January 2, 2020

Episode 79: Marlena Lee: Financial Theory in Practice: Gaining Insight from Models

January 2, 2020

Today on the show we welcome the Head of Investment Solutions at Dimensional Fund Advisors, Marlena Lee. Marlena has a Ph.D. from the University of Chicago where she served as the TA to Eugene F. Fama. She has been at Dimensional for 11 years where a big part of her role is communicating what their research team is doing for the advisors and clients who are using their products. In this fascinating episode, we discuss and define models, factors, and the importance of understanding the risks involved with any investment decision. We talk about the many different reasons why stocks have different returns, and what the research says about underperformance and our expectation of positive premiums. Marlena has some interesting perspectives on whether risk or behavior drives higher returns, and shares some of her biggest lessons gained from working with Eugene Fama, and Dimensional Fund Advisors.

Key Points from This Episode:

The uses and limitations of models when making investment decisions. [0:02:30.0]
Understanding the concept of ‘factors’ and why the word is evolving. [0:04:35.0]
Why Dimensional doesn’t combine Price-to-Book with price sales and cashflows. [0:13:10.0]
Marlena’s thoughts on whether risk or behavior drives higher returns. [0:15:15.0]
The theoretical rationale for why we expect the value premium to be positive. [0:21:00.0]
The role of company size in identifying differences in expected returns. [0:25:10.0]
The split between dividend income and capital gains: What is the trade-off? [0:27:40.0]
How to choose which Factor Model to use for your investing decisions. [0:31:15.0]
The good arguments for owning bonds in your portfolio as a young investor. [0:35:00.0]
Risk factors and equities when it comes to fixed-income and bonds versus stocks. [0:38:00.0]
Questions investors should be asking about fees, risk, and portfolio worth. [0:41:48.0]
Evidence that investors can use Yield Curve Inversions to time the market. [0:43:33.0]
Marlena shares her most fascinating research topics and economic debates. [0:43:33.0]
Marlena shares her biggest lessons gained from working with Eugene Fama. [0:48:13.0]

Read the Transcript:

Dimensional is known for applying theoretical models and empirical research to portfolio management. How do you think about the limitations of models when you're making investment decisions?

It might be helpful to first describe why we even have models. We think of models as something that's useful to simplify reality, but hopefully they're still capturing some salient features of how something works. They're really useful for gaining insights about the world, but inherently they're going to be incomplete. Here's an example that I'd like to borrow from Bob Martin. Let's just consider the number pi. One model for pi is pi equals 3 or pi equals 3.14. By the way, I'm not talking about apple pie, I'm talking about the number pi. Another model for pi is 3.1459. Okay. So which is the right model? It depends on what you need it for. So are you helping your kids with their homework or are you trying to compute the trajectory of a spacecraft? You need to have the right tool for the right job.

Let's go to financial markets and here's another example. Consider the efficient markets hypothesis. That's a model that was developed by Professor Fama and he says that prices quickly reflect available information, but does this model hold exactly? Well, probably not. It's still a useful model in that most of the research says that investors are probably best served by acting as if markets are efficient. There's not a ton of evidence out there that investors can really add any value by identifying over mispriced securities. Another insight that we gain from this model is that investors can take comfort knowing market prices are already doing a fantastic job of incorporating expectations about lots of different stuff. So you don't have to pay a lot of attention to everything that's happening in the news or in markets because generally they shouldn't impact your asset allocation decisions.

You said it's tough based on this model to add value from finding mispriced securities. Dimensional has long been known as a specialist in factor investing. So I'm curious, at a very basic level, how do you describe to people what a factor is?

I think of factor as being one of those words that have changed in meaning over time. There's been a little bit of an evolution in terms of how people use that word. The original meaning and kind of how we use it at Dimensional is it's a long shot portfolio. Think SMB or HML in your Fama and French Three-factor Model. I mean, they're designed in order to explain common variation in a factor model. But it seems like over time folks are using the term factor to describe a variable that explains differences in expected returns, and that's definitely our goal here at Dimensional is to understand expected returns.

We are sometimes using the phrasing drivers of expected returns and our assessment goes way beyond just what's a driver or what's a factor. There's lots of other questions you want to know before you can design the portfolios to efficiently pursue these different drivers of expected returns. So for example, why does it work and what's that mechanism for why it works? When does it work and for how long? Where does it work? Like what markets are different market segments, and how does it interact with the other stuff you have in your portfolio? So the way we think of it is the level of your understanding needs to be much deeper than just what's the factor and how many you have.

Can you actually dig into that a little bit more? Like there are somewhere around 400, maybe the number is more than that now, of these so-called factors. So whatever you want to call them, quantitative characteristics that describe differences in returns. But when Dimensional is looking for which factors do they actually want to use in making real portfolio decisions, what does the process look like? How do you make that decision?

Well, there's a lot of work that's done. We don't necessarily want to recreate the wheel. We want to see what's been done in the academic literature, but anything that we might see in an academic paper, we want to really understand better. We kind of think of an academic paper as maybe showing you 10% of the results that were actually run. We want to understand what those other 90 are, and plus a lot more in order to understand how a premium or a source of returns really behaves. What we tend to find is that there really aren't 400 different variables even though that's kind of a number that's often time sided.

What we find is that a lot of the factors, those 400 are just variations on a few themes. So if you look at the data that's commonly used, they're going to fit into a few broad categories. You have your market prices and you have balance sheet data and those are all stock variables, meaning that you get to measure them at a point in time. And then you have other variables that might come from the income statement, and those are more flow variables, meaning that they occur over a period of time. And then for all of those different data items, you can look at check levels, you can look at changes, and then you can look at different ratios. And in the end, all those different combinations, because they're different themes or different variations on similar themes, they're getting to similar information about expected returns. So it's not really the case that you have 400 independent variables that are describing returns. You really only have a handful.

We think that a framework is really important. So oftentimes we use evaluation framework. So that's, again, another model to think about what are differences in returns that you expect to see in the data before you even look. And what that valuation framework tells you is if you have differences in returns across stock, and you also think that prices are discounted at expected future cash flows, then you're looking for low prices combined with high expected future cash flows as being indicative of higher expected returns. I mean, there's tons of robust empirical support for the predictions of that framework.

You could look at a factor that was developed by machine learning algorithm, but you could look at that and say, well, this doesn't actually make sense. Is that kind of what you're talking about?

If you were thinking about something that was derived from like a machine learning framework, I think a lot of the premise behind some of those types of factors are, can we turn through the data faster than the market? I think that if you're hoping to add value by always being faster than the rest of the market, that seems to be a potentially more tenuous proposition. There's not a ton of evidence that different managers can beat benchmarks in a systematic fashion repeatedly over time. What it could help you do is through that framework, what we know to be related to returns are expected future cash flows. So are there different ways that you can measure the variables that are related to returns in a better fashion? And that's certainly something that's ongoing research.

Here's an example that we have at Dimensional where rather than just taking, for example, balance sheet data or profitability data straight from a data provider where your data might get updated quarterly, for example, because it's an income statement item. If there's an event that happens where we think that those data items are stale and we can find a public source of information that we think leads to a good adjustment in that data, for example, a proforma accounting statement, those are things that we would bring into the process in order to more accurately reflect the book value of a data or the profitability of a data.

I've got a big high level question for you, Marlena. If you look top down, many investors own market cap weighted index funds and don't have any sort of factor tills inside their portfolios. In terms of extra return but also reliability of those returns, how much do you think those investors are leaving on the table in terms of return and reliability of those returns?

Well, let me back up a second because I do you think that depending on your goal, the market portfolio could be a great portfolio. It is well diversified. It tends to be low turnover, typically low expense ratio. But to your point, if your goal is to outperform the market, then overweighting those areas of the market with higher expected returns. So smaller cap securities value or companies with higher profitability is a good way to do that. And when I say good, I mean, there's a lot of theoretical and empirical evidence to suggest that those are robust drivers of returns and that you're tilting the odds in your favor of outperformance by pursuing outperformance using that method.

Investors have to be able to tolerate differences from the market because those premiums are never going to be a certainty. We can't guarantee those premiums are going to be positive over any period of time. If they were certain, then there really isn't a good reason to expect the premium. They are volatile. So being able to cope with that tracking error and periods of underperformance is kind of part of the deal.

What about from the perspective of reliability? Like you can still get a good outcome with the market portfolio. Is it going to be as reliable as having a portfolio that's diversified across different risks?

It's a great question in that we do think that if you have information about expected returns, wouldn't you want to use all of it? And we think that different variables, some market capitalization or price-to-book or profitability or asset growth, momentum, all of those different variables, they bring different information to you about expected returns. So it doesn't matter if you're pursuing a market-wide portfolio or a segment of the portfolio, like let's just say Canadian equities, wouldn't you want to use all of the information available to you as opposed to just pursuing only value or only small caps? That certainly is true that if you have an integrated approach to using all of the available information, that does improve the reliability of outperformance, meaning that even if you have the same level of expected outperformance, that by having an integrated approach, that there will be more periods under even holding the expected outperformance constant that you would actually outperform your market benchmark.

The value factor is one of the earliest. I think it was the second factor, or whatever it was called at that time. Anomaly, I think, as it was first identified, but HML, high minus low, the value factor. Dimensional uses book to market to define value in portfolios, but other firms and other pieces of research have suggested that other metrics might be better, maybe using earnings to price, cashflow to price, sales to price, stuff like that can actually give you a better premium in the data. So, why is it that Dimensional uses book to market?

This is certainly not the first time we've gotten this question. We've gotten this question a ton, so we've looked at it a ton. We do still use price-to-book, but we use it in combination with a lot of other variables like size, like profitability, things like momentum, SEC lending fees even. Why though that we don't combine price-to-book with price-to-sales or price-to-cash flows, at least now, is that our research doesn't suggest that variables like earnings or cash flow really add to our understanding of expected returns. So in a lot of different ways of testing this, what we find is that all of the information contained in a variable like price-to-earnings or price-to-cash flows are already subsumed by size, price-to-book and profitability. So it's not bringing new information, it's kind of bringing overlapping information.

I do want to recognize that there have been studies that find earnings-to-price deliver better results than price-to-book, and over certain periods. In fact, the recent period. So let's just say the last decade. But those aren't controlling for profitability typically. So when we find portfolios that are sorted on earnings-to-price, they pick up a little bit of profitability and that kind of makes sense because earnings is like a profitability variable. So in those periods, like the last decade, when you have a positive profitability premium, that dose of profitability can help and that's what those studies are picking up, but it's no better than directly using profitability, which is the way we've chosen to implement.

No, that's really interesting. So by using one of those other metrics, you're picking up value, great, but you're also picking up some of something else like profitability.

Yeah. That's what we tend to find in our research. So we think that we've captured that profitability, but using it in a much more direct fashion.

Right. The concept of factors is certainly becoming widely accepted, but the explanation of what drives those factors and those higher expected returns, whether it's risk or whether it's irrational behavior I think it's safe to say is pretty hotly debated amongst many people in the industry. You take the approach or Dimensional takes the approach that risk is what's driving these higher expected returns. Can you talk about how you feel that's the best way to think about it?

I think that there's actually a lot of different reasons why stocks may have different returns. Certainly risks are part of it, but you can have investors have different tastes or preferences for certain stocks over others and that would be just one more reason to expect differences in returns. I actually don't think of it necessarily as risks or behavioral. I kind of think of it as risks and behavioral because both contribute or potentially contribute to driving differences in expected returns. The framework we used to think about that is once you have those differences, what's the best way to identify them? And that's where that concept of prices or discounted expected future cash flows comes into play, but valuation theory is actually silent on whether those differences are due to risk or mispricing.

Let me give you an analogy. Let's just say you observe two people with different mortgage rates. Maybe they have different credit scores or maybe the banks just had some non-risk criteria that drove the differences. But regardless, the inputs that determine that mortgage rate are going to be what's their loan amount and what are their mortgage payments? And then I can back up the lending rate. That's very similar to what we do. For stocks, it doesn't tell me anything about why the expected returns or why those mortgage rates are different, just that they are. There's certainly a lot of research that's been done in order to try and ask what are the risks or what are the taste preferences, behavioral considerations that might drive these differences. And I don't think we're ever going to see the end of that debate, but it seems reasonable that on both sides no one's arguing that different stocks should have different expected returns and therefore this premium should exist.

How do you think about persistence in that context? So if we don't know and arguably can't know what's driving differences in returns, if it's behavioral, if it is we can't know, but if it is, I think there's a reasonable argument that it should go away eventually. If it's risk based, it should persist. How do you think about that?

It's a real consideration of how much confidence do you put on whether a premium will continue in the future. And if there's a risk based reason, then people tend to have more comfort that that's something that won't be arbitraged away, whereas if it's completely behavioral, then why isn't the case that it then becomes arbitraged away? So that's certainly a consideration, but even in the risk camp, the different kind of risks that people face also probably change through time or the importance that people put on that change through time. I think what's the important thing here is that what shouldn't change through time is that there should always be differences in expected returns across securities. It seems strange that that wouldn't be the case. How much is due to risk, how much is due to mispricing I don't think we'll ever know.

Maybe it varies through time, but there are always going to be those differences. There certainly should be differences in risks. We certainly see those differences come through in fixed income instruments or fixed interests. And if you think about equities as the residual claimant, then they should just be amplified. So thinking about there certainly has to be a difference in risk component that's driving differences in expected returns. And I kind of think of if there are additional behavioral reasons, then that's just icing on the cake.

But in terms of implementation, you do want to be careful. Momentum is a great example of something that people don't think has great rational risk based reasons for why it exists. So that's certainly something that you look at and maybe you think, hmm, is that something that's going to be continuing the data? And the other piece to that of how momentum behaves is that because it's so high turnover, and if you have more doubts of whether that will continue in the future, maybe you want to treat it differently because a typical momentum portfolio in the academic literature might have 2 to 300% turnover.

What we see in live funds is something more on the order of 75 to over 100% turnover. When you have all of that turnover and maybe not great reasons, great being risk based reasons for why it exists, is there a way that you can pursue them that has kind of lower opportunity costs associated with them so that if the premium does either get reduced in the future or even go away completely, that you're still ending up with a great portfolio. We use it in a different way. We don't implement it the exact same way as the academic literature would implement it in those studies. So there are ways you can still use this information in a variable like momentum and have it be a low cost approach.

Yeah, it's interesting. I guess that kind of speaks to the question that we asked at the very beginning, which was the limitations of models. You can't just take what momentum looks like in the data and put that into a portfolio.

Yeah, you do want to give them some thought. And that's where I was trying to get with the go beyond what's a factor and you do want to understand how it behaves so that you're not just pursuing every single thing the same way. Like how you pursue value might be different from how you pursue momentum because they behave differently in the data. There are different theories for why they exist in the data and that might influence your confidence in how these premiums will behave in the future.

We were just talking about persistence and what is driving factor premiums, whether it's risk or behavioral. The value factor has had a good decade or more now in the US of a negative premium. So value stocks have trailed gross stocks pretty substantially. And we just talked about how we don't really know what's driving any risk premium at all. If it is a risk premium, is it risk or behavior? So thinking about value underperforming and not really knowing what's driving any potential future positive premium, what would it take for us to stop believing that value's going to deliver a positive premium in the future?

When I think about the theoretical rationale for why we expect the value premium to be positive, I think that rationale is evergreen and it really supports the expectation that the value premium should be positive every day. So that rationale is just if you pay a lower price, that should indicate higher expected returns. That seems hard to argue that it is specific to what decade or even what century the investor is investing in. But we do know that these premiums are quite noisy. They are just as volatile, for example, as like the equity premium is also quite volatile.

There's a recent paper by Fama and French called volatility lessons and in it they assume, let's just say you had the historical distribution. So that means you knew that the value premium was going to be, in their paper it was 29 basis points a month. So roughly three and a half percent annually. Even in that case where you knew that the value premium was positive in three and a half percent, there was still more than a 5% chance that value would underperform the market over a 10 year period. So even though we expect these value premiums to be positive every day, the range of outcomes suggests that a 10 year period of underperformance isn't actually that unusual. And that's something that is I think really important for investors who want to pursue these premiums to just understand how much noise there is in them because you really can't interpret too much when you do go through these periods of underperformance.

It's noisy, but the market could do the same thing. I think that's one of the things people often miss is that values underperform for 10 years, but there have been decades where the equity risk premium has been negative as well.

Absolutely. In the US, there was a 17 year period where US equities underperformed T bills, and people have the most confidence in the equity premium. So going through that period, you have to have a framework for how can you stay invested even when you go through those periods because they can and have happened in the past. Certainly not pleasant to go through though.

Definitely not. But that's why combining the different, like if you have market, which can underperform for a decade and you also have value, which could underperform for a decade as it has, they're not necessarily and maybe even probably aren't going to underperform at the same time or deliver a negative premium.

Right. And that's what we've seen is actually people generally, if they're in a well diversified global portfolio, have seen pretty good positive returns over this last decade. It's just they haven't really kept up with, in particular it's been the S&P 500 that's just done really well.

It kind of comes back to the whole concept of the limitation of models. Like one of the questions that we get sometimes is, why don't you guys just go all small cap value?

I think this last decade tells a lot of people why. It might be the portfolio with the highest expected return, but darn there's a lot of noise in it. Are you going to be able to stick with it when you go through one of these periods? Sometimes we get the question of what's the right asset allocation. It's kind of going to be the asset allocation that a person can stick with. If an asset allocation is not one that an investor can stick with and stay invested during a period of disappointing performance, then it wasn't the right asset allocation.

You mentioned having the valuation framework to think about differences in expected returns, and we've talked in a fair amount of detail in the podcast in the past about the valuation equation. It doesn't contain anything about size, like small cap stocks don't show up in the theoretical valuation equation, at least assuming that we're talking about the same one, which I think we are. And other papers have shown that on its own, small size, small cap stocks don't actually have their own premium. So it's not in the valuation equation, it doesn't necessarily show up in the data. How do you think about the role of company size in identifying differences in expected returns?

Well, just like what we were saying earlier, we know that a model is incomplete and that applies to the valuation model as well. So we know not everything we include in our process fits neatly into evaluation framework. It is a model. It's going to be incomplete. Size is tied to price. It's price time shares outstanding, and we believe price is tied to expected returns. So there is a tie, but to your point, I understand that it's not a direct tie exactly to the valuation equation. But empirically, most asset pricing models do find that including size is generally additive in understanding returns. So you see that in different factor models and different cross-sectional regressions that it does matter in the data. And then also in terms of live performance.

Dimensional has had success at delivering the size premium. That's most easily observed in funds that only invest in small caps, such as the ones that we have in the US. So our US domicile, small cap funds have outperformed large cap indices since inception and that's net a fee. And I'll point out that you also get that small cap exposure in the Canadian funds that it's just part of an allocation across the entire market that also includes a focus on value and high profitability.

Speaking about performance, Marlena, dividend growth stocks, for example like the ones in the S&P Dividend Aristocrats index, have tended to outperform the market. And this is one of the hottest topics, certainly on Ben's YouTube channel, for debate. So the obvious question is, why wouldn't we all build portfolios based on dividend growth stocks alone?

Well, first it's important to understand why that index has performed. That portfolio behaves like a portfolio of high profitability value stocks. You can see that through a factor model. And what do we know about stocks that have high profitability and low relative price? We know those are the types of stocks that have been associated with higher average returns. And once you control for those exposures, you've also explained the returns of that type of portfolio. We can see that by their five factor alpha loadings, for example, are close to zero. That means that it's possible to tilt to size value and profitability so that you're getting a very similar expected return profile as a portfolio of dividend paying stocks.

We understand that clients might have a preference for how their return is delivered. So what's the split between dividend income and capital gains? And we think that that's a fine preference, but it's really important to understand that that preference comes with trade offs, specifically a potential loss of diversification. So for example, the S&P 500 Dividend Aristocrats index had just over 50 constituents as of July of this year. So it's not the most diversified portfolio out there. So it's just a matter of trade offs. We think that having a well diversified portfolio that's focused on the drivers of expected returns, there you're still going to get a total expected return profile that will be similar, and it may not have the same focus on the return through yield and more of it would come through capital gains. But we think that that's split. It's a matter of preference, but we think that in terms of how much net income or how much wealth you have at the end of the day, it's going to be driven by the total return.

So do you have an opinion as to why people almost don't want to hear this kind of rational evidence? Like we've given this explanation countless times in our various channels and we still get a fair amount of pushback on this. But in the end, if only a portfolio of dividend paying stocks helps you behave better, is it even worth to continue to debate if someone will perform better based on that belief system?

I mean, if I'm going to speculate, I think that it's a device used in order to put some discipline on the spending approach. I don't want to touch principle and I'm just going to pursue or I'm just going to spend from dividend or yield income. There are different ways to address that. If you're working closely with an advisor, then the advisor can help bring some discipline or awareness around what's a sustainable spending rate or a spending policy. So I think that there are other ways aside from just only allowing yourself to spend from dividend income in a portfolio.

Another consideration or trade off that people just need to be aware of is if it's the case that the preference for spending out of dividends drives one to a riskier portfolio, one that might not be appropriate for their risk tolerance and for their financial goals, then that seems to be a potentially suboptimal outcome from that preference. I do think of it as a matter of trade offs. It's a preference that we hear from a lot of clients. I do think it comes from this idea of not wanting to touch principle. But on the other hand, if it's leading to other investment outcomes, for example, too much risk in the portfolio or inadequate levels of diversification, then it seems better to try and embed spending discipline in another way.

I've always thought it's kind of funny to let corporate dividend policy dictate your retirement spending policy.

Yeah. Especially when you see fewer firms are paying dividends these days, there's other ways that they have to pay back money to investors. So share repurchases are becoming a much bigger way for returning capital to investors.

We've talked about factors and we've talked about the valuation equation, valuation theory. Dimensional is very much aligned with the Fama-French thinking and they've got the Fama-French Five-factor model. We've talked about a lot of the factors in that model, but AQR has a factor model, MSCI has a factor model, and there are probably many, many others. Those are just the ones that came to mind. Once someone's decided to go down this path of factor investing and decided that there are differences in expected returns between securities, how do you choose which factor model you should use to evaluate your investment decisions?

Here's what are similar across all models. We know they're all incomplete. They all require estimates. Those estimates come with noise. And kind of like how we were saying earlier that you need to have the right model for the right job, it kind of depends on what you're using that factor model to do. A common use would be to use a factor model to gain some perspective about how a portfolio is positioned. That's a pretty common use of a factor model. If you're using it for that, then you just have to be aware that there's a lot of noise in these estimates. So don't read too much into the output of any one model. Looking at something in a variety of ways can sometimes be helpful, like at Dimensional we call that surrounding the problem.

So we might look at a Fama and French Three-factor model, a Fama-French Three-factor plus momentum model, a Fama and French Five-factor, five plus momentum. We'll look at a lot of different ways, but then we'll also look at characteristics. We'll look at how the holdings are allocated. That does take a lot more data and analysis in order to do that. But all of those different methods give you different views of the portfolio so you can get a more complete picture. I would just advocate that there's no one model is going to be perfect.

Do you think there's any merit to like, and I don't know if there is, but if you look at one factor, like say the Fama-French Five-factor model, and you look at another model, like the whatever, the AQR factors, are there going to be differences in characteristics like the amount of turnover that you're going to get based on the way you're defining factors? Like is a simpler factor definition going to give you an easier to implement portfolio?

Well, that's a different question because you'll notice that the purpose that I had for my previous example is to understand what's in a portfolio, how it's tilted. If you're after investment implications, there I think it's really important to go well beyond just thinking about a factor model. We would cut the portfolio in a lot of different ways. We do use time series factor models. We'll use cross-sectional regressions. We'll use a lot of portfolio sorts and look in those portfolio sorts how a premium shows up across different segments of the market, how it interacts with other premiums. Certainly how much it turns over is a super important consideration. If you have something like size, value or profitability, those variables tell you something about expected returns for years and years and years; whereas a variable like momentum, momentum is telling you about returns for the next six to 12 months, tends to decay completely somewhere around the nine month mark.

So the turnover that is associated with a particular variable, where it shows up. So like with asset growth, for example, we see companies with really high asset growth as having very poor performance amongst small cap, but it's really among those with extreme high asset growth where we see it. So they all kind of behave a little bit differently and all of those questions are important considerations for deciding the investment implications, which in my mind extend far beyond a factor model.

I have a fixed income question for you, Marlena. Let's assume I'm a young investor, say 25 or 30 years old, have a stable job, stable income, and I'm investing for the long term. If you assume I can handle volatility, so take out the behavioral considerations, what are some of the good arguments that I should own bonds in my portfolio?

If you're a young investor, this kind of gets to that point around what's the perfect asset allocation. For a typical young person, and let's just say there we're talking about saving for retirement, because if it's a near term goal, then maybe fix bonds make sense. If it's saving for retirement and retirement is decades away, then a lot of the both rules of thumb, but even Bob Martin who's done a lot of work in life cycle finance would say, hey, this person can tolerate a lot of additional risk in their financial portfolio. The reason for that is that a lot of that retirement is going to be funded not from that financial portfolio but from their human capital. So it's actually their future savings that would fund the bulk of their retirement if you're talking about someone who's young.

And to the extent that their human capital is less risky than equities, you can think of that as serving as the safe part in their overall portfolio, and therefore they can take on a lot of equity risk. I think that generally what we see is that the younger an investor is, the more allocation they have to equities. But again, if you have a specific investor who says, hey, I really can't tolerate the volatility of equities, and during an equity market downturn they're not going to stick with their asset allocation, then it's not going to be the right asset allocation for them. Maybe that person, because of their risk tolerance, needs a greater allocation to bonds in their portfolio.

And then it's a question of making sure that they're aware of the trade offs. That portfolio will have lower expected returns because we expect a positive equity premium. And what does that imply? Either they're going to need to save more or at least in expectation have less accumulated for retirement. So those are the trade offs. And I think that that's just really important to make sure that we're always highlighting the trade offs associated with any financial decision.

I think you might have just answered the question that I want to ask, but how far can we push that logic? Like 100% equity makes sense for someone who's young and has a stable income, what about leverage?

That's something that you'll certainly hear is that actually someone who's young, maybe they should have a levered equity portfolio. We tend not to see it too often. I'm not sure why that's the case of whether it's the cost of leverage or whether the appetite for having that level of volatility, equity risk, but also then those higher expected returns are worth it for most investors.

Risk factors and equities are talked about a lot and we've talked about all different angles of that already. When it comes to fixed income, I don't think that anybody really talks about factors. Can you apply the same thinking to investing in bonds as you can with stocks?

Yes, a similar concept for sure, but the details of how you implement a systematic approach and fixed income should be different, well, because bonds are different than stocks. Here's what similar is different bonds we think should have different expected returns. And we also still think that prices are discounted in expected future cash flows. It's just that for fixed income, that discount rate is readily observable and it's called yield. So for a bond with no risk of default, the yield is the expected return if you hold that bond to maturity. And then if you don't hold it to maturity, you can actually calculate the expected returns based on the yield and the steepness of that yield.

So those two things combined are called the forward rate and they're readily observable every single day and you can use them to target the highest expected return bonds within a universe, and that's the systematic approach to fixed income that we've applied for over 35 years. There's other variables or other studies that we've seen to try and explain differences in fixed income returns. The important part though is you have to control for the things that you already see in the data. In particular, you get to see yields. So one thing that you have to be careful of is not to use a noisy proxy when you actually have the variable that's readily observable. The analogy I use there is if you want to know the time, look at your watch, don't look at a sun dial.

I have a separate question for you about different asset classes. We all know about stocks and bonds and diverse portfolios and factor tilting. We've talked about that a lot. A question we're getting more and more is, at what point does it make sense to consider adding in other asset classes like liquid alternatives or hedge funds or private equity?

Yeah. Here's my framework. We're big on frameworks for thinking about whether an alternative investment makes sense. Here's the first question is, does it expand my investment universe? I think that that's a really important question if you're asking about diversification potential. In the case of liquid alts, the answer is typically no. A lot of liquid alt strategies, they might go long in short different securities that are already held in a global portfolio of listed stocks and bonds. And then when you add them up to your global stock and bond portfolio, what you get is net exposures that might result in either underweight certain securities or overweight other securities.

At Dimensional, we are doing something very similar in our portfolios. We're underweighting securities with lower expected returns, overweighting those with high expected returns. So we think that there's really good reasons why you might potentially want to do that. But then because it doesn't expand the universe, we don't think of the motivation for that is whether it increases diversification. You want to then understand, well, do I think that we're over and underweighting because we're pursuing a driver of expected returns?

And then all of the other things that we just talked about about how robust is that driver, how confident am I that it will continue in the future? Those are the things that need to be assessed. And the diversification argument is just a red herring. Sometimes you might have the answer to that first question, does it expand my investment universe. And the answer might be yes. For example, private equity I do think as an expansion of the investment universe relative to a public markets only portfolio. So maybe there's something to the diversification argument there, but then the other questions that investors should ask are, what are the fees? How much idiosyncratic risk am I taking? And then also just do I expect it to be additive to the return profile of my portfolio? Is it worth a place in my portfolio?

The idiosyncratic risk question in particular is really important, especially when considering private equity, because the range of outcomes across private equity managers is huge. So there's one academic study that reported a cross sectional standard deviation of over 40%. Let's assume that there are good private equity managers, an investor should also ask, do they think they can identify them in advance? Would they take my money or would they have enough capital raise from big institutional investors? And the economic principles that predict... Well, you wouldn't necessarily expect that even if you found a great manager that you should get higher expected returns because economic theory would suggest that if it's the manager with the skill that's the scarce resource, then they should be the ones who reap the benefits of their skill through higher fees.

So it's kind of hard to get good data for private equity performance and that's because all of those databases tend to be self-reported. So because of that, there's oftentimes biases in the data. So the academic literature is actually pretty mixed. Depends on the data source used and how they accounted for different biases, how they benchmark the funds. But I will point out one study that provides some anecdotal evidence and that's the NACUBO-TIAA Study of Endowments. So their 2018 study included over 800 US endowments and that represented over $600 billion in AUM.

If you just take that aggregate portfolio of all of those endowments, half of it was invested in alternatives. So that includes private equity, hedge funds, private real estate, natural resources, all of your typical alternative categories. And over the 10 year period ending June, 2018, the average endowment underperformed the Dimensional global 60/40 index. So the average endowment earned an annualized return of 5.8% versus 6.8% for the global 60/40. And that's with a higher standard deviation. So again, just anecdotal evidence but it doesn't really support the notion that investors are better off with the addition of alternatives.

People in Canada and probably in the US too have been pretty worried about an economic recession. I think mostly about a US economic recession. I think a lot of that concern has been driven by the US yield curve inverting. Is there any evidence that investors can use yield curve inversions to time either the market or factor premiums?

No. We've done some work with the internal research team at Dimensional. Professors Fama and French have written a piece now where they look to see, hey, can you use something like a yield curve inversion in order to make either our market or factor timing decisions, and the evidence doesn't suggest that you should. There's really no evidence that you can have a better investment outcome by doing that. And even from a theoretical perspective, it doesn't really make much sense. Like how could the expected equity premium ever be negative? Who would invest in stocks if that were the case?

Beyond yield curve inversions, there's been a ton of empirical work. We, for example, have looked at a whole host of different market timing signals, from ones based on mean reversion, aggregate price ratio. So that includes things like Shiller's CAPE ratio, macroeconomic variables that includes yield curve inversions, but also things like interest rates. And what we've consistently found is that the best approach is a disciplined one. And remember, in order to time market successfully, you have to make two right decisions. So even if your odds are 70% chance of getting it right, which I would say is pretty darn high, what are the odds of doing that twice? That's 49%. So that's just a little numerical way to show why it's so hard to tie markets.

I really like that example. I haven't heard it framed that way before. Even if you have a positive expected outcome on one side and on the other side your overall expected outcome is negative.

Yeah. It's hard to do it two times in a row.

You're having an incredible career research in this industry obviously. I'm just curious, what has been for you the most interesting research finding since you've been at Dimensional? Do you have a favorite? Is there like a favorite child?

Yeah, exactly. I was going to say it's kind of like your favorite child. You're supposed to love them all, right? And yeah, I think that that will actually be my answer is I have been surprised many times with the research process at Dimensional. I wouldn't say it's a finding, it's the process. I did my PhD at Chicago Booth and the finance seminars at Chicago Booth are kind of known for having really lively debates, and I think a lot of that stems from a healthy skepticism about just patterns in the data. So when I decided to go into industry, instead of doing academics, I don't know, I guess I assumed that I'd be leaving a lot of that behind. That's certainly what people kind of tell you or lead you to think.

But what I've found is that's not at all true, especially at Dimensional, or I guess I can only really speak for Dimensional because that's where I went straight out of my PhD program. Here it's where I really learned to appreciate that there's a lot of noise in empirical asset pricing research and the way you have to deal with that is you look at the data a lot of different ways in order to try and surround that problem. And I think our approach is unique. When you look out at the landscape, I don't think there's anyone who really implements the research in the same way that we do. We all have access to the same data, the same academic studies. The difference is really in the interpretation and the implementation of those ideas. That's where I think that Dimensional really does interpret and implement differently.

You've spent more time with Eugene Fama than most people will ever get to in their lives. What's the most interesting thing that you've learned from working with Gene?

There's a lot of different things. Here's one that I'll point to, and this is something I learned really early on in my experiences with Gene. As a first year PhD student in his class, and then later as his TA, there is a lot of thought that goes into his language. Precision and language I think can really only result from precision in thinking. So I remember grading exams where there would be a whole essay in order to circle around an answer that should really just take 10 words. We would doc a lot of points for that because it showed a lack of clear thinking. And that's something that I certainly noticed as a student and really started to appreciate once I got to Dimensional where we're also really big on precision and accuracy of language, and that I think comes from all of these decades of working with Ken and Gene.

I also think of it as a reflection of our approach. We want to deeply understand how things work. You're never going to be able to prove things with data and making sure that we're not making statements that we know aren't the case or are stretching what the data can say with how we're describing it in our words is something that we are really careful about. It shows in our communication with clients as well. We want people to understand here's the whole range of outcomes that you might expect because we think that it's really important to understand here's what your experience might look like and that setting those expectations are important in order to have a good investment experience so that people can stay disciplined because you can't control the markets, but you can control how you react to them.

How do you define success in your own life?

This is going to sound corny, but I think of success as a mindset instead of like a destination. Some people might say I'm going to be successful when I do X, Y, and Z. But success for me is having an appreciation for where I've been, kind of appreciation for where I am today and being optimistic about the future because we should always be striving to be better in the future. So using that definition, today I feel really successful. I'm incredibly fortunate for all of the experiences and opportunities that I've had. And there's been so many people in my life who have helped me along the way. I have two wonderful boys. They're crazy. They keep me busy. My super dad husband who lets me have the career I have. My job's never boring. There's always new challenges.

But in the end, all of the work that we're doing at Dimensional is to help lots of people put their kids through college and have a secure retirement. So I think that the work we're doing is also really important from that perspective and that makes it really rewarding and really cool. I know that I'm not perfect. So there's a lot of things that I have to work towards in the future and certainly have all of the people around me helping support me and pointing out those flaws every day. So yeah, using that definition of success, where I am today I feel pretty darn successful.

Marlena, this has been great. We really appreciate you coming on the podcast. Again, thank you.

Thanks for having me.