Episode 306 - Wei Dai: Fighting for Every Basis Point

24:07And then there's a multiplier, a kind of a fixed multiplier assigned to each one of those categories. And then stocks are sorted daily by those three dimensions. And then weights are assigned to the individual stocks. Super interesting. What about investment? You talked about the three dimensions that firms are sorted by. How does investment, the fourth dimension, factor into weights? Some work we did a few years ago related to basic firms with high investment tends to have low expected returns. We see that if you use the asset growth as a variable to measure investment, that's more of a catch-all variable that gives you good information about future investment.

24:43So what we see, though, is the effect is mostly in small caps, and also it's not a continuous effect. It's more like a cliff-like pattern that we see. So for example, if you sort small cap stocks from those with lowest investment to the highest investment, let's say decals, the first eight, nine decile, they're pretty similar, but it's more the last decile with the highest asset growth. That's where you see the big cliff-like pattern, basically a big decrease in the average returns. So because of that, we feel that lends itself pretty well to more of an exclusion approach. So rather than using that as part of a multiplier, going to basically exclude those stocks with high asset growth within our small caps or any portfolios that hold small caps, that's kind of additional exclusion that we did in addition to the small growth, low profitability.

25:30I used in the example earlier, so that's another exclusion we use. So I feel that's also kind of where the research meets the implementation. You have the research findings, but then you also need to think about how you actually implement in the actual world, depending on the finding and depending on the behavior of the premiums. So there's a three-way sort, there are multipliers, and then there's an exclusion for asset growth. Are there any other exclusions? For small growth, low profitability, that's another exclusion we do within small caps. And then there's also, so these are more exclusions from our eligible universe.

26:02We're not going to buy them in the first place. But then there are also other short-term signals we use that's more related to buying stocks daily. So maybe there are certain stocks we want to delay buying if the stock is on downward momentum, let's say, or maybe has very high sliding fee on special or latest addition being reversals if it's also have done well recently in terms of our reversal signals. So those are the ones we might want to refrain from buying more. And then we can actually emphasize buying others. So definitely a lot of details when you think about combining all of these different signals.

26:37Here's a question that's very popular in Canada and I'm sure many other countries. How should investors decide whether to currency hedge their foreign asset exposures? I'm happy if actually investors are asking that question. So at least that means that they are going beyond their home markets, they are enjoying a bigger opportunity set, and they are getting that global diversification. I think then the currency hedging decision will depend a lot on their asset allocation and also their investment goals. We did do a few papers on the impact of currency hedging on your overall portfolio. In terms of the impact on volatility, that very much depends on basically how volatile your assets is relative to currencies.

27:18So typically, you know, with equities, you know, equities already move around quite a bit. So at least on par or even more volatile than currencies, then hedging currency exposures will have very little impact on your equity investments or for any portfolios that's already very heavy in equities. And in contrast, if you think about fixed income, fixed income tends to be less volatile. Everyone knows that and much less so than currency movements. So then currency hedging can be a very effective tool to reduce the portfolio volatility for fixed income heavy portfolio. So that's on the volatility side.

27:53Now on the return side, it gets a little bit more interesting. I think you can actually decompose the return difference between the hedge portfolio versus on hedge portfolio. So the difference has two terms. One is basically your four currency premium. The second is the currency return. So the currency return parts, based on our analysis, it's not really predictable over the short term. So basically, you can treat it as a very noisy zero. And then the for currency premium, basically, that's what you can calculate from for rate and spot exchange rate today. So that actually can tell you then something about the expected return difference between hedged versus unhedged version of your assets.

28:34So that means you can actually use a dynamic hedging approach in the attempt to increase expected return. So essentially, you can say, well, let's just hedge the currency exposure in markets where the four currency premium is positive. So hedging, expected return of the hedging hedge portfolio is higher. And then you leave the currency exposure unhedged otherwise. And that's an approach that can increase our expected return. And also it does not require forecasting exchange rate movements. But in terms of volatility, it will give you more volatility for at least for fixed income investors. So it will make sense for fixed income investors, if you are willing to take on that additional return volatility for equities, the volatility impact will be minimal.

29:19Interesting. So if someone's currency hedging their foreign equities, does it make more sense, do you think, to have like a constant currency hedge or should it be dynamic using forward rates, like you said? It kind of depends. I feel there are a few different things at play here. Let me take a step back to say, you know, typically what we do, we would actually leave the equities unhedged and then we hedge on the fixed income side more as a default. This is mostly coming from that volatility impact perspective. And that's also, frankly, where we see a lot of client demands. And in some regions we will offer, you guys know very well, we offer hedged and unhedged versions of equities and a 50-50.

29:54Those are mostly also driven by client preferences, at least in aggregates. So those are more constant hedgings. But we do have selectively hedge equity strategies, also select a hedge fixed income strategies in some regions. I think those haven't been the most popular strategies or maybe not the default for people, partially because it does require more education, more understanding, right, how dynamic hedging actually works, you know, what really drives realized returns. And also, you probably need to consider if there's any impact on tracking error, right, relative to benchmark, which are typically unhedged.

30:28So, you know, we're definitely happy to provide those strategies. But then I think in the end, it also depends on what clients are asking us for. I want to move on to premium timing, which you've done a bunch of really good work on. Can you talk about why timing exposure to premiums is so tempting? Well, there is a lot of economic incentive for doing so, right? It's also not our first time to look at premium timing. So basically, if you can figure out precisely, right, when to go in and when to get out, that's phenomenal returns. Just before this, I took a quick look at just S &P 500 return, say from 1926 to end of last year.

31:02Your$1 will grow to$15 ,000. That's pretty good. That's 10 % annualized. But then let's assume you have perfect timing strategy, say not monthly, but let's say just calendar year. I know when S &P 500 will be up and that's when I'm in the market. And then otherwise, if the market is down, I go into T-bills. that$1 instead of growing to$15 ,000 will grow to 2 million. So that's 16 % annualized. That's a huge difference. So there's a lot of incentive in between that difference, which is why I think that attempt to time the market or to time the premiums that we pursue will be never ending. That's a forecast I'm making here, but you will always see some of that attempts.

31:44So what are some of the ways that investors do try to time the premiums? There are lots of different things I see. One of the most popular ones is using valuation ratios, because you probably often hear people talking about, oh, the U.S. market is being so expensive based on aggregate valuation. Is it now time to get out? Those type of statements you hear quite a lot. And also the same concept you can apply to any two sides of the premium. If you think about value versus growth, you can calculate, say, the price book ratio of growth stocks relative to that of value stocks. Basically, then the difference is called the valuation spread.

32:20You can say, well, the spread is high. Maybe that's the time we want to switch to value. When the spread is slow, it's time to switch back to growth. So that's definitely one popular signal I see. The other ones, I think people inevitably also look at past performance a lot. There are definitely people who are betting on mean reversion. So essentially, if a premium has done really well, maybe then the premium will tank in the future. But then equally, there's also people betting on momentum. So if the premium has done really well, maybe then the momentum will continue. It's the time to continue to be in that premium or get into the premium.

32:53So definitely seek both sides of that spectrum. I would also say that we do get a lot of questions from our clients in different kind of shapes and forms, but a lot of them are actually timing questions in disguise. People ask you about interest rate, Fed actions, inflation, business cycles. Ultimately, I feel everyone wants to know if there's anything we can say about the future direction of the premiums or the magnitude of the premiums and then what that means for their asset allocation. So definitely lots of different methodologies that we see. What kind of parameters need to be defined to implement one of those timing strategies?

33:28So a lot of the concepts I was just referring to can be quite straightforward, but actually translating them into an actual strategy will involve a lot of parameter choices to a point, Ben. Essentially, what you want to know is when to go in and when to get out. Let's use, say, valuation ratio as an example. People say, well, we should go in when it's cheap, get out when it's expensive. But what is cheap? What is expensive? People say, well, then maybe you can look at the current valuation relative to the historical distribution. That's reasonable. but then how do you then construct a historical distribution?

34:02Do you look at the most recent 10 years or 20 years? Or do you look at all the data available to construct that historical distribution? So basically an expanding window approach versus a rolling window approach. And then once you have that historical distribution, you still need to ask, you know, what is the threshold for being cheap versus expensive? Do you look at the 10th percentile or 20th percentile or maybe the median? And also, how often do you want to look at that signal and then to decide whether or not to rebalance. Is it monthly? Is it quarterly? Is it annual? So you can probably already get the sense that for every strategy, there's lots of parameters involved.

34:36And also the same strategy can be applied to different premiums, different regions, which is why in our recent study, like on timing, we looked at 720. So that's a combination of three signals. So valuation spread, mean reversion, momentum, and five different kind of thresholds for switching in and out. and two rebalancing frequency, annual versus monthly, and then also two historical distribution construction using rolling versus expending, four premiums, equity, size, value, profitability, and three regions, US, develop XUS, emerging markets. Hopefully that's all. If you multiply all of them, you should get 720.

35:15That's what we did in our analysis. Pardon me, on behalf of all of the listeners, which timing strategies worked in your research? Yes, that's actually typically what people are most interested in, which ones work. And I'm happy to report that 30 out of those 720 worked. By kind of worked, I meant that delivered reliable outperformance relative to just staying invested, staying consistent, focused on the premiums. One of the best strategies we saw was a timing strategy that was used to time the market premium in developed US market and based on the valuation ratio that we were just talking about.

35:56And that strategy was able to actually avoid major market downturns in 2001, 2008, 2012 as well. So that led to a excess return on top of that average market return, let's call it 10%. So you get the excess return of 5.5 percentage points. So that's quite impressive. But there are definitely a lot of caveats. What we also saw is that, yes, it's very easy to focus on the winner without asking, you know, how much of that is actually due to luck versus skill. I often like to use that coin flipping analogy. If I ask you, Ben or Cameron, if I have, say, 10 ,000 people flipping coins, how many of them will flip 10 heads in a row?

36:36Any guesses? I don't know. Far more than I'd guess, I'm sure. Yeah, our expectation, there will be 10. Should then we then be calling them, say, coin flipping masters and expect them to continue that performance? Probably not. So in terms of the best strategy I was referring to, it does rely a lot on that exact combination parameters, annual frequency, rolling 10-year window, like a 10th percentile to switch out, 50th percentile to switch back in. And then we did some sensitivity analysis. we saw that if you just change one single parameter of that strategy, for example, change one of the percentiles you use, or maybe change the rebalancing frequency to annual to monthly, that alone would already reduce that excess return by more than half and also make that excess return no longer reliable.

37:25And also the same strategy doesn't really work in other regions or doesn't really work if you apply to other premiums. So I think just a good example showing that, yes, shining results can happen. At the same time, they can be quite tenuous, especially if they are due to random chance. So you can find a strategy that works, but it's super sensitive to like the region you tested it in, the parameters used to define it, and even the time period. Yeah, absolutely. Very interesting. In our paper, we definitely have another example showing how some of the results can be very sensitive to time periods.

37:57I think it was related to, you know, timing the US size premium using momentum. So essentially betting the performance to continue. And a lot of that excess return is due to just two years. It was 1972 when size premium was down by a lot, 73 again down by a lot, like negative 20%. So of course, the momentum strategy worked by avoiding that second negative year, but that alone contributed half of that excess return. And if you look at maybe the second half of the sample period without those years I was describing, then none of the strategies would work. I guess the idea then being that it wouldn't be reasonable to expect that type of event to happen again in the future with that exact strategy.

38:38Yeah. If your results are driven by, say, some outliers, very small part of the sample, then I think you do need to ask yourself how confident you are to kind of put your money into a strategy that has worked well in historical data, but actually very sensitive to just a few outliers. I guess there are other trade-offs too, right? Like you'd introduce more turnover, probably more tracking error, all that kind of stuff. Actually, in our analysis, we didn't even consider tax considerations or trading costs. So on top of that, it will be even less appealing to go with any of these timing strategies.

39:12On valuation ratios, can you talk about why theoretically it makes sense they would be related to differences in expected returns? Why are they not useful in timing premiums? That's a great question. I've been thinking about that myself as well. Generally speaking, I do think there is a distinction between identifying changes in expected stock returns over time and differences in expected returns across stocks at any point in time. So essentially, that time series aspect has always been a lot harder than the cross-sectional aspect, possibly because there are a lot of changes from time to time, whereas more things can be controlled for when you're looking at the same cross-section at each point in time.

39:52But when it comes to valuation ratio, there is a theoretical link, price being the discounted future cash flows. But that also means that if you see a lower valuation ratio, that can result from higher discount rate, meaning higher expected return, but it can also be due to lower expected future cash flows or usually a combination of both. So empirically, I think, at least in the time series, we see it's not really possible to cleanly isolate that clash flow versus discount rate effects from the data. That's on the equity side. But actually, good to point out that in contrast on the fixed income side, the future cash flows are a lot more certain, right?

40:30If you're thinking about the promised coupons and part payment. So you can actually say something about the variation of expected returns over time, the variation of the term premiums and credit premiums over time. And that's what we do in our strategies. But that is more like on the equity side. I think because of the noise, you are better off behaving as if the expected premiums are constant. There is another broader point I want to maybe just quickly mention is that there's also a distinction between a statistical relation that you see from the data versus a profitable strategy you're able to run.

41:05Because oftentimes you see regressions, scatter plots showing you some in-sample relation using all the data. But when you run a strategy, you need real-time out-of-sample prediction that's based on only data available to you at that point. So I think that's a much harder hurdle. Sometimes you see valuation ratios, people say, things like CAPE ratio, right? Cyclically adjusted price to earnings ratio. They have been shown to be related to future returns, usually over the long term, right? Over the next five years, 10 years, or 20 years. Interesting thing is that even if you assume there's some long-term return predictability, but then you use it for short-term asset allocation decisions, you know, oftentimes we see people deciding on allocation on an annual basis.

41:47That misalignment in the horizons means that oftentimes you're still not better off. Basically, the precision that's required for you to have a successful timing strategy is nowhere near what we can observe from some of the popular valuation ratios. I found that pretty interesting from our analysis as well. So what are the main implications of this research for pursuing premiums? In short, it's time to stop timing, if that makes sense. So in the end, I think premiums, they are positive on average. Good to keep in mind that whenever you switch out of the market or switch out of these premiums, the odds are already stacked against you.

42:27In recent years or just over the longer history, we see that the premiums can also turn very quickly. So you really need to be there to capture them whenever they show up. But then I also understand, we said that very earlier, it's very tempting to do so. So definitely staying invested will require a lot of discipline. That's why we see a lot of value from financial advisors to help people stay disciplined. One thing I found useful to point out to people is that just keep in mind that as you're going through those ups and downs in terms of the volatility of the premiums, the volatility itself should actually give us some comfort for why the expected premiums are positive in the first place.

43:07If things just go up every day, you should be worried. And why are you even being compensated? So to some extent, that willingness to bear that uncertainty and that potential underperformance is rewarded with higher expected returns. So stop timing, but think about other ways to put the odds in your favor. Think about integrating multiple premiums. Think about being broadly diversified. These are much more effective ways to put the odds in your favor. Just on the idea of timing, I tweeted, I don't know, a week ago about how the death of the value premium had been declared in 2021. And I looked at the value premium from then until April 2024.

43:43And it's like, especially in Canada, very, very positive. Staying in your seat is definitely important. Even equity premium back in the last decade, probably 2009, people say, oh, is the equity dead? These went different ways. I think there's definitely something to be said about not being carried away by some of the headlines and then stay disciplined through those ups and downs. Can you talk about how important diversification is to capturing the return premiums? So we're talking about, you know, what are the ways to actually increase your odds, at least improve your odds of capturing the premiums?

44:16I would say diversification is a very important consideration. People already know that diversification is very useful when it comes to reducing aquatic risks associated with single stock or a country and sector and so on. But our analysis also shows that it can play a very important role in increasing your reliability of capturing those premiums. The reason is that if you just imagine, let's say, a year, you said value premium was strong in Canada. Let's say we have a year with a positive value premium, let's say 4%. But if you look at all the different value stocks, it's not the case that all value stocks have outperformed growth benchmark by exactly 4%.

44:53A subset has done really well. They are the ones driving the positive premiums. Some probably had done average and some poorly. The issue is that you cannot really predict which stocks will end up driving that premium. And that subset will also change over time. So that means if you have a more concentrated portfolio, meaning that you kind of just simply sample from the value universe, then you might miss out on the subset that actually drive the premiums. And in contrast, if you have more broadly diversified exposure, and that will help you to capture the premium whenever they occur. In our analysis, more concretely, not sure if you have seen that, basically we say, let's look at a fully diversified portfolio that's tilted to a size, value, profitability.

45:38And then we also look at a few sampled portfolios, meaning that they maintain the same level of tilt to the premium. So basically, you can think about their expected returns being similar, but then they have fewer stocks, maybe 500 or even 50 on average. And then what we show is that the more diversified portfolio will actually give you a lower deviation from the market, so lower tracking error for the same level of outperformance. And that leads to a higher probability of outperformance, which is why you see that in our portfolios, we want to be as diversified as possible, given the investment universe and then given our investment goals.

46:17I really liked that paper. I thought that the way that you set it up is really intuitive, but I will say that it's a paper that's been discussed at length in the rational reminder community. One of the nerdiest places at length, there's pages and pages and pages of discussion on it. And the main point of discussion is the fact that you did what you said, which is held expected returns for the different levels of concentration constant. Do you have thoughts on what that analysis would look like if you were increasing expected returns with concentration, like a more diversified portfolio with a lower expected return versus a more concentrated one, but with a higher expected return?

46:50Does that make concentration look better? I see what you're saying. Yeah. I need to check out the debate and I'm very curious, but it is a very interesting question because ultimately the results, right, in terms of the the probability of outperformance will be driven by two things. One is your expected outperformance, and the other is the uncertainty around that outperformance. So you can think about basically the distribution of your excess return. You have the mean of that distribution. You have the width of that distribution. To your point, yes, in our analysis, we control for the mean of the distribution.

47:22That's basically usually what we do when it comes to research to say, let's isolate the effect of diversification so that you can show the impact more cleanly. I understand oftentimes there are also portfolios with different level of tilts, or maybe you, even for us, we have four portfolios with different level of tilts, or maybe even exclude certain parts, they exclude mega caps, so four versus a vector type of portfolio. So then they will come with different expect return. Then there are two effects. You are moving the mean to the right, more positive, but you are probably also increasing the width of that excess return distribution because your tracking error will be bigger.

48:00I would say that if the strategies are well-designed, you can actually balance the trade-off between those two effects so that to actually achieve very comparable results. That's typically what you see in terms of, you know, just the average asset return numbers we see for these portfolios versus the tracking error we see. But I would say there's still a limit to it, how much you want to concentrate. You can still say, oh, I can find 20 stocks, the smallest, deeper value, highly profitable. Yes, maybe on paper, they will have the highest expected return, but it's still not a robust portfolio and the trading costs will be super high if you have such a tiny opportunity set.

48:39I would say you can achieve comparable results, but you need to design thoughtfully. I think that's probably where that discussion in the rational reminder community ended up. There's a trade-off between concentration and expected returns, and it's really hard to define where that optimal point is. So there's lots of fund managers pursuing factor premiums today. What do you think investors should look for when choosing a systematic investment manager? I definitely see there are more managers trying to do a systematic active approach and also more investors embracing this type of approach. So I think it's good to see the world is moving in our direction.

49:17But as we have been talking a lot today, not all systematic strategies are created equal. And when it comes to assessing different systematic managers or strategies, oftentimes people start with performance, which is understandable. But I think good to keep in mind that returns are noisy. So every manager will have maybe some strategies that have done well in a certain period. So you need to be able to differentiate luck versus skill, which is not an easy task. A few things I think can be helpful. One is when you look at track record, try to look at longer term track record over different market environments.

49:50and also look at not just one strategy, even if you only are interested in one strategy, because it's a systematic approach, you need to look at the full set of strategies. That's a reflection of the process and how you are able to capture the premiums. And also importantly, not just look at the strategies that you can buy today, that are still around today. You should also include those that are closed, you know, merged or liquidated, because that gives you a much better sense of the performance that's not biased by survivorship. And also shows, to some extent, the commitment of the manager to a particular investment approach.

50:26And you can also look at conditional performance. By that, I mean that maybe look at periods when the premiums the managers are pursuing are positive. Even in the challenging period, like last decade with value on average on performing, we've seen periods of outperformance. So then you want to see what the managers are able to maintain their exposure and actually capture your fair share of the premiums when the premiums show up. And ultimately, it will be great if people can actually look beyond just performance and really think about what drives a good systematic strategy. In my mind, it's really a combination of two things.

51:04One is, do you have rigorous research that's underpinning your strategies? And the second is, do you have that implementation expertise to translate those research insights into actual strategies? Now, on the research side, I would say, especially these days with the rise of big data, a lot of computing power, it's getting even easier to find patterns in the historical data we talked about in timing. So I feel there's actually not only you need the technical skills to be able to differentiate noise versus signal, you also need to have the right incentives for researchers and to have that discipline to not to make constant tweaks to your portfolios.

51:43So on that side, I would say, especially dimensional, we have a culture that's very heavily influenced by our academic heritage. For me, it's quite scary. You have your work judged by such a crowd, including Nobel laureates. No kidding. At the time, I would say it's very tremendously helpful. Lastly, on implementation, I would say we already talked a lot about that in terms of portfolio design. Even if you know the same premiums, how you design the portfolio matters, and also how you do the daily implementation matters. The process, the infrastructure, and also how you think about technology versus human oversight.

52:17That matters a lot in today's environment as well. And for us, there has always been a very relentless focus on implementation. probably because, you know, our very first strategy is a microcap strategy, right? So it has been part of our DNA if we are not careful. If we're not practical, then we would have been killed in the market. So we wouldn't have a business that focused on research and implementation has really served us well until today. Continuing on manager selection for systematic strategies, can you talk about the downsides of performance fees specifically for systematic managers? So that's actually a research paper we did with Professor Bob Merton, who is a resident scientist at Dimensional.

52:58That's a few years ago. Maybe just to give you a little bit of background why we're even looking at that. Because traditionally, if you think about long-only investments, the typical fee structure is just a fee expense ratio or fee as a percentage of your assets under management. So performance fee is more for alternative investments, like hedge fund you do here, 2 and 20, so that 20 % being charged on outperformance against some threshold. That said, we started to see performance fee structure kind of trickle into the traditional long-only space within separate manager accounts, which is why we decided to do a deeper dive and just to understand the incentives induced by performance fees and whether it's suitable, especially for systematic strategies like the ones we are running.

53:44And the reason I mentioned Bob was involved, the reason that Uncle Bob was involved, because actually the performance fee from the manager's perspective, the payoff very much looks very much like a call option. If you have outperformance above certain threshold, the manager get paid a portion of per profit. But then the managers don't usually share the downside. So there's the floor. So it's like constant and going up. It's a call option. That's why you can also price the value of performance fees using standard option pricing models. And if you remember the option pricing models, the pricing of the option doesn't really rely on the expected return of the underlying, but it relies on the volatility and other inputs.

54:26So by analogy, the valuation of the management fee from the manager's perspective, it doesn't rely on the strategies expect return or alpha. It actually relies on the volatility or tracking. So if you're thinking about absolute return, it's volatility. If it's excess return, it's tracking. So then you can see there is some incentive to increase your deviation from the benchmark in order to make the payoff more valuable. But as we have talked about earlier in many examples, not all deviation from the market is justified by higher expected return. If you sample the universe, I can give you higher tracking error, but you might actually get worse outcomes.

55:06And then in the end, we kind of discussed that incentive structure in different types of strategies. I would say that I can see in some cases the performance fee structure could make sense or even preferred by manager. If you are a traditional active stock picker, your strategy is already high volatility or high tracking error, then maybe you will like it. And also, if you have some capacity constraint, you cannot increase revenue through growing your AOM, then maybe getting that through the performance fee will be a better route to do it. But then for systematic managers, we've talked a lot about striking that right balance between expected return, risk control, and cost, and diversification.

55:46So I don't think that the incentive to increase tracking, it's probably, if anything, will be detrimental, right? It will probably lead to less robust portfolios. And then we also talked about earlier how the premiums can be very volatile. So that means that if the manager is not getting paid when the premiums are negative, you're to some extent penalized for staying focused on the premiums. But we all know it's important to stay focused in order to be there when the premiums turn around. So I feel in that sense, having that fixed fee is probably more appropriate just to recognize the effort that's required to stay focused on the premiums.

56:23What was it like writing a paper with Bob Merton? Is that pretty cool? Yeah, it is pretty cool. I actually feel quite fortunate that not only with that paper, I get to do a lot of research update calls with him. Of course, you benefit a lot from his insights and his guidance. But at the same time, I'm just very impressed by how excited he is about another paper. Because he probably doesn't need yet another paper to show how good he is. But he's still very excited. And when it gets published at the journal, he's very happy. he sent us those emails. And also he was very careful even sending us some edits on wording in the footnote.

57:00So I feel all of that is very inspiring for me. I feel if I can continue to have that passion, I'll be pretty happy. We had Bob on in episode 234. He was very enthusiastic about the nerdiest topics. I'm sure. And such a gracious, he was a great guest. Did it last like five hours? It was like two and a half hours. It's one of our longest episodes. And he kept apologizing. I'm not surprised. Apologize for taking so much time. We said, you're Bob Mert. Talk as long as you want, Bob. Yeah, his generosity is quite impressive too. Yeah, it really is. It really is incredible. I want to come back to something we talked about briefly earlier, which is short run reversals.

57:37Can you talk about how short run reversals are different from momentum? I kind of said very earlier, you know, what is short run reversal? As a recap, it's basically the tendency for recent winners to underperform, recent losers. over a very short horizon. So we're talking about ranking performance over the last couple of weeks up to a month. And then you see the reversal kind of over the next few weeks as well. Now, momentum is to some extent the opposite, the tendency for continuation. So the winners over the last six to 12 months, typically that's the horizon, so longer horizon, to continue to outperform losers.

58:14They're related in a way that typically when people construct momentum signals, you would skip the most recent month, right, for individual stocks. Reversals is the reason for skipping that most recent month, because the most recent periods, that's more reversals. How are reversals related to liquidity? So that's a paper that has done kind of looking at the reversals and the liquidity in the time series. And we looked at in the cross-section. But more broadly speaking, if you think about the returns to a reversal strategy, that basically approximates the compensation to liquidity providers. It could be designated market makers or any market participants that provide liquidity basically for the service they offer.

58:56So you can imagine, let's say, if an investor wants to quickly sell a large amount of a particular stock, that demand for immediate liquidity will have some downward pressure on the stock. They will generally be forced to sell at a discount because then the price will later revert back to a higher level once the liquidity demand is alleviated. And then that reversal then is basically the compensation for market makers for providing that liquidity. Because when they provide the liquidity, they are exposed to the risk of the price falling even further, right, before they can actually offload their inventory.

59:31And you can also imagine the opposite happening when investors want to quickly buy a large amount of stock. And I would say that the understanding of this link to liquidity provision is actually quite important. because one of the contributions of our study is to actually construct a cleaner signal to isolate the liquidity-driven reversals. Because you know, price change for many reasons, right? Could be due to liquidity, could be due to information. And if it's due to information, it shouldn't be causing reversal. So that's why if you look at the literature, if you just rank on stocks raw returns over the last month, it's polluted by some of the news-driven effects.

1:00:12So that's why in earlier papers, some of the effects, reversal effects, is not very strong. So what we did was actually to adjust the past performance for two news-related effects, post-earnings announcement drift and also short-run industry momentum. What that means is that basically we don't just look at the raw returns of the stocks. We basically look at their returns relative to their industry returns. And we also take out the returns during the earnings announcement window. And by using that signal, you get much stronger reversal effects globally, not just in the US, but also outside of the US, including emerging markets.

1:00:47Yeah, it's a really cool idea. This one's published in the Financial Analyst Journal, right? Yeah, earlier this year, it got published. We're very happy about that. Hopefully more people get to see that and give us some feedback and comments. Yeah, very cool. How do reversals vary across different types of stocks? Oh, yeah. So that's definitely the focus for our analysis. Essentially, you think about how it's related to liquidity. Then you can imagine the less liquid the stock, then the higher the compensation for providing liquidity for it. As a result, reversal should be stronger among stocks that are less liquid.

1:01:20But the problem is liquidity is a very nebulous concept. It's very complex, very multidimensional. There's a huge literature on that. So basically, there's not necessarily a single variable that is a comprehensive measure of liquidity, which is why we kind of looked at a few different aspects. We looked at the market capitalization of stocks. We looked at volatility and turnover. So the reason that they are related is that intuitively, you know, the smaller cap stocks and the more volatile stocks, they're probably riskier to hold in the market maker's inventory. So reversals should then be stronger among those stocks.

1:01:56And then similarly, if you think about turnover, if a stock trades less often, then that means it probably will stay in the inventory for longer. We see some evidence of that from other papers. So that means the reversals should also persist longer because it takes longer for alleviating that demand. So then what we confirm is that that's indeed the case empirically. We find that in particular, the effects on volatility and turnover are quite dramatic. So basically, you see that four stocks are very volatile. the reversals happen very fast and initially very strong. And then for stocks that have lower turnover, the reversal effects tend to be a lot more persistent.

1:02:35Ultimately, the cumulative effect can be pretty strong as well, which also means, and you related to my comment earlier about how you measure reversal signal, if you just use the standard one-month horizon, that's typically in the earlier literature, that's simply too long for some of those stocks. For high turnover, high volatility stocks, the reversal effect has already dissipated before you kind of see it. So that's why it's more sensible to use a shorter look back period, which is why in our implementation, we will use the cleaner signal I was describing, but also kind of over a shorter look back period.

1:03:08Can you talk more about how Dimensional will use this in implementing portfolios? We actually just started implementing this year. All of our global equity strategies are very exciting for us. The general idea, though, is just to say, well, if it's related to liquidity demand, the compensation for providing liquidity, what you want to do as an investor is to reduce the cost of demanding liquidity and then to increase the compensation for providing liquidity. But then how you implement matters. If you think about a standalone reversals, let's say you just chase the most recent losers and then sell the most recent winners, that's not practical.

1:03:47The turnover will be extremely high and then the cost will be probably eating into any premium you can get, even if you just look at monthly rebalancing frequency with monthly signal, you're already looking at a thousand percent one-way turnover. So that's not practical, which is why in our approach, we will use reversals basically as a reason not to trade instead of as a reason to trade. So we can delay buying the recent winners and then we can delay selling the recent losers until the reversal effect dissipates. So that's what we call reversal screen. It's very similar to how we incorporate momentum.

1:04:20We call it momentum screen. So basically, you can then use the screen on your buy and sales, and that way you can increase the expected return of your strategy, but really without impacting the strategy's long-term focus on size, value, profitability, and also that our performance will come without additional turnover and costs. We actually, in addition to the FHA paper, did a shorter block just to estimate the value odds. So we looked at new data from 2000 on, basically the period. Post-decimalization is a little bit more representative of today's market microstructure. We see that across different strategies, across different regions or core versus component, the value as is somewhere between one to 10 basis points.

1:05:03We think that's quite meaningful if you can get that without additional turnover. Because in the end, if you think about these type of systematic strategies, you're fighting for every basis point. We'll keep fighting whenever there's something that really clears that hurdle for us that we're able to implement, that's always very exciting for us. Can you talk about how just the process at Dimensional of there's an idea for something like this, how does it go from that to research, to publication, to implementation? Every single project, I think the process will be a little bit different, but by and large, you start by getting your hands dirty with data.

1:05:36With the reversals analysis, we looked into the US data first, and then we decided to look into DevEx US and emerging markets. And then we started with monthly data, and then we really wanted to look at shorter signals. So, you know, looking into daily data, which is definitely non-trivial. And because Robert Novy-Marx, that's going to work with him, he has been very interested in this topic and he has done some work on his side as well. So it was natural for us to collaborate with him. It was very close collaboration. We were kind of talking every single week and then to compare results and think about next steps.

1:06:07So that's more the research part. Then once we get comfortable with the results, think, oh, there is something actually interesting and maybe potential useful for implementation, then there's the process of bringing that to the Investment Research Committee. So that's basically the committee setting the big picture. Basically, you also need to approve, right, what the changes we need to add to our investment process or enhancements we put into our strategies. And that Investment Research Committee, I was talking about, you know, how scary it is to get judged by such a crowd. There are top financial economists, academics on that committee.

1:06:40So people like Ken French, InFama, Robert Merton, but also our seasoned investment folks, not just the research, but also portfolio managers and training, because then they have much holistic view of what that may mean for our investment process. And then there will be some back and forth. And then when it comes to the implementation, it will be truly cross department work to kind of move that into something that's operationalized. And then once it's in the system, the portfolio managers will actually not turn it on immediately, but actually monitor and see what kind of orders we'll get, what kind of changes will happen.

1:07:15And just to make sure everything in good shape before it gets live in our portfolios. I hope that makes sense. Oh, it makes a ton of sense. And it sounds like just an incredible process to work on something like that with Professor Novi Marks and have the people at Dimensional and on the committee reviewing it. That's crazy. I will remind everyone that not everything will go into like an enhancement in our portfolio or even into a paper. But that's part of our understanding to begin with. When you look at data, look at research, a lot of things will be, even if it's negative results, will be additive to our understanding of investment.

1:07:47Basically giving us good reasons for why we don't do this, why we don't change and so on. And so I feel all of that research is useful. And that's how our researchers are incentivized just to kind of get a better sense of as a pricing and then how things work. Yeah, it's just incredible to think about. I do have one more question on this. So this is published in the Financial Analyst Journal. People can read the paper. How important is Dimensional's ability to, because this is public, like anybody can read the paper and try and implement it. How important is Dimensional's ability to execute on something like that?

1:08:18What makes Dimensional different from someone else who could read the paper and try and implemented in their own portfolio. I've actually got similar questions at one of the panels recently asking, you know, with AI and the chat GBD, does it mean that people can easily replicate your factors or factor models? You would not need Dimensional to do this. Kind of similar to the spirit of your question. I think in the end, what I said actually at the panel is that you can already use those factor models for free. The French website, you can find those models. But But at the end of the day, just the factors themselves are not investment solutions.

1:08:51Just knowing the signals by themselves are not investment solutions. So there's still a lot of expertise to interpret the data and then also think about how you want to implement into your strategies. It also depends on what your investment process, what your investment approach is in the first place. The reason we can incorporate some of these shorter term signals very effectively is one, because we run a systematic strategy, right? We're not in love with a particular stock. So that means we can actually be more indifferent across stocks with similar characteristics. That helps us to use that short-term signal more effectively without impacting the long-term focus.

1:09:29The other thing is we have already built up that daily investment process. It uses a lot of technology infrastructure, but then there's also the human oversight on top. So that means, you know, every day we can do the evaluation. If these stocks are not great today, we can look at them tomorrow. So I feel kind of setting up the whole process is not something you can do overnight. So in the end, there's a lot of devil in the details. So we feel very comfortable to share these results, partially because we feel there's a lot of value adds in how we implement these results. And then we share these papers for people to have a better understanding of what we do in our investment process.

1:10:02Because I feel the more you understand what we do, the more likely that you can actually buy into the idea and become a long-term investor, because that's what's also required to actually reap that higher expected return over the long term. I will say that you guys have gotten much better. I don't know if better is the right word, but you've been publishing much more research like this and research is getting published. I, for one, love it. I'm glad. Yeah. Sometimes I send a lot of papers to our clients and say, okay, here are their weekend reading. Enjoy. For a lot of these papers, we also write shorter term blogs.

1:10:35So if you just want to get an executive summary of things, there's something in that format as well. So hopefully that way we can explain better what we're doing in our portfolios because those investment processes are getting more complex, you know, over time as we refine and we constantly enhance. So hopefully those papers are useful for increasing that understanding. Our final question for you, Wei, for what has been an incredible conversation. So I'm grateful. But here's our final question. How do you define success in your life? I have to say I'm even humbled to be even asked that question because sometimes, especially with two young kids and then busy work, sometimes I feel just surviving the day is already a huge success, you know, believe me.

1:11:16But I do think that the definition of success has evolved over time. It used to be a lot more binary, right? Do I get a good school? Do I get my job offer and so on? But now it's much more about the process than the outcome. What I hope I can do is, you know, similar to what I was talking about, you know, Bob Mertz and being so passionate about research, I hope I can maintain that curiosity about the world, whether it's a new data set I get to work on, a new paper, a new client request, or maybe a new challenge my kids are facing, just still having that passion and really to put in my best efforts to embrace all of them without even thinking about the outcome.

1:11:58If I can do that, I will be very happy. That's an incredible answer to the question. Thank you. It's been a pleasure. It's a fun conversation. I don't know how long it took. Wei, that was great. Thank you so much for joining us. We learned, I learned a ton. Great, great conversation. Thank you. Yeah, thank you.