Get with the Program: Fintech Meets Regtech in the Light-Touch Sandbox
Posted: 31 Mar 2017
Date Written: March 31, 2017
Previous papers at TPRC and elsewhere have contributed to our understanding of the significance of automated, data-driven decision systems, especially when used for regulation. The objective of this research is to analyse two specific and interrelated developments in order to ascertain the extent to which they call for the development of new types of model and new approaches to algorithmic regulation. The first of these is ‘fintech’ - in particular the consolidation and transfer of investment decisions from human to algorithmic control and the migration of financial intermediation, payment services and data curation beyond traditionally regulated financial entities. The second is regtech – automated and real-time provision of compliance-related information to regulators. Both pose challenges: practical, legal and ethical difficulties in auditing and assurance; moving beyond individual reporting entities to systems as units of analysis and control; maintaining the efficiency and efficacy of contracts, markets and regulatory obligations; and ensuring that the resulting system is as low-cost and flexible as possible (but not more so). These difficulties are not just pragmatic; recent work on data ethics highlights the difficulty of detecting and correcting incorrect decisions (let alone dealing with their indirect and systemic impacts) and the inevitable fragmentation of responsibility among those officially responsible, technology players, programmers, data sources, etc. The research draws on a set of cases to include (at least) digital wealth managers and automated spectrum use reporting and trading. The ‘red thread’ is the extent to which algorithms can usefully be considered as more than just faster and more consistent was of implementing decision rules. Four alternatives are considered: i) the impact of interactions among algorithms and the persistence and manipulability of the ‘histories’ on which they operate (in other words, the difference between source code and running instances); ii) application of principal-agent analysis to entities that are imperfectly auditable, act on the basis of imperfectly observable information and/or can only be controlled be rules that operate at the ‘speed of information’ (slower or faster than the algorithm itself); iii) modelling algorithms as links in a network (connecting people, data, and institutions) using the tools of network game theory; and iv) an evolutionary approach which views algorithm types as ‘species’ whose populations interact in a market ecosystem. The first three alternatives apply existing theoretical methods to stylised versions of the cases; the fourth uses simulations based on exemplary decision rules (moving averages, CAPM/OPM, Kalman filters and Bollinger band strategies) and different regulatory options. An interesting outcome is the possibility that continuous reporting can enable new types of dynamic regulation that are less burdensome and more effective than today’s static alternatives.
Keywords: fintech, regtech, complexity, algorithms, regulation, modelling simulation, platforms
Suggested Citation: Suggested Citation