50 Years of Investing in the World (Part 2)
Many North American investors tend to look carefully at historical returns in the US and in Canada and draw various conclusions. Occasionally, some references are made to Japan and the UK, but few people look any further. The world changes though. The UK was undoubtedly the world economic leader at the end of the 19th century, while the US clearly dominates nowadays. Japan was on a roll in the 80s, with a bigger market capitalization than the US at some point, and yet badly faltered since then. The world changes in ways we cannot predict and it would be naive to assume that a few decades in the future, the situation will be similar to today’s environment. One thing we can do to get some perspective is to analyze what happened to a larger sample of sizable economies.
This article focuses on the historical returns from 16 developed countries over the past 50 years, looking from the perspective of a local investor and assuming a strong home country bias to begin with (i.e. solely using domestic stocks and domestic bonds). We will look at more diversified portfolios mixing domestic and global investments in Part 3.
Quick Reminder about Part 1
This article is the second part of a multi-part study, click here for Part 1. We will leverage data sources described in Part 1, mostly based on MSCI data series for per-country stock historical returns (in USD and in local currency) and OECD for per-country inflation rates. The evolution of per-country market capitalization over the past century was illustrated by this graph.
Part 1 studied the idea of investing 100% in the world (global stocks and global bonds). The outcome proved surprisingly diverse, due to very distinct (and hard to predict without hindsight) patterns in inflation and exchange rates in the various countries being investigated. We concluded with the desire to look at other types of asset allocations, involving domestic stocks as well as global stocks. Part 2 (this article) explores the view of exclusively investing with domestic assets. Part 3 will seek a middle ground.
As a fully diversified ‘Invest in the World’ approach is our general theme and starting point, most charts below will compare side-by-side a ‘domestic only’ approach with a ‘global stocks’ approach, while using domestic bonds on both cases.
Assembling Domestic Bonds Data Series
Stock returns from 1970 are fairly easy to assemble, leveraging MSCI per-country data series. Bonds historical returns aren’t as readily available on a per country basis. Barclays provides global indices for treasury bonds in each country starting in 2000 in most cases (and a decade or two earlier for the other cases). An exception is Switzerland, which isn’t tracked by Barclays, but this is easily addressed by the SIX SBI index. For most countries, this leaves several decades of missing data starting from 1970.
OECD and IMF both provide government bonds (10-yrs) and bills (3-month) interest rates for all countries of interest. The IMF database is fine-grain (monthly rates) and provides extensive history for bonds and bills, although recent updates have issues (some data series disappeared; lagging updates for past few years). The OECD database is fine-grain (monthly rates) and more timely updated, but it has some major gaps in the early years, notably for bonds. For overlapping years, OECD provides bond numbers which are consistent with IMF (bills are slightly different, notably due to the use of ECB rates instead of per-country rates for most of Europe in the past two decades). For the few remaining gaps, the Macrohistory database provided the missing data (annual bill and long-term rates). We settled on using IMF data when available, then OECD, then Macrohistory. Finally, one last gap remaining (a few early years for Austria/bills) was approximated by averaging the other countries, to obtain a full 1970-2019 dataset of bonds and bills interest rates.
Extending the same logic to go back farther in time, here is a chart displaying the evolution of treasury bonds interest rates for the 16 countries of interest, between 1940 and 2020 (rates as of the beginning of the year). As you can see, Japan (JPN) and Switzerland (CHE) acted somewhat independently, while the other countries largely followed each other, at least starting from the 90s, with a collective drift towards low interest rates. Most of Europe has been using very low (and occasionally negative) interest rates in recent years, while both Italy and Denmark peaked around 20% (!!) in the early 80s, while inflation was raging. Things change…
The next step is to use those treasury bond and bills interest rates and derive (nominal) returns. Here AlohaJoe came to the rescue, leveraging his extensive experience with the bond fund simulator and clever ways to derive bond returns from bond rates while simulating a bond fund (as opposed to individual bonds). This allowed to construct full returns data series for each of the 16 countries, approximating a treasury bond fund. A sanity check with known US treasury bond returns and some known international treasury bond returns expressed in local currency provided reasonably satisfying results when comparing growth charts of known returns against modeled returns.
Please keep in mind that simulated numbers are just an imperfect baseline for analysis, not actuals. Actuals (the Barclays/SIX indices previously described) were used whenever possible in lieu of modeled numbers in the following analysis.
Finally, inflation rates determined in Part 1 of this study allowed to derive real (inflation-adjusted) bonds returns for all countries and years of interest.
The first instinct of many investors is often to stick to known quantities, i.e. domestic equities. This presents the great advantage of reducing currency risk and the disadvantage of making a narrower bet, restricting diversification. Let’s investigate a (somewhat extreme) 100% domestic stocks asset allocation and see what would have happened between 1970 and 2019 to a local investor with an initial investment made end of 1969. The following charts use returns expressed in local currencies, using real (inflation-adjusted) returns.
The outcome of the domestic-only approach (left side of the display) is very diverse, to say the least. Italy stands out as the worst under-performer (followed by Austria, Spain and Japan). While Sweden, Denmark and the Netherlands did remarkably well. As to the US and Canada, they stand out by displaying solid growth and low volatility (relatively speaking). Also note the extreme volatility of Norway (probably due to its concentration on natural resources industries, e.g. petroleum). Comparing with the global stocks approach is quite striking, providing solid mitigation to the poor performers (and dampening a bit the results for the best performers) and making outcomes much closer to each other.
Let’s look at maximum drawdowns now (using annual returns, hence not accounting for intra-year events; also remember those are real returns, inflation-adjusted).
Holy smokes. Can you imagine drawdowns of 70% or 80% in countries as diverse as Australia, Japan, Norway, UK, Spain, Italy and more? It is probably a safe guess that few investors outside North America would want to be 100% domestic stocks after such a ride in the past five decades. Note how most countries (e.g. besides Canada and Denmark), even the US, would have seen drawdowns improved by a global stocks approach, although drawdowns of 50% to 60% remain dizzying. The bottomline is clear, diversification is in order, using a mix of bonds and global equities.
Portfolio CAGR computations are notoriously sensitive to start and end dates though. The PortfolioCharts author (Tyler) was kind enough to provide a customized version of his excellent ‘Heat Map’ representation, allowing to analyze our country-specific data series while varying start years and years held. Here are four heat maps, the truly frightening Italy and Japan, the impressive Sweden and the familiar USA. As Tyler says, a picture is worth 10,000 numbers.
Please check here for explanations on how heat maps work. Quite obviously, if you see some level of red on a row after 15 years or so, then it is an extremely severe drawdown period. Also, if you see white after 25 to 30 years, then you can probably forget about the “4%” withdrawal rule.
Outcomes were quite diverse, to say the least. Overall, besides Sweden, Canada, the Netherlands and Denmark did remarkably well, while Spain struggled as mightily as Italy. There is something to be said for cold weather, as an early reader pointed out!
Hopefully, it would never come to anybody’s mind to invest in 100% bonds, but it is interesting to look at those numbers in isolation if only to acknowledge that bond returns, when expressed in real terms, can be as crisis-prone as stock returns (or at least have been in the past). Here are the corresponding charts. The biggest problem was of course the oil crisis in the 70s, with inflation devastating bond returns all over the world (it wasn’t the first time this happened, the aftermath of World War II was even more devastating for bond holders). Let’s check volatility vs. returns, domestic bonds on the left side, global bonds on the right side.
Volatility appears well contained with domestic bonds (except Spain and Italy to an extent) , much better than with global bonds. Clearly, inflation and currency rates forces made global bonds much more volatile, which is definitely not what typical investors look for with bonds.
Let’s check the maximum drawdowns now (remember, this is expressed in real terms, adjusted for inflation). Sobering, isn’t it?
Some readers might be surprised by treasury bonds drawdowns of 25% or more for most countries (including the USA), plus the extreme cases of Spain and Italy domestic bonds. Inflation and rising interest rates can be truly devastating for bonds and this becomes very visible when expressing returns in real terms. Using global bonds would have helped Spain and Italy, but would have been very counter-productive otherwise.
In numerous cases, it actually took between 10 and 20 years for bonds to recover in inflation-adjusted terms. Here is the drawdown trajectory of US domestic bonds (more than a decade before recovery in 1984) and Italy (OUCH!) based on annual returns. We didn’t witness any bond crisis in recent decades, which makes us subject to recency bias, but truth is bond crises can be even more frightening than severe stock crises, if only because of their duration and somewhat insidious nature (harder to perceive in day-to-day nominal terms).
Let’s look at Japan and Australia now. Domestic bonds would undoubtedly have helped a Japanese investor during the recent Asset Price Bubble crisis which unfolded in recent decades (starting in the late 80s), but they didn’t behave as well during the oil crisis, with a decade-long 20% to 30% drop. Australia went through even bigger troubles though, with a deep bonds crisis lasting nearly 20 years (Norway displayed a similar pattern). Which is itself nothing compared to Spain (an eye-popping 30 years drawdown which went deeper than 70%). Spain wasn’t a one-time event, something similar happened in multiple major countries after World War II, but that is another story…
In conclusion, bonds investors have been subjected to devastating crises, much more visible when expressing charts in real terms. Global bonds did significantly worse than domestic bonds, damaging bonds’ ability to act as a ballast, and will not be considered in the rest of this article.
A More Balanced Asset Allocation
Clearly, investing 100% in domestic stocks or 100% in bonds isn’t quite reasonable, notably when seeing the devastating crises which occurred in some countries for either asset class. Let’s examine a more balanced 70/30 portfolio (30% bonds) like we did in Part 1 of this study, comparing a domestic-only approach to a global stocks approach. On the bonds side, we will stick to a domestic-only allocation in both cases.
Looking at the domestic-only charts (left side), volatility is more contained than with 100% stocks, but still quite dispersed. Norway still stands out as highly volatile and Canada/USA as remarkably stable (relatively speaking). Italy remains disturbingly low growth. Drawdowns are improved compared to 100% stocks, but a good number of them remain frightening (50% or more).
A 70/30 allocation shows better results than 100% stocks, but a domestic-only approach remains very painful for multiple countries. And given the disturbing trajectory of bonds during the oil crisis, a 50/50 asset allocation would NOT have saved the day (quick test: 5 countries displayed drawdowns of 50% or more). The comparison with a global stocks approach (right side of the display) seems to make a rather compelling case, dampening both volatility and drawdowns in most cases, and only ‘lottery winning’ domestic investors (e.g. Sweden or Denmark) could have had cause for complaining… on hindsight!
Balanced Asset Allocation for retirees
Let’s take the perspective of retirees and compute the Safe Withdrawal Rate (vs. portfolio CAGR) for the worst 30-years periods a local retiree could have faced since 1970, while keeping a 70/30 asset allocation to begin with. Note that this chart shows the 5% percentile number (i.e. 95% success rate) typically used in SWR computations.
As you can see on the domestic-only side, Spain, Italy and Japan were in a race to the bottom (2% to 3%) while Australia struggled too. Interestingly enough, quite a few countries did better than the USA in this respect and readers will not be surprised by now to see Sweden and Denmark in the top right corner.
The global stocks approach nicely mitigated Italy (barely visible on the graph, now at 4.0% SWR) while improving the situation for other poor performers (e.g. Spain and Japan), but not very convincingly. Unfortunate side-effects occurred though, Norway and Switzerland dropping around 3.5% and other countries losing ground on the horizontal axis (SWR).
Most retirees tend to increase their bonds exposure, perceiving that this will be a ‘safer’ approach, so let’s check a 50/50 allocation. The results are not quite convincing as is easily seen in the graphs below, reflecting the issues with bonds and real returns we’ve discussed earlier, and with mixed signals again. Note that using global bonds would have helped Spain, but would have been counter-productive for most other countries.
A fully domestic investment can deliver fairly good results, as anybody having invested in the US or Canada knows. It can even deliver very impressive results as Swedish and Danish investors experienced in recent decades. But it can also put local investors in devastating situations, with decades-long drawdowns for both stocks and bonds (in real terms), as Spain, Italy and Japan investors went through. Trying to predict without the benefit of hindsight in which camp a given country will fall in future decades appears extremely speculative.
Comparing a domestic-only approach with a global stocks (+ domestic bonds) approach, the latter appears to come significantly ahead, mitigating severe issues with the worst performers without overly impacting the middle of the pack. Still, the Safe Withdrawal Rate chart (as well as the discussion in Part 1 of this study) showed that only betting on global stocks may not be entirely optimal.
We will explore in Part 3 if we can find a better compromise by keeping some domestic exposure in addition to solid global exposure, and see what impact this would have had in the various countries we’ve been studying.