The FDIC Summary of Deposits, explained

There is a habit in writing about banks of treating a branch count as a fact of nature — a number that simply exists, the way a building exists. It is not. Almost every deposit figure attached to a place comes from one survey, taken on one day of the year, under rules that say which buildings are inside it and which are not. That survey is the Federal Deposit Insurance Corporation’s Summary of Deposits, and three separate documents govern what it contains: a regulation that sets the dates, a set of reporting instructions that defines the office types, and a live register that answers a query on any day at all and follows neither.

This piece reads the three in that order. It is written for a reader who has a number in front of them and wants to know what the number is a count of.

One survey, one date

The as-of date and the deadline are set in the same clause of the same regulation. The text reads: “Reports as of June 30 of each year must be submitted no later than the immediately succeeding July 31.”

CODE OF FEDERAL REGULATIONS · 12 CFR 304.3 · law.cornell.edu

Two dates, one sentence, and the order matters. June 30 is the day the survey describes. July 31 is the day the description is due. Everything between those two dates — an office opened on July 2, an office closed on July 20 — belongs to the following year’s survey and to no part of this one. A deposit figure carrying a June 30 date is not stale because it is a month or a year old; it is a photograph, and a photograph is not out of date, it is dated.

CODE OF FEDERAL REGULATIONS · 12 CFR 304.3 · law.cornell.edu

The same regulation carries the one exemption a reader needs to know about: “institutions with only a main office are exempt from reporting.” An institution that operates a single office has nothing to distribute between places, and the survey excuses it. That exemption is the first reason a national tally of survey filers is smaller than a national tally of insured institutions, and it is worth holding on to, because the second reason is larger and much less obvious.

CODE OF FEDERAL REGULATIONS · 12 CFR 304.3 · law.cornell.edu

What the survey calls an office

The word office does not mean a place with a door and a sign. In the FDIC’s own location register, every office record carries a service-type code, and the codes present for a single certificate are not all the same kind of place. For certificate 12437 the register returns five distinct values — 11, 12, 23, 24 and 27 — and 243 of the 266 records share one of them, while the four remaining codes cover everything else.

FDIC LOCATION REGISTER · CERT 12437 · api.fdic.gov

The register does not leave the codes bare. It names the levels as well, in a field of its own, with strings such as "FULL SERVICE - BRICK AND MORTAR", "LIMITED SERVICE - LOAN PRODUCTION" and "LIMITED SERVICE - MESSENGER". A reader who has only ever seen the phrase “branch network” meets three different things at once here, and only the first of them is the thing the phrase brings to mind.

FDIC LOCATION REGISTER · CERT 12437 · api.fdic.gov

The survey’s own reporting instructions define the levels, and the definitions are not interchangeable. Two of them are quoted in the table below exactly as the instructions print them, numbers and all. A loan production office “Processes loans and does not accept deposits.” A limited service messenger office is “Used by courier services to make deposit drops; no retail customer interaction.” Neither definition describes a counter a member of the public walks up to, and one of them explicitly describes a place where no deposit is taken at all.

FDIC SUMMARY OF DEPOSITS · REPORTING INSTRUCTIONS · fdic.gov

FDIC Summary of Deposits reporting instructions, read September 22, 2026. The instructions define each level; the two above are quoted because they carry the piece’s argument. fdic.gov

The offices the instructions take out

Defining the levels is one job. Deciding which levels belong in the survey is another, and the instructions do that job in a sentence addressed directly to the institution filling in the form: “The offices below are not in your survey.” The survey population, in other words, is not the register population. Somewhere between the register and the form, records are removed by rule.

FDIC SUMMARY OF DEPOSITS · REPORTING INSTRUCTIONS · fdic.gov

One of those removals is stated outright rather than left to be inferred from the definition. The instructions say: “Loan Production Offices are not considered branches for purposes of this survey”.

FDIC SUMMARY OF DEPOSITS · REPORTING INSTRUCTIONS · fdic.gov

Read that against the register, which carries a level named "LIMITED SERVICE - LOAN PRODUCTION" and returns records under it, and the whole shape of the problem appears. A place can be a record in the FDIC’s location register and not be a branch for the FDIC’s own deposit survey. The two statements are both true and they are about the same building. They are not in conflict, because they answer different questions: the register answers “what offices does this certificate have on file today,” and the survey answers “among the offices this survey counts, where were the deposits on June 30.”

FDIC LOCATION REGISTER · CERT 12437 · api.fdic.gov

Why a live register and an annual survey disagree

Two mechanisms drive every disagreement between a register count and a survey count, and neither of them is an error.

The first is the date. The location register is a separate dataset from the annual survey and can be queried at any time; on the day this piece was verified it returned a total of 266 records for certificate 12437. A survey taken on June 30 and a query run on September 22 are separated by every structure change in between.

FDIC LOCATION REGISTER · CERT 12437 · api.fdic.gov

The second is the population, and it is the one that catches people. Even if the two were read on the same morning, the register would include office types the instructions remove — the loan production office above all. A difference of a handful of offices between a bank’s own page, a register query and a survey table is usually not a contradiction to be resolved. It is three documents doing three different jobs, and the honest response is to cite the one that answers the question being asked, with its date attached.

FDIC SUMMARY OF DEPOSITS · REPORTING INSTRUCTIONS · fdic.gov

Reading the register for one institution

The institution behind certificate 12437 is Renasant Bank, of Tupelo, Mississippi. Its office records are the ones quoted above, and its institution-level record answers the same question from a different part of the same system: the field "OFFICES" returns 266, which is the figure the location register returns by counting the records themselves. Two independently produced fields, one number.

FDIC INSTITUTION REGISTER · CERT 12437 · api.fdic.gov

The deposit figure on that same institution-level record carries its own report date, and the date is the June 30 the survey uses: deposits of $22,166,216 thousand as of June 30, 2026. The coincidence of dates is not a coincidence at all — June 30 is a reporting date the whole system is built around — but it is worth stating plainly, because it is the reason a deposit total and a survey can be laid side by side without adjusting either.

FDIC INSTITUTION REGISTER · CERT 12437 · api.fdic.gov

FDIC location register and FDIC institution register, certificate 12437, both read September 22, 2026. The codes present are 11, 12, 23, 24 and 27. api.fdic.gov

A note on reading a register literally

One more feature of the data deserves a line, because it is the kind of thing that turns a count into a wrong count. The register’s wording is not typographically consistent. Of the 243 records carrying the full-service brick-and-mortar level, 242 spell it in capitals and one does not: "FULL SERVICE - BRICK AND MORTAR" against "Full Service - Brick and Mortar". A count filtered on the exact capitalized string returns 242 and looks perfectly plausible.

FDIC LOCATION REGISTER · CERT 12437 · api.fdic.gov

That is why the counts in this piece were computed from a saved copy of the register response rather than by eye, and cross-checked against the institution-level field, which a different part of the system produces. A single record with different capitalization is not a scandal. It is a reminder that a register is a machine-readable file with a history, and that the discipline of reading it literally is not pedantry — it is the only way the number comes out the same twice.

What the survey figures are, and are not

The state-by-state distribution of the same 266 records is set out separately in where Renasant Bank operates, which counts the register rather than the survey and says so. The quarterly form that produces the balance-sheet figures beside these office records is explained in Call Reports; office and deposit-by-branch data do not come from that form.

Three limits apply to everything above. First, no deposit market share is stated here for any institution in any market: the market-share tables of the survey were not read, and a figure that has not been read is not published. Second, an office in the live register is not, by that fact, a branch — some service levels are removed from the survey by its own instructions, and a loan production office is not a branch for this purpose at all. Third, no institution is compared with any other. The survey supports comparison; this piece does not make one, because making one would mean publishing numbers that were never opened at their own source.

What is left after those three limits is still the useful part: a date, a deadline, an exemption, a definition, and the reason two published counts of the same bank can differ by more than a rounding error without either of them being wrong.

Charterbend does not provide investment advice. It does not rate, recommend or value any security or institution, and nothing here is an offer or a solicitation to buy or sell anything.