County banking markets

The consolidation did not reach the median county

Between 1994 and 2025 the number of institutions operating branches fell 66% and the branch network shrank 6%. The median county market did not move. It had 6 institutions in 1994 and 6 in 2025, and its branch-count HHI is 2,222 against 2,222. It fell to 2,000 in 2008 and came back. The mean county, which is not confined to the lattice the branch HHI takes in small markets, does the same: 2,993 in 1994, 2,690 in 2009, 2,988 in 2025. This is a null result, and it is the finding. What the module cannot tell you is why.

Why the deposit HHI cannot carry a county page

The Fed has a name for the problem. “Central booking occurs when a bank records deposits at a central office and does not book the deposits to a branch that is connected to the location of the depositor. As a result, centrally booked deposits of an applicant can overstate the presence of an institution in the local banking market.” SOD is a survey of branches of record, so it contains the distortion unadjusted. Below are the 2025 counties where the two indices disagree most, each with at least 20 branch offices, ranked by deposit HHI minus branch HHI.

Source: FDIC Summary of Deposits | Federal Reserve, Competitive Effects of Mergers and Acquisitions FAQs June 30, 2025. Counties with 20 or more branch offices, ranked by deposit HHI minus branch HHI. The Fed also drops credit-card banks, trust companies and internet banks from local market shares and weights thrift deposits at 50%; neither index here does any of that. Methodology

Most and least concentrated county markets, 2025

By branch-count HHI, among counties with at least 20 branch offices. The floor is doing work: 236 counties in the 2025 survey are served by a single institution, which puts every one of them at the 10,000 ceiling by construction, so without a floor the most-concentrated table would be a list of those and nothing else. Deposit HHI is in the last column for comparison, not as the ranking key.

Source: FDIC Summary of Deposits June 30, 2025. Branch-count HHI, counties with 20+ offices. Methodology

The counties that did lose banks

Of the 3,172 counties in both the 1994 and 2025 surveys, 1,188 have fewer institutions in 2025, 1,183 have more, and 801 are unchanged. 87 that had more than one institution in 1994 have exactly one now. The ten counties below are those with the largest fall in institution count; they are not a random sample, and nothing here measures where losses fall in general.

Source: FDIC Summary of Deposits Counties present in both the 1994 and 2025 surveys, ranked by change in institution count, ties broken by 1994 branch count then FIPS. 3 of these 10 lost institutions while their branch count grew. Methodology

Codes that leave the survey, and what leaving does not prove

3,237 county FIPS codes appear across the panel; 3,194 are in the 2025 survey. That leaves 43 absent. Reading all 43 as counties that lost their last bank would be wrong for 16 of them, which are not county equivalents in the 2025 geography: 14 are recodes, where the geography still exists under a different FIPS code, and 2 are codes Census does not treat as county equivalents. By jurisdiction: 8 Connecticut, 4 Alaska, 1 Florida, 1 Marshall Islands, 1 Northern Mariana Islands, 1 Virginia.

Codes that are not counties in the 2025 geography (16)

Source: FDIC Summary of Deposits | US Census Bureau, Substantial Changes to Counties and County Equivalent Entities Every successor code above carries at least one branch in its first survey. Each successor is matched by FDIC cert, not by proximity: 69010's last branch is First Hawaiian Bank (cert 17985), and cert 17985 files under Saipan (69110) in 2025, so Saipan is the successor and Tinian, which is merely new in 2025, is not. The switch is not clean either: 2 of the 4 retired Alaska codes keep receiving branches after their successors appear (02261 for 2 surveys, 02280 for 5), so a raw count of distinct county_fips double-counts those geographies in the overlap years. Methodology

Absent codes the register does not flag, with no SOD branch since the year shown (27)

These are the 27 absent codes the register above does not flag. The claim each row supports is the one printed in it: no SOD branch reported under this code after that year. That is not “no banking access”. SOD counts branches of FDIC-insured institutions only, so a county served entirely by credit unions carries no row here, and a resident may bank at a branch one county over, or online. American Samoa left the survey outright: none of its 3 codes appears after 2023, when its last 2 branches reported.

Source: FDIC Summary of Deposits | US Census Bureau, Substantial Changes to Counties and County Equivalent Entities Codes with no SOD branch after the year shown. Leaving is not always permanent: 19 FIPS codes in the panel dropped out of the survey for one or more years and came back. Methodology

What this cannot tell you

• A county is not a banking market. The Fed defines markets from commuting and shopping patterns; some are MSAs, some are groups of counties, some are parts of counties. Its actual definitions live in CASSIDI, not here. County HHI is a proxy for the merger-review lens, not the lens.
• Neither HHI on this site is the Fed’s screen HHI. The Fed weights thrift deposits at 50%, excludes credit-card banks, trust companies and internet banks from local shares, and adjusts case by case for central booking and large government deposits. These indices are unweighted and include every reporting institution. Do not read a number here as a merger-review outcome.
• The deposit HHI is not a measure of local competition. Central booking makes it an upper bound in the county where a bank books, and a lower bound in every county the money actually came from. That is why it is never the headline here.
• SOD counts branches, not access. Credit unions are not in this survey (they file with the NCUA), branchless banks hold deposits that no county sees, and a branch office is not the same thing as available credit. A county with no SOD branch is not proven to be a banking desert.
• There is no county master list in this estate. Every code here is one that reported at least one SOD branch at some point since 1994. A county that never reported one is invisible to this module and cannot be counted by it, so the absent-code tables above are a lower bound on counties without a branch, not the full set.
• Not every code in the county file is a US county. The FDIC surveys the branches of insured institutions in the Compact of Free Association states (Federated States of Micronesia, Marshall Islands, Palau), which are sovereign countries and whose FIPS codes Census does not treat as county equivalents. 6 such codes are in the 2025 survey, carrying 12 branches, and 1 has left it. They are kept, labelled on their own pages, and counted in the medians. Read “county” on this page as a row of the SOD county file.
• Medians describe counties, not people. Every median here is unweighted: a county of 10 million and a county of 500 count once each. This module holds no population data, so it cannot say how many people live in the counties that lost institutions, and the flat median above is not a statement about the typical depositor.

Every geography in the 2025 survey

3,194 codes, grouped by jurisdiction. Each has its own page, covering only the surveys that code appears in: 3,159 of the 3,194 appear in all 32 surveys, and the other 35 in fewer. The figure beside each is its 2025 branch-count HHI. 6 of them are in the Compact of Free Association states, which the FDIC surveys and Census does not count as county equivalents. The 43 codes absent from this survey have pages too, and are listed above.