June 2, 2025 – Welcome to another installment in my Safe Withdrawal Series, please check the landing page for all posts so far. Today’s topic is about Small-Cap Value (SCV) stocks and whether they should have a prominent role in retirement portfolios. Some financial experts recommend adding Small-Cap Value to your retirement portfolio, which will miraculously and automatically increase your safe withdrawal rate from 4% to 5% or even 5.5%.
In today’s post, I first would like to present some simulations using historical data. Those simulation results look pretty impressive. Thus, investors in 1926 who had somehow been aware of the Fama-French research, published almost 70 years later (maybe through time travel!?), could have done remarkably well.
Of course, if you are familiar with my blog, you will know that I am skeptical of SCV. I’ve written two posts, one in 2019 and one last year, where I outline my main concern: the Small-Cap Value engine that generated extra returns worth several percentage points between 1926 and about 2006 started sputtering about twenty years ago, and it’s unlikely that now when everybody is aware of SCV, we will repeat those impressive investing results so easily. Thus, I also want to provide some simulations that factor in more realistic small stock and value premia going forward. Alas, once we scale back those factors’ return expectations, your retirement portfolio will have very little to gain from small-cap value stocks.
Let’s take a look…
Before we begin, please take a look at my new project. If you prefer a slightly less intimidating web-based simulation tool (no spreadsheet!), then please check out this project I started with a friend of mine:
With this tool, you can replicate most of the functionality of the Google Sheet, but it’s all web-based. You can also save your parameters for future use and upload them again when you run the toolkit at a future date! The Fama-French factors adjustments I use in this current blog post (more on that below) are not (yet) implemented in the tool, though. You’d need to use the Google Sheet to do so.
Small-Cap Value in Safe Withdrawal Rate Simulations
In my Google Sheet, I provided a large selection of asset returns, including the Fama-French SMB and HML factors. Let’s put these series to work and simulate how a Small-Cap Value portfolio would have performed over time.
Let’s start with the simple baseline case safe withdrawal scenario without any Small-Cap Value stocks:
- A 30-year horizon.
- 0% final asset value target, i.e., asset depletion.
- A portfolio with 75% large-cap stocks and 25% intermediate government bonds (10-year benchmark bonds).
- A 0.05% weighted expense ratio.
Next, let’s assume our investor added Small-Cap Value stocks to diversify the equity portfolio. Specifically, assume we replace one-half of the equity portfolio with SCV. To simulate this in my toolkit, I assume that we keep the 75% equity allocation but replace half of the S&P 500 index fund with SCV. You might be tempted to set both Fama French SMB and HML factors to 37.5%. However, most SCV funds I am aware of don’t achieve full 100% exposure to those factors. In my Google Sheet, I provide factor regression slope estimates for a wide range of ETFs, mutual funds, and even a few individual stocks; see the tab “ETF Factor Exposures.” Most of the SCV ETFs and Mutual Funds have SMB exposures of around 90% and HML exposure of around 60-70%.
Let’s assume our SCV equity fund has an SMB exposure of 90% and an HML exposure of 70%. I then assign 0.9×0.375=0.3375=33.75% exposure to SMB and 0.7×0.375=0.2625=26.25%.
To account for the slightly higher management fees in your average SCV fund, e.g., 0.31% p.a. in the DFSVX), I also apply a 0.20% p.a. additional management fee to the SCV portion only. For example, if the baseline portfolio has a 0.05% weighted expense ratio, I assume that the 37.5% move to SCV necessitates an additional 0.375 × 0.20% = 0.075% per annum. Thus, the weighted expense ratio is now 0.05% + 0.075% = 0.125%.
Now, let’s examine the simulation results and compare the worst-case withdrawal rates by decade; please refer to the table below. If we ignore the safe withdrawal rates in the earlier decades (the 1900s and 1910s), where the SMB and HML returns were not available (i.e., set to zero for most of the retirement horizon), there is a universal improvement in the failsafe withdrawal rates in every single decade. That’s impressive! The 1920s failsafe at the pre-Great-Depression market peak is more than a whole percentage point higher (4.98% vs. 3.84%). The worst-case scenario for the SCV portfolio is in the 1960s, specifically the December 1968 cohort. However, even in that cohort, we can achieve an impressive 4.46% withdrawal rate and never run out of money during a 30-year retirement.
So, what’s not to like about SCV then? Should I now shift half of my equity portfolio over to an SCV ETF and sail into the sunset? Not so fast. Here is a cumulative return chart of the two Fama-French factors since 1926; please refer to the chart below.
I also report average return data of the SMB and HML factors. For completeness, I also include the Fama-French Market Factor and the Risk-Free rate series; please see the table below.
SMB (the small-cap factor) had only two major return boosters: one in the 1930s and 1940s and the other in the 1970s. Since the early 1980s, approximately 45 years ago, small-cap stocks have not consistently outperformed large-cap stocks. If you ask market historians, the second boost likely came from institutional investors rushing into small-cap stocks, a sector they had previously avoided. I suspect that the first boost in the 1930s and 1940s was a response to the market shock in the 1920s and 30s and investor participation, i.e., retail investors who were burned by the Great Depression rushing into the market again and picking up the pieces, and especially beaten down small-cap stocks after the big drop.
Absent any new stark market and investor preference changes that make small-cap stocks even more attractive and mainstream, I don’t see any sizable return advantage in the future. If anything, small-cap stocks are now widely traded, and there is great ETF and mutual fund coverage. Where is that new rush into small-cap stocks going to come from?
HML (the value factor) had a very impressive run from 1926 to approximately 2006. During that time, we observed frequent drawdowns, typically coinciding with recessions and bear markets; however, the subsequent recovery consistently pushed the HML factor to new highs. However, it’s as if someone flipped a switch in 2006, and the red line is now trending downward. If you’re a chartist, you’d hate the HML chart: since 2006, you got successively lower lows and lower highs.
I should stress that I’m not saying that this recent 19-year HML and SMB dumpster fire will continue forever. I invest in broad-market index funds that cover both growth and value stocks, as well as large and small-cap stocks, because I believe in efficient markets. There is no free lunch, neither on the growth nor the value side of the market. Any “alpha,” i.e., uncorrelated, free, and reliable outperformance, will likely be arbitraged away by investors who are much smarter and faster-moving than some finance blogger in his home office in Washington State. The “alphas” we’ve observed historically, such as 5%+ for HML between 1926 and 2006 and 3+% in SMB from 1926 to 1979, are no longer realistic. We should not rely on those significant alphas going forward.
So, how can we account for the likely regime shift in SMB and HML returns in my SWR toolkit? That brings me to the next section…
How to model a more “realistic” Small-Cap Value premium
Imagine you are a retiree today who subscribes to the idea that the 1926-2006 performance in SCV was an anomaly. But you still like to use my toolkit, including the SMB and HML factors. How can we use the SMB and HML factor returns without artificially inflating our retirement simulation results? Specifically, we want to achieve two tasks: 1) maintain the salient features of the SMB and HML factors, i.e., correlations with the business cycle and other asset classes, but 2) remove the likely ill-gotten average excess returns that we will likely not replicate going forward. Not even Paul Merriman would assume 5%+ annually for HML and 3%+ for SMB for the next thirty years! And, of course, I’m not saying that exactly zero excess return must prevail. We can always add back a “reasonable” alpha to those SMB and HML factors.
To achieve the first task, I apply the Hodrick-Prescott Filter (HP-Filter) to separate the trend from the cycle of both the SMB and HML series. The HP-Filer was developed by Robert J. Hodrick and 2004 Nobel Laureate Edward C. Prescott (who taught me macroeconomics at the University of Minnesota, by the way) and was developed exactly for this task, i.e., to separate economic cycles from trends in time series.
I plot the results in the two charts below. The black line is the cumulative Fama-French Factor return, the green line is the HP-Filter trend, and the red line is the detrended, zero-average factor return series. Well, close to zero; more on that in just a minute.
And I do the same with the HML factor. Notice the trend return hovering around 5% for the first 80 years! But now that everyone and their grandmother knows about this “secret” return booster, the party is over, and the trend has even turned negative!
I made an additional adjustment to further clean the return data. The monthly HP-filtered SMB and HML returns could still have a slightly non-zero mean. To fix that, I added one more adjustment to ensure that the following three portfolios all have the same Compound Annual Growth Rate (CAGR) over the period with available SMB/HML data, i.e., 7/1926-3/2025: 100% S&P 500, 100% S&P 500 + 100% SMB, and 100% S&P 500 + 100% HML. To confirm that, please see below:
If you want to check out the Fama-French return data in my Google sheet, please see “Asset Returns,” columns AB-AC for the HP-filtered data, and columns AD-AE for the raw data from Ken French.
Next, I simulate the SMB and HML factors without any additional excess returns. To do so, I keep the SMB and HML weights at 33.75% and 26.25%, respectively, but change the Data Series to “HP-Filtered.” I also keep the additional alpha for those factors at 0% for now. Furthermore, I set the expense ratio to 0.125%.
And here are the results, all compared to the plain and simple 75/25 benchmark. Not surprisingly, your safe withdrawal rates stink when using zero-alpha SMB and HML factors. Especially during the historical worst-case scenarios (1929 and 1968), you would have further lowered the already lean withdrawal rates. Notably, even a zero average factor return could have greatly benefited your retirement in the 1980s and 1990s because HML hedged the 2001-2003 bear market quite nicely. Specifically, HML dropped precipitously before the market peak and then recovered during the bear market. So, I’m not bashing HML. There were a few historical retirement cohorts that may have benefited from this style. However, they were the cohorts that weren’t significantly impacted by Sequence Risk, with SWRs already above 4%. All cohorts with high Sequence Risk performed worse with Small-Cap Value.
Of course, our next step should be to assign some “reasonable” SMB and HML factor outperformance to see how much traction we get from SCV in our SWR simulations. This brings us to the next section…
What’s a “realistic” Small-Cap Value alpha going forward?
Vanguard’s asset class return expectations model indeed anticipates a marked outperformance of some of the equity flavors. For example, Vanguard believes that small-cap stocks will outperform large-cap stocks by 120 basis points over the next 30 years. Additionally, U.S. value is expected to outperform growth by 160 basis points; therefore, one would use half that, or 0.80%, as the outperformance of value stocks over a blended index (i.e., half value and half growth).
I then apply the 1.20% and 0.80% annualized alphas to the HP-filtered returns; please refer to the Google Sheet screenshot for the implementation details.
Here are the results; please see the table below. Several decades saw an improvement in your SWR, but only those that already had no major Sequence Risk issues. But the 1920s and 1960 saw very little movement. In the 1960s, you even managed to reduce your SWR by two basis points. Furthermore, you create a new, previously unknown Sequence Risk headache in the 1930s, which is the cohort that would have retired around the 1937 market peak. The SWR in the 1930s dropped by 22 basis points or approximately 5.02% of the retirement budget. I would not increase my SWR when including SCV!
How about a longer horizon?
Let’s repeat the same exercise with a 50-year horizon. All other parameters are the same, including the 0% final asset target. I simply changed the horizon to 600 months. Qualitatively, we obtain very similar results, although all SWRs are lower by approximately 30-50 bps due to the longer horizon.
- Using historical returns for the 37.5% SCV portion, you shift up the failsafe in every single decade. Though, I would not recommend a 5% and certainly not a 5.5%. The 1920s and 1930s remain solidly below 5% SWR. The 1960s failsafe (December 1968, but also several more cohorts from 1964/65) would have only a 4.16% failsafe rate.
- Using the zero-alpha, HP-filtered SMB/HML returns, you reduce your SWR by about 12 bps (3.27% vs. 3.39) in the post-1926 era.
- However, with the calibrated extra returns in SMB and HML, the picture brightens a bit. The post-1926 SWR rises by 12 bps (3.51% vs. 3.39%). The 1920s and 1960s now both have a higher SWR at 3.51%. But you also perform worse in the 1930s again.
So, what do I make of this? If you believe that going forward, both HML and SMB will fetch a modest extra return over the boring large-cap blend index, I don’t fault you for tipping your toes into the Small-Cap Value pond. But don’t expect miracles. The increase in your SWR is about 12 basis points or about half of what you’d expect from a reverse glidepath (see Parts 19 and 20). Don’t expect safe withdrawal rates of 5% or more!
A different Small-Cap Value strategy: Golden Butterfly
People come up with new asset allocation schemes faster than I can simulate them. One asset allocation style that’s become very popular is the so-called Golden Butterfly Portfolio, which adds gold, long-term bonds, and small-cap value stocks into the mix. Specifically, the Golden Butterfly portfolio has five equal 20% shares of 1) large-cap blend stocks (e.g., S&P 500), 2) small-cap value stocks, 3) long-term government (e.g., 30-year) bonds, 4) short-term government bonds, and 5) gold.
I need to point out two caveats, though! 1) US investors were not allowed to own gold between 1933 and 1974, and 2) the nominal gold price was pegged to the dollar for extended periods, only with some occasional jumps. So, take the simulations with a grain of salt.
A side note: The initial version of my post stated only the first caveat, and people took issue with that, both in the comments section and elsewhere. Thus, for completeness, yes, the Gold Standard and Bretton Woods also raise concerns when using gold prices in historical simulations. By the way, eliminating convertibility in 1971 and the subsequent gold rally, from around $35 in the late 1960s to over $600 in 1980, to compensate for pent-up demand is also a concern. I don’t expect an almost 20x gold rally during the next bear market. Specifically, I would not extrapolate the unique 1970-1980 gold return pattern to all the upcoming market volatility events during the remainder of my early retirement.
As before, I like to start with the good old and simple 75/25 portfolio and then go through the progression of four different Golden Butterfly-style portfolios:
- Start without the Small-Cap Value allocation, but use a 40% large-cap blend initially, along with 20% long-term bonds, 20% short-term bonds, and 20% gold. (Call that Golden Butterfly Light)
- Replace 20 percentage points of the S&P 500 with Small-Cap Value stocks and use the historical returns for SMB and HML.
- Instead of the historical SMB and HML returns, use the zero-alpha SMB and HML factors.
- Add the 1.2% annualized alpha to SMB and 0.8% annualized alpha to HML.
Please see the simulation results below. The safe withdrawal rates are pretty disappointing. You can tell that this strategy was maximized to make the 1970s look as pretty as possible. Why 1970? It’s because the inventor of that rule has monthly simulation data only going back to 1970, thus missing actual historical worst-case scenarios, such as 1929, 1937, and the 1960s. To nobody’s surprise, in exchange for better results from 1970 to 2025, you will experience worse results elsewhere. It’s like squeezing a balloon. Let’s go through the different portfolio’s withdrawal rate stats:
- The Golden Butterfly without SCV (Golden Butterfly Light) would do slightly better in 1929 and even in 1968. However, the 1930s and 1940s, decades that posed little to no headache with the 75/25 baseline portfolio, now look awful. The post-1926 failsafe was only 2.80%, much lower than the 3.39% in the baseline. It’s a 17.4% lower retirement budget.
- The standard Golden Butterfly, with SCV and using actual historical SMB/HML returns, performs significantly better than the Butterfly Light but still underperforms terribly in the 1930s and 1940s. Yes, you get an absolutely amazing 5.22% SWR in the 1970s, even for a 50-year horizon. That’s what data-snooping, in-sample bias, and hindsight bias look like. But it comes at the cost of far lower sustainable withdrawal rates in other decades and market conditions.
- Results only get worse from here. For example, when the Fama-French Factors have zero alpha, the 1930s and 1940s are now the new worst-case historical cohorts. The overall failsafe is only 2.70%, much lower than what the 75/25 would generate.
- Even with the Vanguard-approved factor alphas, you still can’t manage to bring the overall failsafe above 3%.
The punchline here is that the Golden Butterfly portfolio was optimized specifically with the 1970s, especially the December 1972 cohort, in mind. It will perform poorly outside of that time frame and in different market environments. The biggest problem I see with this approach is that you have only a 40% equity allocation. With or without SCV, that might be too low to sustain a 50-year retirement. Stay away from these fashionable, shiny objects that are only cockamamie, overfitted data snooping exercises.
Conclusion
Can small-cap value stocks miraculously increase my safe withdrawal rate to 5% or even 5.5%? If I add the historical and unadjusted SMB and HML factor returns to my portfolio, i.e., I believe that the historical outperformance will repeat, I can certainly increase my safe withdrawal rate, though not all the way 5%. But I’m doubtful the historic SCV performance will repeat. I don’t even have to assume that the poor returns of SCV of the last twenty years need to continue. If I merely assume that small-cap stocks and value stocks add zero extra alpha going forward, then safe withdrawal rates will appear worse because we add more equity volatility with little extra return, which is unfavorable from a Sequence Risk perspective. I called that “Di-Worse-fication” in my post last year.
Admittedly, shifting about half of your 75% equity allocation into small-cap value can marginally improve your SWR simulations if the Vanguard expected return projections materialize as planned. If that’s your thing, I wish you good luck, but I would not take that chance. I would also not recommend raising the withdrawal rate by much more than a few basis points. Don’t go to 5%, and certainly not 5.5%, as recommended by some financial influencers, though.
I also issue a stern warning about any exotic and overfitted portfolio allocations that were optimized to make only(!) the 1970-2020 interval appear favorable, such as the Golden Butterfly portfolio. A repeat of some of the other recessions and bear market scenarios, such as the 1920s, 1930s, 1940s, and 1960s, could significantly and negatively impact your retirement safety if you had picked a withdrawal rate above 4%. Therefore, my recommendation remains the same: Don’t listen to the noise and hype. Instead, keep your retirement portfolio simple. 75/25 is a great starting point. Customize your withdrawal rate to account for your unique parameters, such as future pension and Social Security benefits. Monitor market valuations, i.e., scale back withdrawal rates in light of expensive CAPE ratios. You should have a relaxed and safe retirement.
Please leave your comments and suggestions below! Additionally, be sure to explore the other parts of the series; see here for a guide to the different parts so far!
Title Picture Credit: WordPress AI
To be fair, gold has only strated trading freely in August 15th 1971 after Nixon ended dollar/gold convertibility. Before that the price of gold was fixed under the Bretton Woods system (1944) at $35 per ounce. Before the Bretton Woods system, the price of gold was fixed/ or heavily controlled. 1900-1933 : $20.67 per ounce. 1934-1944 : 35$ per ounce.
Good point. I mentioned that issue already in Part 34, the post about gold. So, we should take all gold simulation results with a grain of salt. But I wanted to give the Golden Butterfly strategy the best chance to shine.
Very interesting to see the small-cap and small-cap value filtered to include the more recent decades of underperformance. I’m curious if you have evaluated large-cap value relative to the total market index or large-cap blend portfolios. I suspect, given the underperformance of large-cap value in recent years (due largely to tech company outperformance), that while adding an allocation to large-cap value may not produce much alpha, it may reduce volatility in down markets and could help sequence risk given today’s high valuation of the market as a whole — a not insignificant impact for those of us already in the retirement period. Your reference to Vanguard return projections made me think of this, in that Vanguard has a large-cap value index option…
Between small and value, the main driver of the SCV underperformance was value not small over the last 20 years. So, I don’t see much hope for large-cap value either.
Given the current valuations, what is the safe withdrawal rate for an early retiree at age 40 in 100% world stock index fund?
Sub-3%.
Thank you, Karsten, for your great analysis. As always it is well done and very informative.
Maybe a couple of thoughts on Tyler’s Golden Butterfly: One compelling advantage of the permanent portfolios are a lower drawdown percentage and “ulcer index” I don’t think I could withstand the drops of a 75/25 portfolio over 50 years no less, or over 30 years. But I can control my spending and reduce withdrawals should the markets drop. Another great advantage to me is the ability to balance between 5 asset classes, and not just 2 in times of market swings. I respect and understand the desire to “guarantee” a safe withdrawal rate, but most of us FIRE people have way more assets than needed to support a 4% SWR.
I also like investing my money now into SWR which are much less overvalued than the large caps at a CAPE of >36. But that is a minor quibble and a bit of a timing discussion.
Thank you again for making all you work public, this is such a great service for us all and it is a great learning tool. All the best to the “Goldgrube”
It’s compelling indeed. But only post-1970. You would have had some bad ulcer-inducing drawdowns pre-1970.
I retired in January 2022 at 59. I have constructed my Rollover IRA for a 3.5% Withdrawal rate over 40 year. I withdraw on a Quarterly basis to coincide with the Tax man rules. I rebalance my portfolio on a yearly basis…or when prudent-> Taco trumps Massacre day is a good example.
For this article…IMO, it is prudent to have a true Value Stock ETF and Growth stock ETF in one’s Retirement portfolio. No need for Small caps. Just take a look at the past 4 years. I will use PVAL and QQQ to prove a point.
2022 Total Return: PVAL (-2.6%), QQQ (-32.5%), SPY (-18%).
2023-2024 Total Return: PVAL (+41.5%), QQQ (+95%), SPY (+53.5%).
Those drawdowns and eventual rebounds are Significant. To me worthwhile to maintain Value/Blend/Growth ETFs in my portfolio. I try to hold Category beating/ SPY beating ETFs in my portfolio.
You can also look at Decades for this Value/Growth stock relationship.
January 2000-January 2010: VASVX (+111%), QQQ (-49.6%).
January 2010-January 2020: VASVX (93%), QQQ (+411%).
Again that is a pretty significant return difference.
My Rollover IRA Stock portfolio: CGDV, CGGR, PVAL, IDVO, DIVO.
Impressive fund! Let’s hope that this PVAL ETF is not just lucky and can keep up that winning streak. History is not on its side. A lot of lucky funds then fizzle eventually.
Also: the 0.56% expense ratio is tough to overcome in the long-term.
Stephen, you make an interesting point, and the decade return comparison you’ve provided is very useful to readers. Although it could be that the specific fund you reference may be an anomaly, as Karsten points out, the concept of large-cap value index funds providing some ballast in a drawdown period relative to the total market index is reasonable. This is especially true given recent years’ underperformance of value stocks and the potential for a sequence risk problem given the high CAPE ratios in today’s overall
market.
I would hope that other Stock Asset classes would be considered also to see if the SWR could be increased. I have read Randy L. Thurman’s book: The All-Weather Retirement Portfolio.
70% Stocks: 49% US-> 23% LCV, 12% LCG, 14%, SCV. 21% Intern’l-> 15% DM, 6% EM.
30% Bonds: 17% US short-intermediate term, 10% Intern’l, 3% HY.
The only other stock asset class I incorporate is International Stocks. I use IDVO in Rollover IRA, DIVI in Roth IRA.
I use DODLX/BINC/ICSH for Bond portfolio.
Yes, there are more exotic flavors than I can simulate and write about on this blog. Some will do well, some will suck even more. It’s hard to know which other even more overfitted style will do well going forward.
Helpful information, thank you! Might small-caps, mid-caps, (and international) add diversification when CAPE for the S&P is high?
International may help. But don’t get your hopes too high. If go through another deep recession, it’s going to impact the whole world, like 1929, 2001, 2008, 2020.
I hear you. I have a remaining question of why a high CAPE means lower returns for US stocks–but mid/small cap and international are more reasonably priced. Does this not factor into long-term portfolio sustainability? I believe CAPE plays into Vanguard’s forecast stating international will outperform US long-term. In your view, is CAPE only useful in forecasting long-term S&P performance? Or am I missing something?
The CAPE is a useful measure to predict index returns over a 10-15-year horizon. From a statistical point of view, that’s a time series phenomenon.
Using differences in CAPE to decide the relative valuation of different parts of the market to help in stock picking is a different ballgame: statistically, yo are now playing the cross-section instead of the time series. I don’t think the CAPE is useful for that purpose.
Did not find this article particularly convincing. Completely ignores global equities outside of the US, but of course Big ERN is not an advocate of international investing, so I guess we can just ignore it?
International developed markets and emerging markets post-2006 have featured robust premiums in Small Value.
And why is Big ERN not an advocate of international investing? Because US is “good enough?” or because it has done better recently? Not clear, but the same arguments he uses against Small Caps can be levied against the whole US vs exUS debate – the “US premium” has largely been the result of very narrow periods post WW1 and WW2, with of course the last decade as well (but nobody knows if that will be given back, with time).
And of course we have to tweak our model to remove the “ill gotten gains” from factor investing (what about all the “ill gotten gains” from the valuation and currency expansion favoring US stocks the last decade?) in order to make Factors look worse.
I will continue including Factor products in my portfolio and use the larger global set of history to drive my decisions rather than this very cherry picked article by Big ERN.
International is a different issue. I distilled the SCV issue for US investors. I can’t include the whole kitchen sink. I didn’t include real estate either, oh no, run for the exits.
Recommendation to you: start your own blog and then you have control over what you include in your simulations.
I believe US investors should have International stocks to be adequately diversified. So from the perspective of a US investor, the topic of Factor products is still relevant.
However, even the case for US SCV for those that choose to be US only investors is still highly relevant even in the context of what HAS happened outside the US. For instance, Japanese only investors that had decided to use market cap weights had to bear a significantly poor outcome whereas Japanese Small Value investors made respectable returns post-1990 Japanese bubble. Valuation and Bubble risk is real, and can have very longstanding consequences if the market re-rates negatively.
The mere fact that Factors “stopped working” for US SCV only post-2006 isn’t necessarily something that Factor investors should discount. If Factors always worked all the time, they would stop being risk factors. And we see this in the data. While relatively rare, there are certainly cherry picked start/end dates where SCV underperforms a MCW approach. Of course this is also true for the biggest factor of them all – the Market factor. Certainly, US market cap weighted Stocks in the late 60s went on to underperform bonds for nearly two decades. That’s risk for you, whether you choose to spread that risk amongst just the Market, or if you also include smaller companies with higher discount rates.
All fair points. I can imagine that SCV may be better intact abroad, so if you still have faith that for international stocks you have some additional beta exposure from SCV and you deserve an additional expected return, be my guest. But international SCV will not offer you free alpha.
Dear Karsten, Can you please talk about if you think it is worth it to get into DFA and factor style ETFs opposed to just simple index funds? I am at Vanguard primarily and I don’t think they are available there. I have been following DFA for a while now, but never wanted to go to an advisor to have access to them, but I believe you can now access them from perhaps even Fidelity and Schwab? – Thank you for your time and thoughts, I am a huge/long time fan and listen to every interview you do.
Thanks!
No, I would NOT use DFA advisers. Again: SCV is an old style whose time is now over. I would not want to spend an additional 1% annual management fee to have access to those now only average return expectations. I am sticking with Fidelity index funds.
Thanks for your comment, Karsten! This video is what prompted me to ask you the question; https://www.youtube.com/watch?v=mqIHa6URUPk – He talks about some research/papers and how the DFA index based ETFs buy and sell funds with just slight “management” to avoid obvious pitfalls like IPOs. I am a 100% passive investor, but this just made me think twice, and some of the academic topics and technical transactions that have to take place are outside of my realm of understanding. Thanks for your feedback if you have any other thoughts, we appreciate it!
He’s a mutual fund salesman and wants to steer you into active funds, so there is no surprise he doesn’t want to advertise passive funds. Some of the issues mentioned there are non-issues. For example, the S&P 500 addition/subtraction issue can be easily avoided by smart index providers that will already front-run those changes. A lot of providers also set up their own proprietary indexes and avoid two issues that way: the index constituent changes and the license fees for S&P.
So, this video is a total nothing-burger.
From this comment it seems you think proprietary indexes are a good thing. Isn’t that exactly what DFA and Avantis do?
Let me explain it slowly:
A proprietary index …
… is good if you replicate the S&P 500 with a correlation of 0.999, but you save the S&P license fee and just use your own, e.g. Fidelity 500.
… is bad if you replace a broad index with some active stock picking index, and the index vastly underperforms the S&P 500 for about 20 years and also creates an insane tracking error and much larger overall variance.
DFA has been around since early 90’s, so you can go back and look at the data and see for yourself and see if it’s worth it vs. small cap value indexes. For instance, DFSVX has a return of 1,092.71% since 7/31/2000, while IWN (Russell 2000 Value Index ETF) has a return of 704.81%. DFA and Avantis both have ETFs now so you can buy them without paying costly 1% AUM with a financial advisor.
Yes, I know that. I did go back that far and determined that DFA is not worth it.
And again, DFSV underperforms the DFSVX, so the funds are not comparable.
LOL DFA and Avantis ETFs are now available widely without going through an advisor and you know it. Crazy you would pretend 1% AUM is still required to invest in these funds.
I knew that.
I also know that DSFV underperforms DFSVX by about 0.40% a year, so the two funds are not 100% comparable.
You can access great Factor funds from both DFA and Avantis through low-cost systematic ETFs.
Do you NEED factors? Absolutely not. You also don’t “NEED” exUS stocks (or US stocks, for that matter). There are many paths to taking diversified equity risk that will likely provide you with a positive real return over time.
I would argue that the overwhelming evidence does suggest that Factors can provide increased returns and more reliably positive equity premiums but it’s not a free lunch – benchmark anchoring (tracking error), and increased risk.
If you are able to withstand those downsides, Factors are likely to provide a significant increase in returns to your overall portfolio if held over the entire lifetime. Certainly, basic index funds (even if they aren’t as diversified and only invest in the US) will likely be just fine too, but you’re more than likely going to have to withstand longer periods of sub-par equity performance (less reliable).
Excellent post as always, Karsten!
Very timely and you provide important context, data, and counter-arguments, given the recent surge of proponents of SCV tilt to raise SWR.
Your SWR series is truly an impressive (and I believe unmatched in the SWR space) body of work!
Thanks for your kind words! I’m glad there are still a lot of folks who agree with me on this controversial topic! 🙂
And now I remember: you were the commenter who inspired me to do this SMB and HML adjustment! Thanks for the suggestions!
You’re most welcome!
And thank you for adding the SMB and HML adjustment (HP-filter), the writeup in part 28 explaining it, and the in depth analysis here.
Thank you for sharing your insights and inspiring for FIRE strategy adaptation. Lot has been going on due to market volatility – tariff wars, recessions fears, Generalized AI emergence. You are a beacon of light in these tough time. Kudos for that. Keep inspiring!
Thanks for the kind words!
Karsten,
Thank you for your work, I have read every single post over the years. You were my intro to the SWR field and tremendously valuable to how I think about it. However I fear this post confirms dogmatic beliefs as much as it offers a data driven analysis.
You say the SCV premium will not materialize in the future because everyone knows about it… ok, maybe, my crystal ball’s cloudy. But why 2007? Fama and French published their first model in 1992, Dimensional Fund Advisors launched its first fund leveraging the model in 1993, and Fama and French won the Nobel in 2013. There’s no reason to pick 2007 as the date the switch turned off for SCV other than data snooping…
As an alternate hypothesis, there were huge economic forces at work starting around 2007: the Great Financial Crisis and ensuing regulation taking the oomph out of financials (SCV indices typically have ~2x the allocation to financials as the S&P), and years of ZIRP launching big tech stocks to the moon. Those are similarly plausible explanations for the shift. Again my crystal ball does not say whether they’ll continue, but there are certainly historical precedents for economic cycles like this emerging and then reversing.
To go back to the idea of everyone and their grandma learning about SCV: What about the tremendous overperformance of US Tech Mega-Cap (ie the bulk of S&P returns since 2007)? That’s hugely more obvious than the SCV premium; why should we expect that to continue and why should we remain invested in S&P? For that matter, everyone knows indexing is better than active management, so why is buying a broad index still a good idea? These are straw-man arguments, but my point is if we’re going to base an analysis on historical data, picking and choosing which historical data to use without a clear hypothesis is akin to using a crystal ball and is not compelling to me. The SCV premium as proposed by Fama and French is not a mathematical arbitrage, it’s additional risk factors that investors require compensation to hold, and there’s a solid theoretical underpinning to expect it to persist indefinitely.
The headline and takeaway paragraphs also undersell the benefits of SCV in your analysis. Using your HP-filtered, plus Alpha SCV returns, allocating to SCV reduced the SWR in one decade (1930s), had effectively no change in 3 decades (1910s, 1920s, 1960s), and increased the SWR in the other 6 decades. Sounds great to me! I understand the absolute worst-case SWR (1960s) doesn’t improve, but it makes more sense to take a trade that leaves you better-off or the same 90% of the time, and think about ways to adjust if you find yourself in that absolute worst case scenario (I.e. flexible withdrawal rules). Your point seems to be simply that SCV does not push the SWR up to 5%, which seems correct, but your headline and other commentary throughout implies SCV is bad (diWORSEification), which is not correct.
Throwing CAPE ratios back into the mix, although there’s a lot of debate I find reasonable about whether they’re mean-reverting and therefore predictive, SCV looks hugely better than S&P on this metric. The CAPE ratio for SCV seems to be about 15 currently (based on a quick Google search; perhaps someone has a better source), while for the S&P it’s 35. If we’re going to rely on CAPE and do market timing, this is the best time to tilt towards SCV in a generation!
And regarding the drive-by against gold: back in Part 34 you found that owning a bit of gold helps SWR, but… you weren’t going to buy it because… I guess because you didn’t want to? This post abandons all that analysis. As you noted in Part 34, “gold seems to work well both during inflationary recessions (1970s) but also during the bad demand shocks that conjured up fears of a deflationary scenario (2008/9).” Of course the data on gold only goes back to the 1970s because before that the dollar was pegged to it. Analyzing the Golden Butterfly’s performance pre-1971 essentially throws away 1/5 of the portfolio’s value since gold had no return at all in that time. It is a valid complaint to say we can’t back test the Golden Butterfly in this timeframe; it is not a valid finding to back test it with garbage data and say its performance was disappointing.
The analysis seems to be picking nits with the data the opposite way for gold vs SCV: For SCV we have 80 years of great data, but 18 years of iffy results, so you’re arguing to throw it out. For gold the data series effectively starts in 1971, then we have 55 years of pretty good results, but again you’re throwing it out. Essentially the hurdle you’ve set is to only consider assets with solid data and strong performance going back to at least 1926, if not the 1800s. There’s a valid argument for that position (I always picture the art you had in an early post of the man covered head-to-toe in bubble wrap), but I’d say it’s quite conservative and forecloses almost all attempts to improve portfolio design.
All that said, no one has to like SCV or gold or any asset, clearly it’s everyone’s prerogative to hold whatever they want in their portfolio. Even when I disagree I find your research a robust long-sample perspective, and I agree many others excessively overfit their findings (to do a drive-by of my own, I always puzzled over whether the “Weird Portfolio” would really hold up in the future). But I believe in this case you’re over-discounting both gold and SCV, and with solid data and theoretical underpinnings for why they should help a portfolio, there’s a good case to hold both.
Agreed.
The “everyone knows about it” argument does not stand up to reason. If SCV represents genuine additional risks, the fact that everyone knows about those risks is largely irrelevant. Investors are still pricing the securities in a way that should more often than not generate a premium for bearing that extra risk.
Market Cap Weighting, the first original factor, was also discovered and was then popularized into index fund products. Has the equity risk premium gone away simply because it was discovered? Far more inflows are going into Index products than SCV funds, and it would not surprise me if the equity risk premium for the US market in particular (despite being great for the last decade), is now markedly lower than what we have historically experienced. SCV in comparison has valuations roughly in line with their own history, with the Value spread relative to the Market still being at near historical levels. This suggests expected premiums today should be much higher. It’s not really any surprise why premiums post-2006 were low; valuations for SCV were extremely high relative to their own history and the Value spread compared to the market was small.
Certainly, there may be some factors that weren’t genuine proxies for additional risk that may indeed be nothing more than anomalies that get arbitraged away from the market, but I can look at the volatility and drawdowns historically for SCV compared to the market and realize that these do indeed have quantifiably higher measures of diversified risks compared to the Market.
Fama and French didn’t win Nobel Prize for the factor research. French didn’t win the Prize at all, and Fama won it for other unrelated work.
About the timing, here’s my theory: in response to the Fama-French research even more money rushed into the market and built up a bubble that peaked in 2006 and has since deflated. Since 1/1993 to today, SCV and LCB had an almost identical CAGR.
https://www.portfoliovisualizer.com/backtest-asset-class-allocation?s=y&sl=39zdp0XGxNWMTLX4t9QS6x
SCV kept outperforming until 2006 and then the bubble deflated again. It’s a theory.
I never meant to bash gold. I acknowledged that gold has good diversification potential. I stressed that Part 34. I also believe that gold is the one redeeming feature of the Golden Butterfly. But much of the historical outperformance is due to SCV and the performance of SCV during the 1930s will not so easily repeat.
75/25 and chill. Love it
Wasted years listening to the noises aka risk parity radio for nothing.
Big ern rocks
But 75/25 of what? Can I use 75% ITOT and 25% IEF Etfs? (anyone with alternatives, – all but Vanguard which I’d never do business with again)
75% ITOT is great. 75% FXAIX or FSKAX or a mix works for me.
Thanks! Stop wasting time and read 62 blog posts on SWRs. 🙂
Thanks Karsten. As always it is very interesting. I would love you to write about the case of your fellow europeans. What would be the SWR in Europe for 75% MSCI world index/ 25% Bond (speaking of which, would you recommend EU bond or international hedged?). Thanks a lot
I don’t have returns for EU stocks and I don’t have CPI numbers. Maybe some other EU-based blogger wants to tackle that.
If, as you hypothesize, the small cap and value premiums have gone away because they are now well-known and widely traded, what would we expect that to look like from a return or valuation standpoint? Wouldn’t that mean that value stocks would need to get more expensive compared to growth stocks and small cap stocks more expenses vs large cap? Is that what we’ve seen? No. In fact, we’ve seen the opposite. Value spread between value stocks and growth stock is near all time highs. Its large-cap growth stocks that have gotten more expensive. The reason there has been no premium for these factors the last 20 years isn’t because small cap and value stocks have severely underperformed, it’s because large growth stocks have drastically outperformed historical norms (again largely due to increasing multiples). This fact actually should imply a stronger premium for these factors going forward, no?
Part of the earnings multiple expansion can be explained by poor earnings measurement quality. A lot of tech companies have seemingly low earnings, but they also invest a lot. Which will account for future growth. On the other hand, a lot of value stocks are late-stage corporations with sclerotic growth.
So, I agree that there is some concern about relative earnings multiples, but some of that is also legit in today’s economy.
How much of the SCV alpha boiled down to simply buying at a lower valuation?
According to Yardeni(1), the small cap S&P600 currently has a forward PE ratio of 14.7, compared to the large cap S&P500 at 21.2. In other words, earnings from these small caps are now over 30% cheaper than earnings from large caps!
And as others noted above, the underperformance of SCV could be explained as the overperformance of large cap tech companies. So why not interpret small caps today as an opportunity to lift the SWR, based on what we know about valuation?
(1) https://yardeni.com/charts/stock-market-p-e-ratios/
It could certainly mean that Small will outperform large going forward. But it could also mean that small is cheap for a reason. Relative valuation of stocks with just that one year forward snapshot is hugely unreliable to perform stock picking. The PE value differences could be due to different growth outlooks. Unless you do a full-scale long-horizon earnings forecast and you perform a dividend discount model (DDM) or residual earnings growth (REG) model, I’m not taking these cross-sectional comparisons very seriously. In fact, even with DDM and REG models, stock picking is hard enough, as I’ve observed when working at BNY Mellon Asset Management.
At current valuations, what SWR would you recommend for a very early 30 year old retiree using 75% VTI and 25% T-bills?
I would run my simulation toolkit with 75% US stocks + 25% Cash and see what’s the historical failsafe. Today’s equity valuations are again similar to the historical worst-case scenarios.
Karsten, thank you for this post and the many excellent ones before. Using your toolkit now as is considering the high valuation is sensible. I was wondering if after a drawdown the depth of the drawdown could be an alternative input to the CAPE ratio, to allow us to be more aggressive in the SWR that the historical worst-case, considering mean-reversion. In that case, instead of simulating worst-case starting today, would it be safe enough to exclude all the starting point that were closer to all times high (% of all times high or a smarter method) when calculating the SWR? I would love to read a post about that, and if that is valid see this as an option in the toolkit. Hopefully we have enough time until we would need to use this option 🙂
That’s what I recommend: the drop from the all-time-high equity level should be considered.
For example in Part 54, I went through a case study. See the section “Even in the static withdrawal simulations, 4% and even 4.5%+ may work again!”
Your caution about the Golden Butterfly’s selective data 1970 on is valid. But then doesn’t a similar criticism apply to your 1927-today Schiller data used in overlapping sequences? Wouldn’t you at least want to bootstrap it a bit to try to get around the limitations?
No.
I use data since 1871 for my SWR series.
The equity risk premium is much more statistically significant over time than the SCV premium.
If the answer to the question in thr post title is NO. What will then? Simple question
Nothing. Markets are efficient. The average stock picking scheme (that’s than SCV really is) will not outperform the market.
Then again, I trade options as a reliably alpha source. But that’s not suitable for everyone.
This response made me laugh but is really one of the best statements that can be made. I wasted a lot of time trying to optimize different equity flavors before finally settling on just looking at it as a single Equity bucket. Do you think it’s market timing to keep ~20% of the portfolio in a Dynamic bucket that can move between stocks to bonds and back based on equity risk premium (1/PE-10 year) being close to zero or negative and moving back once it’s a more positive value? Looking for a way to pare down risk but still using a rules based approach using valuations and rates.
Market timing is just as difficult as stock picking. And it can go against you for a long time before it finally pays off. But yes, I generally like the idea of a Stock-Bond allocation based on the differential between the stock vs. bond yields. But caution: the 1/CAPE is a real return, while the 10y yield is nominal.
When I read about this it was described using the normal P/E instead of the CAPE. Is there a downside to that? Was considering going between 60/40 and 40/60 using this rule so would always be invested to some extent, but with (in my mind) a lot less volatility. Switching rule would have both entry and exit points and a wide band between them where you “do nothing” (hold what you have) – low amount of switching. I tried some simulations and it looked OK.
The same problem still applies: 1/PE is real, while the bond yield is nominal. But It’s a good start, for sure.
Fascinating. You only prove “bullshit in, bullshit out”. You assume what you want to prove, that markets are efficient, although this was debunked over and over again. If value does not provide alpha, although it did historically – besides the last nearly two decades, which will revert – then you can’t increase the withdrawal rate.
Someone with your poor reading comprehension should refrain from calling other people’s research BS. If you had read the post and understood it, you would know that I very generously attached a sizable alpha to SCV, in line with what Paul Merriman assumes and what the Vanguard expected asset return model uses. And even assuming that, there’s no clear benefit from SCV. Certainly not enough to raise the SWR by a full percentage point.
Regards, Karsten. It’s as fascinating as ever. Writing on the situation of your fellow Europeans would be very appreciated.
Well, to my fellow Europeans, I’m now US-based. So, I’m not in a rush to make this toolkit more EU-relevant. Sorry! 🙂
I really like reading through a post that can make men and women think. Also, thank you for allowing me to comment!
You bet. Thanks for commenting.
Hi Big ERN – I’ve found your writing to be hugely useful; thank you. In one of your earlier glidepath posts, you mentioned almost as a throwaway that you’d like to simulate CAPE based dynamic asset allocations, but unless I missed it I’m not sure that you have.
It seems to me that there may be opportunities to eke out nontrivial gains on the margin by using an approach like the Merton rule to target allocations based on the forecast equity premium (e.g. CAEY vs. TIPS) with risk tolerance a factor of time to (or from) projected retirement.
Is this something you’ve looked at, or something you’d consider looking into?
I have that on my to-do list: essentially a tactical asset allocation rule, where you allocate S/B and maybe even S/B/cash based on valuation and momentum rules.
It’s not a panacea but it may help a little bit. I will just have to wrap that up and write a post about it, haha.
Hey Big ERN,
Long time reader and big fan of the blog.
If you have 10 minutes to spare, I’d love to hear your take on Ben Felix’s new video comparing DFUS to VTI. Seems too good to be true?
https://www.youtube.com/watch?v=lMdGA0ePXzI
I was completely unaware of some of the (minor) shortcomings of index funds so this fund seems like a promising way to squeeze out a few extra basis points in return with no extra work.
Thanks again for the blog,
Ryan
It’s an actively managed fund. I will stick with the advice from Jack Bogle rather than Ben. That said, the DFUS is probably one of the least offensive funds. The expense ratio is low and it tracks the overall market extremely closely. I don’t expect this fund to deliver sustained and reliable outperformance. YTD it underperformed in 2025. So, like a lot of hyped funds with some great performance in the past, this one will fizzle, too. Likely right after Ben Felix gives a buy recommendation.
But again, it will not destroy your financial picture.
I noticed the Vanguard expected return chart has “total stock market”.
Wouldn’t it be easier just to find find vanguard’s alpha by Value – tsm; Small – tsm?
Your suggestion is also a spread, so its not easier; it’s the same complexity of calculation. Your method would also be incorrect. The HML and SMB spreads are exactly the way I constructed them, i.e., Fama-French HML is value minus growth and SMB is Small minus Large.
Of course, you could also calculate (Value-TSM) – (Growth-TSM), which is still Value-Growth, as I calculated it.
Hi Karsten, the community is lucky to benefit from your well-researched and well-written posts. I read through your SWR-62 paper and really appreciated the effort that went into it. A few points gave me pause. The two-sided HP filter builds in look-ahead because it uses future data, which can make the factors look different than what investors would have seen at the time. The step where you equalize CAGRs across portfolios also pulls down the factors more than intended since it penalizes them for volatility rather than just removing alpha.
It also seems optimistic to treat SMB and HML as if they were investable tilts. Real funds do not behave the same way, and their costs can vary quite a bit, so a single 0.20% fee may not capture reality. The use of Vanguard’s long-term premia forecasts as direct factor alphas is another leap since those forecasts were not designed for the academic factors after detrending. Finally, the call that HML went negative after 2006 depends a lot on the filter settings and endpoints.
All that said, the main message comes through clearly. Historical SMB and HML lifted withdrawal rates in certain periods, but looking forward with modest premia, the effect is small.
Thanks for the compliment. And I got good news. The two points that gave you pause shouldn’t. They are both complete nothing-burgers:
1: You take a true feature of the HP-filter, but misinterpret it in the context of this exercise. Yes, the HP-Filter will indeed look different in expanding windows versus the whole window. However, that’s irrelevant for this exercise. I am the researcher, and I detrend the returns, assuming that those were the returns everyone faced starting in 1926. It is a proper thought experiment, and the investor wouldn’t need the entire history of SMB+HML to invest in SCV. The investor doesn’t make the HP-Filter calculations. I did.
2: HML and SMB are indeed investable factors in multiple ways. 1: You can look up what stocks fall into the SMB and HML long/short portfolios. 2: You can certainly run factor models of some fund and estimate exposures and invest in an ETF that will give you x% HML and y% SMB.
Maybe what you’re implying in your #2 point is that HML and SMB are not that cheap to replicate. Investing in SMB and HML would have been crazy expensive due to illiquidity and transaction costs back in 1926. I’d agree with that, of course. But again, as a thought experiment, I’d assume that the FF factors are investable for today’s investors and we can certainly study how a 2025 investor with today’s technology and today’s brokerage account features and ETF lineup would fare if we ever had a repeat of past return windows.
Article suggestion (I know you get plenty):
We are in the 1990s tech bubble, somewhere between 1997 and early 2000, with between zero and three years to go until the bear market. We don’t know exactly where we are in relation to the peak, but we plan to retire within 2-3 years.
What is the ideal asset allocation, assuming the availability of all the tools an investor has at their disposal in 2025 but with late-2025 interest rates on bonds? Which performs best over the next 5-6 years and does not fail (defined as: only 25% of original balance remaining after 25 years)?
-all bond portfolio?
-barbell portfolio of bonds plus call options?
-short portfolio (e.g. SQQQ, SPXU, or SH)?
-cash / CD centric portfolio?
-foreign currency / ex-US bonds
Good suggestion.
I’d recommend a glidepath to increase the bond share until retirement and then a GP back into equities when you start retirement.
But it would be nice future post. Thanks for suggesting that!
Thanks for your blog! As someone who runs a small creative blog I appreciate how much time and effort put into creating work like this – for free – for others. Especially after reading some of the snarky critiques in the comments.
Anyways, I was curious, I have this hypothesis about SCV. I have no data to back it up so perhaps it’s just a poor opinion (😂), but it seems to me that perhaps the reason small-cap underperforms, is the increase of public-private bailouts and partnerships, plus what seems like a lack of regulatory interest in limiting monopolistic practices or mergers, and the sheer amount of capital large cap companies have to throw their weight around and sustain games over their smaller competitors. To me it feels like a unique time, although uttering a “this time is different,” never really sits right. But if we look at 2008 bailouts, the fed money injections post COVID and the widening gap between Wall Street and main street and it doesn’t seem at least to me – that SCV is a good bet (beyond what you already lay out).
Anyways, I was curious with your economist hat what you think of that conjecture. Thanks!
I can see that theory at work. Crony capitalism is getting worse, which helps politically connected, large corporations.
Of course, this would only explain the SMB effect, not the HML, which is the real head-scratcher. The SMB has been flat, but the HML has performed terribly since 2006.