Key takeaways
- Across the 409 months from July 1992 to July 2026, the Fama-French value premium averaged 2.17% a year in the US, 5.29% in developed markets outside it and 7.86% in emerging markets.
- Momentum ran in the same order with a wider gap: 5.18% a year in the US, 7.90% developed ex-US and 10.21% in emerging markets, all of it before trading costs.
- Size paid almost nothing anywhere. The small-minus-big premium averaged 1.12%, 0.36% and 0.21% a year, on t-statistics of 0.62, 0.32 and 0.17.
- Three regions are not three independent tests. Monthly US and developed ex-US value premiums correlated 0.66, and the momentum pair correlated 0.77.
- Emerging value paid 14.82% a year in the 2000s and 1.89% a year in the 2010s, so one decade carries much of the 34.1-year average.
Outside the US the premiums were larger, and size paid nothing in any region
You've read that the value premium died somewhere around 2007, and that the case for factors rests on one long US backtest. So the question worth asking isn't whether value worked in America. It's whether the same sorts paid anything in the rest of the world, where the researchers who found the effect were not looking.
They did, and by more. Over the 409 months from July 1992 to July 2026, the value premium averaged 2.17% a year in the US, 5.29% a year in developed markets excluding the US, and 7.86% a year in emerging markets. Momentum came in at 5.18%, 7.90% and 10.21%. Size paid 1.12%, 0.36% and 0.21%, and not one of those three is distinguishable from zero.
That ordering is the easy part of the story. The harder part is that the three samples cover the same calendar months, correlate with each other, and put the biggest premiums in the markets that cost the most to trade. One decade also does an uncomfortable share of the work. That's what the case for international factor investing has to survive, and here's the arithmetic behind each objection.
What international factor investing measures, and which files it comes from
Kenneth French's data library publishes the same five factors region by region. The developed ex-US set runs monthly from July 1990, the emerging set from July 1989. The emerging investment factor is blank until the middle of 1992, so the common window used throughout this piece starts in July 1992 and ends with the July 2026 file. That's 409 months, or 34.1 years, with every factor present in all three regions.
Coverage is wider than most people assume. The developed table names 23 countries, the United States among them. The emerging list currently names 24, running from Brazil and Chile through India and Indonesia to Taiwan, Thailand and Turkey. All returns are in US dollars and include dividends. The Fama-French factor files behind every number here are public, which is why this is one of the few corners of international factor investing you can audit yourself.
The five factors are long-short paper portfolios, not funds. HML is high minus low book-to-market, the value premium: the average return on the two value portfolios minus the average return on the two growth portfolios. SMB is small minus big. RMW is robust minus weak operating profitability. CMA is conservative minus aggressive investment. The momentum factor, WML, is winners minus losers on the prior 2-to-12-month return. Internationally, the sorting breakpoints for book-to-market, profitability and investment are the 30th and 70th percentiles of the big stocks in each region. RMW and CMA are the two factors Fama and French added to their model in 2015, and what the five-factor model gained over the three-factor version depends on which portfolios it is asked to price.
Value: 2.17% in the US, 5.29% developed ex-US, 7.86% in emerging markets
The gap is not marginal. Developed ex-US value paid roughly two and a half times the US premium over the same months, and emerging value paid more than three times it. What matters more than the averages is how reliable they were. The t-statistic, which is simply the average divided by its own standard error, was 1.12 in the US, 3.74 developed ex-US and 5.76 in emerging markets. A reading near 2 is the conventional line at which a sample average stops looking like chance.
So the US value premium over these 34.1 years fails the usual test, while the developed ex-US value premium and the emerging one clear it comfortably. The volatility numbers point the same way. Annualised, the value premium ran at 11.3% in the US, 8.3% developed ex-US and 8.0% in emerging markets. Outside the US it was both larger and steadier, which is why the t-statistics diverge so much more than the raw averages do. Our read on the value premium after 1992 covers what happened to the US series on its own.
Momentum had the widest regional gap and the worst tail risk
Momentum is the factor where the international data looks most flattering and most dangerous at once. The premium averaged 5.18% a year in the US on a t-statistic of 1.84, 7.90% developed ex-US on 3.88, and 10.21% in emerging markets on 5.48. Volatility went the other way: 16.4% in the US against 11.9% developed ex-US and 10.9% in emerging markets.
Then there's the tail. The worst rolling twelve months in the sample returned -56.5% for US momentum, over December 2008 to November 2009. Developed ex-US momentum returned -38.8% across the same twelve months, and emerging momentum -36.8% over November 2008 to October 2009. These crashes arrive together, which is the point: a portfolio holding momentum in three regions held one exposure, not three. The mechanics of that reversal are set out in our piece on the momentum premium and its crashes.
The most recent month in the file is its own warning. Emerging momentum's worst single month across all 409 months was July 2026, at -16.84%.
Size paid nothing in any region, and the breakpoints make that result harder, not easier
The size premium averaged 1.12% a year in the US, 0.36% developed ex-US and 0.21% in emerging markets, on t-statistics of 0.62, 0.32 and 0.17. Those are null results in every region tested.
The construction detail makes that more striking rather than less. The US factors split large from small at the median NYSE market equity, applied to a universe of NYSE, AMEX and NASDAQ stocks. The international factors don't. French's description states that big stocks are those in the top 90% of June market cap for the region, and small stocks are those in the bottom 10%. The international small leg is therefore a far more extreme cut of the market than the US one, and it still returned less. Our look at the size premium after Banz traces the US series from Banz's 1981 publication onwards.
Profitability and investment travelled better than their headline premiums suggest
These two get less attention and deserve more. The profitability premium averaged 3.65% a year in the US, 2.80% developed ex-US and 2.67% in emerging markets. Read only those numbers and the US looks best. Read the t-statistics and it doesn't: 2.23 in the US against 3.46 developed ex-US and 2.85 in emerging markets. The reason is volatility. The US profitability premium ran at 9.5% a year, the developed ex-US one at 4.7%.
The investment premium is the most uniform of the five. It paid 2.22% a year in the US, 1.99% developed ex-US and 2.81% in emerging markets, on t-statistics of 1.71, 1.89 and 2.42. A factor that pays about the same modest amount everywhere is a different kind of evidence from one that pays a lot in one place.
The premium by decade, where the averages stop being comforting
A 34.1-year average hides the thing a holder actually experiences. Split the same 409 months into decades and the picture changes in every region. All figures are annualised averages of monthly factor returns, in percent a year.
| Period | US value | Dev ex-US value | EM value | US momentum | Dev ex-US momentum | EM momentum |
|---|---|---|---|---|---|---|
| Jul 1992 to Dec 1999 | 0.07% | -0.34% | 5.66% | 14.79% | 11.15% | 13.39% |
| 2000 to 2009 | 7.95% | 12.68% | 14.82% | 1.01% | 4.54% | 5.07% |
| 2010 to 2019 | -2.35% | -0.97% | 1.89% | 3.18% | 8.78% | 10.74% |
| 2020 to Jul 2026 | 2.64% | 10.00% | 8.86% | 3.60% | 7.97% | 13.60% |
The 2000s carry the value result almost everywhere. Developed ex-US value paid 12.68% a year that decade and -0.97% the next. Emerging value went from 14.82% to 1.89%. The US went from 7.95% to -2.35%. A reader who bought the international value case in January 2010, on the strength of the previous decade, collected -2.35%, -0.97% and 1.89% a year for the next ten.
Momentum outside the US is the steadier of the two. Developed ex-US momentum paid between 4.54% and 11.15% a year in every period shown, and emerging momentum's weakest decade, the 2000s at 5.07%, still beat the US figure of 1.01%.
The strongest objection: three regions are not three independent tests
This is the criticism that does the most damage, and it is right. If the ex-US result were an independent confirmation, the regional premiums would move separately. They don't. Over these 409 months the US and developed ex-US value premiums correlated 0.66, the US and emerging pair 0.40, and developed ex-US against emerging 0.46. Momentum is tighter still, at 0.77, 0.56 and 0.60. Only size looks genuinely independent, at 0.27, 0.04 and 0.32, and size is the factor that paid nothing.
Asness, Moskowitz and Pedersen made the constructive version of this point in Value and Momentum Everywhere, published in the Journal of Finance in 2013. They found "consistent value and momentum return premia across eight diverse markets and asset classes, and a strong common factor structure among their returns", and argued the results "present a challenge to existing behavioral, institutional and rational asset pricing theories that largely focus on U.S. equities". A common factor structure cuts both ways. It makes the effect harder to dismiss as a US data-mining artefact, and it makes a three-region factor portfolio less diversified than three separate numbers imply.
There's a second objection worth stating plainly. The regions where the factor premiums were largest are the ones where the market itself paid least for the risk taken. The market risk premium over these months averaged 9.28% a year in the US, 5.87% developed ex-US and 7.65% in emerging markets. An investor who tilted a portfolio away from US equities to chase a 7.86% value premium gave up part of a market return that was 9.28% a year to do it. Our comparison of emerging markets allocation weights runs that trade-off at three different portfolio sizes.
What this data cannot tell you
These are the limitations, and they are not small. Every figure here is a gross return on a long-short paper portfolio. There is no borrow cost on the short leg, no spread, no market impact, no turnover charge and no tax. Momentum reforms monthly, so its costs are the highest of the five, and emerging markets are where trading costs and shorting constraints bite hardest. The measured emerging premium of 10.21% a year is therefore the most overstated number in this piece, not the most attractive one. Emerging market factor premiums are the ones a net-of-cost record would cut hardest.
The data itself has seams. French notes that the emerging raw data incorporate data from both International Finance Corporation and Bloomberg for 1989 to 1994, and from Bloomberg alone from 1995. The emerging investment factor doesn't exist before the middle of 1992, which is why the sample starts there. And 34.1 years is one sample of international factor investing, not a law. Backtests describe what happened to a set of sorting rules in a particular window; they are not forecasts, and historic premiums carry no promise about the next decade. Live products fare worse again, as our audit of factor ETF returns against their backtests shows.
What would change the conclusion
Three things would falsify the reading above, and each has a number attached. If the developed ex-US value t-statistic fell back below 2 from its present 3.74, the international evidence would be no stronger than the US evidence it is meant to corroborate. If emerging value converged on the US figure of about 2.17% a year, the geography argument would be dead. And if the next ten years looked like the 2010s in all three regions at once, at -2.35%, -0.97% and 1.89%, the 2000s start to look like a single regime rather than a premium.
One of those tests is already running. Emerging momentum's worst month in 409 months of data was the last one in the file, July 2026 at -16.84%. Whether that turns out to be noise or the start of the crash pattern of late 2008 is the thing to watch, and the July 2027 update to these files is where the answer shows up first.