Annex I. Structural Model of Mortgage Risk for Austria
1. We study the impact of borrower-based macroprudential tools in Austria by extending the Reserve Bank of New Zealand (2008) model. The model simulates mortgage default rates and losses in a tail risk event. It assumes that borrowers will continue to service a loan if they can afford to pay off the scheduled interest and principal payments, even when they have negative equity. The drivers of borrowers’ debt servicing capacity include changes in (i) house prices; (ii) income; (iii) unemployment rate; and (iv) mortgage interest rates. We extend the model to incorporate Austria-specific characteristics, and to allow macroprudential tools to affect the risk profile of new mortgages.
2. To simulate mortgage losses under borrower-based macroprudential limits, we proceed in three stages:
Assessment of tail risks to the macroeconomic outlook. To calibrate the tail risk event, we use the Growth-at-Risk (GaR) and House Price-at-Risk (HaR) methodology.
Introduction of macroprudential tools. We assume macroprudential policies are introduced 4 or 8 quarters before the tail risk materializes. They affect the distribution of LTV, DSTI, and DTI ratios of new mortgages granted during that period.
Simulation of mortgage defaults and losses. We simulate default rates and losses on banks’ mortgage portfolios in the tail risk event for different levels of macroprudential limits.
3. The use of GaR and HaR ensures the tail risk event reflects the current level of systemic risk. As calibration of macroprudential tools should be guided by the current level of risks to macrofinancial stability, we employ GaR and HaR to link the desirable tightness of macroprudential tools with cyclical position:
Following Adrian et al. (2016) and IMF (2017), downside risks to both GDP growth and house price growth are proxied by the fifth percentile of the forecasted distributions of future GDP and house prices. The methodology employs a quantile regression estimation on macrofinancial data between 1993Q3–2018Q4 to calculate the predicted distribution one year ahead. The explanatory variables used for the house prices-at-risk analysis include: house price growth, financial conditions index, private sector credit growth, house price misalignment, and GDP growth. Explanatory variables for the growth-at-risk include lagged GDP growth and financial conditions index.
Estimates from GaR and HaR models are used to inform the size of income and house prices decline in the tail risk event.
The increase in the unemployment rate is estimated based on the past relationship with GDP growth, and the increase in the interest rate on housing loans in the tail risk event is calibrated based on evidence from past recessions.
4. Borrower-based limits reduce the riskiness of new loans in the model. To study the impact of borrower-based macroprudential tools, we simulate the portfolio of mortgages for four or eight quarters into the future, before the tail risk event is triggered. During this time, the macroprudential limits are binding and affect the LTV, DTI, and DSTIs of new flows of mortgages, while some of the outstanding loans mature. We assume that in the absence of macroprudential measures, the new mortgage flows would be similar (in terms of volume, LTV, DTI, and DSTI distributions) to the average flows observed in the last four quarters of data. The income and house prices are assumed to grow at the median values predicted from the GaR and HaR exercises, and we assume no change in the unemployment or in the mortgage interest rate in the period before the tail risk event.
5. In the last stage, we simulate mortgage losses during the tail risk event. The model assumes that a borrower defaults on housing loan if two conditions are satisfied:
- The borrower is in distress: he is unable to repay the debt on time due financial difficulties. The probability of the financial distress, Pr (FSt), is a function of i) the borrower’s initial DSTI, ii) the change in the debt servicing capacity, ΔDSTI (due to changes in income and interest rate), and iii) unemployment rate U and growth in unemployment, ΔU1
The net value of collateral, after disposable costs, is less than the value of the loan: the borrower cannot sell the collateral to service the loan. The value of collateral is proxied by the house price
net of fees to be paid when selling the house, C. The net present value of the loan, NPV, depends on the outstanding value of the loan Lt, the interest rate, rt, and the remaining maturity of the loan T — t.
6. Conditional on defaulting, a bank’s loss on a mortgage is driven by the discounted sale price of the house. The loss-given default (LGD) is calculated assuming that the sale occurs at time t+s (where s denotes the time to sell the collateral, calibrated at 1.75 years) and that the sale proceeds net of transaction costs, δ, are discounted at a rate reflecting the risk premium of the foreclosed asset (rf + spread):
7. The default rate and LGD for the banking system-wide mortgage portfolio are generated through simulations. Using information on system-wide distribution of LTVs, DSTIs, and DTIs, we simulate probabilities of default (PDs) and LGDs for each LTV-vintage bucket of mortgages and calculate system-wide PDs and LGDs as weighted averages of bucket-specific values. For a given bucket of mortgages, the model generates 10,000 draws of individual house prices (from a normal distribution with the mean equal to the fifth percentile estimate from the HaR model). For each of these house prices, the model determines whether a borrower defaults and computes the LGD. We simulate each bucket 2,000 times and take the averages across iterations to compute the bucket-specific PDs and LGDs.
8. We study several alternative calibrations of borrower-based limits. The model is flexible in allowing several alternative calibrations of borrower-based macroprudential limits. In particular, we consider a range of (i) LTV limits; (ii) DSTI limits; (iii) a combination of LTV-DSTI limits; and (iv) a combination of LTV-DSTI-DTI limits. We also distinguish between hard limits and speed limits, where, in the latter case, a pre-defined share of loans is allowed not to comply with the regulatory limit. For each choice of the macroprudential tools, we compare the losses on the mortgage portfolio generated in the tail risk event with the losses observed in the absence of any limits. The results suggest that reasonable calibrations of combined LTV-DSTI or LTV-DSTI-DTI limits with speed limits could cut the expected losses on the new loans in a tail risk event by over a half.
9. The model can incorporate several Austria-specific characteristics of the housing market, but caveats apply. The model distinguishes between fixed and floating-rate loans when considering the impact of interest rate increases on debt servicing capacity in the tail risk event. The LTV and DSTI values for different mortgage vintages are updated to reflect changes in house prices, real income, and loan repayments since the mortgage origination. At the same time, the model has some important limitations. In particular, the calibration of macroprudential tools should incorporate a broader cost-benefit analysis, that would capture the impact of the new limits on bank credit and house prices. These considerations remain beyond the scope of the model developed here.
Adrian, Tobias, Nina Boyarchenko, and Domenico Giannone. 2016. “Vulnerable Growth.” Federal Reserve Bank of New York Staff Report No. 794.
Central Europe and South Eastern Europe.
The improvement in economic conditions in the CESEE region has helped too.
The Financial Stability Analysis, Stress Testing and Interconnectedness Technical Note provides a complementary analysis of key vulnerabilities and provides an in-depth assessment of the banking system’s resilience to severe macrofinancial shocks.
European Union Capital Requirement Regulations No. 575/2013 and European Union Capital Requirement Directive 2013/36/EU, respectively.
When a measure is intended to be undertaken by a national authority, the ECB should be notified within 10 working days in advance of the relevant decision, and the ECB can object to the proposed measure within 5 working days, stating its reasons for the objection in writing. Where the ECB objects, the national authority is required to consider the ECB’s reasons prior to proceeding with the decision as appropriate. Similar notification requirements apply to the ECB decision to apply higher requirements.
The Austrian authorities inform the ECB and the European Systemic Risk Board about the exercise of those powers.
Defined in Articles 13 and 13 lit. a of the FMABG for the FMSB, Article 13 paragraphs 2 and 9 of the FMABG for the FMA, and Articles 44 lit. c of the NBG for the OeNB.
The FMSB advised the MoF to create a legal basis for borrower-based macroprudential instruments in June 2016.
Vienna Initiative was established in 2009 to bolster coordination among home and host country authorities of EU-based cross-border banks active in the CESEE region and in order to avoid disorderly deleveraging. The initiative continues to this day as Vienna Initiative 2.0.
Press releases are subject to the FMSB´s decision making, whereas the Financial Stability Report is a document of the OeNB.
The six categories include; financial markets, mispricing of risks, strength of bank balance sheets, credit and debt developments, price pressures, and external imbalances.
The FMA is also actively participating in the EU-level initiatives and discussions related to the development of macroprudential policy tools for the insurance sector.
For example, sensitivity of debt servicing capacity of households to declines in house prices was presented in FSR no. 31; sensitivity of adjustable rate mortgages to interest rate increases was presented in FSR no. 32.
Dynamic Stochastic General Equilibrium model.
AnaCredit is an initiative of the European System of Central Banks. It is a dataset containing detailed information on individual bank loans in the euro area, harmonized across all member states.
The credit-to-GDP gap follows the BIS (Bank for International Settlements) definition. Additionally, the OeNB’s aggregate indicator of systemic risk, based on variables from six risk categories (following ESRB recommendations) does not point to a broad-based build-up of vulnerabilities.
See Financial Stability Analysis, Stress Testing, and Interconnectedness Technical Note.
A corporate provision fund is a company authorized to pursue severance and retirement fund activities.
The European Commission (EC) has asked European Insurance and Occupational Pensions Authority (EIOPA) to assess whether the existing provisions of the Solvency II framework allow for appropriate macroprudential oversight. EIOPA is also analyzing whether to integrate macroprudential measures similar to the banking framework into the Solvency II regulation.
Parallels exist in the European Union, for example, between 2008–2010, when investors’ redemptions requests to German open real estate funds exceeded funds’ liquid assets, several German real estate funds suspended redemptions while others announced liquidation.
Guiding Principles on lending in FX refers to the commitment of Austrian banks to refrain from the riskiest forms of FX lending in CESEE.
This is when treating the Raiffeisen sector as one banking group.
The 2018 extension of the EBA guidelines also included the introduction of four new sufficient conditions for being identified as an O-SII: i) based on a threshold approach identifying institutions whose score is particularly high in any one of the indicators used; (ii) based on a high EBA score when substituting institution’s share in EU deposits and EU loans with domestic equivalents; (iii) based on the institution’s market share of covered deposits (above 3.5 percent); and (iv) based on unconsolidated banking data.
In 2015, the Unicredit decided to consolidate its CESEE subsidiaries in Italy rather than in Austria.
A speed limit of 10 percent means that 10 percent of new mortgage flows do not have to comply with the macroprudential limits.
Parameters β0, β1, β2, β3 are calibrated as in RBNZ (2008), parameter D is calibrated to match the Austria-specific average probability of default of a mortgage during normal times.