A recent survey by the ABA finds nearly 1 in 5 banks outsource their ALCO and other risk management modeling/reporting. We know lots of bankers that enjoy running models every month and we know others who despise it. No matter which group you are in, there are many things to consider.
To most, outsourcing is a personal decision where bankers must ask themselves, "do I want to commit the time and resources to do this task in house?" If the answer is yes, other steps must follow. Today, we focus our discussion on the group of bankers that likes to run their own models.
Before we go down that path, however, it is important to understand the "why" of bank modeling. Specifically, models can be used for many purposes. Bankers rely on models because they allow us to estimate how risks will change in the future, based on a set of assumptions. The analysis is generated from the bank's own data set (which can have its own issues) and reports are then provided to the management and board of directors. In short, bankers use models as a way to describe in mathematical, accounting or other terms; how their bank's risk profile would look (under a series of conditions), projected into the future. What can we say - bankers love to look at numbers.
It is important to review how the regulatory community also views modeling within the industry. No longer can a quantitatively-minded person within the bank simply wander from their cubicle (after running a model only they understand), carrying reams of reports, espouse the risk profile of the organization and get immediate management buy-in. Managers have learned that regulators won't accept a high-level review anymore because errors are common and modeling complexity has increased. To make more informed decisions about the level of risk the institution is willing to assume, management and the board must be able to rely on the information they are presented (i.e. demonstrate an understanding). The entire modeling process has become more complex because banking has become more complex.
Examiners roll into banks with an already heightened focus on results, because they know management and the board rely on these results to make strategic decisions that can radically affect each of the CAMELS components. Examiners want to make sure everyone in the organization understands not only the output of the model, but also the model's limitations. Specifically, regulators put a microscope on what actions bank management takes based on the results of the model and how those models are integrated into the strategic planning process of the organization.
Given this backdrop, bankers that run internal models have multiple steps they must take. These include employing robust due diligence in developing the model, regular and independent validation (of not only the model, but also its logical and conceptual soundness) and back-testing (to ensure the results being presented to management and the board make sense).
As if that weren't enough to consider, banks running their own models must enhance their auditing processes as well. Audit teams must not only make sure that appropriate controls are in place, but also that the person or team doing the modeling is following the policies and procedures. They must verify the data being used is valid and that it has come directly from core/source systems. Finally, audit teams should review the models, assumptions, results and any independent review materials. When completed, audit should then report back to the board of directors on the effectiveness and accuracy of the model (as well as compliance with policies and procedures). Only then can management and the board feel good about making decisions based on the reports they have received.
Running models internally that have been purchased from software vendors provide flexibility in processing frequency, but they also demand robust controls. Banks pursuing this path must not only be prepared to commit resources to the process, but be capable of evaluating the validity of results generated and be able to translate the results into actionable items.