Non-Maturity Deposit (NMD) Modeling

Depository institutions are prone to believe that their deposits have above-average value, that this value is intrinsic and impervious to rising interest rates, above-market asset growth rates, and desperate competitors. As a result, they are not prepared for the adverse impact rising interest rates have on earnings and economic value.

Despite the lack of contractual cash flows for NMDs, depository institutions must establish meaningful behavioral assumptions in order to quantify their exposure to interest rate risk (IRR) and liquidity risk (LR); the computation of product profitability measures also requires behavioral assumptions. Poorly founded assumptions, in either case, will necessarily lead to arbitrary, inaccurate, and inconsistent measures of risk and profitability.

DGA has developed a unique and comprehensive approach to modeling NMDs which is specifically designed to align the behavioral assumptions used in a variety of risk, profitability, and product management processes. Only by aligning these processes can an organization establish measures of risk and profitability which are robust through the entirety of the business cycle.

The Problem

For any depository institution, an internal models or third-party study is required to estimate repricing and liquidity cash flows such that subsequent measures of duration and weight-average lives (WAL) can be loaded into the ALM model which calculates IRR and LR. Similarly, cash flow estimates are also required to calculate FTP rates on deposits, which are necessary to produce margin estimates for budgeting and forecasting exercises.

It is not unusual for the assumptions around NMD behaviors supporting these risk and profitability measurement exercises to be developed and managed by separate functional groups within the firm, with each using different systems from different vendors. Not only are the behavioral assumptions for NMDs not aligned, but there are also material impediments in place that prevent a meaningful reconciliation of the underlying behavioral assumptions. This almost certainly means that the strategic recommendations emanating from these two areas are in conflict with one another.

In general, ALM models are specifically designed to model repricing and liquidity cash flows at a very granular level. While this is helpful, many ALM models have not been designed to calculate FTP rates, either for the current position or pro forma periods. Even when they may have this capability, many depositories rely on FTP calculations performed in profitability management systems that may not utilize or require granular or dynamic estimates of repricing and liquidity cash flows; they are limited to simplifying assumptions which are inevitably inconsistent with the repricing and cash flows used to estimate product duration and WAL.

Why Should This Be a Concern?

If the way in which deposits are managed by the business units (which should follow logically from their views on the absolute and relative value of each product) produces behaviors that are not consistent with the assumptions contained within the ALM model, then measures of firm-level IRR and LR will not be correct. Worse yet, if the actual behavior of the deposits is not correctly reflected in the FTP methodologies, then the margin projections used in the budget and forecasts will be arbitrary and misleading. Woe be to the product manager whose compensation or bonus is tied to such projections.

When interest rates are increasing (and depositors are more cognizant of deposit product rates) and excess liquidity has been consumed (because depository institutions are growing their loans faster than the market is providing deposits), these disconnects can quickly lead to material miscalculations of risk and profitability which will lead to poorly-informed decisions around balance sheet management. If this sounds like the experience at a number of depositories in 2018/2019, it shouldn't come as a surprise; in fact, their frustration and disappointment has been inevitable.

The Solution: The DGA NMD Model

DGA has designed an NMD Model which simultaneously addresses the needs of risk AND profitability management. The NMD Model produces empirically-based estimates of rate betas, balance decay functions, and balance volatility functions to produce repricing and liquidity cash flows. The NMD Model also contains its own FTP engine to compute FTP rates, which leverage these same cash flow estimates.

Any changes to the behavioral calibrations are reflected simultaneously in adjustments to the cash flow output used in the ALM model AND the FTP rates used in profitability management; in fact, these can never be out of sync. The NMD Model is also designed to be operated monthly in order to back-test actual behaviors to expectations because changes in behavior can occur at any time (especially when interest rates are increasing and competitive market dynamics are shifting). When behavioral variances are identified, a re-calibration of the behavioral assumptions may be warranted.

Let's Walk Through an Example

Assume that the DGA NMD Model has been used to feed the ALM model AND the budget with a consistent set of behavioral cash flow assumptions. If deposit gatherers begin to increase deposit rates faster than had been anticipated, actual FTP spreads (FTP rate minus the customer rate) will narrow. Because the FTP spread will have tightened relative to the budget, Finance and the business units will certainly take notice (as should the ALM/FTP manager). If the sensitivity of the deposit rates cannot be re-aligned with the original expectations, the assumed behaviors in the NMD Model will need to be re-calibrated. A lower deposit duration will then be computed by the ALM model which will result in an increase in liability-sensitivity and a reduction in the economic value of equity; the re-calibration of the NMD Model will simultaneously lower the current deposit FTP rate, but this FTP rate will subsequently increase at a faster rate when market interest rates increase. The deposit provider will realize a lower FTP spread, but the spread will be more stable (and hence more predictable) when rates change.

If the depository is also computing and managing earnings correctly in the mismatch center, when the FTP rates for NMDs are reduced, mismatch center earnings will increase (or become less negative), indicating that a larger amount of firm earnings are attributed to IRR and LR. ALCO, which is accountable for these earnings, should then be challenged to determine if the change in the risk profile is acceptable; if the increase in the risk profile is unacceptable, then hedging or re-balancing activity may be required. Regardless of their response, the risk, profitability, and product management processes remain aligned because they have all been fed with the complete output from a single NMD model.

Is Such a Model Necessary?

Absent this approach to quantifying and managing the behavior of NMDs, the firm in the example above will still see a reduction in profitability and realize an increase in IRR, but it will be unable to properly attribute it to a discrepancy between expected deposit behaviors and actual deposits behaviors. If this discrepancy is not properly identified in a timely manner, it is highly unlikely that the tactical or strategic response to the earnings reduction will be correct.

NMD Model Design Features and Output

  • Product hierarchy recommendations (for risk, profitability, and performance management)
  • Beta estimates and recommendations
  • Decay and curtailment analytics
  • Dynamic volatility process (stable/non-stable attribution), which stabilizes NMD dollar-duration (and by extension DOE)
  • Economic value of NMDs
  • Effective duration of NMDs
  • ALM model feeds (functional forms or cash flow schedules)
  • Complete time series of FTP rates and spreads (they span the full length of the empirical study, as well as the budget and forecast scenario horizon)
  • Extensive reports of KPIs (concentration risk, average balance, and measures of absolute and relative value)
  • Cannibalization dynamics (to determine if growth is real or if it is just coming from elsewhere on the balance sheet)
  • Real-time product pricing inputs (which can be reconciled to balance sheet strategy and market conditions)

We would also be happy to put you in touch with clients who have adopted the NMD Model and would love to tell you how much of a difference it has made in the interactions between treasury, finance, and the business units.