The growing sophistication of behavioral models: a challenge and an opportunity in ALM, liquidity and balance sheet modeling

Defining ALM

ALM is a diverse area (see Figure 1) that addresses the panoply of market risks associated with the banking business. In simple terms, it can be defined as the process of managing cash flows and calculating funding costs to ensure that there is a balance between the asset side of the book, that is, the loans made, and the liability side of the book, which is the deposits taken.

Of course, there are many layers of complexity to this process. According to the nature of the business involved, it may entail lending to wholesale businesses, small and medium enterprises (SMEs), or retail firms – each with different interest rates and market risk profiles. It may also involve certain kinds of business that have moderate or very low market risk, such as cash management or treasury services. Finally, it may involve certain kinds of treasury services, such as access to liquid markets or markets more generally. This can include derivatives trading, which, depending on the strategy chosen, can have very substantial amounts of market risk. On top of this, there are also several regulatory constraints to consider, from the Basel liquidity provisions to the requirements of domestic central banks and regulators.

ALM in practice

ALM can be broken down into two types: strategic ALM and operational ALM. Strategic ALM, which is a proactive, future-focused approach to forecasting the balance sheet, cannot happen without a strong operational ALM foundation. Operational ALM is a combination of the tactics the bank follows on a day-to-day basis: how it charges for products, the different types of interest rate risk it may be subject to, its interest rate risk across various products with different liquidity profiles, and how it handles aspects of its customers’ behavior, particularly around retail deposits and mortgages.

These are all essentially operational activities, so operational ALM starts with a good grip of the products a particular institution offers by business line and jurisdiction and across the banking hierarchy. Banks then develop the detailed and granular cashflows that these products generate, which will then be modified by behavioral rules for different pools.

Recent years have seen significant improvement in operational ALM processes. Systems are now increasingly capable of generating cashflows across a very broad range of loans and other instruments. Cashflow generation and management have also become more sophisticated, granular and capable of creating more rapid and flexible pools against which risk can be analyzed and measured.

Nonetheless, challenges remain. For these systems to be operational, it is important to understand precisely the size, structure and nature of a bank’s balance sheet. However, for various reasons, this may be hard to determine, not least because the optimum structure of the balance sheet changes according to the division using the data. Two challenges stand out in particular, the growing importance of behavioral frameworks in determining customer behavior and the need for strong data management capabilities to service behavioral frameworks and other data requirements.

Behavioral frameworks

Behavioral models are a vital tool in helping banks to understand their current balance sheet and how it will evolve. They help to provide an understanding of clients’ behavior, which is paramount in this volatile environment when there is little prospect of returning to persistently low interest rates. In addition, behavioral models also play a key role in determining liquidity risk exposure (see Figure 2).

Segmenting, profiling and bucketing clients is a central aspect of developing a behavioral framework. Different slices of the market and different customers behave very differently when confronted with similar risks and challenges based on their specific context, including borrowing history, geography, net income and so on.

Developing segmentation models is both an art and a science, involving many questions and tradeoffs. These include: how many segments does the customer pool break down into? And how effective is the segmentation? What is the trade-off involved in breaking up customers into many smaller segments? And should there be one segmentation structure or several?

Much of this involves a set of econometric or other functional mapping frameworks and models, often employing machine learning, that must be continuously adjusted, calibrated and tested against the evolution of customers’ actions.

A fly in the ointment: retail contingent claims

Retail contingent claims represent a particular challenge when it comes to ALM and highlight the importance of strong behavioral modeling capabilities by customer segment. In general, contingent claims are hard to analyze and value because they have many features that are difficult to model, including challenges in mapping to simpler structures, illiquidity and a lack of market price for risk factors.

However, retail-contingent claims have particular and special challenges, not least the additional complexity involved in determining the sensitivity and reaction of the claimholder. Each deposit (or liability) contract is entered into by a retail individual whose decision-making is often driven by a whole range of variables. In particular, retail individuals make decisions that are not purely economic, for example, someone might terminate their deposit because they’ve run out of cash, or they fear the bank itself may be running out of cash, or they want to make an unexpected purchase or pay unexpected bills. As such, financial institutions need to factor in several important noneconomic drivers in the case of retail assets and liabilities with embedded optionality.

Sourcing and managing data

The availability and nature of data used in behavioral frameworks is a further challenge. The availability of data depends heavily on the market ecosystem and formal structural frameworks adopted by financial institutions.

In the US, a significant amount of these termination-dependent assets and liabilities are often pooled by banks and other financial institutions and sold as capital market-driven securities. A capital market record can enable firms to potentially value other comparable assets. This is not a straightforward operation, one issue being the vast number and range of securitized papers, but it can provide some market indicators to factor into models.

In other markets, such as Europe, acquiring external market data is tough. However, successful firms tend to be large, regional banks, with a single institution often serving a broad range of customers. They can, therefore, develop reasonably accurate segmentations and adopt a point of view whereby their customer base at the margins can represent the national market. The main test is how to build, test and continuously calibrate models while also ensuring that the whole model lifecycle of testing, calibration and benchmarking is managed efficiently.

Yet, regardless of whether external or internal data is used, a strong data management infrastructure is a must. Risk systems need to be able to handle large volumes of data, while data quality, data access, integration, storage and movement are all critical success factors. The strongest solutions leverage modern data management technologies, architectures and delivery methods, such as in-memory databases, complex event processing, and cloud technology.

Oracle – an ALM category leader

With a strong operational ALM foundation, banks are much better placed to establish strategic ALM processes. To enable strategic risk management, firms must ensure that the basic operational framework, cashflows and behavioral dynamics are in order. Behavioral dynamics must have the appropriate aggregation and bucketing mechanisms, and firms must be able to calculate interest across a very large range of product profiles, and perform essential market risk calculations in this context. Together with modern data management capabilities, this strong operational foundation facilitates balance sheet optimization, which in turn puts institutions on a much stronger footing in a volatile environment.

Oracle is a leading player in the ALM market. With its ALM Cloud Service offering, it is well-placed to help banks overcome the challenges of a market that has changed considerably since 2021. A true software as a service (SaaS) offering engineered specifically for the cloud, the Oracle ALM Cloud Service is underpinned by the Oracle Financial Services data model. This provides a common repository for capturing the precise nature of customer relationships, which, in turn, drives the modeling assumptions.

It offers advanced analytics capabilities, which enable users to model any product in any currency and to calculate a number of measures, from market value to liquidity levels, in a range of scenarios. It also allows for a flexible time horizon modeling, with users able to configure different time buckets for net interest income (NII), interest rates and liquidity gap reporting. Users can also readily capture behavioral characteristics, such as multi-factor prepayment, early redemption and default assumptions. These are stored independently so that today’s risks are separated from tomorrow’s actions.

Alongside the Oracle ALM Cloud Service, Oracle offers a range of other balance sheet management solutions, including Funds Transfer Pricing, Liquidity Risk Management, and Profitability Management. These all share a common data model and cash flow engine with the Oracle ALM Cloud Service, which means users can focus on data analysis as opposed to data reconciliation between the systems. These also provide users with a full audit trail, so they can understand how numbers have been reached, unlike with a black box approach.

Finally, thanks to the power of Oracle Analytics, users can run balance sheet management reports without the need to perform data transformations. Users simply need to choose to run a process and then can immediately view the results. These reports make use of predefined templates or can be tailored according to an organization’s specific needs.

Sid Dash, Research Director concludes, ‘Oracle ALM Cloud Service provides banks with a wide range of capabilities. It provides best-in-class data management, as exemplified by the launch of its new balance sheet paradigm, Multi-dimensional Balance Sheet Structure (MDBSS), which allows users to create and organize balance sheet structures as per their needs. Behavioral modeling is also catered for, with users able to capture multiple factors into balance sheet simulations. Our research also identified business planning, cash flow projection, integration and reporting as particular strengths of Oracle’s ALM offering.’