Yes, but the pullback is concentrated in correspondent/remittance account services, not physical branch presence — border-county branch decline has tracked the national trend almost exactly

Summary

Yes, but selectively. Banks did pull back sharply from account-level correspondent, money-services-business (MSB), and cross-border-trade banking relationships in Southwest border counties during roughly 2010-2016 — the U.S. Government Accountability Office's 2018 survey found an estimated 80% (±11 percentage points) of Southwest border banks terminated at least one account for Bank Secrecy Act/anti-money-laundering (BSA/AML) risk reasons in 2014-2016, and similarly large shares limited or refused accounts to foreign businesses, MSBs (the primary U.S. channel for remittances), and cross-border-trade businesses. However, this pullback was mostly a service-line and account-level phenomenon, not a story of the border region physically losing bank branches faster than the rest of the country. Recomputing the same 33-county Southwest border definition GAO used, against FDIC branch data through 2025, shows border-county branch counts fell 22.7% since 2012 (1,444 → 1,116 branches) versus 21.8% nationally over the same span — essentially identical rates. A difference-in-differences test confirms the border-vs-national branch gap did not widen after 2016 (interaction coefficient +0.39 index points, p=0.92, statistically indistinguishable from zero) — extending, with newer data, GAO's own 2018 finding that the region has been "losing bank branches since 2012, similar to national and regional trends." Treasury's 2023 De-risking Strategy, drawing on more recent (2022) bank/MSB consultations, found that de-risking of MSBs has largely "plateaued" nationally since a mid-2010s peak, though small MSBs serving high-risk immigrant/remittance corridors still face significant difficulty, while large and medium MSBs are less affected than they were in 2015.

What actually pulled back: correspondent, MSB, and cross-border-trade accounts

GAO defined the "Southwest border region" as the 33 counties with at least 25% of their landmass within 50 miles of the U.S.-Mexico border: Cochise, Pima, Santa Cruz, and Yuma (AZ); Imperial and San Diego (CA); Doña Ana, Hidalgo, and Luna (NM); and Brewster, Brooks, Cameron, Culberson, Dimmit, Edwards, El Paso, Hidalgo, Hudspeth, Jeff Davis, Jim Hogg, Kenedy, Kinney, La Salle, Maverick, Presidio, Starr, Terrell, Uvalde, Val Verde, Webb, Willacy, Zapata, and Zavala (TX).

GAO surveyed a nationally representative sample of 406 banks, including all 115 banks operating in this region, on 2014-2016 account activity. Key findings:

• ~80% (±11 pp) of Southwest border banks terminated at least one account for BSA/AML risk reasons.
• ~80% (±11 pp) limited or did not offer accounts to customers considered high risk for money laundering because those customers drew heightened regulatory scrutiny — the survey's own working definition of de-risking behavior.
• Specifically by customer type, Southwest border banks reported limiting or not offering accounts to: 76% for non-trade-related foreign businesses, 75% for money services businesses (the main U.S. remittance channel), and 72% for foreign businesses engaged in cross-border trade.
• The most-cited reason (88% of banks) was that the customer type fell outside the bank's risk tolerance; 69% cited inability to manage the BSA/AML risk given resource constraints; 80% cited the heightened regulatory oversight that comes with these customer types.

In the three border communities GAO visited in depth — Nogales, AZ; San Ysidro, CA; and McAllen, TX — local businesses and money-service-business owners reported longtime accounts terminated "sometimes without notice or explanation," cross-border produce and used-clothing exporters losing accounts specifically because of large cash volumes moving across the border, and Mexican-national customers finding it harder to maintain U.S. accounts used while shopping in border towns. GAO also documented that these communities lost 12 bank branches from 2013-2016 (9 of them branches of large/extra-large banks): Nogales lost 3 of its 9 branches (33%), San Ysidro lost 5 of its 12 branches (42%), and McAllen lost 4 of its 63 branches (6%) — illustrating that the branch impact, while real, was concentrated in specific smaller communities rather than spread evenly across the whole 33-county region.

What did NOT disproportionately pull back: physical branch presence region-wide

To check whether the account-level pullback GAO documented through 2016 corresponds to the border region losing physical bank presence faster than the rest of the country — and to extend GAO's analysis nine years past its original window — we queried FDIC's own Summary of Deposits data (the same source GAO used) for all 33 GAO-defined Southwest border counties, 1994-2025.

Branch counts in these 33 counties fell from a peak of 1,444 in 2012 to 1,116 in 2025, a decline of 328 branches (-22.7%). Nationally, branch counts fell from 97,341 to 76,097 over the same period (-21.8%). We then ran a difference-in-differences test on the two annual branch-count indices (border-county vs. national, both rebased to 100 in 2012), with 2017-2025 as the "post" period following the end of GAO's original data window. The border × post-2016 interaction term is +0.39 index points (standard error 3.71, p=0.92) — not distinguishable from zero, meaning the gap between the border-county and national branch-decline paths did not widen (or narrow) after 2016. Total branch deposits in the border counties actually rose over the same period, from $109.8 billion to $205.4 billion (nominal), consistent with nationwide deposit growth, not regional retreat.

This confirms, with newer data than GAO had, GAO's own 2018 conclusion: "Counties in the Southwest border region have been losing bank branches since 2012, similar to national and regional trends" — branch consolidation at the border is overwhelmingly a story of the same digital-banking-driven, nationwide branch-network shrinkage every U.S. county has experienced, not a border-specific retreat, and that has remained true for the nine years since GAO's report.

The distinction that matters, and why

The apparent tension — branches not disproportionately closing, but account-level services sharply curtailed — is not a contradiction. FDIC's branch/deposit data can only observe whether a branch exists and how much money sits in it; it cannot observe which categories of customer a still-open branch is willing to serve. GAO's report separately states, as one of its own findings, that money-laundering-related indicators (High-Intensity-Drug-Trafficking-Area status, and the volume of Suspicious Activity Reports filed per dollar of deposits) were among the factors GAO judged likely more consequential in the Southwest border region than elsewhere, given that the region had a higher concentration of them — SAR filings there were 2.5 times the national high-risk-county average in 2016. That is GAO's own reported conclusion, not something recomputed here. Read alongside the account-level survey results and the diff-in-diff test above, it suggests money-laundering-related risk plausibly explains part of why border branches close at all, even though the region's overall closure rate does not exceed (and has not diverged from) the national rate; the larger, more visible effect of that same risk shows up as account terminations inside branches that stay open, which is exactly what GAO's bank survey captured and branch-count data cannot.

Treasury's 2023 AMLA De-risking Strategy, drawing on 2022 consultations that GAO's 2018 report could not have captured, found the intensity of MSB de-risking has changed since the mid-2010s peak: as of 2022, "large MSBs do not experience widespread denial of access to banking services," and "medium-size MSBs have not experienced de-risking to the same degree as they did in 2015" (though they still face high fees and lengthy onboarding). Small MSBs serving high-risk immigrant/remittance-corridor communities, however, "continue to have significant issues accessing banking services" — the same population GAO's border-community visits identified as most affected. A parallel 2015 World Bank global survey of money transfer operators (MTOs) found that by that year, roughly 28% of MTO principals and 45% of MTO agents worldwide had already lost banking access, and a 2017 Financial Stability Board survey found respondent banks terminated services to MTOs at least 70% more often than to other high-risk client types — both consistent with, though not specific to, the U.S. Southwest border. No public U.S. source found (including a search for post-2018 border-county-specific MSB/correspondent-account counts) provides a border-county-level, post-2016 update to GAO's specific account-termination percentages; the most recent authoritative account is Treasury's 2023 national-level assessment.

Bottom line

Banks pulled back substantially and measurably from correspondent-banking-adjacent, MSB/remittance, and cross-border-trade account relationships in Southwest border counties, concentrated in 2010-2016 (roughly 80% of border banks terminated or limited such accounts) and driven by BSA/AML compliance cost and risk-tolerance decisions rather than a broader physical retreat — branch counts and deposits in the same 33 counties tracked the national branch-consolidation trend almost exactly, both through 2016 and (per a diff-in-diff test) in the nine years since. The pullback appears to have eased somewhat nationally by 2022 for larger MSBs but persists for the smallest, most remittance-dependent operators — the ones serving the border communities GAO singled out.

Every query behind this report

10 warehouse calls ran in this session, in order. Each is reproducible against the same snapshot.

SELECT s."year",
  COUNT(DISTINCT s.uninum) AS border_branches,
  SUM(s.deposits_branch_thousands) AS border_deposits_th
FROM banking.sod s
WHERE s.county_fips IN ('04003','04019','04023','04027','06025','06073','35013','35023','35029','48043','48047','48061','48109','48127','48137','48141','48215','48229','48243','48247','48261','48271','48283','48323','48377','48427','48443','48463','48465','48479','48489','48505','48507')
GROUP BY s."year"
ORDER BY s."year"

SELECT s."year",
  COUNT(DISTINCT s.uninum) AS national_branches,
  SUM(s.deposits_branch_thousands) AS national_deposits_th
FROM banking.sod s
GROUP BY s."year"
ORDER BY s."year"

SELECT county_fips, "year", COUNT(DISTINCT uninum) AS branches, SUM(deposits_branch_thousands) AS deposits_th
FROM banking.sod
WHERE county_fips IN ('04023','06073','48215','48479','48061','48141')
  AND "year" IN ('2012','2016','2020','2025')
GROUP BY county_fips, "year"
ORDER BY county_fips, "year"

SELECT county_fips, "year", COUNT(DISTINCT uninum) AS branches, SUM(deposits_branch_thousands) AS deposits_th
FROM banking.sod
WHERE county_fips IN ('04023','06073')
  AND "year" IN ('2012','2016','2020','2025')
GROUP BY county_fips, "year"
ORDER BY county_fips, "year"

SELECT county_fips, COUNT(*) FROM banking.sod WHERE county_fips IN ('04023','06073') GROUP BY county_fips

SELECT DISTINCT county_fips FROM banking.locations WHERE state_abbr='AZ' AND city ILIKE '%Nogales%'

SELECT county_fips, COUNT(*) FROM banking.sod WHERE county_fips LIKE '4023' OR county_fips LIKE '6073' OR county_fips = '4023' OR county_fips='6073' GROUP BY county_fips

SELECT county_fips, city, state_abbr, "year", COUNT(*) FROM banking.sod WHERE state_abbr='CA' AND "year"='2020' GROUP BY county_fips, city, state_abbr, "year" ORDER BY county_fips LIMIT 20

SELECT s."year",
  COUNT(DISTINCT s.uninum) AS border_branches,
  SUM(s.deposits_branch_thousands) AS border_deposits_th
FROM banking.sod s
WHERE s.county_fips IN ('4003','4019','4023','4027','6025','6073','35013','35023','35029','48043','48047','48061','48109','48127','48137','48141','48215','48229','48243','48247','48261','48271','48283','48323','48377','48427','48443','48463','48465','48479','48489','48505','48507')
GROUP BY s."year"
ORDER BY s."year"

SELECT * FROM (VALUES
(2012,1,0,100.0),(2013,1,0,99.4),(2014,1,0,97.8),(2015,1,0,95.0),(2016,1,0,93.4),
(2017,1,1,91.3),(2018,1,1,90.8),(2019,1,1,90.2),(2020,1,1,88.0),(2021,1,1,85.0),(2022,1,1,83.0),(2023,1,1,79.8),(2024,1,1,77.8),(2025,1,1,77.3),
(2012,0,0,100.0),(2013,0,0,99.0),(2014,0,0,97.3),(2015,0,0,95.8),(2016,0,0,94.3),
(2017,0,1,92.3),(2018,0,1,90.5),(2019,0,1,88.8),(2020,0,1,87.3),(2021,0,1,84.0),(2022,0,1,81.3),(2023,0,1,79.9),(2024,0,1,78.8),(2025,0,1,78.2)
) AS t(yr, is_border, post2016, branch_index)