The CAPM had never been tested and probably never will be because the market portfolio at the heart of the model is theoretically and empirically elusive.
Roll (1977) and Fama-French (2004)
One of the key concepts in Finance is the Market Portfolio. This means a universal benchmark portfolio containing all investable assets in a market participant's investment universe. As an ultimate anchor, the Market Portfolio thereby provides investors with an objective performance benchmark and enables them to judge their degree of active deviation as well as the success of their bets.
Among other things, it serves as a key pillar influential works such as the Capital Asset Pricing Model (CAPM) and the Black-Litterman Model. However, despite its ubiquitousness, this Market Portfolio has remained somewhat of a Chimere and some academics even reckoned that it is generally unobservable and will never be clearly defined.
As Gadzinski et al (2021) pointed out, "investors still lack a global composite portfolio benchmark that includes a broad spectrum of assets with weights in line with a fair representation of the stock of capital for each asset class".
In their influential 2018 paper in the Journal of Portfolio Management (JPM), Gadzinski et al already tried to tackle this problem by gauging the global capital stock mainly using data from the Bank for International Settlement (BIS), Eurostat, and the OECD. Their study included estimates of the global value of assets that are not traded in public markets such as private businesses and real estate (see chart below).
In 2021, they followed-up with another article in the JPM, providing two version of the Market Portfolio, including one portfolio consisting of 11 asset classes replicable through ETFs on 87 different indices (the full composition is available here). While previous, more theoretical versions were mostly relevant to institutional investors, this finally makes the Global Market Portfolio a viable concept for the retail market as well.
I last wrote about the topic in September 2020 (see attached paper), introducing a global multi-asset portfolio (Fixed Income, Equity, Gold) derived bottom-up from Bloomberg data. This follows a slightly narrower approach, ignoring private assets.
At Amadeus Capital we then used this to construct a proxy portfolio through ETFs and simulated its performance over the past 20 years with the help of index mapping. In practice, we tend to follow the global market capitalization-weighted portfolio relatively closely on the equity side but prefer a more active approach to Fixed Income.
Challenges on the Fixed Income side include the problem that a market capitalization-weighted approach automatically overweights heavily indebted issuers as well as the complexity and cost of foreign currency hedging, and subdued liquidity in many corners of the market. Beyond this, the massive interventions of Central Banks raise the question of whether this market portfolio is an objective reference point, reflecting the investment community's allocation decisions.
In the equity market, things are a bit more straight forward and Bloomberg publishes a range of indices tracking the market capitalization of basically all countries' stock markets making it computationally less burdensome to derive weights. With the bottom-up approach introduced in my 2020 article, it is possible to differentiate further between sectors and industries.
However, not everyone has access to a large-scale database and this is why I decided to write the World Equity Market Portfolio application for the quantamental platform.
This application builds on the open API of the Worldbank. Apart from providing market capitalization weights (by country), it also supports a GDP weighted approach. Furthermore, it makes it easy to compare the historical development of single or multiple countries' weights using both approaches. Some investors argue that market capitalization weights steer investors away from less developed markets, most important markets with large insider holdings (such as recently privatized state-owned companies).
Investors following a market capitalization-based approach may also be more at risk of getting trapped in markets that are severely overvalued (for instance Japan in the 90s). I would like to note though, that the famous Buffet Indicator comparing a country's market capitalization to its GDP is already questionable as a market timing indicator and most likely even worse in the context of country allocation decisions. Generally, there are huge structural differences between different economies and they tend to be pretty persistent. Among other things, this includes differences in the capital structure such as a higher portion of debt financing or higher share of privately held firms vs public equity financing.
GDP-weighted portfolios may also take overly large positions in countries with weak governance standards, low profitability, and a small export sector. In this context, we recently asked ourselves, which approach would have done better in the face of China's crackdown on various industries. The slightly dissatisfying answer is, that the country's weight would have been approximately the same in both portfolios (don't forget Hong Kong in this context).
Unfortunately, Worldbank data is not as timely as the market capitalization data provided for instance by Bloomberg. However, even changes in market capitalization-based weights tend to be modest over time.
Hence, while we rely on Bloomberg and other more timely data providers in our day-to-day investment operations, this little tool can probably provide some guidance and interesting insights.