Banks Cut Headcount and Noninterest Expense Rose 10 Percent. The Cost Base Is Not People.
Two numbers from the same quarterly filing move in opposite directions. Read together they say something specific about where a bank’s cost actually lives.
Insured institutions employed 2,034,151 full-time equivalents at the end of the second quarter, down 27,662 from a year earlier. Over the same twelve months, noninterest expense rose $15.0 billion, or 10.0 percent.
Fewer people, materially more cost.
What moves cost when headcount does not?
A cost that grows while the workforce shrinks is attached to the work rather than to the workforce. That is the whole finding, and it is worth sitting with before reaching for an explanation.
The convenient reading is compensation: fewer people, paid more each. That accounts for part of it and cannot account for all of it, because a 1.3 percent reduction in staff would need an extraordinary per-head increase to produce a 10.0 percent rise in total noninterest expense. Something other than salary is carrying the growth.
The rest is the cost of running processes. Technology contracts that price on volume rather than seats. Third-party services absorbing work that used to sit inside the building. Vendor platforms with an annual uplift. Occupancy that does not fall when a team gets smaller. And the recurring internal cost of every process that still needs a person to move it from one system to the next.
Why the shrinking headcount makes it worse
Removing people from a manual process does not remove the process. It concentrates it.
When a reconciliation team loses two of nine, the breaks do not stop appearing. The same volume arrives against less capacity, so the queue lengthens, the oldest items age further, and the work that used to happen in the same week now happens in the following one. Nothing on the income statement records that. What does get recorded is the overtime, the contractor brought in for the quarter, or the outside firm engaged to clear the backlog before a filing deadline.
Each of those lands in noninterest expense. None of them appears as headcount.
This is why the two lines can move apart for several quarters without anyone inside the institution experiencing it as a contradiction. Everybody is busier. The staff number is down. The cost number is up. All three are true at once and they are the same fact seen from three positions.
Where the money actually goes
Consider what a single unreconciled break costs to resolve. Somebody notices a difference between the core and an ancillary system. They pull extracts from both. They compare them, usually in a spreadsheet, and identify the transaction responsible. They determine which side is right, correct the wrong one, and record what they did somewhere that may or may not be searchable next quarter.
That is a person-hour or several, and it produces no durable asset. The same break, from the same cause, appears again in the next cycle because nothing captured why it happened the first time. The institution pays for the same resolution repeatedly and books it as staff time or as outside help, depending on how far behind the queue has fallen.
Multiply that by the exception queues, the manual bridges between platforms, the reports assembled by hand each period, and the approvals that live in a mailbox. Those four are the places we see the banking operating day back up. None of it is a project. All of it is cost, and it grows with transaction volume rather than with the number of people employed.
Loans grew 6.8 percent over the same year. Volume went up, staff went down, and the work in between was absorbed by processes nobody had rebuilt.
What changes the shape of the line
A process only stops costing money when it stops requiring a person to run it. Reducing headcount against an unchanged process moves cost between line items; it does not remove cost.
The work that responds to this is unglamorous and specific. Reconciliation logic captured as software rather than as a comparison somebody performs. Exception queues with named owners, aging, and escalation, so the backlog is a number that gets managed rather than a mailbox that gets deeper. Scheduled routines with a defined failure mode, so a job that does not run raises an alert instead of a silence that surfaces three weeks later as a variance.
CXO builds those systems through three practices, scoped and priced before work begins, with delivery measured against the process being replaced. That measurement matters more than usual here, because the alternative is another year in which the expense line moves and nobody can say which process moved it.
The test worth running on your own numbers is simple. Take your noninterest expense growth for the last four quarters and set it against your headcount change over the same period. If expense is rising faster than staff, the difference is the price of processes that still need a human in the middle.
Start with the process that costs you most. A conversation first, and a structured assessment when it earns one. Book a discovery call at https://cxocorporation.com/contact.