Gene

Leverage effect

Engineering

Philosophy

Volatility (finance)

Stock (firearms)

Biology

Statistics

Predictability

Stock market

Robustness (evolution)

Economics

Biochemistry

Computer science

Paleontology

Mechanical engineering

Market sentiment

Mathematics

Realized variance

Predictive power

Epistemology

Econometrics

Leverage (statistics)

Horse

Artificial intelligence

Financial economics

Chemistry

Autoregressive conditional heteroskedasticity

Stock market index

International Review of Financial Analysis

This study investigates the predictability of sentiment measure on stock realized volatility. We propose a new investor sentiment index (NISI) based on the partial least squares method. This sentiment index outperforms many existing sentiment indicators in three aspects. First, in-sample result shows that the NISI has greater predictive power relative to the others. Most sentiment indicators show predictability in the non-crisis period only while the NISI is also effective in the crisis period. Furthermore, the NISI exhibits more prominent superiority in longer horizons forecasting. Second, further analysis indicates that the NISI has robust predictability before and after the Chinese stock market turbulence periods while the others not. Importantly, the NISI is still effective significantly after considering leverage effect while most of the others not. Finally, out-of-sample analysis demonstrates that the NISI is more powerful than other sentiment measures. This result is reproducible in different robustness checks.

Investor sentiment and stock volatility: New evidence