Do the Factors Work Abroad? Fama and French Check

Every factor you have ever heard of was discovered in the United States. That is not a small problem. It raises the possibility that the entire edifice of factor investing is an elaborate description of one country's stock market over one particular stretch of the twentieth century.

The only real cure is to go somewhere else and look. In 2012, Eugene Fama and Kenneth French did exactly that, taking size, value and momentum to four regions: North America, Europe, Japan, and Asia Pacific. The results are a useful mixture of reassuring and unsettling, and the unsettling parts are the more interesting.

The problem: out-of-sample means out of the country

A factor found by searching US data has, in a sense, already used up its evidence. Testing it again in the same market proves little. Genuine out-of-sample evidence means new data that could not possibly have influenced the original discovery, and a different country is about as clean as that gets: different investors, different institutions, different accounting, different history.

So Fama and French built size, value and momentum portfolios in each of the four regions and asked two questions:

1. Are the premia there at all?
2. Can any asset pricing model explain them, and does the world price assets as one integrated market or as separate local ones?

The findings, including the awkward one

Value shows up everywhere. In all four regions there are value premiums in average returns. Cheap stocks beat expensive ones, in North America, in Europe, in Japan and in Asia Pacific. That is real, meaningful out-of-sample support for the most famous factor in finance.

Value is stronger among small stocks. In every region except Japan, the value premium is larger for smaller companies and shrinks as you move up the size scale. This is a persistent, robust pattern, and it is one reason so many value strategies end up quietly tilted small.

Momentum shows up everywhere except Japan. And here is the awkward exception that has caused years of argument. Japan has essentially no momentum premium. Everywhere else, past winners keep winning. In Japan, they do not.

This matters more than it sounds. Momentum is supposed to be the most universal effect in finance, present across centuries, asset classes and countries. A large, developed, heavily traded market where it simply does not appear is a genuine problem for any confident theory of why momentum exists. It is either evidence that the momentum premium is partly luck, or evidence that something specific about Japan suppresses it, and the literature has not settled which. Practitioners note, fairly, that a combined value-and-momentum strategy still works in Japan because value is strong there, which is precisely the diversification argument. But the bare fact remains uncomfortable.

Global models do not work. The most technically important finding is about integration. If the world's capital markets were one integrated market, then a single set of global factors should explain returns everywhere. They do not. Fama and French find that global models do a poor job of describing regional average returns, and that local models, built from local factors, work considerably better in North America, Europe and Japan.

That is a significant negative result. It means asset pricing does not appear to be integrated across regions, which is awkward for the theory (capital is supposed to be mobile) and useful for the practitioner (if you are building a factor model for European stocks, build it from European data, not from a global blend).

Why it mattered

• It is the out-of-sample test the value factor needed. Value appearing in four independent regions is about the strongest defence available against the charge that it was mined from US data.
• It documented the Japan momentum problem honestly. Fama and French are famously skeptical of momentum, and they reported the exception rather than burying it. Any honest account of momentum has to deal with Japan.
• It settled a practical modelling question. Use local factors for local portfolios. That guidance is now standard, and it came from this paper's failure to make global models work.
• It tempered the size factor further. The evidence for a standalone international size premium was underwhelming, adding to the accumulating doubts about size as an independent factor, though size continued to matter as a dimension along which the value premium varies.

The honest limitations

• The samples are shorter and the data is worse. International data begins later and is patchier than US data, so the tests have less statistical power. Failing to find something in a shorter sample is weaker evidence than finding it.
• Four regions are not four independent draws. Developed markets are correlated, share investors, and are exposed to the same global shocks. The out-of-sample evidence is real, but it is not four times the evidence.
• Japan is unexplained, not explained away. The paper reports the anomaly-within-the-anomaly without resolving it. Explanations offered elsewhere, ranging from the long post-bubble deflation to differences in corporate ownership, remain speculative.
• Rejecting integration raises more questions than it answers. If pricing is not integrated globally, why not? Capital is mobile. The finding is credible and awkward, and the paper does not attempt to explain the mechanism.

The one-line takeaway

Fama and French took their factors abroad and found that value works in all four major regions and momentum works in three of them, with Japan the stubborn exception, and that a single global model fails: markets are priced locally, not as one integrated whole.