A Dynamic Factor Model of Price Impacts

Abstract

We model the response of mean-variance optimizing investors to noise trading flows within a multi-asset framework, where these flows exhibit dynamic predictability. In our model, these flows impact asset prices via factors, which are portfolios of underlying assets. Unlike in standard mean-variance models, the price sensitivity to flows can differ across factors, as it depends on the specific factor's flow predictability. This dynamic feature allows our model to flexibly fit price impacts at various aggregation levels and to uncover cross-asset substitution patterns not present in standard models.

Keywords: dynamic, factor, flow, predictability, price impact

JEL Classification: G12