Over the years, I’ve devoted many posts to the idea that stocks aren’t as “scary” as typically assumed. And sometimes I’ve approached it from the flipside and argued that government bonds don’t entirely deserve their reputation as a “safe” asset.
So, I’m constantly on the lookout for new arguments or research that emphasizes the scary aspects of stock investing. Most of what I find leans too heavily on rhetoric and perfunctory analysis to be worth discussing. But occasionally, something more insightful shows up like the 2021 paper, Stocks for The Long Run? Evidence from A Broad Sample of Developed Markets by Aizhan Anarkulova, Scott Cederburg, and Michael S. O’Doherty (ACO Study for short).
The ACO Study makes three major conclusions, which I’ll paraphrase as follows:
- The long-term outcomes from diversified stock investments are highly uncertain.
- Catastrophic stock investment outcomes are common even with a 30-year horizon.
- Stock market performance across dozens of developed markets is notably worse than historical U.S. experience.
The title of the paper seems a bit cheeky because the authors clearly think that stocks are pretty scary, even for the long run.
I conducted a similar but less robust analysis for developed market stocks using MSCI developed market data going back to 1970 earlier this year. And I reached a nearly opposite conclusion that “the risks from global¹ stocks are pretty similar to the risks from U.S. stock investing”. So, at first glance, I found the ACO Study as a bit of mystery, and I love a good mystery.
Examining the Evidence
A good way to start solving a mystery that comes from historical returns is to examine the data that are included and excluded from any given analysis. Here’s a graph from the ACO Study showing the timespans of return data from each Developed Market country that they used.
It’s an impressively large data collection. And the ACO Study used specific economic criteria to expand their dataset to include some countries that have only qualified for developed market status in the last 5 to 30 years (see the bottom of the graph).
The ACO Study randomly sampled from this entire dataset to generate statistics on long-term stock performance. One of the most interesting statistics they developed was the chance of losing inflation-adjusted money over different investing timeframes. This statistic is essentially identical to the “permanent loss” stats that I’ve generated in the past based on other datasets.
This bar graph shows the chances of losing inflation-adjusted money (in local currency) in stocks using the ACO Study dataset and random sampling methods for various buy-and-hold timeframes, which I calculated using summary data from their paper.
The researchers found a greater than 20% chance of losing inflation-adjusted money when holding stocks for 10 years. So, they reasonably concluded that global stocks are scarier than U.S. stocks, which have historically exhibited a roughly 10% chance² of losing money over a decade. Perhaps even scarier, the ACO Study found a greater than 10% chance of losing inflation-adjusted money in global stocks even with a relatively long 30-year investing timeframe.
Interviewing Witnesses
While the evidence of the ACO Study leads to one conclusion, most mysteries are solved by interviewing multiple witnesses and cross-checking their stories for consistency. Over the years, I’ve compiled several different estimates of permanent loss statistics for both U.S. and global stocks. The stories from these other witnesses are compared to the ACO Study in this graph.
I wrote about the JST Study here, which contains return data going back to at least 1900 for 16 Developed Market countries. And I wrote about MSCI stock data here, which includes data back to 1970 for 22 Developed Market countries. Finally, the U.S. statistics come from an analysis I conducted of S&P 500 data going back to 1928.
For the three studies of global stocks, I gathered or calculated the permanent loss statistics for 1-year, 5-year, 10-year, and 20-year timeframes and then extrapolated in between these timeframes. For the MSCI data, I stopped at 10 years. I should also note that the line for U.S. stock data is a bit more fine-grained because I calculated the permanent loss statistics for every timeframe from 1 to 20 years.
So, all three of these global datasets agree that the chances of losing money in global stocks are greater than investing in U.S. stocks, with the JST and ACO Studies both putting the chances of a loss for 10-years of investing at greater than 20%.
Comparing Stories
Although our three global witnesses agree that a crime occurred, they disagree on some of the details given that the 10-year loss estimates ranged from 25.1% (JST Study) to just 16.2% (MSCI data). I previously concluded in a more detailed head-to-head match-up that the MSCI global data and the U.S. data since 1970 are nearly identical. For example, the U.S. resides right in the middle of the pack of 22 MSCI countries in terms of chances of a permanent loss over a 10-year timeframe. So, why do the global statistics vary so much?
War Stories – The key difference between the MSCI data and the two other global studies is the timespans of data used. Both the JST Study and ACO Study include a substantial amount of data for European and Scandanavian countries as well as Japan that comes from World War 1, World War 2, and the global economic tumult that occurred in between. While these events also impacted the U.S., those effects were mild as compared to the market and economic devastation in Europe and Japan. You can review some of these stock market horror stories in my post on the Investing Disasters and Deep Risk.
Specifically, of all the return years from all the countries included in the JST Study, nearly a third of them (29%) come from the war years where most of the fighting occurred. For the ACO Study, more than 16% of the data come from these same times and regions. In other words, for every 30-year random return sequence evaluated in the ACO Study, on average, nearly five of those years would be expected to come from war-torn Japan, Europe, and/or Scandanavia.
Fictional Stories – Further, for the JST Study, I calculated simple averages of loss statistics across all countries. So, my summary of the JST results provides a central tendency of more and less extreme outcomes across the individual countries.
But the ACO Study methods potentially compound the significance of the war years data because of their random sampling approach. The ACO Study throws all the return data from every country into one basket and then picks “blocks” of return sequences at random from that basket. They then cobble together those blocks into 1 million longer return sequences that are intended to each represent one stock investor’s experience.
Therefore, almost all of the return sequences in the ACO Study are fictional; they never occurred at any one known place or time. Worse, it’s not only possible but pretty much guaranteed that many of these million sequences include war years from one country, followed by the same war years from another country, and perhaps even more war years from a third country after that.
In other words, the ACO Study generated a large number of sequences that were hugely worse than any sequence of stock returns that has ever occurred. And by the same token, they generated many wildly optimistic sequences as well.
Because the JST Study data are freely available, I was easily able to calculate permanent loss statistics that exclude the war years by using return data after 1948 only. However, I don’t have access to the ACO Study raw data. So, for that study, I estimated permanent loss statistics for a post-war dataset by proportional extrapolation from the JST Study post-war data³. This graph compares permanent loss percentages using all data from the two studies to the calculated (or estimated) percentages when only post-war data are included.
Focusing on the 10-year statistics again, the chances of an inflation-adjusted loss from investing in global stocks drop by about three percentage points when only post-war data are used. And the lower range of these post-war percentages is pretty similar to the 16% chance of loss from global stocks that I previously calculated from the MSCI dataset. Similarly, the chances of a loss for 10 years of post-war investing for 5 of the 16 countries in the JST Study (Denmark, Sweden, Finland, Australia, and the U.K.) are less than the 13.3% chance of a loss for U.S. stocks noted previously.
You might accuse me of “mining” the data to support a prejudiced opinion that stocks aren’t so scary. But regardless of any potential bias, this exercise shows that a substantial portion of the differences between U.S. and global stock returns can be attributed to a single catastrophic event that had a relatively minor impact on the U.S.4
What’s The Alternative?
Moving beyond the realm of stock returns, most of these scary stock mysteries fail to address another major piece of the investing puzzle. Namely, if we should be scared of and avoid stocks, what assets should we choose instead for long-term investing?
For example, the ACO Study plugs their results into a standard asset allocation model based on “risk aversion” and concludes that investors should allocate more to bonds and less to stocks. Long-time readers know that I’m averse to risk aversion as a useful method for asset allocation.
But more to the point, such an analysis fails to judge stocks and other “safe” assets like bonds using the same standard of permanent losses. Here’s the same permanent loss graph as before but I’ve added the chances of an inflation-adjusted loss from investing in U.S. Treasury bonds and U.S. cash (3-Month T-Bills).
So, a retreat from stocks and towards bonds or cash is an irrational march towards ever higher chances of a permanent loss over every investing timeframe. That’s because the relatively poor returns of bonds, and particularly cash, have a difficult time keeping pace with the long-term erosive effects of inflation.
Some might say that it’s more useful to instead think of bonds as a guard against rare investing disasters such as world wars. But I can refute that with a few graphs showing the inflation-adjusted performance of one unit of local currency (“$1”) invested in several countries from my post on investing disasters and deep risk. These examples should be adequate to remind everyone that bonds have historically performed enormously worse than stocks during and after catastrophic events.
Granted, I picked four countries that were particularly devastated by the world wars. But I couldn’t find a single country in the JST dataset where bonds performed substantially better than stocks through the war period.
Conclusions
Considering all that, here is my solution to the mystery of scary global stocks when we invert the three ACO Study conclusions into question form.
1. Are the long-term outcomes from diversified equity investments highly uncertain? Yes. Stock returns are “uncertain” simply because they are more volatile than bond and cash returns. But I disagree that stock returns are “highly” uncertain because the ACO Study methods compound the apparent uncertainties with stocks by creating many extreme outliers that never existed in history.
2. Are catastrophic stock outcomes common with 20 to 30-year horizons? No. Even if we take the worst permanent loss statistics from all three global stock studies, the chances of any loss at all are only 20% to 10% over these longer timeframes. This necessarily means that the chances of catastrophic losses are far less. Again, the ACO Study methods create a greater proportion of more catastrophic outcomes than the return histories from any individual country would indicate.
3. Is the historical developed market stock market performance notably worse than the U.S. experience? After excluding the extreme impacts of the war years, it appears that global stock markets, in aggregate, still have a somewhat higher potential for permanent losses (16% to 22% chance of a loss for a 10-year timeframe) as compared to U.S. stocks (13% chance). However, given that the U.S. results often fall somewhere within the pack of developed market country results, it seems like a stretch to say “notably worse”.
And all this still ignores the question, “What’s the alternative?” It seems clear that trading the uncertainty in stocks for the greater certainty of losing inflation-adjusted money with bonds and cash is an irrational response to a relatively minor problem.
1 – When I use the term “global”, I’m referring to Developed Market Stocks and not Emerging Market or Frontier Market Stocks. By their very nature, the historical data for emerging and frontier markets is too erratic and covers too short a timeframe to be useful in these types of historical data analyses. So, Developed Market Stocks are the necessary representative for “global” determinations about stocks.
2 – A little further on in this post, you may notice that I present the chances of losing inflation-adjusted money when investing in U.S. stocks for 10 years at exactly 13.3%. One reason I round this number to “about 10%” is that if U.S. return data going back to 1879 are used instead, the result is right around 10%. Another reason is that all these data are based on historical estimates with inherent uncertainties that deserve such rounding in any headline introduction.
3 – I’ll freely admit that this is a rough and ready estimate of the impact of the war years data on the ACO Study because of the differences in the datasets and methods between the two studies. But without the raw ACO data, it seemed like a reasonable way to generate a ballpark estimate.
4 – I think it’s safe to define WW 1, the interwar period, and WW2 as one event given that many historians view this period as two acts of the same play with a terrible intermission in between.
