Skip to main content

Algorithmic Finance: Inquiring into the Reshaping of Financial Markets

Periodic Reporting for period 2 - AlgoFinance (Algorithmic Finance: Inquiring into the Reshaping of Financial Markets)

Reporting period: 2018-11-01 to 2020-04-30

Financial markets are increasingly dominated by fully automated computer algorithms: whereas in the past, human traders would place orders to buy or sell financial assets on the bustling floors of financial exchanges, today that type of activity is essentially carried out by algorithms. These algorithms are programmed by human beings, but once they are active in markets, they operate independently of humans. The aim of the AlgoFinance research project is to understand the consequences of this transition toward algorithmic trading. The project has three main objectives. First, we are interested in understanding how trading firms are (re)organized in response to algorithmic trading: what types of people do they hire for this type of work? What organizational structures are designed in order to cater for this type of trading? Second, we are interested in understand what happens once algorithms start interacting with other algorithms in markets. It is widely acknowledged that such forms of interaction take place and some argue that they might lead to negative feedback loops and momentary market crashes. However, little is known about the ways in which these interactions occur; the conditions that may trigger crashes; and what might be done to prevent them. The project hopes to be able to address such questions through a novel simulation of financial markets. Doing that will help us reach the third objective which is to examine whether fully automated algorithms engage in behaviors that qualify as ‘social’, i.e. forms of behavior otherwise reserved for human beings.

The project is important for society because it offers new insights into an opaque field of algorithmic trading which is at the center of present-day financial markets, and because the research may help to better understand how we might design markets so that algorithmic market crashes can be avoided.
The work performed so far has mainly focused on three things: (1) conducting interviews with firms specializing in algorithmic trading; (2) setting up a computer simulation of present-day financial markets; and (3) theorizing the interaction of fully automated algorithms.

As per (1), we have conducted more than 170 interviews, mainly with firms specializing in algorithmic trading. The majority of these firms are from Chicago, New York, London or Amsterdam, all central hubs in financial markets. In addition to trading firms, we have interviewed exchange officials, regulators and non-algorithmic firms such as institutional investors who might be adversely affected by algorithmic trading. In addition to interviews, we have been granted access to follow two trading firms ethnographically, one from Chicago and one from London. The interviews and the fieldwork have given us a unique view into the world of algorithmic finance. We are currently analyzing the data and hope to be able to publish findings over the next year.

As per (2), we have developed the contours of our simulation framework and its so-called agent-based model. Developing a realistic simulation of present-day financial markets is no trivial feat, given the complexity of these markets. We have a simple model in place now and hope to add further complexity to it over the next years.

As per (3), the PI has completed a monograph (forthcoming with Cambridge University Press), in which he theorizes the interaction of trading algorithms, arguing for conceiving of such interaction dynamics on the basis of a vocabulary of “social avalanching” that is derived from late-nineteenth-century reflections on crowd behavior.
The project hopes to present progress beyond the state-of-the-art on several dimensions. First, we expect to provide novel insights into the ways in which algorithmic trading firms are organized (on an intra-firm level). In particular, our ethnographic fieldwork in two trading firms is unique, with one firm allowing us to examine the use of sophisticated AI-based trading.

Second, our simulation framework is truly beyond the state-of-the-art: it is an advanced agent-based model that contains dimensions of algorithmic trading never included in academic simulations of financial markets before. We expect to be able to explore an array of important features (network effects, market design effects, market crashes, etc.) once the framework is fully developed.

Third, the project offers a novel theorization of the interactions of algorithms. We have already presented novel theorization on this topic during the project’s first two years and expect to be able to add new significant findings on the basis of some of the simulation work.