How effective are narratives for pricing Chinese stocks?

We study whether narrative factors extracted from Chinese financial news are informative for asset pricing in the Chinese stock market. Using latent Dirichlet allocation and instrumented principal component analysis, we construct narrative factors that admit an interpretable topic-based representation and evaluate their empirical pricing performance. Our analysis reveals that narrative factors (NF) capture systematic risks and deliver lower pricing errors than China-specific benchmark factor models (CH), although their overall performance is weaker than the benchmark evidence reported for the U.S. market. We further find that the NF models perform particularly well among small-cap stocks and state-owned enterprises, and narrative factors provide incremental pricing information beyond standard observable factor models. Moreover, integrating narrative factors with firm characteristics leads to stronger mean–variance–efficient Sharpe ratios and cumulative excess returns, highlighting the complementarity between textual narratives and firm-level information in empirical asset pricing.