Momentum & factors

Value's worst decade in a century: the value premium, decade by decade, 1927–2026

Momentum & factors · Published 2026-11-22

Is the value premium dead? The question has hung over factor investing since the late 2010s, when cheap stocks lagged expensive ones year after year. We went back to the public source most of that debate relies on, the Kenneth R. French Data Library, and recomputed the classic factors decade by decade, from January 1927 to September 2026. The answer is less dramatic than either side of the argument: value had the worst decade in its recorded history, it has partly recovered, and long stretches without a premium are part of its normal behaviour.

Educational analysis of historical data; not a forecast or investment advice. Updated 22 November 2026 with monthly data to September 2026.

Key figures

What the factors measure

The data come from the Kenneth R. French Data Library at Dartmouth's Tuck School, which publishes the factor returns used in thousands of academic papers. Each factor is the monthly return of a long-short portfolio of US stocks listed on the NYSE, AMEX and Nasdaq. According to the library's description of the Fama/French factors, HML is the average return of two value portfolios minus two growth portfolios, sorted on book-to-market; SMB (small minus big) does the same for size; Mkt-RF is the value-weighted market return minus the one-month Treasury bill. The momentum factor buys stocks with high returns over the prior twelve months, skipping the most recent one, and sells those with low returns. The construction goes back to Fama and French's 1993 paper in the Journal of Financial Economics; our article on the five-factor model explains how the framework grew from there.

Decade by decade

Annualised compound return of each factor by calendar decade, % (and HML's deepest drawdown inside the decade)
PeriodMonthsMarket − T-billSize (SMB)Value (HML)MomentumHML max drawdown
1927–293612.0−12.52.325.0−17.3
1930s120−0.88.81.0−6.9−43.7
1940s1209.04.59.66.4−11.6
1950s12016.2−0.93.310.9−12.3
1960s1204.24.53.211.1−13.5
1970s120−0.32.87.79.3−17.7
1980s1207.3−0.45.68.6−22.8
1990s12012.5−2.1−0.413.7−32.5
2000s120−3.14.27.4−2.0−28.3
2010s12013.0−0.5−2.62.6−32.6
2020–268111.9−2.80.72.5−33.7

1927–29 covers 36 months; 2020–26 covers January 2020 to September 2026 (81 months). Source: Kenneth R. French Data Library, our calculations.

Value (HML) and momentum (Mom) premiums by decade, annualised, 1927–2026Bar chart. HML was negative only in the 1990s, 2010s and close to zero in the 2020s; momentum was negative in the 1930s and 2000s.-10%0%10%20%30%1927–29 HML: +2.3%1927–29 Mom: +25.0%1927–291930s HML: +1.0%1930s Mom: -6.9%1930s1940s HML: +9.6%1940s Mom: +6.4%1940s1950s HML: +3.3%1950s Mom: +10.9%1950s1960s HML: +3.2%1960s Mom: +11.1%1960s1970s HML: +7.7%1970s Mom: +9.3%1970s1980s HML: +5.6%1980s Mom: +8.6%1980s1990s HML: -0.4%1990s Mom: +13.7%1990s2000s HML: +7.4%2000s Mom: -2.0%2000s2010s HML: -2.6%2010s Mom: +2.6%2010s2020–26 HML: +0.7%2020–26 Mom: +2.5%2020–26Value, HMLMomentum, Momcompound annual return, %
Value and momentum premiums by decade, annualised compound return. Source: Kenneth R. French Data Library; calculations by Out of Sample.

Three things stand out. First, value paid in most decades but not all: it was slightly negative in the 1990s, clearly negative in the 2010s and roughly flat in the 2020s so far, because the gains since 2021 came after a terrible 2020. Second, the factors rarely fail together. Momentum's bad decades were the 1930s and the 2000s, exactly when value did well; that negative relationship is the reason many systematic funds hold both, a point made in Asness, Moskowitz and Pedersen's work on value and momentum and in our piece on combining diversified alpha. Third, size has been the weakest of the four over the whole sample, and SMB was negative in both the 2010s and the 2020s so far.

The longest drawdown in value's history

Value factor (HML): distance below its previous peak, monthly, 1927–2026Area chart of HML drawdowns. The deepest, about minus 58 percent, ran from December 2006 to September 2020; in September 2026 HML was still about 33 percent below its 2006 peak.-60%-40%-20%0%1930194019501960197019801990200020102020
HML's distance below its previous peak, monthly, January 1927 to September 2026. Source: Kenneth R. French Data Library; calculations by Out of Sample.

Plotting how far HML sat below its previous high shows why the 2010s felt different. Earlier drawdowns were deep, notably in the 1930s and around the dot-com bubble (−39.8% from May 1998 to February 2000), but every one of more than 20% had been recovered within six years at most. The one that began in December 2006 lasted almost fourteen years to its trough and has not closed since. Split the sample at the start of 2007 and the contrast is stark: HML compounded at 5.0% a year from 1927 to 2006 and at −2.0% a year from 2007 to September 2026.

The debate this produced was public. In 2020 Fama and French themselves reported, in a paper later published in the Review of Asset Pricing Studies, that value premiums were "on average much lower" in the second half of the July 1963–June 2019 period, while noting that volatility made it impossible to reject equal expected premiums in both halves. Chicago Booth Review summarised it under the headline "The Value-Stock Premium Is Shrinking". On the other side, AQR researchers concluded in the Journal of Portfolio Management that the suggestion value investing is dead was "premature".

The comeback, and its limits

Then the factor turned. HML returned 25.6% in 2021 and 25.7% in 2022 in the library's annual figures, as rising interest rates hit expensive growth shares. The Irish Times put the 2022 move in perspective: "Value stocks didn't actually make money in 2022 – it's more that growth stocks suffered an annus horribilis." That is exactly what a long-short factor captures: HML can gain in a falling market if growth falls further. The run did not last in a straight line, with −14.0% in 2023 and −8.4% in 2024, before +8.7% in 2025.

Our reading of the numbers is deliberately modest. A decade of underperformance is not proof that a premium has gone, and a few good years are not proof that it is back. With annual volatility of around 12% and an average premium of around 4%, ten years of data cannot pin down the true premium tightly; the same arithmetic underlies our deflated Sharpe ratio calculator. What the record does show is that every one of these factors has gone through ten-year stretches without paying, and that investors who hold one should expect to live through such a stretch. Momentum, the subject of our article on momentum crashes, is itself still about 31% below the peak it reached in November 2008, after losing 83.5% in 2009.

Download the data

factor-returns-by-decade-1927-2026.csv (2 KB): for each period, the annualised compound return, annualised volatility and maximum drawdown of the four factors, computed by us. The file does not contain the library's monthly series, which remain under Fama and French's copyright; download them from the source. Free reuse of our table, charts and CSV, with a link to this page.

Methodology

Sources