Heterogeneous Return Predictability from Order Flow

return predictability study using retail order flow

Paper Metadata

• Publication Date: 2025-10-07

• Source: SSRN

• Link: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5567999

Keywords

• market microstructure

• return predictability

• retail flow

Notes for Review

This paper investigates the relationship between information and inventory effects on return predictability from retail and total order flow. The author builds a model that combines the asymmetric information impact of investors with the inventory effect of market makers to analyze how lagged order flow can forecast future returns. The model suggests that the difference in predictive power between retail and total order flow can be attributed to the varying informativeness of different investor groups. The paper provides empirical evidence using data from stocks in the banking sector and finds that return predictability can turn negative when the market maker is extremely unwilling to provide liquidity. The study has important implications for market makers and traders who seek to understand the dynamics of order flow and return predictability. The use of WRDS Intraday Indicator and TAQ data provides a robust and reliable dataset for the analysis. Overall, this paper is a valuable contribution to the field of market microstructure and high-frequency trading. Key findings include the importance of inventory capacity and risk aversion in determining return predictability, as well as the interaction between information and inventory effects. The paper also highlights the need for further research on the relationship between market makers’ risk bearing capacity and return predictability.

---

Abstract

This study investigates how information and inventory effect jointly determine return predictability from retail and total order flow. I build a model that combines the asymmetric information impact of investors with the inventory effect of market makers to analyze how lagged order flow can forecast future returns. The model illustrates that the difference in predictive power between retail and total order flow can be attributed to the varying informativeness of different investor groups. The focus of this paper is to empirically test how market makers’ varying inventory capacity affect this predictive power. While previous literature has theoretically demonstrated that the predictive power of past returns is positive and increases with a market maker’s risk aversion, such a monotonic relationship requires specific model parameter constraints and lacks empirical support. My framework suggests that when predictibility remains positive, the magnitude increases when the market maker has lower risk bearing capacity, but this monotonic relationship is only within a certain range. Specifically, when the market maker is extremely unwilling to provide liquidity, return predictability can turn negative, as the price impact channel dominates. I empirically test this theoretical prediction using data from stocks in the banking sector and the results align with the model. The rationale is that these stocks exhibit the strongest positive correlation with the market maker’s business, and therefore they have lowest inventory capacity for these stocks. This finding is supported by the observation that negative predictability becomes more pronounced in times of higher market volatility or for stocks with lower liquidity.