The author studies the “failure” of the momentum strategy in Japan.
He finds that the failure, which occurred only in Japan, does not prove that
momentum strategies are ineffective. Rather, momentum strategies combined with
the value effect are powerful across the globe, including in Japan.
Various studies conducted during the early to mid-1990s have documented that the success
of momentum strategies is one of the strongest empirical regularities in finance.
Momentum has joined size and value as one of the main risk factors in modern investing.
The author notes that both momentum and value have shown consistent out-of-sample
success when tested across geography, asset class, security type, and time. One notable
exception, however, has been the empirical failure of momentum for stock selection in
Japan.
The author examines the failure of momentum in Japan and seeks to discover whether this
exception actually proves the rule or exposes momentum strategies as being flawed. He
also argues that value and momentum have a negative correlation and must be viewed
together as a system as well as studied together.
The author uses international equity returns from Datastream aggregated across four
regions: the United States, the United Kingdom, Europe (ex United Kingdom), and Japan.
The data cover the period from July 1981 to December 2010, from which value and momentum
portfolios are obtained for each of the four regions. Eight data series are created: the
returns of a large-capitalization, liquid, dollar-balanced, and long–short value
strategy for each of the four regions and the returns of a similarly constructed
momentum strategy for each of the four regions.
From an empirical analysis of the data series, the author obtains results of the value
and momentum strategies for each of the four regions and for an “All” region
that invests an equal dollar amount in all four regions. For each region, he reports the
average annualized returns, annualized standard deviations, Sharpe ratios, and
correlation between value and momentum, he creates a portfolio for each region with
equal weights in value and momentum.
Studying the degree of correlation between value and momentum in each region, he uses the
correlation value to create optimal (Sharpe ratio–maximizing) portfolios for each
region by assigning different weights to the value strategy and the momentum strategy.
Finally, the author performs a factor regression analysis using the Fama–French
three-factor model (using market exposure, size, and value as the three factors) on the
momentum strategy across the four regions and the All region.
The key finding is that the results in Japan do not disprove the momentum strategy. The
author shows that, instead, the results are ultimately supportive of momentum
strategies. He confirms the basic finding that value and momentum work nearly everywhere
but that momentum does not work in Japan. He shows that given the success of both value
and momentum strategies around the globe, the chance of one strategy delivering poor
results in one region is quite high. An optimized portfolio that uses both value and
momentum would still do well investing heavily in the Japanese momentum strategy,
despite its poor performance. Using the three-factor model, the author finds the size of
the momentum strategy’s intercept in Japan to be economically and statistically
large, meaning that net of market exposure, size, and value, momentum strategies have
added considerable return over the period of study.
The supposed failure of Japanese momentum is not at all statistically significant, and
there is a considerable chance that it is just random noise. The author concludes that
the portfolio of a Japanese investor with access to the value strategy would have
performed much better if it had included momentum. By viewing value and momentum as a
system that has negative correlation and using a three-factor model for support, the
author provides indications that momentum in Japan is a strong empirical success.