Factors (Fama-French, value, momentum) in the withdrawal phase

For thirty years, academic research has been cutting stock returns into "factors": systematic characteristics that have, historically, paid a premium above the market. The main ones are cheapness (value), small size (small cap), quality, and momentum. It is one of the most solid bodies of evidence in finance, and one of the most heavily marketed. Someone living off a portfolio has a narrower question to settle than a saver does. Do these factor tilts improve a decumulation plan, where the worst path is what counts and the average is not (Sequence of returns risk: the retiree's real enemy)?

The answer comes with conditions, and this page works through them. What a factor really is, and why the premium might still be there. What the numbers say once you subtract the erosion that follows publication. Then the retiree's own case file, where small-cap value plays the lead (it is the Golden Butterfly's pick, and the subject of ERN's Part 62, which cuts its claims down to size) and where value brings a valuable regime affinity. Then the real costs: the 2010s were a purgatory for value, and the gap to the index gets paid in patience. Then what you can actually buy, and how much of it, because here more than anywhere else a factor tilt is an optional refinement, worth taking only if you understand what you are buying.

The gap in annualized real return between US small-cap value and the S&P 500 total return, over rolling ten-year windows, each window dated by the month that closes it. The small value leg is Kenneth French's academic portfolio, which pays no fees and no brokerage: 1.0 point a year is taken off here, the gap measured between that portfolio and a real fund over their thirty-three overlapping years. Both series are deflated by US CPI and run from 1963 to 2026.

What a factor is: the academic base, in brief§

The history fits in four dates.

In 1964, Sharpe's CAPM recognized a single factor, the market. Everything else was supposed to be noise.

In 1992, Fama and French showed that two characteristics predict a stock's return. Size first: small caps beat large caps on average. Cheapness second (value): stocks that are cheap against their fundamentals beat expensive ones. That is the three-factor model, market, size and value.

In 1997, Carhart added a fourth, momentum: stocks that have gone up recently keep going for a while. It is the cross-sectional cousin of the trend following in Managed futures and trend following: the diversification that works in a crisis.

In 2015, Fama and French went to five. They added profitability, because at the same price profitable companies beat unprofitable ones. And investment, because frugal companies beat spendthrift ones. Practitioners file those two under one label, quality.

Those five are not alone. Research has published hundreds more, what people call the factor zoo. Most of them evaporate the moment another researcher looks for them in other countries or other periods. Only a handful survive replication: market, size, value, momentum, profitability and investment. That small robust core, and nothing else, is worth your attention.

Historically, these factors paid something on top of the market. That something has a name, the factor's premium (the value premium, the small cap premium, and so on). Two questions follow. Why did the premium exist? And why would it survive, now that research has made it public and everybody knows about it?

Two families of explanations, probably both true. The first is risk. Small caps and value stocks are more fragile in a recession. The premium is then the wage of that risk. On that story it does not go away, but it costs dearly, because you pay for it in the bad moments, the ones where you would rather not suffer.

The second explanation is investor behavior. People chase recent winners and neglect boring companies. The premium then comes out of that collective mistake. On that story it survives as long as few players go after it.

One fact is well measured. Once a factor is described in an academic paper, investors pile in and its premium shrinks. McLean and Pontiff put numbers on that erosion across 97 anomalies: 26% of the return lost as soon as the study period ends, 58% after publication. So the raw numbers get read with a haircut. The value premium was worth 3 to 5% a year in the studies (academic long-short). In practice, a long-only value ETF captures more like 1 to 2% of expected excess a year (Forward-looking expected returns (Morningstar, Vanguard, the investment banks)).

The retiree's case file: three exhibits§

Exhibit 1: Bengen and small caps. In his later work, Bengen showed that adding US small caps to the reference portfolio raised the SAFEMAX, from 4.15% in 1994 to 4.3% in 1997 with 30% small caps in the equity sleeve, then to 4.5% in 2006 with a wider palette of assets (Bengen, the Trinity study, and the birth of the safe withdrawal rate). His 2025 book goes to 4.7%. Stocks make up 55% of that portfolio, split evenly across five styles: US large, mid, small and micro caps, plus international stocks.

For years this was read as a diversification effect. ERN's Part 65 (2026) offers another reading, built on a factor-model reconstruction of the portfolio. With the historical premia left whole, Jeske gets 4.23% over 30 years, not 4.7%. With the premia set to zero he gets 3.82%, the rate of the same portfolio without its four added styles. On that reading the gain depends on the size premium, which paid close to 3.6% a year from 1926 to 1981 and next to nothing since.

Two reservations belong here. Bengen has not published his series, and the gap between 4.7% and 4.23% is unexplained. More important, a zero premium is an assumption, not a fact. The debate is open, and this book for its part works with premia that are marked down and uncertain (Where returns come from: risk premia). So the answer sits between the two bounds. One point needs no assumption at all, though. Funds that are highly correlated with the S&P 500 and more volatile than it diversify little, and from 2007 to 2026 the five-style portfolio was more volatile than its plain version, 8.8% against 7.9%. So this exhibit proves less than it seems to: small caps helped through their premium, and what they add from here is worth whatever that premium turns out to be worth.

Exhibit 2: ERN, Part 62, and small-cap value. Jeske devoted a whole part to this exact question, small-cap value (SCV) in decumulation. SCV is small caps that are cheap, the most studied crossing of two factors. What he found in the long data cools the enthusiasm, and you want to know it before you buy the thesis. With factor alpha neutralized, the average safe rate across vintages since 1926 goes from 3.39% to 3.27% when SCV goes into a 75/25. Grant SCV the forward premia the forecasters publish, 1.2 points for size and 0.8 for value, and it climbs back to 3.51%, 0.12 points better than the untilted portfolio. More to the point, the vintages with the worst sequence risk tend to do worse with SCV, and the 1930s start years give up 0.22 points.

The argument on the other side is real, and it is historical. The great valuation disasters are concentrated. The 2000 bubble was a bubble in large growth stocks, and 1966 was the Nifty Fifty. An SCV tilt moves you away from that point of concentration (Valuations, the CAPE, and what they say about your withdrawal rate), and the retiree of 2000 who held SCV had a decent decade while the S&P lived through its lost one. The flip side is part of the deal. From 1929 to 1932, and again in 2008, cheap small caps suffered more than the market, because that is when the factor pays for its recession risk. The honest verdict: SCV moves the lost decades around, it does not raise the safe rate.

Exhibit 3: value's regime affinity. This is the most modern exhibit. Historically, value is the style that holds up best in inflationary regimes. Its companies have cash flows that arrive soon and tangible assets, and their debt erodes with inflation. Growth stocks suffer instead, because their earnings are expected far out in the future and get discounted harder. In 2022, world value returned roughly 0% while growth lost 25%. For a portfolio whose worst enemy is the inflationary regime (Market regimes (growth × inflation, sticky bears) and why they matter), a value tilt is half a defensive building block, housed inside the equity sleeve. It costs nothing in expectation, and sometimes it even pays, where gold and linkers cost you carry (Defensive assets: the map and the roles).

⚠ The price of admission: the 2010s, and every decade like them

Every factor premium is paid for the same way: years, sometimes a whole decade, of doing worse than the index everybody else owns. The 2010s are the example to frame and hang on the wall. World value underperformed growth by about 5% a year for ten years, the worst episode in its history, worse even than 1999. The same purgatory shows up in the rolling gap for US small-cap value, which ran as much as 8.0 points a year behind the S&P 500 over the ten years ending in April 2020, its worst gap since 1963. Investors capitulated in droves, right before the turn of 2021 and 2022.

The parallel with the CTA winter (Managed futures and trend following: the diversification that works in a crisis) is exact, and so is the one with gold's purgatory from 1980 to 2000 (Gold in a retirement portfolio). Every real diversifier has its years of shame. That is how you recognize one: what never strays from the index adds nothing to the index.

The practical rule is the same everywhere. Take a tilt only if it is documented, written down in black and white, and sized so you can hold it for ten years with no proof. Otherwise take the broad index and the quiet life (The psychology of spending: why spending it is the hard part).

The menu has improved a lot over the past decade, and how deep it goes depends on where you buy your funds.

What exists and works. Small-cap value is bought through funds tracking a small value index, around 0.30% a year, with a serious methodology; in the United States the Avantis and Dimensional ranges are the usual reference, and elsewhere the choice is thinner. Broad value goes through the enhanced value funds; be wary of diluted "value" indexes, common in the cheap ranges. Momentum comes as world momentum funds, useful but heavy on turnover and on internal costs. Quality has funds of its own. And multifactor in a single fund exists too. It buys real convenience at the price of an opaque construction. Read the methodology before you buy that one.

The three checks. First, the fund's capacity: small-cap value is a niche market, and the large vehicles are the safer place to sit. Second, turnover, above all on momentum, the source of invisible internal costs. Third, the real exposure to the factor, because plenty of "value" funds carry no more than a homeopathic tilt. You check it by comparing their P/B and P/E with the parent index's. Where you hold them is a separate question, settled by the account rules and not by the factor (Building it with US-listed ETFs: the retiree's shopping list).

Testing a tilt in a simulator. A tilt gets judged by an A/B test, like any other building block. The usual test swaps 20 to 30% of the world equity sleeve for US or international small-cap value, then compares the two plans. That takes long data, and here is the good news of the subject. Factor funds are young, but the academic series are deep, and Kenneth French's library goes back to 1926 for the United States. The whole point of a tilt plays out in exactly those old vintages. So do not read the central case, where the tilt's expectation, already marked down, is nearly invisible. Look at the replayed start years (2000, SCV's field of glory, and 1929, its field of pain) and at the dispersion of the paths. The tilt widens the fan a little, because it adds idiosyncratic risk, while it moves the worst paths. That tradeoff is exactly what you have to judge (Reading a fan chart without fooling yourself).

Sizing it, and the honest verdict§

The size that practitioners use. A tilt of 10 to 30% of the equity sleeve, so 7 to 20% of the whole portfolio. Build it from one or two simple blocks: US small-cap value plus international small-cap value for the standard version, or a broad value tilt for the gentle one. Below 10% the effect is cosmetic. Above 30% the portfolio becomes a style bet, and its gap to the index will end up dominating the whole life of the plan.

The verdict, no dodging. Factors are second order in decumulation. They matter far less than audited spending, the initial withdrawal rate, the withdrawal rule and the allocation (Choosing your strategy: criteria, comparison, worked case). They weigh about as much as fine-tuning your spot on the allocation plateau (The stock and bond allocation in retirement). The case for is real but modest: diversifying which regimes hurt the equity sleeve moves the lost decades around (the 2000 exhibit and the value-inflation affinity show it), without raising the safe rate for all that. The case against is real too: marked-down and uncertain premia, long years of lagging the index, complexity, fees, and a menu that stays thin in most markets. This book's default position is simple. The tilt is legitimate and optional. A plan with no factors, well built otherwise, is missing nothing essential. A plan with a documented 20% in SCV, and held, gains a little regime robustness. And a plan that walks away from its factors in 2019 will have paid the premium for somebody else. Pick your side once and for all, in writing (Building your plan step by step).

🧮 A tilt judged on the evidence

The starting plan: $1.5M, $51,000 a year, 70% world stocks and 30% defensive. The variant tests an equity sleeve of 50% world, 10% US small-cap value and 10% international small-cap value. The verdicts read line by line. Central case: 3.9% to 3.8%, so the tilt's expectation, marked down, is barely visible. The 2000 vintage replayed: clearly softer, because SCV came through the broad market's lost decade. The 1929 vintage: slightly worse. Sequence stress and the broad sample: 0.3 to 0.5 points off the failure probability, with value's inflation affinity earning its keep in the 1970s blocks. Dispersion of the fan: 5% wider. One quantitative couple takes the tilt, with a written clause. "The tilt is judged over rolling ten-year windows against its thesis, regimes, never against this year's Nasdaq." Another couple, wanting peace and quiet, says no, and that is just as defensible. The robustness gained is worth about 0.3 points, and the certainty of tracking the broad index is worth every bit as much.

The essentials§

The robust core is six premia: market, size, value, momentum, profitability and investment. They are real, but marked down once published (roughly 1 to 2% a year for a long-only index tilt), and they come from a mix of risk (paid in the bad moments) and behavior.
The retiree's file has three exhibits. Bengen's small caps raise the SAFEMAX from 4.15% to 4.3%, then to 4.5% and 4.7% with a wider palette, a gain that ERN's Part 65 ties to the pre-1981 size premium more than to diversification. ERN's Part 62 cools SCV down: it comes through the broad market's bubble vintages (2000), suffers more in a pure recession (1929), and does not raise the average safe rate. The value-inflation affinity buys half a defensive building block, free, inside the equity sleeve.
The price of admission is the gap to the index over long stretches. Value lost 5% a year to growth from 2010 to 2020. Every real diversifier has its years of shame, hence the binary rule: a tilt written down and held, or no tilt at all.
Buying it: small value funds for SCV, enhanced value funds for broad value, with checks on capacity, turnover and real factor exposure. Size it between 10 and 30% of the equity sleeve.
Verdict: legitimate and optional. The tilt is second order, behind spending, the rate, the rule and the allocation. Judge it in a simulator on replayed vintages and on the tails of the distribution, never on the central case.

Going further§

Fama & French, "The Cross-Section of Expected Stock Returns" (1992) and "A Five-Factor Asset Pricing Model" (2015): the sources.
Larry Swedroe, Your Complete Guide to Factor-Based Investing: the honest practitioner's synthesis (a credible factor is persistent, pervasive, robust, investable and intuitive).
Early Retirement Now, Parts 62 (small-cap value in retirement) and 65 (Bengen's five-style portfolio) (ERN's Safe Withdrawal Rate series: a reader's guide); McLean & Pontiff (2016) on the decay of premia after publication.
In this book: The stock and bond allocation in retirement (what goes inside the equity sleeve), Market regimes (growth × inflation, sticky bears) and why they matter (the regime affinity), Building it with US-listed ETFs: the retiree's shopping list (the vehicles), Managed futures and trend following: the diversification that works in a crisis (time-series momentum, its more powerful cousin), Where returns come from: risk premia (the general frame for premia and what becomes of them after publication).