Special Issue of Quantitative Finance on ‘Financial Data Analytics’
This Special Issue is to draw together connected papers which use the astonishing breadth of data available today, together with advances in computational techniques, to deliver insights into sectors of the financial world which could previously only be speculated upon.
Abstract
Data are everywhere. The data universe is far richer and more far-reaching than it was a decade, five or even three years ago. Data are collected routinely on platforms and via media undreamed of a short while ago. And so much of these data are relevant to finance. Who would have thought, in the pre-Twitter world, that discussion threads could be connected to market risk appetite? Or those algorithms could begin to detect how much information asymmetry exists in a market? The focus of this Special Issue is to draw together connected papers which use the astonishing breadth of data available today, together with advances in computational techniques, to deliver insights into sectors of the financial world which could previously only be speculated upon. We begin with papers which have done novel work in the area of textual analysis. Rönnqvist and Sarlin study the Reuters online news archive, which contains approximately three million articles, to understand the degree to which banks are connected and interdependent—this could prove to be a useful indicator of risk in an era where the phrase ‘too big to fail’ has come into common parlance. Conceivably too much connectivity could serve as an indicator of needed structural changes. Yang, Mo and Liu have focused on the Twitter financial community and whether it can predict stock market movement. This elegant paper first constructs the financial community within Twitter using network techniques, and then shows that critical ‘nodes’ within this community are significantly correlated with major financial market indices. Yaros and Imieliński use text analysis of equity analyst coverage and news articles to show that market sentiment expressed in these media gives a good indication of the future similarity and correlations of different equity markets, a technique which could have significant utility for the construction of investment portfolios. The next paper is more relevant to those who study the activity, motivation and methodology of the trading community. Algorithmic trading, where deals in the market are executed according to predefined rules and without human intervention, has become commonplace in the market. However, the impact of this algorithmic execution on market dynamics is uncertain, and thus a rich data universe has been gathered on trading patterns from different institutions and individuals. Yang, Qiao, Beling, Scherer and Kirilenko utilize these data to spot trading patterns unique to individuals or particular algorithms, enabling them to identify which traders or execution platforms are active at different times. This additionally allows them to identify to some extent the drivers behind traders’ actions under different market conditions. After this, we enter the world of machine learning. In recent years, there has been a degree of discussion about whether hedge fund returns are really due to the skill and experience of their portfolio managers, or whether they could be created more cheaply from simple underlying assets. Payne and Tresl perform a comprehensive study on over 4500 stocks to show that genetic algorithms can be used to create replicating portfolios of simple underlying assets like stocks, bonds and mutual funds, which show good out-of-sample correlation with the hedge fund returns. Next, Wilinski, Cui, Brabazon and Hamill analyse individual trades on the London Stock Exchange. They include an examination of time-of-day effects which is new and significantly shows that price impact is the highest at the start of the trading day and the lowest towards the end, important information for those with large flows to execute. Finally, Panayi, Peters and Kosmidis offer a large-scale study of liquidity and liquidity resilience, using high frequency data to show that resilience is only partly explained by market factors, but at extreme levels is dominated by individual asset factors. We hope this Special Issue provides valuable insights into the rapidly growing area of data analytics in finance and contributes to the advances of this field. We thank the contributing authors; the referees for their valuable input and the Editors-in-Chief for their support.