Click to Login

Deposit Modeling

Wed Aug 29, 2012

In our last post, we discussed the issues around why a bank might want to use a simulation system rather than closed-form equations or simple algorithms to drive their internal thought exercises, and satisfy examiners. To recap, while a formula-driven approach may provide a "first order" estimate of risk or other factors, simulation is the only way to capture real-world phenomena such as prepayment estimates (whether they are cyclic, provided by a third-party, or otherwise generated), instrument interdependencies (e.g., overdraft protection, which can be a significant source of fee-based revenue), and defaults.

Part and parcel of a simulation based approach, however, is the need to have a firm handle on depositor liquidity based on both a quantitative analysis of each depositor's past behavior, and an understanding of their business model within the context of the bank's overall deposits. In other words, insofar as a single depositor or a group of "connected" (correlated) depositors can "move the needle" of the bank's funds, the bank will need to understand how that depositor or group of depositors behaves.

Since our Cashflow Modeling Engine actually runs a day-by-day simulation of each account or record at the bank -- both assets and liabilities -- we can uniquely tell a bank what the expected behavior of a single depositor may be under different circumstances. Before running a simulation, we collect all of the past history for an account and build several different models for that account's balance based on (a) short term (6 months) or (b) long term (3 years) histories. In both cases, we then build a "straight (trend) line" and a more exotic distribution to fit that account's balance history in order to drive our simulations.

As a result, we can work with a client to either look at their depositor balances on a simple, intuitive basis, or a more quantitative basis. What is the benefit of this "either/or" approach? While the intuitive approach may yield an "extremely likely" result with less computation, the distributions that we build for each account allows us to operate a fully-functional and correctly implemented stochastic simulation that can be run several times until confidence intervals for the bank's future performance can be identified. Without performing this type of statistically valid analysis on each account, a bank cannot fully appreciate how quickly its position may be impacted in times that the financial system may be stressed. Capitalytics is proud to be the first service provider to make this level of power and analysis available to the community banker in a cost-effective manner.

If you and your bank would like to benefit from this type of service, contact Capitalytics today. We would be glad to work with you to understand the risks and the strengths of your institution.