Enhancing Financial Latent Factor Analysis: Application of Instrumented Principal Component Analysis (IPCA) and Procrustes Alignment

This study examines the application of Instrumented Principal Component Analysis (IPCA) to financial datasets, including synthetic data and real-world data from the Wharton CRSP/Compustat Merged Database. By leveraging observable characteristics like volatility, moving averages, and trading volume, IPCA enhances the estimation of latent factors in high-dimensional financial datasets. To address the non-uniqueness of factor loadings across cross-validation folds, Procrustes alignment is employed to standardize the Gamma matrices, ensuring consistency and interpretability across time periods. Findings indicate that, while IPCA captures certain latent structures, the variance explained by the identified factors is modest, highlighting the influence of unobserved dynamics. The alignment of Procrustes significantly improves consistency, reducing discrepancies by more than 70 %. The research underscores the potential of IPCA for financial analysis, particularly when combined with alignment techniques, while emphasizing the limitations of using a restricted set of characteristics.