AnaCredit forbearance — status values and FINREP

Forbearance is one of the most-queried and worst-implemented corners of AnaCredit — and the commonest defect is a data model that knows about four status values when the register defines five. The register asks, per instrument, whether the terms were modified because the debtor was in — or about to be in — financial difficulty, and when that status arose. The answer must reconcile with the forbearance definitions your FINREP process already applies and with the impairment staging your accounting runs. When the three disagree, the national central bank asks why.

1. Where forbearance sits in the data model

AnaCredit is established by Regulation (EU) 2016/867 of 18 May 2016 (ECB/2016/13). Forbearance is carried in the accounting dataset through two linked attributes: status of forbearance and renegotiation, the classification of the instrument, and date of the forbearance and renegotiation status, recording when the current status arose. The accounting dataset runs on a quarterly cadence rather than the monthly cadence of the financial dataset, and both attributes must move consistently with the rest of it — impairment stage, carrying amount, cumulative recoveries.

One scoping point shapes what a reporting agent actually has to produce. The reporting threshold is €25,000, measured on the debtor’s commitment amount — the sum of outstanding nominal and off-balance-sheet amounts across all that debtor’s instruments — and under Article 5(1) instruments continue to be reported until at least the end of the quarter in which that amount falls below the threshold. Forbearance frequently coincides with a shrinking exposure, so the instrument you most want to stop reporting is often the one you must keep for another quarter.

2. The five status values

The register defines five values, and the third is the one most implementations omit.

The refinancing value has its own definition to satisfy: refinancing is the use of debt contracts to ensure the total or partial payment of other debt contracts whose current terms the debtor is unable to comply with. That definition is what makes it a forbearance value rather than an origination event, and it is why a refinancing generates two reportable instruments carrying the same status, not one.

The test for the forborne values is difficulty plus concession, the same as the definition applied in FINREP. Both legs must be present: the debtor faces, or is about to face, financial difficulty, and the institution grants a concession it would not otherwise offer. A payment holiday granted as a marketing feature is renegotiation. A maturity extension for a struggling debtor is forbearance even if pricing is unchanged.

3. Worked example — a rate cut that is not forbearance

Facts: a healthy corporate borrower produces a competitor’s term sheet and asks for a lower margin. The relationship manager agrees a 60 basis point reduction to retain the client. Separately, a second borrower who has missed two instalments is granted the same reduction plus a six-month interest-only period.

Which rules apply: the attribute captures all modifications of an instrument’s terms, irrespective of whether they meet the forbearance criteria — so both events are reportable, but not identically. The first is a contractual change for commercial reasons made with the active involvement of the debtor: renegotiated instrument without forbearance measures, with the status date updated to the effective date. The second satisfies difficulty plus concession and includes a rate modification below market conditions, so it takes forborne: instruments with modified interest rate below market conditions — that value applies where the package includes at least a rate modification, even though a payment concession was granted too.

What the practitioner does: records the difficulty determination as a structured decision at the point of approval, not as a narrative in a credit paper. Without it, the two cases are indistinguishable downstream: identical pricing change, identical documentation, opposite reporting treatment.

Outcome: a same-size concession lands in two different values, and the register shows a renegotiation count that is not silently inflating the forborne population — the error that makes a portfolio look worse than it is, and invites a data-quality query all the same.

4. The reconciliation triangle

Three frameworks look at the same event, and supervisors cross-check them:

• AnaCredit — instrument-level status and date;
• FINREP — aggregate forborne-exposure reporting in template F 19.00, under Annex V of Commission Implementing Regulation (EU) 2021/451;
• IFRS 9 — modification accounting and impairment staging, where a forbearance event is a strong indicator of a significant increase in credit risk and typically drives Stage 2 or Stage 3.

The sum of forborne instruments in AnaCredit should track the FINREP forborne totals, and instruments flagged forborne should rarely sit in Stage 1. Persistent mismatches invite a data-quality letter. The clean implementation drives all three from a single forbearance flag set in the credit-risk workflow at the moment the concession is approved — not from three parallel processes that happen to agree in a good quarter.

5. Probation, and the clock that stops

Exit from forborne status is governed by the FINREP framework, and AnaCredit reflects the classification that framework produces. Under Annex V, Part 2 of Implementing Regulation (EU) 2021/451, a minimum two-year probation period must pass from the date the forborne exposure was considered performing. Where an exposure previously had a concession event and was classified non-performing, it serves a one-year non-performing probation from the date of the concession event, followed by the two-year performing probation.

Two mechanics catch implementations out. The two-year probation applies to all forborne exposures, irrespective of whether they were performing or non-performing when the measures were granted. And classification as non-performing interrupts the counting of days for the probation period — so probation is not a date-plus-24-months calculation, it is an accumulator that pauses. Once the discontinuation conditions are met the status reverts to not forborne or renegotiated, and the date attribute carries the discontinuation date; where that date differs from the instrument’s inception, it is itself the signal that the instrument was previously forborne.

6. Worked example — an exit that was not an exit

Facts: an instrument is forborne in March of year 1, becomes performing in September of year 1, is classified non-performing again for four months during year 2 after two missed payments, and recovers. In September of year 3 the reporting team removes the forborne flag on a straight two-year reading.

Which rules apply: the performing probation started in September of year 1, but the four months of non-performing classification in year 2 interrupted the count. Two years of qualifying probation therefore complete around January of year 4, not September of year 3.

What the practitioner does: implements probation as a day accumulator keyed to performing status, driven off the same performing / non-performing classification that feeds FINREP, and does not build a separate AnaCredit-only exit rule. Where a second concession is granted during probation, the position typically moves to non-performing forborne, and the AnaCredit status, date and impairment attributes should all show that same picture in the same reporting period.

Outcome: the flag comes off four months later than the naive calculation, the AnaCredit and FINREP populations agree, and the firm avoids a finding that recurs in on-site inspections — premature exit from forborne classification, which understates forborne volumes in both returns at once.

7. Operational wiring that works

• Put the forbearance decision in the credit-workflow tool as a structured field — difficulty flag, concession type, effective date — approved together with the restructuring itself.
• Feed AnaCredit, FINREP and IFRS 9 staging from that single field. Never from three parallel spreadsheets.
• Handle refinancing as a two-instrument event: the new refinancing contract and the old outstanding repaid contract both carry the forborne status.
• Reconcile quarterly: instruments flagged forborne in AnaCredit against the F 19.00 population and against the Stage 2 and Stage 3 lists. Investigate every one-sided entry before the national central bank does.
• Keep the history. The date attribute is the audit trail the NCB checks first, and the discontinuation date is what evidences a legitimate exit.

FAQ

What to do, today

• Reporting lead: confirm your data model carries all five values. If “totally or partially refinanced debt” is missing, refinancings are being reported under the wrong value today.
• Credit risk: structure the difficulty determination as a field, not free text, and record it at approval.
• Engineer: build probation as an interruptible day accumulator keyed to performing status, and check the €25,000 exit rule keeps reporting to the end of the quarter.
• Finance: run the quarterly three-way reconciliation against F 19.00 and IFRS 9 staging, and treat forborne instruments in Stage 1 as exceptions requiring an explanation.

Related: The AnaCredit data model · AnaCredit for payment firms · AnaCredit phases 1, 2 and 3 · How an AnaCredit submission is built · IFRS 18 — NIIF-UE 18 and the 2027 accounts reset