Forecasting Japanese Equity Returns Using Equi-Correlation Structures and Component Selection
A new study finds that standard asset pricing models like CAPM and Fama-French fail to predict Japanese stock returns. Researchers introduce two novel risk factors: the Industry Equi-Correlation index, which measures how industries move together, and a principal component factor from intra-industry correlations. These factors, independent from traditional ones, greatly improve prediction accuracy. The principal component factor consistently outperforms others. The study uses advanced statistical methods but does not fully explain the economic reasons behind these findings.
What it examines
This study tests if adding new correlation-based risk factors to traditional asset pricing models improves the prediction of Japanese stock returns. Using advanced statistical methods and 100 sorted portfolios, it examines both market-wide and industry-level comovement to better explain equity returns in Japan.
What it concludes
Adding correlation-based factors, especially those from industry-level comovement, greatly improves model performance in explaining Japanese stock returns. These findings can help investors and risk managers build better portfolios. Future research should explore why these correlation risks matter and test the approach in other markets.
Evidence objects
A new study finds that classic asset pricing models like CAPM and Fama-French ($FF3$, $FF5$) fail to explain Japanese stock returns, prompting the introduction of two innovative risk factors.
key_findings bullet 1 · key_findings · validation V0
The Industry Equi-Correlation (IEC) index and a principal component (PC) factor, derived using advanced techniques like PCA and Block DECO, are largely independent from standard factors and dramatically improve prediction accuracy.
key_findings bullet 2 · key_findings · validation V0
Surprisingly, the PC factor, capturing hidden intra-industry stock patterns, consistently outperforms traditional factors across all tests, setting a new standard for Japanese asset pricing but leaving economic explanations for these effects unexplored.
key_findings bullet 3 · key_findings · validation V0
This paper innovatively integrates correlation-based factorsIndustry Equi-Correlation and intra-industry PCAinto traditional asset pricing models like CAPM and Fama-French, specifically for Japanese equities. Employing Block DECO and principal component analysis, it offers a novel empirical approach, demonstrating improved model performance and providing unique, compelling insights for asset pricing in Japan.
key_findings bullet 4 · key_findings · validation V0
Raw abstract and provenance
- … Using 100 size-and-characteristic sorted portfolios as test assets, we evaluate 16 distinct models with a comprehensive set of econometric methods, including Fama–…
Source row: 902 · abstract type: snippet