational point of view, you cannot say, okay, so this is the rational explanation why it's there because the premiums are found in all economic scenarios. That was a cool paper. Thanks. You got another recent paper that I also liked on gold. Can you talk about how an allocation to gold holds up as a downside hedge in the 1975 to 2022 sample?

52:52yeah maybe some explanation on the gold paper so it was sort of a idea that gold comes and you have to call haters the gold lovers it's the crypto of the ancients it's a store well blah blah so there's something with gold it's also enough for a colabury and i did lots of lowval papers i think more than a dozen we thought let's take a look at gold because gold also is seen as a safe haven like lowval so that's sort of the reason we looked at it also reactions to the paper some gold bug said bim don't you like gold gold haters said bim why are you promoting gold so i got the same paper completely different which means we found a middle way i think if that's a critique so we looked at the safe haven hypothesis of gold in this paper what are the downsides of getting downside protection from gold yeah so it offers downside protection not fully it offers protection with a longer horizon so that's the first thing we found so on a monthly basis, it doesn't offer much protection to start with.

53:50You only start seeing it when you move to one year or longer. And the downside is you give up return. So you get protection, but it costs you money. And I guess, since you looked at low vol on the paper, is the suggestion that low vol is a better downside risk protector than gold? Yeah. In this period, for me, it was a short period. So we only started in 1975, which is like, how many cycles are we talking about then? But still, and then for that period, we found evidence that low vol was indeed much more effective. So if you want to protect your capital, gold is a way to do it. But it's more of a second order effect.

54:26Because if you do it, do it carefully. Don't overdo like 5%. It's like putting salt on the food. Don't overdo it. So a bit of gold. But with low vol, you can go all in. You can say, okay, let's just take my equities out and put defensive in. If you really don't want to lose money on a one-year horizon, then going defensive equities, low-vol, is more effective. The good thing is they strengthen each other. So it's not either or. So it's a different source of protection. So in years when low-vol offers protection, it's not necessarily that then gold is also offering it. So they correlate positively.

55:04You mentioned the sample size. Do you think this sample is sufficient to inform expectations? And maybe to expand on that question, do you actually think a 5 % allocation to gold makes sense or only in this sample? In this sample, for sure. Going forward, it's very ambiguous. I do tend to like the story that when you go back to Rome 2000 years ago, you were in the army and you had your monthly wage. You had this golden coin. You could go to a shop and buy yourself a toga, you know, a nice suit. You throw it in the sand in the Colosseum. You put some sand on top of it. Fast forward 2000 years. You pick it up.

55:37It's still clean. It's not gone. You don't have a password which you lost with your crypto or your stock account. You go out to the shop, modern shop, and you buy yourself a modern suit because it's worth a couple of, it's like$400, this coin. So in that sense, it's an ultra inflation edge. However, as a quant, the observations are too short. It's difficult to say something about that statistically. So also whether it should be five, one, or 10, it's very difficult to say. Personally, I do have a little bit of golden coins I like to look at them and touch them and discuss them with my kids. So you collect an emotional dividend from your gold?

56:16A little bit, yeah. Our final question for you, Pim. You've obviously done some incredible research. You've also written a very accessible book on low-risk investing. How do you define success in your life? I think success is achieving your goals. For me, my goal is to have a long-term positive impact through family and finance. if I call it two Fs. So let's start with finance. I believe wealth can be enhancing your well-being. For example, it gives you more freedom. Well-being can also increase if you give to charity, something my wife and I like to do, and we enjoy it. One of my passions is applying academic investment research and then applying it in a simple way, not too complexity.

57:01And we have one insight close to me is the fact that conservative stocks beats speculative stocks. In the long run, And I like to read a lot and do a lot of research. And success for me is also to bring this wealth of knowledge found in the literature to a broad audience. For example, low-volve or defensive investing is now one of the common four factors in the industry, whereas about 15 years ago, it was clearly not. And as a fund manager, I'm also managing funds based on all those scientific insights and quantitative factors. financial success is to deliver a high risk adjusted return to our clients and then significantly statistically so that's really financial success i think the second one is the f of family that's most important being faithful to them i've got three boys they will probably outlive me yeah they're great kids success for me is to be a good father a good husband and a good brother and son besides this even broader i think it's important to be a good citizen and provide positive to society.

58:06The golden ethical rule, I like. It's simple. Treat others as you want to be treated. Simple, not easy, I can say. For me, Christianity is a secure base, treasure of hope. Wealth can be a blessing, but it's ultimately that you remain faithful to the people around you. If you can do that, I think those two things, you're very successful. What a great answer. This has been a great conversation, Pim. Thanks so much for coming on. It was my pleasure. Thanks. Thanks, Pim.