High Frequency Trading and Ghost Liquidity – GHOST

The originality of our project is based on a unique database provided by the European Securities Markets Authority (ESMA) and covering most active European trading venues for a large sample of stocks in May 2013. The ESMA dataset contains confidential detailed information about traders which will allow us to track them across venues. We have processed the raw ESMA data to create cross-market consolidated order books synchronized with trades, and to classify traders as fast (HFTs) or slow (non-HFTs) according to several criteria. Those proprietary data are used to build original GL measures and to design GL proxies that could be computed from public intraday data. On the basis of those measures, we will conduct three empirical studies. The empirical work will be complemented by theoretical work aiming to model GL in a market setting with two order-books, with fast and slow traders and to build theoretical predictions on the determinants and effects of GL.

At this stage, we have finished the structuring of the database and we have classified market members in different categories. We have established some descriptive statistics on GL. They show that GL is greater for HFTs and proprietary traders. Those preliminary findings have not been published yet..

Based on our preliminary results, we are conducting a multivaraite analysis of GL to identify its determinants..

Forthcoming publication of an article on market fragmentation and HFT:
Gresse, Carole (2017). Effects of lit and dark market fragmentation on liquidity. Journal of Financial Markets, forthcoming.

Today high frequency trading (HFT) plays a central role in financial markets. The rise of HFT has tremendously increased the frequency of messaging in stock markets and has changed the way liquidity is supplied to stock markets. At the same time, markets are fragmented and stocks trade on several venues. Some HFT strategies consist of supplying liquidity on several trading venues simultaneously and then withdrawing that liquidity as soon as some orders from the strategy are executed on one of the venues. This makes the true level of market liquidity different from its perceived level – the difference being ghost liquidity (GL). Understanding GL is important as it causes investors and regulators to mis-estimate the likely effects of trades and thus may lead to sub-optimal investment decisions. Our project aims to establish empirical measures of GL, to quantify its magnitude in European stock markets, to identify its determinants and understand its links with HFT, and to assess its impact on market quality and fairness. The originality of our project is based on a unique database provided by the European Securities Markets Authority (ESMA) and covering most active European trading venues for a large sample of stocks in May 2013. The ESMA dataset contains confidential detailed information about traders which will allow us to track them across venues. We will process the raw ESMA data to create cross-market consolidated order books synchronized with trades, and to classify traders as fast (HFTs) or slow (non-HFTs) according to several criteria. Those proprietary data will then be used to build original GL measures and to design GL proxies that could be computed from public intraday data. On the basis of those measures, we will conduct three empirical studies. First we will undertake a study identifying the determinants of GL. Second, we will conduct analysis of the impact of GL on liquidity for different subsets of traders and how GL links to HFT and market fragmentation. Last, we will investigate the impact of GL on price stability, efficiency, and integration across markets. A challenge in assessing the impact of GL on market quality will be to solve endogeneity and identification issues using an instrumental variable approach. Further, due to the size of the data to be exploited and the technicality of the measures to be implemented, funding research assistance on the project is crucial for its success. The empirical work will be complemented by theoretical work aiming to model GL in a market setting with two order-books, with fast and slow traders and to build theoretical predictions on the determinants and effects of GL. The project should produce four publishable papers with theoretical, methodological, and empirical contributions. Our findings will have implications for regulatory bodies and market practitioners.