Can the new Bill Bengen Portfolio Increase My Withdrawal Rate to 4.7%? – SWR Series Part 65

September 28, 2026 – Welcome to a new installment in the Safe Withdrawal Rate Series. You can find the series landing page here. Today’s topic is the new portfolio allocation recommended by Bill Bengen, the inventor of the traditional 4% Rule. He’s making the rounds recommending an innovative portfolio allocation philosophy (his words, not mine) that will ostensibly raise your withdrawal rate to between 4.7% and 5.5%, depending on how flexible you are. And potentially 7% once equity valuations normalize again. I’m mostly unconvinced of this new approach, and I’ve responded to inquiries about it here and there. I also posted a short note on the ChooseFI site. But it’s always best to have a dedicated post in my series with detailed simulation results.

Let’s take a closer look…

The new Bengen Portfolio

Bill Bengen proposes a portfolio with 55% equities, 40% intermediate bonds, and 5% short-term government securities, such as 3-month T-Bills. He further specifies that the 55% are in equal shares in:

• US Large Cap Stocks (e.g., S&P 500)
• US Mid-cap Stocks (e.g., S&P 400 Mid-Cap index)
• US Small-cap stocks (e.g., S&P 600 Small-Cap index, Russell 2000)
• US Micro-cap stocks (e.g., Russell Micro-Cap index)
• Non-US Stocks (e.g., MSCI World Ex-US)

Here’s an article and an interview with Bill on MarketWatch with more details. The justification for spreading the equity portion like that is that broader diversification across markets and indices raised the safe withdrawal rate from 4.15% to 4.7%. He further argues that if you no longer require a 100% success rate and have some flexibility in spending, you can increase your withdrawal rate to about 5.25% or even 5.5% per year. There is no explicit mention of what he means by “flexibility,” i.e., would the 5.5% withdrawal rate imply an explicit ex ante success rate target and/or how flexible a retiree would need to be, i.e., how deep and how long were the retirement spending cuts?

Preliminary Calculations: Implementing the new Bengen Portfolio with ETFs

How would one implement this new Bengen portfolio today with our current menu of ETFs? On the T-bill side, several options exist. The ETF with the longest history seems to be the SPDR ETF (ticker BIL), with return data going back to 2007. iShares started its SGOV (with a much lower expense ratio) in 2020. Vanguard’s VBIL only started in 2025, but has the most competitive expense ratio. Because it covers more years, I will stick with the BIL ETF for now.

For intermediate bonds, I use the iShares IEF, which holds 7-10 year Treasury bonds. I know that lots of you prefer Vanguard, which has the VGIT ETF in this space. However, VGIT has a shorter history, so I can’t simulate a window long enough to include the Global Financial Crisis (GFC). Moreover, VGIT has a slightly shorter duration because it includes Treasuries with 3-10 year maturities. So, the IEF better fits the traditional 10-year U.S. Government benchmark bond.

For the U.S. large-cap portion, I use the iShares IVV (since 2000) rather than the Vanguard VOO (only since 2010) to cover the Global Financial Crisis. Sure, the SPDR offering SPY has an even longer history, dating back to 1993, but it also has the highest expense ratio of the trio, and I don’t need the simulation going back to 1993, since my calculation is constrained by the ETFs’ availability.

For the exotic equity allocations, I pick the following ETFs:

• Vanguard’s VO for the Mid-Cap US exposure.
• Vanguard’s VB for the Small-Cap US exposure.
• iShares’ IWC for the US Micro-Cap index exposure. I found only two options from prominent providers with decent return histories. The other would have been a mutual fund from the DFA network, which isn’t really useful because you need an advisory relationship with a DFA-approved advisor to access it. IWC seems like the obvious choice here.
• Vanguard’s VEU for the non-US stock market exposure. To my knowledge, it has one of the lowest expense ratios available. VEU also has a very long return history going back to pre-GFC.

Before we even simulate any historical retirement cohorts, let’s perform some preliminary calculations to understand how the Bengen portfolio, implemented with actual ETFs, would have performed recently. Let’s pull some stats from my ETF return database. I have monthly total return data (nominal total returns, i.e., dividends reinvested) since 2007, so we have close to 20 years of returns, covering two recessions (2008/9 and 2020) and three bear markets (2007-2009, 2020, and 2022), as well as quite a few corrections and other drama along the way.

I like to start with a correlation table of the seven ETFs:

Clearly, there could be some diversification potential: the exotic ETFs all have correlations well below 1.0 with the IVV (US Large-cap blend) and very low correlations with bonds and bills. Among themselves, Bengen’s exotic indices also had relatively benign correlations: as low as 0.736 between non-US and US Micro-Cap.

But if I calculate the standard deviation of the two competing 55/40/5 portfolios, both with 40% IEF and 5% BIL, one with 55% in IVV and one with 11% each in VOO, VO, VB, IWC, and VEU, we get this surprising result: The standard 55/40/5 portfolio with just one equity ETF (IVV) had an annualized risk of 7.9%, while the Bengen ETF portfolio’s risk was 8.8%. Even worse, the annualized return has also lagged since 2007.

Well, clearly, the underperformance is simply because the exotic ETFs lagged the S&P 500 during this period. But the higher risk of the ostensibly diversified portfolio deserves some additional analysis. So, how is it possible that we created inferior risk characteristics, despite the seemingly attractive correlation matrix? Very simple: Despite their less-than-zero correlation, the new ETFs had a standard deviation so far above that of the S&P 500 (17.77% to 22.26% vs. the low 15.52% in the IVV) that even with a correlation slightly below 1.0, there is no diversification benefit; please see the table below with the individual ETF return stats:

Will less exposure to Bengen’s exotic asset classes help?

I can already foresee one objection: Maybe old Bill Bengen went a little too far, allocating too much to volatile funds, and a smaller allocation to exotic equity flavors would have performed better? So, what happens if we add the “di-WORSE-fying” ETFs in smaller increments? In my first exercise, I remove four percentage points from IVV and allocate 1 percentage point each to the new ETFs; please see the bar chart below. Even from the first step, the new Bengen portfolio is detrimental to diversification. So, the Bengen portfolio’s poor performance over the last 20 years is not due to over-diversification. It never created any diversification, even in the smallest possible dosage.

Alternatively, I can also study how adding the four different ETFs one at a time changes the portfolio volatility; please see the chart below. Not a pretty picture either. Every single new ETF was detrimental at the margin: higher risk and lower CAGR. Thus, every one of Bengen’s ETFs “di-WORSE-fies” your portfolio over 2007-2026.

Calling this new portfolio “more diversified” is the greatest misnomer since calling FTX a great crypto exchange.

The lesson here: simply adding more ETFs with equity correlations in the 0.80s and 0.90s will not necessarily diversify my portfolio, especially if the new ETFs have significantly higher standard deviations than the S&P 500. This “di-WORSE-fication” effect isn’t my opinion; it’s not some ambiguous or esoteric effect up for debate (even though some ditzy, mathematically illiterate financial influencers may think so). Rather, it’s a mathematical certainty that adding all those junk ETFs is truly detrimental if certain easy-to-check, unambiguous conditions are met. For the math geeks, I derived the formula for this effect last year in my “How to Lie with Diversification” post; please see below:

So much for my preliminary calculations. Now I’ll turn to the part everyone has been waiting for: actual Safe Withdrawal Rate simulations. Before that, here are a few more tasks:

How to Simulate the Bengen Portfolio ETFs before 2007

Of course, no Micro-Cap ETF goes all the way back to 1871, or even 1926. For example, the Russell Micro-Cap index has been available only since June 2005, and the corresponding IWC ETF started a few months after that. The MSCI World-ex-USA index only started in 1970. So, we have to make some assumptions about how we simulate Bengen’s new portfolio with my SWR toolkit.

Let’s start with the easy ones. I assume that in my simulation toolkit, the three broad asset classes 1) T-bills, 2) Intermediate US Treasuries, and 3) the S&P 500 index map one-for-one into my already existing asset classes, i.e., the simple 55/40/5 is just an allocation of 55% to the large-cap index, 40% to intermediate bonds, and 5% to Cash (i.e., short-term T-bills). I also assume that there is a 0.05% annual drag on the portfolio due to expense ratios, trading costs, and other fees and inefficiencies.

For the exotic equity flavors, I will stick with the ETFs I used above in the preliminary calculations: VO, VB, IWC, and VEU. However, none of these funds have any data available going back to 1926 and certainly not 1871, which are the popular starting points for historical safe withdrawal simulations. What am I to do? Very simple: I will put on my old financial analyst, statistician, and econometrician hat again and replicate the ETF returns during the time span when they did have return data, i.e., determine what mix of the return series that I already have on my Google Sheet would optimally – from a statistical perspective – replicate these ETFs. That allocation matrix is in the table below. Notice how well this factor model replicates the four exotic ETFs. All adjusted R^2 values are well above 0.9, so the factor models produce a very nice replication fit. See the technical appendix below for more charts and stats.

The estimates are also quite intuitive: Going from mid- to small- to micro-cap equities increases your Fama-French SMB exposure from 42% to 82% and then 126%. They all have mild HML (value factor) exposures, which is also exactly as I would have expected. The equity portion (US LCB and International) adds up to roughly 100%. But of course, the US-based ETFs have mostly US exposure, while the VEU has mostly non-US equity exposure. So this is exactly what you’d expect, but it’s nice to confirm these patterns in the estimation exercise.

We found the exact factor loadings to properly simulate the ETF returns in the early period before the ETFs were available. Thus, we’re now ready for the simulations…

Safe Withdrawal Rate Simulations

I will simulate seven portfolios, starting with my personal preferred portfolio: 75% US equities and 25% US Intermediate bonds. I then successively change one parameter at a time to arrive at the new Bengen portfolio

1. My favorite portfolio: 75/25.
2. 60/40: Shift 15 percentage points from equities to bonds.
3. 55/40/5: Shift another 5 percentage points from equities to cash/Tbills.
4. Shift 11 percentage points from S&P 500 to mid-cap stocks.
5. Shift 11 percentage points from S&P 500 to small-cap stocks.
6. Shift 11 percentage points from S&P 500 to micro-cap stocks.
7. Bengen’s Portfolio: Shift 11 percentage points from S&P 500 to international stocks to arrive at the 5×11% weights in the five equity styles.

Summary:

To translate the factor beta percentages into the parameters in my SWR toolkit (see Part 28 for an intro), here’s the list of parameters I use in the seven different portfolios:

Why would I study the whole progression of portfolios this way? Instead of throwing everything at the wall and hoping something sticks, we should understand what each portfolio component actually contributes. It serves two purposes: 1) it’s academically and intellectually proper to disentangle the effects that way, and 2) knowing what each new component does would also aid the individual investor in both implementing and maintaining this strategy. That’s especially important during volatile market phases, so an informed investor can stay the course and avoid overreacting. For example, my personal finance buddy Joe Saul-Sehy made a similar point recently in the awesome Afford Anything Podcast; see this YouTube video at around the 56:00 mark, where Joe describes his reasons for skepticism about convoluted Risk Parity portfolios. He thinks that Risk Parity portfolio weights are arbitrary and unintuitive, and 85% of users have no clue where they came from. Side note: I would set that percentage to 100% and suspect that not even the inventors of Risk Parity know where those came from; see my recent post on the Risk Parody strategy.

In any case, what were the return and risk statistics of the seven portfolios? I plot the average real (i.e., CPI-adjusted) compound return and risk over the 100-year span from 1926 to 2026 below. Moving from 75/25 to the basic 55/40/5 certainly boosts your CAGR, except for the non-US fund. Not a huge surprise, as non-US stocks have modestly underperformed since 1970. But intriguingly, the most diversified, lowest-risk portfolio would have been the basic 55/40/5 portfolio. We added the exotic ETFs not to lower risk, but to increase average returns.

I will also simulate two retirement scenarios:

1. The standard traditional retirement case: a 30-year horizon and zero final asset target, i.e., our retirees are fine with depleting their assets. This is the old-fashioned Bengen and Trinity Study exercise.
2. The FIRE scenario with a 50-year horizon, plus a final asset target of 25% of the initial portfolio (CPI-adjusted). This is the assumption in my own retirement planning right now.

Both scenarios assume a flat spending profile and no additional cash flows. Of course, every retiree I know, myself included, will have different and additional assumptions: some with a spending smile, some with increasing spending, some with positive supplemental cash flows, and some with negative supplemental cash flows. But any changes in these assumptions will affect the seven portfolios roughly in the same direction and likely make little quantitative difference in their relative performance. Hence, for today’s exercise, I like to keep the analysis as generic as possible.

Let’s look at the safe withdrawal rates for the seven portfolios in the two different scenarios; please see the chart below. The good news is that Bengen’s new portfolio generates better outcomes in both retirement scenarios. But let’s look at the evolution of SWRs step by step. First, shifting from 75/25 to the basic 55/40/5 has very little impact on the failsafe. But adding the four exotic ETFs is the part that’s really crucial to the higher Safe Withdrawal Rate.

However, notice that my simulated results for the 30-year simulation horizon are noticeably worse than Bengen’s SAFEMAX estimates. For example, Bengen claims that before introducing the exotic styles, he already achieved a 4.15% failsafe rate. My basic 55/40/5 only has a 3.82% safe rate. And his 4.70% rate is also much higher than my 4.23%. How can we explain this 33- to 47-basis-point difference in failsafe withdrawal rates? To be honest, I can’t tell for sure, but there are several potential reasons:

• Different frequency: I check the performance of all monthly retirement cohorts, starting in 1926 (and even 1900 or even 1871 if needed). If you use a lower frequency, i.e., only quarterly or even annual retirement start dates, you might “jump over” the true historical worst-case scenarios like August 31, 1929, or November 30, 1968. In his interviews, he’s consistently stated that he simulated 400 retirement cohorts. That’s peculiar. At monthly frequency, that’s only 33.3 years. Why would you start in January 1929 and end your simulations in April 1959? You could have simulated many more retirement cohorts. In fact, his purported worst-case retirement cohort, October 1968, isn’t even in that sample of 400. So, he must have used a less-than-monthly frequency. Likely quarterly, where he would have certainly missed some of the historical worst-case retirement cohorts. Of course, then you can’t simulate 400 quarterly retirement cohorts because that would go up to 2026. Maybe he misspoke and meant he simulated 400 retirement return quarters with his roughly 100 years of return data, but only 280 of those quarters are actual retirement-starting cohorts with 30-year (=120-quarter) horizons. This alone would explain about 10-20 basis points of SWR difference.
• Different return data series: I’m completely transparent with my simulations, return series, and methodology. I post the simulation results in this Google Sheet for anyone to check (read-only, as usual, so make your own copy if needed). I have no idea what series Bill Bengen used.
• Ignoring expense ratios and other “drag” factors: It would not be proper to just simulate raw index returns. You need to account for expense ratios, which can be substantial for exotic ETFs. For example, the IWC has an expense ratio of 0.60%. Other hidden costs like mis-tracking and bid/ask spreads might also be at work. This oversight can easily explain 15 basis points in your safe withdrawal rate.
• Sloppiness and mistakes: Again, I post my methods very transparently. If you find a bug, please let me know. I have no way of checking Bengen’s results because his research is a complete black box.
• Confirmation Bias: If you’re writing a new book with a new shtick about how to raise your safe withdrawal rate, then, well, biases could creep into your simulations. You will face multiple design choices along the way; intentionally or unintentionally, you might always opt for the assumptions that reach a higher final goal, i.e., better safe withdrawal rates.

The answer is likely a combination of all of the above, which can easily generate a few dozen basis points in SWR difference.

What were the historical failure rates of Bengen’s 4.7% and 5.5% Withdrawal Rates?

Let’s start with the 30-year horizon. I plot the failure probabilities of the 7-portfolio sequence in the chart and table below. Let’s start with the unconditional failure probabilities. First, I want to concede again that the Bengen portfolio over 30 years would have given you an impressively low unconditional failure rate of only 2%. So the discrepancy between Bengen’s purported 4.7% and my 4.23% failsafe translates into only a tiny failure probability at a 4.7% withdrawal rate. Notice that at a 5.5% withdrawal rate, you’re already at an unconditional 20% failure rate. The path along the transition is also quite intuitive. Over 30 years, the impact of moving from 75/25 to 55/40/5 is relatively small, while over 50 years the lower equity allocation is clearly detrimental, raising your failure probability from 26.2% to 34.2%. Adding the exotic asset classes generally lowers the failure probabilities, with one exception: over 50 years there would have been a slight increase in the last step, i.e., adding international stocks.

But the results look much less appetizing when you consider elevated equity multiples. Conditional on a CAPE above 20, the 4.7% withdrawal rate had a higher, but likely still acceptable failure probability of 6.9% over 30 years, while the 5.5% withdrawal rate would have failed in almost half the historical cohorts (47.5%). Side note: The Shiller CAPE is now at about 40.6, and even the ERN-adjusted CAPE stood at 36.3 as of 9/25/2026, so using the unconditional failure probabilities will vastly underestimate the risk you’re taking!

For early retirees with a longer horizon and a sizable bequest target (25% of the initial portfolio), the new Bengen portfolio combined with 4.7% and 5.5% withdrawal rates would have had substantial failure probabilities, even in the best possible case where the Fama-French factors were fully intact; please see the chart below. In fact, all failure probabilities of the Bengen portfolio are worse than those of the 75/25 portfolio. Over the longer horizon, you also raise your failure probability in all four cases going from 75/25 to the third portfolio (55/40/5, but before adding the exotic equity flavors). That’s an intriguing result, because even when equities are expensive, it would still be unhelpful to pick a permanently lower equity share (though a glidepath might help, see Parts 19 and 20). Even the unconditional failure probabilities of the Bengen portfolios, 32.3% for the 4.7% rate and 58.1% for the 5.5% initial withdrawal rate, are unacceptable. Conditioned on today’s CAPE ratio, your failure probabilities rise to 78.0% and 98.7%. Anybody who extrapolates the 4.7% and 5.5% to early retirees in today’s market environment is entering financial malpractice and clownshow territory!

How important is “Small-Cap Alpha” to Bill Bengen’s New Portfolio?

If you pin your retirement hopes on Bill Bengen’s new portfolio allocation, look at the cumulative return of the Fama-French SMB factor (Small Minus Big, a.k.a. small-cap alpha). Between 1926 and 1981, the SMB factor added almost 3.6% annually. However, after 1981, when Rolf Banz published his seminal paper “The Relationship between Return and Market Value of Common Stocks” in the Journal of Financial Economics, the SMB factor has essentially just moved sideways, albeit with significant volatility. Critics of the SMB alpha story have conjectured that the excess return (alpha) of this Fama-French style factor has been arbitraged away after it became common knowledge in the finance community. So, while the SMB premium may have existed for historical retirement cohorts, we may not want to extrapolate this alpha source into the future. In other words, for your retirement portfolio, you have to live with today’s SMB alphas, because you can’t teleport past SMB alphas from the 1930s into the 2020s.

Additionally, even historical retirement cohorts might not have been able to harvest the full SMB premium, as several researchers have pointed out: transaction costs would have eaten up much (and potentially all) of the excess return (Joel L. Horowitz, Tim Loughran, N.E. Savin: The disappearing size effect, Research in Economics, Volume 54, Issue 1, 2000, Pages 83-100). Some researchers have pointed out methodological flaws that call into question how reliably one could have harvested SMB alpha: the non-synchronous trading issue of thinly traded stocks comes to mind, i.e., some of the quotes used in the calculation were stale, and you could not have traded at those prices. The so-called Bid-Ask bounce messes up return calculations (Marshall E. Blume and Robert F. Stambaugh: Biases in computed returns: An application to the size effect, Journal of Financial Economics, Volume 12, Issue 3, 1983, Pages 387-404).

My personal view: We should ignore the small-cap alpha for the following reasons: Everybody knows about the SMB factor today, and it’s easy to verify, so it’s likely arbitraged away. I’m not saying that SMB will underperform consistently, but there is no reliable outperformance either, and certainly no 3.6% p.a.! And even the historical outperformance is slightly suspect.

Which begs the question: what if the small-cap alpha we observed from about 1926 to 1981 doesn’t repeat itself and just remains flat going forward, as it did during the last 45 years (and potentially even before 1981 due to transaction costs and faulty methodology)? How do we even simulate this scenario? Well, avid ERN blog readers know the answer, because in Part 62 of the series I introduced this feature in my retirement simulation toolkit. Instead of using raw Fama-French factor returns, I propose using Hodrick-Prescott (HP)-filtered return series that maintain volatility, correlations, and business-cycle correlations but remove the (time-varying) mean excess return. This would help us gauge the performance of the exotic ETFs with less rosy assumptions.

Let’s look at the returns stats below (all are CPI-adjusted). Not surprisingly, your portfolios’ risk doesn’t really change. Going from 75/25 to 55/40/5 reduces both risk and CAGR. But CAGR changes very little once you add the exotic ETFs: you raise risk, but CAGR ends up at 5.18%, only marginally higher than the 55/40/4’s 5.13%. So, all the extra returns go out the window.

Let’s repeat the same retirement simulations, but I set the SMB and HML Fama-French factors to their HP-filtered return series with zero additional alpha. Adding the exotic equity ETFs adds no further gain in the SWR. Quite the opposite, over 50 years, the Micro-Cap fund would have dragged the SWR to only 3.12%. Over the shorter horizon, Micro-Cap stocks were also very detrimental. International stocks lifted the SWR again, but only to 3.82%, right where you started and the same as the 75/25 and 55/40/5 portfolios.

Next, let’s look at the failure probability charts, first for the 30-year retirement. Not much action here. The failure probabilities of the Bengen portfolio are very similar to those of the 75/25. Because the 75/25, 60/40, and 55/40/5 portfolios don’t use any Fama-French factors, the chart/table above doesn’t differ, of course. Adding the various exotic equity ETFs has a relatively small impact and leaves the failure probabilities uncomfortably high. Maybe the unconditional failure rates of the 4.7% rate still look OK at 11.0%, but conditional on an elevated CAPE ratio, you had 29.2% and 55.0% failure rates in the historical cohorts with 4.7% and 5.5% withdrawal rates, respectively. Not a pretty picture.

Next, let’s study a FIRE retiree’s 50-year horizon. The six-step transition from 75/25 to the Bengen portfolio makes the already unacceptable failure probabilities even worse. Much worse. Conditional on a high CAPE ratio, the Bengen portfolio would have failed in 87.8% of cohorts when using a 4.7% initial withdrawal rate. And a cool, impressive 100% of the cohorts would have failed when using 5.5% initially. The one (slightly) redeeming feature of the Bengen portfolio in this chart was that the exotic ETFs weren’t the main culprit in sabotaging your retirement success. It was mostly the low equity portion, i.e., moving from 75/25 to 55/40/4. Adding the exotic equity ETFs after that doesn’t change the results very much, but only because the 55/40/4 results were already totally atrocious and the new Bengen ETF recommendations couldn’t make the outcome much worse.

How important is the 1926 start time to the Bengen Portfolio?

Sure, we have Fama-French data only starting in July 1926. But that doesn’t stop me from simulating the pre-1926 retirement cohorts. I have all the returns except the Fama-French SMB and HML factors, which are only alpha contributors, so we can set them to zero and use the other return streams as usual, implicitly assuming that due to data availability issues nobody could have implemented the small-cap and micro-cap ETFs pre-1926 and everyone just resorted to using the large-cap blend index. And, obviously, a retirement cohort in 1907 would have still had exposure to ten years of small-cap returns toward the end of their horizon. The 50-year window would have covered a full 30 years of returns with the benefit of small-cap alpha, so cutting off the retirement simulations in 1926 seems arbitrary. We should at least include the retirement cohorts going back to 1900. The 1901 and 1907 bear markets were banking crises not much different from the Global Financial Crisis, so they’re relevant events we should potentially include in our analysis.

Then, how do the safe withdrawal rates look if we include the pre-1926 data? Here are the failsafe rates for 30 and 50 years when starting the retirement cohorts in 1900. I also included the failsafe for the Post-1926 simulations for comparison. The picture is disappointing for the Bengen portfolio. Over 30 years, adding the Bengen portfolio components one at a time noticeably deteriorates performance. While the first few steps, i.e., lowering the equity portion and adding mid-cap and small-cap stocks, only lowered the SWR from 3.82 to 3.77, the micro-cap and international stocks allocation hurt you more and dragged the SWR down to 3.68%. You get a very different picture in the 50-year retirement. Here, the biggest drag on retirement performance came from the shift to 60/40, and beyond that, the impact on safe withdrawal rates is relatively small.

Summary so far: the Bengen methodology only works under very narrow assumptions. Relax any of those assumptions and the whole thing falls apart; you’re then back to the safe withdrawal rates we’ve known since, well, the younger version of Bill Bengen, i.e., about 4% for 30 years and 3.25% for a 50-year retirement.

If you like the Bengen small-cap stock portfolio, though, here’s a better way…

If you want to argue that small-cap stocks will produce reliable outperformance again going forward, one can certainly entertain that idea. In my opinion, the Fama-French SMB risk-factor alpha is arbitraged away now that every stock picker is aware of the effect. The same is true for HML (High-Minus-Low Book Value, aka the value factor). But who am I to predict future stock returns? If you believe that there is again significant outperformance in small-cap stocks, sure, go ahead. But even then, I wouldn’t follow Bill Bengen’s recommendation either.

Avid readers of my blog will know where I am going with this: if you believe those exotic equity flavors are so great, why shortchange yourself and invest only 55% in that new and improved portfolio? So, rather than reducing the equity share down to 55%, just replace a portion of the 75% equities with that amazing new investing style. Hence, I propose changing the order in the attribution exercise, i.e., shift into the four equity styles first, before reducing the stock percentage.

1. The 75/25.
2. Shift 15 percentage points from the S&P 500 to mid-cap stocks.
3. Shift 15 percentage points from the S&P 500 to small-cap stocks.
4. Shift 15 percentage points from the S&P 500 to micro-cap stocks.
5. Shift 15 percentage points from the S&P 500 to international stocks.
6. Shift 15 percentage points from equities to bonds. (3 percentage points from each of the five equity styles)
7. Bengen’s Portfolio: Shift the remaining 5 percentage points from equities to T-bills (1 percentage point from each of the five equity styles)

Let’s now check how the sequence of portfolios would have performed in the historical cohorts. And again, this is under the (optimistic) assumption of investors being able to fully harvest the entire Fama-French SMB and HML premia. Not surprisingly, the first four steps up to portfolio 5 enhance my retirement safety. But moving out of equities and down to the Bengen 55% share will take away a good chunk of that advantage again for the 50-year retirement, while a lower fixed-income allocation has a negligible effect on the shorter horizon with asset depletion. In a 30-year retirement with asset depletion, you can get away with fewer stocks and more bonds. But over longer horizons, you need the return power of equities to get the maximum safe withdrawal rate.

The results also fit into a greater pattern I have noticed in my safe withdrawal rate series: If you have confidence that a new investing flavor will enhance your portfolio, it’s best to maximize this impact and not water it down with other junk, as I’ve shown over the years:

• If you believe gold is a good diversifier, just add it to the 75/25 portfolio. Stay away from junk portfolios like the All-Weather and other Risk Parity flavors that may have gold, but you’d throw out the baby with the bathwater and also reduce the equity portion too much, which will likely jeopardize a long retirement. See Part 34 of my series.
• If you believe value stocks and/or small-cap value stocks will outperform again, add them to a 75/25 portfolio. Stay away from junk portfolios like the Risk Parity + value stocks portfolios peddled by clowns on the internet. See Part 64 of my series.
• If you believe that momentum strategies are a thing, just add a massive portion of momentum to your portfolio. I showed that 50% equities plus 50% in the ERN momentum strategy would have performed quite nicely in historical simulations. See Part 63. Mixing in just 10% momentum plus some other junk (as recommended in some asinine Risk Parity portfolios) will not help you.

Conclusion

Bill Bengen is a legend. He wrote the first serious research paper on sustainable withdrawal rates in the 1990s, ahead of the Trinity Study. He’s a nice guy; I met him in 2024 at the Bogelheads conference in Minneapolis and served on a panel with him and Christine Benz. Respect where respect is due. So I feel terrible that Bill Bengen’s most recent research insights aren’t that impressive, to put it diplomatically.

In the early 2020s, he made illogical and poorly reasoned claims about the importance of inflation on safe withdrawal rates. I thoroughly debunked that in parts 41 and 51; inflation, especially the 12-month trailing inflation rate – the one Bengen focuses on – has zero significance to the safe withdrawal rate in my statistical/economitric models. His results back then relied on heavy data massaging, such as ignoring CAPE ratios above 22, which really throws out the baby with the bathwater.

Now he’s touring the financial podcast circuit touting his new portfolio with a 4.7% and even 5.5% withdrawal rate. I’m unconvinced again. Even in the best possible case (short horizon, asset depletion, and the full Fama-French alphas), it seems quite risky to withdraw 4.7%, and it’s outright irresponsible to recommend 5.5% without mentioning the roughly 50% historical failure rate if you condition on the elevated (i.e., above 20) CAPE Ratio. Over a 50-year horizon, even with the Fama-French raw data, the historical failure rates of a 4.7% and 5.5% withdrawal rate were 58.1% and 98.7%, respectively, conditioned on an elevated CAPE ratio. All members of the FIRE community should avoid overly optimistic withdrawal rates, even if you have deluded yourself into believing the small-cap and value flavors will miraculously pick up again.

The wheels completely come off the Bengen portfolio if the rosy assumptions about small-cap premiums don’t hold and you believe that smart and educated investors operating in (mostly) efficient markets have now arbitraged away the SMB factor alpha; you’re basically back to a roughly 3.8-4.0% safe withdrawal rate as in the traditional 30-year Trinity Study or about 3.25% over the longer 50-year horizon. I don’t buy the hype about higher sustainable rates, and I recommend my readers avoid this flawed approach to retirement planning. You can certainly try the Bengen allocation, but my expected SWR from that portfolio is roughly the same as with a standard 55/40/5 without the exotic junk. So even if you use the Bengen portfolio, keep the SWR the same as before, because whether Small-Cap stocks keep underperforming and start outperforming again is really a crapshoot.

Of course, if some stock market fairy whispered in my ear today that the small-cap premium will return to its old glory, and we can extrapolate those impressive 1926-1981 returns to 2026 to 2081, I would like to try a Bengen-style portfolio. But even then, I would substantially deviate from Bengen’s recommendations: First, I’d stay away from the 55/40/5 idea and instead max out that stock-picking alpha, keeping the equity portion at 75%. Second, I’d adjust my safe withdrawal rate to about 3.8% as a failsafe or about 4.5% with some flexibility in my personal 50-year retirement scenario. But that’s a whole percentage point lower, i.e., a 20% lower retirement budget than the overly optimistic 4.7% to 5.5% recommended by Bengen. I summarize my thinking in the decision tree below:

Most importantly, I wouldn’t call Bengen’s approach diversification. In portfolio management, we define diversification as spreading investments across a variety of assets, sectors, and geographies to reduce overall risk and volatility. True, moving a portion of your equities to non-US markets can likely help. But not the mid-, small-, and micro-cap ETFs. Adding Bengen’s exotic equity asset classes is the opposite of diversification; you move away from a broadly diversified index fund (e.g., VTI, ITOT, VOO, IVV, etc.) and overweight tiny pockets of the market you hope will outperform, increasing volatility in the process.

Bengen essentially used stock picking and called it diversification to make it more palatable to the gullible public. Attributing (or I should say misattributing) better safe-withdrawal performance to diversification certainly sounds better than “Hey, I went through the return history with a fine comb and picked the series that had stronger returns than the large-cap index!” I call a Spade a Spade; it’s not diversification, but stock picking, and actually the worst type of stock picking, i.e., retroactively looking for the best portfolio. I’m amazed that there isn’t more pushback against this shell game. So, I urge my readers to use caution before raising your safe withdrawal rate based on stock picking. It may not work as well in the future!

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 + ERN edits

Technical Appendix: Replicating the new Bengen Portfolio in Historical Simulations

The betas from SPX-TR to Gold plus Cash are constrained to sum up to 1.00. SMB and HML are alpha factors and are unconstrained. I constrain the ERN momentum beta to be zero. Here are the detailed replication stats for the four ETFs:

Of course, I also studied the t-stats for those slopes (not displayed here for brevity), and all the relevant betas are indeed highly significantly different from zero. Likewise, all the F-tests reject the hypothesis that the regression betas are all zero.

We can also look at how well the replication compares to actual returns. First, a scatter plot of actual (x-axis) vs. replicated (y-axis). These all look excellent

I also plot the cumulative returns: All series are actually pretty adequately captured through the replication. In fact, we get slightly higher final values with the replicated returns because the factor model targets the arithmetic average return; with a slightly below 1.0 R-squared, the replication has marginally lower volatility, so the replication’s CAGR is slightly better than the original series’ CAGR. I could have adjusted the intercept to account for this well-known effect, but to give the Bengen portfolio the best possible chance to outperform, I kept the intercept as is.

For the purists, the same chart but with a y-axis in log scale:

Another note about the replication: I have international equity return data only starting in 1970. For the pre-1970 period, I use the S&P 500 return series again, not only because of the lack of monthly data but also because it would have been challenging for the average US investor to access non-US stock market indexes early in the sample. Besides, the returns during WW2 would have looked really awful in Europe and Japan, so I’m giving the Bengen portfolio again the best possible chance to outperform.

I can also verify with returns provided by Portfolio Visualizer how the Bengen portfolio would have performed going back a little further, i.e., 1972 for a subset of the indexes and since 1986 for the entire portfolio:

• Since 1972: I use 22% US Large-Cap, 11% each in Mid, Small, and Micro-Cap. No allocation to VEU/Non-US stocks, because the site doesn’t have available data for that index. Then 40% 10Y Treasuries and 5% Cash. The CAGR according to PortfolioVisualizer was 9.64%. My simulation during the same time span fetched 9.57%, or about 7 basis points less annualized. But the weighted expense ratio of the seven ETFs was 14.6 basis points, so my simulated returns were even better than expected.
• Since 1986: I use the exact weights as in Bengen’s portfolio: 11% each in US Large, Mid, Small, Micro-Cap and non-US equities, plus 40% 10Y Treasuries and 5% Cash. The CAGR between 1/1986 and 6/2026 was 8.75%, according to PortfolioVisualizer. My simulation over the same period returned 8.62%, about 13 basis points lower annualized. But the weighted expense ratio of the seven ETFs was 14.7 basis points, so my simulated returns were even slightly better than expected.