Behavioural Maturity: a Worked Example

On contract, every transaction account matures tomorrow. A maturity ladder built only from contracts would show a bank owing most of its deposits overnight while lending them out for years, a gap that would look fatal and in ordinary times is not. Most balances stay. Customers pay wages and suppliers from them, receipts refill them, and the total moves far less than any one account does. The same gap runs the other way on the asset side: a fixed-rate loan written for five years is often refinanced in two, and a revolver with a three-year term is renewed before it expires.

So a bank keeps two calendars. The contractual one says what it is legally owed and owes. The behavioural one says what is likely to happen, and three decisions rest on it: how much of the deposit base can fund long-dated loans, which is the maturity transformation at the centre of banking; how long the deposits' interest rate is effectively fixed, which sets how treasury hedges them; and what funds transfer pricing credits to the team that gathered them.

The method has three steps. First, the balances are split into groups that behave alike: household against business, transaction against savings, accounts used for payments against money parked for yield. Second, each group's stable part is measured from its history, typically the share of the balance that has stayed through a long run of years. Third, the stable part that is also unlikely to reprice, called core, is spread across time buckets as if it matured over several years; the rest is treated as overnight.

Supervisors review the estimate, and the Basel Committee's 2016 standard on interest rate risk in the banking book shows one way to bound it. Its standardised framework, which a supervisor can require a bank to use, takes ten years of balance data and caps the result: core may be at most 90% of retail transactional balances with an average maturity of five years, 70% and four and a half years for other retail balances, and 50% and four years for wholesale, which includes deposits from companies unless a small business is managed as retail. Each market applies the standard its own way. In Australia, banks approved to model the risk for capital under APS 117 must document and support their assumptions for balances with no repricing date, and APRA can impose others; the PRA expects UK firms to justify how they slot current accounts and sight deposits.

Liquidity rules make their own, harsher assumption about the same money. The Basel liquidity coverage ratio expects a stressed month to take 5% of insured, stable household deposits, 25% of deposits that companies hold for payments and cash management, and 40% of other deposits from non-financial companies. The UK applies that ratio to its banks and Australia to all but the ADIs with simple retail models, which hold a minimum share of liquid assets instead; New Zealand's Reserve Bank currently uses its own mismatch and core funding ratios.

A bank, constructed for this page like every figure in it, holds US$800m in business transaction accounts that pay little or no interest. Ten years of balances show that about 60% of the total has stayed through every downturn in the period.

The history would support 480 as core; the cap allows 400, because these are company deposits. Spread evenly over seven years, the core averages four years, which is the limit for this category. The non-core 400 is treated as able to leave at once. The stress line treats every balance as non-operational, the example's assumption; balances that qualified as operational deposits would carry 25%.

The two views do not contradict each other. The interest rate view says that over a cycle about half the balance behaves like four-year money. The liquidity view says that in a bad month as much as 320 could go. The bank funds and hedges on the first and holds liquid assets against the second. If treasury credits core at a rate 0.40 points above overnight, a gap that is the example's, the business that gathered the deposits earns US$1.6m a year more than it would if they were all treated as overnight.

The neighbour readers confuse it with is how quickly a balance's interest rate follows the market. Behavioural maturity is about whether the money stays; repricing behaviour, measured by deposit beta, is about what the bank has to pay to keep it. A bonus savings account can stay for years and still reprice every month, so it is long in one sense and short in the other. The Basel definition of core needs both: balances that are stable and unlikely to reprice even when rates move a long way.

The two are linked under stress. Depositors who stayed put while rates were low begin to move when better rates appear elsewhere, so a rising cycle shortens behavioural maturity and lifts beta at the same time.

DEFINED IN ONE PARAGRAPH → · WHERE BANKS GET THEIR MONEY → · FUNDS TRANSFER PRICING →

For a business customer, the estimate is why a bank values the operating account more than a large balance parked for yield. Payroll and supplier payments run through the first, so it is likely to stay, earns a better transfer-pricing credit and counts in the relationship's return. A single large deposit is often treated as less stable than many small ones of the same total, however long it has been there, because one decision can move all of it.

For a bank, the risk is the estimate itself. In March 2023 a US regional bank whose deposits were about 94% uninsured, most of them held by companies, lost over US$40bn of deposits in a day and expected more than US$100bn to follow the next. Its group had tripled in size in the three years before, so little of its deposit history came from a stress. When the depositors doubted the bank, that history stopped describing them within hours, which is why behavioural assumptions are capped and tested against a run rather than taken from the good years alone.

NET INTEREST MARGIN → · THE BANK TREASURY ANALYST PAPER →