Not Another Investment Podcast
Understand investing beyond the headlines with Edward Finley, sometime Professor of Finance at the University of Virginia and veteran Wall Street investor.
Not Another Investment Podcast
Charting the Course: Strategic Asset Allocation and Portfolio Optimization (S1 E20)
Ever wondered why some investors seem to ride the waves of market volatility with grace, while others get swept away by the current? Our latest episode of Not Another Investment Podcast offers a lifeline to those floundering in the choppy waters of strategic asset allocation. I'm your host, Edward Finley, guiding you through the complexities of matching your risk appetite with your financial aspirations. We tackle the psychology of loss aversion, the tug-of-war between liquidity and potential gains, and the distinction between systematic and active risks that could make or break your portfolio's performance.
Venture with us into the realm of portfolio optimization where the elegance of Markowitz's mean-variance curve meets the gritty reality of our financial markets. Together, we'll dissect the science of combining different assets to construct a portfolio that not only sings to your individual risk tolerance but also waltzes along the efficient frontier of returns. But the path to investment nirvana is strewn with estimation errors and the ever-present specter of diversification. So, we'll arm you with the knowledge to navigate these pitfalls and manage risk like a seasoned pro.
To cap off, we peer into the Black-Litterman model's crystal ball, revealing how it draws from the CAPM's market efficiency theory to craft portfolios that resonate with current market sentiments. Comparing this approach with its mean-variance counterpart, we illustrate how it shapes conservative and growth-oriented investments. Finally, we shift into the gears of risk control strategies, from the equity-heavy Black-Litterman to the harmonious risk parity approach, promising to inform your next investment move. Tune in for an episode that doesn't just scratch the surface, but charts a map through the investment management labyrinth.
Episode Notes: https://1drv.ms/p/s!AqjfuX3WVgp8un20bLpr6Q2agiEm
Thanks for listening! Please be sure to review the podcast or send your comments to me by email at info@not-another-investment-podcast.com. And tell your friends!
Hi, I'm Edward Finley, a sometime professor at the University of Virginia and a veteran Wall Street investor, and you're listening to Not Another Investment podcast. Here we explore topics and markets and investing that every educated person should understand to be a good citizen. Welcome to the podcast. I'm Edward Finley. Well, if you've been listening to this podcast serially and going through each of the episodes, I want to congratulate you. You've finished learning all we can about asset classes. If you've just joined the podcast here because the topics sounded interesting, let me encourage you to go back and listen selectively, maybe if you want to pick some spots where you feel you need to brush up a little, but to think a little bit more about asset classes more carefully. I think we'll help as we talk about this next topic, which I think is really, for many people, a really important part of understanding markets and investing, that is, portfolio construction. We're going to break portfolio construction down into a couple of different categories in this podcast. Today We'll talk about strategic asset allocation, but then in future episodes we'll talk about dynamic asset allocation and we'll talk about different ways of evaluating our allocations to see if they're going to accomplish our goals.
Speaker 1:But without further ado, let's turn our attention to strategic asset allocation. It's perhaps the most important part of investment management. Over 90% of the variants in observed returns and portfolios can be explained by a variance of the asset allocation. 90% it's huge. More than choosing managers that beat the market, more than having lots of private investments or no private investments, or only passive or only active. 90% is about the allocation to asset classes. That is a hugely important impact. Therefore, it's the thing we ought to take most seriously, it seems to me. But how do you determine the strategic asset allocation? What are the factors that go into it? Primarily, I would say it's about understanding risk tolerance, and really risk tolerance comes down in a couple of different categories, with three really big ones Preferences, cash flow and liquidity needs. So first, preferences Well, recall the behavioral tendencies that we talked about when we did some finance theory, which is things like loss aversion.
Speaker 1:Secondly, we find, as humans, we find losses far more painful than the elation that we get from gains, and so we have to kind of be self-aware about our preferences. We have to be self-aware about how painful would it be for us to experience losses if, even over the long run, we earn a lot of return. In addition, we can think of preferences in terms of terminal wealth values. Right, we have goals. Our goals are going to sometimes be defined by the terminal wealth value.
Speaker 1:Let's say I'm saving up for retirement. Well, that means that my goal is probably to make sure that I grow in real terms so that I'll have enough when I retire. But how about once I get to retirement? Well, now my preferences have changed for that portfolio, because now I'm probably interested in making sure I don't run out, and so that's going to suggest a different set of preferences and, accordingly, a different strategic asset allocation. Or let's say I'm really very, very wealthy, I have more money than I could possibly spend, and this will pass to future generations. Or let's say I'm an endowment or a foundation where I know my assets are going to be almost perpetual. Well, there, my preferences might be subtly different, though still described in terminal wealth values. So, for example, I might want to just make sure that I can support a 5% distribution each year and maintain purchasing power. That may be my goal. Alternatively, my goal may be to support a 5% distribution and to grow in real terms.
Speaker 1:So, again, when thinking about a strategic asset allocation, first we want to understand behavioral tendencies. How do we react as individuals to a set of situations that could include drops in the portfolio's value? How do we think about our goals in terms of terminal wealth? Are we building this portfolio to retire? Are we making sure we don't run out? Do we have things like goals of preserving purchasing power or growing in real terms while still supporting distributions? And then, lastly, we should always bear in mind the importance of relative wealth values. So relative wealth values are different than terminal wealth values in the sense that relative wealth values may be relative to my peers.
Speaker 1:So back to my example of an endowment or a foundation. You might be pretty pleased with your performance, but if other endowments and foundations are earning greater return and taking similar sorts of risk, that would be bad, and so your preference would be not to be that person, right? Or if we take it out of endowments and foundations and just think about everyday people, that importance of funding a college education, for example, or funding retirement, that's a relative wealth value goal. Because relative? Why? Well, because where I send my kid to college or how much I need to spend in retirement will differ person to person, and so it will really be important to think about it in your terms, not in some abstract terms. So, first off, preferences that's the key. Second, cash flow. Now, cash flow is going to be a little different than liquidity needs, and I'll explain why. Cash flow can mean am I adding to this portfolio every year or am I withdrawing from this portfolio every year? And depending on which is the case, my strategic asset allocation will have to be different. Why? Well, if I've got a portfolio with really large net cash inflows meaning every year I add money to the portfolio on a net basis that's a lot of college endowments and foundations. Every year they spend 5% to support their college or university, but they also get contributions and for a lot of them the contributions are greater than the distributions.
Speaker 1:Well, if you've got large net cash flows, then your risk tolerance is higher. Why? Well, precisely because if you've got lots of money coming in on a net basis, you can afford those periods when markets have declined. You can afford those periods when even asset classes all move in unison down, and the reason you can afford it is because you're actually buying assets in that moment that are cheaper than they normally would be. So think about it Every year, you have additional money that you're adding to your portfolio If markets take a nosedive. In a way, you're delighted because that new cash that just came in is going to be invested into the market at lower values, which means future expected returns will be higher. That's great, that's a really good thing, and so, as a consequence, we can afford to take more risk, that is, we can afford to have the portfolio experience greater down drafts like that, precisely because we add on a net basis to the portfolio each year. But what about if it's the opposite? What if we have portfolios with significant net cash outflows? Well then I can't afford to take nearly as much risk. Why? Well, if every year I'm net spending money out of the portfolio, the more risk I take.
Speaker 1:Think about risk, at least defined as volatility. If I define risk as volatility, it's the observed monthly returns. How dispersed are they around the mean? And if it's very high volatility, they're very greatly dispersed, which means we might have months with really big drawdowns and other months with really big increases in value. If I've got that kind of dispersion, but every year I take 5% out of the portfolio or more 7%, 10% out of the portfolio, it means that in down markets I'm selling things at losses rather than holding them for them to recover, and that's going to lock in those losses. Same token In up markets I'm selling assets to fund my distributions, assets that are probably going to go even higher than when I'm selling them. So I'm minimizing the up part of the market and in combination that means that even though my asset allocation for that level of risk might expect an 8% return, when I'm spending at that clip and I'm deepening the troughs and I'm capping the peaks, means that probably I'm going to average something less than 8% but still be taking all that risk. So in those portfolios we want to think about taking less risk.
Speaker 1:Last is liquidity needs. Beyond the ability to take more risk or less risk, it's important to consider the required outflows from a liquidity point of view. That is, do I have assets that create sufficient cash liquidity distributions so that I can fund my outflows and then I can take more risk in the portfolio? So illiquid assets, for example, may have and when I say illiquid assets, right, I'm talking about, we've discussed them thoroughly venture capital, growth, equity buyout, hedge funds, even, to some extent, commercial real estate, farmland, timberland these are all very illiquid assets. Well, these illiquid assets might have really beneficial risk return profiles, we might think when we're building a strategic allocation oh, we want to own that, but if they comprise too much of the portfolio, then there's not really going to be sufficient cash flow to support your distributions Every year. If you're net distributing, you're going to have to sell your liquid assets and that will make for really fat left tails, because there are going to be moments when you don't want to sell those liquid assets but they're down, and they're down significantly, and you haven't got a choice.
Speaker 1:So when we talk about liquidity needs, it's a little different than the cash flow part. Cash flow part is really about adding or subtracting each year from the portfolio, and that tells us something about risk tolerance, but also the degree to which we have sufficient cash distribution coming out of our assets, like from bonds, whether they're high yield bonds, investment grade bonds, treasury bonds, municipal bonds or other kinds of investments that pay dividends, like public equity. That is a terribly important part of the equation in order to see how much risk we can take. It's not to say one's good and one's bad, it's just to say it's going to inform our choice about how much risk we can take. All right, so those are the three things we take into consideration when determining our risk tolerance, our preferences, whether we're adding or subtracting from the portfolio, and how much liquidity is the portfolio producing on its own if we are spending.
Speaker 1:Well, there are some caveats to forming a strategic asset allocation using these inputs. So first, by construction, the strategic asset allocation is just a single period framework. In a moment I'm going to walk through the different ways that we can build a strategic asset allocation, but all of them have in common the very same thing, which is they make assumptions about asset class returns and asset class volatilities and asset class correlations that are dynamic in time. Over time they move, they change. Over a long period they will average out, we hope, to our expectations, but in the short run they move around. The strategic asset allocation doesn't take into account the moving around part. It builds the allocation using just those estimated average return volatility correlation statistics and that is why we call it a single period framework.
Speaker 1:We say imagine a fictitious single year. The portfolio is going to exist only for one year. What do we want to own? That's how we're going to build it, but that means we don't consider any of the changes in expected means, volatilities and covariances that we know will happen in real time. That's the first caveat. The second caveat is there's increased empirical evidence that markets exhibit quite significant variation in means, volatility and covariances due to changes in the way investors approach the market or experience wealth. This has nothing to do with the risk of the asset class. There's increasing evidence that tells us that, while we might expect a certain asset class with a certain risk to earn 8% a year, because historically that's what it's earned, if there have been changes in relative risk aversion that's just generally people are greater risk takers or less risk takers and it lasts for a long time five, eight years or if there are changes due to consumption patterns or wealth patterns. All of those things are going to affect the way people invest in markets and, as a consequence, those same risks that historically might have earned 8%, in the future might only earn 6%. So when we think about building a strategic asset allocation, we've got to take great care in remembering that there's a lot of time variation in these three statistics, and so not only is it a single period framework, it doesn't take into account the ordinary changes, even if the average will be the average. We also have to remember that the average might change in the future compared to the past.
Speaker 1:Another caveat is that, by construction, any strategic asset allocation thinks only about systematic risk. It doesn't take into consideration any return that you might earn for skill, only systematic risks. And that's what we talked about in all of our episodes just systematic risk. But it's important to bear that in mind because, as we'll discuss in a little while, there's a way to bring into your thinking things like active risk, which is different than systematic risk. Active risk is sort of the risk that one takes in doing something different than the overall market risks and who maybe earn return for those differences? We can take that into account. But we got to remember that at least in the first moment. Strategic asset allocation only systematic risks are all we're thinking about.
Speaker 1:So, as a result, when we take those caveats together, any strategic asset allocation needs to be understood as just the starting point of an investment process. It is not the end to sit down, either by yourself or with the help of an advisor, and to say, okay, what are my goals? What is my risk tolerance? Let me think about my preferences, cash flow and liquidity needs. Here's the strategic asset allocation Great, see you in 20 years. That is wrongheaded for all of the reasons that I just mentioned. It's a single period framework. It doesn't take into account that it time varies, even though it might converge on its average. There's a lot of empirical evidence to say that the average is change, the historic average isn't going to be the future average, and that it is not taking into consideration any active risk and you ought to take into consideration. Okay, so with that backdrop, what are the ways in which investors have historically built strategic asset allocations? Let's just say they've determined their risk tolerance. They understand where they're headed.
Speaker 1:Now. How do we build it? Well, we're going to start with our old friend mean variance theory, and here we're going to talk about something called Markowitz optimization, or sometimes called mean variance optimization. So remember back to some of the earlier episodes in which we talked about finance theory. We talked about something called mean variance utility. We talked about this idea of two very famous economists, levy and Markowitz, in the 1950s, who were able to say that the average returns on a portfolio are a good approximation for utility. And utility is just what economists talk about when they talk about how happy we are. The goal is to maximize happiness. And so Levy and Markowitz said. Well, we think that returns are a good approximation for utility and we think that volatility is a good approximation for risk. So what Levy and Markowitz conclude is if we understand how risk averse an investor is and we understand that tradeoff for them between earning returns and taking risk measured as volatility, means that we can find a way to maximize the return for a given level of volatility that makes the investor the most happy. The tradeoff between return and volatility is really the risk aversion of the investor, all right. So how do they do it?
Speaker 1:Again, if you have access to the slides, which are in the show notes, you can take a look. If you don't, not to worry, you can look at it later. I'll just describe them in words. So the first thing that you look at in a mean variance optimization model is something called the mean variance indifference curve. So what's it mean? Well, it's really just a chart where the volatility is on the x-axis and the return is on the y-axis and you can draw a bunch of lines that are upward, sloping from left to right, and they describe all of the different utilities that you might experience. So each line represents a given level of happiness and so on one line. That level of happiness might be earning 1% return and taking 1% volatility, and it also may be earning 12% return and taking 15% volatility. You'd be equally happy with either of those. The next line up, as they sort of build on each other, are greater levels of happiness, and so on. The next curve up that may make us even happier if we earn 2% return for 1% volatility, or if we earn 14% return for 12% volatility, and so on.
Speaker 1:So if you imagine this is just a theoretical construct we're going to get real in a minute but if you imagine in theory you can draw these curves, then for somebody who's very risk-averse they really don't like to take risk their indifference curves are going to be very steep. They've got to earn a lot of additional return to take an extra dollop of risk. In contrast, an investor that is a risk-taker they're not very risk-averse their indifference curves are going to be rather flat. You would be willing, if you're that person, you'd be willing to take more risk for a more modest increase in return Still an increase, but a more modest increase in return. And so the way that each of these curves are drawn is they represent the combination of return and risk over which the investor would be equally happy, or, to put it another way, over which they would be indifferent. And so that's why we call them indifference curves. You'd be equally happy anywhere along a curve. The steeper the slope of the curve, the more risk-averse the investor is. They don't like to take risk. And the flatter the curve, the less risk-averse. And each successively higher curve represents greater and greater happiness, and so is preferred over the lower curves.
Speaker 1:Alright, so that's step one in Markowitz's mean variance optimization. What's step two? Well, step two is what's the investment universe? What is the entire range of choices of all risky assets that I can invest in? Again, thought of in terms of, on the y-axis, volatility and, on the x-axis, expected return. And here I've reproduced the chart that we looked at way back in the beginning when we talked about finance theory and we talked about the CAPM.
Speaker 1:We talked about this idea that any two assets in combination, if they perfectly co-moved, you could draw as a line between those two points where zero of an asset and 100% of the other asset would give you the return and volatility of the second asset, and 50-50 would give you a point somewhere between the volatility of the first and second, halfway between and somewhere between the return of the first and second, halfway between, and then finally, zero of the second asset. 100% of the first asset gives you the other. But we know that assets don't perfectly co-move. We talked about this in many, many formats and what we found is that not only do they not perfectly co-move, but when they don't perfectly co-move, combining them means that the range of outcomes is not a straight line between those two points, but a curve, and the curves fatter or slimmer depending on how much these assets co-move with each other. And therefore we can build up again a highly theoretical model.
Speaker 1:But we can build up a model in our head that says if I imagine a world in which I can gather together all the risky assets of every single kind and I put them together in a portfolio, then I can imagine that there is a curve, sort of a bullet-shaped curve, which represents all the possible combinations of return and volatility when you have the entire universe of assets to choose from, and all that varies on that curve is how much of each you're using. So the farthest to the right on the bullet has the riskiest assets mostly, and very few of the less risky assets. And as you work your way down the bullet to the point of the bullet. The point of the bullet is where you own the mixture of assets that has the least volatility in combination and then, as the bullet reverses on itself and starts to go down and to the right, those are portfolios that you wouldn't bother owning because they earn lower return for higher volatility and the higher volatility portfolio can just be converted into the one on the top of the bullet. And so we ignore that bottom part of the curve and we just think about the top part of the curve and that's what's called the efficient frontier. And so when we take the efficient frontier and we map that onto indifference curves again all highly theoretical.
Speaker 1:I promise this is going to get real in a minute, but I think it's important to understand the theory. When I map the efficient frontier onto all of the indifference curves, what I find is pretty logical that there are some indifference curves. I would be made a lot happier if I could earn 15% with 2% volatility. I would be way happier. But the efficient frontier tells me that that's just simply not in the realm of the possible. It doesn't add up. You can't get there. You might want it but you can't get it.
Speaker 1:So the indifference curves tell me what would make me happy, the efficient frontier tells me what's possible, and when I map them, there's going to be a portfolio that sits at the tangency. That is where an indifference curve just touches the efficient frontier and where that's the case, that's the portfolio that an investor wants to own. The tangency portfolio is the portfolio that, for a given level of risk aversion, for a given level of utility, this is the portfolio that is the most efficient at accomplishing that level of happiness for the investor. And so in a Markowitz or mean variance optimization, what we're doing is we rely on estimates of expected return, expected volatility and expected correlation in order to choose that tangency portfolio. We're going to choose the portfolio that earns the optimal return for our goals with the minimum amount of volatility.
Speaker 1:Well, that sounds great. How's it work? Well, it's extremely difficult to estimate returns. Sure, they're very predictable over very long periods of time, but in general, in short periods of time, returns are not terribly predictable. And so that's a problem, because in a mean variance model remember, those are the two main things that are the inputs going into how you build the portfolio that plus covariance. And if your mean estimate is wrong, then you own the wrong portfolio, right by definition. If you're trying to maximize return for a level of volatility that's going to be problematic. If it's wrong and it turns out that the way the mean variance model works, it's most sensitive to errors in the estimated return and the sensitivity is on an order of 11x, which is to say, if you have, you know, sort of proportional sensitivity, that would be 1x and then disproportionate sensitivity, 11x means that the outputs, the terminal value of the portfolio, is going to be 11 times in the wrong direction. If your estimate is wrong Now that might be it'll be 11 times bigger. If your estimate is too low, it could be that it's 11 times smaller if your estimate is too high. It's just to get across the point that mean estimates are very, very hard and they have a really disproportionate impact on what the model tells you is the right portfolio.
Speaker 1:Covariance meaning correlation and variance meaning volatility. Also, time vary and in the case of covariance it also switches signs, as we saw earlier in asset classes, that even though stocks and bonds typically, over a long period of time, seem to have an independent relationship, at various sub periods we would see negative correlation and other sub periods positive correlation. So they time vary and they switch signs, so that kind of makes it hard to predict as well. But the effect on the terminal value is less severe if you get the covariance wrong there. The effect on the terminal value is only 4x. And so, yeah, an incorrect assumption about covariance will absolutely mean the model produces the wrong portfolio, but it will be wrong by a factor that's about a third.
Speaker 1:As wrong as, say, return estimate errors and volatility time varies as well. But, as we've talked about before, it tends to mean revert, it tends to cluster, so it makes it a little bit more predictable. And it's not only a little bit more predictable, but the effect on outcomes is only 2x. It's not nearly as important in getting things right. So we know that financial markets are not normal already. We know that higher moments like skew and kurtosis make volatility a poor measure of risk, and so that's really a problem, right, if our mean estimates play this huge outsized role and they're very hard to forecast, and if we know that financial markets are not normal and what we're trying to do is minimize volatility as a measure of risk. But we know that volatility is going to be artificially depressed if we own assets in financial markets where there's kurtosis or where there's skew, kind of makes the output of the model a little suspicious. And likewise we saw with, for example, hedge funds.
Speaker 1:A lot of asset classes also have exposure to nonlinear risk. And when you have exposure to nonlinear risk here, for example convexity and a lot of hedge fund strategies, volatility is a yet poorer measure of risk. And so the mean variance optimizer, while it has a lot going for it in theoretical terms, is liable to produce a portfolio that is going to be the optimal portfolio for high sharp ratios, but it isn't necessarily going to be the optimal portfolio for the future, and evidence of that is the fact that when you do design using mean variance, you design portfolios for different goals. You see weird looking portfolios, and I'll compare them in a minute, these weird portfolios, but for the moment I just want to leave you with the thought that mean variance optimization is complicated, and it's complicated particularly because of these problems in estimating returns and problems with nonnormality and problems with nonlinearity, and, as a consequence, you're going to get a portfolio that you think is the efficient portfolio, but it may not be.
Speaker 1:How would I know that? Well, let's just remember what we read in headlines in 2023, where people were moaning about the death of the 6040 portfolio, and the death of the 6040 portfolio 60% stock, 40% bonds is because in 2023, we saw stocks and bonds move together down. They were not non-correlated that year and you'll recall from earlier episodes that that happens and that happens not infrequently. But in fact, the reason that was so troubling and the reason why people wrote these articles about the death of the 6040 is because all of these models assume that's not going to happen and so the portfolio that you own, that you think is designed to weather the storm and is going to do well when there are down markets and so on, turns out that it does really poorly, makes you sort of doubt. The whole enterprise Mean variance optimization is pretty much exactly what you get when you use any kind of FinTech or AI software to build a portfolio.
Speaker 1:So if you build a portfolio using I don't use these, so I'm not mentioning them because I'm supporting them or saying they're good or bad, just they're out there. But if you use things like Betterment or Robinhood or E-Trade or any of that kind of stuff and they purport to ask you some questions about your goals and your risk tolerance and they say here's your optimal portfolio, or if your company's 401k has a tool that lets you pick what the asset allocation should be, based on your age and some goals. In both of those cases, what they're giving you is a market with mean variance optimized portfolio, and I would tell you, okay, we're going to spend a little time in a minute evaluating them statistically, but I would tell you, okay, but what's wrong with it? Well, what's wrong with it is that whatever estimate they're using for average returns for asset classes is probably going to be wrong for the next 10 years. And it's not only probably going to be wrong, but it will have the biggest impact on your results. It's probably the case that markets will not be normal over the next 10 years, and so maximizing return for a given level of volatility is probably missing the risk that you own, and so it's probably not the optimal portfolio at all. And that if it does include asset classes with nonlinear risks, like hedge funds or others like them, then you are further compounding the problem. And so you want to just. It's not that it's bad, it's just you want to be aware of what someone is showing you and what they're saying. It is and what it does, and to make sure that you understand exactly what it is.
Speaker 1:Well, is there a solution to the problem of the mean variance model? Well, one solution was proposed by two economists who, at the time of their development of this technique, were working at Goldman Sachs, and their names, black and litterman, are now graced with the name of the model, so it's often called black litterman optimization, or sometimes people just call it global equilibrium optimization. All right, so what? What do these guys do? Well, they're tackling the problem that we just talked about. They're tackling the estimate return problem. That's the biggest effector of portfolio success. That's going to get the. That's going to get the investor the most upset. If it's wrong, that's going to have the biggest output on terminal values. So how can we build a portfolio where we get a better estimate of returns?
Speaker 1:And black and litterman start with a cap M world. They start in a world in which capital markets are efficient and investors allocate their wealth to risk assets on the basis of what each of them thinks the expected return and expected volatility of an asset is going to be. So you just think about a world in which all risky assets are available. Every investor is making her own decision about what she thinks the expected return is on that stock or that hedge fund or that private equity fund. She's making her own determination about what she expects the return to be, what she expects of volatility to be. And all of us are doing that simultaneously because it's one giant market for risky assets and in that world, if you look at the current state of market capitalization, like how much of all us investable wealth is in each asset class, that's telling you the markets expectation of return and volatility of all of those asset classes for an average risk aversion, because it's everybody in the market. So some people in the market very risk averse, some people not at all risk averse and everywhere in between. So if we're looking at the capital market weights we're saying, sure, that tells us something about the expected return that the market has for each asset class, but it tells us that expected return for the average level of risk aversion.
Speaker 1:And so what this method does is it kind of works backwards from the market capitalization weights to determine what the implied expected return is for each asset class. So it doesn't look at historic expected returns as you would do in Markowitz, mean variance to predict the future. It looks at historic returns but it adjusts those historic returns for how the market currently owns all risky assets and it comes up with implied expected return. The model is also one where you can add your own personal, forward-looking views of the market. You can modify those estimated returns and volatilities or you can impose constraints based on your own personal view of what you think the market is doing. And then, once you have those return estimates, the implied return estimates, or sometimes they're called the global equilibrium estimated returns.
Speaker 1:Now the Black-Litterman model runs a mean variance optimizer. So it does exactly the same thing as Markowitz mean variance optimizer would do, but instead of using historical returns and historical volatilities and historical correlations to forecast the future, they work from the current market capitalization weights of assets and work backwards to implied expected returns and implied expected volatilities. This process, this Black-Litterman optimization process, is the industry standard if you're working with a wealth management firm. If you're working with a wealth management firm, it is the industry standard by which portfolios are designed, this Black-Litterman optimization. And it tends unlike the I mentioned earlier that the mean variance optimization model is just going to push you to the highest sharp ratios, so it will produce some very odd-looking portfolios.
Speaker 1:This one will not push you towards the highest sharp ratio historically. It's going to push you towards the highest sharp ratio, but constrained by the current allocation of assets in the market, and that means that the portfolio isn't going to look so wonky. It's going to look a little bit more like you expect a diversified portfolio to look. However, it's going to still be true in a Black-Litterman optimization model. It's going to still be true that the model is sensitive to errors in return estimation, or more than any other errors, and it's to the tune of 11X. So if the implied return is wrong, you don't necessarily own the most optimal portfolio. It's also going to be the case that it invests in assets that are non-normal, which means skew and kurtosis artificially depress volatilities, which means that the model is maximizing return for a given level of volatility, but it's mismeasuring the volatility. And it's also going to mean that assets that have non-linear risks, which depress volatility, will itself likewise see that the portfolio is going to push in the direction of those assets with returns with lower volatility, because it assumes the volatility is the correct measure of risk.
Speaker 1:So where mean variance? Gives you quirky-looking asset allocations that are kind of peculiar Black-Litterman. In contrast, gives you allocations that don't look so peculiar but will be much more weighted towards those assets that have depressed volatility for the reasons that we described. I've put some charts in the slides for you to look at so that we could compare a Black-Litterman portfolio and a mean variance optimized portfolio, just to get a flavor of the differences. Again, if you've got it handy you can look at it, that's great, but if not, I'm just going to describe it verbally. So the first slide talks about someone's portfolio where the goal is to be very conservative, have very little volatility 5% volatility, so the very conservative portfolio.
Speaker 1:So what's the difference between the Black-Litterman and the mean variance portfolios? Well, what you find is that the Black-Litterman model has significantly more in T-bills and diversifies bonds with more international bonds. It overweights international bonds to US bonds and it diminishes, but it doesn't eliminate the bias towards value stocks in this model. And so when you just glance at the allocation, you see that the mean variance allocation has only one, two, three, four asset classes and that's mostly investment grade bonds and, to a lesser extent, us small cap growth stocks, us large cap value stocks and US small cap value stocks. The Black-Litterman portfolio has far, far less in investment grade bonds. It mostly has treasury bills, but then it's got an allocation to a lot of different things. It's got some allocation to large cap growth, it's got some allocation to international bonds, it's some allocation to commodities, some allocation to international stocks. It just essentially looks more like what you would expect it to look like.
Speaker 1:What about a moderate portfolio allocation, a moderate risk portfolio allocation, where maybe you're targeting 7.5% volatility? Well, here the Black-Litterman portfolio reduces US bonds further in favor of T-bill exposure. It again overweights international bonds to US bonds and it to some extent diminishes the bias towards value equities and owns a more mixed bag of equity risk. And, last but not least, I have a slide where the goal is 10% and this is an aggressive portfolio, so 10% volatility. And here it overweights international bonds to US bonds. The value bias is a little less peaking, and so we get the point that it is increasing risk is not necessarily creating a wonkier portfolio. It's still creating a portfolio in a Black-Litterman model that looks a little bit more like what we expect, and in some respects, that's really the value of the Black-Litterman model is it gives us what we expect to see.
Speaker 1:But, as I mentioned a moment ago, there's a lot going for these two methods, but anybody who thinks that choosing a strategic asset allocation using them means you have the optimal portfolio or the most efficient portfolio or the portfolio most perfectly tuned to your goals is misunderstanding the problem with these models. All right, let's turn our attention then. So we said that mean variance optimization is pretty bad because historic returns are really poor judges of future risk, because financial markets are not normal, because a lot of asset classes are non-linear, making volatility poor measures of risk, and so we get crazy wonky portfolios. Black-litterman comes along and says we can fix that. We'll use a CAPM approach and we'll estimate what returns in the future should be based on the market's views of asset classes. Well, you get more asset classes there, but it's not clear to me that you're getting those asset classes because those estimates are correct. It's still very hard whether the whole market thinks it's true or not, it's still very hard to predict returns. It's still true that financial markets are non-normal. It's still true that we're gonna have certain asset classes that are non-linear, all of which make volatility tough to measure.
Speaker 1:So what are some alternative ways of going about it? Well, in principle, the main alternative to those two methods is something called risk control. In general, risk control strategies seem the most likely to overcome the vagaries of mean estimation and the large effect that that has on allocation decisions by not relying on means to build a portfolio, but instead relying on the more reliable estimate of volatility and its smaller effect. And so the strategies then seek to build modeled portfolios for a given set of objectives, a given set of volatility, but where all they're doing is trying to build a portfolio that has that level of volatility. Now, the problem, of course, with that is that they're still relying on historical data to decide which asset is going to have which volatility. And a good example of how that can be dangerous is that we know that for the last 30 years in the US there's been a bull market in bonds, that is, bond returns have been very strongly positive and the volatility has been rather muted. I am rather sure that that is not true in 2024, and it will not be true for some years in the future.
Speaker 1:And so if you're building a portfolio based on trying to predict or forecast the volatility, you say, ha, I forget returns. I'm not gonna estimate those, they're too hard to estimate and the effect on outcomes is too significant. I'm just gonna estimate volatilities. That's fine, but it's the case that you might really be building into your model some historical anomalies that aren't really true. It's also the case that the strategies that build based on volatility estimates may actually be loading on risks, even though it doesn't seem like they're loading on risks. They're just loading on different risks, and so you have to ask whether the exposure to those risks is really diversified or not. And then, finally, since all of these methods tend to shoot for low risk tolerance portfolios that's what they're designed to do the allocations are very sensitive to the volatility and correlation estimates. So, even though those errors are, they have fairly constrained effects on outcomes. They still have effects and they're still not easily predictable. It's just they're more easy to predict than return estimates. So we wanna be careful about thinking of it as a kind of complete solution to the problem.
Speaker 1:In 2015, an economist called Hellerbach did a study in which he evaluated risk control portfolios and tried to make some sense, and he created a model. That's very, very basic, very, very simple, and the model is gonna try to evaluate risk control strategies and compare them to things like black littermin, but using only four assets equities, treasuries, investment grade bonds and high yield bonds, and it considers risk as volatility. And then it builds these portfolios and then evaluates the kind of risk you own and what the returns look like over the risk-free rate, et cetera. And so let's talk a little bit about them, because I think it's instructive to see the some of the vagaries in using risk control strategies. I should point out, by the way, that risk control strategy is, the most famous of which is the Ray Dalio All Weather Fund. His strategy there began in its origin and I think continues to this day to be a risk control strategy, and so if you're wondering what that looks like, that's an example of what that looks like.
Speaker 1:So, as a baseline, hellerbach starts by just evaluating a black littermin-style portfolio, says, okay, what's the optimal allocation for a given level to minimize risk? Owning equities, treasuries, investment grade and high yield, where he has a targeted return. He has a targeted return and he wants the minimum volatility and in his model he builds with that target return a portfolio with 53% equities, 29% treasuries, 14% investment grade, 4% high yield. So that sounds like a pretty diversified portfolio and in his case, from 2005 to 2014, it earned a premium over the risk free rate of 6%. So let's call it about 9% return with volatility of about 8%. And so in this case, because they were measuring the premium, not the return, it's the sharp ratio, not the information ratio. So the premium divided by the volatility is the sharp ratio and that's a sharp ratio of 0.67. So pretty good, right, makes a lot of sense and you earn a return that's sort of consistent with most goals. A premium of 6% over the risk free rate is pretty good. It's gonna be consistent with most goals and it's got a relatively high sharp ratio and so that's kind of his baseline.
Speaker 1:This is the classic, the classic Black-Liderman model. But what's he seeing is that equities represents 92% of the risk contribution. That is to say, when you look at these asset classes, it's still the case that equity risk, even that 53% allocation that equity risk is so big that even 53% allocation to equities means that 92% of the portfolio's volatility, 92% of its volatility or its risk, is coming from equities with only 53%. And that's the thing that risk control strategies are trying to avoid. They're trying to avoid the fact that that's that's probably gonna be a pretty aggressive portfolio if you think about it.
Speaker 1:So the next one that he builds is he builds an equal weight portfolio, he says, all right, what if we have no opinion about estimated returns, we have no opinion about estimated volatility. If there are, in his model, four assets, we own 25% of each asset. Equal weight, perfect. Well, what's he find? Well, here he finds that the equal weight portfolio earns slightly less than the market cap weight or the Black-Liderman portfolio. It earns 5% premium instead of 6% premium. It's got 7% volatility instead of 8% volatility and its sharp ratio is a higher 0.79.
Speaker 1:So you kind of think, well, okay, that's, that seems pretty good. What about its risk contribution? Well, 49% of the portfolio's risk is attributable to its equity stake Again, 25% in equities. But that drives 49% of the changes in the value of the portfolio. And what's the risk contribution? High yield has 35% risk contribution and investment grade 16%, and treasuries doesn't contribute to the portfolio's volatility at all. In other words, by construction, 51% of the risks in this portfolio is interest rate risk and 49% is equity risk. So that seems like a pretty sensible. This is just equal weight, right? Nobody has minimized volatility or tried to do anything special or fancy, just 25% in each of these asset categories. And that sounds pretty logical. And what you find is that that's more or less consistent with growth in real terms when you're supporting spending. So, like the Black-Liderman model, pretty sensible, except that you are not taking 92% of the portfolio's risk in equity risk and you're not really making any estimates of return or any estimates of volatility. You're simply owning an equal weight of each asset.
Speaker 1:Then let's compare it to one of the first of two risk control strategies. This is what he's trying to study. The first of the two is something called minimum variance. Here he's gonna build portfolios that minimize the overall volatility of the portfolio by making sure that each asset has an marginal contribution of volatility that's equal. I'll repeat that. So he's gonna build a portfolio that minimizes the overall volatility by making sure that there is equal marginal contribution to volatility. So if you're looking at the slides, that's the one called minimum variance. All right, so what's the outcome? Well, the minimum variance portfolio owns 5% equities, 82% treasuries, 0% investment grade, 13% high yield.
Speaker 1:Remember that model came up with that allocation because you're trying to minimize the overall volatility of the portfolio. Not there's no target return like you would have in the Black-Litterman model. It's just minimize the volatility in the portfolio and that's the mix that you would get. What is the premium is a lower 4%, so now 4% over the risk-free rate. That's about 1% less than, say, the equal-weighted portfolio, but the volatility is 3%, and that's about half of the volatility of the equal-weighted portfolio, so it gives it a very high, sharp ratio of 1.07.
Speaker 1:If we take a look at where the risk comes from, though, 82% of the risk is from interest rate risk and 18% of the risk is from equity risk, and so, in other words, this portfolio is kind of the mirror image of the Black-Literman portfolio, where 92% of the risk of the portfolio in Black-Literman was equity risk, even though it only had some 50% in equities. Here, 82% of the risk is coming from interest rate risk and about 18% is coming from equity. So in that case, it's just the mirror image, but with only a 4% premium over the risk-free rate. So let's say that's about 7%. We're starting to get into a category where it might be hard if the goal is to grow the portfolio in real terms, if you're also taking distributions, and so typically, when strategies like minimum variance are being put into work, they're usually paired with leverage, so you might use leverage in order to amplify the returns, because, while the leverage will amplify the returns. With such a low level of volatility, you're still going to have a portfolio in the end. That makes sense as a practical matter. That's a lot harder to do than it might sound.
Speaker 1:All right, the last of the risk control strategies is something called risk parity, so let's talk about risk parity. Risk parity builds the portfolio not to minimize overall volatility, with equal marginal contribution to volatility. Instead, risk parity builds a portfolio that minimizes volatility with equal total contribution to volatility from each asset class. So, in the case of the minimum variance, we talked about how 82% of the risk came from interest rate risk. In the risk parity portfolio, each asset class will contribute exactly the same proportion to the risk. What does that turn out to be? Well, it means 11% in equities, 55% in treasuries, 19% in investment grade bonds and 14% in high yield bonds, and if you have that allocation, then it means that 25% of the portfolio's volatility comes from equities, 25% comes from treasuries, 25% from investment grade bonds and 25% high yield. Thus risk parity. Unfortunately, though, like the minimum variance portfolio, this, while it delivers a really robust, sharp ratio, there's only a 4% premium on this portfolio over the risk-free rate, and so that means that you might find this as a more complicated portfolio to live with. If you also have to make distributions each year, and or you want to grow in real terms, all right, well, let's put it all together then. What does it all mean In the end? I think what it means is first, don't believe anybody who tells you that this is the optimal portfolio, because it doesn't exist. It's simply a fantasy.
Speaker 1:Each of these methods of building a strategic asset allocation has pros and has cons, and therefore, which one you choose to use will depend a little on your own sensibilities. Mean variance portfolios that is the standard stuff that you would get on an app or something like that. They may not be optimal, but they might be a perfectly good starting point, and then you can sort of adjust it using your intuition. They tend to load on equity risk. We saw that in the Hellerbach study. They tend to pretty much own equity risk, but that might not be such a bad thing. It may arguably be the case that, as we talked about when we talked about equities as an asset class, that equities are the most remunerative. They're also the most risky, so it may not be a bad thing that it loads on equity risk and we also want to make sure we bear in mind that, while mean variance might be a good starting point, diversification is only as good as the additional assets adding expected real returns. So we don't want to have a mean variance portfolio that has everything plus the kitchen sink. If there are asset classes that you don't think will earn positive real returns that are consistent with your goals, you shouldn't want to own them and they shouldn't be in your diversified portfolio. Just because someone says it's diversified, you should not have them in your portfolio because you don't need to own them. You only need to own things that are non-perfectly correlated where they expect to earn positive real returns.
Speaker 1:The second takeaway from all of this is to be modest about what these strategic asset allocations do. If we have a really high confidence in our mean estimates let's say you work with an advisor who says I've studied this for years and this is what I think each asset class is going to earn in the future, and you totally trust that person you think they're always going to get it right. A black littermin optimized portfolio is going to give you the most diversified portfolio and it's pretty easy to execute. So if you really believe that the mean estimates are right, don't mess around with other stuff. That's kind of the way to go, and that means either finding one online or working with one of the established wealth management firms.
Speaker 1:If you have high confidence not in your mean estimates, you're like I don't think anybody knows but you definitely have confidence in estimates of volatility and correlation, then risk parity is probably going to be the most diversified portfolio and relatively easy to execute. But, depending on your goals, you might have to use leverage and that can be rather complicated. Last but not least, if I have no confidence in my mean estimates and I have no confidence in my volatility estimates, it seems to me that an equal weight portfolio is a perfectly good place to start and then tweak it, adjust it, moving things around until you get an expected return that's a little more consistent with your goals, but it's a perfectly good starting point. I'm going to leave it there. When we come back next time, we're going to continue talking about portfolio construction, but we're going to introduce the notion of dynamic allocation, which is to take strategic allocation and adapt it to time. Until then, thanks for listening and I look forward to talking to you then.
Speaker 1:You've been listening to Not Another Investment Podcast hosted by me, edward Finley. You can find research links and charts at NotAnotherInvestmentPodcastcom. And don't forget to follow us on your favorite platform and leave comments. Thanks for listening.
.jpeg)