Banks and regulation
How often banks must revalue collateral under CRR3
CRR Article 208(3)(a) requires banks to monitor the value of collateral at least once a year for commercial property and once every three years for residential property, and more often when the market changes significantly. Monitoring can use an index or a property-level model; both must be granular and built on observed transactions and appraisals (EBA paragraph 223). Since CRR3, the monitoring results also feed the six- or eight-year average that caps any upward revaluation (Article 229(1)(e)), so the cadence of monitoring now limits what a bank can recognize.
The difference between monitoring and revaluation
- Monitor. Watch the value of the book on a schedule and flag properties whose value may have moved. CRR Article 208(3)(a).
- Revalue. Put a new value on a property. EBA paragraph 224 lists the approaches: "desktop valuation, drive-by valuation, full visit with internal and external assessment of the property, statistical models".
- Review. The valuer's look at a property when information "indicates that the property value may have declined materially relative to general market prices". CRR Article 208(3)(b), which since CRR3 counts ESG considerations as such an indication.
Monitoring decides which properties get revalued. Revaluation produces the number. A review is the revaluation the regulation reserves for a valuer.
How often
| Rule | Residential | Commercial | Where |
|---|---|---|---|
| EU minimum | At least every 3 years, more often when markets change significantly | At least every year | CRR Article 208(3)(a) |
| EU, by risk | More often for unfinished buildings, high carrying amounts, high LTV and weaker credit quality | Same | EBA/GL/2020/06, paragraph 222 |
| EU, large loans | Valuer review at least every 3 years for loans above EUR 3 million or 5 % of own funds | Same | CRR Article 208(3)(b) |
| Poland | At least every 3 years where the current LTV is below 80 %, otherwise yearly | Same rule | KNF Recommendation S, 16.1 |
| Slovakia | Internal valuer revalues at least every 3 years, and within 1 year of a significant change in market conditions | Not covered by Measure 10/2016 | NBS Measure 10/2016, Section 8(1) |
| Hungary | Good practice: at least every 3 years; for loans above EUR 3 million and 5 % of own funds, a full new valuation every 3 years | At least every year | MNB Recommendation 15/2021, points 21 and 24 |
Index or model
A bank can monitor in two ways. It can apply a price index for the area to the last known value, or it can value each property again with a model that reads the property's own characteristics.
| Question | Index on the last value | Property-level model |
|---|---|---|
| What it needs | The last value and its date; an index for the area | The property's address and characteristics; recent evidence nearby |
| Granularity | As fine as the index: region, district, sometimes town | The property and its immediate area |
| Handles a new building next to an old one | No: both move with the area | Yes, if age is an input |
| Handles renovation or damage | No | Only if the model receives that information |
| Drift over long periods | Grows: an error in the starting value is carried forward | Each value starts fresh from current evidence |
| Cost and explanation | Low cost, easy to explain | Higher cost; needs documentation and validation under Article 208(3a) |
| EBA paragraph 223 test | The index must be "sufficiently granular" and built on "a sufficient time series of observed empirical evidence of previous transactions and appraisals" | The same test, plus the model criteria in Section 7.4 |
The two can work together. The index moves the whole book each quarter, and the model takes the properties the index flags or cannot handle.
What the national rules say about indices
Hungary has the most detailed index rules of the four. The MNB's Recommendation 15/2021 accepts statistical revaluation of residential collateral (point 25) and names as an acceptable method indices built from "the change in property prices reflecting realized sales prices" between years (point 29; quotations from Hungarian, Slovak and Polish texts here are our translations). It then sets limits:
- The index should rest on a representative transaction database, and regional or county indices should account for differences within the area (point 30).
- The bank is responsible for the result even when it buys the index (point 31).
- In settlements with fewer than five sales a year over the previous two years, a regional or county index, or at most a national one, is good practice; increases above that should be recognized only after an individual valuation (point 32).
- A fall in value beyond the bank's own threshold, or an increase the bank wants to recognize above that threshold, sends the property to an independent valuer (point 26). The MNB considers a threshold of no more than 20 percent good practice (point 27).
- After a significant change in market conditions, statistical revaluation is allowed, but the new value may not exceed the last value approved by an independent valuer (point 39).
Slovakia allows a "procedure involving monitoring of the property value, identification of the property that needs revaluation and revaluation of the property" alongside the internal valuer's expert estimate (Measure 10/2016, Section 8(2)). The procedure uses a statistical method, and the internal valuer checks material differences from general market prices (Section 8(4)). The data set behind it must hold properties from the bank's own contract documents, each only once, with three different characteristics that matter for value (Section 8(5)). The sample must also be large and homogeneous enough and cover a long enough time series.
Poland ties the frequency to LTV (Recommendation S, 16.1) and, in a footnote, reads monitoring as the same concept as in CRR Article 208, "allowing the use of statistical methods to monitor property values and identify properties that need revaluation". Recommendation J asks banks with at least 2 percent of the national mortgage market to use statistical models built on reliable databases for assessing the risk of changes in collateral value (recommendation 10).
How monitoring now feeds the property value
CRR3 caps any upward revaluation at the higher of the value at origination and the average value "over the last six years for residential property or eight years for commercial immovable property". The average is taken "across property values observed at equal intervals", with "at least three data points", and banks "may use the results of the monitoring of property values in accordance with Article 208(3)" to calculate it (Article 229(1)(e)).
Two consequences follow for a monitoring design:
- 1
Cadence
A bank that monitors residential collateral only at the three-year minimum has three points in a six-year window at best, the minimum the rule allows. If one point is missed or late, the bank cannot revalue upward at all, because "the property value shall not be revalued upward if institutions do not have sufficient data to calculate the average value". Monitoring once a year gives six or seven points and leaves a margin for a missed one.
- 2
Level
An index that is too coarse lifts or holds down every property in its area alike. Under the old Article 229, monitoring results could only reduce the recognized value. Under CRR3 they also enter the average that caps any increase, and through it the risk weight.
Our page on [CRR3 property value](/guides/crr3-property-value) works the cap through with numbers.
A worked example
The figures below are made up to show the mechanics. They are not market data.
A flat in a regional city was valued at EUR 150,000 in 2019. The regional index rose 55 percent between 2019 and 2026. The flat's own block, an older building with no lift, saw prices on comparable flats rise 38 percent over the same years.
| Method | 2026 value | Gap to the block's evidence |
|---|---|---|
| Index on the 2019 value | 150,000 × 1.55 = EUR 232,500 | +EUR 25,500 |
| Property-level model | 150,000 × 1.38 = EUR 207,000 | 0 |
The index overstates this flat by EUR 25,500, 12.3 percent of the model value. On a loan of EUR 120,000, the LTV reads 51.6 percent on the index value and 58.0 percent on the model value. For a new building in an area of older stock the error runs the other way, and the index understates the value.
Asking price indices
A mix-adjusted index built from asking prices, such as [Bytero Live](/live), moves earlier than a sale-price index, because a flat is offered for sale months before its sale is recorded. It is useful as an early signal, since a fall in asking prices in a district is a reason to look at the collateral there.
It is not, on its own, the "time series of observed empirical evidence of previous transactions and appraisals" that EBA paragraph 223 asks for, because asking prices are neither transactions nor appraisals. A bank that uses one should pair it with sale prices or valuations, where they exist (our page on [asking and sale prices in Slovakia, Czechia, Poland and Hungary](/guides/asking-vs-transaction-prices-cee) lists what is published).
A short checklist for a monitoring design
- Monitor residential collateral at least yearly if you want to revalue upward under CRR3.
- Store each monitoring value with its date and method; never overwrite the value at origination.
- Write the triggers down, as EBA paragraph 224 asks, for example a fall beyond a threshold, low model confidence and property-specific information.
- Check the index's granularity against the spread of prices inside each area it covers.
- Send new buildings, renovated buildings and thin markets to the property-level model or a valuer.
Where Bytero Engine fits
Bytero Engine values each flat from its own inputs and nearby evidence, which is the property-level column in the table above, and Bytero Live provides an asking price index by region with monthly steps. A bank can run the model across its book through the API and keep the index as an early signal. Both are built from asking prices, so we recommend pairing them with the bank's own sale prices and valuations for back-testing.
Sources
- Regulation (EU) No 575/2013 (CRR), Article 208(3)(a), unchanged by CRR3
- Regulation (EU) 2024/1623 (CRR3), amended Articles 208(3)(b), 208(3a) and 229(1)
- EBA, Final Report on Guidelines on loan origination and monitoring (EBA/GL/2020/06), paragraphs 221 to 226
- Magyar Nemzeti Bank, Recommendation 15/2021 (X.29.), points 21 to 39
- Národná banka Slovenska, Measure 10/2016, consolidated text, Section 8
- KNF, Recommendation S, amended June 2023, recommendation 16 and footnote 10
- KNF, Recommendation J, March 2023, recommendation 10






