Decomposing anomalies


  • Boubaker Sabri
  • Li Bo
  • Liu Zhenya
  • Zhang Yifan


  • Asset pricing
  • Anomaly variable
  • Factor model
  • Functional principal component analysis
  • Eigenfunction

document type



This paper introduces the functional principal component analysis approach for decomposing the panel returns of the anomaly-sorted portfolios. Using the US stock market data covering July 1963–July 2020, our findings indicate that the Fama–French (F–F) market factor can be captured by the first empirical functional principal component in the time-series. For the other F–F anomalies, market capitalization (Size), book-to-market ratio (B/M), profitability (OP), investment (Inv), and price momentum (Mom), the cross-sectional features remain in the monotonicity of the second principal component and in the curvature of the third principal component. Furthermore, a time-varying framework shows two neglected reversals of the F–F anomalies Inv and Size in the 1970s and the 1980s.

more information