On 30 April 2025 BaFin lowered the sectoral systemic risk buffer on German residential mortgages from 2 per cent to 1 per cent, effective the next day. The countercyclical capital buffer stayed at 0.75 per cent. The reasoning was explicit: the overvaluation that had built up in German housing through the cheap-money years had largely been worked off, the market had cooled, and the capital held against a correction could be partly released — about €2 to €2.5 billion, roughly 0.4 per cent of the banking system’s Common Equity Tier 1.
Twelve months later, the first loan-level data Germany has ever collected on its mortgage market told a less reassuring story. The question that opens is uncomfortable for a macroprudential authority: did the buffer come down against the right risk?
Two different measurements
The case for the cut rested on a stock measure. Overvaluation is a statement about the gap between house prices and their long-run relationship with rents, incomes and interest rates. By the Bundesbank’s own Financial Stability Review 2025, that gap had narrowed to under 10 per cent for Germany as a whole, down from the double-digit overvaluations of 2022. On that metric, the housing market genuinely looked safer in April 2025 than it had two years earlier. Prices had stopped running away from fundamentals.
But a buffer against residential mortgage risk is not only insuring against a price correction. It is insuring against the leverage in the loans being written now — the flow. And until this year, German supervisors could not see the flow at loan level. They inferred average loan-to-value from ECB surveys and industry samples, useful for spotting a trend but useless for measuring a distribution. The data that would show whether new originations were getting riskier simply did not exist in harmonised form.
That changed on 11 May 2026, when the Financial Stability Committee published its first assessment built on WIFSta, the Bundesbank’s new granular housing-loan collection. Clarqo covered the debut as a new instrument — Germany finally acquiring the surveillance infrastructure to calibrate a borrower-based limit. The point worth returning to is narrower and more awkward: the first WIFSta reading describes loans written after the buffer was cut, and it does not look like a market that needed less of a capital cushion.
What the flow shows
For Q4 2025 new originations, WIFSta records an average loan-to-value of 83 per cent, an average debt-service-to-income ratio of 38 per cent, a debt-to-income ratio of roughly 6.3 times annual income, and an average maturity of 29 years. On the means alone, the FSC’s verdict was that “lending standards were therefore, on average, within an acceptable range.”
The averages are not the story. The tail is. Fourteen per cent of new loans were written at a loan-to-value above 100 per cent — borrowers in negative equity from the day the contract was signed, before a single price tick moves against them. The committee flagged the “relatively large share of new loans with high debt ratios,” noting it stands out in European comparison. This is a cohort that a modest house-price correction pushes underwater and that an income shock cannot easily absorb, because a meaningful slice of monthly income is already committed to servicing the debt.
Here is the mechanical problem with the timing. A high-LTV, high-DSTI origination written in late 2025 carries its risk for the life of the loan — 29 years, on these numbers. The overvaluation that justified the buffer cut is a statement about the stock of housing today. The leverage in the new loans is a liability the banking system will hold into the 2050s. The buffer was eased against the first and is exposed to the second, and the two had moved in opposite directions: prices back toward fundamentals, origination leverage stretching at the top end.
The institutions hedged — but on instinct, not data
To be fair to BaFin and the Bundesbank, neither eliminated the buffer. They halved it and left 1 per cent standing, and the Financial Stability Review was unusually candid about why. The buffer “remains necessary,” the Bundesbank wrote, citing “increased uncertainty as to the lending standards applied to loans granted in the past.” Translated: we have cut against the price risk we can measure, but we are holding capital against a lending-standards risk we cannot yet see clearly.
That is a defensible posture — and the WIFSta data has now partly vindicated the instinct to keep 1 per cent rather than go to zero. But it also exposes the sequencing. The authority moved its capital setting in April 2025 on the strength of a backward-looking price metric, while explicitly admitting it lacked visibility on the forward-looking flow metric. The flow data arrived a year later and pointed at exactly the risk the hedge was meant to cover. The macroprudential stance was not wrong so much as it was operating blind on the dimension that turned out to matter, and it loosened anyway.
The defence that the cut was small — 0.4 per cent of system CET1 — cuts both ways. If the release was that marginal, the resilience it bought back for banks was marginal too, against a tail the data has since put a number on.
What to watch
The honest read is not that the buffer cut was a mistake. It is that the sectoral buffer is the wrong tool for the problem WIFSta has surfaced, and the authorities now have the data to use a better one. A broad sectoral capital buffer taxes every residential mortgage equally, prudent and stretched alike. The risk in the WIFSta distribution is concentrated in a tail — the high-LTV, high-DSTI cohort. The targeted instrument for a tail is a borrower-based limit: an LTV cap, a DSTI cap, an amortisation requirement, the toolkit the FSC has held under §48u of the Banking Act since 2017 but never deployed for want of data to calibrate or enforce it.
Three things will tell you which way this goes. First, whether the FSC’s next assessment hardens from “closer analysis of mitigating factors” toward consideration of those borrower-based instruments. Second, whether BaFin opens supervisory dialogue with the regional Sparkassen and cooperative lenders that have taken mortgage share since 2023 and are the likeliest holders of the high-LTV book. Third, the Q1 2026 WIFSta reading, due in August, which will show whether the stretched tail was a Q4 seasonal effect or a structural drift.
And there is a fourth, quieter signal: whether the 1 per cent buffer gets revisited upward. Cutting a buffer is a statement that risk has fallen. If the loan data Germany now collects keeps describing a market writing leverage at the top of the distribution, the harder question is not whether the April 2025 cut was premature. It is whether the next move has to be a reversal.
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.