Rate repricing will rely on institution-based data

NW_RISK_4_10_19.jpg

Two weeks ago, we considered the general background factors that play a critical role in influencing your rate repricing decisions. This week, we move on to the institution-based data required for your model.

Financial institutions generally will have monthly historical data on deposit pricing and deposit balances readily available. Those data provide an excellent starting point for developing an institution-specific model for deposit repricing and the resulting deposit behavior. But to fully understand what is going on with deposit balances, you should delve deeper into your depository data and consider what additional information it contains that would be critical for analyzing depositor and deposit behavior.

How total balances respond to rate changes is critical for financial institutions making informed decisions, but it represents only part of the story. It is also necessary to analyze how depositor behavior changes. Total balances in general will change when new depositors open accounts and existing depositors change balances or close accounts. Regulators are particularly sensitive to the behavior of existing depositors because in the event of default, regulators would face the responsibility of insuring depositors against some losses. But financial institution should also be concerned about those depositors and their retained balances, independent of the regulators’ needs, because prior behavior of existing depositors likely provides important information about their likely future behavior. And their future behavior and rate sensitivity will impact the institution’s future earnings. Looking at retained balances tells you what your deposit base of customers are doing and how that changes over time; looking only at total balances tells you net what your depositors are doing with their balances and how successful you are at generating additional depositors and balances. If you want to focus on your depositors’ choices, you need to examine their retained balances.

Looking at the behavior of existing depositors means that a complete analysis of rate repricing should include three basic factors – historical deposit rates, total balances and retained balances from a fixed set of accounts at some point in the past. How has your depositor base behaved historically? To answer that question, institutions need to choose a set of accounts at least three years – and preferably more – in the past. The history of the balances in those accounts should then be compiled and will represent a critical part of the data employed to determine the impact of your deposit rate repricing.

Before moving to the modeling process, however, it is critical to consider two examples of how additional factors discussed in the prior section may affect the modeling process. First, accounts that opened at different times may exhibit different behaviors, e.g. deposits in accounts post-financial crisis may differ systematically from accounts in older accounts. For example, they could be more transient or more rate sensitive.

This is the so-called vintage effect with deposits of different ages or vintages behaving differently. If that is the case, analysis only of total balances will not provide a good insight into the behavior of deposits. Suppose older accounts are less rate sensitive than newer accounts. As time passes, with a growing deposit base, newer accounts will make up a greater proportion of deposits and deposit-rate sensitivity will increase. Estimating rate sensitivity without considering the gradual switch in depositor mix towards newer accounts would systematically understate the estimated rate sensitivity and by so doing may distort earnings forecasts. Exhibits 1 and 2 present two examples that consider the possibility of a vintage effect.

RISK_4_10_19_ex_1.jpg

Exhibit 1 presents total balances, rates paid and two vintages of an institution’s interest-bearing checking account deposit balances. Vintage 1 represents the balances of all accounts that were in the institution as of August 2008 and shows how their balances have gradually declined over time. Vintage 2 is the difference between the two downward sloping lines and represents the balances of all accounts that were opened between August 2008 and August 2010. Exhibit 1 suggests that the retention rate for the newer vintage is very similar to the retention rate for the original vintage. A graph cannot show whether the interest rate sensitivity of deposits differs between the two vintages, but a more sophisticated statistical analysis can easily make such a determination.

RISK_4_10_19_ex_2.jpg

Exhibit 2 presents similar information for an institution with a very different deposit behavior and a major difference between vintages. The original vintage drops off at about the same rate as in the prior example, but the new vintage drops off much more rapidly, judging again by the difference between the downward-sloped lines. Once again, we cannot determine whether rate sensitivity differs between the two vintages, but Exhibit 2 suggests that the retention rate for the newer vintage is much lower than for the original. Both the difference in retention rates, clearly visible in Exhibit 2, and the potential difference in interest rate sensitivity of the two vintages, could have major implications for interest rate risk (IRR) applications and for earnings.

The second example considers whether there was a temporary increase or surge in deposits due to the flight to safety with the financial crisis. Exhibits 1 and 2 do not suggest any surge deposits during the financial crisis. Checking accounts typically focus on services rather than rates so no appreciable surge balances are expected. Exhibits 3 to 5 present three different cases examining money market deposit accounts (MMDA) balances focusing on the financial crisis period with a flight to safety and potentially a temporary increase in deposits. High-tier MMDAs are more likely to be sensitive to the returns on alternative financial instruments, e.g. stocks or bonds. During a period like the financial crisis, some of the balances in this product may be parked and waiting for a better return on other assets. Alternately, in periods of low returns, depositors may attempt to maintain a higher yield by moving funds to more risky assets, and MMDA balances could decrease rather than increase with a surge of deposits out of MMDAs. While conventional analysis of temporary or surge balances focuses on an increase in deposits, you should keep in mind that a surge could be positive or negative.

RISK_4_10_19_ex_3.jpg

Exhibit 3 presents the case of a financial institution that shows no sign of a surge in MMDA balances in the financial crisis. There is strong growth in MMDAs through the recent period with what appears to be a suspension of that growth during the financial crisis period. In contrast, Exhibit 4 shows an institution with a dramatic increase in MMDA balances at the start of the crisis followed by a gradual unwinding of those balances over the next four years. Balances 10 years later are up somewhat from their initial values, but the bulk of the run-up in MMDAs from late 2008 into 2010 appears to be temporary.

RISK_4_10_19_ex_4.jpg
RISK_4_10_19_ex_5.jpg

Exhibits 3 and 4 suggest that the process of distinguishing surge balances is relatively straightforward, with a large increase in deposits expected to reflect only a temporary change to the balance sheet. Exhibit 5 should make it clear that such a conclusion is unwarranted. In this case, there is a sharp increase in MMDA balances during the financial crisis, but there is no subsequent increased runoff. In fact, MMDA balances continue to grow albeit at a substantially slower rate. The pattern in Exhibit 5 suggests that the run-up to the crisis may have seen a drop in the trend growth rate that was recouped during the crisis with the long-term increasing trend resuming after the crisis period. The general conclusion from Exhibits 4 and 5 would be that if you had to make repricing decisions in late 2010 and were relying only on total balance data available at that time, you could be easily make a serious mistake. Could you have identified which one represented surge balances and which represented a permanent change?

The key feature of these exhibits is the behavior of total and retained balances, but no financial institution is immune to the vagaries of the market as a whole. Just in the past 12 years we have seen a sharp drop in market rates, a period of unprecedentedly low and unchanging rates followed by moderate rate increases. Those changes will affect both your repricing decisions and your depositors’ decisions. In addition, as noted in the first part of this series, banks currently have a total of $1.5 trillion in excess reserves. With that amount of excess reserves in the system, the general need to pay higher rates to attract deposits will differ substantially in 2019 from what it was in 2006 when the banking system had less than $2 billion in excess reserves.

To date, we have considered the general factors that will potentially impact your deposit base and thus your appropriate repricing, and then examined the data required to generate a best-in-class model. Next up: the modeling process itself.