AgPa #4: Measuring the World’s Assets (2/2)

Historical Returns of the Market Portfolio (2019)
Ronald Doeswijk, Trevin Lam, Laurens Swinkels
The Review of Asset Pricing Studies 10(3), 521-567, URL

The Global Multi-Asset Market Portfolio, 1959–2012 (2014)
Ronald Doeswijk, Trevin Lam, Laurens Swinkels
Financial Analysts Journal 70(2), 26-41, URL

As I promised in the last post, this week’s AGNOSTIC Papers also examine the global market portfolio. Again, there are two papers on the topic (see boxes above). The first one is mainly about the composition of the market portfolio. In the second one, the authors expand the analysis and also present return time series. I mainly focus on the more comprehensive 2019-paper, but still want to give you the full reference.

This post is designed to be a sequel of last week. Therefore, I skip the part on the Setup and Idea of the Paper. The objective remains to develop an estimate for the global market portfolio. That is the collection of all investment opportunities with each asset weighted by its market value. Simple in theory but very hard to implement in practice. In last week’s post, I provide all the details why the market portfolio is such an important concept. So I recommend to read that one first.

I will refer to last week’s papers as GSV (Gadzinski, Schuller, Vacchino). Similarly, DLS (Doeswijk, Lam, Swinkels) or just “the authors” stands for this week’s papers. I hope this makes the post more readable.

Everything that follows is only my summary of the original paper. So unless indicated otherwise, all tables and charts belong to the authors of the paper and I am just quoting them. The authors deserve full credit for creating this material, so please always cite the original source.

Data and Methodology

Collecting data on all of the world’s assets is very hard and subject to many assumptions. As a consequence, two people estimating the same market portfolio will almost certainly get two different results.1That sounds unfortunate, but this is nothing new for empirical social sciences. GSV and DLS are no exception from this.

DLS focus exclusively on invested global assets that are traded on financial markets. For example, they mention that the estimated value of global real estate was about twice the value of their entire market portfolio in 2015. But since most of it is not traded, they do not include it. In contrast, GSV try to estimate the aggregate stock of capital in the world economy. Regardless if it is traded on public markets or not.

Those different perspectives lead to very different results, but both deserve their spot in the literature. They are just different and depending what you are interested in, one is more relevant than the other. Admittedly, however, the approach of GSV captures more assets and is therefore closer to the theoretical ideal. But before we get into a deeper comparison, let’s see how DLS construct their proxy for the market portfolio.


First, the authors divide the world into the following 5 asset categories, which are again subdivided into 10 asset classes. Notably, the authors are skeptical about the data quality of appraisal-based prices.2In simple terms, appraisal-based valuation means you higher someone who tells you how much your asset is worth. It is often criticized that appraisers are somewhat biased and not completely objective. As a consequence, they only include listed private equity and real estate. In absolute terms, their market portfolio is therefore considerably smaller than GSV’s.

  • Equities Broad (Stock Markets, Private Equity)
  • Real Estate
  • Non-Government Bonds (Investment-Grade, High Yield, Leveraged Loans)
  • Government Bonds (Standard, Inflation-Linked, Emerging Market Debt)
  • Commodities

In the second step, the authors approximate the return and market value of each asset class with investable market indices. To construct their sample, they use well-known data vendors like MSCI or Bloomberg, but also results of prior research. The final data set covers the period from 1960 to 2017. This is substantially longer than the data of GSV which only ranges from 2005 to 2020. All else equal, more data and a longer history is always better. So this is clearly an advantage of DLS’s paper.

Like I also mentioned last week, such a patchwork of data sources almost certainly suffers from some kind of errors and inconsistencies. However, the authors address a lot of potential issues in great detail and explain all of their steps logically. They also relate and stress-test their data against prior research. Similar to last week, this is no criticism. Data on all of the world’s assets is not readily available and I appreciate the authors’ effort to collect it.

The authors also mention the problem with double-counting. They exclude all types of investment funds because such vehicles are ultimately holding the underlying asset classes. However, they do not address double-counting in the sense of corporation-owned assets. For example, including commodities probably overestimates the value of the market portfolio because some of them belong to companies. At least to some extent, their value should be already reflected in public equity markets. GSV explain this issue in great detail and exclude commodities.

In my opinion, this is a matter of perspective. GSV estimate the aggregate stock of capital in the world. That is the total value of all unique assets. For this purpose it is necessary to exclude commodities.3Ideally, one would distinguish between corporation- and non-corporation-owned commodities. However, the data for that is hardly available in practice. If I own all companies and all outstanding debt, I ultimately end up owning most of the world’s commodities indirectly.

In contrast, DLS take the perspective of an investor who can choose from all traded assets to build a portfolio. For such an investor, it is of course possible to invest in both commodity companies (e.g. gold miners) and physical commodities (e.g. gold). Therefore, it is logical that DLS include them. As you see, the global market portfolio is a tough empirical challenge. There are different approaches that lead to different results. But as long as the methodology is logically justified, the results deserve their spot and we cannot discard them.

Important Results and Takeaways

Although they are different, the market portfolio estimates of DLS and GSV also share some similarities and I try to compare the two as good as possible. However, I will not include the charts of last week again. This post is devoted to the papers of DLS but I encourage everyone to open the other post in split view to see the differences.

Value and Composition of the Market Portfolio

In Table 5 of their paper, DLS provide detailed annual statistics on their proxy for the market portfolio. Specifically, they list the total value in trillion USD and the weight for each of their five asset categories. I show these numbers for the period from 2005 to 2017 in the chart below. Given that the sample period from GSV ranges from 2005 to 2019, this is the largest possible overlap.

Own illustration based on Table 5 of Doeswijk et al. (2019). Values in trillions of USD.

According to DLS’s market portfolio, global assets were worth about $57T in 2005 and grew to more than $126T by 2017. In each year, the Equities Broad category (stock markets and listed private equity) accounts for the largest individual share. Despite the considerable drawdown during the financial crisis in 2008, the value of Equities Broad almost doubled from $30T in 2005 to about $57T in 2017. For the other asset categories, the pattern is very similar or even stronger. The value of Real Estate, Government- and Non-Government Bonds more than doubled between 2005 and 2017. All of this is consistent with the last decade of low interest rates and steadily rising asset prices.

As mentioned before, this market portfolio is way smaller than the proxy from GSV. For example, GSV estimate the value of the global market portfolio to be around $597T in 2017. This is five times the estimate of DLS. What are the reasons for this? The first and biggest part are unlisted assets. About $352T of the $600T are illiquid assets like Land, Real Estate, Private Equity, and Non-Securitized Loans. DLS only consider the listed portion of those asset classes, which is rather small. For example, their value for Real Estate is only $7T vs. $118T for GSV. In addition to that, GSV also include more asset classes. For example, Cash and Money Market instruments made up almost $50T of their market portfolio in 2017.

What is interesting though, the proxies even disagree for “easy” asset classes like equities. For instance, DLS estimate the global value of stock markets and listed private equity to be around $57T in 2017. This is considerably smaller than the $85T presented by GSV for stock markets alone. And there are similar deviations for the value of Government and Non-Government bonds. So who is right and who is wrong? Unfortunately, I think it is impossible to answer this question without examining the data sources in detail.

I know this is unsatisfying, so let me try to be a bit more specific. As I read it, the methodology of the two authors differs at a particular point. GSV first estimate the aggregate value of each asset class and rely on a variety of financial and macroeconomic data to do this. Only subsequently, the select investable indices and funds to approximate the return of their market portfolio. In contrast, DLS already start with investable indices. To estimate the value of each asset class, they just use the aggregate market capitalization of the respective indices.4See Appendix A of the 2014-paper for a detailed explanation. For this reason, I believe that GSV’s estimate for the total value of the market portfolio is better and closer to reality.

Own illustration based on Table 5 of Doeswijk et al. (2019).

Of course, the different methodology also affects the composition of the two market portfolios. Unsurprisingly, the weight of equities and bonds are way larger in the proxy of DLS. Due to the different asset classes, I think it is hard to compare the numbers beyond that. So I leave you with the charts to draw your own conclusions.

Returns of the Market Portfolio

Unfortunately, DLS do not provide a monthly return chart of their market portfolio that is comparable to those of GSV.5In Figure 2 of the 2019-paper, they provide a long-term chart of real returns for the market portfolio and each asset category. However, this is hardly comparable to the nominal returns from 2005 to 2020 that are presented by GSV. However, they offer a lot of statistics in the following table.

From 1960 to 2017, their global market portfolio (GMP) generated an average annualized return of about 8.36% per year. This was about 3.39 percentage points more than the risk-free return over the same period of time. After inflation, the real return was about 4.45% per year. The annualized volatility (standard deviation) was about 11.2%, leading to a Sharpe-Ratio of 0.36. Before we continue, I just want to mention that, in my opinion, those are great numbers. The annual return of 8.36% turned $1 into about $97 over the 57 year time period. That is a remarkable creation of wealth and we should celebrate capitalism for that!

Unsurprisingly, the more risky Equities Broad (EB)– and Real Estate (RE) category generated even higher annualized returns (9.76% and 10.45%, respectively). However, the standard deviations of returns are also higher. The risk-adjusted performance (Sharpe-Ratios) were therefore almost identical to the global market portfolio (0.37 for both).

These are of course long-term numbers that are not comparable to the return series from 2005 to 2020 by GSV. To somehow compare the two, I calculate the annualized return for DLS’s market portfolio from the aggregate value (first chart above). The growth from $57.2T in 2005 to $126.5T by 2017 is equivalent to an annualized return of 6.3%. This number is higher than the 4.36% for the market portfolio of GSV.6According to their estimates, global assets grew from $343T in 2005 to $597T by 2017. This is a compound annualized return of 4.36%. Again, without having the data it is hard to find the precise reasons for this difference.

But for me, the results are still plausible because the larger market portfolio of GSV contains more low-yielding assets like cash, money market instruments and loans. In addition to that, the share of equities (which performed really well in this period) is lower. Unfortunately, I can’t do the same comparison for the risk metrics because I just don’t have the data.7Calculating standard deviations from 12 annual returns is absolutely not meaningful. However, I suspect that the larger market portfolio of GSV is less volatile because of more multi-asset diversification.

Conclusions and Further Ideas

What is the market portfolio? Who is right?

For a moment, I think it makes sense to go back to the very beginning. In theory, the market portfolio is the collection of all investment opportunities with each asset weighted by its market value. The closer we get to this theoretical idea the better the empirical estimate. From this perspective, the estimate of GSV is clearly better. It includes more assets and also emphasizes the important role of illiquid assets like private equity or real estate. So this is a point for GSV.

But all four papers are based on transparent empirical methodology and high-quality data. In addition to that, all of them are published in peer-reviewed journals. So there should be some truth in all of them. To get an overview, I summarize the key differences in the table below.

Gadzinski et al. (2018, 2021)Doeswijk et al. (2014, 2019)
Sample Period2005 – 2020/Q11959 – 2017
FocusMacroeconomic Perspective
Aggregate Stock of Capital
Investor Perspective
Aggregate Market Capitalization
Market Portfolio Value
2005
2017
2019

$343 trillion
$597 trillion
$667 trillion

$57.2 trillion
$126.5 trillion
not covered
Market Portfolio Composition
Equities
Bonds
Loans
Real Estate & Land
Private Equity
Commodities
Cash

14%
16%
18%
24%
20%
0%
8%

36%
55%
0%
5%
4%
0%
0%
Market Portfolio Return p.a.
2005 – 2017

4.36%

6.30%
Own illustration of data by Doeswijk et al. (2014, 2019) and Gadzinski et al. (2018, 2021). The “Composition” row is taken from Exhibit 6 of Gadzinski et al. (2021).

Unfortunately, there is no right or wrong here. The global market portfolio is a though empirical challenge and data on the world’s assets is very noisy. So it remains a matter of perspective. If you are interested what assets are truly existing in the world, you should go for the estimate of GSV. If you are more interested in the total capitalization of financial markets, the proxy of DLS appears to be the better choice.

Personally, and that’s truly just my opinion, I would go for the estimate of GSV. Since it captures more asset classes and in particular illiquid ones like private equity and real estate, it is just closer to the theoretical idea of the market portfolio. In addition to that, the authors also provide a list of 87 ETFs to approximate the return of their market portfolio and to make it investable. To me, the availability of this investable proxy debunks the arguments of DLS to include only listed assets.

What I like about the papers of DLS is the long history. It is clearly valuable to see how a reasonably good proxy for the market portfolio performed over long periods of time. That’s clearly a shortcoming of GSV as they cover only the years from 2005 to 2020. As we all know, this were very special years for asset prices and essentially the longest bull market ever.

In the end, the global market portfolio remains a hard problem. And as always, there are no easy solutions for hard problems. Therefore, I think all four papers deserve their spot in the literature and enhance our understanding of the market portfolio. I don’t want to fool anybody. I also hate the fact that there is no clear answer and that we can’t just buy the damn market portfolio. But that’s just not the way empirical finance works.

What do we know about the market portfolio?

Before it gets too depressing, let me try to conclude this little series on the market portfolio more positively.8There are of course more papers out there. But to the best of my knowledge, those four represent the most recent and most comprehensive work on the topic. Perspective and methodology matters a lot, but there are some robust insights.

Most assets are not publicly traded

Both, DLS and GSV agree on this fact and it is also supported by many studies before. Most of the world’s assets are held privately and the ordinary investor has no access to them. As of 2019, GSV estimate that more than 60% of the global market portfolio consist of (mostly) unlisted assets such as real estate, private equity, and unsecuritized loans. That is indeed unfair, but it is just the way the industry currently operates.

To get access to those so called alternative investments, you need a lot of money and some kind of institutional setup.9For example, most alternative asset managers require minimum investments of $25M or even more. There are some start-ups which try to change this and to democratize alternative investments. While I appreciate the effort, I think it will take more time until we see meaningful change. In the meantime and likely even beyond, institutional investors probably still have exclusive access to the best alternative investments out there.

Stock markets are relatively small

Although they receive a lot of attention from the media and the public, stock markets are actually a relatively small asset class. According to the estimate of GSV, stock markets were only 14% of the market portfolio in 2019. These are “just” $95T of the total $667T.

This has some important consequences. First, investing is never only about stock markets. For example, you can also buy real estate or just hold cash. For institutional investors, this is nothing new but some individual investors often forget this.

Second, the stock market is not the economy. In fact, global stock markets account for less than half of the world‘s total business value.10According to the estimate of GSV. In addition to that, public companies are a curated sample of mostly large and successful businesses. This is not an adequate proxy for the private sector of the economy. The composition of the market portfolio reflects this fact.

Recent historical returns were 4.4% to 6.3% per year

Different papers present different results, but most estimates for the return of the market portfolio are somewhere around “mid- to upper-single-digit” returns per year.11This is the analyst-jargon for “I think the number is somewhere between 5 and 10% but I don‘t want to give you a specific answer that you can pin me down on.” I don‘t want to do macro forecasts but I think something in the ballpark of 5% is a reasonable number. Of course, this is just the aggregate average and the number is different for each asset class. For example, stocks should continue to outperform government bonds because they are more risky.

DLS estimate a long-term (1960-2017) average annual return of 8.36% for their investable market portfolio. Given that virtually all assets became much more valuable over the last years, this figure is probably too ambitious for the future. For instance, AQR Capital Management estimates that the expected return for the famous 60/40 portfolio currently stands at less than half of its historical average.12AQR is an asset management firm that is famous for its close relation to financial research. Check this website for the study on expected returns. Again, a rather depressing conclusion but we can’t change the environment we live in and need to make the best of it.

The market portfolio is a good starting point

This is probably the most important one. As I show in the example of last week’s post, the market portfolio is the theoretical base of passive investing. It represents the average portfolio weights of all active investors and therefore captures a lot of knowledge. Nobody want’s to be average, but in financial markets this is usually not a bad strategy. At least historically, you have beaten most active investors by just sticking to the market portfolio.13See for example, Carhart (1997) or Fama and French (2010).

While the four papers are complicated because the authors try to estimate the true aggregate market portfolio, the concept itself is very flexible. There is a market portfolio on every level – for each asset class, for each region, etc. The criteria are always the same: hold all securities and weight them by their current market value. In most cases, this is a very good starting point and ensures that nothing too bad happens. If you are convinced that you are smarter than the rest, you may over- or underweight some securities from there. However, the beauty of the market portfolio is that nobody can escape. My overweight is the underweight of someone else and only one of us can win.



This content is for educational and informational purposes only and no substitute for professional or financial advice. The use of any information on this website is solely on your own risk and I do not take responsibility or liability for any damages that may occur. The views expressed on this website are solely my own and do not necessarily reflect the views of any organisation I am associated with. Income- or benefit-generating links are marked with a star (*). All content that is not my intellectual property is marked as such. If you own the intellectual property displayed on this website and do not agree with my use of it, please send me an e-mail and I will remedy the situation immediately. Please also read the Disclaimer.

Endnotes

Endnotes
1 That sounds unfortunate, but this is nothing new for empirical social sciences.
2 In simple terms, appraisal-based valuation means you higher someone who tells you how much your asset is worth. It is often criticized that appraisers are somewhat biased and not completely objective.
3 Ideally, one would distinguish between corporation- and non-corporation-owned commodities. However, the data for that is hardly available in practice.
4 See Appendix A of the 2014-paper for a detailed explanation.
5 In Figure 2 of the 2019-paper, they provide a long-term chart of real returns for the market portfolio and each asset category. However, this is hardly comparable to the nominal returns from 2005 to 2020 that are presented by GSV.
6 According to their estimates, global assets grew from $343T in 2005 to $597T by 2017. This is a compound annualized return of 4.36%.
7 Calculating standard deviations from 12 annual returns is absolutely not meaningful.
8 There are of course more papers out there. But to the best of my knowledge, those four represent the most recent and most comprehensive work on the topic.
9 For example, most alternative asset managers require minimum investments of $25M or even more.
10 According to the estimate of GSV.
11 This is the analyst-jargon for “I think the number is somewhere between 5 and 10% but I don‘t want to give you a specific answer that you can pin me down on.”
12 AQR is an asset management firm that is famous for its close relation to financial research. Check this website for the study on expected returns.
13 See for example, Carhart (1997) or Fama and French (2010).