CRSP US Treasury Database covers daily data since 1961 and monthly data going back to 1925.
@ricke Maybe Mr. B's Figure 8.5 would be a better set of data to look at. Starting in about 1970 to present, international did do better in the 1980's.
Bengen's new results are interesting as a data point, but I don't find them that useful overall. My problems are with what he does within his methodology, and his methodology itself.
1) I am curious about how reliable the international data is before 1970. If his results are higher using these specialized assets like micro-cap, it is likely due to having taking more risk (that happened to pay off in our particular history.) But how does his using this particular 11X5, 40, 5 portfolio avoid data picking? What reason do we have to think that mix would do better than generic going forward?
2) For a very similar analysis, I prefer Karsten Jeske's spreadsheet (big Ern's: Early Retirement Now). Not only does ERN go back to 1871 (compared to Bensen's 1926--i.e.50 years more cohorts), but he breaks things down various cohorts who retired at CAPE record high, and the SP500 being record high.
If you look at my quick n dirty attempt to incorporate Bengen's portfolio in ERN's failsafe withdrawal rate spreadsheet, you'll see that overall his results agree with Bengen on the average. (In fact, ERN's failsafe withdrawal amounts are much higher for many cohorts--something Bengen notes as well.) What Bengen calls "Safe Max" is what Jeske calls "Failsafe".
But.... Look what happens for the cohorts who retire at CAPES > 20 and/or SP500 close to record high: the safe-max drops to 3.98%. (3.81% if we just don't use micro-caps and mid-cap.)
(I took some screen shots, attached at bottom of message. But feel free to copy the spreadsheet, link right above, and play around with it yourself.)
3) I've said this before, but my biggest problem with the whole methodology (Bengen and ERN) is that it draws conclusions from exactly one unfolding of history in a particular order. I mean it looks impressive to say you had cohort 1 going 1928-1958, cohort 2 going 1929-1959 all the way up to 1975-2014. But there is so much overlap in the same events in the same order that really isn't as much "testing" as it seems. The Depression > US dominance > Vietnam/Korea > Crises > Covid etc. For each cohort events unfold in exactly the same way. And that is why 1965 has such a big importance to Bengen in his safemax. But if I'm looking at the future I'm fearful of many different types of "1965s" if you will.
If we're going to look at past history, I prefer a bootstrap that "shuffles" the order of history's 5 year blocks. It's still limited by the way history unfolded (which is only one way history could have unfolded), but at least it generalizes that. And we know no matter how bad things were, go long enough and something worse will come along.
So, yes, the new Bengen is interesting. But I don't trust my retirement on his results, especially with Jeske's analysis of cohorts that retired with record-high SP500 and/or CAPEs.
Note in ERN's data that "All World Ex US" doesn't have its own data prior to 1970, it just uses the S&P500. I have seen others claim advantages for international as well, but I think researchers have discovered that their datasets can be monetized, so unlike Shiller and Damodoran that publish US results, I don't know where to find a long term dataset for All World Ex US.
From a diversification standpoint, it makes sense to have some international (and I do), I just wish we could see the data being used as it seems like the big wipeouts due to world wars would crush the historical returns.
@ricke There is an interesting free international data set from the authors of “the Return on Everything” article, Jorda et al 2017. It’s country by country, not an index per se. I forget how deep you have to go into the notes to get the data. It might be an appendix. It’s very crude, yearly, with only a few asset classes; but fascinating stuff. The article has a big impact on me, and i lowered my expected returns given how poorly many various country’s stock markets have done in history.
https://doi.org/10.24148/wp2017-25
i think pedersen et al have come to a similar conclusion, but they don’t share their data
Thanks, I recall seeing that paper (but couldn't recall the authors or title, thanks for pointing to it) and having the same reaction, that it was sobering how special US growth has been.
Like I said in an early post, I have not talked about everything that's in Mr. B's new. I think it is a little premature to discount some of his ideas unless his book has been read cover to cover.
1) There is foreign market history going back to 1900 https://mindfullyinvesting.com/historical-returns-of-global-stocks/ (also see reports by Credit Suisse), information that everybody could have access to if they look for it. I find it interesting that almost everybody on the Pralana forum has a fairly big potion of their equity asset allocation (AA) in international markets. Almost no one has questioned it's reliability before.
I don't see how his 11X5, 40, 5 portfolio is data picking, anymore then than generic, whatever that means. Can you expound?
2) It's nice to see that Karsten Jeske's spreadsheet has similar results. But you really can't complain about the lack of historical results not going back to the 1800s for stock markets and then talk about high CAPE and high stock market numbers as a reason to based your AA on. Prior to about 1997, CAPE only exceeded 20 twice, once in 1929, and again over a few years in the mid 1960's. So there is not a lot of history with high CAPE values. If you look at the 17 retires between July 1997 and July 2001 (even though they do not have 30 years of data), when the Cape was between 32 and 44, the SAFEMAX did not go below 5.4%.
3) That's a whole different discussion. History is the best data set we have. Everything else is just guessing. The world has gone though a lot in the past 100 years and we are still here. The S&P 500 has literally hit record highs over 1,250 all-times since 1950, and only recently has it done so during high CAPE values. How do you plan for an occurrence that is worst then anything that has happened in the last 100 years? I say all this in the spirit of debate knowing that I don't know anymore then anyone else 😀.
@ricke I found the link to Jorda's excel database. https://www.macrohistory.net/database/ . I was fascinated at how real estate did as well as stocks in many countries. None of this magically answers the question about US exceptionalism (that we've discussed in pizzaman's threads) of course. My own lens is to avoid the US exceptionalism debate (though I'm skeptical that I just happen to live in the one exceptional country for stocks--"where everyone is handsome and smarter than average") but rather to try to understand the inherent nature of the asset itself, in its range of returns and its volatility, and how it reacts under crises. Stocks seem very sensitive to the larger geo-political context compared to bonds and real estate. (Maybe that was obvious all along?) Which for me implies I want to diversify geographically in order to avoid that particular risk for such an vital asset.
All of this is getting a bit off topic; but, coming back to Bengen, is he correct that adding international 11% allows a larger SWR? That is a very conservative amount compared to e.g. Vanguard, Fidelity. So I think it would be hard to fault him. The more controversial issue is his adding mid, small and micro US at 11% each. (Interesting he doesn't ride the popular value/factor train.) Intuitively it seems like this adds risk, since smaller firms are less stable than established giants. So intuitively, this would lead to higher returns at the price of risk, which may or may not be relevant over 30 year time spans.
Like I said in an early post, I have not talked about everything that's in Mr. B's new. I think it is a little premature to discount some of his ideas unless his book has been read cover to cover.
Sorry if I came off as dismissive. I think this is a good topic, and I'm open to learning more. I was just offering my initial reaction based on your posts and listening to two interviews with him about the new book.
There is foreign market history going back to 1900 https://mindfullyinvesting.com/historical-returns-of-global-stocks/ (also see reports by Credit Suisse), information that everybody could have access to if they look for it. I find it interesting that almost everybody on the Pralana forum has a fairly big potion of their equity asset allocation (AA) in international markets. Almost no one has questioned it's reliability before.
The data you cited only goes back to 1970. This is not enough time to form a judgement about international for purposes of SWR.
Maybe cherry picking was too strong a term; but I would like to know more about his reasoning why he chose to add international, mid, small and micro US in equal quantities to large US. If he were trying to match the cap-weight of the total world market, it would be closer to the AA of VT etf. Is he trying to mimic the "total stock market" rather than the Sp500? If so, he is overweighing small and micro I think. Micro-cap seems especially unusual, since it is way more risky than large stocks. Why not value? Why not factors? Why not market cap? Why not foreign small? Why not real estate or gold? I'm not saying he's wrong, just wondering why he picked the exact mix he did, and why he feels that will reflect higher returns in the future.
2) It's nice to see that Karsten Jeske's spreadsheet has similar results. But you really can't complain about the lack of historical results not going back to the 1800s for stock markets and then talk about high CAPE and high stock market numbers as a reason to based your AA on. Prior to about 1997, CAPE only exceeded 20 twice, once in 1929, and again over a few years in the mid 1960's. So there is not a lot of history with high CAPE values. If you look at the 17 retires between July 1997 and July 2001 (even though they do not have 30 years of data), when the Cape was between 32 and 44, the SAFEMAX did not go below 5.4%.
Good points about CAPE. It does go higher than 20 more times if you go back to 1871, but point taken. But do note that ERN also tracks how Safemax (=Failsafe) does when one retires at SP500 at its high. This is independent of CAPE. This isn't an argument against Bengen: in his book he talks about the fact that the high valuation of the market now means one might expect less than the "average" 4.7% I think. But I like the way that ERN analyzes it out to exact numbers. ERN also uses monthly returns not just annual, so captures crises like 1929 better than Bengen.
3) That's a whole different discussion. History is the best data set we have. Everything else is just guessing. The world has gone though a lot in the past 100 years and we are still here. The S&P 500 has literally hit record highs over 1,250 all-times since 1950, and only recently has it done so during high CAPE values. .
I disagree. History is the best data we have of one flow of world events. (What would have happened if vaccines had not been successful against COVID for example? What would have happened if "monetary easing" hadn't rescued us from the total world economic meltdown? What would have happened if the US had lost WW2?)
The issue isn't data about the past; it's how much it predicts the future. And that is where assuming the future will be like the past is even more of a "guess" (to use your word) than generalizing it.
How do you plan for an occurrence that is worst then anything that has happened in the last 100 years?
There is a limit of course, since you can't plan for what has never occurred. ("Black swans"). And this is a deep question that goes beyond the topic of this thread. But one simple answer is good old monte carlo simulation: the idea behind parametrically using a normal curve is that over time the rare events ("tails") will happen. The normal curve gives you a sense of how often that will happen statistically. 1000 runs (that is 1000 alternate runs of history) gives you some concrete results in the bands. I can nitipick that the Student-t curve is a better approximation of stocks than a Normal curve. I can argue about bootstrapping. But either way, these both allow you to run 1000s of examples of history unfolding so you have a better sense of how often these rare occurrences might occur in the unknown future.
@jkandell No problem 🤗 I always assume the best intentions in what people post. As you can tell I love to debate, I find it fun, and realize that debating has a built-in adversarial base to it, but it can and should be friendly.
Mr. B doesn't go into detail as to how he picked his AA as far as small and micro-cap other then to say "In this book, I intend to focus on assent allocation only from the perspective of increasing SAFEMAX without increasing investment risk." He is a big fan of portfolio diversification which he calls a "free lunch." He also wants to keep it simple. He also makes it clear you can change your AA and get higher returns if you are willing to accept a lower SAFEMAX, which then would REQUIRE you to make adjustments in the future when things change (like a bear market).
Not sure retiring at a high S&P 500 has any bearing on long term retirement results. I think that gets into market timing.
The whole point of SAFEMAX is capturing crises. Retiring in 1929 was not the worst case situation as the 1930's experienced significant deflation. The 1970's saw several bear markets combined with high inflation making the 1968 retire the worst case. .
The "unfolding of history in a particular order" I think is too far off topic. I think we just agree to disagree 😎. There is a whole thread on Pralana forum debating historical simulation vs monte carlo (MC) simulations.
What do other people think???
I think that most portfolios are good if you stick with them.
I recently read the following post:
https://www.whitecoatinvestor.com/150-portfolios-better-than-yours/
The post starts with the following sentence:
“To help new investors realize there is no perfect portfolio and that the best one can only be known in retrospect. Therefore, they should pick something reasonable and stick with it.”
…
And ends with this:
“A good investment portfolio is broadly diversified; low-cost; mostly or completely passively managed; regularly rebalanced; and consistent with its owner's need, ability, and desire to take risk. Every portfolio (except the Kiplinger ones) in this post meets those qualifications. Pick one you like, or design your own. Just don't go looking for the best one. As Prussian General Carl Von Clausewitz said, “The enemy of a good plan is the dream of a perfect plan.”
1) There is foreign market history going back to 1900 https://mindfullyinvesting.com/historical-returns-of-global-stocks/ (also see reports by Credit Suisse), information that everybody could have access to if they look for it. I find it interesting that almost everybody on the Pralana forum has a fairly big potion of their equity asset allocation (AA) in international markets. Almost no one has questioned it's reliability before.
The data at the link you cited only goes back to 1970.
Not sure retiring at a high S&P 500 has any bearing on long term retirement results. I think that gets into market timing.
Take a look at the chart I posted in my post, screen shot from ERN's spreadsheet, reposted below. Going 1871 to 2025 Jeske look at the 30 year SWR starting when the market was at it's high. There is a direct correlation.
Retiring when the market is at a "new high" has a maxsafe of 3.98%. Retiring when the market has dropped 10% has a max safe of 4.36%. Retiring when the high as dropped 25% has a max-safe of 4.92%. Retiring when the market has crashed 50% has a max-safe of 6.38%.
Yes, you're actually safer after a crash because valuations are so low.
@jkandell Sorry about the international link, I assumed the Credit Suisse link would work, it does not 😫.
@jkandell I guess I am just dense. I still don't see how retiring on the day the S&P 500 hits a record high has long term consequences to your retirement financial planning. I think it is just a correlation with no bases. In 1946 in England, the stork population (the bird) increased 25%. During the same time period the human birth rate increased 25%, proof that storks delivered baby's. The S&P 500 has hit records highs 40 times in just 2025 alone. What if the S&P 500 hits a record high on Tuesday (don't retire). The very next day the S&P 500 drops 1% because the Feds decided not to drop interest rates (time to retire). What am I missing??
