<<< A YouTube video is available for this blog post. >>>
From Gerd Kommer and Daniel Ganowski
If you want to set up a globally diversified passive stock portfolio, you should ask yourself how you want to weight the stocks, countries or regions in your portfolio. The following eight methods are the best known and most important in practice:
○ Market Capitalization Weighting (“MCW”): With market capitalization weighting, a stock in the index (and therefore in the ETF that tracks this index) is weighted according to the ratio of its MC to the MC of all stocks in the relevant universe. [1] Indirectly, at the level of individual stocks, MCW also leads to the MCW of the individual states to which those stocks belong (the countries in which the companies have their primary stock exchange listing). The MC method is the most widely used method and market standard in passive investing. It is easy to understand and based on intuitively plausible economic logic. It represents the “natural” weighting that results from the market itself. No other method produces lower stock turnover in the index, which in turn contributes to the low transaction costs (cost of buying/selling) in the ETF tracking such an index.
○ Equal Weighting (“EW”): All stocks in the index are weighted exactly the same. The EW method (also called 1/N method or Naïve Weighting) only works in practice in the large cap equity segment (but works very well there), as it leads to unacceptably high transaction costs in the small cap equity segment. In an international EW index, the weight of the USA would decrease significantly and that of the emerging countries would increase. There are currently 15 EW ETFs offered in Germany, most of them based on national USA indices.
○ Price weighting (“PW”): Here the companies are weighted according to the ratio of their share prices (prices) to one another - a rather logic-free weighting method, as this ratio is completely arbitrary in economic terms. Nevertheless, the PW method is used for historical reasons, e.g. B. applied in the Dow Jones Industrial Average Index (“Dow Jones Index”) and the Japanese Nikkei Index. There are ETFs offered in Germany based on both indices.
○ Purely fundamental weightings: This primarily includes weightings based on company sales (revenue weighting) or company profits (profit weighting). These methods are also called Fundamental Indexation known. As with equal weighting and price weighting, there is no connection to market prices (stock prices). It has been shown for revenue weighting that stock returns are typically higher than with the MC method (Cohen et al. 2019, Hao 2023). Because of the strong fluctuations in profits from period to period, profit weighting can lead to high portfolio turnover and therefore high costs at the fund level. ETFs that implement these two weighting methods are not offered in Germany.
○ Factor Premia Weighting/Factor Investing: Factor investing (also called smart beta investing) is the most common deviation from MC weighting in passive investing practice. Of around 1,600 stock ETFs that were offered in Germany at the end of July 2023, 283 fell into the factor investing category. In factor investing, a combination of two criteria determines the weight of a stock in the portfolio (and therefore also the weight of the country to which the stock belongs). The first criterion is the intensity of the expression of a “factor” – for example the small size factor, the value factor or the momentum factor – relative to this intensity for all other stocks in the underlying stock universe. After a specific subset of stocks from the overall universe has been preselected using criterion 1, in the second step the market capitalization determines the weight of an individual stock in this subset (criterion 2). If criterion 1 is defined by several factor premiums at the same time, it is called integrated multifactor investing.
○ GDP weighting: The weighting according to Gross Domestic Product (GDP). Here, the MC weight of each stock in the index is adjusted up or down based on a GDP factor. Specifically, the adjustment is made based on the percentage share that the GDP of the country of that stock (the country of primary listing) has in global GDP. This results in the weights of stocks from a country that has a large stock market relative to its economy (today's example being the USA) being reduced and the weights of stocks from a country that has a small stock market relative to its economy (like most emerging markets) increasing. There is currently only one stock ETF available in Germany that takes the GDP weighting element into account (see at the end of this text). That being said, GDP weighting is reasonably easy to implement by a private investor through the appropriate composition of several individual conventional regional ETFs, albeit with additional work. The second part of this blog post is about comparing the BIP weighting method with the MC method.
○ Random weighting: Although this method seems strange, it has also been researched and - before costs - apparently leads to results that are just as good as the MC method (Levy 2013, Arnott et al. 2016). However, depending on the specific implementation, random weighting could cause high transaction costs. Either way, there are no ETFs to implement this method as a private investor.
○ Mean Variance Optimization (“MVO”): The MVO weighting method is only used for individual portfolios, not for indices (perhaps because it performs so consistently poorly). The MVO was formulated in 1954 by the US researcher Harry Markowitz (1927 - 2023) and was conceptually an important milestone for the further development of scientific financial economics. Markowitz received the Nobel Prize in Economics in 1990 for this research. MVO weighting is mathematically comparatively sophisticated, but in most scientific studies it performs poorly in terms of absolute and risk-adjusted returns (Sharpe Ratio), often even the worst among all analyzed methods (Marques Mendes/Santos 2019). The main reason for the empirical failure of the MVO method: Hardly any other method is based more heavily on short and medium-term forecasts of prices, volatilities and correlations. Forecast-based investing is known to work poorly due to the error-prone nature of forecasts and the high transaction costs of implementing them.
Of course, the methods mentioned here can also be combined.
The fact that MC weighting generally or even predominantly leads to a better return-risk combination than the alternative weighting methods listed here (with the exception of MVO, which performs poorly universally) has been frequently and consistently refuted by scientists over the last 20 years [2]. These analyzes show that most alternative weighting methods have historically produced more attractive absolute returns or return-risk combinations (Sharpe Ratio) than MCW, provided the analysis period is sufficiently long. MCW only tends to perform very well or even best in comparison with relatively short data series that cover the immediate past decade. We'll talk about the reason for this below.
Before we come to the return comparison between the MC method and the BIP method, we would like to briefly address the topic of “cluster risk with the MC method”.
Because the MC weighting has the least influence on market developments, it can lead to particularly strong “clumping”, i.e. – depending on the perspective – undesirably high “imbalances” or concentrations in relation to individual values, sectors and countries. In this regard, the following table compares the MC-weighted and the GDP-weighted alternatives of two well-known international stock indices.
Table 1: Comparison of two market cap-weighted equity indices with their GDP-weighted alternatives with regard to the cluster risk aspect (as of June 30, 2023)
► Source: Fact sheets of the respective indices. ► Most percentages have been commercially rounded.
Table 1 shows that the GDP-weighted index variants (GDP Weighted) are less concentrated in general and specifically with regard to the weight of the largest individual state (here USA).
We believe that high concentration in a single country creates unnecessary black swan risk, which passive long-term investors can and should avoid. Below are four historical mini case studies. These case studies are intended to illustrate that even countries with a comparatively large stock market can go “from 100 to 0” in a short period of time.
○ Case study Russia from 1917: The Russian stock market suffered an uncompensated loss of 100% in less than 12 months in 1917 - the year of the Russian Revolution - due to the nationalization of all listed companies. The Russian stock exchange then remained closed for 76 years, until 1994. The share of the Russian stock market in the world stock market in 1917 was around 11%.
○ Case study Germany from 1941: The German stock market suffered a loss of 96% in the six years from the end of 1941 to the end of 1947. The index level of 1941 was reached again in 1953 (although not for shareholders who lived in the eastern part of the former German Empire, i.e. in the GDR and the former German areas in Poland and Russia). At that time, the German stock market's share of the world stock market was around 5%.
○ Case study China from 1949: The Chinese stock market suffered a 100% loss from 1937 to the end of 1949, when the Chinese Stock Exchange was closed in the context of the communist revolution. All listed companies were nationalized without compensation in 1949/1950. The Chinese stock exchange only reopened 43 years later in 1992. The share of the Chinese stock market in the world stock market was around 5% in 1949
○ Case study Japan from 1989: With the bursting of the then Japanese real estate and stock bubble, the Japanese stock market suffered a slow-motion crash starting in March 1989, which only reached its maximum drawdown of minus 74% (real in USD) after 14 long years by mid-2003. Adjusted for inflation, the Japanese stock market has not yet fully made up for this loss in dollars or euros to this day, 34 years after the crash began. Japan's share of the world stock market at that time was an astonishing 44%, surpassing the USA share.
In the last 100 years, numerous other countries temporarily became “failed states” whose economies and stock exchanges (if there was one) suffered serious and severe damage as a result, including Austria from 1914 onwards, which had a small stock market at the time.
Against this background, a rhetorical question to the readers of this blog post: What impact on returns would it have if the weight of such a “failed state” in one’s own stock portfolio had previously been 65%?
It is not obvious that the USA is fundamentally exempt from such a risk. No country is fundamentally exempt from this. The “storming of the Capitol” (Parliamentary Building) in the USA in January 2021 could be viewed as worrying in this regard. The killing of the Vice President (Mike Pence) and the Speaker of the House (Nanci Pelosi) were within the realm of possibility. At that time, leading politicians in the USA spoke of the danger of civil war.
Such black swan risk must be reduced for one's own portfolio to the extent that this can be implemented with reasonable costs and effort. The BIP method is a rational, easy-to-implement option for this with ETFs.
In practice, the GDP method will tend to lead to more balanced country weightings by lowering the weight of countries in the portfolio where the stock market significantly exceeds the state's share of the global economy. This freed-up weight is then redistributed to those countries whose stock exchanges are below average relative to the size of the national economy and have the potential to catch up.
In Table 2, we compare the historical returns of MC and GDP weights using the longest data series available for two global stock market regions.
Table 2: Comparison of returns of two regional stock indices: country-weighted index by market capitalization versus index weighted by gross domestic product - in USD, adjusted for inflation
► “EAFE” = abbreviation for Europe, Australasia, Far East = the non-North America part of the MSCI World Index. ► The longest data series for the two index pairs are shown. ► GDP-weighted indices going back further than 1970 are not available. ► Inflation-adjusted returns in USD excluding costs and taxes. ► Implementation via ETF would lead to slightly higher transaction costs with the BIP method due to the need for rebalancing, which would have to be deducted from the index returns.
The table shows that the GDP weighting method produced a slightly higher return than the MCW globally in developed countries (MSCI World Index) for the full 53.5 year period from 1970. If you divide the entire period into two halves, the GDP method was slightly ahead in the first and slightly behind in the second. If one were to only show the last ten years up to June 2023, the MC method would show a clear advantage, since the USA outperformed almost all other national stock markets in the world during this period and because the USA is given a particularly high weighting in the MC method due to its large stock market.
The separate consideration of only the industrialized countries exclusive USA and Canada using the MSCI EAFE index (see table) provides an interesting nuance of information: It is the individual case of the USA among the 23 developed market countries (according to MSCI definition) that led to the market cap method's slight lead in the MSCI World in the second half of the period. Among the 21 EAFE countries, the GDP method was also ahead in the second subperiod.
The data is sufficient for emerging markets MSCI EM GDP Weighted Index only back to June 2000. In the 23 years from then to June 2023, the GDP-weighted index produced an average return almost one percentage point per annum higher than the MC index.
The statistics for the three indices considered illustrate that the GDP method delivers good results in terms of returns.
Because the risk indicators Volatility (Fluctuation in monthly or annual returns) and Maximum drawdown (maximum cumulative loss) were very close to each other in all periods shown here for the two index variants (the two weighting methods), we do not show these risk figures for reasons of space.
Our historical analysis shows that the BIP method has a return advantage, but overall it is not large enough to recommend it over the MC method for this reason alone.
Our main motivation for preferring the BIP method is that it is less likely to result in “national cluster risks” in the portfolio and therefore less black swan risk in relation to the countries in which these companies are located.
A welcome side effect of the BIP method is that it also reduces the cluster risk in relation to the five or ten largest individual stocks in a portfolio.
Currently (as of June 30, 2023), a portfolio weighted according to GDP is also valued significantly more favorably than a portfolio based on MC weighting (see bottom line in Table 1), which has a positive effect on the expected return of a portfolio.
We have shown above using numbers that the BIP method is equal to and perhaps even superior to the MC method in terms of return and risk. We have tried to argue through our historical digression that it carries less black swan risk.
Finally, there remains the refutation of four weak arguments against GDP weighting that we come across from time to time.
○ Anti-GDP method argument 1: "The world's largest listed companies have the best growth prospects and should therefore be given the highest weighting in a passive portfolio. This can only be achieved with the MC method."
As far as we can see, there is no scientific support for this thesis, at most for its opposite (Dimensional Fund Advisors 2020). The fact, which is largely undisputed in research, that almost all weighting methods, including the radical opposite of this argument - the equal weight method - lead to return-risk figures that are as good or better than the MC method in the long term also refutes the argument. The thesis that mega-large caps have particularly attractive returns is the result of a common misconception Recency Bias, i.e. mistakenly considering what happened in the recent past to be particularly representative of the future. In other words: the recent past is extrapolated more or less linearly into the future. Among the over 50 cognitive errors, which has researched the field of behavioral finance over the past few decades, recency bias is probably the most serious because it causes the most financial damage.
○ Anti-GDP method argument 2: “The best, fastest-growing companies in the world tend to list in the USA because the USA offers the most attractive stock exchanges in the world” (NYSE, Nasdaq).
This is a variation of the false anti-GDP argument mentioned above, but argument #2 is even shakier. First of all, it suffers from the recency bias just as much as argument 1. Until the end of the 1980s, argument 2 could have been applied to the Tokyo stock exchange with unpleasant consequences for returns. [3] The fact that the USA has particularly attractive stock exchanges today does not have to stay that way and will not in the long term because there is too much competition between the 195 countries in the world and their stock exchanges. In addition, a tendency to list where the company expects the highest valuations (if this tendency existed) would be detrimental to returns from a shareholder perspective. High valuations reduce expected returns.
Furthermore, the statement ““The best, most growth-producing companies in the world tend to be listed in the USA.” doesn't agree with the facts anyway. The number of listed companies in the USA has been falling for years, while it is increasing in other countries, especially in the emerging markets. The main reason for the decline in the USA is the strict US regulations in this regard, which are less strict, less legally risky and less costly elsewhere. It is rare for an already large, successful company to move its primary listing to the USA. [4] We estimate that 98% of companies in developed countries have their primary listing in the country where their corporate headquarters are located. They are therefore particularly exposed to the political, tax, regulatory and macroeconomic risks of this country. Emerging market companies that have their primary listing abroad are more likely to list in London and other stock exchanges than in New York.
Furthermore, a company can be listed on several stock exchanges through a “secondary listing”, which is the case for many large companies. So if a large, thriving non-US company wants access to the liquid US stock market, a comparatively inexpensive “non-primary listing” in the USA is sufficient. (However, the country allocation in a stock index is done via the primary listing, which is almost always in the country of the company's headquarters.)
○ Anti-GDP method argument 3: “The Internet and technological progress, including the “network effect,” have fundamentally changed the rules of the game in all major markets. Therefore, insights and rules for investing that were valid until a few years ago are now obsolete.”
In terms of its basic theme, this “new” argument is actually as old as the stock market itself. For example, it was put forward in almost exactly the same form in the run-up to the Dot.Com bubble at the end of the 1990s. Readers of this blog know what happened to the shares of the “New Economy” from the beginning of 2000 (the bursting of the Dot.Com bubble).
In general, the advocates of the age-old “this-time-it’s-different” argument always provide a lot of stories combined with a good portion of “regulars’ table economics”, but little in the way of hard numbers or established scientific findings. The high-tech industry has historically not delivered higher stock returns than the overall market, but has provided more risk (see here).
Incidentally, the “disruptive” economy was not discovered by Frank Thelen; it has been around for 250 years. It began with the Industrial Revolution around 1770. Disruptive “high tech” existed 150 or 75 years ago just as it does today - i.e. new, “revolutionary” technologies that often seemed like miracles to contemporaries and often structurally favored the market leader due to scale or network effects [5] and promoted economic structural change. But even structural advantages do not help in the long term against the forces of competition, “sclerosis” of large companies (“complexity costs”) and anti-monopoly measures by the state.
In any case, technological change has slowed down rather than accelerated since around 1970, but many media professionals and YouTubers apparently haven't noticed this yet. This slowdown is seen by macroeconomists as a key reason for the decline in global economic growth in recent decades (Gordon 2018).
○ Anti-GDP method argument 4: “Because the largest companies in the world according to MC – Apple, for example – are known to export to many other countries, it is not necessary to have every country or region in the portfolio; the largest companies in the USA or worldwide are sufficient.”
Although this argument is often used against the GDP method, if you look closely, it is not directed against it at all, but against global diversification per se. (However, the advocates of this argument do not always seem to notice this.) In general, the argument is so weak that it is hardly worth spending much space refuting it. If you followed him consistently, you should have always only invested in the ten, 20 or 50 largest companies in the world. The result would be a portfolio that would have produced a worse risk-adjusted return (Sharpe ratio) and predominantly a worse absolute return than a better diversified world stock market portfolio over most ten-year periods over the past 50 or 100 years and over the entire period. Yes, over the last ten years or so, a mega-large cap portfolio would have looked good, which brings us back to the basic fallacy of recency bias.
Conclusion
Anyone who wants to make their portfolio “ultra-stable” also thinks about black swan risks, which perhaps only occur every 20 to 40 years and which cannot be calculated or predicted. One of these “tail risks” is that a country with a particularly large stock market, due to its own internal failure or for external reasons, falls into a sudden or gradual malaise that does not affect the other 194 countries in the world in the same way. This risk could exist for a passive equity investor today with respect to the USA because the country accounts for approximately 65% of a market cap-weighted portfolio. A rational way to mitigate this risk and other manifestations of cluster risk in the portfolio is to use GDP weighting or a combination of GDP and MC weighting - as we do in the L&G Gerd Kommer Multifactor ETF (WKN WELT0A).
Endnotes
[1] Market capitalization for stocks corresponds to the market value of the company's free float (not the market value of the total capital, as some believe). Mathematically: MC = number of shares outstanding × share price.
[2] See e.g. B. DeMiguel et al. 2009, Chow et al. 2011, Clare et al. 2013, Taljaard/Mare 2020, Dai/Saito 2022, Swedroe 2023.
[3] In the 15 years from March 1989 (when the Japanese stock bubble began to burst) to February 2004, the MSCI World GDP Weighted had a return 1.4 percentage points p.a. higher than the conventional MSCI World.
[4] The move of the former DAX share Linde to the NYSE stock exchange at the beginning of 2023 was not an example of such a “move”, although this is sometimes claimed. Rather, it was the result of a merger between Linde and the US company Praxair. The merged company understandably wanted to reduce the high costs of a dual listing. Incidentally, four years before the delisting in Frankfurt, Linde had moved its headquarters from Germany to Dublin in order to avoid what Linde considered to be overly restrictive provisions of the German Co-Determination Act.
[5] Think e.g. B. the steam-powered railways that emerged in the first third of the 19th century, first in Great Britain and some time later in other countries, all of which were owned by private companies (not the state).
literature
Arnott, Robert/Jason Hsu/Vitali Kalesnik/Phil Tindall (2013): “The Surprising Alpha From Malkiel’s Monkey and Upside-Down Strategies”; In: Journal of Portfolio Management; 39; No. 4; March 2013
Chow, Tzee-man/Jason Hsu/Vitali Kalesnik/Bryce Little (2011): “A Survey of Alternative Equity Index Strategies”; In: Financial Analysts Journal; 67; No. 5; September 2011
Clare, Andrew/Nick Motson/Steve Thomas (2013): “An Evaluation of Alternative Equity Indices – Part 1: Heuristic and Optimized Weighting Schemes”; 01 April 2013; Internet reference: SSRN/Social Sciences Research Network
Cohen, Laura/Mo Haghbin/Christopher Malloy/Matthew Schilling (2019): “Revisiting Fundamental Indexation: A Deep Dive into Revenue”; March 14, 2019; Internet reference: SSRN/Social Sciences Research Network
Dai, Wei/Namiko Saito (2022): “Weighting for the Right One: Weighting Scheme Design for Systematic Equity Portfolios”; Aug 18, 2022; DFA; Internet reference: SSRN/Social Sciences Research Network
Dimensional Fund Advisors (no author) (2018): "Large and In Charge? Giant Firms atop Market Is Nothing New"; May 17, 2020; Internet reference: www.dimensional.com
DeMiguel, Victor/Lorenzo Garlappi/Raman Uppal (2009): “Optimal versus Naive Diversification: How Inefficient Is the 1/N Portfolio Strategy?” In: Review of Financial Studies; 22; No. 5; May 2009
Gordon, Robert (2018): “Why Has Economic Growth Slowed When Innovation Appears to be Accelerating?”; NBER Working Paper; April 2018; Internet reference: https://www.nber.org/papers/w24554
Hao, Wenli Bill (2023): “Revenue-Weighted Indices: An Alternative to Core Equities”; S&P Dow Jones; March 29, 2023; Internet reference: S&P Dow Jones website
Kanzler, Daniel took (2022): “How to cleverly deal with the potential cluster risk in the USA in your ETF world portfolio”; YouTube video; March 20, 2022; Internet source: YouTube channel Gerd Kommer
Levy, Moshe (2016): “It’s Easy to Beat the Market”; In: Journal of Investment Management; 14; No. 3; Third Quarter 2016
Marques Mendes, António/Dinis Santos (2019): “Portfolio Weighting Methods: Naïve vs. Scientific Diversification”; Oct 25, 2019; Internet reference: SSRN/Social Sciences Research Network
Taljaard, Byran/Eben Mare (2020): “Why Has the Equal Weight Portfolio Underperformed and What Can We Do About It?” May 23, 2020; Internet reference: SSRN/Social Sciences Research Network
Swedroe, Larry (2023): “Is a Naive 1/N Diversification Strategy Efficient?” March 31, 2023; Internet reference: https://alphaarchitect.com/2023/03/naive-diversification/