Internalization in a Post-MiFID 2 World

Project Start:01/2019
Researchers:Satchit Sagade, Erik Theissen, Christian Westheide
Category: Financial Markets
Funded by:LOEWE

Internalization - the practice of broker-dealers executing client orders against their own capital - has been a feature of financial markets for a long time. It has also been highly controversial due to potential conflict with broker-dealers' best execution obligations and incentive to engage in cream-skimming by only internalizing uninformed orders. Consequently, it can potentially lead to a deterioration in public market liquidity which, in equilibrium, hurts all investors, including clients whose orders are internalized, as internalizers’ quotes are affected to public market liquidity.

MiFID 2’s implementation has led to a 10-fold increase in internalizers’ market share in the EU. We propose to examine the determinants of SI activity and its impact on market quality. We plan to use anonymized transaction data from BaFin (they have already approved this project). It includes information such as transaction price, size, time, execution venue, SI ID, and customer ID. We will combine it with high-frequency data from Thomson Reuters Tick History (TRTH). First, we will classify SIs based on their identity (High-Frequency Traders, Investment Banks, etc.) and their mix of clientele (retail vs. institutional, large vs. small, etc.). Next, we will examine how their activity varies by market conditions (e.g., liquidity and volatility) and market’s trading protocols (e.g., tick sizes). We will also focus on the relationship between SIs’ level of price improvement vis-à-vis trade sizes and public market tick sizes. Finally, we will determine the impact of SI activity on market quality. Here we will focus on two channels: (i) the extent to which SIs compete with public markets by cream skimming; and (ii) the extent to which they utilize their institutional characteristics to engage in cost competition.

We will employ a combination of panel and logit regressions. Panel regressions allow us to simultaneously capture changes in variables across stocks and over time. Logit models are suited to examine the internalize/not internalize decision. The impact of rule changes and changes in tick sizes could be examined by employing quasi-experimental techniques such as difference-in-difference or regression discontinuity design. We are requesting funding for: (i) 50% of Satchit Sagade’s position as postdoc; (ii) TRTH subscription; (iii) travel expenses to present this and other papers from team projects (see below); (iv) publication expenses; and (v) 1 student assistant.