Monitoring liquidity is a burdensome task undertaken at most financial firms. But advances in cloud computing, machine learning and artificial intelligence are making the process increasingly real-time and precise.
The hope is that technology will transform decades’ old and siloed liquidity systems into a single, sleek, fully-automated platform with a strong arsenal of statistical analytics for liquidity monitoring and risk management.
The world’s largest asset manager BlackRock, for example, is in the midst of such a project.
The behemoth fund manager is building an end-to-end liquidity risk management platform that uses machine learning and cloud computing to teach robots how to recognize patterns in data. With machine learning, the robot isn’t so much “thinking” as it is learning by rote. Tech savvy market data vendors like Bloomberg, State Street’s and ICE data business, and index provider MSCI have also incorporated machine learning into the liquid risk products they sell to the buy side.
These firms are using machine learning to better forecast the cost of liquidating fund positions in the case of redemptions—a massive undertaking in itself. In days gone by, vendors and asset managers would analyze bid/ask spreads to calculate the cost of liquidating positions, but new thinking in liquidity risk modelling, suggests liquidation, transaction cost and volume are also crucial factors, which is where machine learning comes in—to wade through these complex patterns and correlations.
But this is just the start. Technology is being used to illuminate many areas of liquidity from managing counterparty exposure, to understanding what forces drive liquidity; what kinds of firms are liquidity providers; and even to model what happens when liquidity dries up and redemptions occur. Other applications include using machine learning to estimate the liquidity of bonds, and filling in gaps for where data on illiquid instruments are not available.
As well as new technology, there’s a treasure trove of new data types that firms are processing to better understand their liquidity. In the past, inferring trends, recognizing behaviors and crucially, spotting anomalies in anything close to real-time has been impossible due to the volume of data being consumed. But all this is changing thanks to the availability of relatively cheap cloud computing, where firms use servers hosted on the Internet to process data, rather than a local server. This combination of machine learning and cloud computing means incredibly data intensive processes can be completed in hours rather than weeks, and minutes rather than days.
Due to the multi-dimensional nature of liquidity, some firms are also dipping their toe into deep learning, a more advanced artificial intelligence capable of identifying complex patterns in data without being taught, akin to the way the human brain operates. However, while the predictions made by these models are reportedly more accurate, they are also “black box” meaning it’s impossible to understand the inner workings.
This is making watchdogs wary. Regulators haven’t spoken publicly but asset managers say they are being asked to defend their use of opaque kinds of artificial intelligence. And so, the march of progress towards a fully-automated liquidity monitoring station continues.