50 Years of Investing in the World (Part 1)
Bogleheads know the power of diversification. And yet many such investors (including John Bogle himself!) are reluctant to diversify beyond domestic investments. Japanese investors could have perceived the same thing, with a remarkable run in the 80s in terms of market capitalization (nearly half of the world, higher than the US at some point) and in terms of stock returns. Unfortunately for Japan, as discussed in a previous blog article, this impressive success was followed by the worst equity crisis in modern history. This study about Japan also showed how some level of international diversification could have helped a local investor to mitigate this extremely painful crisis.
This raises an interesting question. Could one simply invest in the world, using global stocks and global bonds? And if this proves unsatisfying, is there a proper middle ground between domestic and global allocations?
A first version of this study was published in 2017. At the time of writing (early 2020), we now have 50 years of data to analyze, which provided a good incentive to update the study.
Morgan Stanley Capital International (MSCI) started to track international markets, as a whole and per country since 1970. One can easily download corresponding indices, per country or per region, from this Web page and derive monthly or annual returns. Countries are categorized according to the following system.
A couple of important clarifications:
- MSCI World actually means Developed Markets (while MSCI ACWI is more representative of the global market, including emerging countries).
- Emerging Markets (and the ACWI index) started to be tracked in 1988.
- Only 18 countries have historical returns tracked since 1970 and they are all in the developed markets category.
- Two of those countries (Hong-Kong and Singapore) are rather peculiar, Hong-Kong is now part of China, plus full information (e.g. inflation, bond rates) is harder to find for those, so we’ll focus on a primary set of 16 developed countries (Australia, Austria, Belgium, Canada, Denmark, France, Germany, Italy, Japan, Netherlands, Norway, Spain, Sweden, Switzerland, UK, USA).
The historical returns of those 16 countries can be obtained in USD and in local currency. This provides an easy way to figure out the dynamics of exchange rates, by comparing USD and local currency data series for a given country year over year.
Downloading the MSCI World Index series in USD currency and using the variations of exchange rates determined by the previous step, one can easily derive the MSCI World returns expressed in each of the 16 local currencies, hence from the perspective of a local investor in each of those countries. Such MSCI World data series in local currencies will be the cornerstone of our study, just bear in mind that this is solely about Developed Markets.
The Organisation for Economic Co-operation and Development (OECD) provides a rich set of historical inflation numbers, which can be downloaded on this Web page. Such inflation data per country allows to refine the perspective of a local investor, analyzing returns in real (inflation-adjusted) terms, to better compare apple to apple.
For both exchange rates and inflation rates, it is important to line up the numbers with calendar year boundaries, to be consistent with annual stock & bond (nominal) returns. Unfortunately, many studies use annual averages for such quantities, distorting the calculation of real (inflation-adjusted) returns as well as currency-adjusted returns. Care was taken in this study to avoid this pitfall.
Finally, what about bonds? The Barclays Global Treasury index was created by the end of 1986 and tracks the global treasury bond market. Such (unhedged) data series can be found on Morningstar, expressed in USD. Unfortunately, earlier data isn’t available. As a rather coarse approximation, one can use the Barclays US Treasury index (here on Morningstar) for 1973-1986, with a historical extension consistent with the Bogleheads Simba spreadsheet for 1970-1972 (10-4 bond fund model). The assumption is that the oil crisis in the 70s triggered high interest rates in most developed countries, and US bond returns should be a decent proxy for global bonds (not great, but hopefully good enough), allowing to fill the 1970-1986 gap. Using the same approach as previously explained, one can then convert such bonds historical returns in local currencies.
Note: in Part 2 of this study, we will see the trajectory of interest rates for the various countries of interest and the US trajectory was usually roughly in the middle of the pack (at least for 1970+). This reinforces the choice of US bonds as a decent proxy for global bonds in the early years, in absence of better data.
The evolution of worldwide market capitalization
Before we dive too much in numbers, it might be useful to establish some context about the respective size of the various countries in terms of market capitalization, and how this evolved over time. The following graph from the Credit Suisse Investment Yearbook 2013 does an excellent job of illustrating such evolution, which is surprisingly dynamic. It is worth noting that the Japan market weight was larger than the U.S. for a brief period of time.
The extraordinary growth of the US in the first part of the 20th century explains the US ‘premium’ over the rest of the world for the past 100+ years as an aggregate. Such premium didn’t really exist in more modern times, i.e. after WW-II, as the graph shows.
To put things in perspective, Gross Domestic Product (GDP) numbers split the world in a very different manner than market capitalization. The following graph was assembled by the author based on GDP historical data from IMF. At the time of writing, the US remains the leader, representing 26% of the world’s GDP (a far cry from its ~50% part of the overall market cap), while China isn’t far behind (18%).
A look at inflation and exchange rates
Thanks to the data assembled in the previous steps, the following chart compares inflation rates and exchange rates for the 1970-2019 time period for all countries of interest. More precisely the compound growth (CAGR) of inflation in local currency, and of exchange rate variations against the US dollar. The red square in the middle is the US. Click on the image to see a larger version of it. Let’s start by the first 25 years, 1970 to 1994.
We can observe a few things:
- Some countries suffered from quite an inflation problem, e.g. Italy and Spain at more than 10% a year, while others did a better job of containing it around 4%, e.g. Netherlands, Austria, Germany, Japan and Switzerland. Inflation risk was running pretty wild by then, and the oil crisis clearly triggered severe issues in this respect.
- For countries with high or low inflation, exchange rates changed quite dramatically against the US dollar, either appreciating strongly or depreciating strongly during this period of 40+ years. This presented quite a currency risk for local investors. The purchasing power parity (PPP) theory seems to be a very coarse tool, with notable exceptions.
- The general shape of the chart (from top left to bottom right) seems to indicate that the forces driving inflation and the forces driving exchange rates were quite linked.
Now let’s examine the next 25 years (1995-2019). How things changed… Note the change of scale of the vertical axis, compared to the previous graph.
In this second time period, it appears that central banks have been doing a much better job of controlling things and syncing up (at least for the time being):
- Most countries succeeded to put inflation in control, between 1.5% and 2.5% (which is Australia). The two big exceptions were Japan and Switzerland, where inflation ran surprisingly low, which is probably not healthy.
- Changes in exchange rates became much milder as well, although -1% or +1% CAGR for 25 years does end up compounding in a significant manner.
- The pattern observed on the previous graph seemingly linking forces driving inflation and exchange rates surprisingly disappeared.
Investing globally with stocks and bonds
To start getting an idea about our local investors (from the perspective of the 16 countries of interest), let’s look at what would have happened to an initial investment over 50 years (from 1970 to 2019), using a 70/30 Asset Allocation, i.e. 70% global stocks and 30% global bonds. All returns track the same MSCI World (global stocks) and Barclays Global Treasury (global bonds) data series, expressed in local currencies. This chart is expressed in real terms (inflation-adjusted) to better reflect the actual purchasing power of each local investor.
Unsurprisingly, the US investor was very well positioned, with solid returns and less volatility than most. Most other countries (to the exception of Canada) suffered from higher volatility, as a direct function of inflation and exchange rates, Japan and Switzerland being paramount examples at the bottom right of the chart (as Wikipedia explains, Japan acted forcefully on its currency, with rather unfortunate consequences). Please note the scale of the horizontal axis though, volatility ranging from 12% to 17% isn’t that much volatility for a stock-heavy asset allocation, a definite virtue of global investing.
More disturbingly, let’s look at maximum drawdowns (in real terms, based on annual historical returns -no intra-year dynamics-), with the same asset allocation.
While the US investor and a few other countries (incl. Canada) had to sit tight during major crises (oil crisis in the 70s, Internet bubble in 2002, financial crisis in 2008/09), investors from numerous other countries would have experienced much more dramatic drawdowns with drops exceeding 50%. In Japan and Switzerland, the situation would have been especially dire. It is fair to say that currency and inflation risks can double down with financial shocks and add insult to injury.
To finish, let’s discuss a long-term form of risk (‘deep risk’ as Dr. Bernstein would put it). Let’s look at the Safe Withdrawal Rate for a retiree, assuming the usual 30 years retirement timeframe, and fixed (inflation-adjusted) withdrawals. To do so, we can look at multiple 30 years (or close) cycles, first cycle starting in 1970, second one in 1971, etc. The following chart shows the portfolio returns (CAGR) and the Safe Withdrawal Rate (SWR) for the worst cycles (5% percentile, i.e. 95% success rate), for each country. The asset allocation remains 70/30 (global stocks and bonds).
This chart is clearly concerning, as the so-called “4% rule” didn’t quite hold, to say the least. Not only did Japan and Switzerland display an abysmal 3% or less, but a significant cluster of countries hovered around 3.5%. While Canada and the USA did quite well in comparison, but they seem to be exceptions, not the rule. Note that the Japanese ‘asset bubble’ crisis started in 1990, exactly 30 years ago.
Is it because of an overly aggressive asset allocation? Not really, running tests with a 50/50 asset allocation showed fairly similar results. As to the ‘bonds are for safety’ theory, well, 30/70 would have made matters worse, as shown below.
Back to inflation and exchange rates
The only varying factors in the previous charts were local inflation rates and the vagaries of local currency rates against the US dollar. Clearly, a higher inflation hurts real returns and an appreciating currency does the same (all other things being equal). In some cases, one counter-balances the other, while in other cases, they ‘conspire’ against the local investor.
The following table summarizes the corresponding numbers, and concisely explains why some countries didn’t perform well while others fared much better in our previous tests. Note the horizontal conditional formatting, green is good, red is not. The third row combines (geometrically) the inflation and currency rate factors, the lower the value the better.
Now it is one thing to acknowledge the performance of the US, but who would have thought (in 1970) that Sweden and Canada would end up the winners of this little test while Japan and Switzerland would be the losers? Guessing the future of a given country’s currency and inflation doesn’t seem any easier than guessing the future of an individual stock or industry sector.
Please also remember that the dynamics in the past 25 years (1995-2019) were quite different from the dynamics from the first 25 years (1970-1994) of our test. The key point is really that the dynamics of exchange rates and inflation rates are very impactful and impossible to predict. This is yet another reason to hedge bets, in other words.
Domestic bonds are typically used in a portfolio to reduce volatility and drawdowns (notably in nominal terms). One could wonder if global bonds could act as such a ballast.
Looking at a 100% global bond portfolio from the perspective of various local investors would show return volatility (standard deviation) within a range of 7% to 14%, depending on the country. While 100% global stocks would show volatility within an interestingly narrower range of 16% to 20%. The difference is significant, although a bit different from the numbers that US investors typically have in mind (US bonds volatility was less than half of the US stocks volatility, at least since 1970; post WW-II was squarely different though). Clearly, currency risk and inflation risk have a significant impact on global bonds when converting to (inflation-adjusted) returns expressed in a local currency.
Here is a regular growth chart comparing global stocks and global bonds, expressed in inflation-adjusted (real) terms and using US dollar as currency. Using another currency would (roughly) shift the entire graph up or down.
We can see that global bonds displayed decent growth, with a significantly smoother trajectory than stocks, while still displaying some level of volatility. Over 50 years, global stocks CAGR was 5.5% with a stand deviation of 17.5%; global bonds CAGR was 3.0% with a standard deviation of 7.5%; all inflation-adjusted numbers. Correlation between global stocks and global bonds was a fairly low 0.2. Global bonds would therefore have been a somewhat decent ballast in a portfolio.
One could question to which extent global bonds performance (per se and as a ballast) was hampered by local inflation and exchange rates though (switching to a Japanese or Spanish perspective pushes the standard-deviation of global bonds to 12%, as a case in point). The logical next step is to conduct similar tests while using domestic bonds in lieu of global bonds, while keeping global equities in the mix. This will be addressed in Part 3 of this study.
Exclusively investing with global stocks and bonds could be viewed by some as an extreme position. Others will view it as the logical starting point for a fully diversified portfolio.
The results of this short study raises questions about such ‘Invest in the World’ approach. One could have thought that currency risk was more of a mid-term risk, with limited impact on 30 years time periods. While true on average over the entire set of countries, this wasn’t quite true for multiple individual countries, including some with a very sizable economy.
Furthermore, currency risk appears somewhat linked to inflation risk, compounding the problem and possibly translating into various forms of short-term (drawdowns) and long-term (safe withdrawal rate) damaging issues for a local investor and notably retirees. Finally, who is to say that the currency and inflation issues which happened to Japan and Spain will not happen in other large economies in the future, including the US.
In the second and third parts of this study, we will explore other approaches, mixing some level of domestic exposure with some level of global exposure.