Did the famous factors pay, and when did they stop?

The full sample

Each factor over every month it exists. The start dates differ: RMW and CMA begin in Jul 1963, momentum in Jan 1927, the rest in Jul 1926. Growth is the geometric product of the monthly returns; the mean and its t-statistic are arithmetic, because that is the quantity the factor literature tests.

Source: Kenneth R. French, Data Library (Tuck School of Business, Dartmouth) Mkt-RF is the market minus the risk-free rate, not a long-short spread. Compounding it gives the growth of the excess-return series itself, which is not the same as the market's total return minus cash. Methodology

The last 240 months

The same six factors over the 240 months from Jun 2006 to May 2026. 1 of the 5 long-short factors compounded above 1% a year: RMW, at 2.70%. 3 of the 6 factors ended the window with a lower index than they started it: SMB 0.80x, HML 0.70x, CMA 0.98x. The market’s excess-return series (Mkt-RF) compounded at 9.72% a year.

Source: Kenneth R. French, Data Library (Tuck School of Business, Dartmouth) The window is the trailing 240 months ending at the last month in the data, derived at build time, not a fixed date. Methodology

Growth of $1, Jul 1963 to May 2026

All six factors from the first month all six exist, so the comparison is like for like. Log axis. Compounding a long-short spread as though $1 were invested in it is a convention, not a portfolio: the spread is zero-cost by construction, so the level of a long-short index has no dollar interpretation. Read the levels as an index, not as money.

Source: Kenneth R. French, Data Library (Tuck School of Business, Dartmouth) Methodology

Ten years is not long enough

Trailing ten-year annualised return of each long-short factor, complete 120-month windows only. HML has been below zero in 153 of the 1,080 complete ten-year windows that end between Jun 1936 and May 2026, SMB in 379 of 1,080. In the window ending May 2026, 3 of the 6 factors are negative: SMB -2.3%, HML -1.1%, CMA -1.2%.

Source: Kenneth R. French, Data Library (Tuck School of Business, Dartmouth) Methodology

Before and after Dec 2006

The split is not chosen: it is the month the HML index peaked, computed from the data. The left column covers the 522 months from Jul 1963, the first month RMW and CMA exist, to Dec 2006. The right column covers the 233 months from Jan 2007 to May 2026. Cutting the left column back to the RMW and CMA start is what puts all 6 factors in both columns. 4 of the 5 long-short factors compounded more slowly after the split than before it, while the market compounded faster: 4.73% a year before, 9.55% after.

Source: Kenneth R. French, Data Library (Tuck School of Business, Dartmouth) Both periods are annualised compound returns; t is the t-statistic of the arithmetic monthly mean in that period. Pre-split 522 months, post-split 233 months. Methodology

Where each index stands against its own record

Source: Kenneth R. French, Data Library (Tuck School of Business, Dartmouth) Peak and drawdown are measured on each factor's own cumulative index over its own full sample. Methodology

By decade

Annualised compound return within each calendar decade, over the months that decade contains. The first and last rows are partial: the data begin in Jul 1926 and end in May 2026. A blank cell means the factor did not exist yet.

Source: Kenneth R. French, Data Library (Tuck School of Business, Dartmouth) Momentum contributes 36 of the 42 months in the 1920s row, and RMW and CMA contribute 78 of the 120 in the 1960s row, so those cells cover less of the decade than the others. Methodology

The market’s premium came in the calm months

Mean monthly return of each factor, splitting the 437 months from Jan 1990 to May 2026 at the median of the month's average VIX close. The market’s excess return averages 1.40% a month in the calm half and 0.10% in the volatile half, a gap of 1.3 points. The widest gap among the 5 long-short factors is CMA at 0.6 points, smaller than the market’s. 4 of the 5 average more in the volatile half than in the calm one: SMB, RMW, CMA, Mom. This is a split of realised returns, not a risk model, and it says nothing about causation.

Source: Kenneth R. French, Data Library (Tuck School of Business, Dartmouth) | Cboe, VIX historical index values Monthly VIX is the mean of that month's daily closes. Factor cells are arithmetic means of monthly percent returns, not compounded. Methodology

The 49 industry portfolios

Value-weighted monthly returns for French’s 49 industry definitions. 40 of the 49 carry all 1,199 months of the panel, which runs from Jul 1926; the other 9 carry fewer, 9 of them because they begin after Jul 1926. Interior gaps, 1 of the 49: Rubber and Plastic Products, 12 months. 49 of the 49 compounded positively over their own history, and 19 of 49 beat the value-weighted market over exactly the months they existed. Over the last 240 months, 14 of 49 beat the market and 1 lost money: Coal -5.9% a year.

Source: Kenneth R. French, Data Library (Tuck School of Business, Dartmouth) Compound annual return, % a year. Blue bars beat their own market benchmark (the white tick), grey bars did not. Each industry is compounded and benchmarked over exactly its own months. Methodology

Source: Kenneth R. French, Data Library (Tuck School of Business, Dartmouth) Market is Mkt-RF + RF, the value-weighted total return, compounded over exactly that industry's months. Vol is the annualised standard deviation of monthly returns. Methodology

What this data cannot tell you

• These returns are not achievable. The long-short factors are paper portfolios: no transaction costs, no bid-ask spread, no shorting fee, no borrow constraint, no capacity limit, no taxes, rebalanced without slippage. The 3.56% a year on HML is an upper bound on what any investor could have earned holding it. This data cannot say how far below that bound the achievable number sat.
• US only.Every return series on this page comes from French’s four US files: the 3-factor, 5-factor, momentum and 49-industry monthly returns. Nothing here says whether value, size or momentum paid outside the United States.
• The published factor literature was built by searching this data. Fama and French published the three-factor model in 1993 and the five-factor model in 2015, both from samples inside these 1,199 months. A t-statistic of 3.43 on a series the profession has been mining for decades does not carry the evidential weight it would carry out of sample. Read the full-sample numbers as a description of what happened, not as an estimate of what pays.
• The decline is measured, not explained. HML has compounded at -2.0% a year for 233 months. This data cannot say whether the premium was arbitraged away after publication, whether book value stopped measuring what it once measured as intangibles grew, or whether 233 months is simply a long unlucky run in a series whose annualised volatility over those months is 11.5%.
• A factor index is not a fund. The cumulative levels compound a zero-cost spread as though it were a funded position. They are an index, and the level of a long-short index has no dollar interpretation.
• The industries are portfolios, not sectors. French assigns each firm to one of 49 definitions by SIC code. The parquet carries the 49 monthly return series and nothing else: no holdings, no market values, no firm counts. Which companies sat inside a portfolio in any given month cannot be read off this data.

Method, units, sentinel handling and the two conflicting Mkt-RF series are documented on the methodology page.