The case for a forward-looking approach in savings modeling

The case for a forward-looking approach in savings modeling

A key challenge for risk modelers is to assess whether the available historical data is representative for the (near) future. This determines to a large extent whether a more backward-looking or a forward-looking approach is needed.

Recent economic developments have resulted in an increased focus on this challenge when it comes to modeling non-maturing deposits. In this article we explore how risk modelers can find a balance between relying on historical data and forward-looking scenarios when designing their savings models.

Over the past years, modeling non-maturing deposits has become increasingly complex, due to the combination of various unprecedented market circumstances, each with its own challenges. Additional regulatory reporting requirements, such as the Supervisory Outlier Test for Net Interest Income (NII), further emphasize the relevance of accurate and robust modeling techniques for the associated client behavior.

In search for an appropriate model, banks traditionally turned to historical data, given the objectivity associated with this approach. A more subjective, expert-based, approach requires substantiation to the regulator, while risking an extensive, time-consuming discussion.

Savings modeling in economic uncertainty

However, completely relying on historical data has proven to have certain limitations. The last 30 years only provides insight into client rate dynamics when interest rates are decreasing. On top of that, forecasting power of data for recent years is compromised by the presence of a ‘soft’ floor on consumer accounts. Banks are hesitant to charge negative savings rates, due to a possible negative impact on their reputation.

The last months show some first signs of increasing interest rates. This raises questions regarding the representativeness of the available historical data to predict future behavior. Given the limited information on client rate dynamics when interest rates rise, it is unknown whether the market rate tracking of savings rates is comparable to historically observed behavior. Moreover, it is uncertain how consumers will react to increasing market rates and as a consequence, when banks will be forced to offer positive savings rates once again. As a result, it is not evident that historical behavior will still be representative of future behavior when market rates revert to higher levels.

While the current interest rate environment primarily complicates client rate modeling, it also affects modeling volume outflow. With low savings rates, customers have little motivation to move their money from current to savings accounts, resulting in so-called ‘hidden savings’. It is, however, likely that this situation will be reverted once savings rates start rising, affecting the outflow profile of both product types.

Incorporating forward looking scenarios

Given the downside of using historical data, banks are increasingly incorporating expert judgment in model calibration, to better align model predictions with business expectations. By applying forward-looking scenarios in the calibration process, the expectations of business experts can be explicitly included in the model estimations.

Forward-looking scenarios usually represent the expected pricing strategy in various scenarios, combined with the expected effect on outflow behavior, as provided by pricing experts. By drafting these expectations in a wide range of interest rate scenarios, the estimation process covers business expectations under various circumstances. This also allows for considering interest rate scenarios that did not yet occur in history, such that the full range of potential future market developments is included.

The main challenge in defining accurate business expectations lies in the interaction between the expected pricing strategy and the volume developments. For instance, the repricing speed of the savings rate in response to changing market rates likely affects the substitution effect, where clients switch to alternative products. Similarly, when evaluating the pricing strategy in relation to the (soft) floor, the expected strategy of competitors should also be considered. Therefore, when discussing the expected pricing development in a specific scenario, the volume targets of the bank should already be considered.

Convincing the regulator

The subjective character of incorporating forward-looking scenarios tends to complicate the discussions with the regulator, requiring more extensive substantiation. By following three steps, the risks of a time-consuming discussion can be partially mitigated:

  1. Demonstrate the necessity of including business expectations in the calibration (i.e., reasoning why using only historical data is not deemed sufficient).
  2. Consult experts from various departments when drafting the expected behavior in the forward-looking scenarios, aiming to reduce bias and increase ‘objectivity’ of the expectations. In this process, it is crucial to have the Risk and ALM department critically challenge the business expectations. Backtesting earlier business expectations, to the extent possible, could provide valuable insights for this challenge.
  3. Draft a sound governance on the use of forward-looking scenarios, outlining the process and involved stakeholders, and ensure proper documentation on the definition of business expectations as well as the resulting decision-making process.

These challenges will be on the top of the risk modeling agenda. While there is no ‘right’ answer, making a substantiated assessment will be key to maintain a high quality of savings models.