A Dynamic BW Index of Stock Market Sentiment

Abstract

We explore a dynamic design of market sentiment aggregation, augmenting the seminal Baker and Wurgler (2006) index on the basis that aggregate sentiment must cover a wide range of investors or investment activities; such aggregation must adapt to changes in how different sentiment indicators reflect sentiment; and the aggregation must not be informed by future information. We demonstrate that each basis has a measurable impact. Ignoring these points when constructing models of sentiment can cause empirical effects to be over- or under-estimated and constrain the usability of the modelled sentiment in practical applications. Using a market event analysis, we show that our index is much more closely aligned with the boom and bust periods of the stock market. We also test our index against fourteen anomaly portfolio returns and confirm the classic sentiment-induced mispricing effect on short or hard-to-value legs, while observing increasing sentiment-induced mispricing on the long or easier-to-value legs post-2000.

Keywords: Investor sentiment, sentiment aggregation, PCA, dynamic modelling, equity returns, anomalies

JEL Classification: C43, G12, G41