Getting What You Expect?

Market interest rates have been sitting at or near record low levels for almost five years. As a result, credit unions are booking assets at very low rates and, in many cases, lengthening their balance sheets to slow the decline in yield. The resulting added level of interest rate risk undertaken by some institutions has not gone unnoticed from a regulatory perspective.

Indeed, the National Credit Union Administration and state regulators have become increasingly concerned about the composition of credit union balance sheets. Interest rate risk is the most significant risk the industry faces right now, according to NCUA’s Letter to Credit Unions 14-CU-02.

Higher levels of interest rate risk, along with increased focus, put more pressure on understanding ALM model methodologies. From a business perspective, it makes sense for credit unions to be especially focused on their asset/liability management position and their understanding of risk—as well as the assumptions on which ALM modeling is based.

Certain key assumptions can greatly impact the results of some traditional methodologies.

In 2012, NCUA implemented its Interest Rate Risk Policy and Program, which spelled out the requirements for a credit union’s interest rate risk program. Key requirements in this directive deal with the assumptions used in simulations. The regulation states that assumptions should be reasonable, supportable, understood by management and the asset/liability committee, and updated as appropriate.

Can You See the Assumptions?

The Interest Rate Risk Policy and Program states that there needs to be “transparency of changes in assumptions, methods and IRR tests.” Based on our experience performing ALM modeling validations, we have discovered that some products and processes make this fairly easy, and others make it nearly impossible to understand what is happening in the model. We recommend that you look at your ALM model or discuss each of the key assumptions with your ALM provider to ensure you can see each of the key assumptions in the model.

If you cannot see the key assumptions, then you don’t know what assumptions are being used. How, then, would you know when they have changed or if they are reasonable?

Are the Right Assumptions Synced?

The key questions here are:

• When should assumptions used in net economic value and net interest income sync up?
• Are there circumstances in which it makes sense for these assumptions to be different?

First let’s consider two assumptions that should sync up when calculating NEV and NII: deposit pricing and loan/investment prepayments.

Deposit pricing assumptions outline how the credit union believes it will price deposits if rates increase or decrease. This assumption can reasonably be based on prior experience, as data is available for rate increases as high as +500 basis points (which last occurred in 2007).

Credit unions should minimally use his-tory as a guide for the pricing of deposits in NEV and NII simulations. Since there is no precedent for how members might behave when rates increase from such a prolonged low environment, credit unions may want to use even more conservative assumptions.

Loan and investment prepayments are another set of assumptions that should sync up between NEV and income simulations. Prepayment assumptions are critical to the process of discounting cash flows. These assumptions impact income simulations by influencing how quickly existing business will pay down.

When possible, credit unions should use observable data as a starting point to adjust and refine prepayment assumptions used in NEV and NII simulations. Using actual experience may not always be reasonable. For example, the level of mortgage prepayments seen over the last couple of years may be much higher than future expectations.

Some ALM providers will use separate models to calculate NEV and NII. If this is the case for your credit union, check to see if the prepayment speeds used for calculating value are the same as the prepayment speeds used in your income simulation. If there are two different assumptions for an account, it guarantees that, at minimum, one of them is wrong.

In performing model validations, we have seen many cases where the prepayment speeds are not the same when comparing methodologies. Credit union management often did not realize this was happening. If you find yourself in this situation, ask yourself, “How would I reasonably explain that the prepayment speeds for my 30-year mortgages are assumed to be one set of numbers in the income simulation and a different set of numbers for calculating the value?”

There is no way to know what will happen, which is why testing various assumptions in “what-ifs” and stress tests is prudent. Having conflicting assumptions in a base case does not help the objective of testing various assumptions.

Additionally, in some cases, deposit values may come out of a different model. Similar to the issues described for the loans, if the deposit values are calculated from an alternative model, are the assumptions between the two different sources in line?

For example, both earnings and NEV simulations typically need deposit pricing assumptions. If the deposit pricing assumptions used for NII are different from those used for NEV, this should raise a flag.

If there are two different sets of assumptions in a base case, there are challenges beyond the flags described above.

Consider if your process for documenting, understanding and tracking assumption changes is designed to follow two paths for each base case. Clarity in this process can become more difficult in such a situation.

Should Assumptions Ever Be Different?

While deposit pricing and prepayment assumptions should sync up, it sometimes makes sense to use different assumptions between the NEV and income simulations in the case of loan discount assumptions. Credit unions will often use offering rates to represent the rate of new business in an income simulation, and then those same offering rates as the discount rates in the NEV analysis. Note: This approach can result in older loans being modeled at sizable gains—more on this later.

While using offering rates may make sense in an income simulation, it does not necessarily mean it is reasonable to use those same offering rates to discount loans in an NEV simulation. To address this, it can be helpful to review NCUA’s definition of NEV (see box, below ).

According to 12CFR NCUA, Section 703.2, fair value is “the amount at which an instrument could be exchanged in a current, arms-length transaction between willing parties.” Ask yourself, are there likely to be willing parties to buy a 105 percent loan-to-value mortgage made in 2007 at a premium today?

In performing simulations of risks to earnings and net worth, we encourage credit unions to stratify mortgages by coupon rate (to allow for refinement of prepayments, discount rates and to aide in tracking balances). In this process, it is not uncommon to see sizable pools of mortgage loans held at coupon rates above 5 percent.

In the table below, 26 percent of the 30-year fixed mortgages are still paying over 5 percent. We see some institutions that have more, some that have less. It is important for each credit union to understand its mix and how that might inform relevant modeling assumptions.

Note there are many different reasons members hold mortgages at higher rates. In some cases, refinancing is just not possible. The collateral may not be adequate, credit scores may have deteriorated, or employment cannot be verified.

Beyond credit issues, the loan documentation and underwriting used years ago may not be up to current standards. In addition, as loans age, third parties are less willing to purchase these seasoned loans (for many of the same reasons listed above). Even in the relatively liquid mortgage market, it is more difficult to sell older loans.

If a credit union uses current offering rates to discount these loans, higher coupon mortgages would be reflected at material gains. Considering the likely reasons the loan is still at a high coupon, modeling significant gains on these loans may not be reasonable.

Relevant factors that could impact market value should be evaluated when setting discount rates. Even the cleanest portfolio will have some loans that would be virtually impossible to sell without incurring a significant loss. One way to address this is to separate loans by coupon or current loan to value and adjust the discount spread to recognize this reality. Discounting to offering rates is common, but that does not mean it is reasonable.

Other Considerations

With all the data and assumptions required to produce NEV and NII results, it can be easy to lose track of which assumptions impact both sets of analyses and which ones do not. Some differences are more obvious, such as the fact that discount rates do not impact NII. However, some differences are more subtle. For example, is your NII simulation incorporating the risk of deposits leaving?

A key part of asset/liability management is addressing risks to liabilities. Asset/liability committees and boards should ensure that risk exposures due to liabilities, in this case non-maturity deposits, are being appropriately addressed.

Therefore, consider the following question, “When we are looking at the ALM results for policy, are we factoring in the risk of non-maturity deposits leaving?” How would you feel about answering that question with, “No, deposits never leave (static balance sheet NII simulation),” and “Yes, deposits are assumed to leave (NEV with decay).”

Note that the difference between answers is even more dramatic if NEV is modeled as shares at par, where conceptually all non-maturity deposits leave immediately.

Given the confusion in the example answer above, what are the chances of decision makers being confused and losing confidence in the ALM process? One of the indicators of potential modeling risk is if changes in your assumptions have a dramatic impact on the results. This should cause you to question, “How reliable are the results upon which you are basing your decisions?

If your credit union is utilizing a static balance sheet to simulate NII, by definition, the balance of each account never changes. Test and see. Is your income simulation including any decrease in non-maturity deposit balances if rates rise? Look over the length of your simulation and see if the balance of non-maturity deposits is the same when comparing the current and higher-rate environments.

As rates rise or investment alternatives change, there is a strong likelihood (and recent historical precedence) that member deposits will shift from lower-rate products (e.g., shares and checking) into higher-rate products, or shift out of the credit union into an alternative type of investment. (See also http://tinyurl.com/cmyersarticle).

Consider taking a look at your cost of funds in 2007. If you simulated a +500 basis point rate environment today (which would take short-term rates back to 2007 levels), does the income simulation show cost of funds minimally getting back to where it was in 2007? If not, you could be understating your risk.

Credit unions’ income simulations should address this issue. Once your base income simulation’s cost of funds minimally reflects history, it would also be a good idea to run additional stress tests. There could be a lot of pent up demand for yield if rates rise, and members may be more rate sensitive than they have demonstrated in the past.

What about the Bottom Line?

Finally, note that neither NEV nor NII simulations take into account net operating expense. These modeling methodologies do not help decision makers understand if the credit union will make or lose money. This can cause decision makers to not understand threats to net worth.

After more than two decades of working with credit union managements and boards on ALM, forecasting, and strategic planning, we have found most decision makers want to know if they will be positioned to make or lose money. And, if there is risk of negative earnings, does the credit union have enough net worth to absorb losses and remain safe and sound? Understanding this risk could help decision makers to determine what steps could be taken to reduce the risk of loss.

If you are relying on NEV and NII simulations to measure risk, recognize the limitations from a decision-making perspective.

If you only remember two things from this article, remember:

1) The heat is on—it’s more important than ever to understand the assumptions that go into your risk simulations.

2) If you have conflicting assumptions, decide if they make sense and, if not, make it a priority to change the conflict!

Since 1991, c. myers corporation has partnered exclusively with credit unions and has worked with about 25 percent of the CUs over $100 million in assets and 50 percent of those over $1 billion providing strategic planning, process improvement, project management, ALM, and budgeting services. The company’s philosophy is to help clients ask the right, and often tough, questions to create a solid foundation that links strategy and desired financial performance. Reach a principal at 800.238.7475.