What Is an Algorithm? Representational Uncertainty in the German High-Frequency Trading Act

A common theme of literature on financial regulation in the era of financial automation is representation and the limits thereof. Scholars have associated these representational challenges with the overwhelming quantity and insufficient quality of the data trade surveillance officers in financial exchanges are charged with making sense of. Since the rise of algorithmic and high-frequency trading in mid to late 2000s, further challenges related to knowing what algorithms are doing and with interpreting algorithms’ strategies have led sociologists of finance to depict an unfolding crisis of representation and to draw bleak conclusions about the prospects for regulation.

This paper builds upon but comes to different conclusions to this literature by presenting the findings of an empirical study of the implementation of the algorithm tagging rule in the 2013 German High-Frequency Trading Act. The first such regulatory requirement of its kind, the rule requires trading firms to identify in the form a numerical code which algorithm was used to generate a trading decision. The intention was to allow trade surveillance to see the operation of individual algorithms in an exchange’s order-book and to subject them to the sort of scrutiny previously reserved for human traders. In order to implement the rule, however, regulators would first have to define what an algorithm is and what constitutes a ‘material change’ in an algorithm – questions for which no off-the-shelf answers were available.

Based on fifteen interviews conducted in 2014 and 2015 with persons involved in the Act's implementation, this paper elaborates the process by which regulators approached these definitional problems and explains why it resulted in trading firms being required to tag a ‘regulatory algorithm’ different to the algorithms firms themselves identify. The paper then evaluates the consequences of such representational uncertainty and asks what it means for the governability of automated financial markets.