February 12, 2024 – Last week, I wrote about how equities historically outperformed bonds by a comfortable margin. So, the principle of “stocks for the long run” is still valid. Does that mean a portfolio with 100% stocks is a good long-term strategy? That’s a recommendation from another finance research paper that’s gotten a lot of publicity lately. Three finance professors claim that a 100% stocks portfolio, 50% domestic and 50% international stocks, would have consistently outperformed all other conventional wisdom asset allocations, e.g., 60/40, glidepaths in target date funds, etc. Quite a sweeping claim! They claim they have the empirical evidence to prove it.
I have my doubts, though. Let’s take a look…
Will 100% Stocks reliably beat a Stock/Bond portfolio?
The paper in question is titled “Beyond the Status Quo: A Critical Assessment of Lifecycle Investment Advice,” written by Aizhan Anarkulova, Scott Cederburg, and Michael S. O’Doherty. It’s available for free at SSRN. To save space, I will call the paper “BTSQ” (Beyond the Status Quo) from here.
Let me start with a few things I really liked about the paper.
First, with my own asset allocation philosophy, I’m certainly closer to the 100% stocks end of the spectrum. 100% stocks worked well for me while accumulating during my career from 2000 to 2018. Despite the S&P 500 returning only 3.2% between 8/2000 and 5/2018 (after inflation, plus dividends reinvested), my internal rate of return (IRR) was much higher because I kept contributing to my portfolio during the sharp downturns, especially in 2002 and 2008/9. Dollar Cost Averaging worked in my favor. While contributing regularly, investors shouldn’t be too concerned about volatility and drawdowns.
I also never cared much for the traditional glidepath design; see my post “What’s wrong with Target Date Funds?” especially the following items:
- Young investors should hold 100% stocks, while the TDF upper limit is 90% stocks. That’s due to ERISA regulations, I believe. Sophisticated investors can even get better results with some leverage, i.e., reach returns with stock-expected returns but at lower volatility than a 100% stock portfolio. More on that later.
- TDFs likely reduce the equity allocation too early, sometimes 20 years before the planned retirement date. You have two or three more full stock market cycles during that time. It’s best to milk the higher stock returns for a while linger. In fact, I’ve written in Part 43 of my SWR Series (“Pre-Retirement Glidepaths: How crazy is it to hold 100% stocks until retirement?“) that investors with a high risk tolerance and/or some flexibility about the exact retirement date would do better keeping their equity allocation at 100% stocks much longer and maybe even until retirement.
- TDFs use the wrong post-retirement glidepath. Contrary to popular belief, a glidepath that takes equity weights up(!) again during retirement offers a hedge, albeit only a partial one, against sequence risk. See my work in Part 19 and Part 20 of my SWR series. Most TDFs further shift out of equities post-retirement.
Thus, I’m unsurprised when the BTSQ paper finds that the stereotypical target date fund creates whacky results. We should thank the authors for confirming what my readers and I already knew.
I also applaud the authors for proving that Wade Pfau’s calculations for international safe withdrawal rates, the paper referenced last week, appear less scary if people had simply used international diversification. Admittedly, German investors in 1914 would have had difficulties moving their money into a diversified “rest of the world” portfolio. But then again, not many Germans were using a balanced domestic stock and bond index fund portfolio for retirement planning either. So, all of the calculations, whether in my SWR Toolkit or the BTSQ paper, should be viewed not as studying how people in the early 1900s used financial assets for retirement savings – they likely didn’t! – but more of a thought experiment of how today’s retirees would fare if historical asset returns were to repeat themselves.
A case study: Retiring at the 1972 stock market peak
Next, I tested the 50% domestic plus 50% international stock allocation recommendation with my Safe Withdrawal Rate toolkit. Unfortunately, I have only monthly data for non-US equities since 1970, the start of the MSCI-World-ex-US index. But it’s enough data to test how this strategy would have performed if you had retired right at the early 1970s (month-end) stock market peak before the first oil shock, i.e., on 12/31/1972. I simulate four different portfolios:
- The BTSQ portfolio: 50% US stocks (S&P500) and 50% MSCI World-ex-USA
- 100% stocks, US only (S&P500)
- 75% US stocks, 25% US 10-year Treasury benchmark bonds
- 60% US stocks, 40% US 10-year Treasury benchmark bonds
I must concede that the strategy performed surprisingly well. I plot the simulated portfolio values in the chart below, normalized to 1.0 in 1972 when you withdraw 4% p.a., i.e., 1/3 percent monthly. After a tumultuous thirty tears, the BTSQ portfolio comes out far ahead of the other three portfolios, with about 80% of the portfolio remaining, even when adjusting for CPI. In contrast, the other three domestic-only portfolios finished at 30% and 40% of the initial assets. That’s still an impressive outcome because it means that the 4% Rule did not fail during those volatile thirty years with five recessions: 1973-1974, 1980, 1981-1982, 1991, and 2001. But the BTSQ portfolio mopped the floor with the other three portfolios using only US assets!
What caused the impressive performance of the internationally diversified portfolio? Did non-US stocks outperform US stocks (and bonds) by that much? Not really! If we look at the buy-and-hold return of the four asset allocations, they all end up more or less in the same spot after thirty years. The real, annualized CPI-adjusted returns were 5.07% for the BTSQ portfolio, 5.38% for the 100% S&P 500 portfolio, 5.30% for 75/25, and 5.16% for 60/40. So, the BTSQ portfolio would have delivered the worst(!) outcome for a buy-and-hold investor, though by only a small margin. However, because international stocks outperformed the S&P 500 early during the simulation period, the BTSQ portfolio suffered less from Sequence Risk. Also noteworthy is that a large part of the MSCI ex-USA outperformance vis-a-vis the S&P 500 came from a significant US Dollar depreciation between 1973 and 1980. So, the non-US stock markets suffered just as much during the 1970s; it’s just that exchange rate movements cushioned part of the fall. More on that later.
So much for the parts of BTSQ I liked. Let me get to the parts that I find troublesome. However, before I get into this, let me post this important disclaimer: What I write here is polite and very civilized. If some faint-hearted folks complain that I am being mean toward those three academics, I can assure you that a) they likely don’t care what I think about their paper and b) my level of criticism does not even come close to how harsh and cruel journal editors, journal referees, and seminar and conference participants would routinely shoot down academic papers. If you’ve ever sat in an academic seminar with Ed Prescott (God rest his soul), Tim Kehoe, Pat Kehoe, Jose Victor Rios-Rull, and many other characters from my good old academic times, you’ll know what I’m talking about. So, there is no need for fans of the BTSQ methodology to get upset on behalf of the BTSQ authors. Those three finance professors are just fine with or without my analysis here.
With that out of the way, let’s take a look at the parts I didn’t like much:
1: Notice the methodological differences!
If you are familiar with my research on Safe Withdrawal Rates, it’s almost entirely based on historical simulations using U.S. data. I recently added monthly MSCI World return data to my retirement simulation toolbox (see Part 28 of my series for a guide and the link), but the series only started in 1970. Also, notice that all simulations use consecutive historical data, so if you want to simulate a 30-year retirement horizon, you can study the historical cohorts between 1871 and 1993 with an actual 30-year realized return series.
In contrast, the BTSQ paper covers asset returns from several dozen developed nations and then randomly draws blocks of realized return data to simulate potentially millions of years of asset returns. Notice, of course, that the draws in any particular month are from one and only one country. So, for example, the returns for domestic stocks, international stocks, domestic bonds, and bills in month 55 are the actual realized returns in July 1981 in, say, France. You want to keep the returns within the same country to preserve the correlations of assets. Moreover, the authors also draw blocks of consecutive monthly returns from that same country. Thus, month 56 would then be again French returns but from August 1981. This replicates crucial features of serial correlations, crashes, subsequent swift recoveries, etc., that often get lost in plain Monte Carlo simulations with independent draws. The length of blocks is 120 months on average to ensure we cover a full market cycle in one country. Also, notice that the international return is different from the perspective of each country, e.g., from the perspective of German investors, the international return is a weighted return on all non-German stock markets that month, taking into account exchange rate fluctuation and net of the German consumer inflation rate.
The paper also assumes a stochastic earnings and longevity path, replicating both average life-cycle earnings trends and idiosyncratic earnings shocks. The stochastic income assumption is certainly neat. It utilizes the work of Prof. Fatih Guvenen, a world-renowned researcher in this field who works at my alma mater, the University of Minnesota. Since I’m mainly interested in the withdrawal part, the income volatility during accumulation is not really my main concern. And if I ever study simulations during accumulation, I’m fine with using a flat contribution profile. I doubt that stochastic income shocks add much to the analysis and suspect that this feature is in that paper as pure bells and whistles and name-dropping without much actual use in this context. And I say this as someone who has published academic papers on the topic of stochastic idiosyncratic household shocks. Two of my own academic papers, both published in the Journal of Monetary Economics (“Housing, mortgage bailout guarantees and the macro economy” and “U.S. tax policy and health insurance demand: Can a regressive policy improve welfare?“), deal with this issue with much more sophisticated computation methods because we not just simulated. We optimized path-dependent actions. This involves infinite-dimensional optimization problems using sophisticated numerical techniques, which are absent from the BTSQ paper. More on that later.
2: The 100% stocks strategy would have severely backfired in many other historical cohorts!
If I rerun my safe withdrawal rate simulations with a starting date at the peak before the dot-com crash, international stocks no longer look so good. Let’s assume retirement had started on 3/31/2000. We can study how the four different portfolios would have performed. The 100% stocks portfolio would be about 12.5% of its initial value. 50%/50% domestic/international stocks would not have made a huge difference. In contrast, the 75/25 and 60/40 portfolios are doing quite well, with 45% and 55% of the original principal remaining. If we keep withdrawing 4% of the initial amount annually, it looks pretty certain that both all-equity portfolios will run out before the 30-year mark (a little more than six years from 12/31/2023), while the 60/40 and 75/25 portfolios will likely survive. In this particular scenario, bonds would have provided much better diversification.
Just like for the 1973 cohort, let me plot the buy-and-hold returns, i.e., start at a portfolio of 1.0 and let the portfolio grow without withdrawals. Now, the BTSQ portfolio finishes dead last, even behind the 75/25 and 60/40 portfolios, while the 100% U.S. equity portfolio has the highest final portfolio value.
The picture looks even worse for the 2007 retirement cohort. Please see the simulated portfolio values for 2007-2023 below. The 100% US equity portfolio would have performed the best for this cohort, though with a crazy almost 60% drawdown in 2009. The 75/25 and especially the 60/40 portfolio would have beautifully cushioned the fall during the global financial crisis, though they would have also fallen behind the all-US equity portfolio thanks to the strong subsequent rally. A distant last is the BTSQ portfolio.
Instead of looking at retirees, we can also look at the performance of the BTSQ strategy during accumulation. I have return data from 1970 to 2023, so let’s check how 30 years of accumulation would have worked out. I abstract from the stochastic income process as in the BTSQ paper and rather focus on a simple $1 monthly contribution to an equity portfolio for 360 monthly. One portfolio has 100% domestic equities, and the other has 50% domestic and 50% international. Since MSCI index levels start on 12/31/1969, I can only display ending dates from 12/31/1999 to 12/31/2023. Sure, we’re missing the Great Depression and the 1929-1932 bear market. But that shouldn’t be a huge problem because that was a deflationary shock, while the main intuition in the paper for why international stocks diversify so well is the inflation story! And 1969 to 2023 covers the two major inflation shocks in U.S. return history: the 1970s and the post-pandemic inflation shock. So, international stocks should do really well, right? Wrong! A 100% allocation to the S&P 500 TR index would have consistently outperformed the BTSQ-style investment, with a 50% on the MSCI World-ex-USA.
In fact, the BTSQ portfolio would have sucked so badly it even underperformed a 75% US Stocks plus 25% US bonds portfolio most of the time. So much for international stocks being superior at hedging inflation risk! See the chart below:
So, the BTSQ strategy would have had a spotty record for a U.S. investor. During accumulation, you would have done better with 100% US equities. During decumulation, two out of the three market peak retirement cohorts would have fared better with a 75/25 or 60/40 portfolio than the BTSQ portfolio.
3: The study has limited relevance for U.S.-based investors.
Why do the authors propose such a mishmash of return series? They posit that this will alleviate some of the survivorship bias (their claim, not mine) inherent in U.S.-only financial market data. I call b.s. on that one, though. U.S. assets have outperformed some(!) European assets between 1890 and 2019 because we have never been invaded and/or destroyed during the time span studied in their paper. We have never abolished capitalism and become a communist country. So, conditional on living in a (relatively!) safe and well-run country like the U.S. or most of Western Europe today, I see little use in feeding German stock and bond data from the Weimar Republic and Third Reich or Czechoslovakian pre-communism data into my retirement planning toolkit.
The following analogy may help make my point: Introducing motorcycle helmets has certainly increased life expectancy. Should everybody own a motorcycle helmet, then? No, that would be a fallacy! I don’t ride motorcycles and thus don’t need one. Buying one would be a financial mistake. In other words, the benefits of motorcycle helmets depend on one crucial characteristic: do you ride motorcycles or not? The same applies to asset allocation: a large international stock allocation would have hedged your investment risk in the war-torn countries in Europe and Japan. But I don’t believe my home country will suffer the same fate! In fact, not only have we not had any blowups in the past, but the good ol’ US of A is also the one country that’s in the best position to stay that way, as Ben Carlson on the excellent Wealth of Common Sense blog recently pointed out in this nice post. As Warren Buffett always says, “Never bet against the U.S.!”
In any case, it is already a bit of a stretch when I perform robustness analysis for how my current portfolio will fare if we replayed the Great Depression or the 1970s in the U.S. It’s quite another leap of faith to feed WW1 and WW2 German return data into my analysis. It borders on insanity to believe that over the next 40-50 years of my retirement, we will have US stock and bond market returns that are drawn from a distribution that includes WW2 data from Germany, Czechoslovakia, and Japan.
Well, the saving grace for the paper would be that while it’s not that useful for U.S. investors, it should be all the more useful for the rest of the developed world, i.e., Germany, U.K., Canada, Japan, etc., right? Unfortunately, I’m not so sure about that either, which brings me to the next issue…
4: The study isn’t that relevant for other countries either.
Unless you believe that Germany, Italy, and Japan will again go through the same destruction as in the 1940s, we can safely ignore even their own early historical data. In other words, for today’s investors in small European countries, including all the WW1/WW2 data in the bootstrapping process seems inappropriate.
But don’t get me wrong: If your country has a lot of idiosyncratic risk and a market capitalization of only 1% or so of the global equity market, you should be diversified. So, 50% domestic is likely still too high for investors in small countries like Iceland or Portugal. The authors correctly mention that they can push the envelope even slightly higher when they shift the equity portfolio to 35% domestic and 65% international because so much of their 2,500-year county/return sample comes from countries with tiny market capitalizations.
5: The inflation story is suspect.
In my simulation toolkit, you can check where things went wrong if you get unexpected/surprising results. Sometimes bonds are a great diversifyer (1929), sometimes not so much (1970s). With a black box methodology like in the BTSQ paper, your opportunities to learn and understand what’s happening are limited. Even if you dig through the return data, what can you really learn from looking at a million years of return data cobbled together in this unworldly bootstrapping technology? Something like this…
“Oh yeah, here, between June and July of year 457,856, we jumped from the 10-year block of Belgian data from the 1950s to Japanese data from the 1920s. A 100% stocks portfolio just totally killed it. Take that, stupid 60/40 portfolio!”
Said by… nobody ever!
What can you really learn from that? So, when explaining their findings, the BTSQ authors attribute their results to the correlation between asset returns and inflation. Specifically, the authors argue that domestic bonds suffer the worst during inflationary events. Domestic stocks get dinged as well, but to a lesser degree, while foreign stocks have the lowest correlation with domestic inflation. Thus, we should replace domestic bonds with foreign stocks to better protect against inflation shocks. Case solved! That story sounds intuitive at first glance. But it doesn’t make much sense after closer inspection. Here are four reasons:
1: Of the four worst stock market events in the last 100 years of U.S. market history, three were accompanied by deflation or at least disinflation: 1929-1932, 2001-2002, and 2007-2009, while only 1972-1982 saw an extended and significant inflation shock. Those same bear markets were felt all across the globe. It’s hard to argue that international stocks would have offered better diversification than domestic bonds during the deflationary recessions. I already proved that for 2001-2002 and 2007-2009. If someone wants to share their international equity data from 1929 to 1932, I will also gladly confirm that for that period. Inflation alone cannot be the explanation, at least not from a US investor’s perspective.
2: Even during the one event that seemingly fits the BTSQ inflation narrative, i.e., the inflationary 1972-1982 period in the U.S., their story doesn’t hold water. The inflation shock came from the oil embargo and was felt internationally. In the U.S., you would have benefited from international stocks because the USD weakened, and thus, the international equity portfolio recovered faster than all your domestic assets. But that also means that in all those non-US countries (from which BTSQ would have drawn in the bootstrapping method), had you invested in international stocks, you would have felt the flipside of the FX move; international stocks, heavily tilted toward US stocks in US Dollars, would have underperformed your domestic assets.
3: One possible explanation to account for their results: I suspect that many non-US economies experienced more inflationary recessions. But in those countries, inflation was often not the cause but a symptom of a larger problem, i.e., a Weimar Republic-style failure of a country with rampant inflation, where all domestic assets go to essentially zero. If you fear such a scenario in your country, you should certainly diversify internationally. If you live in the U.S., you can probably discard that possibility, again, for the same reason as stated above and in my post last week: We now live in a more integrated, connected, and safer global economy. If you’re afraid that the U.S. will soon experience a Weimar Republic-style economic collapse, then international stocks will get hammered just as badly or worse. If you want to hedge against such a scenario, buy a bunker, guns, ammo, and dried foods, not Belgian equities!
4: Quite amazingly, if I calculate the correlations between inflation and the asset class returns between 1970 and 2023, I get almost identical correlations: -0.44 with 10-year Treasuries, -0.28 with domestic equities, and -0.18 with international equities. I also report the, in my view, more meaningful correlations with changes in annual CPI inflation because you’d expect that last year’s realized inflation is already factored into asset prices. The inflation surprise is more telling than the absolute inflation figure. This is consistent with my findings in Part 51 of the SWR Series: inflation isn’t the problem in retirement. Rising inflation is! High but falling inflation can be a boon to your retirement finances! In any case, despite those correlations, BTSQ would vastly underperform in the 2000 and 2007 retirement cohorts and deliver subpar results for all of the 30-year accumulation windows over the 1969-2023 time span. In other words, I have similar correlations, but the BTSQ 50%/50% equity portfolio performed poorly. So, that tells me that the inflation correlation is not a good explanation for their findings.
6: The Clifford Asness Critique still applies
If you were one of the five or so readers who recently “bugged” about the paper, you would recall that my short answer always referenced Cliff Asness’ paper on the 100% stocks discussion. Asness makes the valid point that a 100% stocks portfolio is likely an inefficient way of raising your risk. You’d do better by leveraging the maximum-Sharpe-Ratio portfolio. How do we do that? I took the monthly (nominal) returns from 1/1970 to 12/2023 and calculated the realized CAGR and risk (measured as annualized standard deviation). Let’s draw the efficient frontier between 100% 10-year Treasury bonds and the BTSQ portfolio; see the chart below. Along the frontier, I maintain a 50/50 domestic/equity constraint, so, for example, for 68% bonds, you’d be constrained to have 16% domestic and 16% international stocks. You could have pushed the frontier a bit higher without that constraint, but let’s roll with the BTSQ assumption for now. Also, notice that the efficient frontier starts at the min-vol portfolio. The part that bends southeast from there (the dotted line) is not efficient. You could beat the BTSQ portfolio by a) matching the CAGR but with much lower volatility (about 77% bonds, 64% equities, and -41% T-bills) or b) matching the BTSQ risk but with a much higher CAGR (97% bonds, 91% stocks, -78% T-bills). Or combinations in the blue slice Northwest of the BTSQ portfolio with a combination of higher returns and lower risk.
Also, notice that the portfolio doesn’t need to be fixed over time. For example, you’d ideally move along the optimal portfolios, probably with more risk/return during accumulation, and then scale back a bit again during retirement.
I concede, though, that this type of financial engineering might be a bridge too far for most retail clients. I don’t think we’ll see this anytime soon in Target Date Funds. But for the record and completeness, I need to point this out.
7: Valuation matters
Another reason I prefer my approach utilizing historical U.S. simulations is that I’m not subject to the common retirement planning fallacy of calculating only unconditional success and failure probabilities. Imagine you are at a fresh stock market peak with a Shiller CAPE Ratio above 30. You know, as in right now. I would find it irresponsible to use the bootstrapping methodology in the BTSQ paper that would randomly pick a 10-year window of return from its historical record. Given the elevated CAPE ratio, the probability of tagging on another ten years of a 1990s-style bull market seems very remote now. The chart I always use to demonstrate this point is the following: see Part 50 of my SWR Series. The 4% Rule fails when the CAPE is high (or 1/CAPE = Shiller earnings yield is low).
The opposite is also true. Imagine you are at the bottom of a deep bear market, with Shiller CAPE in the single digits. The probability of another 80+% drop in the stock market, like in 1929-1932, is very unlikely. Thus, recommendations for safe withdrawal rates should always depend on market conditions, especially equity valuations. The same logic extends to asset allocation recommendations. So, I don’t take BTSQ’s unconditional 100% stocks recommendation in retirement very seriously.
8: Violating the Bellman Principle of Optimality
Whacky results can be due to whacky data or a whacky model. Or both, which is what I suspect we have here. In other words, imagine if BTSQ were to clean up their inputs by including only post-WW2 data from the disaster countries while maintaining all data from the countries that survived WW1 and WW2 relatively unscathed. You would probably still get whacky results due to a significant logical flaw. Let me illustrate this with the following example:
Imagine we had only one single 70-year return data series covering 40 years of accumulation and 30 years of decumulation. Imagine picking between two asset allocations: A1: BTSQ (50% domestic and 50% international equities) and A2: 60% domestic equities plus 40% domestic bonds.
Imagine A1 accumulates $2m during the 40 years, while A2 accumulates $1.6m. It’s expected because you took much less risk and also likely got lower returns with A2. Also, imagine that A1 allows you to withdraw 4% of the initial nest egg (subsequently adjusted for inflation), while A2 allows you to withdraw 4.5%. In numbers, that’s $80,000 p.a. under A1 and $72,000 under A2. For example, in Table X in the paper, the US-based Balanced Portfolio (60% domestic stocks, 40 domestic bonds) had the lowest failure probability (4.5%), lower than 100% domestic equities (5.0%), and the BTSQ 50/50 portfolio (5.2%). See the Table below:
From the perspective of a young saver just starting out, if you must pick an asset allocation and stick with it for your entire life, you’d prefer A1. But that’s a stupid assumption. You’d be better off accumulating under A1 and withdrawing in retirement using A2, essentially reoptimizing at your retirement date. You accumulate $2m and can withdraw $90,000, much more than the $80,000 if you stick with A1. I suspect that a similar dynamic is at work here. The BTSQ allocation will likely beat 60/40 during the accumulation phase by so much that even though 100% stocks is a bit risky during retirement and vastly suboptimal (especially for a CRRA utility function with gamma=3.84), the larger initial nest egg more than makes up for that. So, using the BTSQ methodology, a 100% stocks portfolio looks better than 60/40, even though, conditional on being in retirement and a given nest egg, a 60/40 portfolio likely performs better. At least in the U.S. and likely other non-WW2-destroyed countries.
For the math wonks: The 100% stocks allocation throughout your entire life does not satisfy the Bellman Principle of Optimality:
“An optimal policy has the property that whatever the initial state and initial decision are, the remaining decisions must constitute an optimal policy with regard to the state resulting from the first decision.”
Bellman, R.E. (1957), Dynamic Programming, Dover.
In a nutshell, to convince me that a portfolio is “optimal,” I’d need to see some actual optimization, especially dynamic, i.e., time and even path-dependent optimization. Right now, the BTSQ paper proposes exactly eight different portfolios. And Portfolio #8, 50% domestic and 50% international equity allocation, is set up to beat them all because the competition is designed to fail:
- Portfolio 1 (TDF) shifts out of equities too early. And in retirement, the initial equity portion in retirement is too low and then shifts even lower, which exacerbates Sequence Risk.
- Portfolio 2 (60/40) accumulates way too little during the 40 years of accumulation, even though the strategy might perform all right during retirement only.
- Portfolio 3 (60/40, but with half/half domestic/international): Same problem as Portfolio 2.
- Portfolio 4: (stocks=120%-age): Too meek during accumulation. Shifting out of equities in retirement exacerbates sequence risk. Domestic-only equities are problematic in the WW1-WW2 ravaged countries.
- Portfolio 5: (stocks=120%-age, half/half domestic/international equities): Too meek during accumulation. Shifting out of equities in retirement exacerbates sequence risk.
- Portfolio 6 (100% bills): Destined to fail due to low accumulation and low returns in retirement.
- Portfolio 7 (100% domestic equity): destined to fail in the disaster countries during WW1-WW2.
So, unless you show me a lot of additional dynamic asset allocation strategies, there is no proof that Portfolio 8 is optimal. For example, something like a Kitces Bond Tent (though ideally shifted higher to 100% starting and ending weights, 55% at the low point) would likely hedge some of the Sequence Risk around retirement; see the chart below. This would almost certainly beat the “100% stocks all the time” strategy. Notice that this bond tent is characterized by three parameters: the dates for the two kink points and the equity weight at the retirement date. If we run this annually, there are 40 possible kink points before retirement, 30 after retirement, and 100 different asset allocation percentage points (0% to 99% in 1% steps). Thus, there would be 120,000 different bond tent shapes. And they are just the linear transitions. There are infinitely many more non-linear shapes. And then, for each shape, we’d also have to optimize the domestic vs. international equity allocation.
Or what about a simple 75/25 allocation during retirement, transitioned from a 100/0 during the last ten years of accumulating? I wonder if that would easily beat the 100% stocks strategy. My personal experience with U.S. data is that 75/25 is a nice compromise that often gives you the best failsafe withdrawal rate. Pick a higher bond share, and you do better in 1929, but the 1968 cohort looks bad. And, vice versa, a lower bond share might help you in the 1970s but hurt you in 1929.
Also, notice that I’ve only proposed time-dependent equity allocations so far. Richard Bellman purists (note that there’s a “Society for the Appreciation of Bellman Equations” Facebook Page) would probably shoot me for being so sloppy because truly optimal asset allocation policies would be time and path-dependent (i.e., dependent on the path of past returns and idiosyncratic shocks). But before even going there, it would be worthwhile for the BTSQ authors to at least check some of the newer asset allocation ideas, like the bond tent. You can’t just propose one single strategy and call it optimal when you only compare it with seven other cockamamie and destined-to-fail asset allocation strategies. That’s no proof of optimality! Optimality is vastly more complicated in a dynamic setting than in a one-shot portfolio optimization!
Conclusions
The BTSQ paper is a stark example of the bifurcation between practitioners and academics in finance. None of the research on my blog, even if I were to package it into a nice working paper with all the references, would ever find a willing audience in the academic world. Some practitioner journals, maybe, but the academic world would scoff at my low-tech toolkit. For them, I’m too much of a practitioner. A blue-collar financial economist.
But the opposite is true, too. The white-collar finance folks in academia have graduated from running mere historical simulations. Lots of bells and whistles. Including asset returns from small, irrelevant countries like Portugal, Iceland, Finland, etc. is sold as “adjusting for survivor bias.” Don’t get me wrong; Portugal, Iceland, and Finland are beautiful places. However, their asset returns do not represent what I expect for my personal USA-based portfolio. I understand that academia-finance wants to avoid being called too USA-centric and appeal to a broader audience. But when mushing together 2,500 years worth of country returns, including all the country failures due to wars, communism, and fascism, it’s a step too far. You gain more international appeal but lose the U.S. retail investor market. It’s like in marketing: if you try to appeal to too large an audience, you risk pleasing nobody and losing your most committed and devoted customers.
The drawback of this strange “return data casserole,” namedropping, and computational show-and-tell: the BTSQ research is a black box: inaccessible, non-repeatable, unintuitive, and thus uninteresting to the average U.S. investor. No individual U.S. investor, adviser, or any other practitioner will take this research very seriously. The BTSQ paper is unsuitable for my financial planning needs, and I can’t honestly recommend it to my readers.
And by the way, despite all academic mumbo-jumbo, the BTSQ paper misses the one issue that I would have hoped all these smart academics would have included, i.e., solving for a time-dependent, maybe even path-dependent optimal asset allocation policy function that actually honors the Bellman Principle of Optimality. So, even if I were to put on my old academic hat again, I wouldn’t take this paper very seriously either. Compared to my old research published in the JME 10+ years ago, dealing with far more advanced household decision problems necessitating solving a full Bellman equation with a multi-dimensional state space, the BTSQ methodology is child’s play. I’m glad that large parts of the personal finance blogging community are on the same page:
- Banker on Wheels (Part 1)
- Banker on Wheels (Part 2),
- Cut the Crap Investing
- Clifford Asness trashed the BTSQ paper with a devastating one-liner, “It’s not financial analysis; it is finger painting.” Why didn’t I think of that first? Ahh, that’s why Asness makes the big bucks!
- Larry Swedroe tried to debunk the BTSQ paper on Morningstar, though not very convincingly.
That said, I agree that investors should take more risks during accumulation. Target date funds shift out of equities too early. There is too much “CYA” in the financial world. 100% stocks would be one solution. Sophisticated investors may look into the max-Sharpe Ratio portfolio and lever that up for an even better risk vs. return tradeoff. And you have the bragging rights that you’re doing what Cliff Asness of AQR hedge fund fame proposed. For US-based retirees, you’d be insane to keep 100% stocks, whether domestic, international, or 50%/50% mix. There is certainly a case to be made for a glidepath toward higher equity shares, but starting with 100% stocks would have catastrophically backfired during the deep deflationary events (1929-1932, 2000-2002, and 2007-2009).
The BTSQ paper is mostly a cute academic study with some fancy methodological bells and whistles. It will be published in a good finance journal. But it does not apply to actual retirees today, certainly not in the U.S. But who knows? Maybe one day, I realize that future US equity and bond returns might look like 1914 German returns! If my Fidelity statements start coming in with an early-1900s Kaiser Wilhelm II stamp one day (with a pickelhaube helmet!), I might dig out the BTSQ paper again.
Thanks for stopping by today! Please leave your comments and suggestions below.
Title picture credit: pixabay.com
Safe withdrawal rates, by definition, result from worst-case outcomes in the investigated samples. From this perspective I find it inconsistent to simply dismiss a significant portion of worldwide historic samples in this study. With the same justification you could dismiss the great depression or the stagflation area from the US-samples because no one is going to repeat the same mistakes today.
For similar reasons I find it difficult, trying to refute the results of this study by just looking at post 1970 international data. This is only a small step away from just extrapolating the last years of US tech dominance in the stock market.
In my analysis I mostly study worst-case scenarios, correct. The BTSQ paper uses a CRRA utility function. Whatever you go with, we should avoid worst-case scenarios that are irrelevant today. There will be no Russian tanks (or Chinese tanks, for that matter) rolling through Washington D.C. and there will be no Russian soldiers raising a red flag on the Capitol. Take my word for that. If you see that in store for your country again (Germany, I presume) then go ahead and use the BTSQ paper. But the rest of us can safely ignore that.
You are talking about consistency, but you can’t even stay consistent in your own 2-paragraph comment. On the one hand, you want me to exclude the Great Depression. On the other and you criticize that I use only post-1970s data. What is it? In my SWR Series I look at longer horizons, but due to data availability I presented only 1969-2023 in this post.
The Great Depression is an outlier but a lot more possible than the Russian tanks in D.C. scenario. It belongs in the sample because in 2009 people were genuinely concerned about a repeat of the 1929-1932 scenario.
I also stated the reason why I ran only 1969-2023 simulation: MSCI data availability. The 1929-1932 episode was deflationary, so bonds performed very well and international stocks also got hammered during that period. It’s a scenario that would neither fit the inflation narrative nor actual data. I offered to simulate that date range if anyone can provide me the actual data, but until then, I can only surmise that U.S. Treasuries (+39.6%, CPI-adjusted between 1929 and 1932) did very well compared to US Stocks (-79%, CPI-adjusted). If you can show me non-US stock markets that fetched +40% during that time, please let me know! Until then, watch out for those Russian tanks around the Reichstag!
You did not get my point: Of course you can argue to disregard a subset of international samples but where exactly do you draw the line? The FED will make sure not to repeat the great depression or the stagflation period so why keep those in the sample? Would give all of us a nice boost of the SWR well above 4% …
My point is when you begin to remove samples from your dataset you start twisting your results. Yes, the US is *somewhat* exceptional due to its geographic isolation but nobody can reliably estimate the level of protection when the shit really hits the fan. And using an MSCI series starting 1970 is simply insufficient evidence against the conclusions from the study.
Oh, believe me, I got your point. I’m glad that we agree that certain events should be disregarded and others included. There is merely disagreement over where to draw the line. Let me draw the line:
1: The failures of entire countries, like the Axis in WW2, will not be replicated in the USA. Kansas City will not look like 1945 Dresden. So, that’s out. Likely also the smaller countries that Germany occupied, e.g., Belgium.
2: Last week, I detailed that 1792 data from the US is also irrelevant. Where you draw the line in the U.S. is up for debate. 1871 is the furthest I will go. But 1926 as a starting point is acceptable as well. Not a big difference in results.
I certainly believe that the US government in general (Federal Reserve and fiscal authority) could potentially make the same policy mistakes again as in 1929 and the 1970s. As a worst-case scenario, I like to keep those in. Also, I just proved that the 2000 retirement cohort is definitely going to wipe out their portfolio within the next 3-4 years (unless we observe 25%+ returns for the next 7.25 years until 3/31/2030), so the 4% rule will fail between 2000 and 2030 for the BTSQ approach. So, don’t play ignorant and pretend that post-1970 there is no more risk for the 4% Rule
And again, the BTSQ methodology would have failed even worse in 1929 than in 2000 and 2007. Non-US stocks (likely also -60% or worse) would have fared far worse than US Treasurys in 1929 (+40% between 1929 and 1932). Every major country went through a Great Depression around that time. So, a full 100 years of data and 3 out of 4 market events, show that BTSQ is bad during those 100 years of US data.
I should also stress that with the BTSQ’s own data (starting in 1890, so 80 years earlier than my limited sample) it looks like domestic only stocks did a better job in retirement. The block-bootstrap in the US shows that 100% domestic equities yields a lower failure rate (5.0% vs. 5.2%) and 60%/40% (all domestic) has a lower failure rate than 30% dom. stocks, 30% int. stocks, 40% dom. bonds: 4.5% vs. 6.2%.
I found that McClung in Living Off Your Money took a reasonable approach: He first tested strategies against US data and then saw how they (the more promising ones?) performed on UK data post 1923 and Japanese data post 1950.
In any case, for stocks I regard Europe (incl. the non-€ parts) as home market; for bonds it is slighly more difficult.
True. But again, using only domestic bonds and only domestic equities in Japan post 1990 is very risky.
do we even need to worry about deflationary events anymore? Governments learned to just print money and avoid that. We even tackled a global pandemic. I don’t see the point of nominal bonds anymore. Some intermediate TIPS is all that’s needed to survive supply shock.
Yes we do need to worry. We may or may not have higher trend inflation. Japan accumulated a lot of debt since the 1990 and was stuck in a deflation trap.
But even assuming we have higher trend inflation rate, the general idea of demand shocks (declining inflation) and supply shocks (rising inflation) still works. Bonds will still do well during the demand shock recessions due to rate cuts. There is no deflation, merely lower inflation.
Dumb question: Are your blogs only for US citizens? Would your advice apply to Europeans too? Or it’s a completely different story? Will some type of VT+BND tent work for me too?
Depends on the country. If your local government bonds are a safe haven investment you can use those for the bond tent. If not, you might have to use US gov’t bonds. But yes, the bond tent generally works if the same assumptions are present in your stock/bond returns.
I am in Euroland and use €-gov bonds as reference; the problem is of course there is too much Italy for my taste. Alternatively use Worlds government bonds, €-hedged.
For Safe Withdrawal Rates I have made data series for the period where €-data exist (past 1994); 1999/2000 is one of the very worst retirement cohorts, also when comparing to US 1929 or 1968. My conclusion was that if your currency is not USD you may want to add a little extra safety margin to the SWR. The simulation spreadsheet is still terribly useful, thanks Karsten!
Absolutely. That’s a great point. It must be because the USD is still a global safe-haven currency and in the worst possible economic event, even though all markets tank, at least the USD might appreciate a bit, due to flight to safety.
So, for non-USD retirees, use the SWR toolkit, but build in a safety margin, as you write.
Ok, but what about VT? Do I still go with it (~60% US, 40% ex-US)? Or do I need more/less US?
That works too. It’s not identical to the 50/50 but close enough.
Good morning, ERN. Thanks as always for your excellent work. Just a quick question from a small tidbit of your post (“Thus, recommendations for safe withdrawal rates should always depend on market conditions, especially equity valuations. The same logic extends to asset allocation recommendations.”)
We are constantly advised not to try to “time” the market. Is adjusting allocations based on CAPE not considered timing? More specifically, have you reduced your stock allocations today based on the very high CAPE?
It depends on what your “market timing” entails. Shifting in/out of stocks daily or out of emotions is bad. Shifting from 100% stocks into 100% bond one day is bad.
Good market timing would be advice like this:
In normal times (stocks not too expensive, not too beaten up, normal bond yields) pick a 75/25 portfolio.
In times like right now with stocks very expensive and bond yields very attractive, probably go closer to 60/40, at least initially. Then as retirement progresses, slowly shift back to 80/20. See Parts 19+20 on the glidepath.
If stocks are severely beaten down and bonds yield essentially nothing (e.g., March 2009) You can be more aggressive, maybe 90/10 or even 100/0.
I like this approach, but have you tested it beyond the glide path article? I’d love to see guidelines like the cape tool, but also applied to stock/bond mix, and not just spending, of course recognizing that only portions of most people’s portfolios can be moved tax efficiently.
So my question would be “can you increase the safe withdrawal rate by increasing bond percentage when certain conditions are met?”. eg cape high, rates high. vs cape med rate low, etc.
Seconded! That would be a very interesting article!
Noted. stay tuned.
I have a theory on how this might work. It seems the worst case scenario is having to liquidate your stock holdings when the market is down to live, especially early in retirement, but really at any point..
This procedure is assuming you are in retirement and selling securities each month to live. In a normal market (not a correction or a bear market), you sell only stocks.
When there’s a correction (stock market down 10%), rather than selling stocks to live, you switch to selling 100% bonds until the stock market hits a new high (or you run out of bonds, in a long correction/bear market). When the stock market hits a new high, replenish the bonds by selling stocks.
As a practical matter, you could achieve this by having your bonds in a tax advantaged account (IRA), and instead of actually selling bonds, you could buy stock in the tax advantaged account when you sell shares in the regular brokerage account to live, thereby liquiding the bonds. When the market hits a new high, you can buy them back without a tax hit (and you’ve made at least 10% on your money).
Another term for that is the Glidepath. As I noted in Parts 19,20, this can alleviate (not eliminate) Sequence Risk.
Try the same thing with international stocks and you get hammered in 1929, 2000, 2008, etc.
I guess I’m looking for a glidepath-like solution that doesn’t violate Bellman’s Principle of Optimality. So rather than triggering the increase in the equity-ratio at the retirement date, use some other trigger, whether that be stock valuations, bond rates, or some combination. It seems intuitively that if the Cape Ratio is high and bond yields are high (like today), then we should move some money into bonds. But whenever I try to backtest that using your data, it doesn’t seem to improve the success rate consistently. (Probably because most investors are trying to do the same thing).
I’m really interested to see what you come up with.
Really? When I simulate GPs in the case study tab of my Google Sheet, I pick a historical worst-case scenario (e.g. 9/1929 or 12/1968) and normally a GP will do better than a static allocation.
Right, it works on some worst case scenarios when you retire on those dates. But take a case when you’ve already retired. Let’s say, for example, it was 2 or 3 years before 12/1968. You’re gradually sold your bonds you’re in a glidepath. Then when it’s 12/1968, you’ve already sold most of your bonds, but really, you would want to be just starting the glidepath on 12/1968. It’s not like the cape ratio was particularly high on 12/1968 (it was higher 7/1967). There’s no reason to expect that you’d have to start buying bonds at that moment (except that you had retired and were following the algorithm).
I’m retired now, and I’ve seen my ability to spend go up (because of the increase in the market) which is really nice, and I did sell some stock and buy some bonds a couple of months ago (in an ira), to protect against the market going down, reversing my glidepath. But I don’t think that was necessarily the right decision. I can start a new glidepath, again protecting against this being the height of the market, but it seems like random guessing.
There are two ways to deal with that situation:
1: If you retire so far away from the worst-case cohort, chances are that you will make it and both the GP and the static asset allocation will deliver satisfactory results.
2: You could also do the “active GP” (see parts 19,20 in my series) where you delay the GP as long as you’re still in a strong bull market.
Thank you so much for this response! I thought I had read most of the series, but I must have missed 19 and 20. This blog is really a treasure!
I think it’s really interesting that the glidepath has only worked with CAPE > 20.
Since the CAPE ratio is super high (27+) right now, I reset the clock, loaded up on bonds (as much as I can in tax-deferred accounts ~30%), and started the clock again, as if I’m retiring today and the 5 year glide path is starting.
If you’re not a CAPE high you don’t have to worry about an impending market crash, so that makes sense. A static allocation works just as well.
Good plan with the bonds. Yields are certainly attractive!
I have some notes on using tactical asset allocation methods (valuation and momentum). Yeah, they help a little bit but are no panacea either. Stay tuned for a future post on this matter.
After reading this post, I say that this is a “drop a mic” moment (again).
I wish there was more say, but No! Karsten, yet again helped us all to be better off for the rest of our existance and our children.
Thanks, Karl! Glad you dug through 6,000 words this morning and liked my ramblings again!
Karsten –
Really appreciate your thoughts here. Some challenges below that I welcome your comments on. I ask these question with respect and appreciation for all you’ve contributed. Cheers!
1. 100% stocks would fail many historical cohorts
Note – Cederburg’s findings suggest an 8% failure rate. So it still fails, just theoretically less often than other portfolios. You effectively only cite ~2 examples here (which I wouldn’t qualify as “many”), so that feels insufficient to counter the paper given we should expect no portfolio is risk free.
2. Inflation is suspect (section 5.2)
Actually, this argues for Cederburg’s point in my view. X-US benefits when the dollar declines (inflation) as you show. Thats exactly Cederburg’s observation. Domestic currency decline boosts international stock holdings. The 1970’s hold up as an example where international stocks HELP, and given this is one of the worst retirement cohorts for a US investor, I find this a compelling example of why some X-US position should be held. So in my view, you end up arguing for the other side here.
3.Bellman Optimality
Are you sure you have the paper right?
As I understand it, the accumulation phase and decumulation phase while linked in the paper, are separate in practice because all decumulation phases use the same % of assets from the accumulation phase portfolio. In other words this did not happen:
” A1 allows you to withdraw 4% of the initial nest egg (subsequently adjusted for inflation), while A2 allows you to withdraw 4.5%.”
All good questions.
1: 100% stocks has a certain failure probability. The criterion in the paper is not the failure rate but a utility function over retirement consumption. All I am saying is that a more cautious retirement asset allocation will give you less retirement risk. You can and should still use 100% until retirement, but the paper does not show anything about what’s the best reoptimized asset allocation once you reach retirement.
2: Clearly, Cederburg suffers from the same confusion about economics as you. (no offense I hope!) When we say the value of the USD declines it can be understood as one of 2 ways: a) the USD loses value against other currencies. b) the USD loses value against a basket of goods. a=FX movement, b=inflation. The two are not the same, though often correlated. In the USA between 1973 and 1982 the two coincided. But from the perspective of a European investor, their own currency appreciated against the USD, even though they also had inflation domestically. The two go in the opposite direction. Thus, in the 1970s, Euro investors suffered inflation, but investing in US stocks did worse due to the USD depreciation. This negates the inflation story.
3: This was just for illustration. In BTSQ they calculate a utility function over retirement consumption plus bequest. An analogous way would have been that with A1 you could consume 4% and deplete your portfolio and A2 you consume 4% and keep a sizable bequest due to less Sequence Risk. The conclusion would have been the same: it’s optimal to accumulate with A1, then decumulate with A2. Any paper that assumes the allocation has to stay constant (or some cockamamie life-cycle pattern that exacerbates Sequence Risk) and does not account for the fact that people may want to reoptimize in retirement, is useless.
Thanks Karsten. Appreciate your responsiveness.
I think there is a misunderstanding tying both your points 1 & 3 together.
It was confirmed with Cederburg that the accumulation phase is irrelevant to the ruin probability in the decumulation phase because all strategies use the same 4% withdrawal strategy regardless of starting assets. Accumulation affects total consumption but not ruin probability of any given decumulation approach.
So given that, the investment style on accumulation is effectively decoupled from the decumulation outcomes. E.g. Everyone is starting with a 4% withdrawal of their accumulated wealth. So the original path of how you accumulated wealth is a dont care.
I take your point on 2 and will think on it. Thanks for the helpful, and now obvious, separation of FX and goods-inflation.
If your withdrawal rate is less than 3% and your long term goal is to grow your retirement fund, isn’t it better to be 90% to 100% stock (such as VTSAX, VTIAX)?
Yes, correct: The failsafe would be right around 3%. But the upside potential is astronomic. If you want a hospital wing named after you, that might be the way to go.
But other folks may just take chips off the table and play it safe. Why keep gambling if you already won?
Completely agree with this. And it’s easy to do so right now with real yields around 2% on tbills, and equities at sky high valuations.
Well said. 2% real. Not enough to retire on them alone (at least not for 50+ years) but certainly a good yield for a diversifying asset!
Correct but irrelevant. The objective function is not the success probability. The utility function is on p.15, equation (4). You plug in the simulated consumption values equal to withdrawals plus estimated Social Security benefits. If your portfolio runs out you’re stuck with only SocSec benefits. If you have excess savings, they go to your heirs, which is represented by the second CRRA term in the U function.
So, the accumulation very well determines the final utility and has a huge impact on expected utility and the ranking of different asset allocations. So, my points 1+3 stand as before. Specifically, because a 100% stock portfolio so extremely outperforms everything else during accumulation, even if 100% equities are riskier in retirement, the higher initial withdrawals will make up for that. Thus, again: It is mathematical, logical, and financial malpractice not to allow the household to change the asset allocation to something less risky during retirement after miling the equity returns during accumulation. There is no optimization performed in this paper, the 8 different cockamamie portfolios are not close to spanning the complete set of possible actions in a dynamic stochastic optimization problem. The fact that their Portfolio 8 beats the other 7 means nothing.
Thanks
I think I see the source of our divergence. You are focused on his utility numbers (Table 9) which I personally pay no attention to. I personally dont find anything actionable in there.
I am chiefly interested in his ruin probability for which this issue has no bearing (Table 10) and wealth at death (Table 8). Both of those are driving the whole 100% stock w/ 50% international take-aways as they have the lowest ruin rate and highest wealth at death regardless of percentile.
Anyways amazing responsiveness to your community. Keep up the great work!
Ah, OK, we are getting closer to a resolution. Let me point you to Table X, p.44:
As I said in the post, the global sample is contaminated with government failures, irrelevant from today’s perspective, so let’s focus on the USA-only block-bootstrapping part. I would have preferred merely a cleaned-up sample to eliminate the WW2 Axis countries and a few more war victims around the time (France, Belgium, Netherlands, etc.) but the BTSQ paper doesn’t give any further detail. So, the USA-only sample is all we have for a reasonable simulation without all the “war-victim-bias.”
The BTSQ portfolio has a 5.2% failure probability. The 60/40 portfolio has 4.5% failures. Thus, 60/40 beats 100% Stocks.
Also notice that 100% US stocks beats 50% US stocks + 50% international stocks according to the failure probabilities. 5.0% vs. 5.2%
Also notice that 60% US stocks+40% US Bonds beats 30% US stocks + 30% international stocks + 40% US bonds according to the failure probabilities: 4.5% vs. 6.2%.
This is exactly what I meant when I wrote about the re-optimization.
Granted, lower failure probability doesn’t necessarily mean better utility, but I suspect it’s highly correlated. Also, failure probabilities may underestimate the damage from failure. Because the 100% equity portfolio fails much faster, the length of zero-withdrawals is longer for the 100% equity portion than for the 60/40. So the impact on utility is probably even worse form the 100% portfolio. Thus, all of the utility advantage of 100% equities comes from the accumulation. But you would be better off switching to 60/40 during retirement. Table 10 shows it. Or potentially increase utility and decrease failure probability even further with a more advanced asset allocation strategy, like a Kitces bond tent. See Parts 19-20 for my simulations. And my simulations are not for some whacky bootstrapping simulation. I use actual consecutive returns. Something like a 60/40 to 100/0 glidepath beats every single static asset allocation. And certainly a 100% equity allocation.
(shoot, my post here didn’t take. Resending)
Now there is a powerful argument. Almost worth adding to the blog post 🙂
Building on this, the two most powerful observations against the paper’s findings imo:
1) Domestic bonds help every domestic only equity portfolio in the data you link (E.g. “Stocks” vs. “Balanced”, regardless of bootstrap or IID). E.g. It’s not just a US phenomenon.
2) And most troubling, International equity correlations w/ domestic equities have risen substantially over time as markets have become more integrated, especially since the ~1890’s. Given that, today’s domestic equity holder should expect reduced hedging behavior of adding international to a domestic equity holders portfolio. At the extreme, that would make a modern portfolio of 50 domestic/50 international look more similar to the domestic portfolio in 1) where bonds help.
Food for thought. Your reaction always welcome!
It’s a good conjecture. I don’t have all the country data as in the BTSQ paper to prove that. But I agree, the international diversification benefit might be lower today due to higher correlations.
Not too much conjecture.
Domestic Stock correlations to International Equity over the total dataset: .35 (fig 4 in “Long-Horizon Losses in Stocks, Bonds, and Bills”)
Plenty of academic papers show historic correlations similar to this affect (lower even for the earliest periods) on google scholar.
And of course nowadays that number is ~.8+
(sorry can’t reply to the correct comment, probably too far nested)
Good point. From the perspective of, say, German investors, US stocks would have been a good diversifier in the 1940s. Much better than German bonds. But as markets have matured, I’d suspect that domestic/international correlations are much higher today.
Thanks for your fantastic articles.
Is it bad to do 100% stocks regardless of withdrawal method? I have a big buffer between the essentials and the flexible spending, so I was planning to do 2.5% variable percentage based on portfolio size every year. I’m 38 (not quite retired yet, but within the next year most likely). I wanted to do a variable % to avoid SORR altogether, as well reap the benefits long term with a growing pot. I’m 70% US Total Market and 30% International Total Market.
2.5% VPR is not avoiding SORR. If you have a bad market event you withdraw less. So, it’s SORR, just in a different fashion.
But 2.5% is certainly a low enough rate that will keep the portfolio alive forever and likely increase your real withdrawals over time. If you want to simulate how this would have looked in historical simulations, check out my SWR sheet, the “CAPE” tab. You would set a=2.5%, b=0 (i.e., no CAPE used in WR).
Why isn’t it avoid SORR? I assumed that the risk with poor sequences was that even if you assumed a reasonable average of long term returns, if you got the returns poorly early then it hurt you because the withdrawals were outsized relative to the scenario where you had the gains early. With VPR my withdrawals are always the same relative impact on the portfolio, so after 30 years, it didn’t matter if the gains came in early or came in late.
The implicit/obvious deal i’m accepting with VPR is that I need to be able to spend less (which i can given the buffer between essentials and flexible spending).
Seems almost like a definition difference if anything. I’ve always understood SORR to mean that the order of events in the market has the potential to have an effect on the long term survival of a retirement plan. Whereas different withdrawal models is a measure of willingness to accept variance from year to year (VPR, Guyton-Klinger, etc).
Separate from the definitions, my larger question was: Assuming I’m willing to do 2.5% VPR, is it reasonable to then be in 100% stocks (70/30 us/int) with the goal of having more long term spend potential (especially given the longer horizon), or am I thinking about it wrong with respect to bonds?
Thanks again for all your work, huge fan .
Start with a $1m portfolio. 4% VPW.
Scenario 1: Returns +10%, +10%, -20%.
Scenario 2: Returns -20%, +10%, +10%.
Scenario 1:
year Returns Portfolio Withdraw
Y0 $1,000,000 $40,000
Y1 10% $1,056,000 $42,240
Y2 10% $1,115,136 $44,605
Y3 -20% $856,424
Scenario 2:
year Returns Portfolio Withdraw
Y0 $1,000,000 $40,000
Y1 -20% $768,000 $30,720
Y2 10% $811,008 $32,440
Y3 10% $856,424
Both portfolios are the same after 3 years: $856,424. True, there is no SORR for the portfolio value. But you withdraw 27% less in Y2 and Y3 in Scenario 2. That’s a form of SORR, even though the average return was the same in both scenarios.
Karsten,
Great post. I loved your reference to marketing: “It’s like in marketing: if you try to appeal to too large an audience, you risk pleasing nobody and losing your most committed and devoted customers.” Budweiser comes to mind.
One question – I (and I think others) would like your take on the current US fiscal situation and trends. We are now sitting with federal debt > GDP (nominally~120% and public debt~100%) and fiscal deficits 4-6% of GDP as far as the eye can see (5-10 years). Meaning we may exhaust our borrowing capacity. Then the fiscal deficits can only be monetized. It is looking more like Germany circa 1920-40 – unstable.
A reckoning may be coming, but when?
What signs should we look for?
Can we put our fiscal house in order?
And if not, then what are the results?
– increasing interest rates (the beast must be fed)
– reduced capex
– inflation (excess fiscal stimulus)
– or deflation (gov’t spending will be constrained)
– slower growth (rising taxation)
Excellent comment. That’s why I like keeping US data during the Great Depression. Germany failed due to the combined effect of the excessive burden of the Versailles Treaty, the Global Great Depression, and the descent into fascism and WW2 in 1939-1945. So, I agree that we should prepare for some economic trouble in the future, potentially as bad as another 1929, but nothing as bad as Fascist Germany in the 1940s.
Another great article, Karsten!
Quick question that I don’t believe I’ve seen you cover in the past:
If one is using an early retirement withdrawal rate that is lower than what is deemed “safe” using your toolkit spreadsheet (i.e., 2.5% vs. 3.5%), would that change the case for holding 100% stocks, given that you are presumably subjecting yourself to less sequence of return risk?
I’d love your feedback and even possibly a piece with further analysis on this in the future.
Most articles I’ve seen on the safe withdrawal rate discuss a recommendation allocation for the safemax rate, but if you are using a lower number I’m wondering how that changes the analysis and recommendations.
Also, do you have any thoughts on using the RMD method for setting your early retirement withdrawal rate? I saw Morningstar recently published an analysis showing the RMD method produced the best results of various options they tested.
Thanks so much for all the excellent work you do!
I have heard this question a lot recently. Yes, you can certainly keep 100% equities if you find yourself with excess in your nest egg. You could have retired earlier, but that’s water under the bridge. So, maybe just keep 100% equities and use WR=2.5% and run with that. You have some great upside potential with that too, so maybe they will name a college football stadium after you.
The RMD method is inferior. You start with way too little in the beginning and then grow your portfolio massively in the end and you have way too much retirement income when you no longer need it. Stay away from that!
Thanks so much for the timely and very thoughtful response. And great point on the RMD method that I hadn’t considered. I really appreciate all the work you do! Thanks again.
You bet! 🙂
There is an implicit concept in the post that investors are better off with 100% US stocks vs diversifying with international stocks. My understanding of the argument for owning international stocks was not as an inflation hedge but to try to track the relative market cap of the overall market. Do you feel this isn’t valid?
I’m more troubled by the 100% equity claim than the 50% domestic + 50% international. My main complaint is in #8; investors in retirement should have less than 100% equities. By the way, if we agree on 70% stocks and 30% bonds allocated as 35% domestic stocks + 35% international stocks = 15% each for domestic and international bonds, I can live with that too.
Excellent write up!
I agree with your don’t bet against the U.S. sentiment and going 50/50 US/International seems very extreme to me however based on your #7 that valuations matter and U.S. valuations look very stretched compared to the rest of the world right now. Have you ever looked at 80/20 US/International portfolio or 60% U.S/15% International Stocks/25% Treasuries for retirees? While even at these high valuations, I still think there’s a 99% chance that the U.S. outperforms the field over the next 30+ years, there’s always that small chance of black swan events disrupting the party.
As I mentioned in another comment, I can live with international diversification. But it’s no panacea, especially from a US perspective. If you’re in Iceland, you should have 99.9% international stocks and only 0.1% domestic.
I’m more troubled by the claim that international stocks are a better diversifier than bonds. Ideally, you find an optimized portfolio over the 4 asset classes, domestic/international, for both bonds and stocks, taking into account expected returns and the correlation matrix.
Your blog is the only FIRE blog that I still follow. In fact, it has transcended the niche and is a class all on its own.
Thanks for the kind words, Bob! The niche got too small for me, it’s now a class on its own. That warms my heart! 🙂
I wholeheartedly second this comment! Really love your in depth assessments ERN
You rock! Thanks for your kind words! Makes it all worthwhile to skip the Superbowl on Sunday and writing/editing a 6000-word post instead.
Excellent work.
It seemed fairly obvious that the bond component of their modelled portfolios would look quite strange if you were to replicate it today, as it contains a lot of speculative instruments in problematic currencies. Thus, they were not modelling anything that anyone would likely to be holding, but including highly speculative debt instruments. How many advisors are recommending that retirees hold things like Turkish and Portuguese bonds these days?
This was the same problem they had for the earlier paper involving this database, where they concluded that a viable SWR was only 2.7%.
It would have been more interesting if they would have limited the bond component to only bonds denominated in the world’s reserve currencies (at the time of the data), so really only pounds and dollars, since the alleged purpose of bonds in these portfolios is to be a stabilizer.
I have a feeling that even a large academic backlash to this methodology and the newly christened “Pickelhaube Portfolio” is forthcoming.
Hi Frank! Brilliant analysis, as usual!
Good point. I also wondered why they allow diversification within equities only but not international bonds. At the very minimum, you should also simulate 30/30/20/20 (dom. stocks/int. stocks/dom. bonds/int. bonds). But I like your idea even better. In some of the small unstable countries, government bonds are just as risky as domestic equities and also highly correlated. It’s much more beneficial to use only the bonds from major world reserve currency countries (e.g. USA, UK for most of the part, then maybe add Germany and Japan well after WW2). That’s much better diversification.
Excellent suggestion!
Wait. You guys were in a celebrity death match… you aren’t allowed to be friends!
If you can keep a secret: It’s all for show. Like the WWF. We are all good buddies and have a beer after the deathmatch.
Would your opinion change for countries with specific tax implications? Interest (and dividends) here are taxed at 35% while there is no capital gains tax (under most circumstances, a major one being holding for > 6 months, and if you work the tax paid on interest/dividends is credited to your tax bill while the income from interest/dividends added to taxable income, but in retirement your taxable income will be very low, before pension kicks in maybe even 0 and you don’t get tax back afaik so you’ll end up paying the 35% on the only income you have – interest and dividends). At face value losing 35% of your returns from interest versus 0% from selling equity, does that shift your view on 100% equity in retirement?
This is a hypothetical for now, I would have to ask a tax expert how it works exactly, as in it could be that your taxable income is exactly your interest income and they calculate your tax based on that and since it will likely be below 35% you’d actually get money back because you paid 35% on all income at source but tax return the following year would give a lower income tax amount.
Noted.
At the margin, this would certainly make bonds less attractive. But it’s hard to imagine that 100/0 is optimal even in that case. 100% equities would create serious Sequence Risk concerns in a retirement portfolio. I’d still accumulate 100% equities, but during the withdrawal phase you need at least some diversifying assets.
I also wonder if people in that country may also have some tax-deferred accounts (similar to US IRAs, 401k, etc.) where you can shield your interest income.
Thanks, yea makes sense, I have not yet thought about this enough from the decumulation stage.
Yes, dividends in your pension fund are not taxed, but you don’t keep the fund as such post retirement. Either you get a pension that is based on your fund value at retirement, or you have the fund paid out (or a combination of both). However, if you retire early, then yes, until official retirement age you can invest the money in that fund and it is protected from taxes. However, you cannot freely invest I think, you have to find a provider and most pay very low interest, there are some providers I found that allow you to choose more risky strategies and then the value develops as per investment rather than getting a secure but low interest every year, but not sure if you can pick your own portfolio with them or just pick certain strategies and they do it.
Noted! That would be a backdoor way of keeping safe assets in a tax-deferred account for at least a while during early retirement.
Great as always! Do you have opinions on “I can go back to work” as a viable strategy to cover for the smaller percentage of time windows where a 100% portfolio would fail? I.e. keep to 100% for a more likely chance of higher gain – But cover the catastrophic and smaller chance cases
I can tell you from personal experience that even thinking about going back to work after you really get into retirement will be very hard. . . Think of Maynard Krebs if you want to know how hard.
Same here. Thanks for confirming this!
I had to look up the cultural reference to Maynard Krebs, haha!
It works for the early retirees. But this is a dangerous approach for the 65-yo retiree who might find out that at age 75 he/she will head back to work.
Hi ERN(and to Frank Vasquez as well if he sees this) but my question for ERN is have you ever written about the risk parity withdrawal rates claims made by Frank and often Tyler at portfolio charts other than somewhat in the article you wrote on including gold in the SWR series? I really enjoy Frank’s Risk Parity radio podcast and and also Tyler’s portfolio charts site but they often claim much larger SWR amounts available using more diversified asset classes than the basic stock/bonds makeup your articles usually use and I’m not sure but wonder if the main issue or difference is they are mainly using data only from around 1970 while you are going much further back for all your research? Do you think that is largely causing the discrepancies? And have you ever written your critique of their methods both good & bad like in this article?
Also I wanted to add my many thanks to those above for this great site and your fantastic articles like this one as I always learn so much from them.
About Gold: See part 34 of the series. A small gold allocation (10-15%) can hedge some of the Sequence Risk.
In the same post, I also test some of the often touted portfolios, like the the Permanent Portfolio and Ray Dalio’s All Weather Portfolio (very similar to RP in flavor) and those portfolios really sucked if back-tested far enough.
The problem with RP portfolios is that the proponents constantly revise what’s in them. Commodities were all the rage in the 2000s until the GFC. Then commodities in general really sucked in the 2010s, especially energy due to a crazy contango in the futures curve. Now people skip overall commodities to avoid this and only include gold because gold did really well recently. So, there is hindsight bias and their portfolios are often tailored to fit past history. But then they suck when applying them going forward. Case in point, the Risk Parity Radio sample portfolios. Most of them underperform traditional portfolios.
PS: See here: A https://www.portfoliovisualizer.com/backtest-portfolio?s=y&sl=25l1MCnz8BKEuSliawe6SA
A simple 75/25 portfolio beats the All Seasons and Golden butterfly
Although I didn’t understand much of the math, some things are clear. 100% stocks should never be the way. ERN’s bias and passion for the USA is more clear than never in this post, which is weird coming from a german.
And what in the world did you do to turn the 1st chart into the 2nd? man….people can modify anything these days to prove a point…
I’m naturalized US citizen. Sometimes those are the most patriotic. So, keep than in mind. Maybe I’m just biased.
How the first chart and the second chart can possibly be consistent? It’s called Sequence Risk. There is a difference for buy-and-hold investors vs. retirees with regular withdrawals. That’s how you get that stark difference.
I wouldn’t make personal finance decisions based on American exceptionalism.
I didn’t “make personal finance decisions based on American exceptionalism”
The same principle, i.e., hard to invade and likely not going to look like Dresden in 1945, applies to other countries: Canada, the UK, Australia, NZ, etc.
Shouldn’t “unlikely to be invaded” result in lower expected return given the lower risk premium?
Correct.
Interesting…50% domestic is likely still too high for investors in small countries like Iceland or Portugal. What about a larger and more important country like Brazil which is a big part of the Brics but whose currency is not so strong and is more commodity based. How much should someone there have in US stocks?
Brazil is still only in the low single-digit range of the global stock valuation. You can’t have 50% local Brazilian stocks in such a small country with so much idiosyncratic risk.
Even if I’m living here and no plans to move abroad ever?
Yes, International diversification is not only about hedging where you live. Brazil has much higher idiosyncratic risk than the world index.
Diversification is about a covariance matrix and investors from small countries (or large countries like Brazil with a small equity market, compared to the World) will benefit from shifting asset weights abroad.
Funny that in the past 2 years, all my rebalancing has been to sell stocks and buy more bonds! Hope this is the right move but sometimes it worries me not be buying stocks for so long following this strategy….
It’s not a bad idea to take some of the stock market chips off the table and invest in bonds with relatively attractive yields now.
so, it’s time for me to start adding cash and bonds; I’m considering BSV (short term bond) for my taxable brokerage and BND (total market bond) for my ROTH IRA. based on the graphic in section: “Approximate tax efficiency ranking for major asset classes” from site: https://www.bogleheads.org/wiki/Tax-efficient_fund_placement
I’ve avoided bonds in my taxable account; but it seems based on efficiency, it’s ok?
We can’t really make generalized statements about fund placements. See Part 35.
Stocks in a taxable account enjoy better tax treatment, but there are some extreme cases where the choice is not that clear-cut. Especially when comparing apples-to-apples and keeping the true net-of-tax stock/bond allocation constant between the different scenarios. Maybe check out the worksheet an plug in your numbers to make sure the conventional wisdom advice applies to you.
Big ERN, an honor to be leaving my first comment and wanted to extend a huge thank you. I’m managing my parents retirement (investment portfolio, tax strategy, withdrawal rate) and feel very good about the moves we’ve made delaying social security and using low-income years to convert Trad to Roth free of tax. The SWR series has given me the confidence to encourage my parents to spend their hard earned money in ways that will enhance their quality of life. And in all honesty, their reluctancy to spend has probably been the most challenging aspect of all for me during their retirement!
This article has sent my brain in circles thinking about their current asset allocation [75% US stocks, 25% intermediate treasuries/cash and essential/discretionary spending is more than covered by Social Security]. In my limited worldview, they’re in an enviable position. I’m proud of them and love helping them in this way (plus, they babysit).
I’m reminded of a few other articles… In 2018, you wrote “How useful is international diversification” and demonstrated that it hasn’t necessarily worked when you need it most. Your personal asset allocation in March 2016 was under 1.6% in ACWI ex US, and I just read Ben Carlson’s article you referenced. But I’m also conscious of recency bias, reversion to the mean, and current valuations… I’ve begun to wonder if you have slowly diversified into international over time? Any general feedback or nudges in any direction to expand my thinking around this would be genuinely appreciated.
I have to admit that I haven’t done much more international diversification since then. Currently we’re at roughly 3.5% international.
I can see the reversion to the mean argument and the much lower CAPE in Europe. But there’s also less growth in the old world.
Emerging markets have more growth, but I am afraid of a China collapse. So, maybe EM ex China would be an option.
But in the end, there are too many unanswered questions, and nothing has enticed me to move away form the US-centric asset allocation.
Hi Karsten. Two questions please.
1) If a significant benefit to a portfolio’s bond allocation is not just yield but negative correlation to equities, is that value proposition compromised given that sometimes that is not the case? Are HISA-ETFs, GICs (I’m Canadian) a reasonable alternative–they lack a negative correlation to equities (so does IEF, at times) but they provide a positive real-dollar yield.
2) At what low yield would you stop buying IEF?
Thanks for the help.
As you state yourself: the yield is not everything. When the yield goes down, longer-dated bonds gain in value. That’s the diversification benefit of IEF. You wouldn’t get that with savings accounts. It’s a tradeoff because in an inflationary recession (1970s, 2022) you lose from the duration effect and you’d be better off with a savings account.
Hello Karsten
you said : “Good point. From the perspective of, say, German investors, US stocks would have been a good diversifier in the 1940s”.
Except that I doubt that a German investor at the time would have been able to recover his money invested through a German financial institution. I no longer have any hesitation about staying invested in US equities for my retirement: my main concern is how to make withdrawals (in the best of cases) in the event of a systematic financial crisis. The worst case scenario is that the organisation is insolvent, in which case there is a guarantee fund, but I doubt it will be enough in a systemic crisis. For example, there’s a European law that allows you to block any withdrawals from a life insurance policy and believe me, the French government won’t hesitate to use it. In fact, my dilemma is either to invest in a tax-exempt financial product managed by a French financial institution (bank, insurer, broker) or to open a taxed securities account in the United States (although this is possible).
The systemic part is not modelled in the paper or in my SWR series. There is certainly a risk that even though your money is still there, the government steals it or freezes it. It’s hard to gauge that probability. It’s low in the US, but might be a headache in some other countries, even rich western countries (France, Germany, etc.)
Hi Ern. A follow-up question please. If a 40 year-old retiree, current high-CAPE environment, allocating 20% of their portfolio to bonds could only pick one bond fund for the next 50 years, no predictions, no tweaking, just quarterly divesting and rebalancing to fund lifestyle, would you stick with IEF? I believe the Canadian equivalent is ZGB. Thanks very much.
Yes, IEF is the go-to fund for US investors.
I can’t vouch for ZGB. That seems to be the overall government bond index fund, not the intermediate one. Not a huge difference, but it would have a different duration and different diversification properties (correlations).
Thanks very much. Huge fan of your work. I’ve read the whole SWR series twice now. Please keep the articles coming. The FIRE community is lucky to have you.
Thanks for the feedback! Glad you found this useful!
Great analysis.
There are even more good points coming from Banker On Wheels today:
https://www.bankeronwheels.com/should-you-invest-100-in-equities/
Yeah, good one! That post plus mine combined would be as close to a demolition as you can get.
Very nice writeup by Raph there. I can highly recommend his site, in particular for European based investors.
His background include working in banks cleaning up their bond situation after 2008, from memory.
Yeah, smart guy!
I would like to ask here off topic question (I can’t log in into the forum, to ask this). You have wrote about time and path dependant optimisation. Do you have any opinion on the use of stochastic programming to optimise financial decisions?
After looking a lot of models for goal based wealth management I have found this model:
https://www.researchgate.net/publication/232024887_Asset_liability_management_for_individual_households
This is comparison to another strategies: https://www.researchgate.net/publication/292386994_Lifecycle_Goal_Achievement_or_Portfolio_Volatility_Reduction
Model allows you to optimise decision about asset allocation, saving/withdrawal rate, also do tax optimisation. It is based on the dynamic strategy assumption that you can change future decisions depending on the performance of the portfolio.
Interestingly, the strategy resulting from this model is bond tent.
I have temptation to implement this model.
I appreciate your attitude to investments based on quantitate finance. Do you see any drawback about stochastic programming?
I realise that it depends on the correctness of assumptions about the future distribution of returns.
Also it’s not so popular in finance/investment academic world.
I’m supportive of using dynamic programming. I’ve published academic papers using this method. Both published in the Journal of Monetary Economics.
The problem for retirement planning: you’d need to use Monte Carlo style asset return. So it’s hard to do this with actual historical asset returns. DP is also extremely complicated to solve for optimal paths. Likely a bridge too far for most retail investors.
Hi Karsten. Is there an option to solve for optimal allocation mix if we know our current investable assets and annual consumption? Suppose our cost of living is $55,000. We decide on a SWR of 3%. We have adequate assets in this high CAPE environment of $1.83MM. Will the sheet solve for the most robust asset allocation resulting in the lowest failure rate, the appropriate ratio of stocks : ten year bonds : thirty year bonds : gold, etc? Thanks very much.
There is no generally accepted definition of “most robust.” I math, we can’t optimize unless we agree on the objective function.
But if you pick your objective and play with the asset allocation, say stocks vs. bonds in 5% steps to see where your personal objective looks best, then that would be a start.
I tried this on my spreadsheet and it was very interesting. The optimal stock/bond allocation changed depending on what my final value target was. (optimal being the highest SWR) The higher the target, the more stocks and the fewer bonds. I guess this makes sense, since to have a higher final value, one needs more risk and more return.
Correct. If you’re comfortable with depleting your assets, you’d be seeking more bond exposure.
I just have to note that Morningstar just posted an article with same premise as your last two (https://www.morningstar.com/stocks/should-long-term-investors-be-100-equities). It is based on comparing the exact same two papers. I started to read it and thought “wait, I’ve already read a better version of this somewhere”!
Maybe since they were two recent papers covering similar subjects it is just a coincidence. But since it looks like it was contributed by a third party firm and not Morningstar itself, I’m suspicious. Maybe it’s a coincidence, maybe you should be flattered, maybe you should be annoyed. I don’t know – just wanted to pass that along. Long time reader and I’ve enjoyed all the content since day one.
Thanks for the link. You’re a good forecaster. That’s exactly how I feel: flattered because I wrote my piece first and much better. Annoyed because his piece is pretty unimpressive and he gets the Morningstar fame. He claims the result is due to chasing the highest average return. Not true, the BTSQ authors do use a CRRA utility function which generates a tradeoff between expected returns and risk. Swedroe then tries to disprove the BTSQ with the equally bad paper by McQuarrie. Yeah, let’s include the 1792 US returns because their are more relevant than the 1945 German returns?! Oh, my!
I have the sense that everyone in the industry knows that the BTSQ paper is stupid, but nobody can really put their fingers on it. Swedroe came up with pretty lame reasons. The two Canadiens who had Cederburg on their podcast lamented that maybe because investors don’t have the discipline to keep 100% equities – equally lame. So, they are smart enough to figure that there’s something wrong with BTSQ, but don’t have the technical knowledge to figure out why.
So far, the best takedown is my post. Banker On Wheels also had an excellent post: https://www.bankeronwheels.com/should-you-invest-100-in-equities/
Hi Karsten,
Your blog has been an incredible resource for me — thank you for all of your effort and hard work.
I have been at a 100% equity allocation for over 15 years. I found your reference to the Asness paper very interesting and duly went and read it. This approach makes a lot of sense to me however I do agree that the level of effort/engineering is a bridge too far for most retail investors, myself included.
Apparently WisdomTree has stood up an ETF that implements this approach (NSTX.) They have a few differences, most obviously using futures contracts on treasuries as opposed to corporate bonds. I am curious to know if you think this approach would be appropriate for the retail investor who likes the Asness approach of increasing expected returns for the same volatility as 100% equities.
Many thanks again for your work on this blog!
Best regards,
Mason
NSTX is still 90/60, while the Asness portfoloi would be closer to 80/120, because the max -Sharpe is usually at 40/60 or 35/65.
But I like the idea of skipping corporate bonds and implementing this with Treasury futures. You can use the S.1256 tax treatment, too. So, 60% of your bond income is taxed as LT gains.
This seems like a good strategy, but in the case that one already has equities and there would be considerable tax implications from selling, how would you implement this with treasury futures? It seems like a good way to hedge against market risk. Assuming I had a 70/30 portfolio that I wanted to make 70/100, what would be the types of futures you’d need to purchase?
No, it’s the opposite. Replicating NTSX is quite tax efficient. You keep your 90% in stocks. Buy and hold, no taxable income except for dividends. Then 10% in cash or cash equivalent. Plus 60% in 10y Treasury futures (ticker “ZN” at Interactive Brokers). Those gains are taxed as S.1256, i.e., 60% LT and 40% ST gains/losses, instead of 100% ordinary income.
It seems like this would have been a losing strategy over the last couple years with bond values going down, and yet, somehow NTSX’s 60/90 did better than a standard 60/40 split. I don’t get it.
90% equities beats 60% equities post-pandemic. The drag from bonds is smaller than the equity advantage.
So, shouldn’t everybody sell all stock in all other nations except America? Shouldn’t all non us stock go to $0 as everybody sells? If America is objectively the best place to invest and it is so obvious, then all other stock is worthless. Why do other countries even have a stock market for that matter.
No.
Other countries have a stock market because they have corporations and corprate profits and like to have a (secondary) liquid market for trading productive capital.
You still diversify and keep a sizable amount in non-US stocks.
What percentage do you recommend to be international?
For US investors? Between 0 and 50%.
For investors in smaller countries (e.g., Canada) you’d likely need to be at 50-90% non-domestic.