Financial markets are, by nature, adaptive. This makes them hard to predict, and hard to model. Adaptation is probably most familiar to us in Biological Systems. Shaped by the forces of evolution, the genetic material and encoding of organisms is constantly rearranged in order to optimize their chances of survival in dynamic environments. In Complexity Science, Complex Adaptive Systems change their behavior in response to their environment. Such adaptive changes are often a result of goal attainment (or avoidance) by the entities that comprise these systems.
A flock of birds is a classic Complex Adaptive System. Flocks of Starlings produce mesmerizingly intricate, swooping patterns – so-called murmurations — in the sky, as the individuals in the flock change direction in order to seek out food, or to avoid predators.
This Complex Adaptive Systems paradigm extends to Financial Markets. Each of the entities that makes up the market — for simplicity let’s call them traders — will pursue their own objectives, exhibiting both goal-seeking and fear-avoidance. Whenever a change is introduced into such a system, all of the traders will adjust their behavior in response. Understanding the complex ripple of changes in behavior is key to understanding how an individual’s actions influence such a system.
Executing trades efficiently requires a deep understanding of these types of dynamic. This goes to the heart of transaction cost analysis (TCA) – which buy-side institutions utilize to ensure they receive best execution. When executing trades in financial markets, monitoring transaction costs is essential for practitioners concerned with the bottom line.
TCA can be divided into two types: pre-trade and post-trade. In pre-trade analysis, there is a key trade-off between being in the market for longer (increased market exposure) and reducing time in the market by trading faster (increased market impact). However, market impact analysis has historically made simplifying assumptions about the response of other traders when a trade is placed. In reality, what follows is a set of reactions, and higher order responses to the initial placement of a trade, by the other traders in the market, which generate further orders to be placed, subsequently changing the price – typically against the trader who placed the initial order.
Holding a magnifying glass to the processes (the market microstructure) which gives rise to market impact can help us better model the impact of trades placed into a market. This is where strongly micro-founded modeling approaches, such as Agent Based Modeling (ABM) come into their own. By focusing on the behavioural responses of the set of agents (traders) in the marketplace, these models explicitly capture the changes in agent behavior which ripple through financial markets and produce price changes which impact subsequent orders.
Using ABM, it becomes possible to build generative simulations that produce realistic synthetic market data and gives us a framework for studying adaptation. These approaches are already being used in a commercial setting – prima facie evidence that these models can provide useful information about the potential market impact of orders placed.
