Since our manuscript was accepted for publication, we have become aware of weaknesses in the empirical approach used for parameter estimation and inference. A primary objective of the paper was to document the prevalence and duration, yet also rapidly shifting character, of endogenous microstructure noise in asset price processes. Once we establish a significant degree of endogeneity between the underlying “fundamental” price and the noise process, it is well-known that the different components cannot be identified separately through nonparametric techniques. Consequently, we developed a structural parametric model, assumed to apply only locally over short intraday intervals. The model is parsimonious, yet generalizes most existing empirically oriented microstructure models. It features temporary over- and under-reaction to news, but still allows for a relatively quick error correction, or feedback. The premise is that market participants behave rationally, but have diverse and imperfect information regarding other agents’ knowledge and motives, so the interpretation of, say, increased trading intensity, large price innovations, or incoming public information generates random errors which will persist for some time. By estimating the model over numerous short intervals for many different stocks, we pooled information to generate sufficient test power to formally document the endogenous nature of the fundamental price-noise relation, motivating our structural parametric approach. The subsequent estimation procedures and results, summarized in the paper through the distribution of model parameters across time and stocks, are where the weaknesses emerge. This corrigendum seeks to clarify the issues and discuss their implications. We stress that the corrections below do not change the main messages of the paper, but they do render specific results, summarized in a subset of the figures, incorrect.
ASJC Scopus subject areas
- Economics and Econometrics