Does news sentiment precede liquidity shocks? A liquidity-first event study of Indian equities

This study examines whether changes in the public information environment precede firm-level liquidity shocks in Indian equities, using a liquidity-first event design that first identifies sustained deteriorations in Amihud illiquidity and then examines the preceding information flow. The analysis covers 74 event observations across 68 distinct NSE-listed tickers (2022–2025) and evaluates five information-measure families: three core sentiment measures (broad GDELT tone, headline-explicit GDELT tone and entity-conditioned large language model (LLM) sentiment) and two complementary channels (search attention and investor-forum activity and sentiment). None shows statistically significant aggregate displacement. For broad GDELT, the historical-placebo test gives an observed mean displacement of 1.379 versus a null mean of 1.315 (n = 26; p = 0.381). The shift-based 80%-power minimum detectable effect is 0.589 displacement units at raw alpha = 0.05 (approximately 0.41 event-specific baseline-SD units), so smaller systematic effects cannot be ruled out. At the event level, raw p < 0.05 counts do not exceed nominal benchmarks for measures with adequate empirical-null resolution; no observation survives false-discovery-rate correction, but this non-rejection is not evidence that no individual precursor exists. Reddit mention activity covers all 74 events, while T − 1 FinBERT sentiment remains exploratory (n = 5).