Total Fund Management Part 5.1: The Earth Is (Not) Flat...

This is Part 5.1 of a seven part series on integrated Total Fund Management brought to you by Mihail Garchev, the former Head of Total Fund Management at BCI and I. Mihail has worked extremely hard on Episode 5 but since he covers a lot of material, and since it is the nexus of the series, we decided to break it down into three parts (5.1 today and 5.2 and 5.3 next week).

Please take the time to read Mihail's synopsis below on case studies followed by my comments and a clip where he delves deeply into today's topic (added emphasis is mine):

For those of you who are watching the series for the first time, this is a quick look at the topics we are covering in this seven-episode series. In the first three episodes, we talked about why the Canadian pension funds are increasingly turning their focus toward Total Fund Management ("TFM"). At the core of this emerging trend is the desire further to increase the Canadian pension model's efficiency and effectiveness and sustain the benefits of the previous economies of scale by introducing economies of scope – a top-down efficiency based on an integrated TFM approach.

This TFM approach reflects the need for cohesive management of the total portfolio activities, including short- and medium-term capital and risk allocation, and other total portfolio activities such as liquidity, leverage, currency, and balance sheet management. The TFM approach is also supported by the organizations' strategic plans and proper organizational structure.

We also discussed some fundamental investment beliefs which support the long-term investment thesis of TFM by uncovering the link between managing the short term toward the long-term pension sustainability outcomes. These investment beliefs necessitate rethinking the portfolio management framework and process toward developing an outcome-oriented portfolio management approach. We discussed the key elements of this approach and its implementation in Episode 4.

We are already at Episode 5, where we connect theory and practice and illustrate the TFM process and framework using 11 real-life case studies.

Next, Episode 6 will get even more real and discuss what is required to move from a process and a framework to a capability. The investment process, methodology, process, and framework meet technology and organizational structure. We will conclude the series with Episode 7, where we will discuss the role of TFM as part of the next stage of development of the Canadian pension model. With this episode, we already have almost 500 slides and close to six hours of presentation material.

Key takeaways from Episode 4 last week

Before we proceed with the topics today, let us summarize the key takeaways from Episode 4, which provided the TFM framework and process.The case studies today are based on all the theory in Episode 4 and the conclusions from the previous episodes.

First, we confirmed the difference between TFM and other approaches, such as strategic asset allocation ("SAA"), tactical asset allocation ("TAA"), and broad diversification approaches, such as "All-Weather." A significant differentiation is the integrated management of the short, medium and long terms, which requires a multi-period path-dependent asset allocation. As such, TFM critically relies on the ability to formulate expected returns, different time horizons, or what we called the term structure of expected returns ("TSER"). At its core, the TFM process is about the ability to formulate absolute and relative expected returns at different time horizons as part of a consistent and coherent macro, market, and thematic framework. These expected returns then become a critical element for a multitude of total portfolio decisions, such as rebalancing, strategic tilting, risk mitigation, leverage, and liquidity management.

The ability to evaluate the total portfolio exposures, given the expected returns and the market conditions, and diagnose the portfolio allows for making essential total fund ("TF") decisions. TF decision-making, however, requires a structured and disciplined, data-driven process.

Episode 5 case studies

Episode 5 brings together the framework, process and conclusions from the episodes so far, and real-life examples of how to use the theory of TFM for practical TFM decisions at the investment committee ("IC") or by the chief investment officers ("CIOs").

Thought-leadership, frameworks, processes, they all mean little if the organization cannot implement them. As such, this episode is essential because it is the nexus of thought-leadership and implementation. It bridges what we learned about TFM with what we could do with TFM. This is a needed step of linking theory with practice and a prerequisite for moving toward a functional capability (combining methodology, process, technology, and organizational resources).

This episode is very important for me, as it is not only about the value I can personally bring to the community as a thought-leader but also based on my experience of successfully implementing many of the TFM concepts in practice at two of the large Canadian pension funds over the last 10-15 years.

This brings us to the list of case studies we are going to look at in Episode 5. We have a total of 11 case studies, which we will follow through in three installments of Episode 5 starting today and two more parts next week. The wealth of material is inspired by the many professional experiences I have had and the number of comments, questions and suggestions I received throughout this series (thank you for that!).

Below is a quick snapshot of all the case studies in Episode 5:

Case Study 1—The Earth is (not) flat: The Term Structure of Expected Returns over time

Our first case study is a carryover from the discussion we had in the last episode about the difference between TFM and SAA.

One of the differences between TFM and SAA is the TSER. In some of the slides in the previous episode, we were visualizing this difference as the SAA flat expected returns curve (hence, the title of the episode and the case study, with a slight wink to the irrationality of the flat earth theory), which reflects the assumption of average returns over the investment horizon. By contrast, we depicted the TSER for TFM similar to the stylized representation of the typical yield curve in the fixed income markets but representing expected return over the maturity horizons.

In reality, the SAA expected returns are not a flat line but slowly move in a step-wise or moving average fashion while the TFM TSER fluctuates around the SAA line.

Using data since 1980, we built an example of the S&P500 TSER for every quarter since 1980 and compared these TSERs to the typical average expected return assumed for S&P500 in SAA studies. The figure below illustrates the study's set-up (I like this version but did it after hours, so it is not in the presentation).

Any point on the blue line on the chart above represents the average of each of the expected returns at different maturities on the TSER estimated at that particular time. These are all real calculations, by the way, with no look-ahead bias. For the SAA, the equivalent is just a flat term structure—just one average expected return applied to all the maturities.

This means that if there is predictive informational content in the TSER that could be used in a portfolio context, then using the blue line for portfolio construction would be superior just to use the average SAA approach. This predictive informational content needs not to be a crystal ball. It is more about tipping the odds by capturing direction, possibly magnitude, and, more importantly, relative returns between assets. Remember, in the figure above, this is just one asset. In the case of one asset, there is only the time-series predictive informational content because it is about the absolute expected returns and their term structure. There is no other asset to compare with. In the case of two or more assets, we now have the time-series for each asset on their own and the cross-sectional (relative) aspect. At each point in time, we can also express the relative expected returns between each of these assets. These points would become essential for Case Study 3, where we will look into whether there is sufficient informational content in the TSER.

You have probably noticed by now that the TSERs depicted in the figure above are not uniform—they have different levels, slopes, curvatures, and twists. This is because different maturities might have higher or lower expected returns at any point in time. The usual depiction of a yield curve is upward sloping, which assumes that the shorter-term returns are lower than the longer-term returns. And while for yield curves, this is generally true, this may not be the case for other assets. As you can see, these curves vary a lot. It is not by coincidence that at market peaks, the TSER inverts, as you can see in 1982, 1994, and 2008, for example.

As such, the second part of Case Study 1 is to look at how these TSERs move over time. This example is captured in the figure below. You can see these TSERs shift significantly across time. This shifting across time provides the informational content for both the absolute (time-series) and relative (cross-sectional) use of the TSER.

Case Study 2—How insights are born: Opening the "white box" of how Expected Returns are born

We saw the expected returns in action, but what is the process behind determining the TSER? This is the topic of Case Study 2.

In Episode 4, we discussed the TFM framework, which is based on the ability to construct absolute and relative expected returns at different horizons based on macro, market and thematic insights. We also illustrated the additional use of risk matrices, which define short- and medium-term market and economic conditions.

The question then is, what is the general process to arrive at the current structure of expected returns? Is it some black box with esoteric tools inside, or is it a tried and time-tested approach? The former proposition creates the excitement of the mysterious, while the latter, the skepticism that everybody must have done this already. In the presentation, we bring the point that the value is in the amount and type of information (both traditional and new sources of data) and how this information is interpreted and transformed, in addition to applying this information to a sufficient breadth of asset classes (not only, types, but geographies, sectors, capitalization, capital structure, and factor exposures).

The second lever is the existence of a process itself, not just any process, but an integrated, structured and disciplined one. Indeed, many of the approaches (minus the esoteric machine learning and AI) are already used. Still, in many cases, especially at buy-side institutional investors, these approaches are used here and there in various portfolios, maybe asset classes, but rarely, if at all in at the total portfolio level, let alone in an integrated, structured and disciplined manner. Have such a process and framework first, be confident that it possesses informational content that could be transformed into economically meaningful excess returns, even if it is not that spectacular.

Why? Because you have the economies of scope on your side. It is one thing to have a single strategy and try to extract value from it. It is a different circumstance where you have the multiplicative network effect of TFM impacting asset allocation directly and portfolio maintenance (e.g. rebalancing, FX hedging) and other activities such as liquidity, leverage, or risk management, for example. Because it is not only the multiplicative effect of the excess returns but also the opposite effect on the hidden costs that come with the lack of an integrated top-down process, a topic we discussed at length in Episode 1 where we talked about TFM and the economies of scope bringing second-order efficiencies to the restore the initial economies of scale which might be challenged at the Canadian pension funds.

In this synopsis, I will not go through describing the process of arriving at the expected returns and the TSER, or in other words, opening the proverbial "white box." You have to watch the presentation. The other reason why you need to watch the presentation is that you would see there for the first time a real-life implementation of the expected returns, the term structure, and the short- and medium-term risk matrices across more than 240 markets, and provides further insights of how the matrices are constructed and how to interpret them. This information is the basis for understanding what comes in the case studies next week.

What is essential, however, is to illustrate once again what is the TFM core capability. If you do it right, it will allow you to do many other things more efficiently. The figure below illustrates it. If you have watched all the episodes by now, these four elements would be (painfully) familiar.

The last point in Case Study 2 brings together the notion of the economies of scope and the core TFM capability above in the context of none other than Amazon.

Amazon did one (or two) things right (technology, and client outcome focus, and a competitive internal marketplace), which allowed it to unleash the economies of scope on the world (for better or worse). Put a placeholder on the "competitive internal marketplace – we will revisit this concept when discussing the Canadian pension model 2.0.

In Amazon's case, you can think of it as "efficiencies formed by variety, not volume" (the latter concept being "economies of scale"). Just replace Amazon's "core capability" with the TFM ones, and who knows, maybe defence contracting is not out of reach! P.S. Notice the warehouse setting and the real books – challenging to imagine the amount of progress!

Case Study 3— The proverbial "R-Squared" question: What is the Expected Returns value?

So far, we saw the TSER in action. We opened up the "white box" of how the insights about the expected returns are born and discussed their importance as the core capability of TFM.

The next one is the proverbial "R-squared" question: What is the value of the expected returns? We can formulate expected returns, we can show them, but what is the value of these expected returns to make TF decisions? I will bring again the discussion from Case Study 1 where we established that "if there is a predictive informational content in the TSER which could be used in a portfolio context, then using the blue line for portfolio construction would be superior just to use the average SAA approach. This predictive informational content needs not to be a crystal ball. It is more about tipping the odds by capturing direction, possibly magnitude, and more importantly, relative returns between assets."

In this case study, we are trying to answer whether there is any informational content in the expected returns and whether we can monetize this informational content. To test this, one needs to use a portfolio approach. If using the information in the expected returns, we can form profitable portfolios, with minimum optimization or back-fitting), then this means that we could broadly use these expected returns in various investment-related processes. Of course, the processes themselves need to be well designed as well.

Therefore, in Case Study 3, we evaluate four tests. I will briefly describe them and provide you with the key takeaways. You would need to watch the presentation for all the details and the (many) exciting aspects.

The first test is based on the information in the first case study where we constructed a TSER for each quarter since 1980 and observed how the maturity points on the curve frequently crossed each other over time (this is the second figure in this post). If one has this information, then one can perform a (rather blunt) test of "freezing" each TSER at a particular point in time and observing whether, or how well, the 1,2,3, …9, 10-year expected returns on this particular TSER forecast the subsequent actual 1,2,3, …9, 10-year returns. We did precisely this. This test answers the proverbial "R-squared" question as it shows an r-squared statistics of 0.5 to 0.9 (starting at the 2-year through the 10-year point).

These results are indeed very reassuring, but an extra step needs to be taken and perform portfolio-based tests. You would recall again from Case Study 1, the discussion about the absolute (time-series) and relative (cross-sectional) expected returns. Without going into details (watch the presentation!), we tested two long-short portfolios (and other long-only versions as well) on 43 equity markets (23 developed and 20 emerging) using the underlying signals for the TSER. In these two tests, we did not separate the signals by the predictive horizon (e.g. into short- or medium-term ones). We did this in a separate test, again, with simple portfolios of "long high rating, short low rating" set up across short- and medium-term horizon.

We evaluated all these tests based on the following criteria:

• Performance: Economically meaningful excess returns
• Synergy: Would combining S-T and M-T ratings work better?
• Signal Decay: Optimal horizon and informational advantage decay
• Uniqueness: How different is this from other known strategies
• Impact of Constraints: Long-Short vs. Long-Only Tilted Performance
• Risk and Efficiency: Sharpe, Sortino, Information Ratio, Drawdowns, Turnover, Signal Attrition
• Crisis and Event Performance: 14 significant events studied

Following the battery of portfolio-based tests, it appears that the signals from the macro/market/thematic framework that are used to determine the TSER show informational content and the ability to transform this informational content into economically meaningful outperformance on an absolute and risk-adjusted basis for investment management purposes.

Testing against staple factor strategies shows that these signals do not represent any form of factor investing in disguise (AQR Value, Momentum, and Low Vol explain 1%-10% of the performance).

The outcomes are most similar to global macro sources of return (the HFRI Global Macro Hedge Fund Index explains 50%-60% and underperforms) with a slightly negative beta to the market (-1% to -10%). This is rather important because this is an expensive capability to have (can be accessed only via hedge funds, of course, after paying the fees) and cannot be cheaply replicated or purchased like the factor products, for example.

Second, in the context of risk mitigation (we will talk more about it in Case Study 11), there is a need to substitute somehow or enhance the weakened diversification properties of bonds due to low yields, the prevalence of fiscal over the monetary policy, and extreme valuations (among other things). For this reason, there is an emerging trend of investors to explore global macro processes and products as imperfect and expensive substitutes of bonds. If there were an internal capability, it might still be imperfect, but certainly cheaper and much nimbler.

Testing against control universes (46 Global macro and TAA live funds from Morningstar and 37 popular "paper" strategies published in the Journal of Portfolio Management and similar industry publications) show consistent close to first quartile performance.

The time-series and cross-sectional (or absolute and relative) return profiles compared are quite different, implying that the source of return from cross-sectional investing is different from that of the time-series return, and both absolute and relative expected return signals add value.

Using both short- and medium-term signals combined, especially if directionally aligned, leads to superior results. This confirms the importance of using short- and medium-term risk matrices (market and macro conditions in the investment process).

These signals work in various market conditions but are best at identifying vulnerable markets. They benefit most from short positions and, therefore, are most efficient in an overlay set-up. The long-short strategies have definitive risk-mitigation characteristics. This is important if to further use in dedicated risk-mitigation processes.

Finally, the TSER signals allow for identifying "priced-in" markets – test results show that the best/worst performance is not at the extremes, but just before that.

As such, these tests provide additional confirmation that the signals from the macro/market/thematic framework that are used to determine the TSER could be further used for a multitude of total portfolio decisions, such as rebalancing, strategic tilting, risk mitigation, leverage, and liquidity management.

Case Study 4— The Path of Returns Decisions: The quickest road to Rome

In the Episode 4 summary at the beginning of this post, we confirmed the difference between TFM and other approaches, such as SAA, TAA and broad diversification approaches, such as "All-Weather." In addition to the TSER aspect, a critical differentiation was discussed so far, the integrated management of the short, medium and long terms, which requires a multi-period path-dependent asset allocation. This path-dependence was demonstrated in Case Study 1 (the ever-changing TSER through time). This means that asset allocation decisions depend not only on what we know about (or what we think about) today versus tomorrow, or about the future "on average" (like SAA), but what we know with certain informational content about all the different periods until we reach the final objective in the future. This gets back to the fundamental investment belief of TFM we formulated in Episode 3 that the long-term is a series of short terms in the presence of liabilities.

Again, in Episode 3, we started this analogy with Rome and the daily commute dilemma, and we introduced the notion that what matters for long-term investing is not about long or short, but it is about fast and safe. As such, it is not about the distance, but the fastest way to arrive at the final destination. And this was similar to the daily commute dilemma—although the distance could be shorter, it might take a long time because the traffic is congested (and the traffic being the equivalence of low expected returns, so it would take you much longer to get to the final objective). Time is a function of distance and speed, and speed itself depends on gravity and friction. And you can think of gravity as the expected return and friction transaction cost. We hinted about all this in the physics experiment of competing sliding objects in Episode 3.

The question then becomes the shortest time to arrive at the end goal of terminal wealth (the proverbial "Rome") given the structure of expected returns.

The purpose of the case study is to illustrate if one has the term structure for two assets, equities and bonds, how to think about this allocation decision borrowing from the computer and mathematical programming.

It is a simple example but requires a bit of focused attention to get the idea. You would need to watch the presentation for the detailed explanation. However, the next figure provides you with the (simple) answer – follow the yellow path, and invest today in equities to get faster to Rome. If we only follow what we know about tomorrow, but not about the day after tomorrow and thereon, we would have taken a different and suboptimal path. But more about it in the presentation.

In closing, these are the key takeaways from today's episode:

•  The core capability of the TFM process related to the Term Structure of Expected Returns ("TSER") is central to managing and making decisions on most of the Total Fund processes (e.g. asset allocation, efficient portfolio maintenance, as well as inform decisions around the balance sheet, liquidity, and leverage, among others) 
• Many of these decisions are critically dependent on such a core capability, as decisions, but also as sound risk management, and creating effectiveness and efficiency by economies of scope 
• The core capability requires a data-driven, structured, disciplined, and transparent process
• Such a process further leads to practical informational content in the TSER and the related conclusions about the macro and market environment to enhance the outcomes of the investment process 

This concludes Episode 5 for today. The next part of Episode 5 next Thursday will continue with the case studies. We will talk about Case Studies 5 and 6—Rebalancing Decisions: The Nexus of Multitude of TFM Decisions and Balance Sheet Decisions: Ensuring optimal adequate resources for current operations and financing future growth

***  

Let me begin by thanking Mihail for delivering another great comment and staying up all night to finish the video clip below (next time, get some sleep first!).

If you have not done so yet, please review all three previous parts of this series:

• Introduction to Integrated Total Fund Management: Part 1
• Total Fund Management Part 2: What Nobody Told You About Long-Term Investing 
• Total Fund Management Part 3: When All Roads Lead to Rome
• Total Fund Management Part 4: How to Develop a TFM Framework & Process

As stated at the top, we decided to break Episode 5 into three parts because there's a lot to cover and it's the meat and potatoes of our seven part series, where the rubber meets the road.

As Mihail states, "thought leadership, frameworks, process all mean little if one cannot implement" and this is why this episode is very important, "it's the nexus of thought leadership and implementation, it bridges and connects what we learned about TFM with what we can do with TFM".

And keep in mind, Mihail has great experience implementing TFM and as you'll see below, he shares quite a lot (perhaps too much).

I cannot add much to the analysis below since some of the concepts are way above my head, but he goes over everything in detail and it is extremely well explained. 

Clearly, the Term Structure of Expected Returns ("TSER") is central to managing and making decisions on most of the Total Fund processes (e.g. asset allocation, efficient portfolio maintenance, as well as inform decisions around the balance sheet, liquidity, and leverage, among others) .

The TSER should not be confused with the flat expected return governing strategic asset allocation (SAA), it is more meaningful and powerful and when used properly, it can no tonly unlock economies of scope, it can significantly enhance risk management at a total fund level. 

Anyway, please take the time to watch Episode 5 below, Mihail did a great job explaining the concepts above.

I recommend you watch it on YouTube here and click "show more" where you will see "chapters and timeline" to fast forward to different sections of the presentation:

Below, Episode 5 of the seven episode series "Introduction to Integrated Total Fund Management" presented to you by Mihail Garchev, former VP and Head of Total Fund Management of BCI. 

Great job, thank you Mihail!