The market drives ETFs or ETFs the market: causality without Granger

TL;DR

This paper introduces a deep learning-based econometric method to analyze causality in high-frequency financial time series, revealing that ETF transaction imbalances are more informative than overall market imbalances.

Contribution

It presents a novel deep learning approach for causality detection in high-frequency financial data, focusing on ETF transaction imbalances and their relation to market movements.

Findings

ETF transaction imbalances are more informative than market-wide imbalances

ETF imbalance messages outnumber stock imbalance messages by 8:1

Deep learning effectively uncovers causality in high-frequency financial data

Abstract

This paper develops a deep learning-based econometric methodology to determine the causality of the financial time series. This method is applied to the imbalances in daily transactions in individual stocks, as well as the ETFs reported to SEC with a nanosecond time stamp. Based on our method, we conclude that transaction imbalances of ETFs alone are more informative than the transaction imbalances in the entire market. Characteristically, a sheer number of imbalance messages related to the individual stocks dominates the imbalance messages due to the ETF in the proportion of 8:1.