The authors investigate the value premium puzzle—that is, profitable strategies of buying stocks with a high book-to-market ratio (B/M), or value stocks, and selling stocks with a low B/M, or growth stocks—in six non-U.S. markets: the United Kingdom, Germany, France, Italy, Canada, and Japan. They find that the current B/M has significant explanatory power but is not universally strong for different capitalization sizes.

Following Fama–French (Journal of Finance 2008), the authors recognize that changes in B/M can be decomposed into changes in book equity and changes in market price. For all countries examined, except Japan, the predictive value of B/M appears to be driven more by recent changes in book equity than by recent changes in market price. The effect is generally stronger for smaller capitalization stocks. Finally, the effect dissipates over longer periods, which indicates that recent changes are more important than longer-term changes.

Practitioners will benefit from being able to see that the value premium exists in non-U.S. markets and that the value premium varies between different market capitalizations within the different markets. Furthermore, the effect that appears to matter the most is changes to book equity over the previous year.

Practitioners should also be aware that the effect of the value premium is not always the same across countries and that there are some notable exceptions to this general effect, such as Japan. Also, the investigation does not consider the period after 2007, which excludes the recent financial crisis.

The authors use monthly stock return data from July 1981 through June 2007, ending their dataset just before the recent financial crisis. B/M is calculated from the book value as of the end of June of a given year and the market value as of the end of December in the previous year. B/M is then applied in regression analysis for the following June through July beyond the month in which the book value is calculated.

Market capitalization designations are determined by relative market capitalization within a given market. Firms in the smallest 20% of market capitalization are considered “microcap.” Firms in the 20–50% market capitalization range are considered “small cap,” and the remaining firms are considered “large cap.”

The authors deconstruct the log of the current B/M (BM_t) to be equivalent to BM_t–k + dB_t–k – dM_t–k, where k is a lag term (12 months) and dB and dM are the change in book equity and market price, respectively, over the lagged period. Before estimating regressions, they control for market capitalization and for changes in the number of shares outstanding with the dependent variable always being the next period stock return. Three regressions are performed for each country with the following independent variable combinations: (BM_t only), (BM_t–k, dB_t–k, and dM_t–k), and (BM_t, dB_t–k, and dM_t–k). The regressions also include various criteria for market capitalization, including all stocks, large cap and small cap only, and each separate market-cap designation individually. The three regressions are performed for each of the five different market capitalization combinations.

Cross-sectional regressions demonstrate that BM_t has significant explanatory power but is not universally strong in different capitalization sizes. Further regressions demonstrate that, except for Japan, the predictive value of BM_t appears to be driven more by dB_t–k, the recent change in the book equity, than by the change in market price. The effect is generally stronger for smaller capitalized stocks.

The results are robust for longer-term lag periods (two- and three-year lags), but they decrease as the lag period increases. The results are also robust to controls for momentum.

I think the analysis is interesting in that the value premium puzzle appears to exist in other developed markets. But I am left a little disappointed that I cannot determine whether the results continue after the financial crisis of 2008.