In the shadows of opacity: Firm information quality and latent factor model performance

Little is known about how the performance of latent factor models is affected by the quality of firm-disclosed data. Using Chinese data, we demonstrate the superiority of conditional latent factor models (exemplified by the instrumented principal component analysis, IPCA) over unconditional latent factor models (risk-premium principal component analysis, RP-PCA; cross-sectional and time-series principal component analysis, XS-TS-Target-PCA). IPCA’s outperformance is generally more pronounced in explaining trading-based firm characteristics than accounting-based ones. However, in emerging markets such as China, IPCA’s performance is attenuated by the lower quality of firm-disclosed information and poorer stock liquidity. We make the first attempt to investigate how IPCA’s performance is affected by more opaque information environments in emerging markets.