Finanstilsynet tightens requirements on residential mortgage models
Published: 1 July 2014
Last updated: 11 February 2020
Document number: 22/2014
Finanstilsynet is to impose stricter requirements on the models used by banks to estimate residential mortgage risk. The largest banks use these models to calculate risk weights which determine the capital required to back their mortgage loans. Mortgage risk weights have fallen in recent years while higher house prices and higher household indebtedness have increased the risk present in the mortgage market.
The authorities have on several occasions questioned the ability of the banks' models to adequately capture risk in the mortgage market. After reviewing the models in 2012, Finanstilsynet announced in spring 2013 that stricter requirements would be imposed. In the same year the Ministry of Finance published a consultation paper introducing a range of measures to increase risk weights, and opted to raise the floor on loss given default estimates (LGD floor) from 10 to 20 per cent with effect from 1 January 2014. Finanstilsynet's measures combined with the LGD floor entails a significant increase in the risk weights assigned to residential mortgages.
Rules and measures
The Basel II framework, introduced in 2007, permitted banks to use internal rating based (IRB) models to calculate capital requirements for credit risk. Under the IRB approach, banks use their own estimates of probability of default (PD) and LGD to compute a risk weight that determines the capital requirement for each individual exposure. Finanstilsynet has given eight banks and their affiliated mortgage credit institutions permission to apply the IRB approach to residential mortgages.
The capital requirements regulations impose strict requirements on banks' models and supervisory authorities' assessments: the models are required to predict probabilities of default through an economic cycle along with loss ratios under downturn conditions. Safety margins must be included to allow for the uncertainty inherent in models and data. Finanstilsynet previously made it clear that model estimates should reflect the default and loss experience during the banking crisis in the early 1990s. However, data quality from that period is highly uncertain. Uncertainty about the models' long-term predictive ability is compounded by the fact that other data used in the models reflect boom years for the Norwegian economy. Hence Finanstilsynet is now outlining assumptions on which models are to be based, detailing how default data from the banking crisis are to be taken into account, extra safety margins in the best risk classes and levels of loss ratios.
Finanstilsynet estimates that the requirements for PD models, in combination with the LGD floor, will increase risk weights assigned to residential mortgage portfolios to around 20-25 per cent compared with previous levels of 10-15 per cent. The effect is greatest for banks that have previously estimated the lowest PDs or shown the highest concentration in the best risk classes. The effect on the LGD models will depend on loan to value (LTV) ratios, but estimates are expected to rise to the level of the 20 per cent floor.
Finanstilsynet has informed the banks of the tightening both at on-site inspections and in a letter to Finance Norway which also asked the banks for comments [letter to Finance Norway]. The banks' objections are largely on the choice of parameter values where a large degree of judgement is required. The substantial uncertainty of historical estimates requires safety margins. Finanstilsynet upholds its assessments but has made some adjustments in light of the banks' input. Finanstilsynet requires the banks to adapt their models by the end of 2014.
Finanstilsynet has contacted Swedish and Danish supervisory authorities, who share Finanstilsynet's concerns with regard to mortgage models. The Danish supervisor has announced that the requirements will apply to Danish IRB-banks' Norwegian mortgages while the Swedish supervisor has indicated a similar tightening of the Pillar 2 requirements for Swedish banks.