Which lenders are seeing commercial real estate delinquency turn?
Every lender type in commercial real estate publishes a delinquency rate, and no two
of them mean the same thing. CMBS draws the line at 30 days past due. Banks draw it
at 90 days or an accountant's judgment that the loan will not pay. The agencies draw
it at 60. So a reader ordering the published rates by height is mostly ranking where each
publisher drew its line — a fact about disclosure conventions, not about credit. This
page does something else with them: it asks each series how much it has moved against itself
a year ago, measured over the same borrowers at both ends, and draws those on one
axis. Every rate below is published by somebody else. The ruler is ours.
CMBS conduit
+21.7%
year over year to 2026-09 · 30+ days ·
194 trusts on both ends
Bank CRE
-7.0%
year over year to 2026-06 · 90+ days OR nonaccrual ·
3,596 banks on both ends
Fannie Mae multifamily
-4.7%
year over year to 2026-07 · 60+ days
Freddie Mac multifamily
+30.4%
year over year to 2026-07 · 60+ days
Ginnie Mae multifamily
-47.7%
year over year to 2026-06 · 60+ days
CRE CLO — four disclosing managers
+35.2%
year over year to 2026-06 · nonaccrual ·
4 managers on both ends
What a turn is here
A change is a reading only if the entities on both sides of it are the same entities. Each
figure above re-strikes both ends of its comparison over only the trusts, banks or
managers that reported in both periods — so it is not last year's published rate
subtracted from this year's, and it will not tie out against the rates in the Levels view.
That is deliberate, and it is the whole method.
It matters because the filing population moves. The
CMBS tape is the clearest case: a month in which far fewer trusts file produces a spike on
the published population that then reverts, while the trusts filing in every one of those
months show a smaller step that holds. Opposite shapes, same months, same tape.
And it cuts the other way in a crisis: failing banks leave the
denominator, and they are the sick ones, so the published bank series understates
deterioration exactly when it matters most.
Expressed as a percentage of the year-ago rate rather than in percentage points, for
a measured reason: CMBS publishes a 30-day and a 90-day rate on the same book, which is a free
test of whether a transform is neutral to where the line is drawn. Percentage-point change is
not — the severe bucket moves about two-thirds as far in points, almost every month,
because a deteriorating book adds similar dollars to both buckets while the bases differ.
Relative change is near-neutral and noisier. On a page whose entire job is defeating the
threshold, a systematic bias is fatal and noise is honest.
Year-over-year change in delinquency, same borrowers at both ends
above zero = deteriorating · 1-year lookback
· from 2023-10
CMBS conduitBank CREFannie Mae multifamilyFreddie Mac multifamilyGinnie Mae multifamilyCRE CLO — four disclosing managers
Each point compares a period with the same period a year earlier, over only the entities that
reported in both. The axis is set from the typical move rather than from the largest one and
stops at ±300%;
1 of 155 points fall outside it and
are drawn as triangles at the edge, in their own series' color, rather than dropped or
flattened onto it. A
comparison is not drawn at all where fewer than 25 entities report at
both ends, or where the year-ago figure rounds toward nothing: a relative change over a
near-zero base is not a large move, it is not a move. Direction and timing are comparable
across these lines. Magnitude is not — even under this transform the
two thresholds CMBS publishes on one book still disagree by about 41% of a typical move, so
nothing here should be read as one sector deteriorating more than another.
Year-over-year change in delinquency, same borrowers at both ends
above zero = deteriorating · 1-year lookback
· from 2021-10
CMBS conduitBank CREFannie Mae multifamilyFreddie Mac multifamilyGinnie Mae multifamilyCRE CLO — four disclosing managers
Each point compares a period with the same period a year earlier, over only the entities that
reported in both. The axis is set from the typical move rather than from the largest one and
stops at ±300%;
1 of 259 points fall outside it and
are drawn as triangles at the edge, in their own series' color, rather than dropped or
flattened onto it. A
comparison is not drawn at all where fewer than 25 entities report at
both ends, or where the year-ago figure rounds toward nothing: a relative change over a
near-zero base is not a large move, it is not a move. Direction and timing are comparable
across these lines. Magnitude is not — even under this transform the
two thresholds CMBS publishes on one book still disagree by about 41% of a typical move, so
nothing here should be read as one sector deteriorating more than another.
Year-over-year change in delinquency, same borrowers at both ends
above zero = deteriorating · 1-year lookback
· from 2016-10
CMBS conduitBank CREFannie Mae multifamilyFreddie Mac multifamilyGinnie Mae multifamilyCRE CLO — four disclosing managers
Each point compares a period with the same period a year earlier, over only the entities that
reported in both. The axis is set from the typical move rather than from the largest one and
stops at ±300%;
22 of 429 points fall outside it and
are drawn as triangles at the edge, in their own series' color, rather than dropped or
flattened onto it. A
comparison is not drawn at all where fewer than 25 entities report at
both ends, or where the year-ago figure rounds toward nothing: a relative change over a
near-zero base is not a large move, it is not a move. Direction and timing are comparable
across these lines. Magnitude is not — even under this transform the
two thresholds CMBS publishes on one book still disagree by about 41% of a typical move, so
nothing here should be read as one sector deteriorating more than another.
Year-over-year change in delinquency, same borrowers at both ends
above zero = deteriorating · 1-year lookback
· the whole record
CMBS conduitBank CREFannie Mae multifamilyFreddie Mac multifamilyGinnie Mae multifamilyCRE CLO — four disclosing managers
Each point compares a period with the same period a year earlier, over only the entities that
reported in both. The axis is set from the typical move rather than from the largest one and
stops at ±300%;
25 of 682 points fall outside it and
are drawn as triangles at the edge, in their own series' color, rather than dropped or
flattened onto it. A
comparison is not drawn at all where fewer than 25 entities report at
both ends, or where the year-ago figure rounds toward nothing: a relative change over a
near-zero base is not a large move, it is not a move. Direction and timing are comparable
across these lines. Magnitude is not — even under this transform the
two thresholds CMBS publishes on one book still disagree by about 41% of a typical move, so
nothing here should be read as one sector deteriorating more than another.
The published rates, as each publisher reports them
4 different definitions of “delinquent” · logarithmic axis
· from 2023-10
CMBS conduitBank CREFannie Mae multifamilyFreddie Mac multifamilyGinnie Mae multifamilyCRE CLO — four disclosing managers
These heights are not comparable with each other, and the view is here so
that claim can be checked rather than taken on trust. A line sitting below another may simply
be counting from a later day. The table underneath says, for each series, what day it counts
from and over whom. The axis is logarithmic, which is not a flourish: the
rates in this window span a ratio of about
46 to 1 — 1.8 decades —
and on a linear axis fitted to the top of that range the lower series superimpose into a
single band at the bottom of the plot, present and unreadable. A log axis is also the one
this page's argument implies: if relative change is the comparable quantity, then equal
ratios should be equal distances.
The published rates, as each publisher reports them
4 different definitions of “delinquent” · logarithmic axis
· from 2021-10
CMBS conduitBank CREFannie Mae multifamilyFreddie Mac multifamilyGinnie Mae multifamilyCRE CLO — four disclosing managers
These heights are not comparable with each other, and the view is here so
that claim can be checked rather than taken on trust. A line sitting below another may simply
be counting from a later day. The table underneath says, for each series, what day it counts
from and over whom. The axis is logarithmic, which is not a flourish: the
rates in this window span a ratio of about
171 to 1 — 2.37 decades —
and on a linear axis fitted to the top of that range the lower series superimpose into a
single band at the bottom of the plot, present and unreadable. A log axis is also the one
this page's argument implies: if relative change is the comparable quantity, then equal
ratios should be equal distances.
The published rates, as each publisher reports them
4 different definitions of “delinquent” · logarithmic axis
· from 2016-10
CMBS conduitBank CREFannie Mae multifamilyFreddie Mac multifamilyGinnie Mae multifamilyCRE CLO — four disclosing managers
These heights are not comparable with each other, and the view is here so
that claim can be checked rather than taken on trust. A line sitting below another may simply
be counting from a later day. The table underneath says, for each series, what day it counts
from and over whom. The axis is logarithmic, which is not a flourish: the
rates in this window span a ratio of about
1,199 to 1 — 3.21 decades —
and on a linear axis fitted to the top of that range the lower series superimpose into a
single band at the bottom of the plot, present and unreadable. A log axis is also the one
this page's argument implies: if relative change is the comparable quantity, then equal
ratios should be equal distances.
The published rates, as each publisher reports them
4 different definitions of “delinquent” · logarithmic axis
· the whole record
CMBS conduitBank CREFannie Mae multifamilyFreddie Mac multifamilyGinnie Mae multifamilyCRE CLO — four disclosing managers
These heights are not comparable with each other, and the view is here so
that claim can be checked rather than taken on trust. A line sitting below another may simply
be counting from a later day. The table underneath says, for each series, what day it counts
from and over whom. The axis is logarithmic, which is not a flourish: the
rates in this window span a ratio of about
1,199 to 1 — 3.21 decades —
and on a linear axis fitted to the top of that range the lower series superimpose into a
single band at the bottom of the plot, present and unreadable. A log axis is also the one
this page's argument implies: if relative change is the comparable quantity, then equal
ratios should be equal distances.
What each line is, and why the heights cannot be ranked
6 series, 4 different
definitions of “delinquent”, 2 different
reporting clocks, and records running from 9 observations to
247. This table is the reason the chart above is readable at all, so
it sits with the chart rather than under a footnote. Every figure in it is computed from the
same numbers that draw the lines.
Series
Counts from
Clock
Record
Reporting now
Who is in it
CMBS conduit
30+ daysCREFC's reporting boundary; the 90+ bucket is OPEN (90 days or more)
monthly remittance
2020-01 → 2026-1082 readings
102 trusts
constant panel
securitized conduit loans in trusts filing ABS-EE that month
Bank CRE
90+ days OR nonaccrualthe call report's "noncurrent", which folds an accounting judgment into a day count
quarterly call report
1993-03 → 2026-06134 readings
3,646 banks
constant panel
multifamily, nonresidential and construction CRE at FDIC-insured banks
Fannie Mae multifamily
60+ days"seriously delinquent", UPB-weighted; forbearance counted as delinquent per its own endnote
monthly
2006-01 → 2026-07247 readings
one publisherwhole book
Fannie's own guaranteed multifamily book
Freddie Mac multifamily
60+ daysforbearance EXCLUDED where the borrower complies — the opposite of Fannie's rule, in Freddie's own endnote
monthly
2012-01 → 2026-07175 readings
one publisherwhole book
Freddie's own guaranteed multifamily book
Ginnie Mae multifamily
60+ daysMonths Delinquent ≥ 2, UPB-weighted; blended across a residential and a healthcare book
monthly
2019-01 → 2026-0690 readings
one publisherwhole book
Ginnie multifamily pools (FHA-insured), both books
CRE CLO — four disclosing managersA named panel, not a sector rate — see below.
nonaccrualan accounting judgment, not a day count — the only series here with no threshold at all
quarterly 10-Q
2023-09 → 2026-069 readings
4 managers
constant panel
the four public managers that publish a loan-level aging table
Two of these need a word more. Where a series is a single publisher reporting one number over
a book nobody outside can see inside — Fannie, Freddie and Ginnie each guarantee their
own — holding the entities constant has nothing to hold: there is one entity and it is
the publisher. Those turns are computed from the published rate, and they are marked
“whole book” above rather than quietly passed as though the rule had run.
Starts 1993Q1, not 1984. The table holds 169 quarters back to 1984Q1, but before 1991Q1 the multifamily and nonresidential noncurrent columns are entirely empty — only construction is filed, on about a third of banks, against a denominator carrying all three. When those columns switch on in 1991Q1 the rate steps 1.243% to 7.412% in a single quarter, which reads as the S&L crisis arriving and is the columns arriving. 1993Q1 is the first quarter whose numerator covers its own denominator.
And the ruler has a limit worth saying plainly: it fixes a population that moves. It
can do nothing about one that is permanently a biased subset, which is why the CRE CLO line
is drawn under its own name rather than under the sector's, thinner and dashed. It is the
managers who publish a loan-level aging table, not the market.
What was scoped and refused
These were measured and turned down rather than never considered. A series that is silently
absent from a page like this reads as a series nobody thought of.
Life insurance companies
Not drawable as one line, and the prohibition is this platform's own. The Lenders page already refuses to aggregate these filers, in its own words — twelve of twenty disclose an aging table, so a weighted mean over them is "a statement about WHO DISCLOSES wearing the clothes of a market rate". That reasoning does not stop applying because a different page wants the line. Measured here independently: only ONE filer carries both a 90+ figure and an unambiguous book of its own to divide it by, so the panel would be a single company. The per-filer arrears column on the Lenders page is the honest form of this reading and it already exists.
CRE CLO, as a sector rate
Refused as a SECTOR rate and drawn as a named panel instead — see the four-manager line. Four of the nine tracked managers publish a loan-level aging table in essentially every quarter they appear; the other five publish one in zero of nine. The disclosers hold 21-27% of tracked UPB, and the split is permanent rather than a filing accident, so a constant panel holds it constant and does not cure it. A rate over the disclosers labeled "CRE CLO" would be a selection effect wearing a sector's name.
What this page leaves out, and why
A page that omits something silently is indistinguishable from one that never had it, so this
list is part of the page rather than a note about it. Both of these are limits of a chart with
no controls on it, and each will stop being true when somebody builds the thing that fixes it.
How many trusts, banks or managers stood behind each individual pointrendering
Every comparison here is re-struck over the entities that filed at BOTH ends of it, and each point knows how many that was. On a static chart there is nowhere to put 682 of those counts, so the page publishes the current population for each series in the table below and the floor beneath which a comparison is not drawn at all. The signed-in chart shows the pair for whichever point you point at.
The month-by-month working behind the two examples quoted hererendering
The two readings that justify the method are printed on this page in full - the CMBS spike that reverts against the same trusts stepping and holding, and the 2009 quarter where the published bank series understates its own deterioration. What is behind each is a small table of populations and rates per month. The conclusion is the part that belongs on a page; the working is in the signed-in version for anyone who wants to check it line by line.
Where this sits
This page is national, and it answers which lender type is turning.
It is one hand of a clock with several.
These pages publish the national reads — where the credit cycle sits, whether banks are pulling back from commercial real estate, and what moved this week. The platform carries them down to the market and the loan: the policy frame, the index rates a mortgage prices over, and what any of it touches in the buildings you hold.
