100% Stocks for the Long Run?

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:

1. The BTSQ portfolio: 50% US stocks (S&P500) and 50% MSCI World-ex-USA
2. 100% stocks, US only (S&P500)
3. 75% US stocks, 25% US 10-year Treasury benchmark bonds
4. 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