5 key steps to improve ALM model assumptions

By Joe Kennerson, managing director, Darling Consulting Group

Changing economic conditions create uncertainty. The current rate tightening cycle—first time in over a decade—is creating high levels of multiple forms of uncertainty for bank executives:

• Uncertainty on the direction of funding costs. I recently conducted a webinar and polled over 100 bankers on interest rates. Over half of the respondents said they have increased MMDA rates or introduced a CD special so far this year. Funding cost pressures are real.

• Uncertainty on how margins will be impacted if the yield curve continues to flatten as rates rise. What happens if funding costs continue to rise, but loan yields do not?

• Uncertainty on future liquidity levels. Loan growth has been outpacing deposit growth in the community banking sector since 2012. That trend has started to create concern for bankers.

And finally …

• Uncertainty on how risk models are capturing these potential risks.

ALCO and board members depend on accurate interest rate risk (IRR) models to drive balance sheet decisions. One critical question most bankers face: “What is my capacity to extend assets?” The answer depends on the model results and the assumptions that drive those results.

While it is critical to have confidence in your IRR model to drive balance sheet decisions, it is equally important to be able to defend and articulate model results to your examiners.

Regulators have always focused on the development of assumptions by ALCOs. Regulatory feedback today stresses that, where feasible, bank-specific assumptions be implemented into all risk models.

The focus of this article is on how to develop best-in-class model assumptions for all risk models, specifically:

• Deposit stability

• Deposit pricing betas

• Deposit average lives

• Loan prepayments

• Loan pricing spreads

Here are 5 assumption-related initiatives to improve your ALCO process.

1. Bank-specific assumptions

Industry averages just don’t cut it for assumptions, especially for community banks. Every community bank market is different. So assumptions should match.

Deposits

For deposits, generating support for key rate beta and average life/decay assumptions usually means conducting a study. However, a historical analysis that looks at seven years of deposit trends in the 0-25 basis point Fed funds range can only tell you so much about the sensitivity factors for your customer base.

That’s why you must combine a quantitative analysis with a qualitative approach. This idea comes straight from FDIC:

“Bank management may want to explore qualitative adjustments for some assumptions. Qualitative adjustments are applied to historically based analysis to account for unique bank-specific or environmental characteristics”—FDIC Supervisory Insights Winter 2014 

Examples of qualitative factors that can affect deposit model assumptions include the following:

• Analyzing the change in deposit mix—Most institutions have seen non-maturity deposits increase and time deposits decrease as a percentage of deposits.

• Customer relationship status—How many other accounts do our MMDA customers have with us?

• Change in average balance size for non-maturity deposits— Most banks have seen this trend upward since 2009.

• Generational factors—How does rate sensitivity for millennials compare to rate sensitivity for baby boomers?

Community bankers know their customer base better than everyone else. Use qualitative judgment factors as well as data (see next bullet point) to refine the models’ rate betas.

While net interest income simulations drive strategy discussions, Economic Value of Equity (EVE) simulations are also important.

Virtually every bank has policies surrounding EVE and examiners evaluate this risk sensitivity as well. The most significant assumptions driving EVE sensitivity are the average lives/decay rates of non-maturity deposits.

It is critical not to take an overly conservative approach in developing the average lives for non-maturity deposits (NMDs). (One such assumption is that if the average life is unknown, assume a short average life.) Taking such conservative approaches can create undue sensitivity. The result may be that just a subtle move in interest rates can have a large impact on the EVE position.

As evidence, look back to the fourth quarter of 2016 when the bond market sold off. This sent rates higher and steepened the yield curve. Some banks that utilized average life assumptions that were too short started to push or exceed EVE policy limits in their rising-rate scenarios. This creates unwarranted risk that may drive inappropriate balance sheet decisions.

Loan prepayments

Prepayment risk for a bank will depend on the makeup of the loan portfolio. Mortgage banks may rely on secondary market data for prepayment assumptions. This may be fine in most instances. However, the prepayment risk for a bank in Boston may be much higher than a bank in Montana.

Commercial banks are more challenged in developing prepayment speeds for commercial real estate (CRE) portfolios. Some banks will not apply prepayment assumptions to CRE. However, our prepayment analyses have shown accelerated prepayment activity through most of this decade when rates were low.

The point? It is worth analyzing bank-specific prepayment speeds. The byproduct of this analysis can help identify the true average lives of individual portfolios which may defend the strategy of holding more fixed-rate loans.

Take the analysis to the next level by breaking down the portfolio by size (jumbo vs. conforming), customer age, vintage, and location.

Loan pricing

Look to the past and discuss the future.

At DCG, we analyze all loan originations over the past 90 days with our clients to back-test model assumptions with actual results. We then invite senior lenders to discuss pricing as part of our conversations about the bank’s ALM assumptions.

A recent assumptions discussion with a client revealed that recent floating rate commercial loan originations did not have rate floors. This reignited a strategic discussion with the lending committee on the importance of implementing floors, especially now that prime has lifted 100 basis points off the bottom.

2. Data

I attended many balance sheet management conferences in the second quarter. One common theme throughout all the conferences? Data. In our universe, banks possess a treasure trove of data.

Where to start? Go beyond the basic requirements for a deposit study. Add branch; customer age; relationship attributes (e.g. direct deposit flag) to system downloads for a greater understanding of the full relationship of your customers.

Given the previous discussion on how non-maturity deposit average balances have increased since the beginning of 2009, is it appropriate to utilize the pricing betas from the 2004-2006 rate environment? By dissecting data, banks can evaluate where the growth came from and how “core” the relationships are.

One example is the rewards checking product. Rewards checking services gained significant traction at the height of the last rate cycle. Since then, balances have continued to grow. However, there is no data supporting deposit assumptions for most rewards checking products in a rising rate scenario.

For one DCG client, we used qualitative support for the beta assumption by identifying which customers qualified for the high rate versus non-qualifying customers. Then, we identified the percentage of customers that utilize both direct deposit and automatic bill pay services within this product.

The analysis gave us greater clarity in defending how “core” the funds were to all of the stakeholders of the bank—and ultimately reduced the beta in the model from what was previously established.

For loan prepayments, start to include such data segments as branch and loan officer in your loan file. We have also segmented the database by coupon band and vintage to identify potential credits that are at risk of refinancing.

The next level of refinement would be to capture data at the relationship level. DCG’s Deposits360° tool tracks customer-level data flows to capture deposit cannibalization and the marginal cost of funds of promotional products.

Data collection takes time. However, it will yield strong results in the future. As the Chinese proverb says:

“The best time to plant a tree was 20 years ago. The second best time is now.”

3. Model integration

Integrated risk management is an integral component of banking today, and all potential balance sheet strategies should be vetted through IRR, liquidity, and capital risk analysis. Additionally, assumptions from deposit and loan analyses should be incorporated into all risk models.

The results of a recent DCG deposit study were incorporated into the following ALM risk models:

• Decay rates utilized in EVE

• Betas utilized in NII/EVE

• Non-core in liquidity analyses

• Surge in liquidity and capital stress tests

The integration of modeling assumptions not only makes life easier, but the data continuity goes a long way with examiners.

4. Stress testing

As much work as we put into developing the assumptions, they will be wrong. We know that. Stress testing allows bankers to understand the earnings impact if the assumptions are off more than expected.

Stress tests should be focused on the material assumptions that drive results. This will depend upon a bank’s product mix, business model, and risk position.

5. Document & monitor

Bankers are doing more with less today. Documentation may not seem productive, but it is a necessary evil. If not done effectively, banks could experience pushback from examiners or auditors.

Document how your assumptions were developed; what methodologies were utilized; how employees were involved in developing the assumptions; and how frequently they are tested.

Assumptions should not be on the “set it and forget it” mode. DCG conducts quarterly assumptions calls with ALCOs to review all material assumptions within the model (new volume pricing, deposit sensitivity and prepayments). The monitoring of assumptions will be even more critical as we move through this rising rate environment and deposit sensitivities change.

Removing uncertainty

Banks are navigating through a very interesting time for the industry. There is uncertainty on the future of funding costs, the yield curve, margins, and liquidity levels. Reduce the level of uncertainty in your risk models by taking the best steps in developing your model assumptions.

Examiners are reviewing banks’ assumptions and how they are developed very closely. Failure to take this process seriously can lead to headaches during your next exam or worse, have a negative effect on your future earnings potential.

About the author

Joe Kennerson is a Managing Director at Darling Consulting Group, working directly with financial institution executives to improve the effectiveness of their asset-liability management (ALM) process. In this capacity, he provides tailored solutions for managing interest rate risk, liquidity risk, and capital. Kennerson is also expert in the overall balance sheet management process, including risk model design and implementation, regulatory compliance, and executive-level education.